COMPLIANCE WITH ANTIHYPERTENSIVE DRUG THERAPY *
D. W. Taylor, D. L. Sackett, R. B. Haynes, A. L. Johnson, E. S. Gibson t and R. S. Roberts
McMaster University, Health Sciences Centre Hamilton, Ontario, Canada
and t Dominion Foundries and Steel Company of Canada, Ltd.
Hamilton, Ontario, Canada
The treatment of hypertension shares with other long-term preventive regi- mens, a reliance on the patient for the day-to-day management of therapy. Unfortunately, poor patient compliance with drug regimens is well-documented by recent research. Also, clinicians who previously viewed patient compliance as beyond their area of responsibility and expertise are devoting more attention to compliance problems. With the current awakening of interest in this area has come the demand to provide clinicians and other health care professionals with methods of detecting the existence of compliance problems and with effective strategies for improving patient compliance. Before forging ahead with such efforts however, we must be assured that they are both medically and ethically justified. Four general criteria must be met before attempts are made to alter compliance with any treatment program. These are of immediate importance in the consideration of drug therapy for mild hypertension. First, the disease must represent an important source of morbidity and mortality; second, the prescribed therapy when fully complied with must have proven efficacy in reducing this toll of morbidity and mortality; third, the negative consequences in terms of adverse drug reactions, reduced quality of psychosocial functioning and cost must not outweigh the benefits of treatment; and fourth, poor compliance must be documented at levels that are insufficient for thera- peutic benefit. In the case of drug treatment for mild hypertension only the last of these criteria is well-established. Thus, the widespread implementation of compliance detection and intervention programs must await more definitive data on the three remaining criteria. With this provision in mind, this paper will discuss the measurement of patient compliance, document the magnitude of the compliance problem in hypertension, discuss its clinical significance, identify risk factors for noncompliance and describe compliance improving strategies that have been experimentally tested. Because of the scarcity of papers on compliance with antihypertensive drug regimens we will frequently refer to other compliance studies in order to illustrate important points. As well as reviewing what we know this paper will point out what we need to know and suggest research priorities for the study of compliance.
*Supported in part by the Medical Research Council of Canada, the Ontario Council of Health, the Ontario Heart Foundation and the National Health Grant of Health and Welfare Canada.
390 0077-8923/78/0304-0390 $01.75/0 @ 1978 New York Academy of Sciences
Taylor et al.: Compliance With Drug Therapy 391
Since our knowledge of any phenomenon is in large part determined by the reliability and validity of our methods of observation, all results reported in this paper must be considered in light of current compliance measurement tech- nology. Compliance with drug regimens consists not of a single event that can be easily observed but rather of a cumulation of many events of pill-taking occurring over time. Thus measurement techniques must be capable of sum- marizing pill-taking behavior over periods of time and must not be unduly influenced by single acts of taking or missing an individual dose. Also, if the measurement of compliance is to be clinically useful we should strive toward quick and easy methods that will allow clinicians to identify those patients in need of compliance intervention.
Blood Pressure Reduction
Although it seems reasonable that blood pressure reduction should itself be indicative of compliance this clinical outcome is inadequate as a compliance measure for several reasons. First, the extent to which compliance correlates with blood pressure fall will depend upon the dose of the drugs prescribed. It the dose is too low, even total compliance may not substantially reduce blood pressure and if too high therapeutic effect might be achieved with considerably less than total compliance. Second, if multiple medications are prescribed that vary in their hypotensive effect and if compliance differs across these medica- tions then compliance measured on only one of these drugs or average compli- ance computed across drugs may not coincide with the actual hypotensive effect achieved. And third, compliance with the prescribed regimen is only one factor among many that can influence a change in blood pressure. Changes in body weight and recent ingestion of alcohol or nicotine may also alter blood pressure.
Clinical Judgment and Patient Interviews
Although willing to hold reservations about clinical outcome as a measure of compliance, physicians may feel that they can accurately gauge a patient's compliance through their personal clinical skills. The research evidence does not support this belief.' Caron and Roth2 found that 27 ward residents could not judge patient compliance with antacid regimens at better than a chance level. Furthermore, 22 of these residents overestimated patient compliance. Evidence indicates that patients also tend to overestimate their compliance.' A useful observation in this area however, is that patients who admit to non- compliance are almost always telling the truth.3 Furthermore, these self- confessed noncompliers are more likely to benefit from compliance intervention strategies than are patients who deny their nonc~mpl iance .~~
While patient interviews are useful for identifying some noncompliers in the clinical setting more objective techniques are required both for research
392 Annals New York Academy of Sciences
studies and for clinical practice. Urinalysis techniques for drug and drug metabolites have recently been reported for hydrochlorothiazide,O, 7 alpha- methyldopa and propranolol.* The development of simple urinary or salivary dipstick techniques would be highly useful to clinicians. The development and use of such techniques will depend on the excretion patterns of the substances being tested and on variation among patients in excretion patterns. If excretion occurs too rapidly, the test will be highly influenced by its time relationship to ingestion of the drug and if excretion occurs very slowly, with accumulation of the substance over time, then even low compliance may result in a positive test.
In settings where the investigator has control over the dispensing of drugs, home visits to count remaining pills can be used to compute the percentage of prescribed pills which was consumed. The only assumption which must be made is that pills removed from pill bottles were in fact ingested by the patient and not taken by some other family member or simply discarded. Pill counts are more difficult when the investigator does not control dispensing of the drugs. In this situation he must either perform repeat visits to assess pill counts for at least two points in time or attempt to use pharmacy records of date and number of pills dispensed. Although some investigators have asked patients to bring their pills to clinic visits this request is not always complied with and also runs the risk that patients may empty their bottles beforehand in order to avoid being confronted with their lack of compliance. A simple pill-count procedure recom- mended by Mayer 9 for clinical practice is to prescribe pills per day in multiples of 7 to correspond to the number of weeks between clinic visits. Enquiry during clinic visits into whether the patient needs a new prescription or whether he still has pills left from the old one will help the physician identify noncompliers.
In addition to the problems noted with our current methods of measuring compliance, additional problems exist in the varying definitions and reporting practices of different investigators. The definition of what level of compliance should be used to separate patients with adequate from those with inadequate compliance has varied across studies. For clinical practice this level should be selected on the basis of the degree of compliance required to produce a thera- peutic effect. This will of course differ among drug regimens. Gordis lo reports that 33% compliance is adequate for the treatment of streptococcal infection, while Sackett l1 has identified 80% as the degree of compliance necessary with current antihypertensive regimens to produce consistent progress toward blood pressure control. While such dichotomous measures are desirable for clinical decision-making, the study of compliance behavior should employ continuous measures such as the percentage of pills consumed over a given period, and reporting should include whole compliance distributions. The development of reliable, valid, and simple methods of measuring compliance and identifying patients with inadequate levels, is an imperative for both clinical practice and the conduct of research into the natural history, determinants, and modification of patient compliance.
A true picture of the magnitude of noncompliance with antihypertensive drugs would require a longitudinal study of an inception cohort with multiple
Taylor et af.: Compliance With Drug Therapy 393
determinations of compliance at various points after the initiation of drug therapy. In the absence of such data we will review a number of studies of compliance with long-term regimens for asymptomatic illness with the following two reservations. First, studies that do not report on an inception cohort, or a group of subjects identified