University of Groningen Pharmacotherapy in frail elderly

University of Groningen

Pharmacotherapy in frail elderlyDijk, Karen Nanette van

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite fromit. Please check the document version below.

Document VersionPublisher's PDF, also known as Version of record

Publication date:2002

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA):Dijk, K. N. V. (2002). Pharmacotherapy in frail elderly: pharmacy data as a tool for improvement. s.n.

CopyrightOther than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of theauthor(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

The publication may also be distributed here under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license.More information can be found on the University of Groningen website: https://www.rug.nl/library/open-access/self-archiving-pure/taverne-amendment.

Take-down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons thenumber of authors shown on this cover page is limited to 10 maximum.

Download date: 11-11-2021

https://research.rug.nl/en/publications/pharmacotherapy-in-frail-elderly(3887a8bc-7d07-414c-ad3f-ea4974be80ec).html

Pharmacotherapy in f ra i l e lder ly :

Pharmacy data as a tool for improvement

3

ISBN-nummer: 90-367-1639-x

©2002, K.N. van Dijk

All rights reserved. No part of this book may be reproduced in any manner or by any means

without written permission from the author.

This research was financially supported by the Scientific Institute Dutch Pharmacists

(WINAp). Publication of this thesis was kindly sponsored by the Royal Dutch Association for

the Advancement of Pharmacy (KNMP) and the Groningen Universtity Institute for Drug

Exploration (GUIDE).

Several sections of this thesis are based on published papers, which are reproduced with p

ermission of the co-authors and the publishers. Copyright of these papers remains with the

publishers.

Vormgeving en lay-out: Robbin van Nek, Vos & Libert reclame, Leeuwarden

2

Rijksuniversiteit Groningen

Pharmacotherapy in f ra i l e lder ly :

Pharmacy data as a tool for improvement

Proefschrift

ter verkrijging van het doctoraat in de

Wiskunde en Natuurwetenschappen

aan de Rijksuniversiteit Groningen

op gezag van de

Rector Magnificus, dr. D.F.J. Bosscher,

in het openbaar te verdedigen op

vrijdag 7 juni 2002

om 14.15 uur

door

Karen Nanette van Dijk

geboren op 23 oktober 1966

te ‘s-Gravenhage

5Paranimfen: Petra de Boer-Stoter

Karen Cromheecke-Berghuis4

7Promotores: Prof. dr. L.T.W. de Jong- van den Berg

Prof. dr. J.R.B.J. Brouwers

Referent: Dr. C.S. de Vries

Beoordelingscommissie: Prof. dr. A.C.G. Egberts

Prof. dr. F.M. Haaijer-Ruskamp

Prof. dr. J.P.J. Slaets

6 Rien n’est simple,

tout est possible

Page

Chapter 3 Risk assessment studies in the elderly 115

3.1 Constipation as an adverse effect of drug use in nursing home patients: an overestimated risk 117

3.2 Potential interaction between acenoucoumarol and diclofenac, naproxen and ibuprofen and therole of CYP2C9 genotype 133

Chapter 4 General discussion and perspectives 147

Summary 159

Samenvatting 164

Dankwoord 169

Publications 172

Curriculum vitae 173

9

Contents

Page

Chapter 1 Scope, objective and setting 11

1.1 Scope and objective 13

1.2 Setting 18

Chapter 2 Drug util isation studies in the elderly 23

Part one: nurs ing home res idents

2.1 Background 25

2.2 Use of hospital pharmacy data in pharmaco-epidemiologic research in nursing homes 31

2.3 Drug utilisation in Dutch nursing homes 44

2.4 Occurrence of potential drug-drug interactions in nursing home residents 60

2.5 Prescribing indicators as a tool to evaluate drug use in nursing homes: a pilot study 76

Part two : e lder ly outpat ients

2.6 Concomitant prescribing of benzodiazepines during antidepressant therapy in the elderly 90

2.7 Prescribing of gastroprotective drugs among elderly NSAID users in the Netherlands 103

8

Chapter 1

Scope, ob ject ive and sett ing

1110

1.1 Scope and ob ject ive

Introduct ion

Drug utilisation in the elderly has been subject of many studies [1-4]. In the Netherlands,

people aged 65 and over comprise about 14% of the Dutch population and account for 40% of

the drug prescription costs spent in hospital and community pharmacies [5]. Multiple chronic

conditions and the fact that the possibilities for both preventive and therapeutic medical the-

rapies for many diseases have increased in recent years, contribute to the high frequency of

drug use in the elderly. The prescribing of a drug to counteract adverse effects of another drug

(‘prescribing cascade’) may further increase drug use in the elderly [6].

In view of multiple co-morbidity, changes in drug kinetics and effects and the prescription

of several drugs simultaneously (polypharmacy), elderly people are at an increased risk of

drug-related problems such as drug-drug interactions, drug-disease interactions and adverse

drug reactions (ADRs) [7]. The prevalence of ADRs ranges from 1.5 to 44% in elderly inpatients

and from 2.5 to 50.6% in elderly outpatients [8]. Examples of age-related risks of ADRs are

bleeding from oral anticoagulants and gastropathy associated with non-steroidal anti-inflam-

matory drugs [9]. Apart from the risk of overprescribing in this group [6,10], underprescribing

of effective agents, such as statins, may also be harmful to the elderly [11-13]. Also, underdiag-

nosing of certain diagnoses, such as depression, is reported to be an issue in the elderly [14].

Despite the awareness of the problems associated with drug use in the elderly and the atten-

tion that has been given to rational prescribing, the frequency of drug-related hospital admis-

sions among elderly people aged 65 and over is still considerable [15-17]. The incidence of drug-

related problems is reported to be even higher in nursing home patients, due to the higher

levels of drug use and the complexity of the conditions these patients are cared for [18,19].

Physicians are faced with a complex task when prescribing to elderly people [20-22].

Advanced age leads to increased frailty and altered pharmacokinetics and pharmacodynamics

with large interindividual variability, often leading to unpredictable drug responses [23].

Elderly people are mostly not included in randomised clinical trials, both because of age and

co-morbidity [24], and as a result information on efficacy, optimum drug dosages and toxicity

are frequently lacking in this vulnerable age group [25,26]. Also, in daily clinical practice the

circumstances, under which drugs are used, especially in the elderly, differ from those in ran-

domised clinical trials. In view of the considerations mentioned above, prescribers need a tho-

rough understanding of the risks, benefits and consequences of drug therapy in the elderly,

13

Out l ine

In this introductory chapter, the objective and contents of the thesis are outlined in section

1.1. Section 1.2 gives an overview of the setting in which the studies presented in this thesis

were carried out. A description is given of both the medical and pharmaceutical setting for

ambulatory elderly and nursing home residents aged 65 and over in the Netherlands.

12

outline of the thesis are given. Furthermore, background information is given into the setting

of the studies presented in this thesis. Chapter 2 describes drug use and determinants of drug

use in nursing home patients and elderly outpatients. A description is given of the pharmacy

prescription data that were collected from 6 nursing homes. These data comprised the main

dataset used throughout this thesis. Several drug utilisation studies were performed with this

dataset. Prescribing indicators were used with the aim to study drug use systematically.

Furthermore, using prescription data from the InterAction database, we investigated drug use

in elderly outpatients, focusing on psychotropics and non-steroidal anti-inflammatory drugs.

Chapter 3 describes studies that focus on the outcome of drug use in both nursing home

patients and elderly outpatients. In chapter 4, the main findings of the studies are discussed

and suggestions for clinical practice are given.

15

especially the frail elderly. By studying the uses and effects of drugs in this population in daily

clinical practice such insight into these matters can be obtained. In particular in nursing homes,

where frail elderly people reside with often high levels of drug use, and where adverse drug

effects may lead to serious clinical consequences, it is useful to perform pharmacoepidemiolo-

gic studies. In the United States many such studies have been carried out, addressing three

main issues: the measurement, determinants, and outcomes of drug use. The structure and

organisation of Dutch nursing homes and the type of residents that is cared for, differ conside-

rably from those in other European countries and the United States [27]. As a consequence, the

concern that has been expressed in the United States regarding the excessive and inapprop-

riate use of drugs in nursing homes, especially psychotropics [28], cannot automatically be

applied to Dutch nursing homes.

Relatively few drug utilisation studies in Dutch nursing homes have been carried out.

Several reasons may account for this fact. First, the medical speciality ‘nursing home medicine’

is the youngest of the 34 medical specialities in the Netherlands and has been acknowledged

as an official medical speciality since 1990. In view of this short history, the extent of research

in this speciality is relatively small. Second, the nursing home population in the Netherlands

mainly consists of frail elderly patients. In this group ethical aspects, although this is not typi-

cally a Dutch issue, therefore may play a role. Ethical and practical questions arise when the

value of certain medical interventions is assessed, such as the withdrawal of benzodiazepines.

Furthermore, studies on clinical relevant outcomes are often lacking due to ethical considera-

tions. A third reason why studies on drug use in Dutch nursing homes have not been performed

more extensively, is that until the 1990s medication use on individual patient level was only

registered in the medical chart and not in pharmacy computer systems. Only in the last decade,

medication registration on individual patient level in computerised systems is more common.

Performing pharmacoepidemiologic studies, which often requires large datasets, has therefore

not been feasible until recent years.

Object ive and out l ine of the thes is

The objective of this thesis is to provide insight in drug use, determinants of drug use and

outcomes of drug use in the elderly and in particular in the frail elderly that reside in Dutch nur-

sing homes. The studies that provide this information may lead to a better understanding of the

risks of drug use in the (institutionalised) elderly and may serve as a starting point to improve

prescribing practices. Furthermore, the results of these investigations may provide the tool for

monitoring individual patients at risk for drug-related problems. In chapter 1 the objective and

14

27 Ribbe MW. Care for the elderly: the role of the nursing home in the Dutch health care system. Int Psychogeriatr

1993; 5: 213-22.

28 Beers MH, Ouslander JG, Fingold SF, Morgenstern H, Reuben DB, Rogers W et al. Inappropriate medication pre-

scribing in skilled-nursing facilities. Ann Intern Med 1992; 151: 1825-31.

17

References

1 Heerdink ER. Clustering of drug use in the elderly: population-based studies into prevalence and outcomes

[Thesis]. University of Utrecht, 1995.

2 Lau HS. Drug related problems in the elderly: studies on occurrence and interventions [Thesis]. University of

Utrecht, 1998.

3 Veehof LJG. Polypharmacy in the elderly [Thesis]. University of Groningen, 1999.

4 Avorn J. Gurwitz JH. Drug use in the nursing home. Ann Intern Med 1995; 123: 195-204.

5 Tinke JL. Farmacie in cijfers (SFK): Geneesmiddelgebruik door ouderen. Pharm Weekbl 2001; 136: 589.

6 Rochon PA, Gurwitz JH. Optimising drug treatment for elderly people: the prescribing cascade.

BMJ 1997; 315: 1096-9.

7 Van den Bemt PMLA, Egberts ACG, De Jong-van den Berg LTW, Brouwers JRBJ. Drug related problems in

hospitalised patients: a review. Drug Saf 2000; 22: 321-33.

8 Hanlon JT, Maher RL, Lindblad CI, Ruby CM, Twersky J, Cohen HJ, Schmader KE. Comparison of methods for

detecting potential adverse drug events in frail elderly inpatients and outpatients. Am J Health-Syst Pharm

2001; 58: 1622-6.

9 Beyth RJ, Shorr RI. Epidemiology of adverse drug reactions in the elderly by drug class.

Drugs Aging 1999; 14: 231-9.

10 Gurwitz JH, Avorn J. The ambiguous relation between aging and adverse drug reactions.

Ann Intern Med 1991; 114: 956-66.

11 Rochon PA, Gurwitz JH. Prescribing for seniors. Neither too much, nor too little. JAMA 1999; 282: 113-5.

12 Redelmeier DA, Tan SH, Booth GL. The treatment of unrelated disorders in patients with chronic medical

diseases. N Engl J Med 1998; 338: 1516-20.

13 Avorn J. Improving drug use in elderly patients: getting to the next level. JAMA 2001; 22: 2866-8.

14 NIH Consensus Conference. Diagnosis and treatment of depression in late life. JAMA 1992; 268: 1018-24.

15 Atkin PA, Veitch PC, Veitch EM, Ogle SJ. The epidemiology of serious adverse drug reactions among the elderly.

Drugs Aging 1999; 14: 141-52.

16 Van Kraaij DJW, Haagsma CJ, Go IH, Gribnau FWJ. Drug use and adverse drug reactions in 105 elderly patients

admitted to a general medical ward. Neth J Med 1994; 44: 166-73.

17 Bero LA, Lipton HL, Bird JA. Characterization of geriatric drug-related hospital readmissions.

Med Care 1991; 29: 989-1003.

18 Seppälä M, Sourander L. A practical guide to prescribing in nursing homes. Avoiding the pitfalls. Drugs Aging

1995; 6: 426-35.

19 Monette J, Gurwitz JH, Avorn J. Epidemiology of adverse drug events in the nursing home setting. Drugs Aging

1995; 7: 203-11.

20 Beers MH, Ouslander JG. Risk factors in geriatric drug prescribing. A practical guide to avoid problems.

Drugs 1989; 37: 105-112.

21 Hughes SG. Prescribing for the elderly: why do we need to exercise caution? Br J Clin Pharmacol 1998; 46: 531-3.

22 Denham MJ, Barnett NL. Drug therapy and the older person. Role of the pharmacist. Drug Saf 1998; 19: 243-50.

23 Kinirons MT, Crome P. Clinical pharmacokinetic considerations in the elderly. An update. Clin Pharmacokinet

1997; 33: 302-12.

24 Zhan C, Sangl J, Bierman AS, Miller MR, Friedman B, Wickizer SW, Meyer GS. Potentially inappropriate medica-

tion use in the community-dwelling elderly. JAMA 2001; 22: 2823-9.

25 Turnheim K. Drug dosage in the elderly. Is it rational? Drugs Aging 1998; 13: 357-79

26 Avorn J. Including elderly people in clinical trials. BMJ 1997; 315: 1033-4.

16

made services for elderly people in need of continuous medical care, such as a daily need of

dressing renewals or continuous pain relief. When the need for medical care becomes such that

people cannot live totally in their own home anymore, admission into residential homes or

nursing homes is possible on strict medical indication and depending on the amount of care

needed. Regional pre-selection boards carry out assessment of eligibility for care in nursing

homes and residential homes. However, no measuring instrument has been able to assess the

need for care objectively and in a standardised way that is universally applicable [6].

Inst i tut ional ised e lder ly

In the Netherlands, of the people aged 65 and over 5.5% live in residential homes and

2.7% live in nursing homes [1]. Residential homes offer board, lodging and care to elderly who

are no longer able to cope on their own. These homes do not offer nursing. Medical care is pro-

vided by the residents’ general practitioner and medication is provided by the residents’ com-

munity pharmacy. There are 1425 residential homes in the Netherlands, with a total of 117,500

beds (55 per 1000 inhabitants aged 65 and over) [1]. The Dutch nursing home is a healthcare

institution for chronically ill persons in need of permanent medical and paramedical attention

and complex nursing care and is compared to skilled nursing facilities in the US [1]. The type of

care can be characterised as continuous, long-term, systematic and multidisciplinary [7].

Several features make them different from nursing homes in other countries. First, clear crite-

ria based on medical diagnoses, activities of daily living, behavioural characteristics and men-

tal functioning for nursing home admission exist. On the basis of these criteria, a distinction is

made between eligibility for admission and whether people should be admitted to either a

somatic nursing home or a psychogeriatric nursing home. Residents with predominantly psy-

chogeriatric disorders (mainly dementia) are cared for in psychogeriatric nursing homes, whe-

reas in somatic nursing homes people with predominantly somatic disorders, such as

Parkinson’s disease or diabetes mellitus, reside. Often, psychogeriatric care and somatic care is

provided in the same nursing home, and the division between the two types of care is between

wards. Second, the way care is provided in Dutch nursing homes is markedly different from

other countries. For instance, specially trained nursing home physicians provide medical care

on a continuous basis. This medical speciality requires a two-year postgraduate academic edu-

cation. In other countries, medical care in nursing homes is provided by general practitioners,

mostly on an on-demand basis and not continuously. Furthermore, in the Netherlands, care is

provided by a multidisciplinary team, in which nursing home physicians (1 full-time doctor per

100 residents), nurses, physical therapists, speech therapists, and psychologists closely colla-

19

1.2 Sett ing

Ambulatory e lder ly

The Dutch health care system is geared to facilitate elderly people (aged 65 and older) to

live at home for as long and as independently as possible. Eighty-two percent of the people

aged 65 and over live independently in the community [1]. Most people have healthcare insu-

rance, which covers the costs of primary care and medication as well as the costs of in-hospi-

tal and outpatient treatment. In addition, every Dutch citizen is insured under the Exceptional

Medical Expenses Act (AWBZ). This act provides the general public with insurance for health

risks not covered by normal healthcare insurance, such as admission to nursing homes and

residential homes and the costs of home care (e.g. meals-on-wheels and household help) [1].

General medical care is provided through general practitioners and pharmacies providing pri-

mary care services. In the Netherlands, more than 90% of healthcare problems are dealt with

in primary care. General practitioners can refer patients to 143 hospitals (with almost 60,000

beds). Every academic hospital and each large teaching hospital (approximately 17 in total) has

a specialised geriatric department, together with clinical geriatric teaching and research facili-

ties [1]. In other non-teaching hospitals separate ‘Geriatrische Afdeling Algemeen Ziekenhuis’

(GAAZ)-departments are present, providing medical care for geriatric patients.

In recent years, several initiatives have been issued towards improvements in quality of life

of elderly people. Increasingly, attention has been given to quality of drug use. In Dutch com-

munity pharmacies, pharmaceutical care is gradually implemented in daily community practi-

ce [2]. Community pharmacists have addressed some of the problems of medication use in the

elderly and have given suggestions for improvement [3]. Internationally, several studies have

focused on pharmaceutical care. In a multicenter international study performed in 7 European

countries, it was found that community pharmacy-based provision of pharmaceutical care

improved the well-being of elderly patients [4]. Patients reported better control of their medi-

cal conditions and cost savings associated with pharmaceutical care provision were observed

in most countries. In a 12-month controlled intervention study it was found that community

pharmacy-based interventions improved lung function, health-related quality of life, and self-

management in asthma patients [5].

Several secondary medical care services are provided on an outpatient basis, such as the

home visits of the Outpatient Thrombosis Services throughout the country to monitor oral anti-

coagulant therapy by measuring the prothrombin times. Home care organisations provide tailor-

18

References

1 Hoek JF, Penninx BW, Ligthart GJ, Ribbe MW. Health care for older persons, a country profile: the Netherlands.

J Am Geriatr Soc 2000; 48: 214-7.

2 Van Mil FJW, Tromp DF, McElnay JC, De Jong-van den Berg LTW, Vos R. Development of pharmaceutical care in

The Netherlands: pharmacy’s contemporary focus on the patient. J Am Pharm Assoc (Wash) 1999; 39: 395-401.

3 Van Mil JWF. Results of pharmaceutical care in the elderly, the OMA study. In Van Mil JWF. Pharmaceutical care,

the furure of pharmacy. Theory, research and practice [Thesis]. University of Groningen, 2000.

4 Bernsten C, Bjorkman I, Caramona M, Crealey G, Frokjaer B, Grundberger E, et al. (Pharmaceutical care of the

elderly in Europe Research Group (PEER)). Improving the well-being of elderly patients via community phar-

macy-based provision of pharmaceutical care. Drugs Aging 2001; 18: 63-77.

5 Schulz M, Verheyen F, Muhlig S, Muller JM, Muhlbauer K, Knop-Scheickert E, Petermann F, Bergmann KC.

Pharmaceutical care services for asthma patients: a controlled intervention study.

J Clin Pharmacol 2001; 41: 668-76.

6 Dijkstra GJ. De indicatiestelling voor verzorgingshuizen en verpleeghuizen [Proefschrift]. Rijksuniversiteit

Groningen, 2001.

7 Ribbe MW. Care for the elderly: the role of the nursing home in the Dutch health care system.

Int Psychogeriatr 1993; 5: 213-22.

8 Health Care Inspectorate. Medication distribution in nursing homes (in Dutch). Ministery of Health, Welfare and

Sports, The Hague, the Netherlands, 1997.

9 Dutch Association for Nursing Home Care (Vereniging voor Verpleeghuiszorg (NVVz) in samenwerking met

KNMP, NVZA, NVVA): Guideline Pharmaceutical Care in Nursing Homes (in Dutch). Utrecht, 1998.

10 Gurwitz JH, Soumerai SB, Avorn J. Improving medication prescribing and utilization in the nursing home.

J Am Geriatr Soc 1990; 38: 542-52.

11 Shorr RI, Fought RL, Ray WA. Changes in antipsychotic drug use in nursing homes during implementation of

the OBRA-87 regulations. JAMA 1994; 271: 358-62.

12 Schmidt I, Claesson CB, Westerholm B, Nilsson LG, Svarstad BL. The impact of regular mulidisciplinary team

interventions on psychotropic prescribing in Swedish nursing homes. J Am Geriatr Soc 1998; 46: 77-82.

13 Schmidt I, Claesson CB, Westerholm B, Svarstad BL. Resident characteristics and organizational factors

influencing the quality of drug use in Swedish nursing homes. Soc Sci Med 1998; 47: 961-71.

14 Lunn J, Chan K, Donoghue J, Riley B, Walley T. A study of the appropriateness of prescribing in nursing homes.

Int J Pharm Pract 1997; 5: 6-10.

15 Janzig JGE, Van ‘t Hof MA, Zitman FG. Drug use and cognitive function in residents of homes for the elderly.

Pharm World Sci 1997; 19: 279-82.

16 Wagner C, Van der Wal G, Groenewegen PP, De Bakker DH. The effectiveness of quality systems in nursing

homes: a review. Qual Health Care 2001; 10: 211-7.

21

borate. Medical specialists, such as neurologists and psychiatrists, provide specialised medical

care on consult basis. There are 330 nursing homes in the Netherlands, with a total of 57,000

beds (27 per 1000 inhabitants aged 65 and older) [1]. As mentioned earlier, all nursing home

admissions are financed by the AWBZ. The AWBZ reimburses about 98% of all nursing homes

expenses. In addition to addressing financial stability issues, this act regulates the care stan-

dards for this sector, monitored by regional health inspectors. Another act that relates to quali-

ty issues in nursing homes is the ‘Care Institutions Quality Act’ (Kwaliteitswet Zorginstellingen).

This act aims at adequate quality assurance in health care institutions.

Either community pharmacists or hospital pharmacists can provide the distribution of

medicines in Dutch nursing homes. To investigate the quality of the medication distribution

process and other pharmaceutical activities, in 1997 the Dutch Health Care Inspectorate held a

survey among 33 nursing homes. About half of these nursing homes were served by a hospital

pharmacy, the other half by a community pharmacy. It was concluded that quality aspects

should be more incorporated in the medication distribution processes. The pharmacist should

play a more profound role in the pharmaceutical care to nursing home residents, such as par-

ticipation in drug and therapeutics committees, evaluation of prescribing on patient level, and

regular updates of drug formularies [8]. As a result of this survey, a ‘Guideline Pharmaceutical

Care in Nursing Homes’ was drafted in 1998 [9]. This guideline describes how pharmaceutical

care by the pharmacist in nursing homes is best performed, and serves as a guideline for imple-

menting quality assurance processes. For both economic and therapeutic reasons, a drug for-

mulary is used in almost every Dutch nursing home, constituting of a limitative list of preferred

medications. In several Dutch nursing homes, drug therapy issues are regularly discussed in

Pharmacotherapy Discussion Groups, in which both nursing home physicians and pharmacists

participate. For certification, these meetings should be held at least 6 times a year. Studies

addressing the quality of pharmaceutical care in nursing homes have not been performed

extensively. Mainly this work has been carried out in the United States [10,11], but in recent

years European studies also focus on this issue [12-16].

20

Chapter 2

Drug ut i l i sat ion studiesin the e lder ly

2322

2.1 Background

Introduct ion

Several reviews on drug utilisation and quality of prescribing in elderly outpatients have

been carried out demonstrating high levels of drug use and a considerable number of patients

receiving inappropriate medications [1]. In the Netherlands, drug utilisation studies have

demonstrated that the prevalence of multiple chronic drug use in elderly outpatients is high [2].

Recently, the extent of polypharmacy in Dutch elderly outpatients was investigated and it was

concluded that the problem of polypharmacy was less than in other countries [3]. While drug

use in elderly outpatients has been extensively studied, less is known about drug utilisation

and quality of prescribing in Dutch nursing homes. In the next paragraphs, a review is given of

some major studies on drug use and quality of prescribing in nursing homes.

Drug ut i l i sat ion studies in nurs ing homes

Drug use in nursing homes has been reviewed in several articles [4-7] and a number of drug

utilisation studies have been performed in nursing homes [8-24]. Many of these focus on cer-

tain drugs or drug groups, in particular psychotropics [21-24]. In table 1 a summary of some

major studies on drug use in nursing homes during 1990-2000 is given. The number of nursing

home residents reviewed varies from 60 to 1854, and the average number of drugs prescribed

per resident varies from 2.5 to 8.8. These differences might be due to the different lengths of

time for which drug use was studied. In many of these studies, data on drug use was collected

by medical chart review or by interviews. Three drug utilisation studies have been published

from the Netherlands [12,13,18]. Troost and co-workers used hospital-dispensing data to assess

the defined daily dosages (DDDs) per 100 beddays [12]. Drug use was not analysed on an indi-

vidual patient level. Koopmans and colleagues studied drug use of psychogeriatric patients [13],

and Verkoulen-Wijers [18] reviewed drug use of 60 patients using a cross-sectional design. Two

other Dutch studies were published before 1990 [25,26]. Merkus and co-workers [25] investi-

gated drug use at admission and 3.5 years later in 112 patients aged 75 and over. In people res-

iding in homes providing somatic care, medication increased from 3.8 to 5 prescriptions per

day, and in people residing in nursing homes providing psychogeriatric care, medication incre-

ased from 2.7 to 3.4 prescriptions per day. Van Zuylen [26] showed that 80% of the patients

(n=198) used 2 to 7 drugs (on average 4.4). The drugs used most frequently were psychotropics,

analgesics, cardiovascular drugs, diuretics and laxatives.

25

Out l ine

In this chapter, we present several drug utilisation studies in the elderly. This chapter comp-

rises two parts: in part one (section 2.1 through 2.5) drug utilisation studies in nursing home

residents are described, and in part two (section 2.6 and 2.7) drug use in ambulatory elderly is

described. In section 2.6, both ambulatory elderly and nursing home residents are studied

separately.

In section 2.1 some major international studies on drug use and quality of prescribing in

nursing homes are summarised and the Dutch studies that have been published on this subject

are reviewed. Section 2.2 describes how pharmacy prescription data from nursing homes have

been used to build a nursing home database with the aim to perform drug utilisation and risk

assessment studies. This section also presents a suggestion for the criteria such data should

meet to perform pharmacoepidemiological studies. In section 2.3, a drug utilisation study in 6

nursing homes is presented. In this study, detailed information is given on determinants of drug

use in a cohort of 2,355 residents. The results of this study served as a starting point to inves-

tigate some of the issues on prescribing quality in more detail in the next sections. Section 2.4

describes the occurrence and nature of potential drug-drug interactions in the same cohort of

nursing home patients. Section 2.5 presents a pilot study that used several prescribing indica-

tors, based on the studies in 2.3 and 2.4, to evaluate drug prescribing in Dutch nursing homes.

Section 2.6 presents a study on the concomitant use of benzodiazepines and antidepres-

sants in elderly outpatients and nursing home patients. We assessed whether differences

between co-prescribing with tricyclic antidepressants and selective serotonine re-uptake inhi-

bitors existed. In section 2.7, the concomitant use of non-steroidal anti-inflammatory drugs and

gastroprotective drugs, as an example of a beneficial drug-drug combination was investigated.

24

In conclusion, several studies have investigated the rate of drug use in nursing homes.

Many used cross-sectional designs, and only few studies used longitudinal prescription data to

evaluate effects over time. The number of nursing home residents included in each study varied

considerably. The studies carried out in the Netherlands involved relatively small numbers of

residents. To obtain more insight into drug use in Dutch nursing homes, a study among 2,355

residents was carried out (section 2.3).

Qual i ty of prescr ib ing in nurs ing homes

In particular in the United States, much attention has been given to rational and approp-

riate prescribing in nursing homes. In 1987, the Omnibus Reconciliation Act (OBRA ‘87) was pas-

sed as a result of increasing public concern about the overuse of neuroloptics in nursing homes.

These federal regulations were established to ensure appropriate care in US nursing homes,

and in particular to reduce the unnecessary prescribing of antipsychotic medications [27]. Since

OBRA ‘87 numerous studies have addressed the problem of unnecessary or inappropriate drug

use in nursing homes in the US and other countries and have investigated opportunities for

improvement [28-32]. Several tools have been developed to assess the appropriateness of pre-

scribing [33]. Many of these tools were developed for assessing medication appropriateness in

elderly outpatients, and not nursing homes. Most tools used information on clinical status or

diagnoses. In 1991, Beers and colleagues developed a set of criteria for identifying inapprop-

riate medication use in nursing home residents in the US [34]. These criteria, based on expert

consensus, consisted of a list of 23 medications that should be avoided and 13 medication doses,

frequencies, or duration of prescriptions that generally should not be exceeded. The list of inap-

propriate drugs needs to be updated periodically, as it was based on state of knowledge at that

time (1989). An update, incorporating clinical information, was published in 1997 [35].

Internationally, several studies have used Beers’ criteria to assess medication appropriateness.

In 1992, a prospective cohort study was carried out using Beers criteria among 12 nursing

homes in the US (n=1106) [8]. It was found that 40% of the residents received at least one

inappropriate prescription, and 10% received two or more inappropriate medications concur-

rently. Female residents and residents of large nursing homes were at the highest risk of recei-

ving inappropriate medications. In 1996, Gupta and co-workers [36] investigated the associa-

tion between costs and inappropriate prescribing using a slightly modified version of Beers cri-

teria in a retrospective, cross-sectional study among nursing home residents (Medicaid benefi-

ciaries; n=19,932). It was found that cost of pharmaceutical services for a resident was positi-

vely correlated with the number of different inappropriate drugs prescribed, number of physi-

2726

Table 1: A review of studies on drug use in nursing homes

Ref Country N Study design Average Main findings(year of number ofpublication) prescriptions

8 US 81 Analysis of medication 4.7 (entry) Increase in medication prescribing(1990) prescribed at entry and 6.2 (3 months) due to increase in prn* medications

after 3 months

9 Belgium 198 Cross-sectional chart 4.5 Drug use increased with age but (1992) review in sample of stabilises after 80 yrs

patients aged ≥ 62 yrs

10 US 1106 Cross-sectional study 7.2 Inappropriate prescribing was common(1992) using pharmacy data

of patients ≥ 65 yrs

11 US 120 Effect of drug regimen 7.2-5.3 Pharmacist reduced medication(1992) review on medication use in nursing home by computerised

use drug regimen review

12 Netherlands 147 Analysis of pharmacy - Drug use in general according to(1993) dispensing data drug formulary

(number of DDDs/100 beddays) during 1991-1992

13 Netherlands 390 Retrospective chart review 2.5-2.9 Increase in drug use (mainly laxatives)(1994) among psychogeriatric after admission to nursing home

patients

14 South Africa 85 Intervention study; review 4.8 41% reduction in incidence of(1996) of medical profile charts polypharmacy due to

for drug related problems pharmacist intervention

15 UK 101 Intervention study; review 6.5 53% of residents showed ≥ 1 (1997) of medical records to study inappropriate prescription

appropriateness of prescribing

16 Sweden 1854 Intervention study; review 7.7 No effect of intervention on overall(1998) of medical records use of medication

17 Australia 998 Cross-sectional survey of 6.6 Nursing home culture exerts a major(1998) medication use by chart (prescribed) influence over prescribing and

review 4.8 administration of medications(administered)

18 Netherlands 60 Cross-sectional survey of 4.9 Laxatives, diuretics, psycholeptics and(1999) medication use by chart including prn*; antithrombotics used most frequently

review 4,6 excluding prn

19 Sweden 1001 Cross-sectional survey in 3.4 High prevalence of drug use and(1999) ('87-'89); cohort of subject aged 81 ('87-'89); polypharmacy in very elderly

681 and over 4.6('94-'96) ('94-'96)

20 Sweden 1549 Controlled intervention 8.0 (1995) Intervention homes showed higher(2000) study in 36 nursing homes 8.8 (1998) quality of drug use

# expressed as ‘time pattern of drug exposure’: uninterrupted period of usage of the same drug* prn = pro re nata = as required

formed on the appropriateness or quality of prescribing in nursing homes. In section 2.4 to 2.7

we have made an attempt to identify suboptimal prescribing by using several aspects of pre-

scribing indicators from the international literature, such as the occurrence of drug-drug inter-

actions, the concomitant use of certain drug groups, and selection of formulary drug use.

References

1 Aparasu RR, Mort JR. Inappropriate prescribing for the elderly: Beers criteria-based review. Ann Pharmacother

2000; 34: 338-46.

2 Heerdink ER. Clustering of drug use in the elderly: population-based studies into prevalence and outcomes


3 Veehof LJG. Polypharmacy in the elderly [Thesis]. University of Groningen, 1999.

4 Lamy P. Institutionalisation and drug use in older adults in the US. Drugs Aging 1993; 3: 232-7.

5 Avorn J, Gurwitz JH. Drug use in the nursing home. Ann Intern Med 1995; 123: 195-204.

6 Furniss L,Craig SKL, Burns A. Medication use in nursing homes for elderly people. Int J Geriatr Psychiatry 1998;

13: 433-9.

7 Gurwitz JH, Soumarai SB, Avorn J. Improving medication prescribing and utilization in the nursing home.

J Am Geriatr Soc 1990; 38: 542-52.

8 Wayne SJ, Rhyne RL, Stratton M. Longitudinal prescribing patterns in a nursing home population.

J Am Geriatr Soc 1990; 40: 53-6.

9 Vander Stichele RH, Mestdagh J, Van Haecht CH, De Potter B, Bogaert MG. Medication utilization and patient

information in homes for the aged. Eur J Clin Pharmacol 1992; 3: 319-21.

10 Beers MH, Ouslander JG, Fingold SF, Morgenstern H, Reuben DB, Rogers W et al. Inappropriate medication

prescribing in skilled-nursing facilities. Ann Intern Med 1992; 151: 1825-31.

11 Laucka PV, Hoffman NB. Decreasing medication use in a nursing-home patient-care unit. Am J Hosp Pharm

1992; 49: 96-9.

12 Troost SJ, Smit LH, Wentink DH, Neef C. Gebruikscijfers van geneesmiddelen uit het verpleeghuis ‘De Cromhoff’

te Enschede [Drug utilization figures in a nursing home]. Pharm Weekbl 1993; 128: 1511-6.

13 Koopmans RT, de Haan HH, van den Hoogen HJ, Gribnau FW, Hekster YA, van Weel C. Veranderingen in

geneesmiddelgebruik tijdens een verblijf in een psychogeriatrisch verpleeghuis. [Changes in drug use during

stay in a psychogeriatric nursing home]. Ned Tijdschr Geneeskd 1994; 138: 1122-6.

14 Belligan M, Wiseman IC. Pharmacist intervention in an elderly care facility. Int J Pharm Pract 1996; 4: 25-9.



16 Claesson CB, Schmidt IK. Drug use in Swedish nursing homes. Clin Drug Invest 1998; 16: 441-52.

17 Roberts MS, King M, Stokes JA, Lynne TA, Bonner CJ, McCarthy S, Wilson A, Glasziou P, Pugh John W. Medication

prescribing and administration in nursing homes. Age Ageing 1998; 27: 385-92.

18 Verkoulen-Wijers MJ, Koopmans RTCM, Schimmel W. Geneesmiddelengebruik van somatische verpleeghuis-

bewoners en verzorgingshuisbewoners in Zorgcentrum Tilburg-zuid. Tijdschr Verpleeghuisgeneeskd 1999; 1: 4-7.

19 Giron MST, Claesson C, Thorslund M, Oke T, Winblad B, Fastbom J. Drug use patterns in a very elderly

population. Clin Drug Invest 1999; 17: 389-98.

20 Schmidt IK, Fastbom J. Quality of drug use in Swedish nursing homes. A follow-up study. Clin Drug Invest 2000;

20: 433-446.

29

cians and pharmacies used and geographic region. In 1997, Lunn and colleagues [37] developed

a set of 18 explicit criteria, based on expert’s opinions, for identifying inappropriate prescribing

in 101 nursing home residents in the UK. Fifty-three percent of the residents had one or more

inappropriate prescriptions. Medicines most frequently associated with inappropriate prescri-

bing were cardiovascular drugs and drugs for the central nervous system [37]. In Sweden, an

attempt was made to identify inappropriate psychotropic drug prescribing in 33 Swedish nur-

sing homes (1823 residents) by Schmidt and colleagues [38]. They developed a list of 13 criteria,

based on Swedish guidelines for measuring excessive use of psychotropic drug use in the elder-

ly. A wide variability in the appropriateness of drug use among the 33 nursing homes was

found. Drug use for a majority of residents had deviated from one or more of the drug use cri-

teria. Overall, concern was expressed about the quality of drug prescribing practices in Swedish

nursing homes. The drug use criteria developed addressed three main issues: deviation of

documentation on indication of drug (for example: antipsychotic drug prescribed; psychotic

diagnosis absent), deviations of drug choice (for example: prescribing of non recommended

hypnotic drug), and deviations of excess (for example: prescribing of more than 2 psychotropic

drugs concomitantly). Again, these criteria used in part clinical information. In another Swedish

study [20] prescribing indicators were used to assess the effect of an intervention aimed at

improving drug use through improved teamwork among physicians, pharmacists, nurses and

nurses’ assistants. In the intervention homes, a higher quality of drug use was seen after the

intervention, compared with the control homes. The prescribing indicators addressed three

issues: indication of drug (for example: antipsychotic drug prescribed without diagnoses of psy-

chotic symptom); quantity of drug use (more than 2 psychotropic drugs) and potential interac-

tions.

In conclusion, studies on prescribing appropriateness in nursing homes have shown that a

considerable proportion of the nursing home residents received inappropriate prescribing. The

way prescribing appropriateness was assessed differed among the studies as different prescri-

bing or quality indicators are in use. Prescribing indicators used in one health care system are

not automatically applicable to other health care systems due to differences in national phar-

macotherapy guidelines and drug formularies. Furthermore, for many prescribing indicators

information on clinical status, such as laboratory results or diagnoses are needed, which makes

them unsuitable to apply them based solely on pharmacy prescription data. To assess medica-

tion appropriateness in nursing homes, indicators should be used that reflect deviations from

national pharmacotherapy guidelines and drug formularies. As these vary with time and loca-

tion, prescribing indicators are dependent on the country or even the institution where the

research is performed. To our knowledge, in Dutch nursing homes, no studies have been per-

28

2.2 Use of hospi ta l pharmacy data in pharmacoepidemiologic research in nurs ing homes

K.N. van Dijk 1, 2, C.S. de Vries 3, J.R.B.J. Brouwers 1, 2, L.T.W. de Jong-van den Berg 1

1 Department of Social Pharmacy, Pharmacoepidemiology and Pharmacotherapy, University

Centre for Pharmacy, Groningen University Institute for Drug Exploration (GUIDE),

Groningen, the Netherlands2 Clinical Pharmacy Department, Medical Centre Leeuwarden, the Netherlands3 Department of Pharmacoepidemiology, Postgraduate Medical School, University of Surrey,

Guildford, United Kingdom

31

21 Nygaard HA, Naik M. Use of psychotropic drugs in homes for the aged in Bergen, Norway: a comperative study.

Norwegian Journal of Epidemiology 1998; 8: 133-8.

22 McGrath AM, Jackson GA. Survey of neuroleptic prescribing in residents of nursing homes in Glasgow.

BMJ 1996; 312: 611-2.

23 Cohen-Mansfield J, Lipson S, Gruber-Baldini AL, Farley J, Woosley R. Longitudinal prescribing pattern of

psychotropic drugs in nursing home residents. Exp Clin Psychopharmacol 1996; 4: 224-33.

24 Koopmans RTCM, van Rossum JM, van den Hoogen HJM, Hekster YA, Willekens-Bogaers MAJH, van Weel C.

Psychotropic drug use in a group of Dutch nursing home patients with dementia: many users, long-term use,

but low doses. Pharm World Sci 1996; 18: 42-7.

25 Merkus JW, Deurenberg HP, Pollmann AM. Het geneesmiddelgebruik in 3 verpleeghuizen voor lichamelijk

zieken [Drug utilization in 3 nursing homes for somatic patients]. Tijdschr Gerontol Geriatr 1985; 16: 87-95.

26 Van Zuylen C, Oostendorp FMGM, van Beusekom BR, Cools HJM, Bolk JH, Ligthart GJ. Toenemend genees-

middelgebruik in het verpleeghuis [Increasing use of medication in nursing homes]. Ned Tijdschr Geneesk

1988; 132: 1692-5.

27 Llorente MD, Olsen EJ, Leyva O, Silverman MA, Lewis JE, Rivero J. Use of antipsychotic drugs in nursing homes:

current compliance with OBRA regulations. J Am Geriatr Soc 1998; 46: 198-201.

28 Semla TP, Palla K, Poddig P, Brauner DJ. Effect of the Omnibus Reconciliation Act 1987 on antipsychotic

prescribing in nursing home residents. J Am Geriatr Soc 1994; 42: 648-52.

29 Garrard J, Makris L, Dunham T, Heston LL, Cooper S, Ratner ER, et al. Evaluation of neuroleptic drug use by

nursing home elderly under proposed Medicare and Medicaid regulations. JAMA 1991; 265: 463-7.

30 Rovner BW, Edelman BA, Cox MP, Shmuely Y. The impact of antipsychotic drug regulations on psychotropic

prescribing practices in nursing homes. Am J Psychiatr 1992; 149: 1390-2.

31 Schorr RI, Fought RL, Ray WA. Changes in antipsychotic drug use in nursing homes during implementation of

the OBRA-87 regulations. JAMA 1994; 271: 358-62.

32 Ray WA, Taylor JA, Meador KG, Lichtenstein MJ, Griffin MR, Fought R, et al. Reducing antipsychotic drug use in

nursing homes: a controlled trial of provider education. Arch Intern Med 1993; 153: 713-21.

33 Shelton PS, Fritsch MA, Scott MA. Assessing the medication appropriateness in the elderly. A review of available

measures. Drugs Aging 2000, 16: 437-50.

34 Beers MH, Ouslander JG, Rollingher I, Reuben DB, Brooks J, Beck J. Explicit criteria for determining

inappropriate medication use in nursing homes. Arch Intern Med 1991; 151: 1825-32.

35 Beers MH. Explicit criteria for determining potentially inappropriate medication use by the elderly.

Arch Intern Med 1997; 157: 1531-6.

36 Gupta S, Rappaport HM, Bennett LT. Inappropriate drug prescribing and related outcomes for elderly Medicaid

benificiaries residing in nursing homes. Clin Therap 1996; 18: 183-96.



38 Schmidt IK, Claesson CB, Westerholm B, Svarstad BL. Resident characteristics and organizational factors


30

Data avai lable

Medicat ion order data

The study was carried out in six nursing homes for long-term care in the Netherlands.

These nursing homes each had a bed capacity that varied between 90 and 225, and each was

served by one of three hospital pharmacies. In the drug dispensing system in the nursing

homes studied, all drugs that were dispensed to residents were registered in the hospital phar-

macy computer system (Apho-data®). Routinely, the nursing home physicians have to write

medication orders for each change in the drug regimen, such as a dosage or frequency change,

the discontinuation of the drug and, obviously, the start of a new drug. In this way, any changes

in medication are updated routinely on a daily basis in the hospital pharmacy computer system

and a complete medication history is kept for each individual resident. Every day, nurses dis-

pense medication to individual nursing home residents on the basis of the information recor-

ded in the computer system (drug, dosage, route and time of administration). As a consequen-

ce, the recording of actual drug use can be considered very accurate.

SIVIS data

In the Netherlands, a national information system on nursing home residents (SIVIS) is

operational. The SIVIS database consists of anonymous administrative, nursing and medical

data collected on individual nursing home residents. More than 70% of the Dutch nursing

homes contribute to the SIVIS database. The SIVIS data are collected quarterly in each nursing

home by the nursing home staff and, subsequently, these data are anonymised and registered

nationally. The SIVIS morbidity classification is derived from the International Classification of

Diseases (ICD-9) and contains 99 diagnoses. For each resident, a maximum of three diagnoses

can be registered each time.

Data col lect ion


We collected medication order data of all nursing home residents for a 2-year period

between 01-10-1993 and 01-10-1995. Patient data for each nursing home resident included date

of birth, gender, date of admission to the nursing home, date of discharge, code of the nursing

home, and code of the attending nursing home physician. For every prescription a start date

and, if known, an end date was registered, enabling us to calculate the actual duration of drug

use. Furthermore, the name of the drug, daily dose, dosage interval and route and time of

33

Introduct ion

In pharmacoepidemiology, computerised medication databases are a useful tool. To study

drug use in primary health care in the Netherlands several computerised medication databa-

ses are available, such as the PHARMO-database [1], the InterAction-database [2], and the

IPCI-database [3]. Both the InterAction-database and the PHARMO-database are based on

community pharmacy records. Because patients generally attend only one pharmacy and com-

munity pharmacies are entirely automated, almost complete individual medication profiles

covering several years are available. Community pharmacy records have been reported to be a

reliable source of drug exposure [4]. The PHARMO database is record-linked with hospital

diagnoses data. The IPCI-database is based on general practitioners’ (GPs) prescribing data and

as such the information recorded is comparable with the GPRD-database in the UK [5], al-

though the IPCI-database is much smaller [3]. A variety of drug utilisation and drug safety stu-

dies have been performed using these Dutch databases [6-9].

In the Netherlands, drug utilisation studies in secondary (hospitals) and tertiary (nursing

homes) health care are performed sparingly, partly because computerised individual hospital

medication records have not been available until the mid 90s. Internationally, in particular in

the United States and Canada, several drug utilisation studies have been carried out in nursing

homes using large administrative databases [10-14]. In the US a Minimum Data Set, a data col-

lection instrument containing more than 300 demographic, clinical and treatment variables, is

used for each patient that is admitted to a certified nursing home [15]. Although not all drug use

is included in these databases, the fact that they contain medical diagnoses makes them very

valuable. In a large drug utilisation study in nursing homes in Sweden [16] medication use was

recorded by research nurses. In the Netherlands, computerised medication databases in nur-

sing homes have not been available until recently. To our knowledge, these databases have not

yet been used in drug utilisation studies [17,18]. The applicability of computerised medication

databases in pharmacoepidemiology in this population is therefore unknown. Ideally, both

detailed information on patients’ drug use and health status should be available for drug

research in nursing homes.

In this study we describe how medication order data from nursing homes have been used

to build a nursing home database with the aim to perform drug utilisation and risk assessment

studies. Furthermore, we make a suggestion for the criteria such data should meet to perform

such studies.

32

Data ver i f i cat ion


Accuracy of the medication database was verified by comparing the medication history

from the original pharmacy computer system with the medication history in the newly built

database for 10 randomly selected patients for each nursing home (total of 60 patients). In view

of the drug distribution system described above, no other way of verifying the medication data-

base was possible (for example, patient interviews, or nurses’ charts). We did not find any dis-

crepancies.

SIVIS data

To verify the accuracy and completeness of the SIVIS diagnoses data we performed a sen-

sitivity and specificity analysis in which we compared medication order records and SIVIS diag-

noses records for diabetes mellitus and Parkinson’s disease. These diseases were chosen in

view of the clearly defined drug groups that are used for these disorders; namely antiglycaemic

drugs (ATC code A10) and anti-Parkinson drugs (ATC code N04). The validation of SIVIS diag-

noses data resulted in three outcomes: registered SIVIS diagnosis and registered proxy drug

use (true positive), no SIVIS diagnosis that could relate to proxy drug dispensed (false negati-

ve), and SIVIS diagnosis registered and no proxy drug prescribed (false positive). The positive

predictive value of the SIVIS diagnosis was calculated as the proportion of patients with the

SIVIS diagnosis and using the proxy drug among all patients who are registered with the SIVIS

diagnosis. Sensitivity of the SIVIS diagnosis was the proportion of patients registered with the

SIVIS diagnosis and using the proxy drug among the total number of patients who were pre-

scribed the proxy drug. Specificity of the SIVIS diagnosis was the proportion of patients pre-

scribed non-proxy drugs for other indications than the SIVIS diagnosis among the total num-

ber of patients prescribed non-proxy drugs. The results of these analyses are given in table 1.

35

administration were recorded. Each medication record was given a drug ID (KNMP-number, a

unique code for every formulation and pack size of a drug that is on the market in the

Netherlands). Each drug ID and hence every medication record was record-linked with the

national reference drug database of the Royal Dutch Association for the Advancement of

Pharmacy (KNMP, The Hague), to collect drug-specific information such as Anatomical

Therapeutic Chemical (ATC) codes and defined daily dosages (DDD).

SIVIS data and record- l inkage

The medication database was record-linked with the SIVIS database, by matching 3 patient

characteristics (date of birth, gender and nursing home code), in order to collect data on mor-

bidity, type of care (somatic or psychogeriatric) and mobility (2 categories) from the SIVIS data-

base. Fourteen percent of the patient records could not be record-linked to the SIVIS database.

In figure 1 the structure of the databases is given.

34

Patient database

Patient IDDate of birth*Gender*Nursing home code*Date of admission Date of discharge (if applicable)

SIVIS database

Date of birth*Gender*Nursing home code*Diagnosis codeType of careMobility code

Reference drug database (Z-index)

Drug IDATC codeDDD value

Medication database

Patient IDDrug nameDrug IDMedication IDStart date Stop dateDaily dosageDosage intervalPrescriber code

Figure 1: Structure of the databases (*: used for record-linkage)

Table 1a: Results of the sensitivity and specificity analysis of the SIVIS diagnoses data for diabetes mellitus

Pharmacy record ATC code A10

+ - Total

SIVIS diagnose + 152 24 176 Positive predictive value:

diabetes mellitus 152 / 176 = 0.86

- 207 1972 2179 Sensitivity:

152 / 359 = 0.42

Total 359 1996 2355 Specificity:

1972 / 1996 = 0.99

37

Study populat ion

The source population consisted of 2,966 patients. Of this population, the admission date

was missing in 264 patients (9%) and the date of discharge was missing in 268 patients (10%).

If the admission date was unknown, it was assumed to be the first medication start date. If the

date of discharge was unknown, it was assumed to be the last end date of the medication used

or as the last day of the study period (01-10-1995), whichever was the earliest. 194 of 2,966

(6.5%) people were excluded because they were younger than 65 years (average age 52.6

years (SD 10.6) versus 82.1 years (SD 7.4) (p<0.05)). Compared with the remaining 2,772 patients,

the excluded population included more men (44.8% versus 28.9% (p<0.05). Of the 2,772

patients left, 394 (14.2%) were excluded because they could not be linked to the SIVIS databa-

se. Compared with the remaining 2,378 residents, they had the same average age (82.5 versus

82.0 years), and there were slightly more women present (74.1% versus 70.6%; p>0.05). These

patients were equally distributed over the six nursing homes. Of the 2,378 patients left, 23 (1%)

were excluded because of missing data (for example, the period of stay could not be calcula-

ted). Compared with the remaining 2,355 patients, this population included more men (47.8%

versus 29.3%; p<0.05). The final study population consisted of 2,355 patients. Figure 2 shows

how the study population was constructed. In table 2 patient characteristics are given for the

study population.

36

Table 1b: Results of the sensitivity and specificity analysis of the SIVIS diagnoses data for M. Parkinson

Pharmacy record ATC code N04

+ - Total

SIVIS diagnose + 115 36 151 Positive predictive value:

M. Parkinson 115 / 151 = 0.76

- 76 2128 2204 Sensitivity:

115 / 191 = 0.60

Total 191 2164 2355 Specificity:

2128 / 2164 = 0.98

Source population: N=2,966

Age: 80.1 yrs (SD 10.6)Gender: 70.0% female# of medication records: 42,648 (14.4)¶

N=2,772

Age: 82.1 (SD 7.4)Gender: 71.1% female# of medication records: 39,793 (14.4)¶

N=2,378


Study population: N=2,355


N=194 (6.5%)


N=394 (14.2%)


N=23 (1%)

Age: 79.4 (SD 8.1)Gender: 52.2% female# of medication records: 270 (11.7)¶

< 65 yrs

no SIVIS-data

missing data

Figure 2: Construction of the study population

Discuss ion

This study describes how hospital pharmacy data and SIVIS data can be used to build a

database to perform pharmacoepidemiological studies of drug utilisation and drug safety in

nursing home residents [19-21]. Recommendations are summarised in table 3. We encountered

several pitfalls during our study, which are described below.

Data availability. Detailed information on individual patients’ drug utilisation profiles has

to be available on a continuous basis. In most pharmacy computer systems, it is now common

use to register this information on a continuous basis for longer periods of time (e.g. 5-10

years). In this study we retrieved data for a two-year period. Ideally, it should be possible to

retrieve data at any time for any period of time. Because information on morbidity and mobili-

ty was needed to perform one study [19], we collected these data from the national nursing

home information database (SIVIS). At the moment, SIVIS data are the only source of automa-

ted diagnoses data available.

Data collection. An important aspect of the data collection is that the data are adequately

anonymised. Ideally, the pharmacy computer system should contain special ‘export files’ by

which drug utilisation data can be anonymously collected on an individual level. Collection of

the SIVIS data was done by record-linkage. Because the patient identifier used in the pharma-

cy records was different from the patient identifier in the SIVIS database, we performed a

record-linkage on three variables: date of birth, sex and nursing home code. Fourteen percent

of the residents could not be linked. This could have been due to the fact that these patients

had not been registered in the SIVIS database yet.

Data completeness. Completeness and accuracy highly depends on the organisation and

structure of the (hospital) pharmacy concerned. For example, adequate quality control proce-

dures should ensure that all necessary information is recorded in the pharmacy computer sys-

tem. The professional organisation of Dutch hospital pharmacies facilitates adequate quality

control by both pharmacy technicians (who generally record the data), and hospital pharma-

cists (who generally check and supervise). We found that in 10% of the patients the date of dis-

charge was missing, and that in 9% of the patients the admission date was missing. These per-

centages could be decreased when it is made impossible not to fill in certain database fields at

the data entry stage (by pharmacy technicians). Use of over-the-counter (OTC) medication has

not been included in our database. We expect this to be relatively low due to practical reasons

such as immobility of the residents and continuous medical attention by both nursing and

medical staff and the possibility to receive drugs that are available OTC via prescription by the

nursing home physician.

39

Medicat ion use

Initially, for the source population, an average number of 14.4 medication records per

patient were registered in the medication database. In figure 2 the number of medication

records for each patient cohort is given. In the final study population, the average number of

(registered) medication records per patient was 14.8. The relatively high number of medication

records recorded per patient can be explained by the fact that each change (dosage or fre-

quency change) in the therapeutic regimen is recorded separately [19].

38

Table 2: Characteristics of the study population (n=2,355)

Variable Number of residents (% of total)

Age 82 (SD 7.3)

Gender

Male 689 (29%)

Female 1666 (71%)

Type of nursing

Psychogeriatric 700 (30%)

Somatic 1609 (68%)

Not known 46 (2%)

Morbidity

Parkinson's disease 151 (6%)

Diabetes mellitus 176 (7%)

Depression 40 (2%)

Dementia 689 (29%)

Mobility

Mobile 1370 (58%)

Immobile 985 (42%)

Number of different medications

0-5 626 (27%)

6-10 969 (41%)

> 10 760 (32%)

Average number of different medications per day per resident 4.9

Average number of different medication (based on ATC-codes; fifth level) per resident during study residence in nursing home 8.9 (SD 4.9)

41

Data verification. In our study we used medication order data, which form the basis for the

dispensing of medication by nurses to individual nursing home residents. One important con-

dition of this dispensing system is that every change in drug therapy is known to the hospital

pharmacy. Unlike for community pharmacy records [4], the accuracy and completeness of the

hospital pharmacy data has not been verified. In the nursing homes studied, nursing home phy-

sicians are obliged to record every change in drug therapy and send these changes (by fax)

immediately to the hospital pharmacy. As a result, hospital pharmacy data are a reliable sour-

ce of drug exposure of nursing home residents. We did not find any discrepancies when we

verified the medication data. From the SIVIS sensitivity and specificity analysis it was found

that the positive predictive value (PPV) of the SIVIS diagnoses diabetes mellitus was 0.86, and

0.76 for M. Parkinson. These data indicate that 14% and 24% of the residents who are diagno-

sed in the SIVIS-database with diabetes mellitus and M. Parkinson respectively, do not use

medication that is commonly used for these disorders. There is a discrepancy between the

medication records and the SIVIS diagnoses data. This suggests that the prevalence of these

disorders may be overestimated when using SIVIS diagnoses data alone. The sensitivity valu-

es of the SIVIS diagnoses are relatively low, indicating that 58% and 40% of the residents

using drugs for diabetes mellitus and M. Parkinson respectively, are not registered as suffering

from these disorders in the SIVIS database. This suggests a large underestimation of the per-

centage of residents with diabetes mellitus and M. Parkinson when SIVIS data alone are used,

when pharmacy data are considered the ‘gold standard’. A reason for this could be the fact that

these diagnoses are not one of the three diagnoses registered, however in view of the disabling

consequences of M. Parkinson this seems unlikely. Further research is needed into the reasons

for these discrepancies. The specificity values of the SIVIS diagnoses diabetes mellitus and M.

Parkinson are high (0.99 and 0.98, respectively), indicating that no SIVIS diagnoses are recor-

ded for residents that do not use medication for these diagnoses. We suggest combining both

pharmacy data and SIVIS-diagnoses data to get a more reliable estimate of the true prevalen-

ce. Recently, Van de Vijver and co-workers showed that antiparkinsonian drugs in pharmacy

records in ambulatory patients aged 55 and older can be used as a reliable marker for M.

Parkinson [22]. Misclassification of drug use could occur when medication that is prescribed

and hence registered in the pharmacy computer system, is not actually consumed by the

patient (either by non-compliance or by errors in the distribution process). However, a pilot

study has shown that medication compliance in Dutch nursing homes is generally high (99%)

[23]. Because we did not have information on the use of OTC-medication, this may lead to an

underestimation of exposure to OTC drugs and hence misclassification.

40

Tab

le 3

: Dis

crep

anci

es b

etw

een

data

requ

ired

for p

harm

acoe

pide

mio

logi

cal r

esea

rch

and

data

ava

ilabl

e in

hos

pita

l pha

rmac

y pr

escr

iptio

n da

taba

ses

Aspe

cts

Requ

irem

ents

for

Data

ava

ilabl

e in

hos

pita

l Re

com

men

datio

nsph

arm

acoe

pide

mio

logi

c re

sear

chph

arm

acy

pres

crip

tion

data

base

us

ed in

cur

rent

stu

dy

Popu

latio

n

Sam

ple

size

N

eeds

to b

e su

ffici

ently

larg

e to

By g

roup

ing

patie

nt d

ata

from

Colle

ct d

ata

from

sev

eral

nur

sing

hom

espe

rfor

m p

harm

acoe

pide

mio

logi

c di

ffere

nt h

ospi

tal p

harm

acie

s an

(and

hen

ce h

ospi

tal p

harm

acie

s).

stud

ies.

Min

imum

requ

ired

size

ad

equa

te s

ampl

e si

ze (n

=2,

355)

was

Hos

pita

l pha

rmac

ies

shou

ld p

refe

rabl

yw

ill v

ary

depe

ndin

g on

the

rese

arch

ob

tain

ed fo

r the

pur

pose

s of

this

stu

dy.

use

the

sam

e ph

arm

acy

com

pute

r sys

tem

ques

tion.

(or p

rovi

de th

e sa

me

expo

rt re

cord

).

Conf

iden

tialit

yD

ata

need

to b

e an

onym

ous.

Patie

nt d

ata

wer

e no

t ano

nym

ousl

y U

se e

ncry

pted

pat

ient

iden

tifie

rsre

cord

ed in

ori

gina

l dat

abas

e, b

ut th

ey

(key

mus

t rem

ain

in o

rigi

nal d

atab

ase)

.w

ere

retr

ieve

d an

onym

ousl

y.

Gen

eral

isab

ility

/Po

pula

tion

has

to b

e re

pres

enta

tive

Popu

latio

n ha

d a

sim

ilar d

istr

ibut

ion

Nee

ds to

be

take

n in

to a

ccou

nt;

com

pare

re

pres

enta

tiven

ess

for D

utch

nur

sing

hom

e po

pula

tion.

of a

ge a

nd s

ex a

nd ty

pe o

f car

e al

thou

ghw

ith n

atio

nal d

ata

avai

labl

e fr

om S

IVIS

.th

ey w

ere

slig

htly

old

er th

an th

e D

utch

nurs

ing

hom

e po

pula

tion.

Drug

use

Accu

racy

, Al

l dru

gs th

at a

re re

gist

ered

in

Dat

a en

try

was

acc

urat

e, a

lthou

gh s

ome

Ensu

re d

ata

com

plet

enes

s an

d ac

cura

teco

mpl

eten

ess

med

icat

ion

data

base

for i

ndiv

idua

l in

accu

rate

dat

a an

d om

issi

ons

have

data

ent

ry b

y tr

aini

ng p

harm

acy

tech

nici

ans.

patie

nts

are

used

by

thos

e pa

tient

s,

been

foun

d (10

%).

and

all d

rugs

that

are

use

d by

in

divi

dual

pat

ient

s ar

e re

gist

ered

.

Det

aile

d in

form

atio

n on

W

hich

dru

g, in

whi

ch d

osag

e, in

whi

ch

Det

aile

d in

form

atio

n av

aila

ble,

dat

aTo

obt

ain

spec

ific

drug

dat

a su

ch a

sdr

ug, d

ose

and

dura

tion

freq

uenc

y is

take

n fo

r how

long

and

on

DD

D-a

nd A

TC-c

odes

wer

e ob

tain

edD

DD

-val

ues

and

ATC-

code

s, li

nkag

e w

ithby

whi

ch ro

ute

of a

dmin

istr

atio

n.th

roug

h re

cord

-lin

kage

with

nat

iona

l na

tiona

l dru

g da

taba

se (Z

-ind

ex) c

ould

dr

ug d

atab

ase.

be

nee

ded.

Cont

inui

tyLo

ngitu

dina

l dat

a ar

e re

quire

d to

Lo

ngitu

dina

l dat

a w

ere

avai

labl

eEn

sure

con

tinui

ty b

y ke

epin

g m

edic

atio

nst

udy

drug

use

ove

r a ti

me

sequ

ence

.fo

r 2-y

ear s

tudy

per

iod.

hist

orie

s fo

r lon

g pe

riod

of t

ime.

Dru

g in

take

In

form

atio

n re

gard

ing

med

icat

ion

Med

icat

ion

com

plia

nce

ensu

red

Non

eco

mpl

ianc

e is

requ

ired.

by n

ursi

ng h

ome

staf

f.

Oth

er s

ourc

es

Use

of O

TC-d

rugs

kno

wn.

OT

C-dr

ug u

se w

as c

onsi

dere

d ne

glig

ible

Non

e

References

1 Herings RMC. PHARMO: a record linkage system for postmarketing surveillance of prescription drugs in the

Netherlands [Thesis]. University of Utrecht, 1993.

2 Tobi H, Van den Berg PB, De Jong-van den Berg LTW The InterAction database: synergy of science and practice

in pharmacy. In: Brause RW, Hanisch E (ed) Medical data analysis: first international symposium;

proceedings/ISMDA. Berlin: Springer, 2000: 206-11

3 Integrated Primary Care Information (IPCI), Rotterdam, The Netherlands.

4 Lau HS, De Boer A, Beuning KS, Porsius A. Validation of pharmacy records in drug exposure assessment.

J Clin Epidemiol 1997; 50: 619-25.

5 Garcia Rodriguez LA, Perez Gutthann S. Use of the UK general practice database for pharmacoepidemiology.

Br J Clin Pharmacol 1998; 45: 419-25.

6 Leufkens HGM. Pharmacy records in pharmacoepidemiology: studies on antiinflammatory and antirheumatic

drugs [Thesis]. University of Utrecht, 1990.

7 De Vries CS. Collaboration in healthcare, the tango to drug safety [Thesis]. University of Groningen, 1998.

8 Heerdink ER. Clustering of drug use in the elderly. Population-based studies into prevalence and outcomes


9 Lau HS. Drug related problems in the elderly [Thesis]. University of Utrecht, 1998.


11 Pies R, Gorman JM, Gorenstein EE, Katz IR, Rovner BW, Schneider L, Avorn J. Use of psychoactive drugs in

nursing homes. New Engl J Med 1992; 327: 1392-3.

12 Llorente MD, Olsen EJ, Leyva O, Silverman MA, Lewis JE, Rivero J. Use of antipsychotic drugs in nursing homes:

current compliance with OBRA regulations. J Am Geriatr Soc 1998; 46: 198-201.

13 Johnson RE, Vollmer WM. Comparing sources of drug data about the elderly. J Am Geriatr Soc 1991; 39: 1079-84.

14 Lamy PP. Institutionalisation and drug use in older aldults in the US. Drugs Aging 1993: 3: 232-7.

15 Gambassi G, Landi F, Peng L, Brostrup-Jensen C, Calore K, Hiris J, et al. Validity of diagnostic and drug data in

standardized nursing home residents assessments. Potential for geriatric pharmacoepidemiology.

Med Care 1998; 36: 167-79.


17 Koopmans RT, Van Rossum JM, Van den Hoogen HJ, Hekster YA, Willekes-Bogaers MA, Van Weel C.



18 Koopmans RT, de Vaan HH, Van den Hoogen HJ, Gribnau FW, Hekster YA, Van Weel C. Changes in drug use

during a stay in a psychogeriatric nursing home (in Dutch). Ned Tijdschr Geneeskd 1994; 138: 1122-6.

19 Van Dijk KN, De Vries CS, Van den Berg PB, Brouwers JRBJ, De Jong-van den Berg LTW. Drug utilisation in

Dutch nursing homes. Eur J Clin Pharmacol 2000; 55: 765-71.

20 Van Dijk KN, De Vries CS, Van den Berg PB, Dijkema AM, Brouwers JRBJ, De Jong- van den Berg LTW.

Constipation as an adverse effect of drug use in nursing home patients: an overestimated risk. Br J Clin

Pharmacol 1998; 46: 255-61.

21 Van Dijk KN, De Vries CS, Van den Berg PB, Brouwers JRBJ, De Jong-van den Berg LTW. Occurrence of potential

drug-drug interactions in nursing home residents. Int J Pharm Pract 2001; 9: 45-52.

22 Van de Vijver DAMC, Stricker BHCh, Breteler MMB, Roos RAC, Porsius AJ, De Boer A. Evaluation of antiparkin-

sonian drugs in pharmacy records as a marker for Parkinson’s disease. Pharm World Sci 2001; 23: 148-52.

23 Boogaard RG, Dercksen H. Therapietrouw in het verpleeghuis. Een onderzoek in het kader van de opleiding tot

verpleeghuisarts. Verpleeghuis Het Zonnehuis, Zwolle, 1997 (report in Dutch).

4342

Tab

le 3

(co

nt.

): D

iscr

epan

cies

bet

wee

n da

ta re

quire

d fo

r pha

rmac

oepi

dem

iolo

gica

l res

earc

h an

d da

ta a

vaila

ble

in h

ospi

tal p

harm

acy

pres

crip

tion

data

base

s

Aspe

cts

Requ

irem

ents

for

Data

ava

ilabl

e in

hos

pita

l Re

com

men

datio

nsph

arm

acoe

pide

mio

logi

c re

sear

chph

arm

acy

pres

crip

tion

data

base

us

ed in

cur

rent

stu

dy

Outc

omes

Clin

ical

sta

tus

Info

rmat

ion

on d

iagn

oses

, bio

chem

ical

N

ot a

vaila

ble

Obt

ain

thes

e da

ta th

roug

h re

cord

-pa

ram

eter

s, la

bora

tory

test

resu

lts,

linka

ge w

ith p

hysi

cian

com

pute

r sys

tem

,dr

ug b

lood

leve

ls, d

iagn

ostic

test

resu

lts.

and

othe

r com

pute

r sys

tem

as

appl

icab

le.

Util

isat

ion

of h

ealth

car

eIn

form

atio

n on

util

isat

ion

of h

ealth

N

ot a

vaila

ble

Obt

ain

data

on

hosp

ital s

tay

by re

cord

-ca

re, s

uch

as h

ospi

tal c

are.

linka

ge.

Qua

lity

of li

feIn

form

atio

n on

qua

lity

of li

fe.

Not

ava

ilabl

eTh

is in

form

atio

n co

uld

be c

olle

cted

on

an

ad h

oc b

asis

.

Conf

ound

ers

Dem

ogra

phic

sAg

e, s

ex, s

ocio

-eco

nom

ic s

tatu

s,

Age

and

sex

Reco

rd d

ata

on li

fe s

tyle

fact

ors

such

as

smok

ing.

life

styl

e fa

ctor

s.

Indi

catio

n fo

r dru

g us

eIn

dica

tion

for d

rug

use,

N

ot a

vaila

ble

Reco

rd d

ata

on in

dica

tion

and

dise

ase

seve

rity

.di

seas

e-se

veri

ty.

Co-m

orbi

dity

Dis

ease

s th

at m

ight

be

asso

ciat

ed

Avai

labl

e th

roug

h re

cord

-lin

kage

Veri

fy p

harm

acy

data

and

SIV

IS d

ata

agai

nst

with

bot

h ex

posu

re a

nd o

utco

me.

with

SIV

IS-d

atab

ase.

each

oth

er. C

arry

out

a v

alid

atio

n st

udy

of e

ach

agai

nst m

edic

al c

hart

s.

Abstract

Objective: To quantify and evaluate drug utilisation in a sample of Dutch nursing homes.

Methods: A retrospective analysis of computerised medication data of 2,355 residents aged 65

and over of six nursing homes in the Netherlands was performed. For each therapeutic drug

group, the number of users was determined. The ten therapeutic groups used most frequently

were investigated further. For these, patient characteristics, use of therapeutic subgroups, the

average daily dosages, and the chronicity of drug use were determined. Chronicity was expres-

sed as the percentage of treatment days divided by the number of residents’ days in the nur-

sing home.

Results: During the study period, 89%, 77% and 56% of the study population used a drug

from ATC main group N (central nervous system), A (alimentary tract and metabolism) and C

(cardiovascular system), respectively. Eight of the ten therapeutic drug groups prescribed most

frequently were used for more than 50% of the time. In particular psycholeptic drugs, diurec-

tics, and laxatives were used chronically (83%, 81%, and 80% of the nursing home stay,

respectively). Except for a few drug groups such as laxatives and diuretics, the prescribed daily

dosages were relatively low. Twenty-eight percent of the residents received loop diuretics;

these were prescribed in relatively high dosages.

Conclusion: Drug utilisation in the nursing home was high and many drugs were used chroni-

cally. In view of the risk of possible side effects and drug-drug interactions, the prescribing and

dosage of psycholeptic drugs, laxatives, loop diuretics and ulcer-healing drugs should be re-

evaluated.

45

2.3 Drug ut i l i sat ion in Dutch nurs inghomes

K.N. van Dijk 1,2, C.S. de Vries 1,3, P.B. van den Berg 1, J.R.B.J. Brouwers 1,2,L.T.W. de Jong-van den Berg 1

1 Department of Social Pharmacy, Pharmacoepidemiology and Pharmacotherapy, Groningen

University Institute for Drug Exploration (GUIDE), University Centre for Pharmacy,

Groningen, the Netherlands2 Department of Clinical Pharmacy, Medical Centre Leeuwarden, Leeuwarden, the Netherlands3 Department of Pharmacoepidemiology, Postgraduate Medical School, University of Surrey,


A slightly modified version has been published in

European Journal of Clinical Pharmacology 2000; 55: 765-71

44

Methods

Study populat ion

The study was conducted in residents aged 65 and over from six nursing homes in the nor-

thern part of the Netherlands. Compared to a national register on nursing home residents in

the Netherlands (SIVIS), the study population had a similar distribution of gender and type of

care, although nursing home residents in our study were slightly older (average age 82 in our

study versus 79 in the SIVIS data) than the Dutch nursing home population. Physician care was

provided by nursing home physicians who give medical care on a daily basis. From interview

data it was shown that nursing, physician, and pharmacist care, as well as food and fluid inta-

ke were comparable between the nursing homes [31]. The source population consisted of a

dynamic cohort of 2,772 residents who were present at any time during the two-year study

period from 1 October 1993 to 1 October 1995. The average annual turnover rate (mostly due to

mortality) was 40%.

Data col lect ion

For each resident, pharmacy records and individual resident characteristics were collected.

Pharmacy data included the generic name, strength, dosage, the frequency of use, the route of

administration, and the duration of drug use (in days) of each prescribed drug. This type of data

has been demonstrated to be an adequate source of information of the prescription drugs taken

by elderly people [32,33]. Furthermore, age, gender, date of admission, and date of discharge

were collected from the pharmacy records. Pharmacy records were linked with a national

information system on nursing homes (SIVIS) [34], to collect patient specific data regarding

morbidity (three diagnoses per resident), and the type of care: psychogeriatric or somatic care.

Somatic residents usually are characterised by serious physiological chronic disorders, such as

Parkinson’s disease, diabetes mellitus and rheumatoid arthritis. Psychogeriatric residents are

characterised by serious mental or psychiatric disorders such as dementia. As a consequence,

pharmacotherapy between these groups will differ. Patient specific SIVIS data are collected

four times a year by the nursing home staff, and subsequently these data are registered ano-

nymously on a national basis.

Data handl ing

Residents whose pharmacy records could not be linked to data from the SIVIS system

(14.2%) and subsequently, residents whose period of residence could not be defined (1%) were

excluded. The excluded residents did not differ from the other residents with respect to age,

47

Introduct ion

Many studies have reported drug utilisation in nursing home residents to be disturbingly

high [1-3]. It has been estimated that of prescribed medication, approximately 25-40% is con-

sumed by people over 65 years of age [4,5]. When elderly people are admitted to nursing

homes, their medication use often increases [6,7]. In part this may be due to the reasons for

admission to the nursing home, however other studies suggest that the ‘prescribing cascade’,

i.e. the prescription of a new drug to counteract the side effects of another drug, could also be

a reason [8]. This may lead to an increased frequency of unnecessary polypharmacy and of pos-

sible adverse drug effects [9-11]. Since elderly people often suffer from multiple disorders and

consequently use many drugs, this population receives much emphasis. Also, the nursing home

environment provides an excellent opportunity for comprehensive drug regimen review [12].

Special programs have been developed and implemented to improve drug utilisation in the

community-dwelling elderly [13] and in nursing home residents [14,15]. Furthermore, several

groups have developed or applied prescribing indicators with the aim to create means to sys-

tematically assess medication appropriateness and thus a starting point to improve prescribing

[16-23]. Many of these indicators were developed for assessing medication appropriateness in

elderly outpatients, rather than nursing home residents. Beers’ criteria have been developed

for elderly outpatients in the United States and are not necessarily representative for European

pharmacotherapy standards. For example, according to Beers’ criteria trimethobenzamide,

reserpine, chlorpropamide and dicyclomine use is discouraged [16,23] but in many European

countries these drugs are not even available. Therefore, to measure prescribing of these drugs

in not a good indicator for medication appropriateness in Europe. Typically European issues,

for example, are the prescribing of diuretics, psychotropics and laxatives [24]. To improve pre-

scribing locally, an overview of current prescribing is needed. In Europe, only few studies have

been undertaken to describe and evaluate drug use in the elderly [25] and in nursing homes

[24,26-29] and few settings have the availability of computerised medication data to evaluate

drug utilisation continuously.

The purpose of this study was to evaluate and quantify drug utilisation in six Dutch nursing

homes. In relation to patient characteristics and published guidelines [30], frequency of drug

use, drug groups, daily dosages of relevant drug groups and chronicity of drug use are studied.

This study highlights areas that deserve attention in prescribing practices in nursing homes.

After identifying the pharmacotherapeutic domains that comprise the most serious problem

areas, feedback may be provided to improve pharmacotherapy.

46

gender or type of care. This resulted in a final study population of 2,355 residents.

Drugs were classified according to the Anatomical Therapeutic Chemical (ATC) classifica-

tion system [35]. Dosages were calculated as Defined Daily Dose (DDD) values, defined by the

World Health Organisation (WHO) [36], and Prescribed Daily Dose (PDD) values were deter-

mined for each prescribed drug. The PDD value is the daily dosage of drugs, expressed as the

daily number of DDDs that is actually prescribed; i.e. the daily dosage (in mg) divided by the

DDD (in mg) [37]. The fact that drug volume, or quantities of drug use, are expressed as a com-

parable unit of pharmacological efficacy rather than for example the number of milligrams or

prescriptions dispensed enables us to compare daily dosages between different drugs [38].

When the PDD is greater than 1, the prescribed daily dosage is relatively high relative to WHO

guidelines. A PDD less than 1 indicates a relatively low daily dosage. Drugs that do not have a

DDD value in the ATC-coding system such as injections and dermatological preparations and

drugs that were prescribed ‘as needed’ (5% of all prescriptions), were left out when average

PDD values were calculated.

Drug ut i l i sat ion

Initially, all prescribed main drug groups were analysed on the levels of ATC anatomical

main group (first level) and therapeutic group (second level). The number of users and the

number of users who used these drugs for more than 50% of the nursing home stay were

determined for these drug groups. Of the ten therapeutic categories prescribed most frequent-

ly, further analyses were performed on the level of therapeutic ATC subgroups (third level). For

each therapeutic ATC subgroup the chronicity of pharmacotherapy and average and median

PDD values were studied. To obtain insight in the duration of drug use in relation to the period

of stay in the nursing home, for each resident and for each drug the chronicity of pharmaco-

therapy was expressed as the number of drug utilisation days divided by the number of resi-

dents’ days in the nursing home. Software used was SPSS, version 6.01.

Results

Populat ion character is t i cs

The study population consisted of 2,355 nursing home residents. The mean age of the study

population was 82 years (SD 7.3). Figure 1 represents the duration of stay in the nursing home

for individuals in the study population. The average number of different drugs (based on ATC

codes; fifth level) per person was 8.9 (SD 4.9) during the total residence time; the average num-

ber of different drugs per day was 4.9. Other population characteristics are given in table 1.

48

250

200

150

100

50

0

300

350

400

0-2526-5051-7

576

-100

101-1

25126-15

0151-1

75176

-200201-2

25226-250251-2

75276

-300

301-3

2532

6-350

351-3

75376

-400401-4

25426-450451-4

75476

-500501-5

25526-450551-4

75576

-600601-5

25626-550651-6

75676

-700

701-7

30

Figure 1: Duration of stay in nursing home study population (n=2,355)


Variable Number of residents

(n) (%)

Age (years)

65-69 127 5

70-74 288 12

75-79 404 17

80-84 608 26

85-89 574 24

90-94 274 12

> 94 80 3

Gender

Male 689 29

Female 1666 71

Type of care

Psychogeriatric 700 30

Somatic 1609 68

Not known 46 2

Morbidity

Parkinson’s disease 151 6

Diabetes mellitus 176 7

Dementia 689 29

Depression 40 2

Duration of stay in the nursing home (months)

< 1 349 15

1-5 990 42

6-11 306 13

12-17 205 9

≥ 18 505 21

49

51

Drug consumption

Figure 2 represents the number of residents on drugs from the ATC anatomical main groups,

and the number of residents who received these drugs for more than 50% of their stay in the

nursing home. Anatomical main groups A (alimentary tract and metabolism), B (blood and

bloodforming organs), C (cardiovascular system), J (general anti-infective agents for systemic

use) and N (central nervous system) were the drug groups from which drugs were prescribed

most frequently and, with the exception of antibiotics, on average these drugs were used for

more than 50% of the stay in the nursing home. Gastrointestinal drugs used most frequently

were laxatives and ulcer-healing drugs: in this population, 56% used laxatives and 24% used

anti-ulcer medication. Drug use from group B consists mainly of anti-thrombotic drugs and

iron, drug use from group C consists mainly of cardiac drugs such as digoxin, anti-arrythmic

agents, diuretics and angiotensin-converting enzyme (ACE) inhibitors. Non-steroidal anti-

inflammatory drugs (NSAIDs) are the only drugs that were used from group M. Drug use from

group N consists mainly of analgesics, such as paracetamol and aspirin, benzodiazepines, and

antipsychotics. For the ten ATC main groups prescribed most frequently, drug consumption data

are represented in descending order of frequency in table 2. From this table, differences in pre-

scribing can be seen by gender (column 4 and 5) and by the type of care that residents receive

(psychogeriatric or somatic, column 6 and 7). With respect to gender, striking differences occur

in the utilisation of antipsychotic agents (ATC code N05A; more frequently used by men) and

pain medication (ATC code M01A and N02B; more frequently used by women). With respect to

the type of care, differences occur in psycholeptic prescribing for psychogeriatric and somatic

residents: psychogeriatric residents receive more antipsychotics (59% versus 23%), whereas

somatic residents receive more benzodiazepines (ATC code N05C). Next, we studied chronicity

and the daily dosage of drugs. The average chronicity of treatment (column 8) and average and

median PDD values (column 9 and 10) are given. From column 8, it can be seen that in these

nursing homes, most drugs are used for more than 50% of the nursing home stay. This inclu-

des diuretics and anti-coagulant drugs, but also laxatives, hypnotics and sedatives: on an aver-

age, 74% of the nursing home population use a psycholeptic drug for an average of 83% of

their stay in the nursing home. Adjusting these results for the duration of stay in the nursing

home had very little, if any, effect on the results. Leaving out the residents who only stayed for

a very short period in the nursing home did not alter the results significantly (both statistical

and clinical). Finally, drugs are given in relatively low dosages (column 9 and 10). Exceptions

are laxative drugs, antibiotics, diuretics, NSAIDs, and proton-pump inhibitors. In general the

mean PDD values are higher than the median PDD values.

50

0

500

1000

1500

2000

2500

A B C G H J L M N P R S V

ATC main group

Num

ber o

f res

iden

ts

Figure 2: Number of residents for each anatomical therapeutic chemical (ATC) main group; Dark bars show the number

of residents who use the drug for more than 50% of their duration of stay. A alimentary tract and

meta–bolism; B blood and bloodforming organs; C cardiovascular system, G genitourinary system; H hormo-

nal preparations; J general anti-infectives; L antineoplastic agents; M musculoskeletal system; N central ner-

vous system; R respiratory system; S sensory organs; V various agents

Discuss ion

In the nursing homes studied there is much long-term use of various medication types.

Generally, prescribed daily dosages are low, which indicates that prescribers are aware of the

pharmacokinetic and pharmacodynamic changes in the elderly. To our knowledge, no studies

have been performed that study chronicity of drug use for all drug groups in the nursing home.

This study shows that -with the exception of antibiotics- once they are on a certain drug, most

nursing home residents use these drugs for more than half of their stay in the nursing home.

Even though the authors are unaware of the complete morbidity of the individual residents and

utilisation is studied on a population basis, this raises the question whether the residents’ need

for a drug is re-evaluated from time to time. Psychogeriatric residents use approximately the

same drugs as residents in somatic care except for antipsychotics, anticoagulants and ulcer-

healing drugs. This indicates that, next to their psychological complaints, they have similar

somatic complaints as the other nursing home residents. The results with regard to the drug

groups studied will be reviewed in the following sections.

Psycholept ic drugs

Psycholeptic drugs are used by 74% of the nursing home residents. Since the nursing

homes offer psychogeriatric care facilities, utilisation of this drug group was expected in this

patient group. However, drug use from this group is more than 70% in both psychogeriatric and

in somatic care residents. A major part of this drug group are the benzodiazepines, especially

the hypnotics and sedatives. In accordance with prescribing guidelines in the Netherlands, the

prescribed daily dosages are about half the advised adult dosage [39]. However, Dutch prescri-

bing guidelines also state that it is suboptimal to prescribe these drugs for more than 30 days,

since the drugs become less effective but the adverse effects remain [39]. This part of the gui-

delines clearly is not followed. In view of previous studies that have demonstrated an increased

risk of falls and fractures in the elderly [40,41] and in view of other adverse effects such as

drowsiness that directly influence the residents’ quality of life, it would pay to re-evaluate the

need for these drugs in those residents. However, we are aware of the fact that withdrawal from

benzodiazepines is often difficult to achieve [42].

Laxat ive drugs

Constipation is a well-known complaint in the elderly. It could result from an impaired

bowel function due to immobility, decreased fluid- and fibre intake, and, sometimes, the use of

drugs with anti-cholinergic properties [31]. In this population, laxatives are used by more than

5352

Tab

le 2

: Num

ber o

f res

iden

ts fo

r the

top-

10 A

TC th

erap

eutic

mai

n gr

oups

in th

e st

udy

popu

latio

n (n

=2,

355)

ATC

code

Num

ber o

fM

ale/

Fem

ale/

Ty

pe o

f car

eCh

roni

city

‡Av

erag

eM

edia

nre

side

nts

1000

*10

00*

PDD

(SD)

PDD

Drug

gro

upn

(%)

PG† /

1000

som

atic

/100

0

Psyc

hole

ptic

s N

0517

33(7

4)73

973

579

170

9#0.

830.

71 (0

.54)

0.51

Psyc

hotr

opic

sN

05A

813

(35)

418

315#

589

231#

0.68

0.38

(0.5

1)0.

20An

xiol

ytic

sN

05B

665

(28)

296

277

363

243#

0.58

0.55

(0.5

9)0.

40H

ypno

tics

and

seda

tives

N05

C12

62

(54)

511

546

497

551

0.77

0.74

(0.2

8)0.

59

Laxa

tives

A06

1308

(56)

543

561

543

558

0.80

1.48

(0.9

5)1.0

9

Anal

gesi

csN

0212

41(5

3)49

254

152

452

50.

520.

51 (0

.27)

0.50

Opi

oids

N02

A33

9(1

4)15

813

813

315

20.

210.

33 (0

.21)

0.30

Oth

er a

nalg

esic

s an

d an

tipyr

etic

sN

02B

1112

(47)

415

496#

466

471

0.54

0.39

(0.2

9)0.

33

Antit

hrom

botic

age

nts

B01

1230

(52)

514

526

340

605#

0.85

0.96

(0.3

9)1.0

0

Antib

acte

rial

s fo

r sys

tem

ic u

seJ0

111

83(5

0)51

749

653

148

40.

121.1

4 (0

.42)

1.00

Diur

etic

sC0

397

3(4

1)38

342

637

443

1#0.

811.3

8 (1

.24)

1.00

Thia

zide

sC0

3A74

(3)

2235

2037

#0.

740.

96 (0

.31)

1.00

Loop

diu

retic

sC0

3C66

8(2

8)28

425

623

330

8#0.

761.4

8 (1

.42)

1.00

Oth

er37

2(16

)12

917

0#15

915

70.

720.

61 (0

.37)

0.50

Antii

nfla

mm

ator

y an

d M

0188

3(3

7)32

139

7#38

136

90.

491.0

3 (0

.49)

1.00

antir

heum

atic

pro

duct

s

Anta

cids

, dru

gs fo

r tre

atm

ent o

f pe

ptic

ulc

er a

nd fl

atul

ence

A02

563

(24)

263

229

173

265#

0.72

0.93

(0.4

8)1.0

0An

taci

dsA0

2A19

6(8

)87

8259

93#

0.54

0.51

(0.2

5)0.

50D

rugs

for t

reat

men

t of

pep

tic u

lcer

A0

2B43

7(19

)20

017

913

320

7#0.

751.0

2 (0

.43)

1.00

H2-

rece

ptor

ant

agon

ists

A02B

A34

1(1

4)15

714

011

315

6#0.

690.

87 (0

.25)

1.00

Prot

on p

ump

inhi

bito

rsA0

2BC

126

(5)

5852

2665

#0.

691.4

6 (0

.52)

1.16

Antia

nem

ic p

repa

ratio

nsB0

352

1(2

2)20

622

722

920

90.

640.

65 (0

.40)

0.67

Psyc

hoan

alep

tics

N06

406

(17)

144

184#

156

178

0.61

0.55

(0.3

8)0.

43Tr

icyc

lic a

ntid

epre

ssan

tsN

06AA

283

(12)

110

124

104

127

0.58

0.42

(0.2

5)0.

33SS

RIs||

N06

AB91

(4)

2843

3342

0.56

0.89

(0.2

6)1.0

0ot

her

68(3

)20

3244

22#

0.52

0.48

(0.2

8)0.

48

Lege

nd to

Tab

le 2

:*

mal

e/10

00: f

or e

ach

drug

cat

egor

y, th

e nu

mbe

r of m

ale

user

s of

this

cat

egor

y pe

r 100

0 m

ale

resi

dent

s; fe

mal

e/10

00: f

or e

ach

drug

cat

egor

y, th

e nu

mbe

r of f

emal

e us

ers

of th

is c

ateg

ory

per 1

000

fem

ale

resi

dent

s†

PG: p

sych

oger

iatr

ic c

are

‡ch

roni

city

was

def

ined

as

the

num

ber o

f tre

atm

ent d

ays

divi

ded

by th

e nu

mbe

r of p

atie

nt d

ays

in th

e nu

rsin

g ho

me

||SS

RIs:

sel

ectiv

e se

roto

nin

reup

take

inhi

bito

rs#

Stat

istic

ally

sig

nific

ant;

Chi-s

quar

e te

st, p

<0,0

5

tics in dosages that extend well beyond the low dose range. Ulcer-healing drugs form an inte-

resting group in this population: the average daily dosage increases with increasing effective-

ness of the prescribed drugs. Antacids are used by 8% of the population with an average PDD

of 0.51. Histamine H2-antagonists are used by 14% of the population with an average PDD of

0.87. Proton-pump inhibitors are used by 5% of the population with an average PDD of 1.46.

The reason why proton-pump inhibitors are prescribed in higher dosages than recommended

is unknown. Again, more in-depth study of individual patient records is needed: do prescribers

use antacids, H2-antagonists and H. pylori eradication optimally before the step towards pro-

ton-pump inhibitors is made? According to Dutch guidelines [30], the latter should be saved as

a final alternative; the prescribing pattern in this population suggests that this alternative may

be used to solve reflux and gastric acid complaints. Although they are used quite frequently,

prescribing of antianaemic drugs does not seem exceptional from the data available.

Antidepressant use, finally, stands out because many residents use tricyclic antidepressants

(TCAs). In the treatment of depression, the use of TCAs is questionable in the elderly in view of

unwanted adverse effects [30,43]. In some countries, the selective serotonin re-uptake inhibi-

tors (SSRIs) and monoamine oxidase inhibitors have superseded the TCAs for use in the elder-

ly [43], although data on their efficacy or adverse effects in frail, institutionalised elderly pers-

ons are inadequate [12]. Furthermore, TCAs are prescribed in relatively low dosages, which may

indicate that they are prescribed mainly for neuropathic pain disorders. All antidepressants are

prescribed for more than 50% of the time, which is considered rational in view of the time nee-

ded to establish an antidepressive effect [30].

L imitat ions

Biases, to which this study may be subject, all boil down to the completeness of the data.

All residents for whom data were suspected to be incomplete because they could not be linked

to the SIVIS database or because they were in the database but it was unclear when they ente-

red or left the nursing home, were excluded from the study. Since these residents did not differ

from the other residents with respect to demographic characteristics, no bias is expected from

exclusion of this subgroup. For the calculation of average PDD values, all medications which

had been prescribed ‘as needed’, were not included in the calculations. This concerned 5% of

the prescriptions; they were mainly hypnotics, analgesics and laxative drugs. It will lead to a

slight underestimation of the average PDD calculated for the drug groups concerned.

Calculations of the number of users and chronicity of drug use will not be affected. Other than

that, the data are complete with respect to prescription drug utilisation: the database origina-

tes from a drug-dispensing database and, therefore, it is continuously updated and kept ade-

55

half of all residents for 80% of the nursing home stay. Like other studies, our data suggest that

laxatives are overused in institutions [43]. Although non-pharmacological approaches such as

adequate fluid- and fibre intake and physical exercise are considered beneficial [44], these

approaches are often difficult to perform in the nursing home population. To decrease laxative

use and improve bowel function in the institutionalised elderly, an individual approach that

considers both pharmacological and non-pharmacological interventions is recommended.

Analges ics

The analgesic prescribed most frequently is paracetamol. Again, low dosages are given. As

expected, opiates are given less frequently and less chronically than non-opioid analgesics,

which seems rational [30]. Opioids may be helpful for relieving moderate to severe pain, espe-

cially nociceptive pain [45]. NSAIDs are used by 37% of the residents, which is 10% less than

use of analgesics such as paracetamol. This seems appropriate [46], especially when the

NSAIDs are used as alternatives when paracetamol is not effective enough (e.g. for rheumatoid

arthritis). The average PDD value is 1.03, which, in view of the risk of hypertension and renal

failure, is quite high for the elderly [47,48]. If prescribers are unaware of this, a reminder could

be given that paracetamol may be given in higher dosages (4 g/day maximum).

Other drugs

As expected, anticoagulant drugs are given chronically. Fifty percent of the nursing home

residents use an antibiotic drug at any time during their stay; they are not used chronically, and

they are given in normal daily dosages (PDD=1) which usually also is desirable in this popula-

tion. Use of diuretic drugs is more surprising: together with beta-blockers, thiazides are the

drugs of first choice for the treatment of secondary hypertension [30]. They are used by 3% of

nursing home residents; 16% use ‘other diuretics’ that include combination preparations of

thiazides and potassium-sparing agents such as triamterene. Many more residents use loop-

diuretics, the use of which in the elderly has been discouraged because of too severe blood

pressure lowering resulting in needless postural hypotension, and, in some cases, stroke [49].

The average PDD value is 1.48, which indicates that the drugs are prescribed in relatively high

dosages. This calls for more in-depth research as to why prescribers choose loop diuretics so

often, why they are prescribed in such high dosages and whether prescribers are aware of the

advice to be cautious with loop diuretics prescribing in the elderly. Furthermore, the average

dose of thiazides is surprising. Considerable evidence supports the efficacy of low doses of

thiazide diuretics in the treatment of hypertension in elderly people [30,50]. The average PDD,

found in the study presented here (0.96), suggests that nursing home residents receive diure-

54

References

1 Tamblyn RM, McLeod PJ, Abrahamowicz M, et al. Questionable prescribing for elderly patients in Quebec.

CMAJ 1994; 150: 1801-9.

2 Stewart RB. Poly-pharmacy in the aged: practical solutions. Drugs Aging 1994; 4: 449-61.

3 Gill K, Hermans J, Vermeij P. Multiple drug use by the elderly in need of care and nursing and possibilities of

reducing this practice (in Dutch). Ned Tijdschr Geneeskd 1996; 140: 1076-80.

4 Council on Scientific Affairs, American Medical Association. Report of the Council of Scietific Affairs, American

Medical Association. White paper on elderly health. Arch Intern Med 1990; 150: 2459-72.

5 Chrischilles EA, Foley DJ, Wallace RB, et al. Use of medications by persons 65 and over: data from the

established populations for epidemiologic studies of the elderly. J Gerontol 1992; 47: M137-44.

6 Wayne SJ, Rhyne RL, Stratton M. Longitudinal prescribing patterns in a nursing home population.

J Am Geriatr Soc 1992; 40: 53-6.

7 Van Zuylen C, Oostendorp FMGM, van Beusekom BR, Cools HJM, Bolk JH, Lighart GJ. Increasing use of

medication in nursing homes (in Dutch). Ned Tijdschr Geneeskd 1988; 132: 1692-5.

8 Rochon PA, Gurwitz JH. Optimising drug treatment for elderly people: the prescribing cascade.

BMJ 1997; 315: 1096-9.

9 Ackermann RJ, Meyer von Bremen GB. Reducing polypharmacy in the nursing home: an activist approach.

J Am Board Fam Pract 1995; 8: 195-205.

10 Colt HG, Shapiro AP. Drug-induced illness as a cause for admission to a community hospital. J Am Geriatr Soc

1989; 37: 323-6.

11 Monette J, Gurwitz JH, Avorn J. Epidemiology of adverse drug events in the nursing home setting. Drugs Aging

1995; 7: 203-11.


13 Hunter KA, Florio ER, Langberg RG. Pharmaceutical care for home-dwelling elderly persons: a determination of

need and program description. Gerontologist 1996; 36: 543-8.

14 Schmidt I, Claesson CB, Westerholm B, Nilsson LG, Svarstad BL. The impact of regular multidisciplinary team

interventions on psychotropic prescribing in Swedish nursing homes. J Am Geriatr Soc 1998; 46: 77-82.

15 Beers MH, Fingold SF, Ouslander JG. A computerized system for identifying and informing physicians about

problematic drug use in nursing homes. J Med Syst 1992; 16: 237-45.

16 Beers MH, Ouslander JG, Rollingher I, Reuben DB, Brooks J, Beck JC. Explicit criteria for determining

inappropriate medication use in nursing homes. Arch Int Med 1991; 115: 1825-32.

17 Hanlon JT, Schmader KE, Samsa GP, Weinberger M, Uttech KM, Lewis IK, et al. A method for assessing drug

therapy appropriateness. J Clin Epidemiol 1992; 45: 1045-51.

18 Hanlon JT, Weinberger M, Samsa GP, Schmader KE, Uttech KM, Lewis KL, et al. A randomized, controlled trial of

a clinical pharmacist intervention to improve inappropriate prescribing in elderly outpatients with poly-

pharmacy. Am J Med 1996; 100: 428-37.

19 Oborne CA, Batty GM, Maskrey V, Swift CG, Jackson SHD. Development of prescribing indicators for elderly

medical inpatients. Br J Clin Pharmacol 1997; 43: 91-7.

20 Schmader KE, Hanlon JT, Landsman PB, Samsa GP, Lewis IK, Weinberger M. Inappropriate prescribing and

health outcomes in elderly veteran outpatients. Ann Pharmacother 1997; 31: 529-33.

21 Shelton PS, Hanlon JT, Landsman PB, Scott MA, Lewis IK, Schmader KE, et al. Reliability of drug utilization eval-

uation as an assessment of medication appropriateness. Ann Pharmacother 1997; 31: 533-42.

22 Aparasu RR, Fliginger SE. Inappropriate medication prescribing for the elderly by office-based physicians. Ann

Pharmacother 1997; 31: 823-9.

23 Beers MH. Explicit criteria for determining potentially inappropriate medication use by the elderly. An update.

57

quate as a result of its main purpose. Next, nursing staff makes sure that each patient takes his

medication. Although use of over-the-counter (OTC) medication has not been quantified, we

expect this to be relatively low due to practical reasons such as immobility of the residents and

continuous medical attention by both nursing and medical staff. Further, although fluid- and

food supplies are comparable between the nursing homes, no information is available on fluid-

and food intake on an individual patient level. With the exception of laxative drug utilisation

review, this is not considered highly important for the drug utilisation study described. In the

case of laxative drugs, it would be useful to study fluid- and fibre intake on the individual resi-

dents’ level to determine whether dietary changes could lead to diminished laxative use.

In conclusion, this drug utilisation study shows that drug use in the nursing home is high,

many drugs are used chronically and prescribed dosages are relatively low. Priorities for impro-

vement should be given to the prescribing of benzodiazepines, laxatives, loop-diuretics and

ulcer-healing drugs. In case of benzodiazepine prescibing, the high number of users (54%) and

the long-term use deserves attention. Possibilities for tapering drug use should be investiga-

ted. Attention should also be given to laxative use in nursing homes: besides the high percen-

tage of users (56%), we found that relatively high dosages were used. The number of residents

who received loop diuretics was relatively high (28%) and, again, relatively high dosages were

used. Feedback to prescribers is necessary to evaluate the necessity of this practice. In view of

possible adverse effects, the possibility of parallel prescribing and drug-drug interactions, the

use of these drug groups should be re-evaluated carefully.

Acknowledgements

We express our gratitude to D.A. Bloemhof, RPh, for supplying pharmacy data; A.M.

Dijkema, RPh, for her valuable work interviewing the nursing homes’ staff ; SIG Informatics on

Health and Welfare, Utrecht, for supplying data on patient characteristics; and nursing, medi-

cal and pharmacy staff from all participating nursing homes for their co-operation.

56

nonsteroidal anti-inflammatory drug therapy. JAMA 1994; 781-6.

49 Kumar PJ, Clark ML. Clinical Medicine 2nd Ed. Ballière Tindall, London, 1990.

50 Freis ED, Papademetriou V. Current drug treatment and treatment patterns with antihypertensive drugs.

Drugs 1996; 52: 1-16.

59

Arch Intern Med 1997; 157: 1531-6.

24 Koopmans RTCM, de Haan HH, van den Hoogen HJM, Gribnau FW, Hekster YA, van Weel C. Changes in drug use

during stay in a psychogeriatric nursing home (in Dutch). Ned Tijdschr Geneeskd 1994; 138: 1122-6.

25 Nobili A, Tettamanti M, Frattura L, Spagnoli A, Farraro L, Marrazzo E, et al. Drug use by the elderly in Italy.

Ann Pharmacother 1997; 31: 416-22.


27 Koopmans RTCM, van Rossum JM, van den Hoogen HJM, Hekster YA, Willekens-Bogaers MAJH, van Weel C.



28 Vander Stichele RH, Mestdagh J, Van Haecht CH, De Potter B, Bogaert MG. Medication utilization and patient

information in homes for the aged. Eur J Clin Pharmacol 1992; 43: 319-21.

29 Roberts MS, King M, Stokes JA, Lynne TA, Bonner CJ, McCarthy S, Wilson A, Glasziou P, Pugh John W. Medication

prescribing and administration in nursing homes. Age Ageing 1998; 27: 385-92.

30 Thomas S, Geijer RMM, van der Laan JR, Wiersma T. Guidelines for the physician II (in Dutch). Bunge, Utrecht,

1996.

31 Van Dijk KN, de Vries CS, van den Berg PB, Dijkema AM, Brouwers JRBJ, De Jong-van den Berg LTW.

Constipation as an adverse effect of drug use in nursing home patients: an overestimated risk.


32 Johnson RE, Vollmer WM. Comparing sources of drug data about the elderly. J Am Geriatr Soc 1991; 39: 1079-84.

33 Lau HS, de Boer A, Beuning KS, Porsius A. Validation of pharmacy records in drug exposure assessment. J Clin

Epidemiol 1997; 50: 619-25.

34 SIG Informatics in Health and Welfare, Utrecht, the Netherlands

35 World Health Organisation Collaborating Centre for Drug Statistics Methodology, Oslo. Anatomical

Therapeutical Chemical (ATC) classification index including defined daily doses (DDDs) for plain substances.

Oslo: World Health Organisation Collaborating Centre for Drug Statistics Methodology, 1997.

36 Wertheimer AI. The defined daily dose system (DDD) for drug utilisation review. Hosp Pharm 1986: 21: 233-58.

37 Merlo J, Wessling A, Melander A. Comparison of dose standard units for drug utilisation studies.

Eur J Clin Pharmacol 1996; 50: 27-30.

38 Bogle SM, Harris CM. Measuring prescribing: the shortcomings of the item. BMJ 1994; 308: 637-40.

39 Knuistingh Neven A, Graaff de WJ, Lucassen PLBJ et al. Dutch College of General Practitioners-guideline

‘Insomnia and hypnotics’ (in Dutch) Huisarts Wet 1992: 35: 212-9.

40 Ryynanen OP, Kivela SL, Honhanen R, Laippala P, Saano V. Medications and chronic disease as risk factors for

falling injuries in the elderly. Scand J Soc Med 1988; 21: 264-71.

41 Herings RMC, Stricker BHC, Boer de A, Bakker A, Sturmans F. Benzodiazepines and the risk of falling leading to

femur fractures. Arch Intern Med 1995; 155: 1801-7.

42 Busto U, Sellers EM, Naranjo CA, Capell H, Sanchez-Craig M, Sykora K. Withdrawel reaction after long-term use

of benzodiazepines. N Eng J Med 1986; 315: 854-9.

43 Seppälä, Sourander L. A practical guide to prescribing in nursing homes. Avoiding the pitfalls. Drugs Aging

1995; 6: 426-35.

44 Towers AL, Burgio KL, Locher JL, Merkel IS, Safaeian M, Wald A. Constipation in the elderly: influence of

dietary, psychological, and physiological factors. J Am Geriatr Soc 1994; 42: 701-6.

45 Anonymous. New guidelines on managing chronic pain in older persons. JAMA 1998: 280: 311.

46 Ferrell BA. Pain evaluation and management in the nursing home. Ann Intern Med 1995; 123: 681-7.

47 Solomon DH, Gurwitz JH. Toxicity of nonsteroidal anti-inflammatory drugs in the elderly: is advanced age a risk

factor? Am J Med 1997; 102: 208-15.

48 Gurwitz JH, Avorn J, Bohn RL, Glynn RJ, Monane M, Mogun H. Initiation of antihypertensive treatment during

58

Abstract

Objective: It has been suggested that elderly people are at an increased risk of drug-related

problems such as drug-induced adverse effects, drug-drug interactions and drug-disease inter-

actions. This is particularly the case for nursing home residents because of the often complica-

ted and multiple co-morbidity that occurs in these people. The aim of this study was to deve-

lop prescribing indicators to assess systematically the occurrence and nature of potential drug-

drug interactions (DDIs) in a cohort of Dutch nursing home residents.

Methods: The study was conducted in residents aged 65 years and over in six nursing homes

(n=2,355, two-year study period). Computerised medication data for the residents were evalu-

ated with respect to co-prescribing of potentially interacting drugs. All DDIs that were classi-

fied as clinically relevant according to the Dutch National Drug Interaction Database were stu-

died. DDIs were classified into three categories according to their pharmacological mechanism:

1- pharmacokinetic interactions at the level of gastrointestinal (GI) absorption, 2- pharmacoki-

netic interactions at the level of metabolism and excretion and 3- pharmacodynamic interac-

tions.

Results: Thirty-two percent (n=748) of all residents were exposed to one or more combinations

of drugs that could lead to clinically adverse outcomes. The numbers of residents who received

drug combinations with a mechanism of interaction from category 1, 2 or 3 were 73 (3%), 164

(7%) and 612 (26%) respectively. The number of medications prescribed was significantly

associated with the occurrence of a potential DDI (p<0.05). Drug groups most frequently invol-

ved were oral anticoagulants, antibiotics and theophylline.

Conclusion: During the two-year study period, about one-third of the residents were exposed

to at least one drug interaction considered clinically relevant. Adequate surveillance systems

are needed to enable better identifications of these interactions with a view to preventing

potential problems. Using the prescribing indicators developed in this study, such surveillance

could focus on detection and clinical aspects of potential DDIs and possible alternative treat-

ments.

61

2.4 Occurrence of potent ia l drug-druginteract ions in nurs ing home res idents

K.N. van Dijk 1,2, C.S. de Vries 1,3, P.B. van den Berg 1, J.R.B.J. Brouwers 1,2,L.T.W. de Jong-van den Berg 1





International Journal of Pharmacy Practice 2001; 9: 45-52

60

as the number of patients with DDIs per total number of patients [29]. Consequently, the use of

a more uniform way of expressing the incidence or prevalence of DDIs has been recommended

[29]. The duration of concomitant drug use also needs to be taken into account: long-term con-

comitant use may indicate that in daily clinical practice no problems have occurred.

This study aimed to use computerised pharmacy prescription data to develop prescribing

indicators for DDIs and measure their occurrence, nature and duration in a sample of Dutch

nursing home residents. Such prescribing indicators could subsequently be used to audit pre-

scribing practices in the elderly and to monitor potential adverse drug reactions arising from

DDIs.

Methods

Potent ia l drug-drug interact ions

Information on the relevance of potential DDIs was obtained from the Dutch National Drug

Interaction Database [35]. Where available, the DDI information in this database is derived

from international reference books, such as Hansten and Stockley [36,37]. Based on these

books, in the database DDIs are divided into ‘clinically relevant’ DDIs (n=175) and ‘clinically less

relevant’ DDIs (n=74). A ‘clinically relevant’ DDI indicates that immediate action should be

taken, such as adjustment of dosage regimes, or suggesting an alternative drug to the prescri-

ber. For a DDI that is ‘clinically less relevant’ no immediate action is necessary. Only the 175 cli-

nically relevant DDIs were studied. DDIs were classified into three categories according to

pharmacological mechanism: (1) pharmacokinetic DDIs at the level of gastrointestinal absorp-

tion, (2) pharmacokinetic DDIs at the level of metabolism and excretion, and (3) pharmacody-

namic DDIs. The classification is shown in table 2.

Data col lect ion

The study was conducted among residents aged 65 and over in six nursing homes, each

with a 90 to 225-bed capacity. The study population consisted of 2,355 residents present at any

time during the two-year study period (1 October 1993 to 1 October 1995). In this study, phar-

macy records of the nursing home residents were linked with a national information system on

nursing homes (SIVIS), which contains patient-specific information on morbidity and type of

nursing (psychogeriatric or somatic) of virtually every nursing home resident in the

Netherlands. The databases and drug use in this population have been described in detail

[38,39]. The drug dispensing system in the nursing homes results in medication prescribing

being recorded in the pharmacy computer system and updated with medication changes on a

63

Introduct ion

Elderly people are at an increased risk of drug-related problems such as drug-induced

adverse effects, drug-drug interactions and drug-disease interactions [1-5]. This is particularly

the case for nursing home residents because of the often complicated and multiple co-morbi-

dity that occurs in these people [5-9]. To address this problem, several approaches have been

made towards rational and appropriate prescribing of drugs in the institutionalised elderly.

Prescribing indicators have been developed or applied with the aim of systematically assessing

medication appropriateness and thus providing a starting point for improvement of prescribing

[10-23]. One of the prescribing indicators used is the occurrence of potentially inappropriate

drug combinations [15,20-23]. A drug-drug interaction (DDI) is defined as ‘a pharmacological or

clinical response to the administration of a drug combination different from that anticipated

from the known effects of the two agents when given alone’ [24]. DDIs can lead to unintended

responses such as enhanced or reduced drug effects and these effects vary between individu-

als. Drug-drug interactions are commonly divided into pharmacokinetic and pharmacodynamic

DDIs [25,26]. Elderly people may be at a higher risk of the adverse effects of a DDI due to phar-

macokinetic, pharmacodynamic, and disease-related changes that occur with advanced age

[2,3,5]. Risk factors associated with the occurrence of potential DDIs are age, number of medi-

cations prescribed, the number of physicians involved, and the presence of increased frailty [2].

The impact of DDIs in nursing homes is reported to be high: DDIs may account for 22% of all

adverse drug reactions reported, or adversely affect 23% to 53% of all residents [5]. The inci-

dence of hospital admissions caused by DDIs has been reported to range from 0% to 11.5%

[27,28]. These percentages vary due to differences between study populations, use of different

definitions of a DDI, different ways of establishing cause-and-effect relationships, and diffe-

rences in study design.

Several studies have been published on frequencies of potential DDIs in different popula-

tions [28-34]. In a review of 7 studies, the frequency of potential DDIs was found to range from

9.2% to 70% in ambulatory patients [29]. In hospitalised patients, reports of frequencies of

potential DDIs ranged from 2.2% to 60% [28,29]. In nursing home residents, the proportion of

residents exposed to a potential DDIs has been reported to range from 23.4% to 49% [29].

These large variations in reported frequencies are mainly due to methodological considera-

tions, such as the use of different definitions of a clinically significant DDI and use of different

reference lists for DDIs (sometimes no references are given), differences in study design, and

differences in the way of measuring the frequency or incidence of DDIs. For example, the fre-

quency of DDIs can be expressed as the total number of DDIs per total number of patients or

62

ses it was found that the following variables were associated with the occurrence of a DDI:

number of medications prescribed, the type of nursing psychogeriatric or somatic) and

Parkinson’s disease.

In Table 2 detailed information is presented on the potential DDIs. The DDIs to which most

residents were exposed comprised the interaction between loop diuretics and non-steroidal

anti-inflammatory drugs (NSAIDs), and the interaction between oral anticoagulants and

NSAIDs (9.7% and 9.6% respectively). All other DDIs found in this study occurred in less than

5% of the study population. The percentage of index drug users most frequently exposed to a

potential DDI were users of ACE-inhibitors, fluoride, oral anticoagulants, acetazolamide, phe-

nytoin, and loop diuretics.

Table 3 provides an overview of the drugs most commonly involved in prescribing of inter-

acting drugs. The index drugs used most often were oral anticoagulant drugs (involved in ten

DDIs), antibiotics (involved in four DDIs), and theophylline (involved in three DDIs). Interacting

drugs most frequently involved were those with metallic ions (iron salts, antacids; involved in

five DDIs), NSAIDs (involved in three DDIs), diuretics, enzyme inducers (such as carbamazepi-

ne and phenytion), and enzyme inhibitors (such as verapamil and fluoxetine) (all involved in

two DDIs). The number of days that drug combinations were prescribed concomitantly is rela-

tively high. Nineteen out of 32 DDIs were prescribed for an average of 50 days or more per 100

days of index drug use.

daily basis. Nurses dispense medication to individual nursing home residents on the basis of

the information (drug, dosage, route and time of administration) in this computer system. As a

consequence, the recording of actual drug use can be considered very accurate. At the time of

the study there was no automatic signal generation by the system when potential DDIs occur-

red, but nursing home physicians and hospital pharmacists checked medication files on a daily

basis for medication changes.

Data analys is

A retrospective cohort study was performed to estimate the prevalence of potential DDIs

and the possible risk factors associated with the occurrence of potential DDIs. We developed

prescribing indicators based on the frequency, nature and duration of DDIs. For each DDI, an

index drug and an interacting drug were defined. The index drug was defined as a drug of which

the pharmacological or clinical response is altered by the interference of a second drug (the

interacting drug). The outcome of interest was the occurrence of a potential DDI, defined as the

concomitant use of both index drug and interacting drug for at least one day. The number of

residents exposed to each individual DDI was calculated. The prevalence of the DDIs was

expressed as the number of residents exposed to a DDI divided by the number of index drug

users. We also calculated the percentage of all residents (n=2,355) affected. For all patients on

a specific DDI, the number of days that the interacting drugs were prescribed concomitantly per

100 days of index drug use was calculated. A stepwise logistic regression was performed to

determine predictive variables of the prescribing of interacting drugs likely to cause adverse

clinical effects. The statistical software program SPSS 9.0 for Windows (SPSS Inc., Chicago, IL)

was used.

Results

Thirty-two per cent (n=748) of all residents were exposed to one or more combinations of

drugs that could potentially lead to adverse clinical outcomes. Out of 175 clinically relevant

DDIs from the interaction database, 32 drug combinations (18%) were prescribed; the other 143

did not occur in this population. Most DDIs found were based on pharmacodynamic mecha-

nisms. The number and percentage of all residents that received a drug combination from cate-

gory 1 (level of GI-absorption), 2 (level of metabolism and excretion) or 3 (pharmacodynamic

level) was 73 (3%), 164 (7%) and 612 (26%), respectively. Table 1 presents differences between

residents with and without DDIs with regard to age, gender, type of nursing, morbidity, and

number of different medications prescribed. From the multivariable logistic regression analy-

64 65

6766


Variable Number of residents Number of residents ORcrude for DDI ORadjusted¶

with DDI (%) without DDI (%) (95% CI) (95% CI)(N=748) (N=1607)

Age (years) 82 (sd 7) 82 (sd 7) 0.99 (0.98-1.00) 0.99 (0.98-1.00)Gender

Female 549 (73.4) 1117 (69.5) 1 (reference) 1 (reference)Male 199 (28.9) 490 (30.5) 0.82 (0.68-1.00) 0.81 (0.65-1.00)

Type of nursingPsychogeriatric 142 (19.0) 558 (34.7) 1 (reference) 1 (reference)Somatic 592 (79.1) 1017 (63.2) 2.29 (1.85-2.82) 2.34 (1.87-2.94)Not known 14 (1.9) 32 (2.0)

MorbidityParkinson’s disease 29 (3.9) 122 (7.6) 0.49 (0.32-0.74) 0.29 (0.18-0.46)Diabetes mellitus 74 (9.9) 102 (6.3) 1.62 (1.19-2.22) 1.19 (0.84-1.68)Dementia 144 (19.3) 545 (33.9) 0.47 (0.38-0.57) 0.72 (0.52-1.00)Depression 10 (1.3) 30 (1.9) 0.71 (0.35-1.47) 0.53 (0.24-1.17)

Number of different medications prescribed during study period0-4 33 (4.4) 388 (24.1) 1 (reference) 1 (reference)5-9 264 (35.3) 744 (46.3) 4.17 (2.85-6.11) 4.29 (2.92-6.30)10-14 251 (33.6) 376 (23.4) 7.85 (5.32-11.59) 8.58 (5.78-12.73)15 or more 200 (26.7) 99 (6.2) 23.75 (15.46-36.49) 27.21 (17.48-42.35)

¶ adjusted for type of nursing, Parkinson’s disease, and number of different medications prescribed

Table 2: Characteristics of potential DDIs, listed per category of interaction

Index drug‡ (number of users) Interacting drug† (group) Number of % of study % of days Clinical effect of DDIindex drug population of conco-users with (n=2355) mitant drug DDI (%) use§

Category 1 (reduced GI-absorption)

Doxycycline (349) Metallic ions (iron salts) 27 (8) 1.1 88 Risk of subtherapeutic doxycycline serum concentration.

Fluoroquinolones (121) Metallic ions (antacids and iron salts) 18 (15) 0.8 95 Risk of subtherapeutic fluoroquinolone serum concentration.

Thyroid hormones (81) Metallic ions (iron salts) 18 (22) 0.8 42 Risk of inadequate control of hypothyroidism.

Bisphosphonates (48) Metallic ions (antacids, calcium salts and iron) 7 (15) 0.3 34 Risk of inadequate prevention of osteoporosis.

Fluoride (18) Metallic ions (antacids and calcium salts) 7 (39) 0.3 95 Risk of inadequate prevention of osteoporosis.

Oral anticoagulants (646) Bile acids sequestrants 1 (<1) <0.1 100 Reduced effect of oral anticoagulant.

Cefuroxime (8) Drugs for the treatment of peptic ulcer 1 (13) <0.1 100 Risk of subtherapeutic cefuroxime serum concentration.

Total number of residents with category 1 DDI: 73 (3% of study population)

Category 2 (metabolism and excretion)

Oral anticoagulants (646) Co-trimoxazole 42 (6.5) 1.8 4.4 Increased effect of oral anticoagulant due to enzymatic inhibition and protein displacement. Risk of prolonged bleeding time.

Oral anticoagulants (646) Enzyme inducers£ 32 (5) 1.4 60 Decreased effect of oral anticoagulant due to induction of hepatic metabolism.

Digoxin (306) Verapamil 25 (8) 1.1 62 Verapamil inhibits non-renal and renal excretion of digoxin, leading to digoxin intoxication.

Doxycycline (349) Enzyme inducers£ 24 (7) 1.0 95 Decreased effect of doxycycline due to enzyme induction. Risk of subtherapeutic antimicrobial serum concentration.

Phenytoin (74) Co-trimoxazole 21 (28) 0.9 2 Increased effect of phenytoin. Risk of toxic effects such as nys-tagmus, diplopia and dizziness.

Oral anticoagulants (646) Tamoxifen 8 (1.2) 0.3 100 Increased effect of oral anticoagulant due to inhibition of hepatic metabolism. Risk of prolonged bleeding time.

Carbamazepine (87) Enzyme inhibitors¶ 6 (7) 0.3 51 Increased effect of carbamazepine due to reduced clearance. Risk of toxic effects such as drowsiness, nausea.

Theophylline (70) Fluoroquinolones 7 (10) 0.3 8 Increased risk of theophylline toxicity due to inhibition of hepatic (cytochrome P450 1A2) metabolism.

6968

Tricyclic antidepressants (283) SSRIs# 4 (1) 0.2 73 Increased risk of TCA toxicity due to inhibition of hepatic (cytochrome P450 2D6) metabolism

Lithium (14) NSAIDs¥ 3 (21) 0.1 48 Increased risk of lithium toxicity due to decreased lithium excretion.

Theophylline (70) Erythromycin 2 (3) <0.1 4 Increased risk of theophylline toxicity due to inhibition of hepatic metabolism

Oral anticoagulants (646) Amiodarone 2 (<1) <0.1 100 Increased effect of oral anticoagulant due to inhibition of hepatic metabolism. Risk of prolonged bleeding time.

Oral anticoagulants (646) Cimetidine 1 (<1) <0.1 100 Increased effect of oral anticoagulant due to inhibition of hepatic metabolism. Risk of prolonged bleeding time.

Theophylline (70) Enzyme inhibitors¶ 2 (3) <0.1 18 Increased risk of theophylline toxicity due to inhibition of hepatic metabolism.

Oral anticoagulants (646) Metronidazole 1 (<1) <0.1 4 Increased effect of oral anticoagulant due to inhibition of hepatic metabolism. Risk of prolonged bleeding time.

Total number of residents with category 2 DDI: 164 (7% of total study population)

Category 3 (pharmacodynamic interaction)

Loop diuretics (668) NSAIDs¥ 229 (34) 9.7 41 Decreased effect of diuretics due to reduction in renal perfusionand glomerular filtration.

Oral anticoagulants (646) NSAIDs¥ 225 (35) 9.6 41 Inhibition of platelet aggregation leading to increased bleeding risk and risk of peptic ulcer.

ACE-inhibitors (208) Diuretics 197 (95) 8.4 89 Blocking the activated renine-angiotensione-aldosterone-system leads to increased vasodilatation and strong hypotensive effects.

NSAIDs (883) Corticosteroids 78 (9) 3.3 55 Increased risk of peptic ulcer.

Oral anticoagulants (646) Salicylates (low dose) 34 (5.3) 1.4 65 Irreversible inhibition of platelet aggregation leading to increased bleeding risk and risk of peptic ulcer.

Oral anticoagulants (646) Thyreoid hormones 21 (3.3) 0.9 87 Increased thyroid function leading to increased response to oralanticoagulant.

Beta-blocking agents (167) Verapamil/diltiazem 16 (10) 0.7 94 Risk of diminished atrioventricular conductance and reduced contractility of the heart.

Potassium-sparing diuretics Potassium (salts) 14 (4) 0.6 38 Increased risk of hyperkalaemia.(364)

Hypoglycaemic agents (356) Beta-blocking agents 7 (2) 0.3 50 Reduced awareness of hypoglycaemia.

Acetazolamide (8) Diuretics 3 (38) 0.1 4 Increased risk of hypokalaemia.

Total number of residents with category 3 DDI: 612 (26% of total study population)

Table 3: D

rug groups most com

monly involved in prescribing of interacting drugs

Drug (group)N

umber of

Num

ber of patients %

of days of

DDIs implicated

(n=2355)

concomitant drug use

§

Index drug (group) ‡

Oral anticoagulants

10304

61

Antibacterial drugs4

6493

Theophylline3

1110

Interacting drug group†

Metallic ions

571

53

NSAID

s ¥3

397 41

Diuretics

2199

56

Enzyme inducers £

252

79

Enzyme inhibitors ¶

28

44

Legend to Table 3:§

expressed as the number of days of concom

itant drug use per 100 days of index drug use‡

defined as the drug of which the pharm

acological or clinical response is altered by the interference of a second drug, the interacting drug†

defined as the drug that influences the pharmacological or clinical response of the index drug

¥

non-steroidal anti-inflamm

atory drugs£

carbamazepine, phenobarbitone, phenytoin, rifam

picin¶

fluoxetine, fluvoxamine, verapam

il, diltiazem

Legend to Table 2: ‡

defined as the drug of which the pharm

acological or clinical response is altered by the interference of a seconddrug, the interacting drug

†defined as the drug that influences the pharm

acological or clinical response of the index drug§

expressed as the number of days of concom

itant drug use per 100 days of index drug use £

carbamazepine, phenobarbitone, phenytoin, rifam

picin ¶

fluoxetine, fluvoxamine, verapam

il, diltiazem#

selective serotonin reuptake inhibitors¥

non-steroidal anti-inflamm

atory drugs

For individual DDIs the percentage of residents affected was 1% or less. Interactions in this

category are almost always clinically relevant because complex formation leads to subthera-

peutic concentrations of drugs. Sometimes an alternative drug for the interacting drug can be

chosen (for example an H2-antagonist instead of an antacid), and sometimes the interacting

drug can be discontinued temporarily during index drug administration (for example iron salts

during doxycycline therapy).

Practical implications

The DDIs of this category should be avoided by separating the administration times of the

two interacting drugs by a two- to four-hour interval. In practice, this measure is taken fre-

quently, however exact data on the administration times of the drugs in the study population

were not available. A computerised adjustment of dosage schedules could support appropria-

te timing of administration.

Pharmacokinetic DDIs at the level of metabolism and excretion

The DDIs due to metabolism or excretion (category 2) are often considered clinically rele-

vant, in particular those involving inhibition of cytochrome P450 (CYP450) isoenzymes. Overall,

7% of the residents in our study were exposed to one or more of the DDIs from category 2. For

individual DDIs the percentage of residents affected was 3.4% or less. The highest prevalences

were found for the potentially increased anticoagulant effect of acenocoumarol or phenpro-

coumen by the concomitant use of co-trimoxazole and enzyme-inducers, respectively. For these

DDIs, monitoring of the bleeding time (International Normalised Ratio, INR) is warranted, and

this is common practice in Dutch nursing homes. Among the index drug users, the DDIs most

frequently observed were the interaction between phenytoin and co-trimoxazole (28% of phe-

nytoin users) and the interaction between lithium and NSAIDs (21% of lithium users). The lat-

ter may be clinically important, since renal function deteriorates with age and puts elderly at

an increased risk of lithium toxicity. Alternative drugs could be advised, such as paracetamol

instead of NSAIDs and doxycycline instead of co-trimoxazole. Eight out of 15 DDIs from cate-

gory 2 had 50 days or more of concomitant drug use per 100 days of index drug use. This could

mean that, in practice, these DDIs do not lead to clinical problems. Further investigation is nee-

ded to find out if more intense monitoring of drug therapy by means of therapeutic drug moni-

toring is necessary.

Practical implications

The DDIs of this category can be monitored by measuring serum (index) drug levels, for

example digoxine, theophylline, and lithium levels. For the DDIs in which oral anticoagulants

are involved, measuring bleeding time (INR) is warranted.

71

Discuss ion

In this study we developed prescribing indicators, based on the occurrence, frequency and

duration of drug-drug interactions, to describe DDIs in a cohort of nursing home residents.

Although the prescribing indicators were descriptive in nature, they allowed us to identify the

drug groups most frequently involved and residents who were most at risk for being exposed

to a DDI. Thirty-two percent of the nursing home residents were exposed to at least one DDI

classified as potentially clinically relevant. Of all residents, 26% were exposed to a DDI from

category 3 (pharmacodynamic interaction), which could be relevant especially in this popula-

tion because of reduced homeostatic mechanisms. For each individual DDI classified as clini-

cally relevant, no more than 10% of the residents were affected. This means that although one-

third of the population was exposed to a DDI, there was no specific DDI that could be identified

as potentially harmful to many residents. The more drugs residents were prescribed, the hig-

her became their risk of having a DDI. This is in line with other studies [28,30]. Residents who

were nursed in psychogeriatric wards and residents who were diagnosed with Parkinson’s dis-

ease showed a decreased risk for the occurrence of a DDI. Other risk factors were not identified

in this study.

Other studies have reported the frequency and nature of potential DDIs in nursing home

residents [29-34], although the only study published after 1990 was that of Bergendal [30]. In

their study, potential DDIs were investigated in 5125 mainly elderly patients in nursing homes

and homes for the elderly in Sweden. They found that 31% of the patients had at least one DDI

and that these patients were prescribed significantly more drugs than those without DDIs. The

prescribing indicators we present are descriptive in nature and can be used with prescription

databases. Williams and colleagues [20] developed an index of quality prescribing in general

practice by investigating the incidence of potential DDIs. They determined an odds ratio as a

measure of potential DDIs avoided, comparing the use of cimetidine (which interacts with

many other drugs) with that of non-interacting H2-antagonists in users and non-users of inter-

acting drugs (warfarine, theophylline, phenytoin). Although this is an elegant method of audi-

ting H2-antagonists prescribing, it is not by definition applicable to all other drug groups.

Furthermore, drug formularies in nursing homes usually restrict the choice of drugs from a spe-

cific drug class.

Impl icat ions for c l in ica l pract ice

Pharmacokinetic DDIs at the level of gastrointestinal absorption

Overall, 3% of all residents were exposed to DDIs at the level of GI-absorption (category 1).

70

profile, staff pharmacists detected only 20% of the DDIs. The authors recommended compute-

rised drug interaction profiles to be used by pharmacists to ensure recognition of all potential

DDIs. In another study, the value of electronic prescribing for elderly was highlighted [41]. In

particular, on-line detection of DDIs during prescribing and suggestion of non-interacting drugs

could be useful. Recently, national attention has been given to pharmaceutical activities in

Dutch nursing homes. To assess the quality of the medication distribution process and other

pharmaceutical activities, in 1997 the Dutch Health Care Inspectorate carried out a survey

among 33 Dutch nursing homes [42]. A computerised medication surveillance system was ope-

rational in only 9 out of 33 nursing homes. Together with the results of our study this indicates

that computerised detection of DDIs in nursing homes is warranted. Furthermore, more insight

is needed into the clinical relevance of those DDIs classified as ‘clinically relevant’. The fact that

in this study long-term concomitant use of interacting drugs was found, raises the question of

whether clinically relevant side effects are actually seen in the elderly. The prescribing indica-

tors developed in this study provide the tools to audit DDI occurrence in nursing homes syste-

matically.

Acknowledgements

We wish to thank R. Fijn, pharmacist, for his comments on the manuscript. We express our

gratitude to D.A. Bloemhof, hospital pharmacist, for supplying pharmacy data, SIG Informatics

on Health and Welfare, Utrecht, for supplying morbidity data; and nursing, medical and phar-

macy staff from all participating nursing homes for their co-operation. The Wetenschappelijk

Instituut Nederlandse Apothekers (WINAp) financially supported this study.

73

Pharmacodynamic DDIs

Pharmacodynamic interactions are of particular relevance to the elderly [2]. In this study

26% of all residents were exposed to one or more pharmacodynamic DDIs (category 3). The

highest prevalence was found for the interaction between loop diuretics and NSAIDs. This

interaction may be clinically relevant to the elderly, in view of the reduced effectiveness of the

diuretics. Furthermore, the NSAID-induced reduced renal function may influence other drug

therapies. The next interaction seen most often was that between oral anticoagulants and

NSAIDs, which could lead to an increased risk of bleeding and risk of peptic ulcer. Choosing an

alternative analgesic is an option to avoid this interaction. Among the index drug users, the

DDIs most frequently observed were the combination between ACE-inhibitors and diuretics

(95% of ACE-inhibitor users) and again the combination of oral anticoagulants and NSAIDs

(35% of oral anticoagulant drug users). The combination of ACE-inhibitors with diuretics is

only relevant when ACE-inhibitor use is initiated without a temporary discontinuation of the

diuretic (or the use of a low first dosage of ACE-inhibitor). Feedback on this should be given to

prescribers.

Practical implications.

The clinical relevance of DDIs of this category can be assessed by monitoring the clinical

effects of index drugs, e.g. the decreased effect of loop diurectics that (possibly) results in

symptoms of heart failure. Measuring serum potassium levels is warranted in view of the

hyper- or hypokalaemic effects of some DDIs of this category.

Cl in ica l re levance

Tamai and colleagues [31] found that half of the 24 suspected interactions in a group of 138

nursing home residents had potential clinical relevance. However only two residents were

exposed to a substantial degree of risk. Foxall [34] reported 67 potential DDIs in a group of 106

nursing home patients, of which 25 DDIs were potentially life-threatening. Doucet [28] found

that in 14.5% of hospital patients older than 70 years who were exposed to a potential DDI, a

documented side effect was found. Although data on the clinical relevance of DDIs are scarce,

it is estimated that 10% of potential DDIs result in clinically significant events [2]. This would

imply that 3% of our study population, i.e. 71 residents, would be at risk of a clinically signifi-

cant event.

Several methods can be used to alert prescribers and pharmacists to the occurrence of

potential DDIs. Electronic or computer-aided prescribing can be a useful tool in the recognition

and handling of DDIs. Recently, Weideman and colleagues [40] showed that in an eight-drug

72

Drug Saf 1993; 9: 51-9.

28 Doucet J, Chassagne P, Trivalle C, et al. Drug-drug interactions related to hospital admissions in older adults:

a prospective study of 1000 patients. J Am Geriatr Soc 1996; 44: 944-8.

29 Jankel CA, Speedie SM. Detecting drug interactions: a review of the literature. DICP Ann Pharmacother 1990;

24:982-9.

30 Bergendal L, Friberg A, Schaffrath AM. Potential drug-drug interactions in 5,125 mostly elderly outpatients in

Gothenburg, Sweden. Pharm World Sci 1995; 17: 152-7.

31 Tamai IY, Strome LS, Marshall CE, Mooradian AD. Analysis of drug-drug interactions among nursing home

residents. Am J Hosp Pharm 1989; 48: 1567-9.

32 Lang LA, Kabat HF. Drug interactions in nursing home patient prescriptions. J Am Pharm Assoc 1970; 10: 647-7.

33 Cooper JW, Wellins I, Fish KH, Loomis ME. Frequency of potential drug-drug interactions.

J Am Pharm Assoc 1975; 15: 24-7,31.

34 Foxall MJH. Elderly patients at risk of potential drug interactions in long-term care facilities.

West J Nurs Res 1982; 4: 133-51.

35 WINAp Geneesmiddelinformatie, Royal Dutch Society for the Advancement of Pharmacy, The Hague,

The Netherlands, 1998.

36 Hansten PD, Horn JR. Drug interactions and updates. 7th Ed. Philadelphia: Lea & Febiger, 1990.

37 Stockley IH. Drug Interactions. 2nd Ed. Oxford: Blackwell Scientific Publications, 1991.

38 Van Dijk KN, De Vries CS, Van den Berg PB, Dijkema AM, Brouwers JRBJ, De Jong-van den Berg LTW.

Constipation as an adverse effect of drug use in nursing home patients: an overestimated risk. Br J Clin

Pharmacol 1998; 46: 255-61.



40 Weideman RA, Bernstein IH, McKinney PW. Pharmacist recognition of potential drug interactions.

Am J Health-Syst Pharm 1999; 56: 1524-9.

41 Venot A. Electronic prescribing for the elderly. Drugs Aging 1999; 2: 77-80.

42 Health Care Inspectorate. Medication distribution in nursing homes (in Dutch). Ministery of Health,

Welfare and Sports, The Hague, The Netherlands, 1997.

75

References

1 Hughes SG. Prescribing for the elderly: why do we need to exercise caution? Br J Clin Pharmacol 1998; 46:531-3.

2 Seymour RM, Routledge PA. Important drug-drug interactions in the elderly. Drugs Aging 1998; 12: 485-94.

3 Kinirons MT, Crome P. Clinical pharmacokinetic considerations in the elderly. Clin Pharmacokinet 1997; 33: 302-12.

4 Lamy PP. Institutionalisation and drug use in older adults in the US. Drugs Aging 1993; 3: 232-7.

5 Lamy PP. The elderly and drug interactions. J Am Geriatr Soc 1986: 34: 586-92.


7 Monette J, Gurwitz JH, Avorn J. Epidemiology of adverse drug events in the nursing home setting.

Drugs Aging 1995; 7: 203-11.

8 Seppälä M, Sourander L. A practical guide to prescribing in nursing homes. Avoiding the pitfalls.

Drugs Aging 1995; 6: 426-35.

9 Beers MH, Fingold SF, Ouslander JG. A computerized system for identifying and informing physicians about

problematic drug use in nursing homes. J Med Syst 1992; 16: 237-45.

10 Harrison DL, Bootman JL Cox ER. Cost-effectiveness of consultant pharmacists in managing drug-related

morbidity and mortality at nursing facilities. Am J Health Syst Pharm 1998; 55: 1588-94.

11 Williams B, Betley C. Inappropriate use of nonpsychotropic medications in nursing homes.

J Am Geriatr Soc 1995; 43: 513-9.

12 Beers MH, Ouslander JG, Rollingher I, Reuben DB, Brooks J, Beck JC. Explicit criteria for determining

inappropriate medication use in nursing homes. Arch Int Med 1991; 115: 1825-32.

13 Hanlon JT, Schmader KE, Samsa GP, et al. A method for assessing drug therapy appropriateness.


14 Hanlon JT, Weinberger M, Samsa GP. A randomized, controlled trial of a clinical pharmacist intervention to

improve inappropriate prescribing in elderly outpatients with polypharmacy. Am J Med 1996; 100: 428-37.



16 Schmader KE, Hanlon JT, Landsman PB, Samsa GP, Lewis IK, Weinberger M. Inappropriate prescribing and

health outcomes in elderly veteran outpatients. Ann Pharmacother 1997; 31: 529-33.

17 Shelton PS, Hanlon JT, Landsman PB, et al. Reliability of drug utilization evaluation as an assessment of

medication appropriateness. Ann Pharmacother 1997; 31: 533-42.

18 Aparasu RR, Fliginger SE. Inappropriate medication prescribing for the elderly by office-based physicians.

Ann Pharmacother 1997; 31: 823-9.

19 Beers MH. Explicit criteria for determining potentially inappropriate medication use by the elderly. An update.

Arch Intern Med 1997; 157: 1531-6.

20 Williams D, Kelly K, Feely J. Drug interactions avoided - a useful indicator of good prescribing practice.


21 Cantrill JA, Dowell J, Roland M. Qualitative insights into general practitioners’ views of their long-term

prescribing. Int J Pharm Pract 2000; 8: 20-6.

22 Coste J, Séné B, Milstein C, Bouée S, Venot A. Indicators for the automated analysis of drug prescribing quality.

Meth Inform Med 1998; 37: 38-44.

23 Buetow SA, Sibbald B, Cantrill JA, Halliwell S. Prevalence of potentially inapproppriate long term prescribing in

general practice in the United Kingdom, 1980-95: systematic literature review. BMJ 1996; 313: 1371-4

24 Tatro DS (Ed). Drug interaction facts. J.B. Lippincott co., St Louis, 1992.

25 McInnes GT, Brodie MJ. Drug interactions that matter. A critical reappraisal. Drugs 1988; 36: 83-110.

26 Adams P. Drugs interactions that matter. (1) Mechanisms and management. Pharm J 1998; 261: 618-21.

27 Jankel CA, Fitterman LK. Epidemiology of drug-drug interactions as a cause of hospital admissions.

74

Abstract

Objective: We aimed to evaluate drug use in 2 Dutch nursing homes (254 residents) by develo-

ping and evaluating indicators based on pharmacy prescription data.

Methods: We evaluated the prescribing of benzodiazepines, NSAIDs, ulcer-healing drugs, and

diuretics. Prescribing indicators were used to identify prescribing that was potentially not

according to recommendations in national and regional prescribing guidelines. We used both

descriptive indicators, such as the number and percentage of users, and indicators reflecting

potentially suboptimal prescribing, such as use of drugs outside the regional drug formulary,

use of more than one drug from the same drug class and prescription of drug dosages above

recommended values. When potentially suboptimal prescribing was found, we verified the fin-

dings by means of an interview with one of the prescribers.

Results: The prescribing indicators we assessed were generally in agreement with national and

regional guidelines. However, use of benzodiazepines for more than 30 days and prescribing of

NSAIDs without concomitant prescribing of gastroprotective drugs was found in a relatively

high percentage of patients. After prescriber interview and patient chart review it was found

that some prescribing indicators, such as dosages above recommended values, were not always

indicative for suboptimal prescribing.

Conclusion: We found the majority of prescribing to be in line with recommendations upon

which we based our prescribing indicators. The prescribing indicators could be used to evalu-

ate prescribing practices, however appropriateness of prescribing was more difficult to assess.

For this, clinical information from the prescriber was necessary to be able to fully assess pre-

scribing appropriateness.

77

2.5 Prescr ib ing indicators as a tool toevaluate drug use in nurs ing homes: a p i lot s tudy

K.N. van Dijk 1,2, L.G. Pont 3, C.S. de Vries4, M. Franken1, J.R.B.J. Brouwers1,2, L.T.W. de Jong-van den Berg1



Groningen, the Netherlands2 Department of Clinical Pharmacy, Medical Centre Leeuwarden, Leeuwarden, the Netherlands3 Department of Clinical Pharmacology, Groningen University Institute for Drug Exploration

(GUIDE), Groningen, the Netherlands4 Department of Pharmacoepidemiology, Postgraduate Medical School, University of Surrey,


Submitted

76

teria. Although the remaining criteria could be applied using solely pharmacy data, we percei-

ved them as potentially insensitive measurement instruments, which do not always reflect

inappropriate prescribing. For example, the use of multiple antipsychotic drugs that was inap-

propriate according to these criteria, may be clinically beneficial. Also, the fact that a prescri-

bing indication is not documented, another indication of suboptimal prescribing according to

this study, does not necessarily mean that the drug is being prescribed inappropriately. The fact

that in the Swedish studies more than a quarter of the residents received prescriptions that

were classified as potentially inappropriate suggests these criteria may have been too unspe-

cific. The Medication Appropriateness Index (MAI), developed by Hanlon in 1992 [13], was found

to be the closest to a reliable, standardised and valid instrument for assessing medication

appropriateness in elderly outpatients [1]. To our knowledge, the MAI has not been used to

assess medication appropriateness in nursing homes. In view of the differences in drug use and

living circumstances between elderly outpatients and nursing home residents, criteria for

medication appropriateness are not necessarily the same for both populations. The MAI con-

sists of 10 questions assessing the appropriateness of a prescribed medication. For 4 questions,

information on diagnoses is necessary. The other 6 questions might be suitable for use with

pharmacy prescription data only, such as ‘are there clinically significant drug-drug interactions’

and ‘is there unnecessary duplication with other drug(s)’. However, we considered aspects con-

cerning directions of use, such as patient leaflets, not to be as relevant to the appropriateness

of nursing home prescribing, as nurses ensure adequate administration of the drugs. Recently,

Knight and Avorn [5] published a list of 12 quality indicators, based on literature review and

expert panel consideration. For 5 quality indicators, clinical information such as drug indica-

tion, response to therapy or renal function was needed. One indicator concerned a drug not

available on the Dutch market. Another indicator concerned patient education, an item that

could be relevant in view of monitoring side effects by caregivers. The quality indicators that

could be used with pharmacy prescription data only included the availability of a medication

list, periodic drug regimen review, avoidance of drugs with strong anticholinergic properties

and avoidance of barbiturates, although we consider the latter a less clinically relevant problem

in view of the limited use of barbiturates in the Netherlands. To assess medication appropria-

teness in nursing homes, indicators that reflect deviations from national pharmacotherapy gui-

delines and drug formularies should be used [14]. The development of pharmacotherapy gui-

delines specifically for the elderly is limited so far. In the Netherlands, initiatives for Dutch nur-

sing home patients are currently being developed [15].

Our aim was to evaluate drug use in 2 Dutch nursing homes using different prescribing

indicators based on pharmacy prescription data. In our earlier study among nursing home

79

Introduct ion

Appropriateness of prescribing has gained much attention in studies of the quality of health

care [1-5]. This is particularly true for elderly and nursing home patients. In view of the high

rate of drug use, age-related pharmacokinetic and pharmacodynamic changes and multiple co-

morbidity, elderly patients are at a higher risk of adverse drug effects (ADEs) [1,2,6,7]. Schmader

and colleagues defined appropriate prescribing as the selection of a medication and instruc-

tions for its use that agree with accepted medical standards [2]. These standards are based on

efficacy, adverse drug effects and cost-effectiveness and they are derived from national and

international guidelines, clinical trials and expert opinion [2]. Today, the concept of evidence-

based medicine is included in daily medical practice. Evidence-based medicine is not only

based on external clinical evidence, but also on individual clinical expertise [8]. We sought to

evaluate prescribing practices in Dutch nursing homes by assessing tools that could be used

with pharmacy prescription data only.

Several tools have been developed to assess the appropriateness of prescribing in the

elderly [1,5]. Many of these were developed for assessing medication appropriateness in elder-

ly outpatients, rather than nursing home residents. Prescribing indicators used in one health

care system are not automatically applicable to other health care systems due to differences in

national pharmacotherapy guidelines and drug formularies. Furthermore, for many prescribing

indicators information on clinical status such as laboratory results or diagnoses is necessary,

making them unsuitable for use with pharmacy prescription data only. Appropriateness crite-

ria developed by Beers and colleagues [9] were based on expert consensus. They consisted of

a list of 23 medications that should be avoided and 13 medication doses, frequencies, or pre-

scription duration that generally should not be exceeded. An update, including clinical infor-

mation such as information on the prescribing indication and on potassium level monitoring,

was published in 1997 [10]. As Beers’ criteria list several medications that are not available in

the Netherlands or are not in accordance with Dutch pharmacotherapy standards, some of

these criteria cannot be applied to Dutch nursing homes. In 1997, Lunn and co-workers [7] deve-

loped a set of 18 explicit criteria, based on expert opinion, to identify inappropriate prescribing

in 101 nursing home residents in the UK. For 7 of the 18 criteria, information on clinical status

or diagnoses of the residents was necessary, again making them unsuitable for use with phar-

macy prescription data only although some of them could be incorporated. Two Swedish stu-

dies used criteria that were based on Swedish guidelines for measuring excessive use of psy-

chotropic drug use in the elderly [11,12]. In one study [11] the availability of clinical information

for 4 out of 13 criteria was required, in the other study [12] this was the case for 1 out of 10 cri-

78

group of diuretics, as the combination of a loop diuretic and a thiazide diuretic may be someti-

mes therapeutically useful in heart failure and hypertension. The appropriateness of drug

dosage was assessed by comparing the actual prescribed daily dose (PDD) with the recom-

mended dose for the elderly, expressed as the defined daily dose (DDD [20]). For benzodiaze-

pines, the recommended dosage for elderly people is 0.5 DDD [21]. For the other drug groups

the recommended dosage was set on 1 DDD as no specific recommendation for elderly patients

exists. Prescribing of hypnotic benzodiazepines is recommended not to exceed 30 days [18].

This prescribing indicator was assessed for the benzodiazepines only, as for the other drug

groups there were no guidelines concerning the duration of therapy. Furthermore, two drug

combinations were studied, both concerning NSAIDs. First, the co-prescribing of NSAIDs and

loop-diuretics was studied because NSAIDs may decrease the efficacy of diuretics and induce

congestive heart failure [22]. Second, the co-prescribing of gastroprotective drugs (proton pump

inhibitors) during NSAID therapy was studied. In view of the risks of NSAID therapy in the

elderly [23], not to prescribe a gastroprotective drug concomitantly could be regarded as sub-

optimal prescribing.

81

patients [16], we identified signals that the prescribing and use of benzodiazepines, loop diu-

retics, ulcer-healing drugs and non-steroidal anti-inflammatory drugs (NSAIDs) could poten-

tially be improved. Therefore, we focused on these drug groups. To assess the utility of the pre-

scribing indicators we verified the cases of potentially suboptimal prescribing by means of an

interview with one of the prescribers.

Methods

Sett ing

The study was carried out in two nursing homes, one for somatic care (home A; 134 resi-

dents) and one for psychogeriatric care (home B; 120 residents). Five nursing home physicians

(three in nursing home A and two in nursing home B) provided medical care on a daily basis.

Each ward was visited twice a week, and a nursing home physician was on call 24 hours a day.

Both nursing homes were served by the same hospital pharmacy. All drugs dispensed to the

nursing home residents were registered in the hospital computer system. Any changes in medi-

cation were updated routinely on a daily basis in the hospital pharmacy computer system and

a complete medication history was kept for each individual resident. Medication was adminis-

tered to individual nursing home residents based on information recorded in the computer sys-

tem, such as drug, dosage, route and time of administration. Hospital pharmacists carried out

medication surveillance. At the moment of the study there was no computerised medication

surveillance available.

Evaluat ion of drug prescr ib ing by prescr ib ing indicators

We evaluated the prescribing of benzodiazepines, NSAIDs, ulcer-healing drugs, and diure-

tics and compared them with recommendations in national guidelines and a regional drug for-

mulary [17-19]. The drug formulary was based on the International Classification of Primary

Care. We evaluated drug use against this drug formulary [19]. Table 1 presents the drugs listed

in the regional drug formulary.

The prescribing indicators we used fell into two groups, as is shown in table 2. Indicators in

group (a) were descriptive in nature and as a consequence no optimal value was defined. We

used the number and the percentage of users, related to the total number of residents present

in the nursing home. Group (b) indicators reflected potentially suboptimal prescribing.

Examples of these indicators that were applied to all four drug groups were use of drugs out-

side the formulary and use of more than one drug from the same therapeutic drug class (e.g.

two benzodiazepines or two ulcer-healing drugs). The latter indicator was not applied to the

80

Table 1: Drugs listed in the regional drug formulary

Formulary drugs

NSAIDs

Ibuprofen

Diclofenac

Diclofenac + misoprostol (fixed combination)

Meloxicam

Naproxen

Ulcer-healing drugs

Histamine H2-receptor antagonists

- Cimetidine

- Ranitidine

Proton pump inhibitors

- Omeprazole

Benzodiazepines

Hypnotics

- Temazepam

- Nitrazepam

Anxiolytics

- Oxazepam

- Diazepam

Diuretics

Furosemide

Hydrochlorothiazide

83

Ver i f i cat ion of prescr ib ing indicators

In a sample (n=25) of patients, reflecting the range of patients in whom the indicators sug-

gested potentially suboptimal prescribing, the medical charts were reviewed together with

information from the prescribers to ascertain if prescribing for these patients was indeed sub-

optimal. The information was collected during an open-structured 3-hour interview.

Results

Evaluat ion of drug prescr ib ing by prescr ib ing indicators

The results are summarised in table 2 and are reported below.

Benzodiazepines. Of the residents in nursing home A and B, the percentage of benzodia-

zepine users was 31% and 28% respectively. Of the benzodiazepine users in nursing home A

and B the percentage of hypnotic users was 28% and 16% respectively. The percentage of

anxiolytic users was 5% and 13%. The percentage of users that were prescribed daily dosages

above 0.5 DDD was 27% and 24% respectively. Two patients were prescribed a non-formulary

benzodiazepine. Fifteen and 3% of the benzodiazepine users in nursing home A and B respec-

tively received more than one benzodiazepine at the same time. The number of benzodiazepi-

ne users who were prescribed hypnotic benzodiazepines for more than 30 days was 38 (93%)

and 29 (88%).

NSAIDs. Of the residents in nursing home A and B, the percentage of NSAID users was 10%

and 5%, respectively. The percentage of users that were prescribed dosages above 1 PDD was

50% and 17%, respectively. All NSAIDs prescribed were formulary drugs. There were no

patients with concomitant use of more than 1 NSAID. Twenty-one and 17% of the NSAID users

were prescribed a loop diuretic simultaneously. Seventy-nine (home A) and 100% (home B) of

the NSAID users were not prescribed a gastroprotective drug.

Diuretics. Of the residents in nursing home A and B, the percentage of diuretic users was

31% and 13%, respectively. The percentage of users that were prescribed dosages above 1 PDD

was 17% and 19% respectively. All diuretics prescribed were formulary drugs.

Ulcer-healing drugs. Of the residents in nursing home A and B, the percentage of users of

ulcer-healing drugs was 25% and 13% respectively. The percentage of users that were prescri-

bed dosages above 1 PDD was 24% and 25% respectively. All ulcer-healing drugs prescribed

were formulary drugs. There were no patients prescribed more than 1 ulcer-healing drug.

82

Tab

le 2

: Res

ults

of t

he e

valu

atio

n of

dru

g us

e by

mea

ns o

f pre

scri

bing

indi

cato

rs in

two

Dut

ch n

ursi

ng h

omes

Indi

cato

rBe

nzod

iaze

pine

sN

SAID

sDi

uret

ics

Ulc

er-h

ealin

g dr

ugs

Type

of n

ursi

ng h

ome|

Hom

e A

Hom

e B

H

ome

A H

ome

B

Hom

e A

Hom

e B

Hom

e A

Hom

e B

(num

ber o

f res

iden

ts)

(n=

134)

(n=

120)

(n=

134)

(n=

120)

(n=

134)

(n=

120)

(n

=13

4)(n

=12

0)

a. D

escr

iptiv

e pr

escr

ibin

g in

dica

tors

Num

ber (

% o

f use

rs¥

)41

(30.

6)33

(27.5

)14

(10.

4)6

(5.0

)41

(30.

6)16

(13.

3)34

(25.

3)16

(13.

3)

hypn

otic

s: 3

7 hy

pnot

ics:

19

(27.6

)(15

.8)

anxi

olyt

ics:

7

anxi

olyt

ics:

15

(5.2

)(1

2.5)

b. In

dica

tors

ass

essi

ng p

oten

tial s

ubop

timal

pre

scri

bing

(num

ber a

nd p

erce

ntag

e of

use

rs¶ )

PDD

> 0

.521

11 (2

6.8)

8 (2

4.2)

PDD

> 1

--

7 (5

0.0)

1 (16

.7)7

(17.1

) 3

(18.8

)8

(23.

5)4

(25.

0)

[7] #

[5] #

[5] #

Use

of d

rugs

out

side

1 (

2.4)

2 (6

.1)0

(0)

0 (0

)0

(0)

0 (0

)0

(0)

0 (0

)

form

ular

y

Use

of >

1 dr

ug fr

om

6 (1

4.6)

1 (3.

0)0

(0)

0 (0

)-

-0

(0)

0 (0

)

sam

e dr

ug c

lass

Use

> 3

0 da

ys 18

38 (9

2.7)

29 (8

7.9)

--

--

--

Com

bina

tion

with

--

3 (2

1.4)

1 (16

.7)-

--

-

loop

-diu

retic

18

No

com

bina

tion

with

-

-11

(78.

6)6

(100)

--

--

gast

ropr

otec

tive

drug

18[8

] #

|nu

rsin

g ho

me

A pr

ovid

es m

ainl

y so

mat

ic c

are,

nur

sing

hom

e B

pro

vide

s m

ainl

y ps

ycho

geri

atri

c ca

re.

¥de

fined

as

the

num

ber o

f pat

ient

s di

vide

d by

the

num

ber o

f res

iden

ts in

the

nurs

ing

hom

e*10

0¶

defin

ed a

s th

e nu

mbe

r of p

atie

nts

divi

ded

by th

e nu

mbe

r of u

sers

of t

he d

rug

grou

p*10

0#

[ ]=

num

ber o

f pat

ient

s in

clud

ed in

the

inte

rvie

w fo

r ver

ifyin

g pr

escr

ibin

g in

dica

tors

Discuss ion

In this study prescribing indicators based on pharmacy prescription data were used to iden-

tify prescribing that was not in line with regional or national guidelines. We found that pre-

scribing practices in 2 Dutch nursing homes were generally in agreement with national pre-

scribing guidelines and the regional drug formulary. A discussion of the findings is given below.

Evaluat ion of drug use prescr ib ing by prescr ib ing indicators

Number and percentage of users of drug groups

In the nursing home for somatic care, approximately twice as many hypnotics, NSAIDs,

ulcer-healing drugs and diuretics were prescribed compared with the nursing home for psy-

chogeriatric care. This may reflect the somatic disorders these residents are suffering from. This

descriptive indicator reflects overall prescribing practice, and can be used to monitor changes

in prescribing in-house over time. Comparison of prescribing practices between these homes is

difficult in view of the differences in co-morbidity.

Dosage of drug groups

The percentage of the residents receiving dosages higher than recommended (for benzo-

diazepines 0.5 DDD and for NSAIDs, ulcer-healing drugs and diuretics 1 DDD) varied among the

nursing homes, with a minimum of 17% and a maximum of 50% of the residents affected. From

the interview data it was found that often the high dosages were the result of titration of the

dosage based on the clinical effect. This was the case in particular for NSAIDs, diuretics and

ulcer-healing drugs. This indicator does not necessarily reflect suboptimal prescribing regar-

ding these drug groups. Insight in the indication for which the drug is prescribed is needed to

evaluate whether a dosage is too high. In general, monitoring side effects of dosages above 1

DDD is recommended in view of the increased susceptibility of elderly patients to adverse drug

effects.

Use of non-formulary drugs

Overall, 3 patients (1% of the study population) were prescribed non-formulary drugs for

the drug groups studied. Three patients received non-formulary benzodiazepines (flurazepam

and midazolam). For these drugs alternative formulary drugs were available and recommen-

dations with regard to substitution could be made.

Duplication of drugs

More than one drug from the same drug class was prescribed to 0%-13% of the residents

and it concerned benzodiazepines and diuretics. In case of benzodiazepines, it may be worth-

while to limit prescribing to one benzodiazepine.

85


The medication of 25 patients (all from nursing home A) with potentially suboptimal pre-

scribing was reviewed using the medical charts and subsequently discussed with one of the

prescribing nursing home physicians. We selected eight patients who were prescribed an

NSAID and a proton pump inhibitor (PPI) concomitantly. We inquired whether the PPI was pre-

scribed to counteract the gastrotoxicity of the NSAID. Indeed, in two patients the PPI was pre-

scribed to treat gastro-intestinal adverse effects of the NSAID. For the other six patients other

reasons for prescribing a PPI existed. Two patients had a hernia diaphragmatica and were pre-

scribed a PPI to prevent erosive damage due to reflux oesophagitis, and one was diagnosed

with an ulcus duodeni. One patient was diagnosed with reflux oesophagitis, and therapy with

an H2-antagonist was insufficiently effective. One patient experienced nausea and vomiting as

a result of anti-Parkinson drug therapy (levodopa/carbidopa), and was subsequently prescri-

bed a PPI. One patient was bedridden due to spinal-cord injury and was prescribed the PPI to

prevent erosive damage due to reflux oesophagitis. Five patients were prescribed an ulcer-

healing drug (proton pump inhibitor) in dosages higher than 1 PDD (equivalent to 40 mg ome-

prazole). According to the nursing home physician, this might have been due to the fact that

some prescribers tend to start with a high dosage to effectively heal the symptoms and taper

the dosage when acute symptoms have diminished. Three patients were diagnosed with ulcus

ventriculi or ulcus duodeni and were therefore prescribed ulcer-healing drugs in these dosa-

ges. One of these patients was first prescribed an H2-receptor antagonist, but experienced cen-

tral adverse effects. Of the other 2 patients, one patient was diagnosed with reflux oesophagi-

tis and hiatus hernia and was prescribed a proton-pump inhibitor in high dosage by a medical

specialist. This therapy was subsequently continued. The other patient was prescribed metho-

trexate and experienced nausea that responded well to proton-pump inhibitor therapy.

According to the prescriber, side effects were not seen with these high dosages of proton-pump

inhibitors. Seven patients were prescribed NSAIDs above the recommended dosage. According

to the nursing home physician, this was the result of careful dose adjustments that ultimately

led to these relatively high dosages. Three of these patients were prescribed paracetamol in

dosages of 2-4 g before the NSAID was started. Severe rheumatoid arthritis and severe pain

were reasons for prescribing NSAIDs in such high dosages. The necessity for these high dosa-

ges was re-evaluated periodically as well as the occurrence of potential gastro-intestinal and

renal side effects. Five patients were prescribed loop diuretics in a dosage higher than recom-

mended. These patients all had a diagnosis of heart failure. The high dosages were the result

of careful dose adjustments that had ultimately led to these relatively high dosages. Metabolic

disorders such a hypokalaemia were frequently monitored by measuring plasma potassium

levels.

84

because we also reviewed medical charts, most information on prescribing and medical diag-

noses could be traced. Another limitation of our study was that we did not verify all prescribing

indicators used in the drug evaluation, such as whether use of more than one drug from the

same drug class was justified. Indicators that are to reflect suboptimal prescribing should be

sensitive and specific. It is often difficult to derive prescribing indicators solely from guidelines

and drug formularies. This is particularly true for elderly patients, in view of the complex co-

morbidity and often tailor-made pharmacotherapy on the basis of clinical parameters. Efforts

should therefore be directed towards the development of indicators that take into account

these issues.

In conc lus ion

This pilot study showed that in two Dutch nursing homes, drugs were prescribed according

to regional and national prescribing guidelines. Prescribing indicators were used to evaluate

drug prescribing and may reflect potentially suboptimal prescribing. However, clinical infor-

mation from the prescriber was necessary to get insight into the appropriateness of prescribing.

The prescribing indicators used were a useful tool to evaluate prescribing practices. Because

we studied only two nursing homes, that were not necessarily representative of the Dutch nur-

sing home population, generalised conclusion can not be made.

Acknowledgements

We express our gratitude to J. Tideman, for his valuable help in collecting clinical data and

his contribution to the interview. A. Van den Brand, is thanked for his co-operation in carrying

out this study. Furthermore, we are grateful to the nursing home staff of the nursing homes and

the pharmacy staff of the hospital pharmacy for their co-operation in collecting prescription

data.

87

Combination of drugs

Two indicators assessed the combination of drugs. One indicator identified prescribing of

an NSAID and a loop diuretic, which was the case in 17% to 21% of the NSAID users were affec-

ted, and prescribing practices may be improved on this point in view of the increased risk of

renal failure due to this drug-drug interaction. The other indicator assessed potential subopti-

mal prescribing when no gastroprotective drug was prescribed together with an NSAID.

Seventy-nine percent to 100% of the NSAID users were not prescribed a gastroprotective drug

concomitantly. These results indicate that prescribing practices can be improved.

Duration of drug use

For benzodiazepine users, one indicator assessed the duration of drug use. Nearly 100% of

the benzodiazepine users used hypnotic benzodiazepines for more than 30 days, which is

against Dutch guidelines [18]. Although this indicator reflects suboptimal prescribing, it is often

difficult to withdraw benzodiazepines from patients [24,25]. However, in view of the disadvan-

tageous risk-benefit ratio of the benzodiazepines, tapering of benzodiazepine use should be

attempted. Furthermore, restricting prescription of benzodiazepines for a maximum duration of

2-4 weeks, and regularly assessing the necessity of benzodiazepine prescribing can improve

prescribing practice [26].


From the interview and chart review data it was found that the prescribing indicators we

investigated did not always reflect suboptimal prescribing. This has also been found by others

[27,28]. An indicator that performed well was the combination of gastroprotective drugs and

NSAIDs. This indicator may reflect suboptimal prescribing in view of the risks of gastrotoxicity

of NSAIDs in the elderly [23]. Recently, the nursing homes under study have changed their pre-

scribing policies on this point. Currently, guidelines recommend prescribing gastroprotective

medication to all elderly people who chronically use NSAIDs. Indicators that assessed drug

dosages above recommended values for NSAIDs, ulcer-healing drugs and loop diuretics, did not

perform well. Often good reasons for prescribing these high dosages existed, the main reason

being that lower dosages were insufficiently effective. Potential side effects were known to the

prescribers and monitored periodically. Furthermore, drug doses are often dependent on the

indication, and several ‘ideal’ dosages per drug may exist depending on the indication. DDD

values have been developed for other purposes than monitoring prescription appropriateness

and therefore are unsuitable to assess appropriateness of drug dosages of these drug groups.

We interviewed only one nursing home physician, and therefore it was not always possible to

find out the exact reasons for prescribing by colleague-nursing home physicians. However,

86

23 Fries JF. NSAID gastropathy: the second most deadly rheumatic disease? Epidemiology and risk appraisal.

J Rheumatolog 1991; 18 (suppl 28): 6-10.

24 Busto U, Sellers EM, Naranjo CA, Capell H, Sanchez-Craig M, Sykora K. Withdrawal reaction after long-term use

of benzodiazepines. N Engl J Med 1986; 315: 854-9.

25 Van Hulten R. Blue Boy- why not? Studies of benzodiazepine use in a Dutch community [Thesis]. University of

Utrecht, 1998.

26 Eide E, Schjøt J. Assessing the effects of an intervention by a pharmacist on prescribing and administration of

hypnotics in nursing homes. Pharm World Sci 2001; 23: 227-31.

27 Avery AJ. Appropriate prescibing in general practice: development of the indicators. Qual Health Care 1998; 7: 123.

28 Brook RH, McGlynn EA, Cleary PD. Quality of health care. Part 2: Measuring quality of care. New Engl J Med

1996; 335: 966-70. 89

References


measures. Drugs Aging 2000; 16: 437-50.

2 Schmader K, Hanlon JT, Weinberger M, Landsman PB, Samsa GP, Lewis I, et al. Appropriateness of medication

prescribing in ambulatory elderly patients. J Am Geriatr Soc 1994; 42: 1241-7.

3 Coste J, Venot A. An epidemiologic approach to drug prescribing quality assessment. Med Care 1999; 37: 1294-1307.

4 Buetow SA, Sibbald B, Cantrill JA, Halliwell S. Prevalence of potentially inappropriate long term prescribing in

general practice in the United Kingdom, 1980-95: systematic literature review. BMJ 1996; 313: 1371-74.

5 Knight EL, Avorn J. Quality indicators for appropriate medication use in vulnerable elders.

Ann Intern Med 2001: 135; 703-10.

6 Lindley CM, Tulley MP, Paramsothy V, Tallis RC. Inappropriate medication use is a major cause of adverse drug

reactions in elderly patients. Age Ageing 1992; 21: 294-300.



8 Sackett DL, Scott Richardson W, Rosenberg W, Brian Haynes R. Evidence-based medicine: how to practice and

teach EBM. London: Churchill Livingstone, 1998.

9 Beers MH, Ouslander JG, Rollingher I, Reuben DB, Brooks J, Beck J. Explicit criteria for determining

inappropriate medication use in nursing homes. Arch Intern Med 1991; 151: 1825-32.

10 Beers MH. Explicit criteria for determining potentially inappropriate medication use by the elderly. Arch Intern

Med 1997; 157: 1531-6.

11 Schmidt IK, Claesson CB, Westerholm B, Svarstad BL. Resident characteristics and organizational factors


12 Schmidt IK, Fastbom J. Quality of drug use in Swedish nursing homes. A follow-up study. Clin Drug Invest 2000;

20: 433-46.

13 Hanlon JT, Schmader KE, Samsa GP, Weinberger M, Uttech KM, Lewis IK, et al. A method for assessing drug

therapy appropriateness. J Clin Epidemiol 1992; 45: 1045-51.

14 Veninga CCM, Denig P, Pont LG, Haaijer-Ruskamp FM. Comparison of indicators assessing the quality of drug

prescribing for asthma. Health Services Research 2001; 36: Part 1: 143-61.

15 Rohling R, Berman M, Froeling PGAM, Van de Gronde ThW, Van Leen MWF, Pessers RWAM, et al. Richtlijn

Hartfalen [Guideline on heart failure of the Dutch Association of Nursing Home Physicians (NVVA)].

Tijdschr Verpleeghuisgeneeskd 1999; 23: 3-19.



17 Thomas S, Geijer RMM, van der Laan JR, Wiersma T.[Guidelines for the physician II] (in Dutch). Bunge, Utrecht,

1996.

18 Van der Kuy A, editor. Farmacotherapeutisch Kompas 2000-2001 (in Dutch). 17th ed. Amstelveen: Commissie

Farmaceutische Hulp (CFH), 2000.

19 Van der Graaf CJ, Hospes W (Ed). Formularium uitgave Assen (in Dutch), 1998.

20 Anonymous. Anatomical therapeutical chemical (ATC) classification index including defined daily dosages

(DDDs) for plain substances. World Health Organisation Collaborating Centre for Drug Statistics Methodology,

Oslo, 2000.

21 Knuistingh Neven A, Graaff de WJ, Lucassen PLBJ et al. [Dutch College of General Practitioners-guideline

‘Insomnia and hypnotics’] (in Dutch) Huisarts Wet 1992; 35: 212-9.

22 Heerdink ER, Leufkens HG, Herings RM, Ottervanger JP, Stricker BH, Bakker A. NSAIDs associated with

increased risk of congestive heart failure in elderly patients taking diuretics. Arch Intern Med 1998; 158: 1108-12.

88

Abstract

Objective: Concomitant prescribing of benzodiazepines may occur more frequently with selec-

tive serotonin reuptake inhibitors (SSRIs) than with tricyclic antidepressants (TCAs), partly due

to the milder sedating effects of SSRIs. However, drug utilisation studies in this area show con-

flicting results.

Methods: To investigate whether prevalence and incidence of benzodiazepine drug prescribing

is comparable between users of SSRIs and users of TCAs, a follow-up study was performed in

two different cohorts: an ambulatory and a nursing home cohort both aged ≥ 65 years. In each

cohort the incidence and prevalence of benzodiazepine use during antidepressant therapy was

estimated. TCAs and SSRIs were subsequently compared.

Results: The ambulatory population consisted of 14,336 people (58% female), mean age 74 (±

7) years. Co-prescribing of benzodiazepines occurred in 53% of the TCA users and 57% of the

SSRI users (prevalence RR 1.1; CI95 0.9-1.2). Concomitant drug therapy was >65 days per 100

days of antidepressant drug use. During SSRI therapy, significantly more people started with

benzodiazepine drug therapy than during TCA therapy (incidence RR 1.7; CI95 1.2-2.4). Analyses

repeated 5 years later yielded similar results (overall incidence RR 1.6 (CI95 1.3-2.0). The nursing

home population consisted of 2,355 residents (73% female), mean age 82 (± 7) years. Co-pre-

scribing of benzodiazepines was slightly higher than in the ambulatory population: 59% in

TCA users and 57% in SSRI users (prevalence RR 1.0; CI95 0.8-1.2). No difference was found in

the incidence of benzodiazepine starts between SSRI and TCA therapy (incidence RR 0.8; CI95

0.4-2.0).

Conclusion: More than 50% of antidepressant users receive a benzodiazepine at the same

time. In the ambulatory population, it was found that SSRI use is associated with a significant-

ly higher chance of initiating benzodiazepine therapy compared with TCA use. In the nursing

home population, no such difference was found.

91

2.6 Concomitant prescr ib ing of benzo-diazepines dur ing ant idepressant therapy in the e lder ly

K.N. van Dijk 1,2, C.S. de Vries 3, K. ter Huurne 1, P.B. van den Berg 1,J.R.B.J. Brouwers1,2, L.T.W. de Jong-van den Berg 1


University Institute for Drug Exploration (GUIDE), University Centre for Pharmacy, Groningen,

the Netherlands2 Department of Clinical Pharmacy, Medical Centre Leeuwarden, Leeuwarden, the Netherlands3 Department of Pharmacoepidemiology, Postgraduate Medical School, University of Surrey,


A modified version is accepted for publication in the Journal of Clinical Epidemiology

90

study in an ambulatory cohort as well as a nursing home cohort, both aged 65 and over.

Methods

Design

Using computerised pharmacy records a prospective follow-up study was performed to

estimate the prevalence rate ratio (RR) as well as the incidence RR of benzodiazepines pre-

scribed during antidepressant therapy. Differences in benzodiazepine drug use were compared

between SSRI users (index group) and TCA users (reference group).

Study populat ion

Ambulatory cohort. Pharmacy dispensing data from the InterAction database were used for

this study. This database is a collaboration between community pharmacists in the northern

part of the Netherlands and the University of Groningen [20]. The database contains complete

medication profiles of individual patients from a number of pharmacies from 1994 onwards.

Since Dutch patients are generally registered to only one pharmacy and community pharma-

cies in the Netherlands are completely automated, complete individual medication histories

are available in community pharmacies. Community pharmacy records are reported to be a

reliable source of drug exposure [21]. Two 2-year cohorts were defined: one cohort included

those registered in the InterAction database during the years 1994-1995 and one cohort inclu-

ded those registered during the years 1998-1999. The second cohort was selected to investiga-

te whether changes over time led to different outcomes. During 1994-1995, the source popula-

tion consisted of 14,067 people aged ≥ 65 for whom a complete medication history was availa-

ble. During the years 1998-1999, the source population consisted of 17,060 people aged ≥ 65.

The increase in source population was due to an increased number of pharmacies participating

in the database. During 1994-1995 10 pharmacies participated and this increased to 12 pharma-

cies in 1998-1999.

Nursing home cohort. The study was conducted among residents aged 65 years and over in

six nursing homes with a 90 to 225-bed capacity each. The nursing home study population con-

sisted of 2,355 residents residing at any time during the two-year study period (1 October 1993

to 1 October 1995). The databases and drug use in this population have been previously descri-

bed in detail [22,23].

93

Introduct ion

Choosing appropriate antidepressant therapy in the elderly has been subject of debate in

clinical practice [1-4]. Particular attention has been given to the adverse effects of these drugs,

as elderly subjects may be more sensitive to adverse drug effects due to pharmacokinetic, phar-

macodynamic and disease-related changes that occur with advanced age. In elderly patients,

adverse effects associated with the use of tricyclic antidepressants (TCAs) include postural

hypotension, anticholinergic effects, and extrapyramidal symptoms [5]. Adverse effects of

selective serotonin reuptake inhibotors (SSRIs) include syndrome of inappropriate antidiuretic

hormone secretion, gastrointestinal disturbances and insomnia [5]. As a result of the unwan-

ted anticholinergic adverse effects of TCAs in the elderly, the use of SSRIs in the elderly is incre-

asing [3]. However, the preferential use of the more expensive SSRIs in the elderly is still under

debate [6-8]. For example, an increased risk of falls and fractures as a consequence of antide-

pressant drug use has been reported for both TCAs and SSRIs [9,10]. In addition, the need for

concomitant benzodiazepines during antidepressant therapy has been debated [11-15]. These

drugs decrease the non-specific symptoms of depression, such as insomnia, agitation and

anxiety, and are often used during the initiation of antidepressant treatment. After 4-6 weeks

of treatment the benzodiazepine is usually discontinued [16]. From two drug utilisation studies,

it has been suggested that concomitant prescribing of benzodiazepines may occur more fre-

quently in users of SSRIs than in users of TCAs, and this may be partly due to the less sedative

effects of SSRIs [12-13]. However, a more recent study in Sweden, using a cross-sectional design,

demonstrated that among TCA users the frequency of concomitant benzodiazepine prescribing

was higher than among SSRI users [14]. In view of the adverse effects of benzodiazepines in

the elderly, such as drowsiness, excess sedation, memory impairment and risk of falls [17-19], it

is worthwile to know whether use of these drugs differs between TCA and SSRI users in the

elderly. Furthermore, a difference in co-prescribing of benzodiazepines between users of TCAs

and SSRIs could constitute an additional selection criterion for choosing antidepressant thera-

py in the elderly.

A limitation of the cross-sectional study [14] was that no time sequence could be studied

and the benzodiazepines could well have been prescribed before initiation of antidepressant

therapy. To investigate changes in anxiolytic/ hypnotic benzodiazepine drug prescribing during

antidepressant therapy (SSRI or TCA) in more detail, we performed a follow-up study among

elderly subjects. The aim of this study was to investigate whether the prevalence and inciden-

ce of benzodiazepine drug prescribing differed between SSRI and TCA users. To determine

whether differences exist between ambulatory and institutionalised elderly, we performed this

92

son-days both during SSRI time and TCA time at risk. The incidence rate ratio with correspon-

ding 95% confidence intervals was determined as IRSSRI / IRTCA. To investigate whether chan-

ges over time led to different outcomes, the incidence rates of concomitant prescribing during

TCA use, during SSRI use and during use of both drug groups of the 1998-1999 cohort (cohort

II) were compared with the incidence rates of concomitant prescribing of the 1994-1995 cohort

(cohort I). The incidence rate ratio with corresponding 95% confidence intervals was determi-

ned as IRcohort II/IRcohort I. To investigate differences between the ambulatory cohort (1994-1995)

and the nursing home cohort, the incidence rates of both cohorts were compared. The statisti-

cal software program SPSS 9.0 for Windows (SPSS Inc., Chicago, IL) was used for all analyses.

Results


In table 1 characteristics of the study populations are given. More elderly people and more

women were present in the nursing home cohort, compared with both ambulatory cohorts. In

the 1998-1999 cohort, a higher percentage of benzodiazepine users was found, compared with

the 1994-1995 ambulatory cohort.

95

Prevalences

The 2-year prevalence of benzodiazepine (Anatomical Therapeutic Chemical (ATC) [24]

codes N05BA (anxiolytics; benzodiazepine derivates), N05CD (hypnotics; benzodiazepine deri-

vates) and N05CF (benzodiazepine-related hypnotics (zolpidem and zoplicone)) prescribing

among SSRI and TCA users (ATC code N06AB and N06AA respectively) was estimated by divi-

ding the number of SSRI users and TCA users who were prescribed a benzodiazepine for at

least 7 days by the total number of SSRI users and TCA users respectively. Each patient was

included only once in these analyses. The prevalence rate ratio with the corresponding 95%

confidence intervals was calculated by prevalence (SSRI) / prevalence (TCA). Furthermore, the

average duration of concomitant drug use was calculated as the number of days of drug use

divided by the total number of days of antidepressant use (SSRIs and TCAs respectively), for

those residents who used the combination for at least 7 days. To investigate whether changes

over time led to different outcomes, the prevalence of concomitant prescribing during TCA use,

during SSRI use and during use of both drug classes in the 1998-1999 cohort (cohort II) was

compared with the prevalence of concomitant prescribing in the 1994-1995 cohort (cohort I).

The prevalence rate ratio with corresponding 95% confidence intervals was determined as pre-

valence (cohort II) / prevalence (cohort I). To investigate differences between the ambulatory

cohort (1994-1995) and the nursing home cohort, the prevalences in both were compared.

Inc idence rates

Patients were considered to be ‘at risk’ for initiation of a benzodiazepines during the time

period of antidepressant therapy in which no benzodiazepines were used. The start of a ben-

zodiazepine during SSRI or TCA use, and subsequent concomitant use of at least 7 days, was

considered an event. When the start of a benzodiazepine coincided with the start of an antide-

pressant, or with the first day of the study period, the start was not considered an event. In this

way, only incident cases of benzodiazepine drug therapy during antidepressant drug therapy

were included in the analyses. We performed a sensitivity analysis in which we included starts

of benzodiazepines that coincided with the start of an antidepressant, to see whether this led

to different outcomes. For each patient, only the first episode of antidepressant drug use in the

study period was considered. We investigated incidence rates in two 2-year study periods: in

1994-1995 (cohort I) and 1998-1999 (cohort II). We blinded ourselves to drug use prior to the 98-

99 study period for better comparison with the earlier cohort. Thus, incidence rates were cal-

culated in exactly the same way for both cohorts. For each patient, only the first episode of

antidepressant drug use in the study period was considered. We investigated incidence rates

(IRs) by dividing the number of events (start of benzodiazepine) by the total number of per-

94

Table 1: Patient characteristics and use of benzodiazepines and antidepressants in both study populations

Variable Ambulatory cohort Ambulatory cohort Nursing home

’94-’95 ’98-’99 cohort ‘94-’95

(n=14,067) (%) (n=17,060)(%) (n=2,355) (%)

Age (yrs ± SD) 75.2 (±7.1) 76.5 (± 7.1) 82.0 (± 7.3)

Gender

Female 8220 (58.4) 9996 (58.6) 1666 (70.7)

Male 5847 (41.6) 7064 (41.4) 689 (29.3)

Drug use

Anxiolytics/ 4063 (28.9) 6105 (35.8) 1530 (65.0)

Hypnotics¶

TCAs¥ 600 (4.3) 832 (4.9) 283 (12.0)

SSRIs# 298 (2.1) 658 (3.9) 91 (3.9)

¶ ATC-codes [22] N05BA, N05CD and N05CF¥ ATC-code [22] N06AA# ATC-code [22] N06AB

benzodiazepine drug therapy (incidence RR 1.72; CI95 1.23-2.42). In the second ambulatory

cohort, a similar result was found: SSRI users were at a 57% higher risk for the start of a ben-

zodiazepine compared with TCA users. In the nursing home cohort, no association was found:

the incidence RR of benzodiazepine drug therapy during SSRI use compared to TCA use was

0.84 (CI95 0.36-1.96). There were no statistically significant differences between the nursing

home cohort and the first ambulatory cohort in the incidence of concomitant prescribing of

benzodiazepines among among TCA users, among SSRI users, or among both drug classes. No

statistically significant differences were found between the second and the first cohort in the

incidence of concomitant prescribing of benzodiazepines among TCA users, or among SSRI

users. However, when the concomitant use of benzodiazepines during both SSRIs and TCAs in

the 1998-1999 cohort was compared with the 1994-1995 cohort, the overall incidence RRMH was

0.82 (CI95 0.67-1.00), indicating 18% less concomitant drug use in the more recent years. In the

results presented above, when the start of a benzodiazepine coincided with the start of an anti-

depressant, the start was not considered an event. A sensitivity analysis in which we included

these events (‘prophylactic starts’) led to an incidence rate ratio of 1.60 (CI95 1.23-2.09) for the

first cohort (1994-1995) and 1.34 (CI95 1.12-1.61) for the second cohort (1998-1999). In particular in

the second cohort, the point estimate was slightly lower than the IRR without inclusion of pro-

phylactic starts (IRR 1.57 (CI95 1.24-1.98)), however the difference between SSRI users and TCA

users was still statistically significant.

97

Prevalence rat ios

In table 2 the prevalence rate ratios for concomitant use of benzodiazepines during antide-

pressant therapy for both ambulatory cohorts and the nursing home cohort are given.

Furthermore, the average duration of concomitant drug use is given. There was no difference

in prescribing of benzodiazepines for SSRI users and TCA users in the first ambulatory cohort.

In the second ambulatory cohort the concomitant prescribing of benzodiazepines was slightly

higher among SSRI users. In the nursing home cohort, there was no difference in concomitant

prescribing of benzodiazepines between SSRI users and TCA users. There were no statistically

significant differences between the nursing home cohort and the 1994-1995 ambulatory cohort

in concomitant drug use among TCA users, among SSRI users or among both drug classes com-

bined. Table 2 shows that in all cohorts the number of days of concomitant drug use per 100

days of antidepressant drug use was more than 67. There were no statistically significant diffe-

rences between the second and the first ambulatory cohort in concomitant drug use among TCA

users, among SSRI users or among both drug classes.

Inc idence rate rat ios

In table 3, the incidence rate ratios for all cohorts are presented. In the first ambulatory

cohort, use of SSRIs, compared with TCAs, was associated with a significantly increased risk of

96

Table 2: Prevalence rate ratios for concomitant benzodiazepine drug use during TCA and SSRI use

Drug use Number Number of % of days Prevalence CI95

of users concomitant of concomitant RR

(%a) benzodiazepine drug usec

users (%b)

Cohort I TCA 600 (4.3) 319 (53.2) 76.2 1.0

1994-1995

(n=14,067) SSRI 298 (2.1) 169 (56.7) 67.2 1.07 0.94-1.21

Cohort II TCA 832 (4.9) 437 (52.5) 73.2 1.0

1998-1999

(n=17,060) SSRI 658 (3.9) 414 (62.9) 70.3 1.20 1.10-1.31

Nursing TCA 286 (12.1) 170 (59.4) 83.7 1.0

home cohort

1994-1995 SSRI 93 (3.9) 53 (57.0) 82.9 0.96 0.78-1.17

(n=2,355)a % of total study populationb % of TCA and SSRI users, respectivelyc defined as the number of days of concomitant drug use per 100 days of antidepressant drug use

Table 3: Incidence rate ratios for the start of benzodiazepine drug prescribing during antidepressant therapy:

SSRIs compared to TCAs

Drug use Number Number of IR (x 10-3)a Incidence CI95

of events days at risk RR

Cohort I TCA 102 48,237 2.11 1.0

1994-1995

(n=14,067) SSRI 49 13,449 3.64 1.72 1.23-2.42

Cohort II TCA 143 80,187 1.78 1.0

1998-1999

(n=17,060) SSRI 136 48,613 2.80 1.57 1.24-1.98

Overall incidence rate ratio 1.61 1.33-1.96

Nursing TCA 23 10,855 2.12 1.0

home cohort

1994-1995 SSRI 7 3,932 1.78 0.84 0.36-1.96

(n=2,355)

a number of events divided by the number of days at risk

diately. After 4-6 weeks when the antidepressant drug has reached its effect, the benzodiaze-

pine should be tapered and discontinued [16]. The results of our analyses indicate that the

simultaneous use is long-term, and most probably not used as a temporary measure associa-

ted with initiation of antidepressant treatment. The difference of nearly 7% in the percentage

of users of benzodiazepines between the two ambulatory cohorts (table 1) could be the conse-

quence of a higher prevalence of psycho-somatic disorders or sleep disorders in the second

cohort. However, as we do not have information on the indication for which the drugs were pre-

scribed, we cannot prove this assumption. In a Dutch report on the appropriate use of benzo-

diazepines issued in 1998 [25], it was stated that in certain cases it is appropriate to treat elder-

ly people suffering from psychosomatic disorders with benzodiazepines. This may further

explain the higher frequency of benzodiazepine users in the second cohort.

Bingefors and colleagues [14] found a difference in the concomitant prescribing of anxioly-

tic/hypnotic drugs between TCA and SSRI users: concomitant prescribing occurred more fre-

quently in TCA users than in SSRI users (ORadj 1.3 (CI95 1.0-1.6). They concluded that the concern

about increased anxiolytic/hypnotic drug use among SSRI users seems to be unfounded in

Sweden. A limitation of their study was they could only study concomitant prescribing of drugs

that were dispensed on the same day [14]. By using a follow-up design, we were able to esti-

mate the incidence rate ratios for initiating benzodiazepine drug therapy during antidepressant

therapy. In this way, differences between SSRI and TCA users can be measured more accurate-

ly. Several other studies have reported on concomitant prescribing of these two drug groups,

although differences in study design and setting make comparison difficult. Gregor [13] studied

concomitant use of anxiolytics and hypnotics with SSRIs in patients younger than 65 years.

Concomitant anxiolytic and hypnotic use occurred in 9.8% and 2.8% of the patients, respecti-

vely, and was more common among patients treated with paroxetine than among patients tre-

ated with sertraline or fluoxetine. Rascati [12] showed that 35% of patients receiving SSRIs or

clomipramine used anxiolytic/hypnotic drugs concomitantly. Parkes [11] suggested that the use

of newer antidepressants, mainly SSRIs, in a group of veterans resulted in a 48% increase in

prescriptions for benzodiazepines, presumably to manage associated insomnia. The prevalen-

ces in our study are higher than in the above mentioned studies. These differences might be

explained by the differences in study design, such as the setting and age limits chosen. Also, we

studied prevalence during a 2-year period, while most studies used a shorter study period,

giving lower prevalences.

99

Discuss ion

In this study we found that in ambulatory elderly the risk for initiating benzodiazepines

during antidepressant therapy was higher for users of SSRIs than for users of TCAs (overall

incidence RR 1.6; CI95 1.3-2.0). In the second cohort (1998-1999), slightly fewer antidepressant

users started benzodiazepine drugs compared with the first cohort (1994-1995), indicating that

the rate of concomitant prescribing decreased over the years. In both cohorts there was a higher

frequency of concomitant prescribing among SSRI users compared with TCA users. Several

explanations for this finding can be given. The most likely reason for the increased risk among

SSRI users compared with TCA users, also suggested in other studies [12,13], is that the less

sedating effects of SSRIs may contribute to the increased prescribing of benzodiazepines.

Furthermore, elderly people may be more sensitive to the sedative adverse effects of TCAs, the-

refore diminishing the need for benzodiazepine drug prescribing during TCA use. In the nursing

home cohort, no difference in starting a benzodiazepine was found between users of SSRIs

compared with users of TCAs. Partly this may be due to the fact that hypnotic use in this popu-

lation is high already: in a previous study, it was found that 65% of the nursing home popula-

tion used a benzodiazepine during the study period [22]. It can be argued that starting on an

antidepressant together with a benzodiazepine may be due to anticipation of insufficient effi-

cacy of the antidepressant, possibly in view of prior use of the combination (which was un-

known to us because it was outside the study window). We checked for the impact of excluding

these ‘prophylactic starts’ and found that inclusion of these events marginally altered the

results, and it did not change our conclusion.

The prevalence of concomitant prescribing was considerable: in both ambulatory and insti-

tutionalised elderly more than 50% of TCA and SSRI users were prescribed a benzodiazepine

concomitantly. In the ambulatory cohort, concomitant benzodiazepine prescribing among SSRI

users was slightly higher when we studied prevalence (57-63% versus 53% in TCAs). The fact

that we did not find any difference between the nursing home cohort and the ambulatory

cohort suggests that the extent of the problem is the same in ambulatory and institutionalised

elderly. The duration of concomitant drug use was also relatively high: on average, concomitant

drug use lasted for greater than 67 days or more per 100 days of antidepressant drug use. The

high rate of concomitant prescribing seems of concern. In particular, the combination of TCAs

and benzodiazepines may lead to cumulation of adverse effects in the elderly, such as excess

sedation and muscle relaxation leading to an increased risk of falls [18]. In psychiatry, it is com-

mon practice to initiate antidepressant and benzodiazepine drug therapy simultaneously, with

the aim to treat non-specific symptoms of depression such as agitation and insomnia imme-

98

References


2 Flint AJ. Choosing appropriate antidepressant therapy in the elderly. A risk-benefit assessment of available

agents. Drugs Aging 1998; 13: 269-80.

3 Mamdani MM, Parikh SV, Austin PC, Upshur RE. Use of antidepressants among elderly subjects: trends and

contributing factors. Am J Psychiatry 2000; 157: 360-7.

4 Salzman C. Practical considerations for the treatment of depression in elderly and very elderly long-term care

patients. J Clin Psychiatry 1999; 60 (suppl): 30-3.

5 Pollock BG. Adverse reactions of antidepressants in elderly patients. J Clin Psychiatry 1999; 60 (suppl): 4-8.

6 Mittmann N, Herrmann N, Einarson TR et al. The efficacy, safety and tolerability of antidepressants in late life

depression: a meta-analysis. J Affect Disord 1997; 46: 191-217.

7 Williams JW, Mulrow CD, Chiquette E, Hitchcock Noël P, Aguilar C, Cornell J. A systematic review of newer

pharmacotherapies for depression in adults: evidence report summary. Ann Intern Med 2000; 132: 743-56.

8 Steffens DC, Krishnan KB, Helms MJ. Are SSRIs better than TCAs? Comparison of SSRIs and TCAs:

a meta-analysis. Depress Anxiety 1997; 6: 10-18.

9 Thapa PB, Gideon P, Cost TW, Milam AB, Ray WA. Antidepressants and risk of falls among nursing home

residents. N Engl J Med 1998; 339: 875-82.

10 Liu B, Anderson G, Mittman N, To T, Axcell T, Shear N. Use of selective serotonin-reuptake inhibitors or tricyclic

antidepressants and risk of hip fractures in elderly people. Lancet 1998; 351: 1303-7.

11 Parkes A. Starting SSRI antidepressant therapy: its effect on tricyclic antidepressant and benzodiazepine

prescribing (letter). Med J Aust 1996; 164: 509.

12 Rascati K. Drug utilization review of concomitant use of specific serotonin reuptake inhibitors or clomipramine

with anxiety/sleep medications. Clin Ther 1995; 17: 786-90.

13 Gregor K, Riley J, Downing D. Concomitant use of anxiolytics and hypnotics with selective serotonin reuptake

inhibitors. Clin Ther 1996; 18: 521-7.

14 Bingefors K, Isacson DGL. Concomitant prescribing of tranquilizers and hypnotics among patients receiving

antidepressant prescriptions. Ann Pharmacother 1998; 32: 531-5.

15 Pathiyal A, Hylan TR, Jones JK, Davtian D, Sverdlov L, Keyser M. Prescribing of selective serotonin reuptake

inhibitors, anxiolytics, and sedative-hypnotics by general practitioners in The Netherlands: a multivariate

analysis. Clin Ther 1997; 19: 798-810.

16 Joffe RT, Levitt AJ, Sokolov STH. Augmentation strategies: focus on anxiolytics. J Clin Psychiatry 1996; 57 (suppl):

25-31.

17 Holbrook AM, Crowther R, Lotter A, Cheng C, King D. Meta-analysis of benzodiazepine use in the treatment of

insomnia. CMAJ 2000; 162: 225-33.

18 Ray WA, Thapa PB, Gideon P. Benzodiazepines and the risk of falling in nursing home residents.

J Am Geriatr Soc 2000; 48: 682-5.

19 Hanlon JT, Horner RD, Schmader KE, Fillenbaum GG, Lewis IK, Wall WE, et al. Benzodiazepine use and

cognitive function among community-dwelling elderly. Clin Pharmacol Ther 1998; 64: 684-92.

20 Tobi H, Van den Berg PB, De Jong-van den Berg LTW. The InterAction database: synergy of science and practice

in pharmacy. In: Medical data analysis: first international symposium; proceedings/ISMDA (RW Brause,

E Hanisch, eds). Berlin: Springer, 2000: 206-11.

21 Lau HS, de Boer A, Beuning KS, Porsius A. Validation of pharmacy records in drug exposure assessment.




101

Limitat ions

Information for each subject in the study population was obtained from one of the two

datasets: the InterAction dataset and the nursing home dataset. Both databases are pharmacy

prescription databases and contain complete medication profiles of individual patients. We do

not know whether people actually used the drugs that were dispensed. In the nursing home,

nursing staff ensure everyone takes their medication [22]. However, in the ambulatory cohort,

people who collect their prescriptions at the pharmacy, need not necessarily use these drugs.

An overestimation of drug use due to non-adherence could be a consequence of using phar-

macy prescription data. In our study this may lead to an overestimation of drug use in the

ambulatory cohort, thus leading to a greater difference between the ambulatory and nursing

home cohort. Non-adherence may be greater with TCAs than with SSRIs (due to more severe

adverse effects [26]), leading to greater differences between TCAs and SSRIs, which would not

change our conclusions. Antidepressants may be used for other conditions than depression, for

example panic disorder (mainly SSRIs), obsessive-compulsive disorder (mainly SSRIs), and

neuropathic pain conditions (mainly TCAs). This may influence co-prescribing of other drugs.

Since we did not have information on indications for prescribing of antidepressant drugs, we

can not adjust for this possibly confounding effect. Furthermore, we do not know whether the

choice of antidepressant is influenced by the mental state of the patients. For example, TCAs

may be prescribed more frequently to agitated patients.

In summary, in this follow-up study of the elderly we could not confirm an increased risk

for concomitant prescribing of benzodiazepines among TCA users found in an earlier, cross-

sectional study [14]. In fact, we found a higher incidence of benzodizapine drug prescribing

among SSRI users compared with TCA users. In the nursing home cohort, no difference in ini-

tiating benzodiazepine drug therapy was found between users of SSRIs compared with users

of TCAs. We also found a 2-year prevalence of benzodiazepine drug prescribing during both

SSRI and TCA use of more than 50% and that the duration of concomitant drug use was rela-

tively long-term (>65 days per 100 days of antidepressant drug use) in both ambulatory and

institutionalised elderly. In view of cumulation of adverse sedative effects, and questionable

therapeutic benefits of concomitant prescribing of these two drug groups, physicians may bene-

fit from feedback on their prescribing habits.

Acknowledgement

We express our gratitude to D.A. Bloemhof, hospital pharmacist, for supplying nursing

home pharmacy data.

100

2.7 Prescr ib ing of gastroprotect ive drugsamong elder ly NSAID users in theNether lands

K.N. van Dijk 1,2, K. ter Huurne1, C.S. de Vries 3, P.B. van den Berg1,J.R.B.J. Brouwers1,2, L.T.W. de Jong-van den Berg1





Pharmacy World and Science (in press)

103

23 Van Dijk KN, De Vries CS, Van den Berg PB, Dijkema AM, Brouwers JRBJ, De Jong-van den Berg LTW.

Constipation as an adverse effect of drug use in nursing home patients: an overestimated risk.



(DDDs) for plain substances. World Health Organisation Collaborating Centre for Drug Statistics Methodology,

Oslo, 2000.

25 Dutch Health Council: Towards a more appropriate use of benzodiazepines. Publication number 1998/20.

The Hague: Health Council; 1998.

26 Anderson IM. Selective serotonin reuptake inhibitors versus tricyclic antidepressants: a meta-analysis of

efficacy and tolerability. J Affect Disord 2000; 58: 19-36.102

Int roduct ion

The use of non-selective non-steroidal anti-inflammatory drugs (NSAIDs) is associated

with a wide range of gastrointestinal (GI) toxicity, including mild dyspepsia (prevalence 20%),

development of (asymptomatic) duodenal or ventricular ulcera (prevalence of 10-20% within 3

months of NSAID-use) and serious complications such as perforation, ulceration, obstruction

and/or bleeding [1-4]. The risk for serious GI complications increases with advanced age:

NSAID users aged over 60 years are at up to 5 times greater risk of developing serious GI com-

plications (e.g. bleeding, perforation) than those not taking NSAIDs [2]. Other risk factors inclu-

de previous NSAID-related GI adverse effects, previous history of gastrointestinal events, con-

comitant corticosteroid use, concomitant use of anticoagulant drugs, chronic use of NSAIDs and

use of high dosages of NSAIDs [3-7].

To prevent the occurrence of NSAID-induced GI-toxicity, several strategies exist. Prescribing

a gastroprotective agent such as an antacid or an H2-receptor antagonist usually can prevent

mild dyspepsia. Asymptomatic ulcers can be prevented by concurrent administration of H2-

receptor antagonists or proton pump inhibitors. To prevent serious GI toxicity, the prostagland-

in analogue misoprostol [8,9], proton pump inhibitors [9,10] and H2-receptor antagonists in

high dosage [10] are reported to be effective, although a beneficial effect on perforation,

obstruction and bleeding has been shown only for misoprostol [8,11]. In a randomised control-

led trial (n=935) it was found that omeprazole 20 mg and 40 mg daily and misoprostol 800 µg

daily produced similar reductions of endoscopically diagnosed ulceration [12]. Misoprostol

generally causes more adverse effects (diarrhoea and abdominal pain) [12]. It was shown that

misoprostol (used in the combined formulation of diclofenac/misoprostol) was more cost-

effective than proton pump inhibitors [13]. Selective cyclo-oxygenase-2 (COX-2) inhibitors, such

as rofecoxib and celecoxib, are reported to result in 50% less perforation, obstruction and blee-

ding than classic NSAIDs (such as diclofenac, naproxen and ibuprofen) [14,15]. These drugs

might be alternatives to classic NSAIDs combined with gastroprotective drugs [11], however to

our knowledge cost-effectiveness studies have not been performed yet.

In most countries guidelines exist to minimise the risk of NSAID-induced GI toxicity, main-

ly recommending concurrent administration of gastroprotective agents [11]. The American

College of Rheumatology recommends that patients with a risk of developing NSAID-induced

gastropathy should receive concomitant therapy with gastroprotective agents, such as miso-

prostol [2]. In the Netherlands, similar recommendations exist [16]. To investigate whether

these recommendations are being followed in daily clinical practice, we studied the prescribing

of H2-receptor antagonists, proton pump inhibitors and misoprostol in a cohort of NSAID users

aged 65 years and over.

105

Abstract

Objective: Use of non-steroidal anti-inflammatory drugs (NSAIDs) is associated with an in-

creased risk of gastrointestinal toxicity, in particular when risk factors are present. We investi-

gated whether recommendations, suggesting concomitant therapy with gastroprotective

agents for patients at risk of developing NSAID-induced gastropathy, are being followed in

daily clinical practice.

Methods: A study was performed to investigate concomitant prescribing of gastroprotective

agents (H2-receptor antagonists, proton pump inhibitors, or misoprostol) in an ambulatory

cohort of NSAID users aged 65 years and over. The prevalence of concomitant prescribing was

studied, as well as the prophylactic prescribing of gastroprotective drugs. A stepwise logistic

regression was performed to determine predictive variables of prophylactic and concomitant

gastroprotective drug prescribing.

Results: Co-prescribing of gastroprotective drugs occurred in 1,522 (23%) (of which 944 con-

cerned prophylactic prescribing) of the NSAID users (n=6,557), with an average duration of 67

days per 100 days of NSAID use. Co-prescribing of gastroprotective drugs varied among indivi-

dual NSAIDs. Concomitant use of oral corticosteroids (ORadj 2.4; CI95 2.0-2.9), coumarins (ORadj

1.6; CI95 1.3-2.0), and low dose aspirin (ORadj 1.6; CI95 1.4-1.9) were significantly associated with

both prophylactic and concomitant prescribing of gastroprotective agents during NSAID thera-

py.

Discussion: Despite current guidelines recommending gastroprotective drug prescribing among

high risk groups, the rate of concomitant prescribing of gastroprotective agents in NSAID users

aged 65 years and over is low. Feedback to prescribers should be given to improve prescribing

practices in this high risk group.

104

tors were the drugs most frequently used among NSAID users, followed by H2-receptor anta-

gonists and misoprostol (p<0.05).

Co-prescribing of gastroprotective drugs occurred in 1,522 (23%) (of which 944 concerned

prophylactic prescribing) of the NSAID users (n=6,557), with an average duration of 67 days per

100 days of NSAID use. In table 2, the results of the analyses on concomitant drug use of gastro-

protective agents during use of NSAIDs are given for each NSAID that was used in the study

population. The NSAIDs in table 2 are given in order of the GI toxicity ranking of Henry and co-

workers [19], with ibuprofen being the NSAID with the lowest toxicity profile, and ketoprofen

the NSAID with the highest toxicity profile. Meloxicam, nabumetone and the fixed combination

of diclofenac plus misoprostol have been on the market since 1996, and therefore were not clas-

sified by Henry [19]. Prophylactic prescribing of gastroprotective drugs occurred most frequent-

ly among users of ketoprofen (26%), compared to other NSAID users. Ibuprofen users were

prescribed the fewest prophylactic GI-protective drugs (9%). In table 2 it is shown that the rela-

tive risk for prescribing gastroprotective drugs prophylactically was 3.6 times higher for keto-

profen users compared to ibuprofen users. Meloxicam users were 2.3 times more likely to recei-

ve a gastroprotective drug prophylactically. The prevalence of concomitant GI-protective drug

prescribing was highest among users of meloxicam (34%), followed by ketoprofen (31%).

Otherwise stated, meloxicam users and ketoprofen users were respectively 2.4 times and 2

times more likely to receive a gastroprotective drug concomitantly (table 2). The lowest preva-

lence of concomitant GI-protective drug prescribing was found among ibuprofen users (18%).

The duration of concomitant drug use ranged between 58 days per 100 days of indomethacin

107

Methods

Study populat ion

The study was performed with pharmacy dispensing data form the InterAction database,

which is part of a collaboration between community pharmacists in the Northern part of the

Netherlands and the University of Groningen. This database contains complete medication

profiles of 135,000 individual patients from a number of pharmacies from 1994 onwards and

has been described in detail elsewhere [17]. The study population included all patients aged 65

and over who were registered within the InterAction database and who were prescribed an

NSAID at any time during the study period of January 1998 until December 1999.

Drug ut i l i sat ion

Concomitant use of NSAIDs (Anatomical Chemical Therapeutic (ATC) [18] code M01A) and

the following gastroprotective agents was studied: misoprostol (ATC code A02BB01), H2-recep-

tor antagonists (ATC code A02BA), and proton pump inhibitors (ATC code A02BC). The 2-year

prevalence of concomitant prescribing of these gastroprotective agents during NSAID therapy

was calculated by dividing the number of NSAID users who were prescribed a gastroprotective

agent concomitantly for at least 1 day by the total number of NSAID users. Each patient was

included in these analyses only once. The average duration of concomitant drug use was cal-

culated as the number of days of concomitant drug use divided by the total number of days of

NSAID use. These analyses were performed for each NSAID individually. NSAIDs that were

prescribed to less than 50 people were excluded from the analyses. We investigated the pro-

phylactic prescribing of gastroprotective drugs separately, defined as the number of patients

who started with an NSAID and a gastroprotective agent, prescribed by the same physician, on

the same day. Relative risks for both prophylactic and concomitant prescribing of gastroprotec-

tive drugs during NSAID therapy were calculated for each NSAID individually. A stepwise logis-

tic regression was performed to determine predictive variables of prophylactic and concomitant

gastroprotective drug prescribing.

Results

17,060 patients aged 65 and over were registered in the InterAction database during 1998

and 1999. Of these, 6,557 (38.4%) used an NSAID at any time during the study period. In table

1, the numbers of users of gastroprotective medications among NSAID users and non-NSAID

users are given. Proton pump inhibitors, H2-receptor antagonists and misoprostol were used

more frequently in NSAID users compared to non-NSAID users (p<0.05). Proton pump inhibi-

106

Table 1: Use of drugs under study in the study populations

Drug use Cohort 1998-1999¶ NSAID users¥ Non-NSAID users

(n=17,060) (%) (n=6,557) (%) (n=10,503) (%)

Proton pump 2530 (14.8) 1262 (19.2) 1268 (12.1)

inhibitors

H2-antagonists 1998 (11.7) 977 (14.9) 1021 (10.3)

Misoprostol 14 (<0.1) 13 (0.2) 1 (<0.1)

¶ Cohort of patients aged > 64 yrs registered in the InterAction database at any time during 1998-1999¥ Defined as person who uses NSAID at any time during study period.

109

drug use and 82 days per 100 days of ketoprofen drugs use. Table 3 shows the results of the m

ul-

tivariate analyses. It was found that use of N

SAIDs for m

ore than 90 days, concomitant use of

oral corticosteroids, concomitant use of low

dose aspirin and concomitant use of oral antico-

agulants (coumarins) all w

ere significantly associated with both prophylactic and concom

itant

prescribing of gastroprotective agents during NSAID

therapy.

108

Table 2: Use of concomitant GI-protective drugs during NSAID use¶

NSAID Number of patients¥ Number of patients RRcrude (95% CI) Number of patients RRcrude (95% CI) Duration of(% of cohort of on prophylactic on concomitant concomitantNSAID users GI-drug use (%) GI-drug use (%) drug use (%)(n=6,557))

Ibuprofen 2556 (38.9) 229 (9) 1 (reference) 461 (18) 1 (reference) 66.3Diclofenac 3579 (54.6) 475 (13.3) 1.56 (1.32-1.84) 807 (22.5) 1.32 (1.16-1.50) 65.3Naproxen 793 (12.1) 103 (13) 1.52 (1.18-1.94) 176 (22.2) 1.30 (1.07-1.58) 68.7Indomethacin 196 (3) 33 (16.8) 2.06 (1.38-3.06) 53 (27.0) 1.68 (1.21-2.35) 57.7Piroxicam 136 (2.1) 18 (13.2) 1.55 (0.93-2.59) ns 33 (24.3) 1.46 (0.97-2.18) ns 72.3Ketoprofen 104 (1.6) 27 (26) 3.56 (2.25-5.64) 32 (30.8) 2.02 (1.32-3.10) 81.8

Diclofenac+Misoprostol 509 (7.8) 50 (9.8) 1.11 (0.80-1.53) ns 121 (23.8) 1.42 (1.13-1.78) 65Meloxicam 340 (5.2) 63 (18.5) 2.31 (1.70-3.14) 116 (34.1) 2.35 (1.84-3.01) 71.3Nabumeton 279 (4.3) 34 (12.1) 1.41 (0.96-2.07) ns 72 (25.8) 1.58 (1.19-2.11) 73.0

¶ NSAIDs are listed in order of comparative toxicity, ranging from low toxicity (ibuprofen) to relatively high toxicity (ketoprofen)(adapted from [11]), except for diclofenac+misoprostol, meloxicam, nabumeton

¥ total number of patients is more than 6,557 because patients may use several NSAIDsns non significant

Discuss ion

We found that 23% of a cohort of NSAID users aged 65 and over were prescribed a gastro-

protective agent concomitantly. Among two-third of these patients the gastroprotective drugs

were prescribed prophylactically. In view of current prescribing guidelines, these findings indi-

cate low concomitant and prophylactic prescribing of gastroprotective agents among NSAID

users aged over 64 year in daily clinical practice.

The frequency of concomitant prescribing varied between individual NSAIDs. According to

the classification of Henry and co-workers [19], ibuprofen is the NSAID with the lowest GI toxi-

city profile. We found that concomitant and prophylactic prescribing was the lowest among ibu-

profen users, which is in line with Henry’s classification. Ketoprofen (a high GI toxicity NSAID

[19]) users were almost 4 times more likely to receive a gastroprotective drug prophylactically,

compared with ibuprofen users. Also, the prevalence of concomitant gastroprotective drug use

among ketoprofen users was higher than for the other NSAIDs, with the exception of meloxi-

cam. However, percentages of concomitant (31%) and prophylactical (26%) prescribing among

ketoprofen users were still very low. The fact that nearly 75% of the ketoprofen users did not

use any gastroprotective agent raises the question whether these patients are optimally tre-

ated or whether the guidelines need adjusting. Prophylactic and concomitant use of gastropro-

tective drugs among meloxicam users was relatively high (19% and 34% respectively). An

explanation for this finding could be that meloxicam, which is reported to have milder GI side

effects due to preferential inhibition of cyclooxygenase 2 [20], is prescribed to high-risk

patients (‘channelling’). This phenomenon has recently been described in a report by Lanes and

co-workers [21].

Strengths and l imitat ions

Because we used dispensed prescribing data from pharmacies, we eliminated primary non-

compliance. Furthermore, the detail and completeness of information present in the

InterAction database enables the accurate estimation of duration of drug use and the specific

drugs dispensed, for each individual patient. A limitation of our study is that we do not know

whether patients who do not use gastroprotective drugs concomitantly suffer from more seve-

re GI toxicity, such as perforation and bleeding, than patients who are prescribed gastroprotec-

tive drugs. Because we performed a retrospective study on prospectively collected data, sever-

al biases could occur. Information bias could occur when drugs are not taken as dispensed.

Selection bias in this study is not very likely to occur, because this study is based on a popula-

tion-based cohort. Confounding by lifestyle-factors such as smoking and alcohol could have

111110

Tab

le 3

: Var

iabl

es a

ssoc

iate

d w

ith p

roph

ylac

tic a

nd c

onco

mita

nt g

astr

opro

tect

ive

drug

pre

scri

bing

Prop

hyla

ctic

gas

trop

rote

ctiv

e dr

ug u

seCo

ncom

itant

gas

trop

rote

ctiv

e dr

ug u

se

Vari

able

ORcr

ude

(CI 95

)¶OR

adju

sted

(CI 95

)¥OR

crud

e(C

I 95)

ORad

just

ed(C

I 95)

Gen

der (

mal

e vs

fem

ale)

0.66

(0.5

7-0.

77)

0.74

(0.6

3-0.

87)

0.77

(0.6

8-0.

88)

0.83

(0.73

-0.9

4)

Age

(+5

year

s)1.1

0 (1

.04-

1.15)

1.0 (0

.95-

1.05)

1.09

(1.0

5-1.1

4)1.0

3 (0

.99-

1.08)

Dur

atio

n of

NSA

ID u

se (+

30 d

ays)

1.12

(1.10

-1.13

)1.1

0 (1

.09-

1.11)

1.08

(1.0

7-1.0

9)1.0

6 (1

.05-

1.07)

Use

of o

ral c

ortic

oste

roid

s2.

93 (2

.42-

3.56

)2.

18 (1

.78-2

.68)

2.97

(2.4

9-3.

54)

2.44

(2.0

3-2.

93)

Use

of l

ow d

ose

aspi

rin

2.09

(1.78

-2.4

5)1.8

1 (1.5

2-2.

15)

1.83

(1.5

9-2.

11)

1.62

(1.4

0-1.8

8)

Use

of c

oum

arin

s2.

39 (1

.92-

2.96

)1.9

1 (1.5

1-2.4

1)1.9

4 (1

.59-

2.36

)1.6

4 (1

.33-

2.01

)

¶od

ds ra

tio w

ith c

orre

spon

ding

95%

con

fiden

ce in

terv

als,

una

djus

ted

for v

aria

bles

list

ed in

col

umn

1¥

odds

ratio

with

cor

resp

ondi

ng 9

5% c

onfid

ence

inte

rval

s, a

djus

ted

for v

aria

bles

list

ed in

col

umn

1

References

1 Griffin MR, Piper JM, Daugherty JR, Snowdon M, Ray WA. Nonsteroidal anti-inflammatory drug use and

increased risk for peptic ulcer disease in elderly persons. Ann Intern Med 1991; 114: 257-63.

2 Plosker GL, Lamb HM. Diclofenac/misoprostol: pharmacoeconomic implications for therapy.

Pharmacoeconomics 1999; 16: 85-98.

3 Fries JF. NSAID gastropathy: the second most deadly rheumatic disease? Epidemiology and risk appraisal.

J Rheumatol 1991; suppl.28: 6-10.

4 Wolfe MM, Lichenstein DR, Singh G. Gastrointestinal toxicity of nonsteroidal anti-inflammatory drugs.

N Engl J Med 1999; 24: 1888-99.

5 Gabriel SE, Jaakkimainen L, Bombardier C. Risk for serious gastrointestinal complications related to use of

non-steroidal anti-inflammatory drugs. A meta-analysis. Ann Intern Med 1991; 115: 787-96.

6 Griffin MR. Epidemiology of nonsteroidal anti-inflammatory drug-associated gastrointestinal injury.

Am J Med 1998; 104: 23S-29S.

7 Shorr RI, Ray WA, Daugherty JR, Griffin MR. Concurrent use of nonsteroidal anti-inflammatory drugs and oral

anticoagulants places elderly persons at high risk for hemorrhagic peptic ulcer disease. Arch Intern Med 1993;

153: 1665-70.

8 Silverstein FE, Graham DY, Senior JR, Davies HW, Struthers BJ, Bittman RM, et al. Misoprostol reduces serious

gastrointestinal complications in patients with rheumatoid arthritis receiving nonsteroidal anti-inflammatory

drugs: a randomized double-blind, placebo-controlled trial. Ann Intern Med 1995; 123: 241-249.

9 Rostom A, Wells G, Tugwell P, Welch V, Dube C, McGowan J. The prevention of chronic NSAID induced upper

gastrointestinal toxicity: a Cochrane collaboration meta-analysis of randomized controlled trials. J Rheumatol

2000; 27: 2203-14.

10 Agrawal NM, Aziz K. Prevention of gastrointestinal complications associated with nonsteroidal antiinflammatory

drugs. J Rheumatol 1998; 25: S51: 17-20.

11 Boers M. NSAIDs and selective COX-2 inhibitors: competition between gastroprotection and cardioprotection

(commentary). Lancet 2001; 357: 1222-3.

12 Hawkey CJ, Karrasch JA, Szczepanski L, Walker DG, Barkun A, Swannell AJ et al. Omeprazole compared with

misoprostol for ulcers associated with nonsteroidal antiinflammatory drugs. N Engl J Med 1998; 338: 727-34.

13 Brouwers JRBJ, Van Roon EN, Postma M, Al M. Altijd gastroprotectie? Kosten-effectiviteit van NSAIDs

(in Dutch). Pharm Weekbl 2000; 135: 172-5.

14 Silverstein FE, Faich G, Goldstein JL, Shapiro D, Burgos-Vargas R, Davis B, et al. Gastrointestinal toxicity with

celecoxib vs nonsteroidal anti-inflammatory drugs for osteoarthritis and rheumatoid arthritis: the CLASS study:

a randomized controlled trial. Celecoxib Long-term Arthritis Safety Study. JAMA 2000; 284: 1247-55.

15 Bombardier C, Laine L, Reicin A, Shapiro D, Burgos-Vargas R, Davis B, et al. Comparison of upper gastro-

intestinal toxicity of rofecoxib and naproxen in patients with rheumatoid arthritis. VIGOR Study Group.

N Engl J Med 2000; 343: 1520-8.

16 Van der Kuy, A (Ed). Farmacotherapeutisch Kompas (in Dutch). Amstelveen, The Netherlands: CMPC

Ziekenfondsraad, 2001: 940.

17 Tobi H, Van den Berg PB, De Jong-van den Berg LTW. The InterAction database: synergy of science and practice

in pharmacy. In: Brause RW, Hanisch E (ed) Medical data analysis: first international symposium;

proceedings/ISMDA. Berlin: Springer, 2000: 206-11.


(DDDs) for plain substances. Oslo: World Health Organisation Collaborating Centre for Drug Statistics

Methodology, 2000.

19 Henry D, Lim LLY, Garcia Rodriguez LA, Perez Gutthann S, Carson JL, Griffin M, et al. Variability in risk of

113

occurred in this study; this information is not present in the prescription database and is likely

to affect concomitant prescribing. In this study, we defined misoprostol (combinations), proton

pump inhibitors and H2-receptor antagonists as gastroprotective agents. Of these misoprostol

is the only drug with proven effectiveness on perforation, obstruction and bleeding [8]. H2-

receptor antagonists have been reported to be effective in double dosage (2 times the licensed

daily dosage for ulcer healing). Limiting our analyses to misoprostol would have shown lower

percentages of concomitant and prophylactic gastroprotective drug use during NSAID therapy.

We found that H2-receptor antagonists were prescribed in normal dosages (data not shown).

From this study, in daily clinical practice the prescription of double dosage of H2-receptor anta-

gonists does not seem to take place.

Other studies

In a Canadian cohort of NSAID users > 65 years (n= 61,601), 26% were prescribed an anti-

ulcer medication, compared with 11% of the non-NSAID users (n=168,944) [22]. Schubert and

colleagues [23] found that 8.9%-12.2% of patients treated with NSAIDs (n=460) were prescri-

bed an antacid or a drug for the treatment of peptic ulcer. LeLorier [24] found that in a cohort

of elderly aged over 65 (n=773,654) ibuprofen users were prescribed less systemic antiulcer

agents than other NSAIDs, probably due to a channelling of low risk patients towards this drug.

Pertusi and colleagues [25] held a survey among geriatric practitioners (n=229) to investigate

the extent of gastroprotective prescribing among elderly NSAID users and the influence of risk

factors (age, previous peptic ulcer, previous gastro-intestinal bleeding, and heart disease) on

their choices. It was found that 64% of the respondents would not prescribe gastroprotective

agents to elderly patients who use NSAIDs. This was different for nursing home residents using

NSAIDs: only 32% would not prescribe gastroprotective agents for this population.

Furthermore, age and heart disease were risk factors that physicians did not take into account

when choosing gastroprotective agents. Further research is needed into the reasons for not

prescribing gastroprotective agents for high-risk patients. The results of our study are in line

with these findings and raise the question why gastroprotective agents are not prescribed more

frequently among high-risk patients.

Conclus ion

Our study shows that gastroprotective drugs are prescribed to only a minority (23%) of

NSAID users aged 64 and over. This might suggest that physicians do not consider age as a risk

factor for gastrointestinal adverse effects of NSAIDs. In view of the known gastrointestinal side

effects of non-selective NSAIDs, we think it is necessary to give feedback to physicians with the

aim to increase the rate of concomitant gastroprotective drug use in elderly NSAID users.

112

Chapter 3

Risk assessment studiesin the e lder ly

115

gastrointestinal complications with individual non-steroidal anti-inflammatory drugs: results of a collaborative

meta-analysis. BMJ 1996; 312: 1563-6.

20 Hawkey C, Kahan A, Steinbrück K, Alegre C, Baumelou E, Bégaud B, et al. Gastrointestinal tolerability of

meloxicam compared to diclofenac in osteoarthritis patients. Br J Rheumatol 1998; 37: 937-45.

21 Lanes SF, Garcia Rodriguez LA, Hwang E. Baseline risk of gastrointestinal disorders among new users of

meloxicam, ibuprofen, diclofenac, naproxen and indomethacin. Pharmacoepidemiol Drug Saf 2000; 9: 113-7.

22 Hogan DB, Campbell NRC, Crutcher R, Jennett P, MacLeod N. Prescription of nonsteroidal anti-inflammatory

drugs for elderly in Alberta. CMAJ 1994; 151: 315-22.

23 Schubert I, Ihle P, Köster I, Ferber von L. Markers to analyse the prescribing of non-steroidal anti-inflammatory

drugs in ambulatory care. Eur J Clin Pharmacol 1999; 55: 479-86.

24 LeLorier J. Patterns of prescription of nonsteroidal antiinflammatory drugs and gastroprotective agents.

J Rheumatol 1995; (suppl 43): 22: 26-7.

25 Pertusi RM, Godwin KS, House KJ, Knebl JA, Alexander JH, Rubin BR, et al. Gastropathy induced by nonsteroidal

anti-inflammatory drugs: pescribing patterns among geriatric practitioners. J Am Osteopath Assoc 1999; 99: 305-10.

114

3.1 Const ipat ion as an adverse effect ofdrug use in nurs ing home pat ients : an overest imated r isk

K.N. van Dijk 1,2, C.S. de Vries3, P.B. van den Berg1, A.M. Dijkema1,J.R.B.J. Brouwers1,2, L.T.W. de Jong-van den Berg1





British Journal of Clinical Pharmacology 1998; 46: 255-61

117

Out l ine

This chapter focuses on the clinical implications and potential risks associated with drug

use, with reference to two particular issues identified in chapter 2. Section 3.1 presents a study

among a cohort of nursing home residents, in which the association between drug use and con-

stipation was investigated. As was shown in section 2.3, laxative use was high in nursing home

patients and we investigated whether this might be the consequence of use of drugs that have

been associated with an increased frequency of constipation, such as anticholinergic drugs. In

section 3.2, a study is presented in which the clinical effect of a drug-drug interaction was

investigated. The drug-drug interaction studied was between NSAIDs and acenocoumarol, an

interaction that was frequently encountered in nursing home patients as was shown in section

2.4. This study was performed in a cohort of elderly outpatients, using data from the Groningen

Outpatient Thrombosis Service. Genotyping of cytochrome P450 2C9 was performed to deter-

mine whether genotype was a predictive variable for the occurrence of this drug interaction.

116

Int roduct ion

Many studies have reported laxative use in the elderly to be disturbingly high and it has

been suggested that improved pharmacotherapy might reduce the prevalence of constipation

[1]. The prevalence of constipation in ambulatory elderly over age 65 years varies from 16% to

41% [2,3]. Chronic constipation may lead to complications such as faecal impaction, stercoral

ulceration, bowel obstruction, sigmoid volvulus and even syncope [2]. The prevalence of con-

stipation among institutionalised elderly has been reported to be even higher [4,5]. In a popu-

lation of 784 nursing home patients in the Netherlands, 53% were prescribed one or more

laxatives daily [5]. Long-term use of stimulant laxatives may lead to abdominal cramps, fluid

and electrolyte disturbances, malabsorption and cathartic colon [6]. In view of these unwanted

effects and to improve the quality of life of the elderly it is worthwhile to study whether laxa-

tive use can be reduced in this population. Polypharmacy is an important risk factor for consti-

pation, especially in nursing homes where levels of medication use are high [7]. Drugs that are

commonly associated with constipation are opioids, iron salts, calcium channel blockers and

drugs with anticholinergic/antimuscarinic effects [5]. The last group is also responsible for

other potentially dangerous adverse effects in the elderly such as urinary retention, memory

problems, delirium and acute glaucoma [8,9]. In pharmacoepidemiological studies, laxative

administration is used as a proxy for constipation because laxative use has been shown to cor-

relate well with constipation [1]. The association between laxative use and other drug use has

been assessed in several studies [1–3,10,11]. In most of these studies, only some subgroups of

drugs were considered and the majority of these studies used cross-sectional study designs. To

investigate whether the suggested causal association between laxative use and co-medication

could be confirmed in a cohort of nursing home residents in the Netherlands we carried out a

prospective study using prescription sequence analysis. If any such causal association exists,

recommendations can subsequently be given for alternative pharmacotherapy in order to redu-

ce laxative use in the elderly.

Methods

Design

A prospective cohort study was performed to estimate the incidence relative risk of consti-

pation as an adverse effect of drug use. The study was conducted with prescription sequence

analysis of computerised pharmacy records. Prescription sequence analysis is a method to

determine side effects of drugs through individual medication histories. It is based on the

119

Abstract

Objective: To investigate whether results from case control and cross sectional studies, which

suggest an association between laxative use and other drug use, could be confirmed in a cohort

study of nursing home patients.

Methods: A prospective cohort study of 2,355 nursing home patients aged 65 years and over

was performed to estimate the incidence relative risk of constipation associated with drug use.

The study was conducted with prescription sequence analysis of each resident’s detailed phar-

macy records and data on morbidity and mobility.

Results: Use of drugs, which according to the summaries of product characteristics and the lite-

rature on adverse drug effects have moderately to strongly constipating properties, was asso-

ciated with a relative risk of 1.59 (CI95 1.24–2.04) for the occurrence of constipation during

exposure time. Use of drugs with mildly to moderately constipating effects was not associated

with laxative use (RR 1.13; CI95 0.93–1.38). Stratification on the level of age, gender, type of nur-

sing (psychogeriatric or somatic), morbidity, number of medications taken and mobility showed

no confounding effects of these variables on outcome measurements. These variables all acted

as effect modifiers. Effect of age and number of medications taken on the relative risk was non-

linear.

Conclusions: Although an association between drugs that exhibit moderately to strongly con-

stipating effects and occurrence of constipation was found, the risk was not as high as seen in

previous studies. The high prevalence of constipation in nursing home patients is only partly

due to adverse drug effects.

118

Exposure def in i t ion

Drugs were classified into three categories: category 2: drugs that exhibit moderately to

strongly constipating effects (see Appendix A), category 1: drugs that exhibit mildly to modera-

tely constipating effects (see Appendix B) and a reference category which contained all other

drugs. For each drug the summary of product characteristics (SPC), edited and approved by the

Dutch Medicines Evaluation Board [15], provided the main source for the classification of the

constipating effects of the drugs used by the study population, together with specific informa-

tion on adverse drug effects from the literature [16,17]. Each resident’s exposure time was defi-

ned as the duration of drug use from category 1 or 2, respectively. To control for residual effects

we performed both a study in which we defined exposure time as the duration of drug use plus

the first 14 days after every exposure period and a study in which we excluded the first 14 days

after every exposure period from both exposure time and nonexposure time. To investigate if

any differences in constipating properties exist between certain of the subgroups of drugs from

category 2, we performed subgroup analyses on pharmacological subgroups (see table 4). Non-

exposure time was defined as the remainder of the period of stay during the study period.

Exposure days to category 2 drugs, to category 1 drugs and nonexposure days were aggregated

over the study population.

Case def in i t ion

The occurrence of constipation was identified by the start of a laxative, which is considered

a proxy drug. When the start of a laxative coincided with the start of a drug from category 2 or

1, the start was considered as a prophylactic start; these starts were not considered as cases.

When the start of a laxative coincided with the date of admission to the nursing home or with

the first day of the study period, the start was not considered as a case either. Patients were

considered to be ‘at risk’ for constipation during the period of stay in which they did not use a

laxative.

Analys is

Incidence rates during exposure and nonexposure time, respectively, were calculated by

dividing the number of starts of laxative use by the total number of person-days at risk both

during exposure time and during nonexposure time at risk. The incidence relative risk is deter-

mined as Iexp /Inonexp. Mantel-Haenszel relative risks were calculated to control for potential

confounding effects of age, gender, morbidity (Parkinson’s disease, diabetes mellitus, depres-

sion and dementia), type of nursing (psychogeriatric or somatic), number of medications taken

and mobility. All statistical analyses were performed in SPSS for Windows [18]. Incidence rela-

tive risks were calculated with corresponding 95% confidence intervals (CI95).

121

observation that a side effect of drug A is followed by the prescription of drug B (a ‘proxy’ drug)

if drug B is used to counteract the side effect caused by drug A [12]. In this study, laxative drugs

(all drugs in the Anatomical Therapeutic Chemical (ATC) classification A06 and A02AA02 [13])

were considered proxy drugs for constipation.

Sett ing

The study was undertaken in six nursing homes for long-term care with a 1030-bed capaci-

ty in the northern part of the Netherlands. In these nursing homes medical care is provided by

nursing home physicians, who give medical care on a daily basis. Specialists’ medical input is

obtained on demand. A distinction is made between care for psychogeriatric residents and care

for somatic residents. Nursing-, physician- and pharmacist care is comparable between the

nursing homes. In each nursing home nursing staff defined constipation as not having defeca-

ted for more than 3–5 days. Fluid- and fibre intake was comparable between the nursing

homes.

Data col lect ion

For each resident, pharmacy records of a two-year period and individual morbidity and

mobility data were collected. Pharmacy data included the generic name, strength, dosage, the

frequency of use and the route of administration of the drugs, the prescription length (in days)

and the following patient characteristics: age, gender, date of admission and date of discharge.

Drugs were classified according to the ATC classification system [13]. Dermatological prepara-

tions were excluded from the analyses. Pharmacy records were linked with a national infor-

mation system on nursing homes (SIVIS) [14], to collect the following data: type of nursing

(psychogeriatric or somatic), morbidity and mobility.

Study populat ion

All nursing home residents from six nursing homes were initially included in the cohort.

The study population consisted of 2,772 residents aged 65 years and over who were present at

any time during the two-year study period from 1 October 1993–1 October 1995. We excluded

patients who could not be linked to data from the SIVIS-system (14.2%), and subsequently

patients whose period of residence could not be defined as a result of missing data (1%). This

resulted in a final study population of 2,355 patients. Of these patients 65% were newly admit-

ted during the study period.

120

Laxat ive use

Fifty-six percent of the study population used a laxative at some time during stay at the

nursing home. An overview of the laxatives used by the study population is given in table 2.

Seventy-four percent of the residents with Parkinson’s disease used a laxative. At the entry of

the study period, 416 (18%) of the patients used a laxative. After the study entry date 1109

(47%) patients started a laxative for one or more periods. Of these patients 233 (21%) used a

laxative for a period of less than 30 days, and 876 (79%) used a laxative for a period of 30 days

or more. The average duration of laxative use was 154 days (SD 192). On average, people who

were on a laxative drug used it for more than 77% of their nursing home stay. Relatively high

dosages of laxatives were used.

Inc idence rate rat ios

The results from the cohort study are presented in table 3. Use of drugs from category 2

(moderately to strongly constipating drugs) was associated with a relative risk of 1.59 (CI95

1.24–2.04) for the occurrence of constipation and the incidence relative risk of exposure to cate-

gory 1 drugs (mildly to moderately constipating drugs) was 1.13 (CI95 0.93–1.38) compared with

nonexposure. To control for residual effects we performed both a study in which we defined

exposure time as the duration of drug a use plus the first 14 days after every exposure period

and astudy in which we excluded the first 14 days from both exposure time and nonexposure

time. When exposure time was defined as the duration of category 2 drug use plus the first 14

days after this period, the incidence relative risk was slightly higher (RR 1.69; CI95 1.33–2.15),

indicating a carry-over effect from category 2 drug exposure in the nonexposure period. To

exclude this carry-over effect, we deleted the first 14 days after exposure time from both expos-

123

Results


The mean age of the study population was 82 years (SD 7.3). The average residence time

during the study period was 257 days (SD 260). The average number of different medicines

(based on ATC-codes) per person was 8.9 during residence in the nursing home (SD 4.9; der-

matological preparations excluded). The average number of different medicines per patient per

day was 4.9. Most drugs were used for more than 50% of the duration of stay in the nursing

home. In table 1, characteristics of the study population are given. Of the mobile residents 35%

were diagnosed with dementia, while 22% of the immobile residents were diagnosed with

dementia. Forty-six percent of the study population used a drug from category 2; 57% of the

study population used a drug from category 1.

122

Table 1: Characteristics of the study population (n=2355)

Variable Number of residents (percentage of total)

Age (years)

65-74 415 (18%)

75-84 1012 (43%)

≥ 85 928 (39%)

Gender

Male 689 (29%)

Female 1666 (71%)

Type of nursing

Psychogeriatric 700 (30%)

Somatic 1609 (68%)

Not known 46 (2%)

Morbidity

Parkinson's disease 151 (6%)

Diabetes mellitus 176 (7%)

Depression 40 (2%)

Dementia 689 (29%)

Number of different medicines

0-5 626 (27%)

6-10 969 (41%)

> 10 760 (32%)

Mobility

Mobile 1370 (58%)

Immobile 985 (42%)

Table 2: Laxatives used by the study population (n=2355)

Laxative Number of of patients¶ (%)

Lactitol 871 (37%)

Lactulose 346 (15%)

Bisacodyl 336 (14%)

Magnesium oxide 140 (6%)

Docusate sodium 91 (4%)

Triticum 90 (4%)

Ispaghula (psylla seeds) 70 (3%)

¶ Note: patients may use more than one laxative

125

ure and nonexposure time, which resulted in an incidence relative risk of 1.60 (CI95 1.25–2.06).

Results of the subgroup analyses are given in table 4. Point estimates varied from 1.01 (opiates)

to 1.92 (verapamil) but the differences were not all statistically significant. Ninety-six percent

of the people who received opiates received a laxative drug prior to the initiation of opiate use.

Statistical analysis of possible confounding effects of the variables given in table 1 showed no

confounding effects from these variables as shown in table 5. Gender, morbidity and mobility

acted as effect modifiers. There was a non-linear association with age and with the number of

medications taken. Residents with depression and residents with diabetes mellitus were more

at risk for the occurrence of constipation as an adverse drug effect while residents with

Parkinson’s disease showed a markedly lower risk. Residents who were relatively mobile

showed a higher risk for the occurrence of drug-induced constipation.

124

Table 3: Relative risks for the occurrence of constipation with different drug categories

Drug category Events* Time at risk (days) Relative risk (RR) CI95

2: moderately to strongly constipating 84 30931 1.59 1.24-2.04

1: mildly to moderately constipating 179 92339 1.13 0.93-1.38

Reference drug category 236 137835 1.00

* Events: number of starts of a laxative. This was considered a marker for constipation

Table 4: Subgroup analyses of drug groups from category 2

Drugs under study Events* Time at risk (days) Relative risk (RR) CI95

Opiates 5 2880 1.01 0.42-2.46

Morphine, nicomorphine, pethidine,

dextropropoxyphene

Calcium channel blockers 10 3042 1.92 1.02-3.62

Verapamil

Calcium salts and ferrous salts 54 18947 1.67 1.24-2.24

Anticholinergic agents

Atropine, biperiden, orphenadrine, 10 5621 1.04 0.55-1.96

oxybutynine, oxyphencyclimine,

thiazinamium, trihexyphenidyl

Drugs with anticholinergic side effects 58 22244 1.52 1.14-2.03

Amitryptyline, disopyramide,

chlorpromazine, chlorprotixene,

clozapine, clomipramine,

doxepine, flavoxate, imipramine,

maprotiline, nortriptyline, thioridazine

Reference drug category 236 137835 1.00

* Events: number of starts of a laxative. This was considered a marker for constipation

Table 5: Relative risks for the occurrence of constipation associated with exposure to category 2 drugs, stratified for

age, gender, type of nursing, morbidity, number of medications and mobility

Diseased (n) Time at risk (days) Relative risk RR Mantel-Haenzsel

Variable exposed unexposed exposed unexposed (CI95) (CI95)

Age (years)

65-74 17 46 5089 23300 1.69 (0.97-2.95)

75-84 38 95 13820 49055 1.42 (0.98-2.07)

≥ 85 29 95 12022 65480 1.66 (1.10-2.52)

overall 84 236 30931 137835 1.59 (1.24-2.04) 1.55 (1.21-1.99)

Gender

Male 31 69 7428 37380 2.26 (1.48-3.45)

Female 53 167 23503 100455 1.36 (1.00-1.85)

overall 84 236 30931 137835 1.59 (1.24-2.04) 1.60 (1.25-2.05)

Type of nursing

Psychogeriatric 17 66 8614 60166 1.79 (1.05-3.06)

Somatic 64 163 20438 69275 1.33 (1.00-1.78)

Not known 3 7 1852 8394 1.94 (0.50-7.51)

overall 84 236 30931 137835 1.59 (1.24-2.04) 1.47 (1.15-1.89)

Morbidity

Parkinson's disease 5 20 2486 5976 0.60 (0.23-1.60)

Diabetes mellitus 7 12 2276 10790 2.77 (1.09-7.02)

Depression 3 5 355 3077 5.20 (1.24-21.76)

Dementia 17 66 11410 61387 1.39 (0.81-2.36)

overall 32 103 16527 81240 1.53 (1.03-2.27) 1.40 (0.95-2.07)

Number of medications

0-5 9 54 4992 48008 1.60 (0.79-3.25)

6-10 41 101 12721 51 1 31 1.63 (1.14-2.35)

>10 34 81 13218 38696 1.23 (0.82-1.83)

overall 84 236 30931 137835 1.59 (1.24-2.04) 1.45 (1.13-1.86)

Mobility

Mobile 53 137 18272 89477 1.89 (1.38-2.60)

Immobile 31 99 12659 48358 1.20 (0.80-1.79)

overall 84 236 30931 137835 1.59 (1.24-2.04) 1.57 (1.22-2.00)

remained unclear. In a cross-sectional study Monane [1] found a strong association between

laxative use and the use of highly anticholinergic antidepressants (OR 3.12; CI95 1.21–8.03) in

nursing home patients. Also in a cross-sectional study Harari [11] demonstrated that the use of

iron supplements and calcium channel blockers was significantly associated with laxative use

(OR 2.2 and OR 1.9, respectively) in elderly people residing in a long-term care setting. To our

knowledge, our study is the first that uses a cohort design to determine the association

between medication use and laxative administration in a nursing home population. In a cohort

design, laxative use as a result of medication use (sequential use) can be properly assessed

with prescription sequence analysis. With cross-sectional methods, temporal sequences of pre-

scribing are more difficult to assess [12].

Poss ib le b ias

Selection bias. We excluded all nursing home residents for whom data were incomplete or

invalid. Since they represented a small proportion of the population, this is not likely to have

influenced the results. Information for each resident was obtained from the same data set.

Therefore it is unlikely that selection bias played a role in this study.

Information bias. Information bias might occur when a laxative is prescribed for an indica-

tion other than constipation. This could lead to a bias away from the null in both the exposed

and nonexposed group. As the only other indication for the prescription of lactulose is hepatic

(pre) coma, a very rare disease, it is not likely that it leads to differential misclassification. Also

a laxative could be withheld from a patient suffering from constipation. This could lead to a bias

towards the null. This kind of bias is not likely to occur often because residents are frequently

monitored by nurses or carers, so constipation will be noticed at an early stage. However, this

kind of bias could be relevant when the problem of constipation is considered (more constipa-

ted residents), but probably is not relevant when drug-induced constipation is considered (bias

will occur in both exposed and nonexposed groups). Medication use of the individual nursing

home resident is recorded centrally in one of the three computerised hospital pharmacies

involved in our study. Dispensing of drugs takes place when the registration of medication is

complete. Therefore information bias is not likely to occur.

Confounding bias. We tried to control for possible confounding patient characteristics such

as age, gender, type of nursing, number of medications taken, morbidity and mobility. None of

these variables was considered a confounder although some of the variables were considered

effect modifiers (see below). No marked differences were seen in overall fluid and fibre intake

among the different nursing homes. Because we could not collect data on fluid and fibre inta-

ke at individual patient level, this may still confound our results. The influence of fluid and fibre

127

Discuss ion

Our study confirms earlier findings of a risk of constipation as a consequence of drug use.

However, from this cohort study the relative risk appears to be lower than has been suggested

in previous case control and cross-sectional studies. In 2,355 nursing home patients, the use of

drugs that exhibit moderately to strongly constipating effects was slightly but significantly

associated with the start of a laxative (RR 1.59; CI95 1.24–2.04). When the first 14 days after

every exposure period were added to the exposure time, the relative risk was slightly higher

(RR 1.69; CI95 1.33–2.15), which indicates residual effects of these drugs depending on the eli-

mination half-life. The results show that drugs, which according to the summaries of product

characteristics and to the literature on adverse drug effects exhibit a moderately to strongly

constipating effect, in practice are only marginally associated with the occurrence of constipa-

tion. However, the fact that drugs from category 2 are used by nearly half of the study popula-

tion at least once during the study period suggests that this side effect could be clinically rele-

vant in daily practice because it concerns many residents. Drugs that have been reported to

have mild to moderate constipating effects were not associated significantly with constipation

(RR 1.13; CI95 0.93–1.38). This means that although constipation is mentioned as a possible side

effect in the summaries of product characteristics and in the literature, the high prevalence of

constipation is probably not due to use of drugs from this category. Subgroup analyses demon-

strated that the use of the calcium channel blocker verapamil and the use of calcium- and fer-

rous salts, especially, were associated with a high risk for the occurrence of constipation. The

fact that laxatives are given prophylactically with this drug category explains why use of opia-

tes was not associated with a higher risk. Several reports support our results. Mikus and colle-

agues recently showed that the constipating effect of codeine is only seen in extensive meta-

bolizers (CYP2D6 phenotype) [20]. In a literature search (1965–94) concerning adverse events

associated with antidepressant drugs, constipation did not belong to one of the 27 most fre-

quently reported adverse events [21]. In a community based study no significant association

was found between antidepressant drug use and use of laxatives [22].

Previous studies

In the study of Stewart [2], a positive correlation was demonstrated between self-reported

constipation and the total number of drugs used in an ambulatory elderly population, but no

specific drug groups were correlated with constipation. Talley [3] demonstrated that use of non-

steroidal anti-inflammatory drugs was a significant risk factor in elderly subjects with both

functional constipation and outlet delay. However, whether this was a causal association

126

use of newer antidepressants (such as selective serotonin reuptake inhibitors) and antipsy-

chotics with minor or no anticholinergic activity could be considered as alternatives to antide-

pressants and antipsychotics with anticholinergic side effects.

In conclusion, this study demonstrates that medication, which according to their SPCs and

the literature on adverse drug effects exhibit moderately to strongly constipating effects, is

associated with the occurrence of constipation in a cohort of nursing home patients. However,

the risk is not as high as previous studies suggest. Drugs that were classified as mildly to mode-

rately constipating showed no increased risk for the occurrence of constipation. The high pre-

valence of laxative use in nursing home patients is only partly due to adverse drug effects. To

minimize the risk for constipation, alternative pharmacotherapy could be considered for certain

subgroups of drugs.

Acknowledgements

We express our gratitude to D.A. Bloemhof, hospital pharmacist, for supplying pharmacy

data; SIG Informatics on Health and Welfare, Utrecht, for supplying patient morbidity and

mobility data; and nursing, medical and pharmacy staff from all participating nursing homes

for their co-operation.

129

intake on constipation has been assessed in only few studies. Although dietary fibre is often

diminished in the elderly, no clear association has been made with true clinical constipation [7].

There are no data on dehydration as a risk factor for constipation in the elderly although the

beneficial effect of fluid intake has been proven in young male volunteers [7]. A community

based cohort study by Towers showed that constipation was related to caloric intake rather

than fibre consumption or fluid intake [19]. In our study a resident may develop constipation as

a consequence of low fluid and fibre intake. When this patient is using drugs from category 1

or 2, this patient would be wrongly considered as a ‘case’. This might lead to a overestimation

of the relative risk but it does not change our conclusion. Other possible confounders such as

age and co-morbidity did not play a role in our study. Obviously, we can never rule out confou-

ding effects of factors that we are not aware.

Ef fect modif iers

Several other factors have been reported to be associated with constipation in the elderly.

Bowel frequency is decreased with certain neurological and endocrine disorders such as

Parkinson’s disease and diabetes mellitus. Older people and women are reported to be more at

risk for constipation. The type of nursing, psychogeriatric or somatic, can influence outcome. In

several studies, inactivity and immobility have been identified as risk factors for constipation.

Although data in the elderly are scarce, Towers [19] showed that constipated elderly tend to

report less regular activity and exercise. Therefore we stratified for the variables age, gender,

type of nursing, morbidity, number of medications taken and mobility to control for possible

confounding effects. Gender, morbidity and mobility acted as effect modifiers. The non-linear

association with age and number of medications taken suggests a combined effect of effect

modification and confounding by these variables. Men were at a higher risk for the occurrence

of constipation during category 2 drug exposure in comparison with women. Residents suffe-

ring from Parkinson’s disease showed a markedly lower relative risk, probably because these

residents are already constipated. Patients with depression and diabetes mellitus showed a

higher risk for the occurrence of constipation as an adverse drug effect. Finally, residents who

were relatively mobile were more at risk for the occurrence of constipation during category 2

exposure. When pharmacotherapy is needed, the possible constipating effects of category 2

drugs should be taken into account with special reference to the patients’ gender, co-morbidi-

ty, number of medications taken and mobility. In these risk groups in particular the use of cate-

gory 2 drugs is associated with the occurrence of constipation. Furthermore, alternative phar-

macotherapy could be considered with certain subgroups of category 2 drugs. For example, the

necessity of verapamil, calcium salts and ferrous salts should be (re-)evaluated carefully. The

128

131

References

1 Monane M, Avorn J, Beers MH, Everitt DE. Anticholinergic drug use and bowel function in nursing home

patients. Arch Intern Med 1993; 153: 633–8.

2 Stewart RB, Moore MT, Marks RG, Hale WE. Correlates of constipation in an ambulatory elderly population.

Am J Gastroenterol 1992; 87: 859–64.

3 Talley NJ, Fleming KC, Evans JM, O’Keefe EA, Weaver AL, Zinsmeister AR, et al. Constipation in an elderly

community: a study of prevalence and potential risk factors. Am J Gastroenterol 1996; 1: 19–25.

4 Harari D, Gurwitz JH, Avorn J, Choodnovskiy I, Minaker KL. Constipation: assessment and management in an

institutionalized elderly population. J Am Geriatr Soc 1994; 42: 947–52.

5 Brouwers JRBJ, Tytgat G. Laxatives for elderly people with constipation. Formulary criteria, precautions and

complications (in Dutch). Pharm Weekbl 1993; 128: 1483–87.

6 Wald A. Constipation in elderly patients. Pathogenesis management. Drugs Aging 1993; 3: 220–31.

7 Harari D, Gurwitz JH, Minaker KL. Constipation in the elderly. J Am Geriatr Soc 1993; 41: 1130–40.

8 Feinberg M. The problems of anticholinergic adverse effects in older patients. Drugs Aging 1993; 3: 335–48.

9 Riedel WJ, Van Praag HM. Avoiding and managing anticholinergic effects of antidepressants.

CNS Drugs 1995; 3: 245–59.

10 Jones RH, Tait CL. Gastrointestinal side-effects of NSAIDs in the community. Br J Clin Pract 1995; 49: 67–70.

11 Harari D, Gurwitz JH, Avorn J, Choodnovskiy I, Minaker KL. Correlates of regular laxative use by frail elderly

persons. Am J Med 1995; 99: 513–8.

12 Petri JL, De Vet HCW, Naus J, Urquhart J. Prescription sequence analysis: a new and fast method for assessing

certain adverse reactions of prescription drugs in large populations. Statis Med 1988; 7: 1171–5.

13 Anonymous. Anatomical Therapeutic Chemical (ATC) classification index. Oslo: WHO collaborating centre for

drug statistics methodology, 1994.

14 SIG. Informatics in Health and Welfare, Utrecht, the Netherlands.

15 Dutch Medicines Evaluation Board, Rijswijk, the Netherlands.

16 Informatorium Medicamentorum. Royal Dutch Society for the Advancement of Pharmacy, The Hague,

the Netherlands 1994.

17 Dukes MNG (ed). Meyler’s side effects of drugs, twelfth edition, Amsterdam: Elsevier Science Publishers 1992.

18 Norusis MJ SPSS. 6.1. Guide to data analysis. New Jersey: Prentice-Hall Inc. Englewood Cliffs.

19 Towers AL, Burgio KL, Locher JL, Merkel IS, Safaeian M, Wald A. Constipation in the elderly: influence of dietary,

psychological, and physiological factors. J Am Geriatr Soc 1994; 42: 701–6.

20 Mikus G, Trausch B, Rodewald C, Hofmann U, Richter K, Gramatte T, et al. Effect of codeine on gastrointestinal

motility in relation to CYP2D6 phenotype. Clin Pharmacol Ther 1997; 61: 459–66.

21 Egberts ACG, Koning de GHP, Bakker A, Leufkens HGM. Adverse events associated with antidepressant drugs

published in the medical literature. In Egberts ACG Pharmacoepidemiologic approaches to the evaluation of

antidepressant drugs [Thesis]. University of Utrecht, 1997.

22 Egberts ACG, Leufkens HGM, Launer LJ, Grobbee DE, Hofman A, Hoes AW. Characteristics of older subjects

using antidepressant drugs. In Egberts ACG. Pharmacoepidemiologic approaches to the evaluation of

antidepressant drugs [Thesis]. University of Utrecht, 1997.

130

Appendix A: Drugs classified as moderately to strongly constipating [15–17]

ATC-code [13] Generic name

A02BX02 Sucralfate

A03AA01 Oxyphencyclimine

A03BA01 Atropine

A07DA03 Loperamide

A12AA03 Calcium gluconate

A12AA04 Calcium carbonate

A12AA05 Calcium lactate

A12AA20 Calcium, combinations

B03AA02 Ferrous fumarate

B03AA07 Ferrous sulphate

B04AD01 Colestyramine

C01BA03 Disopyramide

C08DA01 Verapamil

G04BD02 Flavoxate

G04BD04 Oxybutynin

N02AA01 Morphine

N02AA04 Nicomorphine

N02AB02 Pethidine

N02AC04 Dextropropoxyphene

N04AA01 Trihexyphenidyl

N04AA02 Biperiden

N04AB02 Orphenadrine

N05AA01 Chlorpromazine

N05AC02 Thioridazine

N05AF03 Chlorprothixene

N05AH02 Clozapine

N06AA02 Imipramine

N06AA04 Clomipramine

N06AA09 Amitriptyline

N06AA10 Nortriptyline

N06AA12 Doxepin

N06AA21 Maprotiline

NO6CA01 Amitriptyline plus neuroleptic

R06AD06 Thiazinamium

3.2 Potent ia l interact ion between acenocoumarol and dic lofenac , naproxen and ibuprofen and the ro le of CYP2C9 genotype

K.N. van Dijk1,2, A.W. Plat1, A.A.C. van Dijk1, G. Piersma-Wichers3,A.M.B. de Vries-Bots4, J. Slomp5, L.T.W. de Jong-van den Berg1, J.R.B.J. Brouwers1,4



Groningen, the Netherlands2 Department of Clinical Pharmacy, Medical Centre Leeuwarden, Leeuwarden, the Netherlands3 Groningen Outpatient Thrombosis Service, Groningen, the Netherlands4 Community Pharmacy ‘t Hooge Zand, Hoogezand, the Netherlands5 Department of Clinical Chemistry, Stichting Klinisch Chemisch Laboratorium, Leeuwarden,

the Netherlands

Submitted

133132

Appendix B: Drugs classified as mildly to moderately constipating [15–17]

ATC code [13] Generic name

A03AB03 Oxyphenonium

A03BB01 Scopolamine

A04AA01 Ondansetron

C02AC01 Clonidine

C03CA01 Furosemide

C03CA02 Bumetanide

G02CB02 Lisuride

N02AD01 Pentazocine

N02AE01 Buprenorphine

N02AX02 Tramadol

N03AB02 Phenytoin

N03AE01 Clonazepam

N04BC01 Bromocriptine

N05AA02 Methotrimeprazine

N05AB03 Perphenazine

N05AD05 Pipamperone

N05AF01 Flupenthixol

N05AF05 Zuclopenthixol

N05AG02 Pimozide

N05AL01 Sulpiride

N05AX08 Risperidone

N05BX01 Mephenoxalone

N06AA01 Desipramine

N06AX03 Mianserin

N06AX05 Trazodone

N06AX11 Mirtazepine

R05DA04 Codeine

R06AB02 Dexchlorpheniramine

R06AD02 Promethazine

Int roduct ion

Oral anticoagulation with coumarin derivatives is widely used in the treatment and pro-

phylaxis of patients with thromboembolic diseases [1]. In Europe, acenocoumarol and phen-

procoumon are the drugs most commonly used. In view of the narrow therapeutic range and

the marked inter- and intra-individual variability, the intensity of anticoagulation is monitored

by measuring the prothrombin time [2]. The result of prothrombin time monitoring is expres-

sed as the International Normalised Ratio (INR). The anticoagulant effect of coumarins is in-

fluenced by many drug-food and drug-drug interactions [3]. Among the non-steroidal anti-

inflammatory drugs (NSAIDs), azapropazon and phenylbutazon are contraindicated in patients

treated with coumarins, because they strongly increase the INR as a result of a pharmacokine-

tic interaction [4]. The inhibition of the cytochrome P450 2C9-(CYP2C9) mediated metabolism

of coumarins probably plays a role. Also, displacement of protein binding sites may lead to ele-

vated coumarin levels. Other NSAIDs such as diclofenac, ibuprofen and naproxen have until

now not reported to result in an increase of the INR of patients treated with coumarins.

However, as a result of a pharmacodynamic interaction the risk of bleeding may also be incre-

ased because of the inhibition of thrombocyte aggregation [4]. Furthermore damage of the gas-

trointestinal mucosa may lead to subsequent risk of gastrointestinal bleeding. In view of these

adverse effects, these NSAIDs should only be used in combination with coumarins when no

alternatives are available and patients are advised to monitor an increased bleeding sensitivity.

In the Netherlands, the monitoring of outpatients on oral anticoagulant treatment is con-

ducted by Thrombosis Services [5]. At the Groningen Outpatient Thrombosis Service, physi-

cians occasionally noticed an increase in INR when NSAIDs such as dicofenac, naproxen and

ibuprofen were added to coumarin therapy. In view of the risks of an increased INR, such as

haemorrhage, we investigated the influence of NSAID therapy on the INR in more detail.

Furthermore, we studied the influence of several patient characteristics. It has been suggested

that drug-drug interactions may be partly due to genetic variability [6]. The role of pharmaco-

genomics, defined as the individualisation of drug therapies based on genetic information, in

the prevention of adverse drug reactions has been highlighted in several reviews [6-8]. It was

found that 59% of the drugs cited in studies on adverse drug reactions are metabolised by at

least one enzyme with a variant allele known to cause poor metabolism [6]. Recently it has

been demonstrated that (R)-acenocoumarol, the enantiomer that contributes mainly to the

pharmacological effect is metabolised by CYP2C9 [9]. Diclofenac, naproxen and ibuprofen are

also metabolised by CYP2C9 [10-13]. Furthermore, it has been described that the age-related

decrease in the content and function of CYP2C9 [14] could contribute to drug-drug interactions,

135

Abstract

Objective: To investigate the influence of diclofenac, naproxen and ibuprofen on the Inter-

national Normalised Ratio (INR) of outpatients stabilised on acenocoumarol therapy. To deter-

mine the role of cytochrome P450 2C9 (CYP2C9) polymorphism on coumarin dosage and INR in

NSAID users.

Methods: The study was carried out at the Groningen Outpatient Thrombosis Service. A retro-

spective study among patients who received both acenocoumarol and one of the NSAIDs under

study was performed. Patients whose INR rose above the upper level of the therapeutic range

(> 3.5) after adding an NSAID under study to the acenocoumarol therapy, were compared with

patients who did not show such an elevation. A two-sample t-test (average acenocoumarol

dosage, age), and chi-square tests (sex, therapeutic range, type of NSAID) were used to test for

differences. Genotyping was carried out by analysing blood samples for the relevant CYP2C9

alleles.

Results: The study population consisted of 112 patients on stable acenocoumarol therapy, of

which 52 (46%) showed an elevation of the INR above the desired therapeutic level (INR 3.5

and 4.0 respectively). In 12 patients, the INR increased above 6. The INR of the other 60 patients

(54%) remained constant after the start of one of the NSAIDs under study. There were no sta-

tistically significant differences between patients with increased INR and patients without

increased INR with regard to age, sex, therapeutic range and average acenocoumarol dosage.

Eighty patients, of whom 36 showed an increased INR as a result of a potential acenocouma-

rol-NSAID drug interaction, were included in the genotyping study. No association between

CYP2C9 genotype and an increased INR as a result of the drug-drug interaction was found.

Conclusion: In nearly half of a cohort of elderly patients, the INR increased beyond the thera-

peutic range (INR 3.5 or 4.0) as a result of a potential pharmacokinetic drug-drug interaction

between acenocoumarol and diclofenac, naproxen and ibuprofen. The average increase in INR

was between 1 and 4. Risk factors identifying patients most likely to be at risk for this potential

drug-drug interaction, such as polymorphism of CYP2C9, could not be found.

134

Data col lect ion

All report forms that were issued during the period 01-10-1999 and 01-05-2000 and concer-

ned one of the three NSAIDs under study, were studied. For each patient, the following data

were collected: age and sex, type of NSAID (diclofenac, naproxen or ibuprofen), dosage of ace-

nocoumarol, INR values before and after the administration of the NSAIDs under study, and the

therapeutic range in which the patient was maintained (low (2.5-3.5) or normal (3.0-4.0)).

Data analys is

For each patient a time-profile of the INR was constructed. Patients whose INR increased

above the upper level of the therapeutic range (3.5 and 4.0 respectively, depending on the the-

rapeutic range) after starting diclofenac, naproxen or ibuprofen in addition to the acenocou-

marol therapy, were compared with patients who did not show such an increase. For each

patient, the increase in INR was calculated as the first INR measurement after adding the

NSAID minus the last INR measurement before adding the NSAID. Also, for each patient the

average dosage of acenocoumarol before administration of the NSAID was determined. The

average dosages of patients who showed an elevation in the INR and patients who did not

show such an elevation were subsequently calculated for each therapeutic range.

The statistical software program SPSS 10.0 for Windows (SPSS Inc., Chicago, IL) was used.

A two-sample t-test (acenocoumarol dosage, age) and chi-square tests (sex, therapeutic range,

type of NSAID) were used to test for differences.

Genotyping study

A genotyping study was carried out to investigate whether CYP2C9 polymorphism was

associated with the occurrence of the potential interaction between acenocoumarol and the

NSAIDs under study. Participants were recruited from patients attending the Groningen

Outpatient Thrombosis Service who were initially included in the cohort study (n=112). The

Medical Ethical Committee of the ‘Stichting Beoordeling Ethiek Bio-Medisch Onderzoek’ at

Assen, the Netherlands, approved the study (July 06, 2001). The general practitioner (GP) of

each patient was contacted by telephone and by fax with details of the proposed investigation

and was asked written consent to contact their patient for inclusion in the genotyping study.

Patients were included in the study after written informed consent was obtained. At their regu-

lar visit for INR control, 4 ml blood was collected in EDTA tubes. DNA was isolated from blood

within 72 hours after blood was collected (High Pure PCR Template Preparation Kit, Roche).

Polymerase Chain Reaction (PCR) for the CYP2C9 variants was performed with subsequent Ava

II digestion (CYP2C9*2) and Nsi I and Kpn I digestion (CYP2C9*3). CYP2C9*1/*3 and CYP2C9*2

137

which could be relevant in an elderly population. Polymorphism of CYP2C9 could therefore be

a risk factor for developing a clinical relevant drug-drug interaction between acenocoumarol

and NSAIDs.

The aim of our study was to investigate whether diclofenac, naproxen or ibuprofen influ-

enced the INR of acenocoumarol users. Furthermore, we determined whether polymorphism of

CYP2C9 is associated with this potential drug-drug interaction.

Methods

Sett ing

The study was carried out at the Groningen Outpatient Thrombosis Service, which serves

about 11,000 patients in the province of Groningen, the Netherlands. In this anticoagulation

clinic, all activities with regard to monitoring and treatment of patients on coumarin therapy

are organised. For example, blood sampling is performed and laboratory determinations are

carried out. An especially trained physician establishes the dose of the coumarin using a com-

puter dosage program. Letters with the recommended dosages are subsequently sent to the

patient. Patient and medical data are stored in a centralised database (Trombosedienst

Information System; TDAS). Two oral anticoagulants are available in the Netherlands: aceno-

coumarol and phenprocoumon. Two therapeutic ranges are distinguished: a first intensity level

(INR therapeutic range 2.5-3.5), and a second intensity level (INR therapeutic range 3-4). The

indication and pathophysiology determines in which therapeutic range a patient is maintained.

Patients with venous thromboembolic disorders, such as atrial fibrillation, are mostly maintai-

ned in the first intensity level and patients with arterial thromboembolic diseases are mostly

maintained in the second intensity level. Patients are maintained in the therapeutic range by

regular checks of their INR. A special feature of the Groningen Outpatient Thrombosis Service

is that all reports concerning potential coumarin-drug interactions are systematically recorded

in a database.

Study populat ion

The study population consisted of outpatients on stable acenocoumarol treatment who

received one of the NSAIDs under study. Stable acenocoumarol treatment was defined as main-

tenance of the INR within limits of the therapeutic range. Patients, who used more drugs bes-

ides the NSAID, were excluded.

136

139

were detected by PCR-mediated site-directed mutagenesis followed by restriction analysis

according to the methods described by Wang and colleagues (*1/*3) [15] and Steward and col-

leagues (*2) [16] with slight modifications. Positive controls were included for method valida-

tion.

Results

The source population consisted of 244 patients. Of these patients, 132 (54%) were exclu-

ded because they used more medications besides the NSAID and acenocoumarol, leading to a

study population of 112 patients. There were no patients who used more than one NSAID. Of the

112 patients, 52 (46%) showed an elevation of the INR above the upper level of the therapeu-

tic range after adding diclofenac, naproxen or ibuprofen to acenocoumarol therapy. The INR of

the other 60 patients (54%) remained constant after the start of one of the NSAIDs under

study. In table 1, patient characteristics of the study population are given. There were no statis-

tically significant differences between patients with increased INR and patients without in-

creased INR for the following variables: age, sex, type of NSAID (diclofenac, naproxen or

ibuprofen), therapeutic range, average dosage of acenocoumarol (before the NSAID was

added) and CYP2C9 genotype. In table 2 the average increase in INR is given, stratified for thera-

peutic range and type of NSAID. In twelve patients the INR was 6 or higher. In figure 1, a time-

profile is given, illustrating the elevation in INR after adding diclofenac in one patient.

138

Table 1: Characteristics of the study population (n=112) and differences between acenocoumarol users with increased

INR and acenocoumarol users without increased INR, and results of the genotyping study (n=80)

Variable Patients with increased INR (n=52) Patients without increased INR (n=60)

n (%) n (%)

Sex#

Male 21 (40.4) 26 (43.3)

Female 31 (59.6) 34 (56.7)

Age; average (± SD) # 72.2 (± 14.1) 70.9 (± 11.3)

Age distribution

<60 9 (17.3) 11 (18.3)

60-69 9 (17.3) 15 (25.0)

70-79 17 (32.7) 24 (40.0)

≥ 80 17 (32.7) 10 (16.7)

Therapeutic range#

2.5-3.5 30 (57.7) 34 (56.7)

3.0-4.0 22 (42.3) 26 (43.3)

Type of NSAID#

Diclofenac 32 (61.5) 34 (56.7)

Naproxen 8 (15.4) 16 (26.7)

Ibuprofen 12 (23.1) 10 (16.7)

Average acenocoumarol

dosage (mg) before

NSAID was added#

overall 2.7 3.1

2.5-3.5 2.8 2.9

3.0-4.0 2.6 3.2

CYP2C9 genotype¶

CYP2C9*1/*1 26 (72.2) 30 (68.2)

CYP2C9*1/*2 6 (16.7) 9 (20.5)

CYP2C9*2/*2 0 2 (4.5)

CYP2C9*1/*3 3 (8.3) 3 (6.8)

CYP2C9*2/*3 1 (2.8) 0

¶ Not all patients of the cohort (n=112) were included in the genotyping study (see also under Results): 36 patients

with increased INR; and 44 patients without increased INR# None of the differences between the two groups were statistically significant (p>0.05)

Genotyping study

All GPs gave consent to contact their patients. Subsequently, 80 patients gave their written

informed consent to participate in the study. Thirty-one patients could not be included, due to

different reasons such as hospital admission (n=5) or lost to follow-up (n=26). One patient did

not give informed consent. In table 3, the characteristics of the patients who were included in

the genotyping study are given. The CYP2C9 genotype distribution is summarised in table 1. Ten

patients (27.8%) in the case group were carriers of one or two of the allelic variants of CYP2C9

compared with 14 (31.8%) of the patients in the control group (Wilson test for difference: CI95

-0.16-+0.23). Among the patients with increased INR, the allele frequencies of CYP2C9 were

84.7%, 9.7% and 5.6% for the variants CYP2C9*1, *2 and *3, respectively. Among patients

without increased INR, the allele frequencies were 81.8%, 14.8% and 3.4% for the variants

CYP2C9*1, *2 and *3, respectively. The differences between both groups in allele frequency

were not statistically significant. For the study group as a whole (n=80), the allele frequencies

were 83.1%, 12.5% and 4.4% for the variants CYP2C9*1, *2 and *3, respectively. When we com-

pared patients having the wild-type genotype (*1/*1) with patients having one of the allelic

variants (*2 and/or *3), it was found that the latter group had a significantly lower dosage of

acenocoumarol (2,3 mg versus 3,0 mg; p=0.023).

141140

Table 2: Average increase of INR stratified for type of NSAID and therapeutic range in the study population (n=112)

Average increase in INR (range)#

Therapeutic range Diclofenac (n=32) Naproxen (n=8) Ibuprofen (n=12)

2.5 –3.5 + 2.3 (0.5-8.4) + 2.8 (0.5-5.3) + 3.8 (2.0-8.6)

(n=17) (n=7) (n=6)

3.0-4.0 + 2.2 (0.3-4.8) + 1.1 + 1.5 (0.4-2.4)

(n=15) (n=1) (n=6)

# defined as the difference between the last INR (among patients who showed an elevation in the INR) before the

start of an NSAID and the first INR after the start of an NSAID.

3,63,8

4,13,9

3,33,3

3,83,7

6,4

0

1

2

3

4

5

6

7

0 28 56 84 112 140 168 196 224 252

Time-intervals (days)

INR

INR

Min.th.value

Max.th.value

NSAID added²

Figure 1: Patient INR-profile showing an elevation in INR after adding diclofenac to coumarin therapy.

Table 3: Characteristics of the patients who were included in the genotyping study (n=80)

Genotype CYP2C9

Variable *1/*1 *1/*2 *2/*2 *1/*3 *2/*3

n(%) 56 (70) 15 (18.8) 2 (2.5) 6 (7.5) 1 (1.3)

Sex

Female 33 (58.9) 12 (80) 0 2 (33.3) 0

Male 23 (41.1) 3 (20) 2 (100) 4 (66.7) 1 (100)

Age 70.8 73.7 70.0 70.8 65.0

Therapeutic range

2.5-3.5 26 (65) 9 (64.3) 1 (50) 4 (100) 0

3.0-4.0 14 (35) 5 (35.7) 1 (50) 0 1 (100)

Type of NSAID

Diclofenac 34 10 1 3 0

Naproxen 13 1 1 2 0

Ibuprofen 9 4 0 1 1

Acenocoumarol dosage (mg)

3.048 2.628 2.071 1.482 1.786

mentioned earlier, phenylbutazone shows a pharmacokinetic interaction with coumarins [3],

resulting in an increase in INR. Other NSAIDs such as diclofenac and naproxen are reported to

increase the risk of bleeding, but no evidence of a pharmacokinetic interaction with coumarins

has been reported [20]. In an open-label study among 8 healthy volunteers no effect of piroxi-

cam on the pharmacokinetics of acenocoumarol enantiomers was found [21]. In a placebo-

controlled study among 56 osteoarthritis patients treatment with nabumetone for up to 4

weeks did not alter INR levels compared with placebo [22]. In an open crossover study among

6 healthy volunteers lornoxicam did not alter the pharmacokinetics of the clinically relevant

(R)-acenocoumarol or the anticoagulant activity of acenocoumarol [23]. Warfarin drug inter-

actions with NSAIDs have been described more frequently, possibly due to the fact that war-

farin is the oral anticoagulant most widely used in the United States. One case-report descri-

bing a patient with a seriously prolonged INR during the combined use of warfarin and indo-

methacin was found [24]. Recently, two case-reports [25,26] have been published showing a

significant increase in INR after adding celecoxib, a cyclooxygenase-2 selective NSAID, to war-

farin therapy. A possible mechanism to explain this interaction could be that both medications

are metabolised through the same cytochrome P450 pathway (namely CYP2C9). Karim and col-

legues showed no significant effect of celecoxib on pro-thrombin times or steady-state phar-

macokinetics of S- and R-warfarin in a small study with 24 healthy volunteers [27]. In an USA

ambulatory care anticoagulation clinic over a period of one year, 28 patients had been prescri-

bed either celecoxib or rofecoxib after being stable on warfarin therapy. Thirteen of these had

increases in INR, within this group 6 patients used only a coumarin (warfarin) with an NSAID

(cyclo-oxygenase-2 inhibitor) [28]. These results are comparable with our results of acenocou-

marol with NSAIDs. For celecoxib that is metabolized by CYP2C9 competitive inhibition of

CYP2C9 metabolism could lead to this potential pharmacokinetic drug interaction. Rofecoxib is

not reported to inhibit CYP2C9. However, rofexocib is approximately 87% bound to plasma pro-

tein, resulting in higher free-warfarin levels and thus possibly accounting for the increasing

INRs [28].

Different mechanisms may account for our findings. As stated earlier due to competitive

inhibition of the NSAIDs of (R)-acenocoumarol 7-hydroxylation, the plasma level of (R)-aceno-

coumarol may be increased. Furthermore, inhibition of NSAIDs of the p-glycoprotein pump

may account for increased acenocoumarol levels. Protein displacement processes may also play

a role. CYP2C9 genotype seems to be not a risk factor for the occurrence of a potential clinical

relevant increase in INR as a result of the interaction between acenocoumarol and diclofenac,

naproxen or ibuprofen. Although polymorphism of CYP-metabolising enzymes may play an

important role in predicting adverse drug effects, including drug-drug interactions [6], we could

143

Discuss ion

In this study we found that nearly half of a cohort of acenocoumarol users on average

showed an increase in INR between 1 and 4 after diclofenac, naproxen or ibuprofen was added

to the acenocoumarol therapy. Considering the estimation that the risk of bleeding increases

with 54% (CI95 44%-65%) for every unit increase in INR [2], these rises in INR should be con-

sidered clinically relevant. Recently it was shown that patients with INRs greater than 6 have

a significant short-term risk of major haemorrhage compared with patients with an in-range

INR [17]. In our study, the INR of 12 patients rose above 6. This means that 11% of the study

population was at risk of major haemorrhage. We did not find an association between CYP2C9

genotype and the risk of an increased INR as a result of the potential pharmacokinetic interac-

tion between acenocoumarol and NSAIDs. However, patients with one of the variant alleles (*2

or *3), had a significantly lower acenocoumarol dosage than patients without one of the variant

alleles. This could mean that acenocoumarol dosage requirements are dependent on CYP2C9

genotype. We are currently investigating this matter in more detail by analysing blood (R)- and

(S)-acenocoumarol levels.

Thijssen and co-workers showed a strong association between acenocoumarol sensitivity

and the presence of the CYP2C9*3 allele possibly due to impaired acenocoumarol clearance, in

particular the (S)-enantiomer [18]. They stated that because of genetic polymorphism, the meta-

bolic clearance of (S)-acenocoumarol is reduced profoundly, and this enantiomer, that is nor-

mally clinically inactive, may now exert main anticoagulant activity. In warfarin users, Aithal

found that patients with CYP2C9 variant alleles were found to require lower warfarin dosages

[19]. Our results seem to be in line with the results reported by Aithal and Thijssen.

In this study data from patients in daily clinical practice were used, reflecting the clinical

relevance of our findings. Furthermore, we restricted our analyses to patients who did not use

any other drugs, thus making a direct cause-and-effect relationship more likely. A limitation of

this study is that we can not rule out confounding by indication. It is known that during episo-

des of fever and inflammation, the breakdown of clotting factors is increased, leading to a

greater response to coumarin therapy and thus an increased INR. The fact that we used a con-

trol group that did not show an increase in INR, reduces the bias as a result of confounding by

indication considerably. Also, we excluded patients who used other drugs simultaneously, such

as antibiotics, further limiting confounding by indication. Finally, by stratifying patients in the-

rapeutic windows, depending on the diagnosis, a further limitation of bias due to confounding

by indication was achieved.

In the literature, several reports on coumarin-NSAID interactions have been published. As

142

References

1 Hirsch J. Oral anticoagulant drugs. New Engl J Med 1991; 324: 1865-75.

2 Van der Meer FJM, Rosendaal FR, Vandenbroucke JP, Briët E. Bleeding complications in oral anticoagulant

therapy. Arch Intern Med 1993; 153: 1557-62.

3 Harder S, Thürmann P. Clinically important drug interactions with anticoagulants. An update.

Clin Pharmacokinet 1996; 30: 416-44.

4 Brouwers JRBJ, De Smet PAGM. Pharmacokinetic-pharmacodynamic drug interactions with nonsteroidal

anti-inflammatory drugs. Clin Pharmacokinet 1994; 27: 462-85.

5 Breukink-Engbers WG. Monitoring therapy with anticoagulants in the Netherlands. Seminars in Trombosis

& Hemostasis 1999; 25: 37-42.

6 Philips KA, Veenstra DL, Oren E, Lee JK, Sadee W. Potential role of pharmacogenomics in reducing adverse drug

reactions. JAMA 2001; 286: 2270-9.

7 Roland Wolf C, Smith G, Smith RL. Pharmacogenetics. BMJ 2000; 320: 987-90.

8 Meyer UA. Pharmacogenetics and adverse drug reactions. Lancet 2000; 356: 1667-71.

9 Thijssen HH, Flinois JP, Beaune PH. Cytochrome P450 2C9 is the principal catalyst of racemic acenocoumarol

hydroxylation reactions in human liver microsomes. Drug Metab Dispos 2000; 28: 1284-90.

10 Klose TS, Ibaenu GC, Ghanayem BI, Pedersen LG, Li L, Hall SD, et al. Identification of residues 286 and 289 as

critical for conferring substrate specificity of human CYP2C9 for diclofenac and ibuprofen. Arch Biochem

Biophys 1998; 357: 420-8.

11 Bliesath H, Huber R, Steinijans VW, Koch HJ, Wurst W, Mascher H. Lack of pharmacokinetic interaction between

pantoprazole and diclofenac. Int J Clin Pharmacol Ther 1996; 34 (suppl 1): S76-80.

12 Leemann T, Transon C, Dayer P. Cytochrome P450 TB (2C9): a major monooxygenase catalyzing

diclofenac 4’-hydroxylation in human liver. Life Sci 1993; 52: 29-34.

13 Miners JO, Coulter D, Tukey RH, Veronese ME, Birkett DJ. Cytochromes P450, 1A2 and 2C9 are responsible for the

human hepatic O-demethylation of R- and S-naproxen. Biochem Pharmacol 1996; 51: 1003-8.

14 Sotaniemi EA, Arranto AJ, Pelkonen O, Pasanen M. Age and cytochrome P450-linked drug metabolism in

humans: an analysis of 226 subjects with equal histopathological conditions. Clin Pharmacol Ther 1997; 61: 331-9.

15 Wang S-L, Huang J-D, Lai M-D, Tsai J-Jl. Detection of CYP2C9 polymorphism based on polymerase chain

reaction in Chinese. Pharmacogenetics 1995; 5: 37-42.

16 Steward D, Haining RL, Henne KR, Davis G, Rushmore TH, Trager WF, Rettie AE. Genetic association between

sensitivity to warfarin and expression of CYP2C9*3. Pharmacogenetics 1997; 7: 361-7.

17 Hylek EM, Chang YC, Skates SJ, Hughes RA, Singer DE. Prospective study of the outcomes of ambulatory

patients with excessive warfarin anticoagulation. Arch Intern Med 2000; 160: 1612-7.

18 Commissie Interacterende Medicatie Cumarines. Standaard afhandeling cumarine-interacties (in Dutch).

Wetenschappelijk Instituut Nederlandse Apothekers (WINAp). The Hague, The Netherlands, 1999.

19 Thijssen HHW, Verkooyen IWC, Frank HLL. The possession of the CYP2C9*3 allele is associated with low dose

requirement of acenocoumarol. Pharmacogenetics 2000; 10: 1-4.

20 Aithal GP, Day CP, Kesteven PJ, Daly AK. Association of polymorphism in the cytochrome P450 CYP2C9 with

warfarin dose requirement and risk of bleeding complications. Lancet 1999; 353: 717-9.

21 Bonnabry P, Desmeules J, Rudaz S, Leemann T, Veuthey JL, Dayer P. Stereoselective interaction beween

piroxicam and acenocoumarol. Br J Clin Pharmacol 1996; 41: 525-30.

22 Pardo A, Garcia-Losa M, Fernandez-Pavon A, Castillo del S, Pacual-Garcia T, Garcia-Mendez E, et al.

A placebo-controlled study of interaction between nabumetone and acenocoumarol.


23 Masche UP, Rentsch KM, von Felten A, Meier PJ, Fattinger KE. No clinically relevant effect of lornoxicam intake

145

not confirm this role in the case of the interaction between acenocoumarol and the NSAIDs

studied. As a result of this finding, at present we are not able to identify patients most likely to

be at risk for a potential clinical relevant interaction between acenocoumarol and diclofenac,

naproxen or ibuprofen.

In conclusion, this study shows that diclofenac, naproxen and ibuprofen can lead to a clini-

cally relevant increase of the INR in patients treated with acencoumarol. In 11% of the study

population, this lead to INRs above 6 with a clinically relevant risk of severe haemorrhage [17].

Genetic polymorphism of CYP2C9 did not contribute to the occurrence of this potential clinical

relevant interaction. In view of the widespread use of NSAIDs and the fact that these drugs can

be purchased without a doctor’s prescription, further prospective studies on this drug-drug

interaction should be initiated. In the meantime, adequate surveillance and close monitoring

of the INR of patients receiving these drugs concomitantly is warranted.

Acknowledgements

We are grateful to C.Th. Smit Sibenga, PhD, and his colleagues from the Bloedbank Noord

Nederland, Groningen, the Netherlands, for their co-operation in transporting patient blood

samples. We thank the laboratory assistants of the Outpatient Thrombosis Service Groningen,

Groningen, the Netherlands, for their co-operation in taking blood samples. Furthermore we

are grateful to the laboratory assistants of the Clinical Chemistry Laboratory at Leeuwarden for

their help in the genotyping part of the study. C.S. de Vries, PhD, RPh, is thanked for her valu-

able comments on the manuscript.

144

147

on acenocoumarol pharmacokinetics and pharmacokinetics. Eur J Clin Pharmacol 1999; 54: 865-8.

24 Chan TY. Prolongation of prothrombin time with the use of indomethacin and warfarin.

Br J Clin Pract 1997; 51: 177-8.

25 Haasse KK, Rojas-Fernandez CH, Lane L, Frank DA. Potential interaction between celecoxib and warfarin

(letter). Ann Pharmacother 2000; 34: 666-7.

26 Mersfelder TL, Stewart LR. Warfarin and celecoxib interaction. Ann Pharmacother 2000; 34: 325-7.

27 Karim A. Tolber D, Piergies A, Hubbard RC, Harper K, Wallemark CB, Slater M, Geis GS. Celecoxib does not

significantly alter the pharmacokinetics or hypoprothrombinemic effect of warfarin in healthy subjects.

J Clin Pharmacol 2000; 40: 655-63.

28 Stading JA, Skrabal MZ, Faulkner MA. Seven cases of interaction between warfarin and cyclooxygenase-2

inhibitors. Am J Health-Syst Pharm 2001; 58: 2076-80.146

Chapter 4

General d iscuss ionand perspect ives

the potential risks associated with drug use in these nursing homes. Pharmacy dispensing data

for 2,355 residents were retrieved. It was found that the hospital pharmacy data were, in gener-

al, accurate, although completeness of discharge and admission date recording could be impro-

ved. With these data, the duration of stay in the nursing home could be calculated, as well as

the duration of drug use. Hospital pharmacy data proved to be an important data source, ena-

bling the performance of drug utilisation and drug safety studies. We demonstrated the impor-

tance of accurate and precise recording in the hospital pharmacy, of prescription data for indi-

vidual nursing home patients in order to reliably determine drug exposure. The main findings

and implications of these studies are described below. First, a drug utilisation study was per-

formed. We found high numbers of drug users and the chronic use of many drugs. This study

revealed several potential problem areas regarding the prescribing of drugs in the nursing

home setting. Prescribing of loop diuretics, laxatives, psychotropic drugs and ulcer-healing

drugs deserved attention, in view of high dosages, long-term use and a high proportion of

users. Prescribing practices in individual nursing homes regarding these drug groups may sub-

sequently be evaluated in pharmacotherapeutic discussion meetings or, on the level of an indi-

vidual patient in a one-to-one discussion with the prescriber. Hospital pharmacists can play a

key role in these evaluations since they have the tools to analyse individual prescription data.

From the drug utilisation studies, several determinants of drug use were found. Sex was found

to be a determinant of drug use in several cases, as was the type of care. It was shown that in

drug utilisation studies it is important to distinguish between nursing home residents residing

in psychogeriatric nursing homes, and those residing in somatic nursing homes as drug use

substantially differs between these populations. As expected, nursing home residents residing

in psychogeriatric nursing homes use more psycholeptic drugs (psychotropics and anxiolytics)

and less antithrombotic drugs, diuretic drugs and antacids than residents residing in somatic

nursing homes. Also, somatic residents showed a higher risk for the occurrence of potential

drug-drug interactions (DDIs). Female residents were more likely than male residents to expe-

rience a potential DDI and were more likely to receive NSAIDs, antirheumatic drugs and psy-

choanaleptic drugs. Male residents were more likely to receive psychotropic drugs. Other deter-

minants of drug use found were the number of medications prescribed. Residents with a hig-

her number of medications were more at risk for the occurrence of potential DDIs than resi-

dents with fewer drugs. Patients with Parkinson’s disease were less likely to be exposed to

potential DDIs.

In view of the high frequency of drug use, we studied the potential risks of polypharmacy

by carrying out a descriptive study on the extent and occurrence of potential DDIs. Several pre-

scribing indicators were used to assess the occurrence and nature of DDIs, such as the number

149148

Pharmacotherapy in f ra i l e lder ly

People residing in Dutch nursing homes are mostly frail, elderly people who are dependent

on continuous nursing and medical attention. Pharmacotherapy in this elderly group is an

important aspect of medical care. Drugs obviously have beneficial effects in the very old. For

example, the use of statins should not be limited to people younger than 70 years, as statins

can reduce the risk of myocardial infarction even in the very old [1,2]. In addition, the benefits

of oral anticoagulant therapy in patients suffering from atrial fibrillation, have been well ack-

nowledged. For some therapies, such as hormone-replacement therapy in elderly women for

reducing the risk of osteoporosis [3], it is not yet clear who will receive the most benefit.

However, the impact of adverse effects of pharmacotherapy in the frail elderly is generally higher

than in other populations. Non-steroidal anti-inflammatory drugs (NSAIDs) may lead to renal

dysfunction [4], anticholinergic drugs to confusion and delirium [5], and psychotropic drugs to

falls and fractures [6], often with significant impact on the quality of life. Therefore, the risks of

drug therapy should be carefully weighed against the potential beneficial effects, especially in

this vulnerable group of elderly persons. Pharmacoepidemiologic studies can provide insight

into this delicate balance. The benefits of drug use are mainly studied in randomised control-

led clinical trials (RCTs). Observational studies are less suitable for evaluating the benefits of

drug use in view of biases and confounding issues such as selection by disease severity.

However, observational studies are often used to study adverse drug effects. This is due to

various considerations. These include differences between the experimental RCT setting and

prescribing in daily clinical practice such as co-morbidity and comedication profiles and diffe-

rences in patient age. This thesis is largely based on observational studies.

Pharmacoepidemiology in f ra i l e lder ly

Organisational differences between nursing homes in the Netherlands and in other coun-

tries, in particular the United States, will have an impact on drug utilisation. Consequently, fin-

dings from one health care setting cannot automatically be considered true for another setting.

Therefore, to gain insight into local drug use, it is important that drug utilisation studies are

performed within the applicable health care system. This thesis reports on an investigation of

drug use and drug-related problems in the elderly in the Netherlands, in particular those resi-

ding in nursing homes. To draw conclusions regarding drug use and safety, a sufficiently large

number of subjects is needed. We collected pharmacy dispensing data from 6 nursing homes.

In this way, we were able to study the extent of drug use, problem areas in drug prescribing and

feasible to perform a prescription-sequence analysis using pharmacy data in the nursing home

setting. Although the constipating effects of certain drugs should be taken into account, this

may not necessarily explain the high rate of laxative use found in the nursing homes we stu-

died. Effects of dietary factors and physical activity on laxative use needs further investigation.

The studies discussed above were carried out to identify problem areas in drug prescribing in

the elderly, especially those residing in nursing homes. Results of these studies may help to

identify patients at increased risk of drug-related problems.

How to ident i fy pat ients at r i sk?

To enable the identification of patients at risk of drug-related problems, more information

than pharmacy data alone is needed. Clinical data and information from prescribers can provi-

de insight into clinical outcomes, as well as into any preventive measures that have already

been undertaken and how these patients are currently being monitored. One way to identify

patients at risk of unwanted pharmacotherapy effects, is to use prescribing indicators to signal

potentially suboptimal prescribing. Although many prescribing indicators have been developed

and used internationally [9], few of these can be used with pharmacy prescription data and few

are specific for drug-related problems common in nursing homes. In a pilot study [10], we found

that a set of prescribing indicators developed on the basis of current pharmacotherapy guide-

lines, could be used to evaluate prescribing practices in nursing homes and to identify patients

at risk of potential suboptimal prescribing. However, information from the prescriber on clini-

cal data was needed to identify patients who were actually at risk for drug-related problems. In

some cases, deviation from national guidelines and drug formularies may not mean that the

patient is treated suboptimally. This discrepancy between potential and actual inappropriate

prescribing has been reported earlier [11] and stresses the importance of taking the patient’s

clinical response and the prescribers’ rationale into consideration. Furthermore, it is often dif-

ficult to define prescribing indicators that actually reflect potential suboptimal prescribing.

Although evidence-based practice guidelines may serve as a basis, these sometimes differ from

expert opinions [12]. In nursing home practice, evidence-based guidelines are just beginning to

emerge, and the lack of specific nursing home guidelines hampers the development of good

prescribing indicators in this setting. In many cases, however, prescribing indicators can be

used to signal potentially suboptimal prescribing, as was also found in recent studies [12,13].

Together with clinical data, such as laboratory values or clinical outcomes, the risks for the

patient can be adequately perceived and preventive measures taken on both an individual

patient and population-based level. In our view, the pharmacological knowledge of the hospi-

151

of residents exposed to a DDI, the prevalence of a DDI, the percentage of days of concomitant

drug use, and the drugs most frequently involved in a DDI. Approximately one-third of the nur-

sing home residents was exposed to at least one clinically relevant DDI. However, for each

clinically relevant DDI registered, no more than 10% of the residents were affected. This indi-

cates that not one DDI was identified that could be considered potentially harmful to many

residents. The interaction between NSAIDs and loop diuretics, and between NSAIDs and oral

anticoagulants were the potential DDIs most frequently encountered (9.7% and 9.6% of the

study population respectively). It is important to know whether these potential DDIs actually

lead to clinically adverse outcomes. At present, the perceived clinical relevance of DDIs is often

based on theoretical considerations and case-reports [7]. However, the actual clinical relevan-

ce is unknown. Studies should be undertaken investigating the clinical effects of DDIs in order

to identify both DDIs that are likely to cause problems and people at a higher risk of developing

these problems [8]. In particular the frail elderly have a high risk of potential DDIs and drug-

disease interactions as a result of polypharmacy and co-morbidity, and knowledge on the

clinical relevance of DDIs could contribute to a safer use of drugs. At present, it is often difficult

to collect data on individual clinical outcomes, such as laboratory test results, on a large scale

because these data are often not stored in easily accessible files. Subsequently, we investiga-

ted in more detail the co-prescribing of several drug groups frequently used in the nursing

home population. We evaluated co-prescribing of benzodiazepines and antidepressant thera-

py to investigate previously reported differences in co-prescribing between tricyclic antide-

pressants (TCAs) and selective serotonin reuptake inhibitors (SSRIs). We found that in elderly

outpatients the risk for initiating benzodiazepines during antidepressant therapy was higher

for SSRI users than for TCA users. However, in the nursing homes patients no such difference

was found. This study also illustrated the need to perform drug utilisation studies in different

elderly populations as the results may differ depending on which population is studied. In

another study we investigated co-prescribing of NSAIDs and gastroprotective drugs in elderly

outpatients. It was found that 23% of NSAID users aged 65 and over were co-prescribed gastro-

protective drugs. Given this finding, initiatives should be taken to increase gastroprotective

drug use in elderly NSAID users. Also, more insight is needed into the reasons concerning why

prescribing gastroprotective drugs were not more frequently prescribed to this high-risk group.

Finally, to investigate the determinants of the high frequency of laxative use in the nursing

homes in more detail, a study was carried out to see whether laxative use was the consequen-

ce of prescribing constipating drugs, such as anticholinergics. It was shown that drugs classi-

fied as moderately to severely constipating were associated with laxative use. However, the

association was not as strong as previous studies have suggested. This study showed that it is

150

pital pharmacy computer systems.

• Clinical status of the patient: this includes diagnoses, data from the nursing home physi-

cians and medical specialists visits, and clinical outcomes. In Dutch nursing homes, these

data are manually recorded in the patients’ medical chart. Sometimes electronic patient

records are available.

Qual i ty of l i fe

It could be argued, that together with clinical outcomes such as the frequency of adverse events

another issue that needs to be taken into account is the quality of life (QoL) in the frail elderly

population. For instance, in clinical practice the impact of side effects from ACE-inhibitors in

this population, is higher than in the general population. This may lead to a more adverse risk

- benefit balance. The impact of various drug-related problems on this populations’ quality of

life, will therefore be different to that in the general population. Measuring QoL, however, is not

something that is routinely done in nursing home patients and such data are not readily avai-

lable. In the elderly, the ‘Short Form-36’ (SF36) is regarded as a suitable, reliable and valid

questionnaire and outcome tool [16,17]. However to our knowledge, this questionnaire has

been used only in community-dwelling elderly and not in Dutch nursing home residents. A high

incidence of non-completion on SF36 questions relating to physical and mental function has

been reported [16]. In nursing homes the prevalence of physical and mental disorders among

residents is much higher and therefore the SF-36 is possibly not a suitable tool for measuring

QoL. It is with these considerations in mind, that we believe that routinely measuring QoL in

nursing home patients is probably not one of the first issues to focus on. Instead, QoL should

be taken into account when the effects of drug use are evaluated on an individual basis where

relevant or when possible.

Future perspect ives

Opt imal use of c l in ica l and pharmacy data to monitor drug effects

Efforts should be directed towards optimal use of existing clinical, pharmacy and laborato-

ry records. This may lead to several advances in the pharmacotherapeutical care of the frail

elderly. First, on the basis of clinical characteristics individual patients could be monitored pro-

spectively regarding potential adverse drug effects and adverse outcomes. Record-linkage of

pharmacy and clinical data provides an excellent opportunity to optimally use both sources of

information. In the nursing home setting, most relevant information is available. Since not all

153

tal pharmacist together with the clinical knowledge of the nursing home physician may act

synergistically in detecting drug-related problems in the elderly. Also other caregivers, such as

nurses, may play an important role in signalling adverse drug effects in clinical nursing home

practice.

The ultimate question, however, is whether potentially suboptimal prescribing leads to

actual clinical problems [14]. We investigated this with respect to a frequently occurring, poten-

tial DDI. The potential DDI between acenocoumarol and NSAIDs constitutes one of the most

frequently encountered DDIs in the nursing home study population and to date its clinical rele-

vance remains unknown. The study presented in this thesis found that this interaction led to a

clinically relevant outcome in about half of all elderly outpatients exposed, namely the increa-

se of the prothrombin ratio expressed as the international normalised ratio (INR) above the

therapeutic window. To be able to identify patients at risk for this drug-drug interaction, sever-

al patient characteristics were investigated as potential determinants. One of the determinants

we included was genetic polymorphism of cytochrome P450 2C9 (CYP2C9). This enzyme is

involved in the metabolism of numerous drugs including acenocoumarol and NSAIDs, such as

diclofenac, naproxen and ibuprofen. We hypothesized that patients with a variant allele on

CYP2C9 (CYP2C9*2 and CYP2C9*3) would be more likely to experience the increase in pro-

thrombin time due to the DDI. We found that none of the determinants investigated was asso-

ciated with the occurrence of increase in prothrombin time. However, patients with one of the

variant alleles mentioned above, required a lower dose of acenocoumarol than patients who

had the wild-type genotype and we are currently investigating this matter in more detail. In the

meantime, all patients who are prescribed acenocoumarol and an NSAID simultaneously

should be monitored, in view of the increased risk for elevated INR levels.

The studies described above, show the importance of capturing data on drug use as well as

data on clinical variables to be able to identify patients at risk for drug-related problems. The

following clinical variables should be included:

• Laboratory test values: insight into potassium levels, creatinine levels, INR, glucose

levels, cholesterol levels and other parameters is often necessary to identify adverse

effects of drug therapy such as decreased renal function as a result of NSAID therapy.

These data are often available in clinical chemistry laboratory computer systems, and

sometimes in the nursing home computer systems. Hard copies of the laboratory test

results are available in the medical chart of the patient. In the near future, results of

genotyping may also be available [15].

• Serum drug levels: information on serum drug levels may be needed to determine the cli-

nical relevance of certain drug-drug interactions. These data are routinely stored in hos-

152

have all been associated with an increased risk of falls in nursing home residents. However,

other clinical outcomes have been studied less frequently. An example is the association

between bowel function and anticholinergic drug use [26]. Quality of life could also be an

important outcome when the risks and benefits of drug use in the frail elderly are balanced.

While the number of life years gained often plays a decisive role in determining the value of a

pharmacotherapeutic intervention, this may not be the case in the frail elderly. The impact of

adverse drug effects is much higher in this group, and quality of life could in some cases deter-

mine whether or not drug therapy should be started. For example, in the treatment of heart fai-

lure, the adverse effects of drug therapies may be more important than the number of life years

gained.

Role of the hospi ta l pharmacist

There are various stages at which hospital pharmacists can play a key role in monitoring

drug-related problems in frail, elderly people. Firstly, hospital pharmacists have the opportuni-

ty to use pharmacy data for drug utilisation research. The studies in this thesis have demon-

strated that hospital pharmacy data are a useful tool for studying the use and effects of drugs

in the elderly, especially in the nursing home setting. Further use of these data for research

should be encouraged, which implies the development of initiatives for anonymous and conti-

nual data collection. Special export files [27] would facilitate the building of a database that can

be continuously updated with current data from pharmacies. The fact that the number of hos-

pital pharmacies using the same computer system is still increasing in the Netherlands contri-

butes to the building of such a database and further collaboration is needed between hospital

pharmacies on this point. In particular, collaboration in the field of nursing home medicine

could contribute to the creation of large databases for adequate performance of drug utilisa-

tion and safety studies. An important prerequisite is that the prescription data cover a suffi-

ciently long period, preferably several years. Initiatives concerning hospital prescription data

arising from the Stichting Farmaceutische Kengetallen (SFK), an organisation that forms part of

the Royal Dutch Association for the Advancement of Pharmacy (KNMP), are currently being

explored. Secondly, hospital pharmacists can play an important role in initiating and conduc-

ting pharmacoepidemiology research in nursing homes. Special interest groups could be for-

med, to facilitate and carry out research projects related both to drug-related problems and

other relevant clinical outcomes in the nursing home. Monitoring of ADR occurrence is also an

issue in which hospital pharmacists could play an essential role as they have the tool to gather

relevant information from prescribers and patients.

Finally, the studies in this thesis show that use of existing clinical and laboratory data is

155

clinical information is stored in automated databases, initiatives should be taken to stimulate

computerised recording of such information. In our study, we used data from a national nursing

home database (SIVIS) to obtain information on clinical diagnoses. Attention should be given

to accurate recording of SIVIS diagnoses, as we found several discrepancies between SIVIS

data and pharmacy data. For example, it is highly unlikely that an antidiabetic drug is prescri-

bed without a registered indication of diabetes mellitus. Data on INR were collected for inves-

tigating the clinical relevance of the potential pharmacokinetic interaction between acenocou-

marol and NSAIDs. Such data were not readily available in computerised databases. It has been

shown that computerised surveillance of adverse drug reactions (ARDs) is feasible in hospitals

[8,18]. Others have suggested that detecting adverse drug reactions would be more effective

and less time-consuming if electronic patient records were available [18]. Ideally, all medical

information of individual patients should be available on an electronic patient record.

Electronic patient records are not usually available in Dutch nursing homes but initiatives are

currently being developed to change this. In primary care, an electronic ‘care card’ is being

developed, on which all patient related medical information is stored. In this way, health care

professionals can gain insight into the patient information needed in order to carry out their

professional duties. This is particularly advantageous when patients are admitted from home

to hospital and or from nursing home to hospital and vice versa. In this way, optimal seamless

care can be provided. An issue that should be considered here, is patient privacy. Information

that is not needed for carrying out professional activities should be seperated from information

that is necessary. This can easily be done once the data is in electronic format. Another advan-

tage of the use of electronically available patient information, is the possibility for monitoring

patient outcomes on real-time basis as information on drug use and clinical outcome can sto-

red in the same database. In the meantime, different databases can be record-linked to provi-

de the opportunity of real-time monitoring for drug-related problems. Internationally, it has

already been shown that it is feasible to record-link patient-specific information that is stored

separately into large databases to monitor and evaluate effects of drug use in large populations

(MEMO [19], SAGE [20]).

A second advantage of combining existing clinical, pharmacy and laboratory records, is the

possibility to study outcomes of drug use in more detail and to establish reliable risk-benefit

analyses of drug use. Outcomes that could be studied are hospital admissions, adverse drug

effects and quality of life, although as discussed earlier, measuring the latter is difficult.

Relatively little is known about the relationship between drug use and clinically adverse out-

comes in nursing homes in the Netherlands. Internationally, many studies have focussed on

falls and fractures [6,21-25]. Psychotropic drugs, such as antidepressants and benzodiazepines,

154

References

1 Rackley CE. CME paper: elderly patients at risk for coronary heart disease or stroke: selecting an ideal product

for lipid lowering. Am J Geriatr Cardiol 2001; 10: 77-82.

2 Hunt D, Young P, Simes J, et al. Benefits of pravastatin on cardiovascular events and mortality in older patients

with cornonary heart disease are equal to or exceed those seen in younger patients.

Ann Intern Med 2001; 134: 931-40.

3 Recker RR, Davies KM, Dowd RM, Heaney RP. The effect of low-dose continuous estrogen and progesterone

therapy with calcium and vitamin D on bone in elderly women. A randomized, controlled trial.

Ann Intern Med 1999; 130: 897-904.

4 Field TS, Gurwitz JH, Glynn RJ, Salive ME, Gaziano JM, Taylor JO, Hennekes CH. The renal effects of nonsteroidal

anti-inflammatory drugs in older people: findings from the established populations for epidemiologic studies in

the elderly. J Am Geriatr Soc 1999; 47: 507-11.

5 Gray SL, Lai KV, Larson EB. Drug-induced cognition disorders in the elderly. Incidence, prevention and

management. Drug Saf 1999, 21: 101-22.

6 Thapa PB, Gideon P, Fought RL, Ray WA. Psychotropic drugs and risk of recurrent falls in ambulatory nursing

home residents. Am J Epidemiol 1995; 142: 202-11.

7 Stockley IH. Drug Interactions. 2nd Ed. Oxford: Blackwell Scientific Publications, 1991.

8 Azaz-Livshits T, Levy M, Sadan B, Shalit M, Geisslinger G, Brune K. Computerized surveillance of adverse drug

reactions in hospital: pilot study. Br J Clin Pharmacol 1998; 45: 309-14.


measures. Drugs & Aging 2000; 16: 437-50.

10 Van Dijk KN, Pont LG, Franken M, De Vries CS, Brouwers JRBJ, De Jong-Van den Berg LTW. Prescribing indicators

as a tool to evaluate drug use in nursing homes: a pilot study. Submitted



12 Zahn C, Sangl J, Bierman AS, Miller MR, Friedman B, Wickizer SW, Meyer GS. Potentially inappropriate

medication use in the community-dwelling elderly. JAMA 2001; 286: 2823-9.

13 Knight EL, Avorn J. Quality indicators for appropriate medication use in vulnerable elders. Ann Intern Med 2001;

135: 703-10.

14 Avorn J. Improving drug use in elderly patients: getting to the next level. JAMA 2001; 286: 2866-8.

15 Ensom MH, Chang TK, Patel P. Pharmacogenetics: the therapeutic drug monitoring of the future?

Clin Pharmacokinet 2001; 40: 783-802.

16 Fowler RW, Congdon P, Hamilton S. Assessing health status and outcomes in a geriatric day hospital.

Public Health 2000; 114: 440-5.

17 Ray WA, Stein CM, Byrd V, Shorr R, Pichert JW, Gideon P, Arnold K, Brandt KD, Pincus T, Griffin MR. Educational

program for physicians to reduce use of non-steroidal anti-inflammatory drugs among community-dwelling

elderly persons: a randomized controlled trial. Med Care 2001; 39: 425-35.

18 Emerson A, Martin RM, Tomlin M, Mann RD. Prospective cohort study of adverse events monitored by hospital

pharmacists. Pharmacoepidemioly and Drug Safety 2001; 10: 95-103.

19 Evans JJ, MacDonald TM. Record-linkage for pharmacovigilance in Scotland. Br J Clin Pharmacol 1999; 47: 105-10.

20 Gambassi G, Landi F, Peng L, Bostrup-Jensen C, Calore K, Hiris J, Lipsitz L, Mor V, Bernabei R. Validity of diag-

nostic and drug data in standardised nursing home resident assessments: potential for geriatric pharmaco-

epidemiology. SAGE Study Group. Med Care 1998; 36: 167-79.

21 Mustard CA, Mayer T. Case-control study of exposure to medication and the risk of injurious falls requiring

hospitalization among nursing home residents. Am J Epidemiol 1997; 145: 738-45.

157

important to estimate the clinical relevance of drug-related problems, such as drug-drug inter-

actions. These can be monitored and studied both on an individual and on a population based

level. Co-operation with other health care organisations, such as Outpatient Thrombosis

Services and Clinical Chemistry Laboratories, can contribute to the optimal use of existing

experience and data.

156

Summary

This thesis focuses on drug use in the elderly, in particular those residing in Dutch nursing

homes: the frail elderly. These elderly are especially prone to drug-related problems because of

their age, frequently occurring co-morbidity and polypharmacy. Relatively little is known of

drug use and drug-related problems in these frail elderly. The studies described in this thesis

aim to increase the knowledge on drug use and drug-related problems in this population. The

thesis comprises four main parts.

In chapter 1, the introductory chapter, the scope and the objective of this thesis are descri-

bed and the problems of drug use in elderly people are outlined. Because of co-morbidity, redu-

ced homeostatic mechanisms and the prescription of several drugs simultaneously, elderly

people are at an increased risk of drug-related problems such as drug-drug interactions, drug-

disease interactions and adverse drug effects. Drugs may also inadvertently be withheld from

the elderly, sometimes as a result of underdiagnosing. In view of these considerations, prescri-

bers need a thorough understanding of the risks and benefits of drug therapy in the elderly,

especially in the frail elderly, most of whom will be residing in nursing homes. In the

Netherlands, relatively few epidemiological studies on drug use in nursing homes have been

carried out. The objective of this thesis is to provide insight in the extent, determinants and

characteristics of drug use, as well as the outcomes of drug use in frail elderly. Section 1.2 gives

an overview of the Dutch health care system for ambulatory and institutionalised elderly. The

Dutch nursing home is a healthcare institution for chronically ill persons in need of permanent

medical and paramedical attention and complex nursing care. The type of care can be charac-

terised as continuous, long-term, systematic and multidisciplinary. Recently it was concluded

that quality aspects should be more incorporated in medication distribution processes and in

pharmaceutical care activities in Dutch nursing homes. Hospital pharmacists play a role in drug

and therapeutics committees, the evaluation of prescribing practices on patient level, and deve-

lopment and implementation of drug formularies.

Chapter 2 describes several studies that investigated drug use in nursing home residents

and elderly outpatients. The first part describes the studies performed in nursing home resi-

dents. In section 2.1 an introduction is given to the field of drug utilisation studies and studies

that describe the quality of drug use in nursing homes. Many studies have investigated the

extent of drug use, whereas only few studies used longitudinal prescription data to evaluate

drug effects over time. The studies showed an average number of drugs prescribed per resident

159

22 Ray WA, Thapa PB, Gideon P. Benzodiazepines and the risk of falling in nursing home residents.

J Am Geriatr Soc 2000; 48: 682-5.

23 Thapa PB, Gideon P, Cost TW, Milam AB, Ray WA. Antidepressants and risk of falls among nursing home

residents. N Engl J Med 1998; 339: 875-82.

24 Yip YB, Cumming RG. The association between medications and falls in Australian nursing home residents.

Med J Aust 1994; 160: 14-8.

25 Wang PS, Bohn RL, Glynn RJ, Mogun H, Avorn J. Zolpidem use and hip fractures in older people.

J Am Geriatr Soc 2001; 49: 1685-90.

26 Monane M, Avorn J, Beers MH, Everitt DE. Anticholinergic drug use and bowel function in nursing home

patients. Arch Intern Med 1993; 153: 633-8.

27 De Vries CS, Van den Berg PB, Timmer JW, Reicher A, Blijleven W, Tromp ThFJ, De Jong-van den Berg LTW.

Prescription data as a tool in pharmacotherapy audit: (II) the development of an instrument.

Pharm World Sci 1999; 21: 85-90.

158

ber of different drugs (based on 5th level of Anatomical Therapeutic Chemical (ATC) code) per

resident was 8.9 (SD 4.9). Duration of drug use was relatively long: eight of the ten therapeutic

drug groups prescribed most frequently were used for more than 50% of the time spent in the

nursing home. In particular psycholeptic drugs, diuretics, and laxatives were used chronically

(83%, 81% and 80% of the nursing home stay, respectively). Except for laxatives and diuretics,

the prescribed daily dosages were relatively low. We concluded that drug use in the nursing

homes was high and many drugs were used chronically. In view of possible adverse effects and

the risks of parallel prescribing and drug-drug interactions, the prescribing of psycholeptic

drugs, laxatives, loop diuretics, and ulcer-healing drugs should be re-evaluated. In section 2.4

a study is presented on drug-drug interactions (DDIs) in the nursing home. We developed pre-

scribing indicators based on the frequency, nature and duration of DDIs to systematically

assess potential DDIs in the cohort of nursing home patients. We found 32% of all residents

were exposed to at least one clinically relevant DDI. The number of medications prescribed was

a strong predictor of the occurrence of a potential DDI. Drug groups most frequently involved in

DDIs were oral anticoagulants, antibiotics and theophylline. The interaction between non-ste-

roidal anti-inflammatory drugs (NSAIDs) and loop diuretics, and between NSAIDs and oral

anticoagulants were the potential DDIs most frequently encountered. The number of days on

which drugs were prescribed concomitantly was relatively high. Nineteen out of 32 DDIs were

prescribed for an average of 50 days or more per 100 days of index drug use. The prescribing

indicators developed in this study provide the tools to audit DDI occurrence in nursing homes

systematically. Finally, in section 2.5 a pilot study is presented that used several prescribing

indicators, based on the studies in sections 2.3 and 2.4, to evaluate drug prescribing in Dutch

nursing homes. We evaluated prescribing of benzodiazepines, NSAIDs, ulcer-healing drugs and

diuretics. Prescribing indicators were used to identify prescribing that was potentially not in

line with recommendations in national and regional prescribing guidelines. Both descriptive

indicators, such as percentage of users, and indicators reflecting potentially suboptimal pre-

scribing, such as use of drugs outside the drug formulary and prescription of drug dosages

above recommended values were used. We found the majority of prescribing to be in line with

recommendations upon which we based our prescribing indicators. Clinical information from

the prescriber was necessary to get insight into actual prescribing appropriateness. The second

part of chapter 2 describes two drug utilisation studies that were performed in ambulatory

elderly, and partly in nursing home patients. Section 2.6 presents a study on the concomitant

use of benzodiazepines and antidepressants in two cohorts of elderly outpatients and the nur-

sing home cohort. We assessed whether differences in co-prescribing between tricyclic antide-

pressants (TCAs) and selective serotonine reuptake inhibitors (SSRIs) existed. Pharmacy dis-

161

ranging from 2.5 to 8.8. The studies that were carried out in the Netherlands involved relative-

ly small numbers of residents, and did not study overall drug use on individual patient level. In

particular in the United States, much attention has been given to rationality and appropriate-

ness of prescribing in nursing homes. Several studies have focused on applying tools, also

referred to as quality or prescribing indicators, to measure medication appropriateness. These

studies have shown that a considerable proportion of the nursing home residents received

inappropriate prescribing. However, prescribing indicators used in one health care system are

not automatically applicable to other health care systems due to differences in pharmacothe-

rapy guidelines and drug formularies. Section 2.2 describes how computerised medication

order data were used to build a nursing home database with the aim to perform drug utilisa-

tion studies. We collected medication order data of all nursing home residents from 6 nursing

homes in Friesland, the Netherlands, for a 2-year study period between 01-10-1993 and 01-10-

1995. These records were subsequently record-linked with a national information system on

nursing home residents (SIVIS). The SIVIS database contains information on medical (such as

diagnoses), nursing (such as activities of daily living and mobility) and administrative data col-

lected on individual nursing home residents. The source population consisted of 2,966 patients.

As a result of the record-linkage with SIVIS, missing data and age-limits (residents aged < 65

years were excluded), the final study population consisted of 2,355 residents. We have made

several recommendations for those who want to collect medication order data from nursing

home residents to perform pharmacoepidemiological studies. For example, an adequate samp-

le size is necessary, and data confidentiality must be guaranteed. Data should be as accurate

and complete as possible, which can be ensured by adequate data entry in hospital pharmacy

computer systems and checks against other sources, e.g. SIVIS data. Furthermore, it is impor-

tant that the data can be collected on a continuous basis, as longitudinal data are required to

study drug use over time. Keeping individual medication histories for several years is therefo-

re a prerequisite. Another important aspect is the registration of admission and discharge dates

in (pharmacy) computer systems so actual duration of stay in the nursing home can be calcu-

lated and provide person time as the denominator when studying person time exposed to

drugs. We found little agreement between SIVIS diagnoses data and pharmacy prescription

data for both diabetes mellitus and Parkinson’s disease, indicating that both pharmacy data and

SIVIS should be verified against each other in order to get the right estimation of disease pre-

valence in the nursing home population. In section 2.3 we performed a drug utilisation study

among 2,355 nursing home residents. During the two-year study period, 89%, 77% and 56%

of the study population used a drug from ATC main group N (central nervous system), A (ali-

mentary tract and metabolism), and C (cardiovascular system), respectively. The average num-

160

drugs that exhibit moderately to strongly constipating effects and occurrence of constipation

was found, the risk was not as high as seen in previous studies. In section 3.2 the clinical effect

of a DDI was investigated. In a cohort of elderly outpatients attending the Groningen

Outpatient Thrombosis Service, we studied the effects of the interaction between three NSAIDs

(diclofenac, naproxen and ibuprofen) and the oral anticoagulant acenocoumarol on prothrom-

bin time, expressed as the International Normalised Ratio (INR). Genotyping of cytochrome

P450 2C9 was performed to determine whether genotype was a predictive variable for the

occurrence of an increased INR as a result of this DDI. The study population consisted of 112

patients stable on acenocoumarol therapy, of whom 52 (46%) showed an elevation of the INR

above the desired therapeutic level (average INR increase between 1 and 4 units). In 12

patients, the INR increased above 6, indicating a clinically relevant risk of severe haemorrha-

ge. No association between CYP2C9 genotype and an increased INR as a result of the DDI was

found, and no other predictive variables were identified. We recommend close monitoring of

the INR of all patients receiving NSAIDs and acenocoumarol as at present we cannot predict

who will and who will not be affected by this DDI.

In chapter 4 the results of the studies described in the thesis are placed in a broader per-

spective and suggestions for clinical practice and further study are given. For example, hospi-

tal pharmacists can play a leading role in the monitoring of drug-related problems in the frail

elderly both on an individual and a population based level.

163

pensing data from the InterAction database were used for the study among ambulatory elder-

ly. We found that in two cohorts of elderly (one during 1994-1995 and one during 1998-1999) the

risk of initiating benzodiazepine drug therapy during antidepressant therapy was higher for

SSRI users than for TCA users (overall incidence RR 1.6; CI95 1.3-2.0). This could be due to the

fact that the less sedative effects of SSRIs may contribute to the increased frequency of benzo-

diazepine prescribing. In the nursing home cohort, no difference in frequency of benzodiazepi-

ne co-prescribing was found between SSRI users compared with TCA users. Partly this may be

due to the fact that hypnotic drug use in this population was already high, as was shown in sec-

tion 2.3. The prevalence of concomitant prescribing was considerable: in both ambulatory and

institutionalised elderly more than 50% of TCA and SSRI users were prescribed a benzodiaze-

pine drug concomitantly. On average, concomitant drug use lasted for greater than 67 days per

100 days of antidepressant drug use. The combined use of TCAs and benzodiazepines seems of

concern in view of the cumulation of adverse effects such as excess sedation and an increased

risk of falls. In section 2.7, we studied a potential beneficially combination of drugs: the conco-

mitant use of NSAIDs and gastroprotective drugs in a cohort of elderly outpatients. Use of

NSAIDs is associated with an increased risk of gastrointestinal toxicity, in particular when risk

factors such as advanced age are present. We studied the prevalence of concomitant prescri-

bing, as well as the prophylactic prescribing of gastroprotective drugs among ambulatory

NSAID users aged 65 years and over. Co-prescribing of gastroprotective drugs occurred in 23%

of the NSAID users (n=6,557), with an average duration of 67 days per 100 days of NSAID use.

Concomitant use of oral corticosteroids, coumarins and low dose aspirin were significantly

associated with both prophylactic and concomitant prescribing of gastroprotective agents

during NSAID therapy. We recommend giving feedback to prescribers to improve prescribing

practices in this high risk group.

In chapter 3, outcomes of drug use are studied in nursing home patients and elderly outpa-

tients. Section 3.1 presents a study among nursing home residents, in which the association

between drug use and constipation was investigated. We performed a prospective cohort study

of 2,355 nursing home patients to estimate the incidence relative risk of constipation associa-

ted with drug use using prescription sequence analysis of each resident’s detailed pharmacy

records and data on morbidity and mobility. Use of drugs that according to the summaries of

product characteristics and the literature on adverse effects have moderately to strongly con-

stipating properties was associated with a relative risk of 1.6 (CI95 1.2-2.0) for the occurrence of

constipation during exposure. Use of drugs with mildly to moderately constipating effects was

not associated with an increased frequency of laxative use. Although an association between

162

aandoeningen. De medische zorg wordt geleverd door speciaal hiervoor opgeleide verpleeg-

huisartsen. De geneesmiddelendistributie en farmaceutische zorg worden geleverd door open-

bare apothekers of ziekenhuisapothekers.

Het gebruik van geneesmiddelen bi j ouderen

Hoofdstuk 2 gaat nader in op het geneesmiddelgebruik bij ouderen. Het eerste deel

beschrijft het geneesmiddelgebruik bij verpleeghuisbewoners. In paragraaf 2.1 wordt een lite-

ratuuroverzicht gegeven van de studies die het geneesmiddelgebruik bij verpleeghuisbewo-

ners beschrijven. Deze studies zijn voornamelijk in de Verenigde Staten uitgevoerd. Aangezien

Amerikaanse ‘nursing homes’ nogal verschillen van de Nederlandse verpleeghuizen, zijn de

resultaten van deze studies niet zonder meer van toepassing op de Nederlandse situatie. In

Nederland is slechts weinig onderzoek uitgevoerd waarin het geneesmiddelgebruik op indivi-

dueel patiëntniveau wordt beschreven. Dit komt mede doordat het geneesmiddelgebruik pas

sinds een aantal jaren wordt vastgelegd in geautomatiseerde gegevensbestanden. Vóór de

jaren negentig werden de geneesmiddelgegevens vaak alleen in de medische status opge-

schreven, hetgeen het uitvoeren van farmacoepidemiologisch onderzoek, dat gegevens van

vele patiënten nodig heeft, bemoeilijkt. Paragraaf 2.2 beschrijft hoe de geneesmiddelgegevens

van alle verpleeghuisbewoners van zes verpleeghuizen zijn verzameld in één databestand. De

gegevens betroffen de periode 1-10-1993 tot 1-10-1995. Dit databestand is gekoppeld aan een

landelijk verpleeghuis-databestand (SIVIS), waarmee gegevens over de diagnosen, de aard

van de verpleging (somatisch of psychogeriatrisch) en de mobiliteit van de verpleeghuispa-

tiënten werden verkregen. De uiteindelijke studiepopulatie omvatte 2355 verpleeghuisbewo-

ners. In deze paragraaf worden aanbevelingen gedaan voor het uitvoeren van farmacoepide-

miologisch onderzoek met behulp van apotheekgegevens. Van belang is dat alleen geanonimi-

seerde gegevens verzameld worden, dat de gegevens een redelijke tijdsperiode (minimaal 2

jaar) beslaan, en dat de gegevens zo volledig mogelijk zijn (bijvoorbeeld een juiste registratie

van opname- en ontslagdata van verpleeghuisbewoners). Met het databestand is een genees-

middelgebruiksstudie uitgevoerd (paragraaf 2.3). Het geneesmiddelgebruik bij de verpleeg-

huispatiënten bleek hoog te zijn. Gemiddeld gebruikte iedere bewoner 4,9 verschillende

geneesmiddelen per dag. Geneesmiddelen die werken op het centrale zenuwstelsel (zoals mid-

delen tegen psychosen, slaap- en kalmeringsmiddelen), laxeermiddelen, pijnstillende midde-

len en middelen ter voorkoming van bloedstolling werden het meest frequent gebruikt. Veel

geneesmiddelen werden langdurig gebruikt, met name psychofarmaca, diuretica en laxantia:

deze werden gedurende meer dan driekwart van de verblijfsduur in het verpleeghuis gebruikt.

165

Samenvatt ing

Dit proefschrift beschrijft het gebruik van geneesmiddelen bij ouderen, in het bijzonder

verpleeghuisbewoners. Vanwege hun leeftijd zijn ouderen gevoeliger voor de effecten en bij-

werkingen van geneesmiddelen. Verpleeghuisbewoners lijden vaak gelijktijdig aan verschil-

lende aandoeningen met als gevolg het gelijktijdig gebruik van verschillende geneesmiddelen

(polyfarmacie). Er is weinig bekend over het gebruik van geneesmiddelen en het optreden van

problemen als gevolg van geneesmiddelgebruik bij deze kwetsbare groep ouderen. De onder-

zoeken die in dit proefschrift zijn beschreven beogen de kennis over het geneesmiddelgebruik

en geneesmiddelgerelateerde problemen bij ouderen te vergroten.

In le id ing

In het eerste hoofdstuk worden de achtergrond en de doelstelling van het proefschrift

beschreven. Oudere patiënten lopen meer risico op het krijgen van geneesmiddelgerelateerde

problemen. Voorbeelden zijn interacties tussen geneesmiddelen, waardoor de werking van een

geneesmiddel verzwakt of juist versterkt wordt. Ook treden bijwerkingen van geneesmiddelen

vaker op bij ouderen, bijvoorbeeld maagklachten als gevolg van bepaalde pijnstillers

(NSAID’s). Ook het ten onrechte niet aan oudere patiënten voorschrijven van geneesmiddelen,

bijvoorbeeld wanneer een bepaalde aandoening (zoals een depressie) niet wordt herkend, kan

voorkomen. Men spreekt dan van onderbehandeling. Om deze redenen is het van belang dat de

voorschrijvend arts een goed beeld heeft van de effecten en bijwerkingen van geneesmiddelen

bij ouderen, met name de kwetsbare ouderen die in verpleeghuizen verblijven. Door middel

van farmacoepidemiologisch onderzoek kan men inzicht krijgen in het gebruik van geneesmid-

delen en de factoren die van invloed zijn op dit gebruik, in bepaalde populaties in de dagelijk-

se praktijk. Ons onderzoek geeft inzicht in het geneesmiddelgebruik, de factoren die dit gebruik

bepalen (‘determinanten’) en de uitkomsten van geneesmiddelgebruik bij ouderen, in het bij-

zonder verpleeghuisbewoners. Paragraaf 1.2 beschrijft de zorg voor thuiswonende ouderen en

voor ouderen die wonen in verzorgings- en verpleeghuizen. Het verpleeghuis in Nederland

biedt verzorging en verpleging aan personen die chronisch ziek zijn en continue medische en

paramedische zorg nodig hebben. Het type zorg dat in Nederlandse verpleeghuizen wordt gele-

verd wordt gekenmerkt als continu, op de lange termijn gericht, systematisch en multidiscipli-

nair. Er is een onderscheid tussen de zorg voor somatische verpleeghuisbewoners, die lijden

aan ernstige lichamelijke aandoeningen (zoals de ziekte van Parkinson) en de zorg voor psy-

chogeriatrische verpleeghuisbewoners, die lijden aan dementie en andere psychogeriatrische

164

namelijk die tussen NSAID’s en maagbeschermende geneesmiddelen. NSAID’s zijn bekend

vanwege maagdarmklachten, zoals maagdarmzweren en maagdarmbloedingen. Met name

ouderen zijn hier gevoelig voor. De aanbeveling is dat mensen ouder dan 65 jaar die een NSAID

nodig hebben, tevens een geneesmiddel moeten gebruiken dat de maag beschermt. We onder-

zochten of dit ook gebeurt in de dagelijkse praktijk. Dit was slechts het geval bij 23% van de

NSAID gebruikers boven de 65 jaar. Verder onderzoek is nodig om te achterhalen waarom 77%

van de oudere NSAID gebruikers niet tegelijkertijd een maagbeschermer kreeg voorgeschreven

en of er bij die patiënten inderdaad klinisch relevante maagdarmklachten optreden.

Bi jwerkingen van geneesmiddelen in de prakt i jk

Hoofdstuk 3 richt zich op de uitkomsten van geneesmiddelgebruik bij ouderen. Paragraaf 3.1

beschrijft een onderzoek naar de relatie tussen het laxantiagebruik in verpleeghuizen en het

gebruik van obstiperende medicatie. In dit onderzoek werd een epidemiologische techniek,

prescriptie-sequentie-analyse, toegepast. Dit houdt in dat we bestudeerden of laxeermiddelen

vaker werden voorgeschreven aan gebruikers van obstiperende geneesmiddelen om de bij-

werkingen van deze geneesmiddelen (obstipatie) tegen te gaan. Verpleeghuisbewoners die

geneesmiddelen met een matig tot sterk obstiperende werking gebruikten, hadden 60% meer

risico om een laxans te gebruiken dan verpleeghuisbewoners die deze geneesmiddelen niet

gebruikten. Dit risico was minder hoog dan in de literatuur is gemeld. In paragraaf 3.2 onder-

zochten we de klinische relevantie van een geneesmiddelinteractie, namelijk die tussen

NSAID’s en acenocoumarol (een bloedverdunnend middel). Beide middelen worden frequent

door ouderen gebruikt. De studiepopulatie bestond uit 112 patiënten die geselecteerd waren via

de Stichting Trombosedienst Groningen. Bij 46% bleek de INR (International Normalised

Ratio, een maat voor de stolling van het bloed) verhoogd te zijn wanneer deze patiënten

behandeld waren met een NSAID. Dit betekent een verhoogde kans op bloedingen. Het ver-

moeden bestond dat een bepaalde erfelijke eigenschap, namelijk een mutatie op het enzym-

systeem cytochroom P450 2C9 (CYP2C9), verantwoordelijk zou kunnen zijn voor de verhoogde

INR als gevolg van de geneesmiddelinteractie. Daarom is bij 80 patiënten het genotype van

CYP2C9 bepaald. Er bleek geen relatie aantoonbaar te zijn tussen de verhoogde INR (als gevolg

van de geneesmiddelinteractie) en het CYP2C9 genotype. Dit betekent dat op dit moment (nog)

niet voorspeld kan worden bij wie de INR stijgt door deze geneesmiddelinteractie en bij wie

niet. Deze studie laat zien dat het nuttig is om klinische gegevens (zoals de INR) en voor-

schrijfgegevens te gebruiken om mogelijk bijwerkingen van geneesmiddelen te kunnen detec-

teren. Daarnaast is duidelijk dat het lastig is om te voorspellen bij wie de interactie zal optre-

den en bij wie niet.

167

De doseringen die werden voorgeschreven waren relatief laag, met uitzondering van diuretica

en laxantia. In paragraaf 2.4 gaan we in op het gelijktijdig voorschrijven van geneesmiddelen

waarvan gelijktijdig gebruik juist wordt afgeraden, hetgeen tot een klinisch relevante bijwer-

king kan leiden. We hebben de frequentie, de aard en de duur van een aantal potentieel

schadelijke geneesmiddelcombinaties in hetzelfde cohort verpleeghuispatiënten onderzocht.

Hier bleek dat 32% van de populatie tenminste één ongewenste geneesmiddelcombinatie

kreeg voorgeschreven gedurende de studieperiode van 2 jaar. De interactie tussen lisdiuretica

en NSAID’s en de interactie tussen orale anticoagulantia (bloedverdunnende middelen) en

NSAID’s kwamen in deze populatie het meest frequent voor. Ook bleek dat per interactie niet

meer dan 10% van de bewoners waren blootgesteld. Paragraaf 2.5 beschrijft een pilot onder-

zoek waarin het geneesmiddelgebruik in twee verpleeghuizen is geëvalueerd door middel van

het toepassen van voorschrijfindicatoren (‘prescribing indicators’). In dit onderzoek wordt de

mate waarin het voorschrijven van geneesmiddelen afwijkt van bepaalde richtlijnen in ver-

pleeghuizen, zoals een geneesmiddelformularium, bestudeerd. Hier bleek dat het voorschrijf-

gedrag weinig afweek van de richtlijnen. Als er werd afgeweken, bijvoorbeeld bij een te hoge

dosering, bestond daar een goede reden voor en werden eventuele bijwerkingen in de gaten

gehouden. De voorschrijfindicatoren bleken goed bruikbaar om het voorschrijfgedrag in kaart

te brengen. Om vanuit voorschrijfgegevens suboptimaal voorschrijven te kunnen detecteren,

was vaak klinische informatie over de patiënt nodig. Het tweede deel bestaat uit 2 onderzoe-

ken die zijn uitgevoerd door middel van analyse van voorschrijfgegevens van ambulante oude-

ren. Hiervoor werd gebruik gemaakt van de InterActie-databank, een databestand waarin de

geneesmiddelgegevens van circa 135.000 mensen in de regio Noordoost Nederland geanoni-

miseerd worden vastgelegd. Paragraaf 2.6 beschrijft het gebruik van benzodiazepinen (slaap-

en kalmeringsmiddelen) bij ouderen boven de 65 jaar en verpleeghuisbewoners die antide-

pressiva gebruikten. Het bleek dat de ambulante ouderen die nieuwere antidepressiva gebruik-

ten (selectieve serotonine heropname remmers (SSRI’s)), een grotere kans hadden om met een

benzodiazepine te starten dan personen die klassieke antidepressiva (tricyclische antidepres-

siva (TCA’s)) gebruikten. Mogelijk komt dit doordat TCA’s een groter kalmerend effect hebben,

waar met name ouderen gevoelig voor zijn. Bij de verpleeghuispatiënten kon een dergelijk

verhoogde kans niet worden aangetoond, wellicht doordat deze mensen al veel benzodiazepi-

nen gebruikten. Opvallend was dat het gecombineerd gebruik van antidepressiva en benzodia-

zepinen groot was: meer dan 50% van de TCA en SSRI gebruikers kreeg tevens een benzodia-

zepine voorgeschreven. Het gelijktijdig gebruik bleek ook langdurig te zijn. Vervolgonderzoek

zou inzicht moeten geven in de redenen waarom deze middelen zo lang gelijktijdig voorge-

schreven worden. In paragraaf 2.7 onderzochten we een wenselijke geneesmiddelcombinatie,

166

Dankwoord

Dat dit proefschrift er ligt is voor het overgrote deel te danken aan de inspirerende, motiveren-

de en ondersteunende bijdragen van velen.

Allereerst mijn beide promotores: Lolkje de Jong-van den Berg en Koos Brouwers.

Lolkje, dankzij jouw enthousiasme, steun en vertrouwen in dit onderzoek is dit proefschrift nu

geworden wat het is. Ik ben ontzettend blij dat ik met je heb mogen samenwerken. Altijd wist

je me weer te overtuigen van de relevantie van het onderzoek. Ik heb veel geleerd van de

manier waarop jij onderzoek doet en steeds het wetenschappelijk aspect van elke studie bena-

drukt. Daarbij was het natuurlijk ook erg gezellig! Dank je voor de tijd die je altijd voor me had.

Koos, met jou begon dit onderzoek toen je me in 1994 de mogelijkheid gaf om het laxantia-

gebruik bij verpleeghuisbewoners nader te onderzoeken. Jij was de ‘linking-pin’ met de uni-

versiteit en degene die me enthousiast maakte voor het onderzoek. Ik waardeer jouw klinisch-

farmacologische blik altijd zeer. Dank je voor je vertrouwen en je grote bijdrage aan dit onder-

zoek.

Zonder mijn referent, Corinne de Vries, had dit proefschrift zeker nog wat langer op zich

moeten laten wachten. Corinne, vanaf het eerste artikel tot en met de discussie: je was altijd

ontzettend betrokken bij dit onderzoek. Kortom, de ‘ideale referent’! Je vertrek naar Engeland

veranderde hier gelukkig niets aan: via de mail heb je talloze versies van artikelen gecorrigeerd

en aangevuld. Jouw epidemiologische inbreng heeft mede voor de publicaties gezorgd. Dank je

voor zoveel input!!

De leden van de leescommissie, prof. dr. A.C.G. Egberts, prof. dr. F.M. Haaijer-Ruskamp en

prof. dr. J.P.J. Slaets, dank ik voor hun snelle beoordeling van het manuscript en de waardevol-

le opmerkingen.

Paul van den Berg, mede-auteur van bijna alle hoofdstukken van dit boekje, was het brein

achter alle analyses. Paul, van jouw expertise op het gebied van databestanden heb ik dank-

baar gebruik gemaakt. Je wist tot ver in het SQL-tijdperk nog met mijn dBase-bestanden om te

gaan, gelukkig zonder al te veel morren. Dank je voor je geduld en je kritische opmerkingen.

Dick Bloemhof en Jan Sijtsma: ik ben jullie zeer erkentelijk voor de hulp met het verzame-

len van de voorschrijfgegevens uit de verpleeghuizen.

Margriet Piersma-Wichers, mede dankzij jouw enthousiasme is paragraaf 3.2 in dit proef-

schrift verschenen. Ik vond het erg leuk dat je betrokken was bij dit onderzoek.

Lisa Pont: gelukkig kwam er toch nog een artikel over de ‘prescribing indicators’. Dank voor

je inbreng hierbij en natuurlijk ook voor de Engelse correcties in enkele stukken.

169

Betekenis voor de prakt i jk

In hoofdstuk 4 worden de resultaten van de onderzoeken in dit proefschrift in breder per-

spectief geplaatst en worden suggesties voor de praktijk en voor verder onderzoek gegeven.

Bijwerkingen van geneesmiddelen kunnen bij deze kwetsbare ouderen grote gevolgen op de

kwaliteit van leven hebben, zoals de kans op verminderd geheugen door benzodiazepinen, de

kans op een delier (acute verwardheid) door bepaalde (anticholinerge) medicatie, en de kans

op vallen door psychofarmaca. Daarom dienen de voordelen van het gebruik van geneesmid-

delen altijd afgewogen te worden tegen de mogelijke bijwerkingen, met name bij ouderen. Om

te bepalen welke patiënten het hoogste risico lopen op geneesmiddelgerelateerde problemen,

is onderzoek zoals in dit proefschrift beschreven nodig. Naast voorschrijfgegevens zijn ook kli-

nische gegevens (zoals laboratoriumuitslagen en gegevens over de diagnosen van een patiënt)

nodig om inzicht te krijgen in de relevantie van geneesmiddelgerelateerde problemen. Een

voorbeeld is de interactie tussen NSAID’s en anticoagulantia: we weten dat deze interactie rela-

tief veel voorkomt en mogelijk tot problemen kan leiden. Op dit moment weten we nog niet hoe

we de personen die het meeste risico lopen op een klinische relevant effect, van te voren kun-

nen herkennen. Door het combineren van individuele medicatiegegevens en klinische data kan

op patiëntniveau het optreden van geneesmiddelgerelateerde problemen bewaakt worden.

Tevens kan door het combineren van deze gegevens op grote schaal farmaco-epidemiologisch

onderzoek worden uitgevoerd waarbij de nadruk ligt op het bestuderen van uitkomsten van

geneesmiddelgebruik (zoals een ziekenhuisopname als gevolg van een bijwerking van een

geneesmiddel). Hierbij kunnen verschillende disciplines, zoals ziekenhuisapothekers, ver-

pleeghuisartsen en klinisch chemici een belangrijke rol spelen. De rol van de ziekenhuisapo-

theker kan zowel initiërend als faciliterend zijn, zowel op het gebied van de verzameling van

voorschrijfgegevens als op het gebied van het in samenwerking met universitaire centra uit-

voeren van farmacoepidemiologisch onderzoek.

168

En dan als laatste, de onmisbare basis die gevormd wordt door familie en vrienden. Lieve mam

en pap: jullie dank ik voor jullie niet aflatende steun en vertrouwen. Paula, Wiebe en Veerle:

Haarlem was (en is) altijd een prima afleiding! Petra en Karen: ik ben heel blij dat jullie mijn

paranimfen willen zijn. Lieve Peter, dank, heel veel dank voor al je liefde en support en voor

zoveel meer dat eigenlijk niet in dit boekje thuishoort.

171

Dan waren er nog vele SFF-ers die mijn pad hebben gekruist tijdens mijn wekelijkse dag-

jes op de universiteit. Roel, het was leuk om met jou en Taco de farepi-cursus in Boston te vol-

gen. Taco, dank voor je hulp bij de voor mij soms ingewikkelde logistische regressie analyses.

Hilde, bij jou kon ik altijd terecht met een statistisch vraagje. Op de valreep heb ik nog veel van

je geleerd tijdens je statistiek cursus. Ada, Bert, Claudia, Eric, Evelyn, Jackie, Janet, Jasper,

Jeroen, Jos, Jos, Maarten, Marijke, Sipke, Willemijn, René, Rogier: dank voor jullie meeleven tij-

dens de vorderingen van dit onderzoek. De SFF-borrels, lunches en de jaarlijkse bbq waren

altijd erg leuk (dit geldt ook voor de ex-SFF-ers)! De leden van de dRUGs-werkgroep dank ik

voor hun interesse en feedback.

Een aantal enthousiaste bijvakstudenten heeft mij geholpen bij de onderzoeken die zijn

beschreven in dit boekje. Anne-Margreeth Dijkema, jij was de eerste ‘bijvakker’ en jij hebt zeker

je bijdrage geleverd aan paragraaf 3.1. Maria Franken, jij hebt een belangrijk onderdeel van

paragraaf 2.5 uitgevoerd. Het was plezierig met je samen te werken. Arian Plat: jij was de (snel-

le) motor achter het onderzoek bij de coumarine-gebruikers. Dankzij jouw inzet bij de recrute-

ring van de patiënten was het mogelijk om het onderzoek naar het polymorfisme van CYP2C9

uit te voeren. Veel succes met jouw promotieonderzoek! Alieke van Dijk was er vervolgens om

de genotypering van 80 patiënten uit te voeren. Alieke, dank je voor je nauwgezette werk!

Aukje Stenekes bepaalde de R- en S-acenocoumarolspiegels (helaas kon dat niet meer in het

boekje). Dank voor je hulp bij het transport van de bloedmonsters.

Dan waren er natuurlijk de collega-ziekenhuisapothekers, die zorgden voor een goede

basis op het werk. Toen ik in 1997 in de apotheek van het Wilhemina Ziekenhuis Assen voor 0.8

begon, wist ik nog niet hoeveel tijd die andere 0.2 me zou kosten. Uiteindelijk was het de moei-

te waard en dat kan denk ik alleen als de basis goed is. Jaap: je was altijd geïnteresseerd in het

onderzoek. Daarbij kwam het goed uit dat ik als aandachtsgebied de verpleeghuizen had. Dank

je voor de mogelijkheid om de zomercursus Epidemiologie in Boston te kunnen volgen en voor

je hulp bij de medisch-ethische toetsing van het coumarine onderzoek. Wobbe, Lous en Hans:

ook jullie dank ik voor de interesse die jullie altijd hadden in mijn onderzoek. Hans: veel suc-

ces met jouw onderzoek. Alle andere apotheekmedewerkers: ik kijk met veel plezier terug op

mijn WZA-tijd! Sinds een paar maanden weer terug in het Friese: mijn kersverse collega’s van

de apotheek Zorggroep Noorderbreedte hebben alleen het staartje van het onderzoek meege-

maakt. Het was een drukke tijd, met veel printjes en ritjes naar Groningen. Bob, dank je voor

het meedenken met het coumarine onderzoek. Eric, Folgert, Jan Peter, Nicole, Rients, Romke:

dank voor jullie flexibiliteit. Dit geldt natuurlijk ook voor alle andere apotheek ZNB-medewer-

kers.

170

Curr icu lum vi tae

Karen van Dijk werd op 23 oktober 1966 geboren in Den Haag. Aan het Rhedens Lyceum te

Rozendaal behaalde zij in 1985 haar VWO diploma. Aansluitend begon zij met de studie

Farmacie aan de Rijksuniversiteit Groningen. Het doctoraalexamen werd in 1991 gehaald,

waarna zij een maand Frans studeerde in Besançon. Het apothekersexamen werd in 1992

behaald. Na een korte tijd werkzaam geweest te zijn in de openbare farmacie in Den Haag

(Waldeck Apotheek), werd in 1993 gestart met de opleiding tot ziekenhuisapotheker in

Heerenveen en daarna Leeuwarden (opleiders: prof.dr. J.R.B.J. Brouwers en drs. R.J. Boskma).

Het registratieonderzoek leidde tot een promotieonderzoek aan de Rijksuniversiteit

Groningen, vakgroep Sociale Farmacie, Farmacoepidemiologie en Farmacotherapie (prof.dr.

L.T.W. de Jong-van den Berg, prof. dr. J.R.B.J. Brouwers en dr. C.S. de Vries), dat in de periode

1995-2001 werd uitgevoerd. Van april 1997 tot oktober 2001 was zij als ziekenhuisapotheker

werkzaam in het Wilhelmina Ziekenhuis te Assen. In 1999 volgde zij de zomercursus

Epidemiologie aan het Epidemiology Research Institute te Boston. Sinds november 2001 werkt

zij in het Medisch Centrum Leeuwarden.

173

Publ icat ions

Publ icat ions re lated to the thes is

Van Dijk KN, De Vries CS, Brouwers JRBJ, De Jong-van den Berg LTW. Farmaco-epidemiolo-

gisch onderzoek in verpleeghuizen: een prospectief vervolgonderzoek naar de associatie tus-

sen medicatie en obstipatie. Ziekenhuisfarmacie 1996; 4: 242-4.

Van Dijk KN, De Vries CS, Van den Berg PB, Dijkema AM, Brouwers JRBJ, De Jong- van den

Berg LTW. Constipation as an adverse effect of drug use in nursing home patients: an overesti-

mated risk. Br J Clin Pharmacol 1998; 46: 255-61.

Van Dijk KN, De Vries CS, Van den Berg PB, Brouwers JRBJ, De Jong-van den Berg LTW. Drug

utilisation in Dutch nursing homes. Eur J Clin Pharmacol 2000; 55: 765-71.

Van Dijk KN, De Vries CS, Van den Berg PB, Brouwers JRBJ, De Jong-van den Berg LTW.

Occurrence of potential drug-drug interactions in nursing home patients. Int J Pharm Pract

2001; 9:45-52.

Van Dijk KN, Ter Huurne K, De Vries CS, Van den Berg PB, Brouwers JRBJ, De Jong-van den

Berg LTW. Prescribing of gastroprotective drugs among elderly NSAID users in the Netherlands.

Pharm World Sci (in press).

Van Dijk KN, De Vries CS, Ter Huurne K, Van den Berg PB, Brouwers JRBJ, De Jong-van den

Berg LTW. Concomitant prescribing of benzodiazepines during antidepressant therapy in the

elderly. J Clin Epidemiol (in press).

Van Dijk KN, Pont LG, De Vries CS, Franken M, Brouwers JRBJ, De Jong-van den Berg LTW.

Prescribing indicators as a tool to evaluate drug use in nursing homes: a pilot study. Submitted.

Van Dijk KN, Plat AW, Van Dijk AAC, Piersma-Wichers G, De Vries-Bots AMB, Slomp J, De

Jong-van den Berg LTW, Brouwers JRBJ. Potential interaction between acenocoumarol and

diclofenac, naproxen and ibuprofen and the role of CYP2C9 genotype. Submitted.

Other publ icat ions

Bloemhof H, Van Dijk KN, De Graaf SSN, Vendrig DEMM, Uges DRA. Sensitive method for

the determination of vincristine in human serum by high-performance liquid chromatography

after on-line column extraction. J Chromatogr 1991; 572: 171-9, Biomedical Applications.

Van Dijk KN, Van der Meer YG. Medicamenteuze behandeling van diabetische neuropathie.

Pharm Weekbl 1992; 127: 1081-2.

Van Dijk KN. Miltefosine: lokaal cytostaticum. Ziekenhuisfarmacie 1996; 12: 106.

Lentelink MB, De Vries TW, Van Dijk KN. Accidental metronidazol overdose in a preterm

newborn [letter]. Clin Pharmacokin 1997; 32: 496-7.

172

Documents

University of Groningen Pharmacotherapy in frail elderly