Cronograma 2020 Farmacología II - Martes 17-21 hs€¦ · Cronograma 2020 Farmacología II - Martes 17-21 hs Laboratorio de Estadística Aplicada a las Ciencias de la Salud (LEACS)

Cronograma 2020 Farmacología II - Martes 17-21 hs

Laboratorio de Estadística Aplicada a las Ciencias de la Salud (LEACS)

Módulo 2: Neuropsicofarmacología

Fecha Tema Artículos Disertante

05/05

Enf. de Parkinson y otros trastornos

degenerativos

Tratamiento inicial Parkinson (JAMA)

Rodrigo Fernández Avello

12/05

Antipsicóticos

CATIE (NEJM) Augusto Feraris

19/05

Antidepresivos / Litio

Litio meta-análisis (Lancet) Nicolás Tentoni

26/05

Anticomiciales

VA-STATUS (NEJM) Federico D’Antonio

02/06 Benzodiacepinas BZD, compuestos Z y fractura de cadera

meta-análisis (PLOS) Santiago Tau

09/06

Antimigrañosos

Meta-análisis triptanos orales (Lancet)

Cintia Cruz

16/06

Lecturas abiertas obligatorias (LAZO) 1er cuatrimestre

ORIGINAL CONTRIBUTION

Pramipexole vs Levodopaas Initial Treatment for Parkinson DiseaseA Randomized Controlled TrialParkinson Study Group

PARKINSON DISEASE (PD) IS A PRO-gressively disabling neurode-generative disorder treated mostcommonly by dopamine re-

placement with the precursor le-vodopa, but agonists at dopamine-receptor sites have recently beenproposed as initial therapy in early stagesof the disease.1 The rationale for initialdopamine agonist treatment derivesfrom the belief that early levodopa ex-posure adversely affects the course of dis-ease and leads to disabling dyskinesiasand motor fluctuations (ie, dopaminer-gic complications).2 These dopaminer-gic complications are likely a conse-quence of underlying nigrostriataldegeneration, elicited by exposure to do-paminergic treatments, including le-vodopa and dopamine agonists.

Two reports have shown a reducedincidence of dyskinesias by initiatingtreatment with dopamine agonists ro-pinirole or cabergoline compared withlevodopa.3,4 The ropinirole trial foundno statistical differences in the occur-rences of adverse events or in changesin the activities of daily living (ADL)between the 2 groups, despite abso-lute differences in favor of levodopa,leaving the relative benefits of thesedrugs in question.

Pramipexole is a nonergoline dopa-minergic agonist5 that has been shownto be safe and effective compared with

Members of the Parkinson Study Group and Finan-cial Disclosure are listed at the end of this article.Corresponding Author and Reprints: Robert Holloway,

MD, MPH, Department of Neurology, University ofRochester, 1351 Mt Hope Ave, Suite 220, Rochester,NY 14620 (e-mail: [email protected]).

Context Pramipexole and levodopa both ameliorate the motor symptoms of earlyParkinson disease (PD), but no controlled studies have compared long-term out-comes after initiating dopaminergic therapy with pramipexole vs levodopa.

Objective To compare the development of dopaminergic motor complications af-ter initial treatment of early PD with pramipexole vs levodopa.

Design Multicenter, parallel-group, double-blind, randomized controlled trial.

Setting Academic movement disorders clinics at 22 sites in the United States andCanada.

Patients Three hundred one patients with early PD who required dopaminergic therapyto treat emerging disability, enrolled between October 1996 and August 1997.

Interventions Subjects were randomly assigned to receive pramipexole, 0.5 mg 3 timesper day, with levodopa placebo (n=151); or carbidopa/levodopa, 25/100 mg 3 timesper day, with pramipexole placebo (n=150). For patients with residual disability, the dos-age was escalated during the first 10 weeks. From week 11 to month 23.5, investigatorswere permitted to add open-label levodopa to treat continuing or emerging disability.

Main Outcome Measures Time to the first occurrence of any of 3 dopaminergiccomplications: wearing off, dyskinesias, or on-off motor fluctuations; changes in scoreson the Unified Parkinson’s Disease Rating Scale (UPDRS), assessed at baseline and fol-low-up evaluations; and, in a subgroup of 82 subjects evaluated at baseline and 23.5months, ratio of specific to nondisplaceable striatal iodine 1232-b-carboxymethoxy-3-b-(4-iodophenyl)tropane (b-CIT) uptake on single photon emission computed to-mography imaging of the dopamine transporter.

Results Initial pramipexole treatment resulted in significantly less development of wear-ing off, dyskinesias, or on-off motor fluctuations (28%) compared with levodopa (51%)(hazard ratio, 0.45; 95% confidence interval [CI], 0.30-0.66; P,.001). The mean im-provement in total UPDRS score from baseline to 23.5 months was greater in the le-vodopa group than in the pramipexole group (9.2 vs 4.5 points; P,.001). Somnolencewas more common in pramipexole-treated patients than in levodopa-treated patients(32.4% vs 17.3%; P=.003), and the difference was seen during the escalation phase oftreatment. In the subgroup study, patients treated initially with pramipexole (n=39) showeda mean (SD) decline of 20.0% (14.2%) in striatal b-CIT uptake compared with a 24.8%(14.4%) decline in subjects treated initially with levodopa (n=39; P=.15).

Conclusions Fewer patients receiving initial treatment for PD with pramipexole de-veloped dopaminergic motor complications than with levodopa therapy. Despite supple-mentation with open-label levodopa in both groups, the levodopa-treated group hada greater improvement in total UPDRS compared with the pramipexole group.JAMA. 2000;284:1931-1938 www.jama.com

See also p 1971 and Patient Page.

©2000 American Medical Association. All rights reserved. (Reprinted) JAMA, October 18, 2000—Vol 284, No. 15 1931

Downloaded From: http://jamanetwork.com/ on 12/09/2016

placebo in early PD,6,7 but has not yetbeen compared with levodopa. Accord-ingly, members of the Parkinson StudyGroup (PSG), an independent aca-demic consortium of investigators, con-ducted a multicenter randomized clini-cal trial comparing initial treatment ofearly PD with pramipexole vs levodopa.Givenrecent invitroandanimaldata thathave suggested that pramipexole may beneuroprotective for dopamine neurons,we also explored, in a subset of the sub-jects, the effects of these treatment strat-egies on dopamine transporter density,amarkerof thedopaminergicneuronter-minal, as measured by single photonemission computed tomography andiodine 123 [123I] 2-b-carboxymethoxy-3-b-(4-iodophenyl)tropane (b-CIT).8,9

METHODSA report detailing the methods of thistrial has been published.10

OrganizationThis multicenter study was organized bythe PSG in conjunction with the spon-sor, Pharmacia Corp (Peapack, NJ; for-merly Pharmacia & Upjohn Inc,Kalamazoo, Mich). Subjects were en-rolled in the trial between October 1996and August 1997 at 22 sites in the UnitedStates (17) and Canada (5). The studywas reviewed and approved by the in-stitutional review board at each of theparticipating sites, and all subjects gavewritten informed consent. An indepen-dent safety monitoring committee wasresponsible for unblinded monitoring ofdata concerning patient safety, with par-ticular attention to patient death, seri-ous adverse events, and adverse eventsresulting in subject withdrawal from thetrial. There were no prespecified for-mal guidelines for recommending modi-fication or termination of the trial, andany decision regarding early modifica-tion or termination would have beenbased on clinical judgment in light of theresults of significance tests.

Recruitment, Randomization,and EnrollmentEligible subjects were adults aged 30years or older who had idiopathic PD for

fewer than 7 years and who required do-paminergic antiparkinsonian therapy atthe time of enrollment. Patients who hadtaken levodopa or a dopaminergic ago-nist in the 2 months prior to enroll-ment were excluded. Subjects were re-quired to be in Hoehn and Yahr stage I,II, or III, a scale that classifies PD into5 clinical stages ranging from mild uni-lateral (stage I) to severe, bed-boundillness (stage V).11 Subjects were ex-cluded if they had (1) history of a pre-vious dopaminergic complication, (2)atypical parkinsonian syndromes, (3) se-rious concurrent illness, (4) treatmentwith methylphenidate, cinnarizine, re-serpine, amphetamine, or monoamineoxidase type A inhibitors in the past 3months, (5) treatment with pramipex-ole in the past 4 months, (6) treatmentwith neuroleptics, metoclopramide, al-phamethyldopa, or flunarizine in the past6 months, or (7) an unstable dosage ofselegiline, amantadine, anticholinergictherapy, or other central nervous sys-tem active therapies (eg, hypnotics, an-tidepressants, anxiolytics) in the past 2months.

Eligible patients were randomized 1:1to pramipexole or levodopa, in combi-nation with carbidopa, using a computer-generated randomization plan that in-cluded stratification by investigator andblocking. A programmer at the Pharma-cia Corp generated a list of the subjectidentification numbers and correspond-ing treatment assignments. The subjectidentification numbers were sent to thePSG Biostatistics Center (Rochester, NY)and incorporated in a computer inter-active randomization module at the PSGCoordination Center (Rochester, NY).Access to the randomization code was re-stricted to 2 programmers, 1 at the Phar-macia Corp and the other at the PSG Bio-statistics Center. When a patient wasjudged eligible and consented to be en-rolled, a telephone call was made to theCoordination Center, which provided aunique subject identification numberfrom the randomization module.

Study InterventionPramipexole was taken as 0.25-mg, 0.5-mg, or 1.0-mg tablets or matching pla-

cebo tablets, 3 times daily, which wereidentical in appearance, taste, and smell.Carbidopa/levodopa was taken as 12.5/50-mg or 25/100-mg capsules or match-ing placebo capsules 3 times a day. Treat-ment assignments included active drugfor one treatment and placebo for theother.

Subjects entered a 10-week dosage es-calation period followed by a 21-month maintenance period. All sub-jects were escalated initially to a dailydosage of 1.5-mg pramipexole or 75/300-mg carbidopa/levodopa (level 1dosage). Subjects requiring additionaltherapy could escalate to 3.0-mg pra-mipexole or 112.5/450-mg carbidopa/levodopa (level 2 dosage), or 4.5-mg pra-mipexole or 150/600-mg carbidopa/levodopa (level 3 dosage). Therefore, allpatients entered into the maintenancephase (week 11) of the trial on level 1,2, or 3 dosing. The pramipexole dos-ages were determined from a previousdosage-ranging tolerability study in pa-tients with early PD.6 Levodopa and pra-mipexole dosages were chosen as thosecommonly used in clinical practice andjudged to be near equivalent.

Throughout the maintenance pe-riod (week 11 through month 23.5),subjects maintained on study dosagelevel achieved in the escalation phase.Subjects with emerging disability wereprescribed open-label carbidopa/levodopa as needed.12 Sustained-release carbidopa/levodopa prepara-tions were not permitted.

Outcome VariablesSubjects were randomly assigned tothe intervention groups at the baselinevisit and were evaluated at 4 and 10weeks, and at 3, 6, 9, 12, 15, 18, 21,and 23.5 months. The primary out-come variable was prespecified as thetime from randomization until the firstoccurrence of any of 3 specified dopa-minergic complications: wearing off,dyskinesias, or on-off fluctuations.

Dyskinesias were defined as an abnor-mal involuntary movement that in-cludes chorea, dystonia, myoclonus, ortics that could be either peak dose or endof dose. Dyskinesias did not include early

PRAMIPEXOLE VS LEVODOPA FOR PARKINSON DISEASE

1932 JAMA, October 18, 2000—Vol 284, No. 15 (Reprinted) ©2000 American Medical Association. All rights reserved.


morning dystonia or other “off” dysto-nias. Wearing-off was defined as a per-ception of loss of mobility or dexterity,usually taking place gradually over min-utes and usually bearing close relation-ship to the timing of antiparkinsonianmedications. On-off effects were de-fined as an unpredictable and generallysudden (seconds to minutes) shift be-tween “on” (mobility) and “off” (immo-bility) not apparently related to the tim-ing of antiparkinsonian medications.10

One blinded investigator at each sitemade the judgment as to the occur-rence of a dopaminergic complication.Subjects reaching the primary end pointcontinued to be followed up through-out the 23.5 months of the trial.

Secondary outcome variables in-cluded changes in scores on the Uni-fied Parkinson’s Disease Rating Scale(UPDRS),13 the Parkinson’s DiseaseQuality of Life scale (PDQUALIF),14 theEuroQol,15 and the need for supple-mental carbidopa/levodopa. Measuresof safety included the frequency and se-verity of individual adverse experi-ences. The UPDRS is a standardized, re-liable, and valid instrument forassessing the severity of the clinical fea-tures of PD.16 The PDQUALIF andEuroQol are disease-specific and ge-neric quality-of-life instruments, re-spectively. The PDQUALIF consists of32 items and is scored on a 100-pointscale including 7 domains: social/rolefunction, self-image/sexuality, sleep,outlook, physical function, indepen-dence, and urinary function.

Single Photon Emission ComputedTomography and b-CIT SubstudyA subset of subjects (n=82) were en-rolled from 17 of the 22 participatingstudy sites to undergo single photonemission computed tomography imag-ing with [123I] b-CIT using methods re-ported previously.17 The imaging out-come measure was the ratio of specificto nondisplaceable striatal [123I] b-CITuptake. Subjects were imaged before thebaseline visit and just before the final23.5-month visit. All imaging evalua-tions took place at Yale University (NewHaven, Conn).

Sample SizeThe planned sample size of 300 sub-jects (150 per treatment group) was cho-sen to provide 94% power to detect a20% difference (70% vs 50%; hazard ra-tio [HR], 0.57) and 77% power to de-tect a 15% difference (70% vs 55%; HR,0.66) in the proportions of subjectsreaching the primary end point be-tween the treatment groups. The as-sumptions underlying these calcula-tions are detailed elsewhere.10

Statistical Analysis PlanThe primary statistical analyses were per-formed by intention-to-treat.18 All sta-tistical tests were 2-tailed and were per-formed using a significance level of 5%.The analysis of the primary outcomevariable used the Cox proportional haz-ards regression model, with treatmentgroup as the factor of interest and strati-fied by the enrolling investigator. TheHR and 95% confidence interval (CI)comparing the 2 treatment groups weredetermined from this model. The as-sumption of proportionality of hazardswas examined with the use of time-dependent covariates.19 Separate analy-ses of the time from baseline to the firstoccurrence of individual dopaminergiccomplications and the need for supple-mental levodopa were performed. Thecumulative probabilities of reaching theprimary outcome and other end pointswere estimated using Kaplan-Meiercurves.

Mean changes in the total UPDRSscore, as well as the mental, motor, andADL UPDRS scores, between random-ization and 23.5 months were com-pared among the treatment groups us-ing analysis of covariance, withtreatment group, enrolling investiga-tor, and the baseline UPDRS score in-cluded in the model. A 95% CI wascomputed for the difference betweenthe adjusted treatment group means.Changes in UPDRS scores betweenbaseline and the other visits were ana-lyzed similarly. These analyses also wereused to examine change scores in thequality-of-life measures. Interactionsbetween treatment and enrolling in-

vestigator were tested but not found.Two-tailed Fisher exact tests were usedto compare proportions of subjects ex-periencing adverse events between the2 groups. Changes in [123I] b-CIT up-take (striatum, caudate, and puta-men) were expressed as percentagechanges from baseline, and means werecompared between the 2 groups usingt tests.

For the analyses of continuous effi-cacy variables, if a subject was missinga response at a particular visit, the lastavailable observation for that subjectwas carried forward and imputed forthat visit. To determine the impact ofdropouts on the results, the analyseswere repeated including only subjectswho had complete data for the re-sponse variable of interest. The resultsof the latter analyses did not differ ma-terially from the analyses of the im-puted data and hence are not reportedhere.

RESULTSPatients Enrolled

Of the 376 patients who were identi-fied as potential participants, 52 werefound to be ineligible and 23 declinedfor no specific reason (FIGURE 1). Theremaining 301 patients were random-ized in the study, 82 of whom also en-rolled in the [123I] b-CIT substudy. Nopatients were lost to follow-up. Twenty-three subjects in the pramipexole group(15.1%) withdrew prior to the planned23.5 months of follow-up comparedwith 19 subjects in the levodopa group(12.7%). In the pramipexole group, 4withdrew due to somnolence and nau-sea/vomiting, and 3 due to hallucina-tions and edema. Six subjects in the le-vodopa group withdrew due to nausea.Two deaths occurred in the levodopa-treated group and neither were judgedto be related to the study drug.

The 2 treatment groups were simi-lar at baseline with regard to demo-graphic and clinical variables and thebaseline characteristics for the 82 sub-jects enrolled in the [123I] b-CIT sub-study cohort were similar to those forthe entire study cohort (TABLE 1).




Pramipexole and Levodopa UseThe numbers of subjects at each dos-age level were nearly identical in the 2treatment groups (Figure 1). Subjects al-located to pramipexole took an aver-age of 2.78 mg/d by the end of the trial.Subjects allocated to levodopa took anaverage of 406 mg/d of levodopa as ex-perimental therapy. Fifty-three per-cent of subjects in the pramipexolegroup required supplemental levodopacompared with 39% in the levodopagroup (HR, 1.54; 95% CI, 1.09-2.17;P=.02). At the end of the trial, subjectsin the pramipexole group who re-quired supplemental levodopa (n=80)were taking a mean (SD) of 264 (245)mg/d of supplemental levodopa com-pared with 252 (245) mg/d for subjectsin the levodopa group requiring supple-mentation (n=58). Subjects in the le-vodopa group thus took an average to-tal daily dosage of 509 mg of levodopa(experimental plus supplemental).

Dopaminergic End PointsTABLE 2 shows that 28% of subjects as-signed to pramipexole treatmentreached the primary end point by 23.5months compared with 51% in the le-

vodopa group (HR, 0.45; 95% CI, 0.30-0.66; P,.001). The reduced risk wasobserved in each of the four 6-monthstudy periods (0-6 month HR, 0.46;6-12 month HR, 0.27; 12-18 month HR,

Figure 1. Patient Flow Chart

Dosage 53 Received Level 1 68 Received Level 2 30 Received Level 3

Dosage 53 Received Level 1 65 Received Level 2 32 Received Level 3

23 Withdrew 4 Somnolence 4 Nausea/Vomiting 3 Hallucinations 3 Edema 2 Syncope 1 Worsening Disease 6 Other

19 Withdrew 6 Nausea 2 Worsening Disease 1 Somnolence 1 Dyskinesias 2 Death 7 Other

0 Lost to Follow-up

376 Patients Screened

0 Lost to Follow-up

151 Received Pramipexole

150 Received Levodopa

128 Completed Trial 131 Completed Trial

80 Received Supplemental Levodopa

75 Not Randomized 52 Not Eligible 23 Declined Participation

58 Received Supplemental Levodopa

301 Randomized

For explanation of dosage levels, see the “Study In-tervention” section.

Table 1. Baseline Characteristics*

Variable

Main Trial b-CIT Substudy

Pramipexole(n = 151)

Levodopa(n = 150)

Pramipexole(n = 42)

Levodopa(n = 40)

Age, y 61.5 (10.1) 60.9 (10.5) 61.9 (10.8) 60.1 (11.1)

No. (%) of male patients 96 (63.6) 99 (66.0) 27 (65.9) 24 (58.5)

No. (%) of white patients 144 (95.4) 143 (95.3) 39 (92.7) 38 (95.1)

Years since diagnosis 1.5 (1.4) 1.8 (1.7) 1.3 (1.4) 1.6 (1.9)

No. (%) of patients with priorlevodopa use

40 (26.5) 30 (20.0) 11 (26.2) 11 (27.5)

No. (%) of patients withselegiline use

50 (33.1) 56 (37.3) 19 (45.2) 14 (35.0)

No. (%) of patients withamantandine use

31 (20.5) 34 (22.7) 6 (14.3) 8 (20.0)

No. (%) of patients withanticholinergic use

7 (4.6) 12 (8.0) 2 (4.8) 2 (5.0)

Unified Parkinson’s DiseaseRating Scale

Mental 1.3 (1.3) 0.9 (1.1) 1.5 (1.6) 0.9 (1.1)

Activities of daily living 9.1 (4.1) 8.3 (4.0) 9.9 (4.2) 8.3 (4.0)

Motor 22.3 (9.2) 22.0 (9.6) 23.2 (9.7) 21.5 (8.8)

Total 32.5 (12.7) 31.1 (12.8) 34.6 (13.1) 30.6 (11.4)

No. (%) of patients in Hoehnand Yahr Stage

1.0 27 (17.9) 33 (22.0) 1 (2.4) 7 (17.1)

1.5 23 (15.2) 17 (11.3) 4 (9.8) 4 (9.8)

2.0 75 (49.7) 78 (52.0) 25 (58.5) 20 (48.8)

2.5 21 (13.9) 13 (8.7) 10 (24.4) 4 (9.8)

3.0 5 (3.3) 9 (6.0) 2 (4.9) 5 (12.2)

Mini-Mental StateExamination

29.2 (1.4) 29.3 (1.1) 29.3 (1.2) 29.2 (1.2)

Parkinson’s Disease Qualityof Life Scale

30.5 (10.7) 28.1 (10.4) 31.3 (10.3) 28.4 (9.2)

EuroQol visual analog scale 75.1 (15.6) 77.6 (12.0) 75.6 (15.2) 79.5 (11.5)

Striatal b-CIT uptake NA NA 3.8 (1.0) 3.6 (0.8)

Caudate b-CIT uptake NA NA 4.3 (1.1) 4.2 (1.0)

Putamen b-CIT uptake NA NA 3.3 (1.0) 3.1 (0.7)

*Values are expressed as mean (SD) unless otherwise indicated. CIT indicates carboxymethoxy-3-b-(4-iodophenyl)tropane; NA, not applicable. The scale ranges are as follows for the Mini-Mental State Examination, 0-30; Parkin-son’s Disease Quality of Life Scale, 0-100; and EuroQol visual analog scale, 0-100.

Table 2. Treatment Effects on Dopaminergic End Points*

End Points

No. (%)

HR (95% CI)† P ValuePramipexole

(n = 151)Levodopa(n = 150)

First dopaminergic complications‡ 42 (27.8) 76 (50.7) 0.45 (0.30-0.66) ,.001

Wearing off 36 (23.8) 57 (38.0) 0.57 (0.37-0.88) .01

Dyskinesias 15 (9.9) 46 (30.7) 0.33 (0.18-0.60) ,.001

On-off fluctuations 2 (1.3) 8 (5.3) 0.27 (0.06-1.32) .11

*All analyses are stratified by enrolling investigator.†HR indicates hazard ratio; CI, confidence interval. The HR is the ratio of the risk of reaching the end point per unit of

time for patients assigned to initially receive pramipexole treatment to the corresponding risk for patients assigned toinitially receive levodopa treatment.

‡Defined as first occurrence of wearing off, dyskinesia, or on-off fluctuations.




0.56; 18-24 month HR, 0.65) and forspecific dopaminergic complications ofwearing off and dyskinesias (Table 2;FIGURE 2).

TABLE 3 shows treatment effects ondopaminergic end points vs timing ofsupplemental levodopa. The absolutenumbers of end points were larger inthe levodopa group. Most of the endpoints occurred after the use of supple-mental levodopa, but in similar pro-portions between the treatment groups.

Of the 5 subjects taking pramipexolewho developed dyskinesias before thesupplemental levodopa, 4 had no priorlevodopa exposure.

Unified Parkinson’sDisease Rating ScaleThe mean improvement in total, mo-tor, and ADL UPDRS scores from base-line to 23.5 months was greater in thelevodopa group compared with the pra-mipexole group (TABLE 4). The le-vodopa group improved significantlyfrom baseline to each follow-up visit rela-tive to the pramipexole group (P#.002)in mean total, motor, and ADL UPDRSscores (FIGURE 3).

Quality-of-Life OutcomesQuality-of-life scores improved in bothgroups initially and then declined overtime (FIGURE 4). At 23.5 months (102weeks), the mean change scores weresignificantly different (P=.006) for the

PDQUALIF with the scores higher (ie,better) for those in the levodopa group.Mean change scores did not differamong the groups at other time points.Analyses of the 7 PDQUALIF sub-scales revealed significant differencesat 23.5 months for 2 subscales in favorof the levodopa group: sleep (P=.004)and self-image/sexuality (P=.02). Qual-ity-of-life scores on the EuroQol scaleshowed a similar divergence betweenthe 2 groups at the 23.5-month visit(P=.06; Figure 4).

Adverse EventsSignificantly more patients in the pra-mipexole group experienced somno-lence (P=.003), hallucinations (P=.03),and both generalized (P=.01) and pe-ripheral edema (P= .002) comparedwith those in the levodopa group(TABLE 5). Of note, the differences insomnolence and hallucinations be-tween the 2 groups emerged during the

Table 3. Treatment Effects on Dopaminergic End Points vs Timing of Supplemental Levodopa

End PointPramipexole

(n = 151)Levodopa(n = 150)

First dopaminergic complicationTotal receiving open-label levodopa 42 76

Before open-label levodopa 13 24

After open-label levodopa 29 52

Wearing offTotal receiving open-label levodopa 36 57



DyskinesiasTotal receiving open-label levodopa 15 45


After open-label levopoda 10 38

On-off fluctuationsTotal receiving open-label levodopa 2 7



Table 4. Mean Changes From Baseline to Month 23.5 in Unified Parkinson’s Disease RatingScale (UPDRS) Scores*

VariablePramipexole

(n = 151)Levodopa(n = 150)

Difference inTreatments (95% CI)† P Value

Total UPDRS 4.5 (12.7) 9.2 (10.8) −5.0 (−7.6 to −2.4) ,.001

Motor 3.4 (8.6) 7.3 (8.6) −3.9 (−5.7 to −2.1) ,.001

ADL 1.1 (4.5) 2.2 (3.2) −1.4 (−2.2 to −0.5) .001

Mental 0.0 (1.6) −0.2 (1.2) 0.1 (−0.2 to 0.3) .72

*Values are expressed as mean (SD). Positive values indicate improvement. ADL indicates activities of daily living.†Difference in treatment is the difference in mean change between the groups (pramipexole minus levodopa) and is

adjusted for investigator effects and the baseline value of the outcome variable in an analysis of covariance model.

Figure 2. Percentages of PatientsExperiencing Dopaminergic Complications

Per

cent

age

60

50

Levodopa

Levodopa

Levodopa

Patients Experiencing First Dopaminergic Complication

Pramipexole

Pramipexole

Pramipexole

40

30

20

10

0

Days From Randomization

Per

cent

age

60

50

40

30

20

10

0 100 200 300 400 500 600 700 800

Per

cent

age

60

50

40

30

20

10

0

A

Patients Experiencing Wearing OffB

Patients Experiencing DyskinesiasC

First dopaminergic complication is defined as the firstoccurrence of wearing off, dyskinesias, or on-offfluctuations.




escalation phase of the trial, whereas thedifferences for edema emerged duringthe maintenance phase of the trial.

Three subjects reported fallingasleep while driving, 2 of whom hadbeen randomized to pramipexole and1 to levodopa. None were takingopen-label levodopa. These eventsoccurred while the subjects werereceiving the level 2 dosing scheduleat 2, 5, and 12 months after random-ization. Two of these events resultedin motor vehicle crashes, one in a sub-ject randomized to levodopa and theother in a subject randomized to pra-mipexole. Two additional subjectscomplained of “abrupt” or “suddenonset” drowsiness unrelated to driv-

ing, both were allocated to pramipex-ole and receiving level 3 dosing.

Single Photon Emission ComputedTomography and b-CIT SubstudyIn the b-CIT substudy, 39 of the 40 sub-jects in the levodopa group and 39 ofthe 42 subjects in the pramipexolegroup had a follow-up b-CIT scan. Themean [123I] b-CIT uptake in the stria-tum, caudate, and putamen at base-line was well below the uptake valuesreported for healthy subjects (Table1).17 The mean (SD) decline in b-CIT

striatal uptake over the 23.5 months didnot iffer between the 2 treatment groupsand was 20.0% (14.2%) in the prami-pexole group compared with 24.8%(14.4%) in the levodopa group (P=.15;FIGURE 5). Caudate and putamen-specific b-CIT uptake during the 23.5-month observation period also did notdiffer between the 2 treatment groups.

COMMENTOur findings demonstrate that prami-pexole, as initial therapy in patients withearly PD, reduced the risk of develop-ing prespecified dopaminergic motorcomplications by 55% compared withinitiating therapy with levodopa overa 2-year period. The absolute risk re-duction of 23% suggests that one wouldneed to treat 4 to 5 patients with pra-mipexole instead of levodopa over a2-year period to prevent 1 additionaldopaminergic complication from oc-curring.

Both pramipexole and levodopa im-proved parkinsonian features, as mea-sured by the UPDRS, but pramipexolewas not as potent as levodopa in im-proving these features. The UPDRSscores remained worse in the prami-pexole group despite the use of open-label levodopa for treating emerging orcontinuing disability. Since the maxi-mum benefit was seen during the 10-week escalation phase, research sub-jects and investigators may havedeveloped a clinical sense of a satisfac-tory response and used this as a bench-mark for measuring the adequacy ofsubsequent treatment; that is, investi-gators added or adjusted supplemen-tal levodopa to maintain function ratherthan to improve it. The findings alsosuggest that although UPDRS scores arenot improved as much with pramipex-ole as with levodopa, the subjectstreated with pramipexole and theblinded investigators judged their ill-ness to be satisfactorily treated.

There are several potential explana-tions for why initial pramipexole treat-ment reduced the risk of developingwearing off and dyskinesias comparedwith initial levodopa treatment. First, thelonger half-life of pramipexole com-

Figure 3. Unified Parkinson’s Disease RatingScale

Mea

n

4540

Levodopa

Unified Parkinson's Disease Rating Scale Total

Unified Parkinson's Disease Rating Scale Motor

Unified Parkinson's Disease Rating Scale Activitiesof Daily Living

Pramipexole

35302520151050

Weeks0 10 20 30 40 50 60 70 80 90 100

A

B

C

Mea

n

454035302520151050

Mea

n

45403530252015105

Scores are expressed as mean (SE). In each case, a lowerscore indicates less severe features of the disease. Scoreswere calculated over the course of the trial by treat-ment assignment. Significant differences (P#.002) wereevident between the pramipexole and levodopa groupsat week 10 and at all subsequent visits.

Figure 4. Mean Changes in Quality-of-LifeScores Over the Course of the Trial

Mea

n C

hang

e Fr

om B

asel

ine

Mea

n C

hang

e Fr

om B

asel

ine

2Levodopa

Parkinson's Disease Quality of Life Scores

Pramipexole

1

0

A

EuroQoL Visual Analog Scale ScoresB

2

1

0

10 26 52 78 102–1

–1

Weeks

P = .06

P = .006

Quality-of-life scores improved by approximately 2units during the first 6 months of the trial. At the endof the trial (23.5 months), the group difference in themean change was statistically significant (P=.006) forthe Parkinson’s Disease Quality of Life Scale and mar-ginally significant for the EuroQol (P=.06), with thescores higher for those in the levodopa group. Dif-ferences in mean changes were not significant at anyother time points.




pared with levodopa (8 to 12 hours vs1.5 to 2 hours) may reduce the pulsa-tile stimulation of the striatal dopaminereceptors thought to be important in thedevelopment of dyskinesias and wear-ing off.20 Second, pramipexole and le-vodopa dosing may not have beenequivalent and the observed differencesmay in part be due to differences in do-paminergic potency between the 2groups. Finally, a neuroprotective ef-fect of pramipexole, if present, could alsoreduce the development of dopaminer-gic complications by preventing the lossof dopamine neurons, although a sig-nificant effect in [123I] b-CIT uptake wasnot seen in this study.

We did not detect significant differ-ences in the quality-of-life scores be-tween the 2 treatment groups duringthe first 78 weeks of the trial, indicat-ing an initial equal satisfaction withboth treatment options. We did detecta significant group difference in thePDQUALIF total score, however, at theend of the trial in favor of levodopa. Itis difficult to judge the clinical signifi-cance of this difference, which oc-curred primarily in the sleep subscaleof the instrument. A similar trend wasseen in the visual analog component ofthe EuroQol. Although generic and dis-ease-specific quality-of-life scales cor-relate with disease severity in patients

with PD,21 the responsiveness of thesequality-of-life scales to a clinicallymeaningful change in function andquality of life remains unclear.

Pramipexole use was associated witha greater likelihood of somnolence, hal-lucinations, and edema. The differ-ences in somnolence and hallucina-tions almost exclusively occurred whensubjects started experimental therapyduring the escalation phase of the trial.Risk factors for these adverse effects arenot known, but they may be of particu-lar concern for the elderly, those withpreexisting sleep disorders, or those re-ceiving multiple concomitant medica-tions.22

This study is the first, to our knowl-edge, to assess the rate of change instriatal [123I] b-CIT uptake in a rela-tively large cohort of patients with earlyPD over a 2-year period of observa-tion. The rate of decline in b-CIT up-take was similar to that found in smallerPD samples8 and was less in the pra-mipexole group than in the levodopagroup, but the group difference at 2years was not significant. We will con-tinue to follow up this cohort to ob-serve the course of neuroimaging out-comes in the next 2 years.

Our findings extend the observa-tions found in the randomized, double-blind trials that compared initial treat-

ment with ropinirole vs levodopa3 andcabergoline vs levodopa4 in early PD.The 268 patients in the ropinirole trialwere slightly older than our cohort (63vs 61 years), but had similar UPDRSscores at baseline and were followed upfor 5 years. The data in our trial aresimilar to the data found in the ropin-irole trial in terms of dopaminergic andUPDRS outcomes. However, in the ro-pinirole trial, only 2 comparisons werereported to be statistically significant:occurrences of dyskinesia and meanchanges in motor UPDRS scores. Theropinirole trial also showed similar pro-

Figure 5. Percentage Reduction in b-CITUptake From Baseline to 23.5 Months

% C

hang

e Fr

om B

asel

ine

to M

onth

23.

5 50

Levodopa

Striatum Caudate Putamen

Pramipexole

30

40

20

10

0

Error bars indicate SDs.

Table 5. Adverse Events by Treatment Group and Study Phase

Total Cohort, No. (%) Escalation Phase, No. (%) Maintenance Phase, No. (%)*


Levodopa(n = 150)


Levodopa(n = 150)


Levodopa(n = 144)

Somnolence 49 (32.4) 26 (17.3)† 35 (23.2) 13 (8.7)† 14 (9.9) 13 (9.0)

Hallucination 14 (9.3) 5 (3.3)‡ 10 (6.6) 2 (1.3)‡ 4 (2.8) 3 (2.1)

Generalized edema 27 (17.9) 12 (8.0)‡ 3 (2.0) 3 (2.0) 24 (16.9) 9 (6.3)†

Peripheral edema 22 (14.6) 6 (4.0)† 7 (4.6) 2 (1.3) 15 (10.6) 4 (2.8)†

Nausea 55 (36.4) 55 (36.7) 45 (29.8) 42 (28.0) 10 (7.0) 13 (9.0)

Dizziness 39 (25.8) 36 (24.0) 21 (13.9) 18 (12.0) 18 (12.6) 18 (12.5)

Insomnia 39 (25.8) 33 (22.0) 20 (13.3) 14 (9.3) 19 (13.4) 18 (12.5)

Headache 31 (20.5) 23 (15.3) 18 (11.9) 15 (10.0) 13 (9.1) 8 (5.6)

Constipation 31 (20.5) 19 (12.7) 24 (15.9) 6 (4.0)† 7 (4.9) 12 (8.3)

Depression 23 (15.2) 20 (13.3) 3 (2.0) 9 (6.0) 20 (14.0) 11 (7.6)

Abnormal dreams 21 (13.9) 19 (12.7) 3 (1.9) 1 (0.7) 2 (1.4) 1 (0.7)

Anxiety 17 (11.3) 10 (6.7) 7 (4.6) 3 (2.0) 10 (7.0) 7 (4.9)

Postural hypotension 9 (6.0) 15 (10) 4 (2.7) 7 (4.7) 5 (3.5) 8 (5.6)

*Refers only to patients entering maintenance.†P,.01 for comparison of pramipexole with levodopa.‡P,.05 for comparison of pramipexole with levodopa.




portional increases in the occurrenceof somnolence (27.4% vs 19.1%) andhallucinations (17.3% vs 5.6%) in theropinirole-treated subjects.

Our trial revealed significant groupdifferences in the occurrences of wear-ing off and dyskinesias in favor of pra-mipexole and in the occurrences ofsomnolence and hallucinations in fa-vor of levodopa. In addition, our trialrevealed significant group differencesin the mean change in the ADL com-ponent of the UPDRS. The smallersample size (n=268) and unbalancedallocation ratio (2:1) in the ropiniroletrial may have contributed to the dif-ferences in statistically significant re-sults seen between the 2 studies de-spite the similarities in the magnitudesof the group differences.

These studies leave several ques-tions unanswered. Does the trade off be-tween motor complications and effi-cacy as measured by the UPDRS favorlevodopa over agonists? What are theimplications of the increased rates ofsomnolence and hallucinations withagonist treatment? Further studyshould help address these questions.Until longer-term data are available, thedecision to initiate treatment of earlyPD with pramipexole or levodopashould be made only after consideringthe favorable dopaminergic motor com-

plication profile associated with pra-mipexole against the more potent an-tiparkinsonian effects associated withlevodopa.

Parkinson Study Group Investigators: Steering Com-mittee: Robert Holloway, MD, MPH, medical director,Ira Shoulson, MD, principal investigator, Karl Kieburtz,MD, MPH, Michael McDermott, PhD, chief biostatis-tician, Pierre Tariot, chair, safety monitoring commit-tee, Cornelia Kamp, MBA, project coordinator, DenniDay, RN, MSPH, administrative manager, Aileen Shina-man, JD, PSG executive director, University of Roches-ter, Rochester, NY; Stanley Fahn, MD, coprincipal in-vestigator, Columbia University, New York, NY; AnthonyLang, MD, Toronto Western Hospital, University HealthNetwork, Toronto, Ontario; Kenneth Marek, MD, prin-cipal investigator, John Seibyl, MD, coprincipal investi-gator for b-CIT, Yale University School of Medicine, NewHaven, Conn; William Weiner, MD, Mickie Welsh, RN,DNS, ex-officio, University of Southern California, LosAngeles.

Participating Investigators and Coordinators: RajeshPahwa, MD, Shantelle Coe, RN, University of KansasMedical Center, Kansas City; Lynn Barclay, MD, LauraSutherland, RN, Kathy Hildebrand, RN, Ottawa CivicHospital, Ottawa, Ontario; Jean Hubble, MD, CarolynWeeks, MT, Ohio State University, Columbus; PeterLeWitt, MD, Clinical Neuroscience Center, South-field, Mich; Janis Miyasaki, MD, Jan Duff, RN,Elspeth Sime, RN, Toronto Western Hospital, Univer-sity Health Network, Toronto, Ontario; OksanaSuchowersky, MD, University of Calgary Medical Clinic,Calgary, Alberta; Mark Stacy, MD, Matthias Kurth, MD,Melanie Brewer, RN, Mary Harrigan, RN, MN, Bar-row Neurological Institute, Phoenix, Ariz; David S. Rus-sell, MD, PhD, Barbara Fussell, RN, Yale UniversitySchool of Medicine, New Haven, Conn; Blair Ford, MD,Sandra Dillon, RN, Columbia University, New York,NY; JohnHammerstad,MD,ClaudiaStone,RA,OregonHealth Sciences University, Portland; David Riley, MD,Pamela Rainey, RN, University Hospitals of Cleve-land, Cleveland, Ohio; David Standaert, MD, MarshaTennis, RN, Massachusetts General Hospital, Boston;Frederick Wooten, MD, Elke Rost-Ruffner, RN, Uni-versity of Virginia Health Sciences Center, Char-lottesville; Stewart Factor, DO, Diane Brown, RN,

Albany Medical College, Albany, NY; Joseph Jank-ovic, MD, Farah Atassi, MPH, Baylor College of Medi-cine, Houston, Tex; Roger Kurlan, MD, Irenita Gardiner,RN, University of Rochester, Rochester, NY; MichelPanisset, MD, Donna Amyot, RN, Jean Hall, RN, McGillCentre for Studies in Aging, Douglas Hospital, Ver-dun, Quebec; Ali Rajput, MD, Theresa Shirley, RN,Saskatoon District Health Board, Royal University Hos-pital, Saskatoon, Saskatchewan; Robert Rodnitzky, MD,and Judith Dobson, RN, University of Iowa Hospitals,Iowa City; Cliff Shults, MD, Deborah Fontaine, RNC,University of California, San Diego, and the Alzhei-mer’s Research Center, La Jolla, Calif; Cheryl Waters,MD,MickieWelsh,RN,DNSc, SuzanneSchuman,RNC,MPA, University of Southern California, Los Angeles;Ronald Pfeiffer, MD, Sarah Rast, RN, Brenda Pfeiffer,RN, BSN, University of Tennessee, Memphis.

Biostatistics and Clinical Trials Coordination Cen-ter Staff: Alicia Brocht, BS, Cindy Casaceli, MBA,Susan Daigneault, Karen Hodgeman, Kathy Hons-inger, MS, Carolynn O’Connell, Arthur Watts, BS, Uni-versity of Rochester, Rochester, NY.Financial Disclosure: In keeping with the ParkinsonStudy Group conflict of interest guidelines, none of theinvestigators have any personal financial relationshipwith the sponsor. All compensation received by inves-tigators for trial-related services was paid through a con-tract between the University of Rochester and the spon-sor that was established before the trial began.Funding/Support: This work was supported primar-ily by the Pharmacia Corp. Support was also pro-vided by the National Parkinson Foundation Centerof Excellence to the Parkinson Study Group, and bythe National Institutes of Health for Clinical ResearchCenter grants RR00044 and RR01066 to the Univer-sity of Rochester and Massachusetts General Hospi-tal, respectively.Acknowledgment: We thank the patients and theirfamilies who participated in this study. We also thankthe Safety Monitoring Committee: W. Jackson Hall,PhD, Pierre Tariot, MD, chair, University of Roches-ter, Rochester, NY; Carl M. Leventhal, MD, Rock-ville, Md; Stephen Reich, MD, Johns Hopkins, Balti-more, Md. Contributions from the following individualsof the Pharmacia Corp are gratefully acknowledged:Leona Borchert MD, MPH, Mark Corrigan, MD, Balta-zar Gomez-Mancilla, MD, Bruno Musch MD, RhondaRagual, MS, Clayton Rowland, PhD, Gene Wright,PhD, Kalamazoo, Mich.

REFERENCES

1. Olanow CW, Koller WC. An algorithm (decision tree)for the management of Parkinson’s disease: treatmentguidelines. Neurology. 1998;50(suppl 3):S1-S57.2. Reardon K, Shiff M, Kempster PA. Evolution of mo-tor fluctuations in Parkinson’s disease: a longitudinalstudy over 6 years. Mov Disord. 1999;14:605-611.3. Rascol O, Brooks DJ, Korczyn AD, et al. A five-yearstudyof the incidenceofdyskinesia inpatientswithearlyParkinson’s disease who were treated with ropiniroleof levodopa. N Engl J Med. 2000;342:1484-1491.4. Rinne UK, Bracco F, Chouza C, et al. Early treatmentof Parkinson’s disease with cabergoline delays the onsetofmotorcomplications: resultsofadouble-blind levodopacontrolled trial. Drugs. 1998;55(suppl 1):23-30.5. Piercey MF, Camacho-Ochoa M, Smith MW. Func-tional roles for dopamine-receptor subtypes. Clin Neu-ropharmacol. 1995;18(suppl):S34-S42.6. Parkinson Study Group. Safety and efficacy of pra-mipexole in early parkinson disease: a randomizeddose-ranging study. JAMA. 1997;278:125-130.7. Hubble JP, Koller WC, Cutler NR, et al. Pramipex-ole in patients with early Parkinson’s disease. Clin Neu-ropharmacol. 1995;18:338-347.8. Marek KL. Dopaminergic dysfunctin in parkinson-ism: new lessons from imaging. Neuroscientist. 1999;5:333-339.

9. Zou L, Jankovic J, Rowe DB, Xie W, Appel SH, LeW. Neuroprotection by pramipexole against dopa-mine- and levodopa-induced cytotoxicity. Life Sci.1999;64:1275-1285.10. Parkinson Study Group. A randomized controlledtrial comparing pramipexole with levodopa in early Par-kinson’s disease: design and methods of the CALM-PDStudy. Clin Neuropharmacol. 2000;23:34-44.11. Hoehn MM, Yahr MD. Parkinsonism: onset, pro-gression and mortality. Neurology. 1967;17:427-442.12. Parkinson Study Group. DATATOP: a multi-center controlled clinical trial in early Parkinson’s dis-ease. Arch Neurol. 1989;46:1052-1060.13. Lang AE, Fahn S. Assessment of Parkinson’s dis-ease. In: Munsat TL, ed. Quantification of Neuro-logic Def ic i t . Boston, Mass : Butterworth-Heinemann; 1989:285-309.14. Welsh M, McDermott M, Holloway R, et al. De-velopment and testing of the Parkinson’s disease qual-ity of life scale: the PDQUALIF [abstract]. Mov Dis-ord. 1997;12:836.15. EuroQol Group. EuroQol: a new facility for themeasurement of health related quality of life. HealthPolicy. 1990;12:199-208.16. Richards M, Marder K, Cote L, Mayeux R. Inter-

rater reliability of the Unified Parkinson’s Disease Rat-ing Scale motor examination. Mov Disord. 1994;9:89-91.17. Seibyl JP, Marek KL, Quinlan D, et al. DecreasedSPECT [I-123] b-CIT striatal uptake correlates withsymptom severity in idiopathic Parkinson’s disease. AnnNeurol. 1995;38:589-598.18. Lee YJ, Ellenberg JH, Hirtz DG, Nelson KB.Analysis of clinical trials by treatment actuallyreceived: is it really an option? Stat Med. 1991;10:1595-1605.19. Kalbfleisch JD, Prentice RL. The Statistical Analy-sis of Failure Time Data. New York, NY: John Wiley& Sons Inc; 1980.20. Olanow CW, Obeso JA. Preventing levodopa-induced dyskinesias. Ann Neurol. 2000;47(suppl 1):S167-S178.21. Rubenstein LM, Voelker MD, Chrischilles EA,Glenn DC, Wallace RB, Rodnitzky RL. The usefulnessof the Functional Status Questionnaire and MedicalOutcomes Study Short Form in Parkinson’s dieseaseresearch. Qual Life Res. 2000;7:279-290.22. Frucht S, Rogers JD, Greene PE, Gordon MF, FahnS. Falling asleep at the wheel: motor mishaps in per-sons taking pramipexole and ropinirole. Neurology.1999;52:1908-1910.




n engl j med

353;12

www.nejm.org september

22, 2005

1209

The

new englandjournal

of

medicine

established in 1812

september

22

,

2005

vol. 353 no. 12

Effectiveness of Antipsychotic Drugs in Patients with Chronic Schizophrenia

Jeffrey A. Lieberman, M.D., T. Scott Stroup, M.D., M.P.H., Joseph P. McEvoy, M.D., Marvin S. Swartz, M.D., Robert A. Rosenheck, M.D., Diana O. Perkins, M.D., M.P.H., Richard S.E. Keefe, Ph.D.,

Sonia M. Davis, Dr.P.H., Clarence E. Davis, Ph.D., Barry D. Lebowitz, Ph.D., Joanne Severe, M.S., and John K. Hsiao, M.D., for the Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE) Investigators*

abstract

From the Department of Psychiatry, Collegeof Physicians and Surgeons, Columbia Uni-versity, New York State Psychiatric Institute,New York (J.A.L.); the Department of Psy-chiatry, School of Medicine (T.S.S., D.O.P.),and the Department of Biostatistics, Schoolof Public Health (S.M.D., C.E.D.), Universi-ty of North Carolina at Chapel Hill, ChapelHill; Quintiles, Research Triangle Park, N.C.(S.M.D.); the Department of Biological Psy-chiatry, John Umstead Hospital, Butner,N.C. (J.P.M.); the Department of Psychia-try and Behavioral Sciences, Duke Univer-sity Medical Center, Durham, N.C. (J.P.M.,M.S.S., R.S.E.K.); the Department of Psy-chiatry, Yale University School of Medicine,New Haven, Conn. (R.A.R.); and the Divi-sion of Services and Intervention Research,National Institute of Mental Health, Na-tional Institutes of Health, Bethesda, Md.(B.D.L., J.S., J.K.H.). Address reprint requeststo Dr. Lieberman at the Department ofPsychiatry, College of Physicians and Sur-geons, Columbia University, New York StatePsychiatric Institute, 1051 Riverside Dr.,New York, NY 10032, or at [email protected].

*The CATIE investigators are listed in theAppendix.

N Engl J Med 2005;353:1209-23.

Copyright © 2005 Massachusetts Medical Society.

background

The relative effectiveness of second-generation (atypical) antipsychotic drugs as com-pared with that of older agents has been incompletely addressed, though newer agentsare currently used far more commonly. We compared a first-generation antipsychotic,perphenazine, with several newer drugs in a double-blind study.

methods

A total of 1493 patients with schizophrenia were recruited at 57 U.S. sites and random-ly assigned to receive olanzapine (7.5 to 30 mg per day), perphenazine (8 to 32 mg perday), quetiapine (200 to 800 mg per day), or risperidone (1.5 to 6.0 mg per day) for upto 18 months. Ziprasidone (40 to 160 mg per day) was included after its approval by theFood and Drug Administration. The primary aim was to delineate differences in theoverall effectiveness of these five treatments.

results

Overall, 74 percent of patients discontinued the study medication before 18 months(1061 of the 1432 patients who received at least one dose): 64 percent of those assignedto olanzapine, 75 percent of those assigned to perphenazine, 82 percent of those as-signed to quetiapine, 74 percent of those assigned to risperidone, and 79 percent ofthose assigned to ziprasidone. The time to the discontinuation of treatment for anycause was significantly longer in the olanzapine group than in the quetiapine (P<0.001)or risperidone (P=0.002) group, but not in the perphenazine (P=0.021) or ziprasidone(P=0.028) group. The times to discontinuation because of intolerable side effects weresimilar among the groups, but the rates differed (P=0.04); olanzapine was associatedwith more discontinuation for weight gain or metabolic effects, and perphenazinewas associated with more discontinuation for extrapyramidal effects.

conclusions

The majority of patients in each group discontinued their assigned treatment owing toinefficacy or intolerable side effects or for other reasons. Olanzapine was the most ef-fective in terms of the rates of discontinuation, and the efficacy of the conventional anti-psychotic agent perphenazine appeared similar to that of quetiapine, risperidone, andziprasidone. Olanzapine was associated with greater weight gain and increases in mea-sures of glucose and lipid metabolism.

The New England Journal of Medicine Downloaded from nejm.org at UNIVERSITY OF ILLINOIS on March 20, 2013. For personal use only. No other uses without permission.

Copyright © 2005 Massachusetts Medical Society. All rights reserved.

n engl j med

353;12


22

,

2005

The

new england journal

of

medicine

1210

ntipsychotic drugs have become

the cornerstone of treatment for schizo-phrenia. The first-generation “conven-

tional” antipsychotic drugs are high-affinity an-tagonists of dopamine D2 receptors that are mosteffective against psychotic symptoms but have highrates of neurologic side effects, such as extrapyrami-dal signs and tardive dyskinesia.

1

The introductionof second-generation, or “atypical,” antipsychoticdrugs promised enhanced efficacy and safety.

2

Theatypical agents differ pharmacologically from previ-ous antipsychotic agents in their lower affinity fordopamine D2 receptors and greater affinities forother neuroreceptors, including those for serotonin(5-hydroxytryptamine

1

A

,

2

A

,

2

C

,

3

,

6

, and

7

) and nor-epinephrine (

a

1

and

a

2

).

1

Although studies indicated that the atypicaldrugs are similar to the conventional drugs in reduc-ing psychotic symptoms and produce few neuro-logic effects, the evidence of their superior efficacyhas been neither consistent nor robust,

3-8

with theexception of clozapine, which repeatedly has beeneffective in patients whose condition is refractory totreatment with other types of agents but has severeside effects that limit its use.

9-11

The newer agentsappear more efficacious than conventional drugsin reducing negative symptoms (e.g., lack of emo-tion, interest, and expression), possibly owing to theabsence of extrapyramidal symptoms

12

or other sec-ondary causes of negative symptoms (e.g., depres-sion) rather than to direct therapeutic effects.

13

The results of studies of the effects of treatment oncognitive impairment and mood symptoms havebeen inconclusive.

14,15

The ability of atypical agentsto prevent relapse and their effects on social andvocational functioning, quality of life, long-termoutcome, and the caregivers’ burden have been in-completely explored.

8,12,16

The safety advantages of the atypical drugs havebeen questioned because of their propensity to in-duce weight gain

17

and alter glucose and lipid me-tabolism.

18,19

Nevertheless, these medications arewidely used and have a 90 percent market share inthe United States,

20,21

resulting in burgeoningcosts. In the wake of this trend, questions havebeen raised about the clinical advantages and costeffectiveness of the atypical drugs. We report theprimary outcomes of a double-blind, active-controlclinical trial sponsored by the National Institute ofMental Health (NIMH) that was designed to com-pare the effectiveness of atypical and conventionalantipsychotic drugs.

22,23

study setting and design

The Clinical Antipsychotic Trials of InterventionEffectiveness (CATIE) study was initiated by theNIMH to compare the effectiveness of antipsychoticdrugs. Its rationale, design, and methods have beendescribed previously.

24-28

The protocol was madeavailable to the public for comment, and a commit-tee of scientific experts, health care administrators,and consumer advocates critiqued the study underthe auspices of the NIMH. The study was conduct-ed between January 2001 and December 2004 at 57clinical sites in the United States (16 university clin-ics, 10 state mental health agencies, 7 Veterans Af-fairs medical centers, 6 private nonprofit agencies,4 private-practice sites, and 14 mixed-system sites).Patients were initially randomly assigned to receiveolanzapine, perphenazine, quetiapine, or risperi-done under double-blind conditions and followedfor up to 18 months or until treatment was discon-tinued for any reason (phase 1). (Ziprasidone wasapproved for use by the Food and Drug Adminis-tration [FDA] after the study began and was addedto the study in January 2002 in the form of an iden-tical-appearing capsule containing 40 mg.) Patientswhose assigned treatment was discontinued couldreceive other treatments in phases 2 and 3.

24

Thepresent report is limited to phase 1 results.

participants

Eligible patients were 18 to 65 years of age; had re-ceived a diagnosis of schizophrenia, as determinedon the basis of the Structured Clinical Interview ofthe

Diagnostic and Statistical Manual of Mental Disorders,

fourth edition; and were able to take oral antipsy-chotic medication, as determined by the study doc-tor. Patients were excluded if they had received adiagnosis of schizoaffective disorder, mental retar-dation, or other cognitive disorders; had a historyof serious adverse reactions to the proposed treat-ments; had had only one schizophrenic episode;had a history of treatment resistance, defined by thepersistence of severe symptoms despite adequatetrials of one of the proposed treatments or priortreatment with clozapine; were pregnant or breast-feeding; or had a serious and unstable medicalcondition.

The study was approved by the institutional re-view board at each site, and written informed con-sent was obtained from the patients or their legalguardians.

a methods



n engl j med

353;12


22, 2005

effectiveness of antipsychotic drugs in chronic schizophrenia

1211

interventions

Identical-appearing capsules contained olanzap-ine (Zyprexa, Eli Lilly) (7.5 mg), quetiapine (Sero-quel, AstraZeneca) (200 mg), risperidone (Risper-dal, Janssen Pharmaceutica) (1.5 mg), perphenazine(Trilafon, Schering-Plough, at the time of the study)(8 mg), or (after January 2002) ziprasidone (Geo-don, Pfizer) (40 mg). The packaging was done byQuintiles. The dose of medications was flexible,ranging from one to four capsules daily, and wasbased on the study doctor’s judgment. Overlap inthe administration of the antipsychotic agents thatpatients received before study entry was permittedfor the first four weeks after randomization to allowa gradual transition to study medication. Concom-itant medications were permitted throughout thetrial, except for additional antipsychotic agents.Patients had monthly visits with study doctors.

Because of product labeling, quetiapine andziprasidone are given twice daily and olanzapine,perphenazine, and risperidone once daily. To pro-tect blinding, half the patients randomly assignedto perphenazine, olanzapine, and risperidone wereassigned to twice-daily dosing and half to once-daily dosing. To minimize initial side effects, pa-tients assigned to quetiapine began treatment byreceiving one 100-mg capsule on days 1 and 2, onetwice daily on day 3, and one for the first dose ofday 4. All patients assigned to twice-daily dosingreceived five identical-appearing capsules to begintreatment. Patients with current tardive dyskine-sia could enroll, but the randomization schemeprevented their assignment to treatment with per-phenazine.

objectives and outcomes

We hypothesized that there would be significantdifferences in the overall effectiveness of olanza-pine, perphenazine, quetiapine, risperidone, andziprasidone in treating schizophrenia that reflectedvariations in efficacy and tolerability. The primaryoutcome measure was the discontinuation of treat-ment for any cause, a discrete outcome selected be-cause stopping or changing medication is a frequentoccurrence and major problem in the treatment ofschizophrenia. In addition, this measure integratespatients’ and clinicians’ judgments of efficacy, safe-ty, and tolerability into a global measure of effec-tiveness that reflects their evaluation of therapeuticbenefits in relation to undesirable effects. The keysecondary outcomes were the specific reasons forthe discontinuation of treatment (e.g., inefficacy or

intolerability owing to side effects such as weightgain, extrapyramidal signs, or sedation as judgedby the study doctor). Additional secondary efficacyoutcomes included scores on the Positive and Neg-ative Syndrome Scale (PANSS) and the ClinicalGlobal Impressions (CGI) Scale. PANSS scores canrange from 30 to 210, with higher scores indicat-ing more severe psychopathology. Scores for theCGI Scale can range from 1 to 7, with higher scoresindicating greater severity of illness. Secondary safe-ty and tolerability outcomes, which were evaluatedat months 1, 3, 6, 9, 12, 15, and 18, included the in-cidence of serious adverse events, the incidence ofadverse events during treatment, the incidence ofneurologic side effects, and changes in weight, elec-trocardiographic findings, and laboratory analytes.

statistical analysis

Randomized patients who received at least onedose of study medication made up the intention-to-treat population. Two hundred thirty-one patientswith tardive dyskinesia were excluded from randomassignment to perphenazine. Ziprasidone was add-ed to the trial after approximately 40 percent ofthe patients had been enrolled. Consequently, com-parisons involving the perphenazine group werelimited to patients without tardive dyskinesia, andcomparisons involving the ziprasidone group werelimited to the cohort of patients who underwentrandomization after ziprasidone was added (theziprasidone cohort). In general, the trial had a sta-tistical power of 85 percent to identify an absolutedifference of 12 percent in the rates of discontinu-ation between two atypical agents; however, it hada statistical power of 76 percent for comparisonsinvolving perphenazine and of 58 percent for com-parisons involving ziprasidone.

We used Kaplan–Meier survival curves to esti-mate the time to the discontinuation of treatment.Treatment groups were compared with use of Coxproportional-hazards regression models

29

strati-fied according to site, with adjustment for whetherthe patient had had an exacerbation of schizophre-nia in the preceding three months and tardive dys-kinesia status (for models excluding perphena-zine). Sites with 15 or fewer patients were groupedaccording to the sites’ health care systems.

The overall difference among the olanzapine,quetiapine, risperidone, and perphenazine groupswas evaluated with the use of a test with 3 degreesof freedom (df ). If the difference was significant ata P value of less than 0.05, the three atypical-drug



n engl j med

353;12


22

,

2005

The

new england journal

of

medicine

1212

groups were compared with each other by meansof step-down or closed testing, with a P value ofless than 0.05 considered to indicate statistical sig-nificance. Each group was then compared with theperphenazine group by means of a Hochberg ad-justment for multiple comparisons.

30

The smallestresulting P value was compared with a value of 0.017(0.05 ÷ 3). The ziprasidone group was directly com-pared with the other three atypical-drug groups andthe perphenazine group within the ziprasidone co-hort by means of a Hochberg adjustment for fourpairwise comparisons. The smallest resulting P val-ue was compared with a value of 0.013 (0.05 ÷ 4).

Successful treatment time was defined as thenumber of months of treatment during phase 1 inwhich patients had a CGI Scale score of at least 3(mildly ill) or a score of 4 (moderately ill) with animprovement of at least two points from baseline.Treatment groups were compared with use of pro-portional-hazards regression.

A sensitivity analysis of the Cox model for thediscontinuation of treatment for any cause evaluat-ed the effects of potentially important baseline co-variates and their interaction with the treatmentgroup.

The PANSS total scores and CGI Scale scoresover time were compared among the groups withthe use of a mixed model including the same fixedcovariates as for the time to discontinuation, plusbaseline value, time, the interaction between treat-ment and time, and the interaction between base-line value and time. Time was classified into months(1, 3, 6, 9, 12, 15, and 18). The results of assess-ments made at the end of phase 1 were assigned tothe next interval. The correlation of the repeatedmeasures within each patient was modeled withthe use of a random subject intercept and an un-structured covariance matrix.

The study was funded by the NIMH. The pharma-ceutical companies whose drugs were included inthe study donated drug supplies, and each provid-ed advice on the dose of its own drug; they wereotherwise not involved in the design of the study,analyses, or interpretation of results. The manu-script was written solely by the listed authors.

characteristics and disposition of patients

Table 1 shows the baseline demographic and clini-cal characteristics of the patients. Figure 1 depictsthe enrollment, randomization, and follow-up of

study patients; 1493 patients were enrolled in thestudy and randomly assigned to treatment. All datafrom one site (33 patients) were excluded beforeanalysis, owing to concern about the integrity of datafrom that site before the end of the study and beforeunblinding. The mean modal doses were 20.1 mgper day for olanzapine, 20.8 mg per day for per-phenazine, 543.4 mg per day for quetiapine, 3.9 mgper day for risperidone, and 112.8 mg per day forziprasidone (Table 2). Seventy-four percent of pa-tients in the intention-to-treat analysis (1061 of1432) discontinued their assigned treatment inphase 1 before 18 months (median, 6).

discontinuation of treatment

The time to the discontinuation of treatment forany cause was longer in the olanzapine group thanin the quetiapine group (hazard ratio, 0.63; P<0.001),the risperidone group (hazard ratio, 0.75; P=0.002),or the perphenazine group (hazard ratio, 0.78;P=0.021) (Table 2). However, the difference be-tween the olanzapine group and the perphenazinegroup was not significant after adjustment for mul-tiple comparisons (required P value, ≤0.017). With-in the cohort of 889 patients who underwent ran-domization after ziprasidone was added to the trial,those receiving olanzapine had a longer interval be-fore discontinuing treatment for any cause thandid those in the ziprasidone group (hazard ratio,0.76; P=0.028). However, this difference was notsignificant after adjustment for multiple compari-sons (required P value, ≤0.013).

The time to the discontinuation of treatment forlack of efficacy was longer in the olanzapine groupthan in the perphenazine group (hazard ratio, 0.47;P<0.001), the quetiapine group (hazard ratio, 0.41;P<0.001), the risperidone group (hazard ratio,0.45; P<0.001), or the ziprasidone group (hazardratio, 0.59; P=0.026), but the difference betweenthe olanzapine and ziprasidone groups was not sig-nificant after adjustment for multiple comparisons(required P value, ≤0.013) (Table 2). There were nosignificant differences between groups in time un-til discontinuation owing to intolerable side effects(P=0.054). The time until discontinuation owingto the patient’s decision (i.e., the patient indepen-dently chose to stop treatment) was similar to thatfor discontinuation for any cause (Table 2).

The duration of successful treatment was sig-nificantly longer in the olanzapine group than inthe quetiapine group (hazard ratio, 0.53; P<0.001),the risperidone group (hazard ratio, 0.69; P=0.002),or the perphenazine group (hazard ratio, 0.73;

results



n engl j med

353;12


22, 2005


1213

* Plus–minus values are means ±SD. Because of rounding, percentages may not sum to 100. SCID denotes Structured Clinical Interview for DSM-IV.† Patients with tardive dyskinesia were excluded from the perphenazine group.‡ Race was self-reported. “Other” includes American Indian or Alaska Native (less than 1 percent of patients), Asian (2 percent), Native Hawaiian

or other Pacific Islander (less than 1 percent), and two or more races (2 percent). Percentages are based on the number of patients with data available: 336 in the olanzapine group, 337 in the quetiapine group, 341 in the risperidone group, 261 in the perphenazine group, and 183 in the ziprasidone group.

§ This category includes patients who were widowed, divorced, or separated. ¶ Percentages are based on the number of patients with data available: 330 in the olanzapine group, 328 in the quetiapine group, 336 in the ris-

peridone group, 259 in the perphenazine group, and 182 in the ziprasidone group.¿ Scores on the Positive and Negative Syndrome Scale (PANSS) for schizophrenia can range from 30 to 210, with higher scores indicating

more severe psychopathology.** The CGI severity score can range from 1 to 7, with higher scores indicating greater severity of illness.††Percentages for baseline medications are based on the number of patients with data on concomitant medications: 333 in the olanzapine

group, 333 in the quetiapine group, 340 in the risperidone group, 259 in the perphenazine group, and 184 in the ziprasidone group.

Table 1. Baseline Demographic and Clinical Characteristics of Randomized Patients.*

CharacteristicOlanzapine

(N=336)Quetiapine

(N=337)Risperidone

(N=341)Perphenazine

(N=261)†Ziprasidone

(N=185)Total

(N=1460)

Demographic characteristics

Age — yr 40.8±10.8 40.9±11.2 40.6±11.3 40.0±11.1 40.1±11.0 40.6±11.1Sex — no. (%)

Male 244 (73) 255 (76) 253 (74) 199 (76) 129 (70) 1080 (74)Female 92 (27) 82 (24) 88 (26) 62 (24) 56 (30) 380 (26)

Race — no. (%)‡White 196 (58) 213 (63) 204 (60) 152 (58) 109 (60) 874 (60)Black 119 (35) 114 (34) 122 (36) 93 (36) 65 (36) 513 (35)Other 21 (6) 10 (3) 15 (4) 16 (6) 9 (5) 71 (5)

Spanish, Hispanic, or Latino ethnicity — no. (%) 42 (12) 48 (14) 38 (11) 24 (9) 18 (10) 170 (12)Education — yr 12.2±2.2 12.1±2.4 12.0±2.2 12.1±2.1 12.0±2.5 12.1±2.3Marital status — no. (%)

Married 36 (11) 34 (10) 37 (11) 43 (16) 17 (9) 167 (11)Previously married§ 105 (31) 90 (27) 101 (30) 68 (26) 61 (33) 425 (29)Never married 195 (58) 213 (63) 203 (60) 150 (57) 107 (58) 868 (59)

Unemployed — no. (%)¶ 281 (85) 274 (84) 288 (86) 219 (85) 155 (85) 1217 (85)Exacerbation in previous 3 mo — no. (%) 90 (27) 89 (26) 95 (28) 68 (26) 60 (32) 402 (28)PANSS total score¿ 76.1±18.2 75.7±16.9 76.4±16.6 74.3±18.1 75.4±18.6 75.7±17.6Clinician-rated CGI severity score** 4.0±1.0 3.9±0.9 4.0±0.9 3.9±1.0 3.9±0.9 4.0±0.9

Psychiatric history

Age at 1st treatment for any behavioral or emotional problem — yr

24.1±9.0 23.6±8.1 23.7±9.3 24.5±8.6 24.1±9.7 24.0±8.9

Years since 1st antipsychotic medication prescribed

14.5±11.0 14.6±10.3 14.8±10.7 13.8±11.0 14.0±10.5 14.4±10.7

SCID diagnosis in past 5 yr — no. (%)

Depression 86 (26) 84 (25) 104 (30) 71 (27) 60 (32) 405 (28)Alcohol dependence or alcohol abuse 74 (22) 81 (24) 92 (27) 74 (28) 37 (20) 358 (25)Drug dependence or drug abuse 86 (26) 95 (28) 110 (32) 74 (28) 57 (31) 422 (29)Obsessive–compulsive disorder 10 (3) 22 (7) 21 (6) 12 (5) 8 (4) 73 (5)Other anxiety disorder 44 (13) 46 (14) 52 (15) 29 (11) 28 (15) 199 (14)

Baseline antipsychotic medications — no. (%)††

Olanzapine alone 78 (23) 69 (20) 76 (22) 58 (22) 41 (22) 322 (22)Quetiapine alone 24 (7) 17 (5) 22 (6) 15 (6) 17 (9) 95 (7)Risperidone alone 57 (17) 59 (18) 63 (18) 64 (25) 32 (17) 275 (19)Any combination including olanzapine, quetia-

pine, or risperidone31 (9) 32 (10) 33 (10) 21 (8) 8 (4) 95 (7)

All others 52 (15) 58 (17) 60 (18) 30 (11) 29 (16) 229 (16)None 94 (28) 102 (30) 87 (26) 73 (28) 58 (31) 414 (28)

Baseline medical diagnoses — no. (%)

Diabetes (type 1 or 2) 36 (11) 40 (12) 32 (9) 29 (11) 17 (9) 154 (11)Hyperlipidemia 56 (17) 44 (13) 42 (12) 36 (14) 26 (14) 204 (14)Hypertension 68 (20) 67 (20) 63 (18) 60 (23) 31 (17) 289 (20)



n engl j med

353;12


22

,

2005

The

new england journal

of

medicine

1214

P=0.013) and was significantly longer in the risperi-done group than in the quetiapine group (hazardratio, 0.77; P=0.021).

adjustment of outcomes for covariates

An exploratory analysis identified the followingpredictors of an earlier time to discontinuation:higher baseline PANSS score (P=0.001), youngerage (P<0.001), longer duration since the first useof antipsychotic medication (P=0.057), and the an-tipsychotic drug taken before study entry (P=0.001).Baseline antipsychotic agents were grouped into sixcategories (Table 1). Patients receiving olanzapineor risperidone before enrollment stayed in phase 1

of the trial longer than those taking no antipsychot-ic agents, those taking combination treatments, orthose receiving a single antipsychotic agent exclud-ing olanzapine, quetiapine, or risperidone; pair-wise hazard ratios ranged from 0.68 (P<0.001) to0.80 (P<0.02). No interactions with treatment groupwere significant at a P value of less than 0.10. Afteradjustment for these predictors of discontinuation,the results of treatment-group comparisons weresimilar to the primary results.

efficacy measures

Total PANSS scores improved over time in all groups(Fig. 2). The mixed model revealed significant vari-

Figure 1. Enrollment and Outcomes.

Patients with tardive dyskinesia were not assigned to perphenazine. Ziprasidone was added to the study after approximately 40 percent of patients had been enrolled.

1894 Screened

1493 Underwent randomization

337 Assigned to quetiapine

8 Did not take drug

341 Assigned to risperidone

8 Did not take drug

185 Assigned to ziprasidone

2 Did not take drug

261 Assigned to perphenazine

4 Did not take drug

336 Assigned to olanzapine

6 Did not take drug

329 Included in analysis 333 included in analysis 183 Included in analysis257 Included in analysis330 Included in analysis

60 (18%) Completed phase 1269 (82%) Discontinued

quetiapine92 For lack of efficacy49 Owing to intoler-

ability109 Owing to patient’s

decision 19 For other reasons

88 (26%) Completedphase 1

245 (74%) Discontinuedrisperidone

91 For lack of efficacy34 Owing to intoler-


decision19 For other reasons

38 (21%) Completedphase 1

145 (79%) Discontinuedziprasidone

44 For lack of efficacy28 Owing to intoler-




perphenazine65 For lack of efficacy40 Owing to intoler-




olanzapine48 For lack of efficacy62 Owing to intoler-



401 Excluded124 Did not meet study criteria109 Declined 33 Decided against changing antipsychotic agent135 Had other reasons

All 33 patients from one siteexcluded before analysisbecause of concern aboutintegrity of the data



n engl j med

353;12


22, 2005


1215

ation in treatment effects over time (P=0.002). Im-provement was initially greatest in the olanzapinegroup, but its advantage diminished over time. Thepattern of change in the scores for the CGI Scalewas similar to that for the PANSS scores (P=0.004for the interaction between treatment and time).

adverse events

The rates of adverse events and side effects are list-ed in Table 3. Fewer patients in the olanzapine groupthan in the other four groups were hospitalized foran exacerbation of schizophrenia (11 percent vs. 15to 20 percent, P<0.001). After adjustment for thedifferent durations of treatment, the olanzapinegroup had a risk ratio for hospitalization of 0.17 perperson-year of treatment, as compared with riskratios of 0.30 to 0.44 in the other groups.

The rates of treatment discontinuation due tointolerable side effects differed between treatments(P=0.04). Risperidone had the lowest rate (10 per-cent), and olanzapine had the highest rate (18 per-cent). Moreover, more patients discontinued olan-zapine owing to weight gain or metabolic effects(9 percent vs. 1 percent to 4 percent with the otherfour drugs, P<0.001) and more patients discontin-ued perphenazine owing to extrapyramidal effects(8 percent vs. 2 percent to 4 percent, P=0.002).

Patients in the olanzapine and quetiapine groupshad lower rates of insomnia (16 and 18 percent, re-spectively) than did patients in the other groups (24percent in the risperidone group, 25 percent in theperphenazine group, and 30 percent in the ziprasi-done group). Quetiapine was associated with a high-er rate of anticholinergic effects than were the otherdrugs (31 percent vs. 20 to 25 percent, P<0.001).

Neurologic Side Effects

There were no significant differences among thegroups in the incidence of extrapyramidal side ef-fects, akathisia, or movement disorders as reflectedby rating-scale measures of severity.

Weight Gain and Metabolic Changes

Patients in the olanzapine group gained moreweight than patients in any other group, with an av-erage weight gain of 2 lb (0.9 kg) per month. A largerproportion of patients in the olanzapine group thanin the other groups gained 7 percent or more of theirbaseline body weight (30 percent vs. 7 to 16 percent,P<0.001).

Olanzapine had effects consistent with the po-tential development of the metabolic syndrome and

was associated with greater increases in glycosylat-ed hemoglobin, total cholesterol, and triglyceridesafter randomization than the other study drugs,even after adjustment for the duration of treat-ment. Ziprasidone was the only study drug associ-ated with improvement in each of these metabolicvariables. Only risperidone was associated with asubstantial increase in prolactin levels.

Other Potential Adverse Events

There were no substantially different effects of themedications on the corrected QT interval on elec-trocardiography, and torsades de pointes did notdevelop in any patients. There were no significantdifferences among the groups in the incidence ofnew cataracts. There were no significant differencesamong the groups in the rates of suicide attemptsor suicidal ideation reported as serious adverseevents.

concomitant medications

There were few substantial differences among thegroups in the rates or types of medications addedduring the study. Patients in the olanzapine and ris-peridone groups were the least likely to have anxio-lytic agents added (9 and 10 percent, respectively,vs. 14 to 15 percent). Fewer patients receiving que-tiapine were prescribed anticholinergic drugs (3 per-cent vs. 8 to 10 percent).

All second-generation antipsychotic drugs were in-cluded in phase 1 of this study except aripiprazole(which was approved by the FDA in November2002) and clozapine, which was included in phase 2for patients who discontinued phase 1 of treatmentowing to lack of efficacy of the assigned drug. Al-though haloperidol is the first-generation agentmost commonly used for comparison, we chose touse perphenazine because of its lower potency andmoderate side-effect profile.

31

Only a minority of patients in each group tooktheir assigned drug for the duration of phase 1 (ratesof discontinuation ranged from 64 to 82 percent).This outcome indicates that antipsychotic drugs,though effective, have substantial limitations intheir effectiveness in patients with chronic schizo-phrenia. Although the rates of discontinuation mayhave been increased by the fact that patients wereparticipating in a blinded, controlled trial, the ratesare generally consistent with those previously ob-

discussion



n engl j med

353;12


22

,

2005

The

new england journal

of

medicine

1216

Tabl

e 2.

Out

com

e M

easu

res

of E

ffec

tiven

ess

in th

e In

tent

ion-

to-T

reat

Pop

ulat

ion.

*

Out

com

eO

lanz

apin

e(N

=33

0)Q

uetia

pine

(N=

329)

Ris

peri

done

(N=

333)

Perp

hena

zine

(N=

257)

†P

Valu

e‡Z

ipra

sido

ne(N

=18

3)§

Dos

e¶

Mea

n m

odal

dos

e —

mg

per

day/

tota

l no.

of p

atie

nts

20.1

/312

54

3.4/

309

3.9/

305

20.8

/245

11

2.8/

165

Max

imal

dos

e re

ceiv

ed —

no.

of p

atie

nts

(%)

124/

312

(40)

137/

309

(44)

122/

305

(40)

98/2

45 (

40)

<0.0

0180

/165

(48

)

Dis

cont

inua

tion

of tr

eatm

ent f

or a

ny c

ause

Dis

cont

inua

tion

— n

o. o

f pat

ient

s (%

)21

0 (6

4)26

9 (8

2)24

5 (7

4)19

2 (7

5)14

5 (7

9)K

apla

n–M

eier

tim

e to

dis

cont

inua

tion

— m

oM

edia

n (9

5% C

I)9.

2 (6

.9–1

2.1)

4.6

(3.9

–5.5

)4.

8 (4

.0–6

.1)

5.6

(4.5

–6.3

)3.

5 (3

.1–5

.4)

Cox

-mod

el tr

eatm

ent c

ompa

riso

ns¿

Ola

nzap

ine

Haz

ard

ratio

(95

% C

I) 0

.63

(0.5

2–0.

76)

0.75

(0.

62–0

.90)

0.78

(0.

63–0

.96)

0.00

4**

0.76

(0.

60–0

.97)

P va

lue

<0.0

01**

0.00

2**

0.02

10.

028

Que

tiapi

neH

azar

d ra

tio (

95%

CI)

1.19

(0.

99–1

.42)

1.14

(0.

93–1

.39)

1.01

(0.

81–1

.27)

P va

lue

0.06

0.21

0.94

Ris

peri

done

Haz

ard

ratio

(95

% C

I)1.

00 (

0.82

–1.2

3)0.

89 (

0.71

–1.1

4)P

valu

e0.

990.

36Pe

rphe

nazi

neH

azar

d ra

tio (

95%

CI)

0.90

(0.

70–1

.16)

P va

lue

0.43

Dis

cont

inua

tion

of tr

eatm

ent f

or la

ck o

f eff

icac

y

Dis

cont

inua

tion

— n

o. o

f pat

ient

s (%

)48

(15

)92

(28

)91

(27

)65

(25

)44

(24

)K

apla

n–M

eier

tim

e to

dis

cont

inua

tion

— m

o25

th p

erce

ntile

(95

% C

I)—

††

6.0

(4.5

–8.0

)6.

0 (4

.4–9

.0)

6.1

(4.5

–9.1

)6.

9 (3

.2–1

2.1)

Cox

-mod

el tr

eatm

ent c

ompa

riso

ns¿

Ola

nzap

ine

Haz

ard

ratio

(95

% C

I)0.

41 (

0.29

–0.5

7)0.

45 (

0.32

–0.6

4)0.

47 (

0.31

–0.7

0)<0

.001

**0.

59 (

0.37

–0.9

3)P

valu

e<0

.001

**<0

.001

**<0

.001

**0.

026

Que

tiapi

neP

valu

e0.

490.

470.

69R

ispe

rido

neP

valu

e0.

590.

93Pe

rphe

nazi

neP

valu

e0.

44

Dis

cont

inua

tion

of tr

eatm

ent o

win

g to

into

lera

bilit

y‡‡

Dis

cont

inua

tion

— n

o. (

%)

62 (

19)

49 (

15)

34 (

10)

40 (

16)

28 (

15)

Cox

-mod

el tr

eatm

ent c

ompa

riso

ns¿

Ris

peri

done

Haz

ard

ratio

(95

% C

I)0.

62 (

0.41

–0.9

5)0.

65 (

0.42

–1.0

0)0.

60 (

0.36

–0.9

8)0.

054

0.79

(0.

46–1

.37)

P va

lue

0.02

70.

051

0.04

30.

41O

lanz

apin

eP

valu

e0.

840.

490.

28Q

uetia

pine

P va

lue

0.97

0.87

Perp

hena

zine

P va

lue

0.19



n engl j med

353;12


22, 2005


1217

*C

I den

otes

con

fiden

ce in

terv

al.

†Pa

tient

s w

ith ta

rdiv

e dy

skin

esia

wer

e ex

clud

ed fr

om th

e pe

rphe

nazi

ne g

roup

.‡

The

over

all P

val

ue is

for t

he c

ompa

riso

n of

ola

nzap

ine,

que

tiapi

ne, r

ispe

rido

ne, a

nd p

erph

enaz

ine

with

the

use

of a

3 d

f tes

t fro

m a

Cox

mod

el fo

r sur

viva

l out

com

es, e

xclu

ding

pat

ient

s w

ith ta

rdiv

e dy

skin

esia

. If t

he d

iffer

ence

am

ong

the

grou

ps w

as s

igni

fican

t at a

P v

alue

of l

ess

than

0.0

5, th

e th

ree

atyp

ical

age

nts

wer

e co

mpa

red

with

eac

h ot

her b

y m

eans

of s

tep-

dow

n or

clo

sed

test

ing

to id

entif

y si

gnifi

cant

diff

eren

ces

(P<0

.05)

bet

wee

n gr

oups

. Eac

h at

ypic

al a

gent

was

then

com

pare

d w

ith p

erph

enaz

ine

by m

eans

of a

Hoc

hber

g ad

just

men

t. Th

e sm

all-

est P

val

ue fo

r th

e pe

rphe

nazi

ne g

roup

was

com

pare

d w

ith a

val

ue o

f 0.0

17 (

0.05

÷3)

. §

Stat

istic

al a

naly

ses

invo

lvin

g th

e zi

pras

idon

e gr

oup

wer

e co

nfin

ed to

the

coho

rt o

f pat

ient

s w

ho u

nder

wen

t ran

dom

izat

ion

afte

r zi

pras

idon

e w

as a

dded

to th

e st

udy,

with

the

use

of a

H

ochb

erg

adju

stm

ent f

or fo

ur p

airw

ise

com

pari

sons

. The

sm

alle

st P

val

ue w

as c

ompa

red

with

a v

alue

of 0

.013

(0.

05÷

4).

¶Th

e m

odal

dos

e an

d pe

rcen

tage

s of

pat

ient

s ta

king

the

max

imal

dos

e ar

e ba

sed

on th

e nu

mbe

r of

pat

ient

s w

ith d

ata

on th

e do

se. I

nfor

mat

ion

on d

ose

was

not

ava

ilabl

e fo

r so

me

pa-

tient

s w

ho d

ropp

ed o

ut e

arly

. The

P v

alue

s fo

r th

e pe

rcen

tage

of p

atie

nts

reac

hing

the

max

imal

dos

e w

ere

calc

ulat

ed w

ith th

e us

e of

a 4

df t

est c

ompa

ring

all

trea

tmen

t gro

ups

from

a

Pois

son

regr

essi

on a

ccou

ntin

g fo

r di

ffere

ntia

l exp

osur

e tim

es, a

nd a

djus

ting

for

whe

ther

the

patie

nt h

ad h

ad a

n ex

acer

batio

n in

the

prec

edin

g th

ree

mon

ths.

¿

For

pair

wis

e co

mpa

riso

ns o

f tre

atm

ent g

roup

s, C

ox-m

odel

haz

ard

ratio

s of

less

than

1 in

dica

te a

gre

ater

tim

e to

the

disc

ontin

uatio

n of

the

first

trea

tmen

t lis

ted.

**P

valu

e is

sta

tistic

ally

sig

nific

ant.

††

The

Kap

lan–

Mei

er 2

5th

perc

entil

e fo

r di

scon

tinua

tion

owin

g to

lack

of e

ffica

cy c

ould

not

be

estim

ated

for

olan

zapi

ne b

ecau

se o

f the

low

eve

nt r

ates

.‡

‡Th

e K

apla

n–M

eier

25t

h pe

rcen

tile

for

disc

ontin

uatio

n ow

ing

to in

tole

rabi

lity

coul

d no

t be

estim

ated

bec

ause

of t

he lo

w e

vent

rat

es.

§§Th

is c

ateg

ory

incl

udes

dec

isio

ns m

ade

by b

oth

patie

nts

and

thei

r ad

voca

tes.

¶¶

Succ

essf

ul tr

eatm

ent w

as d

efin

ed b

y a

CG

I sev

erity

sco

re o

f at l

east

3 (

mild

ly il

l) o

r by

a s

core

of 4

(m

oder

atel

y ill

) w

ith a

n im

prov

emen

t of a

t lea

st tw

o po

ints

from

bas

elin

e.

Patie

nt’s

dec

isio

n to

dis

cont

inue

trea

tmen

t§§

Dis

cont

inua

tion

— n

o. (

%)

78 (

24)

109

(33)

10

1 (3

0)

77 (

30)

63 (

34)

Kap

lan–

Mei

er ti

me

to d

isco

ntin

uatio

n —

mo

25th

per

cent

ile (

95%

CI)

12.3

(8.

0–17

.8)

4.9

(3.1

–7.0

)4.

5 (3

.1–8

.8)

6.2

(4.7

–8.1

)3.

4 (3

.0–6

.1)

Cox

-mod

el tr

eatm

ent c

ompa

riso

ns¿

Ola

nzap

ine

Haz

ard

ratio

(95

% C

I)0.

56 (

0.42

–0.7

5)0.

67 (

0.50

–0.9

0)0.

70 (

0.50

–0.9

8)0.

034*

*0.

63 (

0.43

–0.9

3)P

valu

e<0

.001

**0.

008*

*0.

036

0.01

8Q

uetia

pine

P va

lue

0.21

0.46

0.63

Ris

peri

done

P va

lue

0.95

0.21

Perp

hena

zine

P va

lue

0.27

Dur

atio

n of

suc

cess

ful t

reat

men

t¶¶

Kap

lan–

Mei

er ti

me

to d

isco

ntin

uatio

n —

mo

Med

ian

(95%

CI)

3 (2

–5)

1 (0

–1)

1 (0

–1)

1 (1

–2)

1 (0

–1)

Cox

-mod

el tr

eatm

ent c

ompa

riso

ns¿

Ola

nzap

ine

Haz

ard

ratio

(95

% C

I)0.

53 (

0.43

–0.6

7)0.

69 (

0.55

–0.8

7)0.

73 (

0.57

–0.9

3)<0

.001

**0.

75 (

0.58

–0.9

4)P

valu

e<0

.001

**0.

002*

*0.

013*

*0.

017

Que

tiapi

neH

azar

d ra

tio (

95%

CI)

1.30

(1.

04–4

.63)

1.28

(1.

00–1

.64)

1.06

(0.

85–1

.33)

P va

lue

0.02

**0.

050.

61R

ispe

rido

neP

valu

e0.

720.

74Pe

rphe

nazi

neP

valu

e0.

25



n engl j med

353;12


22

,

2005

The

new england journal

of

medicine

1218

Figure 2 (facing page). Outcome Measures of Effec-tiveness.

The number of patients included at each assessment time point declined over time. Estimates are from a mixed model, which assumed that data were missing at random. Scores for the PANSS and CGI Scale were determined at study entry and 1, 3, 6, 9, 12, 15, and 18 months after ran-domization. Scores for the PANSS can range from 30 to 210, with higher scores indicating more severe psycho-pathology. Scores for the CGI Scale can range from 1 to 7, with higher scores indicating a greater severity of illness. Analyses involving the ziprasidone group were limited to the cohort of patients who underwent randomization after the addition of ziprasidone to the study (the ziprasidone cohort). Thus, the P value for the overall interaction be-tween time and treatment excludes the ziprasidone group and is given separately for the ziprasidone cohort.

served.

5

Within this limited range of effectiveness,the olanzapine group had the lowest rate of discon-tinuation, which might lead one to consider olan-zapine the most effective of the medications stud-ied. Its apparent superior efficacy is also indicatedby the greater reduction in psychopathology, longerduration of successful treatment, and lower rate ofhospitalizations for an exacerbation of schizophre-nia. The results for the other second-generationantipsychotic agents and the representative con-ventional drug, perphenazine, were similar in mostrespects. It is important to note that the differencesbetween olanzapine and perphenazine were mod-erate. Although there were no significant differ-ences in the time until discontinuation owing tointolerable side effects, there were differences inrates. Moreover, olanzapine was associated withgreater increases in weight and indexes of glucoseand lipid metabolism than the other treatments.

Dose could have been a factor in the performanceof the various agents studied. The dose ranges ap-proved by the FDA for quetiapine and ziprasidonemay be below their optimal therapeutic doses,and the recommended doses of risperidone (6 mgper day or less), intended to limit extrapyramidalsymptoms, may not encompass its full therapeu-tic range.

32,33

However, the dose ranges we usedwere based on information from the manufacturerof each medication plus knowledge of clinical prac-tice patterns. Moreover, the average prescribed dos-es of these drugs in the United States for patientswith schizophrenia during the period in which thestudy was conducted (14 mg of olanzapine per day,3.8 mg of risperidone per day, 388 mg of quetia-pine per day, and 125 mg of ziprasidone per day)were generally similar to the ones we used.

34

Thefact that a higher proportion of patients assigned toquetiapine and ziprasidone received the maximaldose allowed in the study suggests that these agentsare either less effective or require higher doses (Ta-ble 2). The dose range of perphenazine was chosento minimize the potential for extrapyramidal symp-toms that may have biased previous comparisons offirst- and second-generation drugs.

4,7,31

The use of low-dose perphenazine appears tohave diminished the frequency of extrapyramidalside effects in patients who received the first-gener-ation drug. In contrast to previous studies,

35

the pro-portion of patients with extrapyramidal symptomsdid not differ significantly among those who re-ceived first-generation and second-generation drugsin our study. Despite this finding, more patients dis-

continued perphenazine than other medications ow-ing to extrapyramidal effects.

As in other studies, we found that risperidonewas associated with hyperprolactinemia and olan-zapine was associated with substantial weight gainin addition to adverse changes in glucose and lipidmetabolism — all features of the metabolic syn-drome. Concerns about potential prolongation ofthe corrected QT interval with ziprasidone and ofcataracts with quetiapine were not realized in thisstudy.

We used broad inclusion and minimal exclu-sion criteria and allowed the enrollment of patientswith coexisting conditions and those who were tak-ing other medications. The study was conducted ina variety of clinical settings in which people withschizophrenia are treated. These “real-world” fea-tures of the study, which were intended to make theresults widely applicable, may account for the dif-ferences in results between this and previous stud-ies comparing first- and second-generation anti-psychotic agents.

In summary, patients with chronic schizophre-nia in this study discontinued their antipsychoticstudy medications at a high rate, indicating sub-stantial limitations in the effectiveness of the drugs.Within this limited range of effectiveness, olanza-pine appeared to be more effective than the otherdrugs studied, and there were no significant dif-ferences in effectiveness between the conventionaldrug perphenazine and the other second-genera-tion drugs. There were no significant differencesamong the drugs in the time until discontinuationof treatment owing to intolerable side effects. How-ever, olanzapine was associated with greater weight



n engl j med

353;12


22, 2005


1219

Olanzapine (N=330)Perphenazine (N=257) Quetiapine (N=329)

Ziprasidone (N=183)Risperidone (N=333)Pr

opor

tion

of P

atie

nts

with

out E

vent

0.8

0.9

0.7

0.6

0.4

0.3

0.1

0.5

0.2

0.00 3 6 9 12 15 18

1.0

P=0.002 for olanzapine vs. risperidoneP<0.001 for olanzapine vs. quetiapine

P=0.008 for olanzapine vs. risperidoneP<0.001 for olanzapine vs. quetiapine

P=0.017 for ziprasidone cohort

P=0.004 for time-by-treatment interaction

P=0.065 for ziprasidone cohort

P=0.002 for time-by-treatmentinteraction

Prop

ortio

n of

Pat

ient

s w

ithou

t Eve

nt

0.8

0.9

0.7

0.6

0.4

0.3

0.1

0.5

0.2

0.00 3 6 9 12 15 18

1.0

P<0.001 for olanzapine vs. quetiapine,risperidone, and perphenazine

Prop

ortio

n of

Pat

ient

s w

ithou

t Eve

nt

0.8

0.9

0.7

0.6

0.4

0.3

0.1

0.5

0.2

0.00 3 6 9 12 15 18

1.0

Prop

ortio

n of

Pat

ient

s w

ithou

t Eve

nt

0.8

0.9

0.7

0.6

0.4

0.3

0.1

0.5

0.2

0.00 3 6 9 12 15 18

1.0

Leas

t-Sq

uare

s M

ean

Cha

nge

in P

AN

SSTo

tal S

core

from

Bas

elin

e

¡4

¡2

¡6

¡8

¡10

¡120 3 6 9 12 15 18

0

Leas

t-Sq

uare

s M

ean

Cha

nge

in C

GI S

ever

ity S

core ¡0.2

¡0.1

¡0.3

¡0.4

¡0.6

¡0.5

¡0.70 3 6 9 12 15 18

0.0

A

C

E

Time to Discontinuation for Any Cause (mo)

Time to Discontinuation Owing toIntolerability (mo)

D

B

Time to Discontinuation for Lack of Efficacy (mo)

Time to Discontinuation Owing toPatient’s Decision (mo)

Months Months

F



n engl j med

353;12


22

,

2005

The

new england journal

of

medicine

1220

Table 3. Outcome Measures of Safety among Randomized Patients.

OutcomeOlanzapine

(N=336)Quetiapine(N= 337)

Risperidone(N=341)

Perphenazine(N=261)*

Ziprasidone(N=185) P Value†

Hospitalization for exacerbation of schizophrenia

Hospitalized patients — no. (%) 38 (11) 68 (20) 51 (15) 41 (16) 33 (18) <0.001No. of hospitalizations/total person-yr of exposure 45/257 80/183 64/210 54/161 40/100Risk ratio 0.17 0.44 0.30 0.33 0.40

Adverse events — no. (%)

Any serious adverse event 32 (10) 32 (9) 33 (10) 29 (11) 19 (10) 0.47Suicide attempt 2 (<1) 1 (<1) 2 (<1) 1 (<1) 1 (<1) 0.99Suicidal ideation 1 (<1) 2 (<1) 4 (1) 3 (1) 2 (1) 0.49Any moderate or severe adverse event identified by

systematic inquiry235 (70) 220 (65) 232 (68) 170 (65) 119 (64) 0.14

Insomnia 55 (16) 62 (18) 83 (24) 66 (25) 56 (30) <0.001Hypersomnia, sleepiness 104 (31) 103 (31) 96 (28) 74 (28) 45 (24) 0.18Urinary hesitancy, dry mouth, constipation 79 (24) 105 (31) 84 (25) 57 (22) 37 (20) <0.001Decreased sex drive, arousal, ability to reach orgasm 91 (27) 69 (20) 91 (27) 64 (25) 35 (19) 0.59Gynecomastia, galactorrhea 7 (2) 6 (2) 14 (4) 4 (2) 6 (3) 0.15Menstrual irregularities‡ 11 (12) 5 (6) 16 (18) 7 (11) 8 (14) 0.17Incontinence, nocturia 18 (5) 15 (4) 25 (7) 6 (2) 10 (5) 0.04Orthostatic faintness 31 (9) 38 (11) 37 (11) 29 (11) 24 (13) 0.08

Any moderate or severe spontaneously reported adverse event

122 (36) 113 (34) 123 (36) 79 (30) 65 (35) 0.10

Neurologic effects — no./total no. (%)§

AIMS global severity score ≥2 32/236 (14) 30/236 (13) 38/238 (16) 41/237 (17) 18/126 (14) 0.23Barnes Akathisia Rating Scale global score ≥3 15/290 (5) 16/305 (5) 20/292 (7) 16/241 (7) 14/158 (9) 0.24Simpson–Angus Extrapyramidal Signs Scale mean

score ≥1 23/296 (8) 12/298 (4) 23/292 (8) 15/243 (6) 6/152 (4) 0.47

Discontinuation of treatment owing to intolerability — no. %

Discontinuation 62 (18) 49 (15) 34 (10) 40 (15) 28 (15) 0.04Weight gain or metabolic effects 31 (9) 12 (4) 6 (2) 3 (1) 6 (3) <0.001Extrapyramidal effects 8 (2) 10 (3) 11 (3) 22 (8) 7 (4) 0.002Sedation 7 (2) 9 (3) 3 (1) 7 (3) 0 0.10Other effects 16 (5) 18 (5) 14 (4) 8 (3) 15 (8) 0.16

Weight change from baseline to last observation¶

Weight gain >7% — no./total no. (%) 92/307 (30) 49/305 (16) 42/300 (14) 29/243 (12) 12/161 (7) <0.001Weight change — lb

Mean ±SE 9.4±0.9 1.1±0.9 0.8±0.9 ¡2.0±1.1 ¡1.6±1.1 <0.001Median 7 1 0 ¡1 ¡2Range ¡14 to 42 ¡25 to 25 ¡24 to 24 ¡29 to 22 ¡24 to 18

Weight change — lb/mo of treatmentMean ±SE 2.0±0.3 0.5±0.2 0.4±0.3 ¡0.2±0.2 ¡0.3±0.3 <0.001Median 0.8 0.1 0.0 ¡0.1 ¡0.3Range ¡1.4 to 9.5 ¡4.4 to 6.3 ¡4.6 to 5.7 ¡4.9 to 4.0 ¡5.3 to 5.9

Change from baseline in laboratory values¿

Blood glucose — mg/dlMean ±SE 15.0±2.8 6.8±2.5 6.7±2.0 5.2±2.0 2.3±3.9Median 7.0 4.3 5.5 1.5 2.5Exposure-adjusted mean ±SE 13.7±2.5 7.5±2.5 6.6±2.5 5.4±2.8 2.9±3.4 0.59

Glycosylated hemoglobin — %Mean ±SE 0.41±0.09 0.05±0.05 0.08±0.04 0.10±0.06 ¡0.10±0.14Median 0.20 0.10 0.05 0.05 0.10Exposure-adjusted mean ±SE 0.40±0.07 0.04±0.08 0.07±0.08 0.09±0.09 0.11±0.09 0.01

Cholesterol — mg/dlMean ±SE 9.7±2.1 5.3±2.1 ¡2.1±1.9 0.5±2.3 ¡9.2±5.2Median 8.5 3.5 ¡3.0 0.5 ¡1.0Exposure¡adjusted mean ±SE 9.4±2.4 6.6±2.4 ¡1.3±2.4 1.5±2.7 ¡8.2±3.2 <0.001

Triglycerides — mg/dlMean ±SE 42.9±8.4 19.2±10.6 ¡2.6±6.3 8.3±11.5 ¡18.1±9.4Median 33.5 17.5 3.0 2.0 ¡7.0Exposure-adjusted mean ±SE 40.5±8.9 21.2±9.2 ¡2.4±9.1 9.2±10.1 ¡16.5±12.2 <0.001



n engl j med

353;12


22, 2005


1221

* Patients with tardive dyskinesia were excluded from the perphenazine group.† P values, presented for descriptive purposes, are from a test with 4 df comparing all treatment groups. P values for reasons of discontinua-

tion are from a chi-square test. P values for percentages are from a Poisson regression accounting for differential exposure times and adjust-ing for whether the patient had had an exacerbation in the preceding three months. P values for a prolonged corrected QT interval and new cataracts are from Fisher’s exact test. P values for laboratory values are based on a ranked analysis of covariance with adjustment for whether the patient had had an exacerbation in the preceding three months and the duration of exposure to the study drug during phase 1. P values for the change in weight and the corrected QT interval are based on an analysis of covariance with adjustment for whether the patient had had an exacerbation in the preceding three months and the duration of exposure to study drug during phase 1.

‡ Percentages are based on the number of female patients: 92 in the olanzapine group, 82 in the quetiapine group, 88 in the risperidone group, 62 in the perphenazine group, and 56 in the ziprasidone group.

§ Scores of 2 or more on the Abnormal Involuntary Movement Scale (AIMS) global severity score indicate at least mild severity of abnormal movements. Percentages are based on the number of patients without tardive dyskinesia who had an AIMS score of less than 2 at baseline and at least one post-baseline measurement. Scores of 3 or more for the global clinical assessment of the Barnes Akathisia Rating Scale in-dicate at least moderate severity of akathisia. Percentages are based on the number of patients who had a Barnes score of less than 3 at baseline and at least one post-baseline measurement. Average scores of 1 or more for the Simpson–Angus Extrapyramidal Signs Scale indicate at least mild severity of extrapyramidal signs. Percentages are based on the number of patients who had an average score for the Simpson–Angus Extrapyramidal Signs Scale of less than 1 at baseline and at least one post-baseline measurement.

¶ Percentages for weight gain are based on the number of patients with a baseline and at least one post-baseline measurement. To convert val-ues for weight to kilograms, divide by 2.2. The range for weight change is the 5th to 95th percentile, which excludes extreme outliers.

¿ Patients were instructed to fast; nonfasting results were not excluded. Change was determined as the difference between the baseline value and the average of the two highest post-baseline values. The exposure-adjusted mean is the least-squares mean from an analysis of co-variance adjusting for whether the patient had had an exacerbation in the preceding three months and for duration of exposure to study drug during phase 1. Since the measurement of glycosylated hemoglobin was added to the protocol as part of a protocol amendment, the numbers of patients are smaller for this test: 151 in the olanzapine group, 137 in the quetiapine group, 139 in the risperidone group, 107 in the perphenazine group, and 89 in the ziprasidone group. The analysis of all other laboratory variables included 286 patients in the olan-zapine group, 268 in the quetiapine group, 262 in the risperidone group, 212 in the perphenazine group, and 143 in the ziprasidone group. To convert values for blood glucose to millimoles per liter, multiply by 0.05551. To convert values for cholesterol to millimoles per liter, mul-tiply by 0.02586. To convert values for triglycerides to millimoles per liter, multiply by 0.01129.

** Percentages are based on the number of patients who had a normal corrected QT interval at baseline (450 msec or less for men and 470 msec or less for women) and at least one post-baseline measurement.

††Percentages are based on the number of patients with a post-baseline assessment.‡‡Percentages are based on the number of patients with data available: 333 in the olanzapine group, 333 in the quetiapine group, 340 in the ris-

peridone group, 259 in the perphenazine group, and 184 in the ziprasidone group.§§ Trazodone was excluded.¶¶Trazodone was included.

Table 3. (Continued.)

OutcomeOlanzapine

(N=336)Quetiapine(N= 337)

Risperidone(N=341)

Perphenazine(N=261)*

Ziprasidone(N=185) P Value†

Change from baseline in laboratory values¿ (cont.)

Prolactin — ng/ml

Mean ±SE ¡6.1±1.2 ¡9.3±1.4 15.4±1.5 0.4±1.7 ¡4.5±1.6

Median ¡0.9 ¡2.7 9.2 1.4 ¡2.4

Exposure-adjusted mean ±SE ¡8.1±1.4 ¡10.6±1.4 13.8±1.4 ¡1.2±1.6 ¡5.6±1.9 <0.001

Electrocardiographic findings**

Mean (±SE) change in corrected QT interval from base-line to last observation — msec

1.2±1.8 5.9±1.9 0.2±1.8 1.4±2.0 1.3±2.2 0.25

Prolonged corrected QT interval — no./total no. (%) 0/231 6/214 (3) 7/218 (3) 2/172 (1) 2/148 (1) 0.03

New cataracts — no./total no. (%)†† 3/272 (1) 1/258 (<1) 2/260 (1) 1/210 (<1) 0/142 0.81

Medications added — no. (%)‡‡

Lithium 1 (<1) 4 (1) 2 (<1) 3 (1) 1 (<1) 0.42

Anticonvulsants 10 (3) 11 (3) 13 (4) 9 (3) 8 (4) 0.63

Antidepressants§§ 40 (12) 28 (8) 54 (16) 28 (11) 26 (14) 0.03

Hypnotics, sedatives¶¶ 22 (7) 14 (4) 32 (9) 23 (9) 17 (9) 0.03

Anxiolytics 31 (9) 46 (14) 33 (10) 38 (15) 27 (15) <0.001

Anticholinergic agents 25 (7) 11 (3) 32 (9) 26 (10) 14 (8) 0.01

Oral glucose-lowering drugs, insulin 12 (4) 7 (2) 8 (2) 5 (2) 4 (2) 0.95

Cholestatin drugs 15 (4) 14 (4) 11 (3) 7 (3) 2 (1) 0.28



n engl j med 353;12 www.nejm.org september 22, 2005

The new england journal of medicine

1222

gain and increases in glycosylated hemoglobin,cholesterol, and triglycerides, changes that may haveserious implications with respect to medical comor-bidity such as the development of the metabolicsyndrome. How clinicians, patients, families, andpolicymakers evaluate the trade-offs between effi-cacy and side effects, as well as drug prices, will de-termine future patterns of use.

Supported by a grant (N01 MH90001) from the NIMH and by theFoundation of Hope of Raleigh, N.C. AstraZeneca Pharmaceuticals,Bristol-Myers Squibb, Forest Pharmaceuticals, Janssen Pharmaceu-tica, Eli Lilly, Otsuka Pharmaceutical, Pfizer, Zenith Goldline Phar-maceuticals, Schering-Plough, and Novartis provided medicationsfor the studies.

Dr. Lieberman reports having received research funding fromAstraZeneca Pharmaceuticals, Bristol-Myers Squibb, GlaxoSmith-Kline, Janssen Pharmaceutica, and Pfizer and consulting and educa-tional fees from AstraZeneca Pharmaceuticals, Bristol-Myers Squibb,Eli Lilly, Forest Pharmaceuticals, GlaxoSmithKline, Janssen Pharma-ceutica, Novartis, Pfizer, and Solvay. Dr. Stroup reports having re-ceived research funding from Eli Lilly and consulting fees from Jans-sen Pharmaceutica, GlaxoSmithKline, and Bristol-Myers Squibb.Dr. McEvoy reports having received research funding from Astra-Zeneca, Forest Research Institute, Eli Lilly, Janssen Pharmaceutica,and Pfizer; consulting or advisory-board fees from Pfizer and Bristol-Myers Squibb; and lecture fees from Janssen Pharmaceutica andBristol-Myers Squibb. Dr. Swartz reports having received researchfunding from Eli Lilly and consulting and educational fees from

AstraZeneca Pharmaceuticals, Bristol-Myers Squibb, Eli Lilly, andPfizer. Dr. Rosenheck reports having received research funding fromAstraZeneca Pharmaceuticals, Bristol-Myers Squibb, and Eli Lillyand consulting fees from Bristol-Myers Squibb, Eli Lilly, and Jans-sen Pharmaceutica. Dr. Perkins reports having received researchfunding from AstraZeneca Pharmaceuticals, Bristol-Myers Squibb,Otsuka Pharmaceutical, Eli Lilly, Janssen Pharmaceutica, and Pfizerand consulting and educational fees from AstraZeneca Pharmaceu-ticals, Bristol-Myers Squibb, Eli Lilly, Janssen Pharmaceuticals, andPfizer. Dr. Keefe reports having received research funding from Astra-Zeneca, Eli Lilly, and Janssen Pharmaceutica; consulting or advisory-board fees from Forest Pharmaceuticals, Eli Lilly, Janssen Pharma-ceutica, Pfizer, and Bristol-Myers Squibb; and lecture fees from EliLilly and Janssen Pharmaceutica. Dr. Sonia Davis is an employee ofQuintiles. Dr. Clarence Davis reports having received consultingfees from Eli Lilly and Quintiles. Dr. Lebowitz is a former employeeand Ms. Severe and Dr. Hsiao are current employees of the NIMH.

We are indebted to the 1493 participants in the CATIE study; tothe late Mahmoud A. Parsa, M.D., of the Department of Psychiatry,Case Western Reserve University, Cleveland; to Grayson S. Norquist,M.D., M.S.P.H., previously director of the Division of Services andIntervention Research, NIMH, and currently chairman of the De-partment of Psychiatry and Human Behavior, University of Missis-sippi Medical Center, Jackson; to Ingrid Rojas-Eloi, B.S., projectmanager of the CATIE study, and Tiffany Harris, staff assistant, De-partment of Psychiatry, School of Medicine, University of NorthCarolina at Chapel Hill; to Allison Andors, Ph.D., director of GrantsDevelopment at the Research Foundation for Mental Hygiene,New York State Psychiatric Institute; and to the Quintiles CATIEproject team.

appendix

The CATIE Study Investigators Group includes the following: L. Adler, Clinical Insights, Glen Burnie, Md.; M. Bari, Synergy Clinical Re-search, Chula Vista, Calif.; I. Belz, Tri-County/Mental Health and Mental Retardation Services, Conroe, Tex.; R. Bland, Southern IllinoisUniversity School of Medicine, Springfield; T. Blocher, Mental Health and Mental Retardation Authority of Harris County, Houston; B. Bol-yard, Cox North Hospital, Springfield, Mo.; A. Buffenstein, Queen’s Medical Center, Honolulu; J. Burruss, Baylor College of Medicine,Houston; M. Byerly, University of Texas Southwestern Medical Center at Dallas, Dallas; J. Canive, Albuquerque Veterans Affairs MedicalCenter, Albuquerque, N.M.; S. Caroff, Behavioral Health Service, Philadelphia; C. Casat, Behavioral Health Center, Charlotte, N.C.; E.Chavez-Rice, El Paso Community Mental Health and Mental Retardation Center, El Paso, Tex.; J. Csernansky, Washington University Schoolof Medicine, St. Louis; P. Delgado, University Hospitals of Cleveland, Cleveland; R. Douyon, Veterans Affairs Medical Center, Miami; C.D’Souza, Connecticut Mental Health Center, New Haven; I. Glick, Stanford University School of Medicine, Stanford, Calif.; D. Goff, Mas-sachusetts General Hospital, Boston; S. Gratz, Eastern Pennsylvania Psychiatric Institute, Philadelphia; G.T. Grossberg, Saint Louis Univer-sity School of Medicine–Wohl Institute, St. Louis; M. Hale, New Britain General Hospital, New Britain, Conn.; M. Hamner, Medical Univer-sity of South Carolina and Veterans Affairs Medical Center, Charleston; R. Jaffe, Belmont Center for Comprehensive Treatment,Philadelphia; D. Jeste, University of California, San Diego, Veterans Affairs Medical Center, San Diego; A. Kablinger, Louisiana State Uni-versity Health Sciences Center, Shreveport; A. Khan, Psychiatric Research Institute, Wichita, Kans.; S. Lamberti, University of RochesterMedical Center, Rochester, N.Y.; M.T. Levy, Staten Island University Hospital, Staten Island, N.Y.; J.A. Lieberman, University of North Caro-lina School of Medicine, Chapel Hill; G. Maguire, University of California Irvine, Orange; T. Manschreck, Corrigan Mental Health Center,Fall River, Mass.; J. McEvoy, Duke University Medical Center, Durham, N.C.; M. McGee, Appalachian Psychiatric Healthcare System, Ath-ens, Ohio; H. Meltzer, Vanderbilt University Medical Center, Nashville; A. Miller, University of Texas Health Science Center at San Antonio,San Antonio; D.D. Miller, University of Iowa, Iowa City; H. Nasrallah, University of Cincinnati Medical Center, Cincinnati; C. Nemeroff, Em-ory University School of Medicine, Atlanta; S. Olson, University of Minnesota Medical School, Minneapolis; G.F. Oxenkrug, St. Elizabeth’sMedical Center, Boston; J. Patel, University of Massachusetts Health Care, Worcester; F. Reimherr, University of Utah Medical Center, SaltLake City; S. Riggio, Mount Sinai Medical Center–Bronx Veterans Affairs Medical Center, Bronx, N.Y.; S. Risch, University of California, SanFrancisco, San Francisco; B. Saltz, Mental Health Advocates, Boca Raton, Fla.; T. Simpatico, Northwestern University, Chicago; G. Simp-son, University of Southern California Medical Center, Los Angeles; M. Smith, Harbor–UCLA Medical Center, Torrance, Calif.; R. Sommi,University of Missouri, Kansas City; R.M. Steinbook, University of Miami School of Medicine, Miami; M. Stevens, Valley Mental Health, SaltLake City; A. Tapp, Veterans Affairs Puget Sound Health Care System, Tacoma, Wash.; R. Torres, University of Mississippi, Jackson; P.Weiden, SUNY Downstate Medical Center, Brooklyn, N.Y.; J. Wolberg, Mount Sinai Medical Center, New York.



n engl j med 353;12 www.nejm.org september 22, 2005


1223

references

1. Miyamoto S, Duncan GE, Marx CE, Lie-berman JA. Treatments for schizophrenia:a critical review of pharmacology and mech-anisms of action of antipsychotic drugs.Mol Psychiatry 2005;10:79-104.2. Kane J, Honigfeld G, Singer J, MeltzerH. Clozapine for the treatment-resistantschizophrenic: a double-blind comparisonwith chlorpromazine. Arch Gen Psychiatry1988;45:789-96.3. Leucht S, Pitschel-Walz G, Abraham D,Kissling W. Efficacy and extrapyramidalside-effects of the new antipsychotics olan-zapine, quetiapine, risperidone, and sertin-dole compared to conventional antipsy-chotics and placebo: a meta-analysis ofrandomized controlled trials. Schizophr Res1999;35:51-68.4. Geddes J, Freemantle N, Harrison P,Bebbington P. Atypical antipsychotics in thetreatment of schizophrenia: systematic over-view and meta-regression analysis. BMJ2000;321:1371-6.5. Wahlbeck K, Tuunainen A, Ahokas A,Leucht S. Dropout rates in randomised anti-psychotic drug trials. Psychopharmacology(Berl) 2001;155:230-3.6. Davis JM, Chen N, Glick ID. A meta-analysis of the efficacy of second-generationantipsychotics. Arch Gen Psychiatry 2003;60:553-64.7. Leucht S, Wahlbeck K, Hamann J, KisslingW. New generation antipsychotics versuslow-potency conventional antipsychotics:a systematic review and meta-analysis. Lan-cet 2003;361:1581-9.8. Leucht S, Barnes TRE, Kissling W, EngelRR, Correll C, Kane JM. Relapse preventionin schizophrenia with new-generation anti-psychotics: a systematic review and explor-atory meta-analysis of randomized, con-trolled trials. Am J Psychiatry 2003;160:1209-22.9. Wahlbeck K, Cheine M, Essali A, AdamsC. Evidence of clozapine’s effectiveness inschizophrenia: a systematic review and meta-analysis of randomized trials. Am J Psychia-try 1999;156:990-9.10. Chakos M, Lieberman J, Hoffman E,Bradford D, Sheitman B. Effectiveness ofsecond-generation antipsychotics in patientswith treatment-resistant schizophrenia: a re-view and meta-analysis of randomized trials.Am J Psychiatry 2001;158:518-26.11. Tuunainen A, Wahlbeck K, Gilbody S.Newer atypical antipsychotic medication incomparison to clozapine: a systematic re-view of randomized trials. Schizophr Res2002;56:1-10.12. Rosenheck R, Perlick D, Bingham S, et

al. Effectiveness and cost of olanzapine andhaloperidol in the treatment of schizophre-nia: a randomized controlled trial. JAMA2003;290:2693-702.13. Tollefson GD, Sanger TM. Negativesymptoms: a path analytic approach to adouble-blind, placebo- and haloperidol-controlled clinical trial with olanzapine. AmJ Psychiatry 1997;154:466-74.14. Keefe RS, Silva SG, Perkins DO, Lieber-man JA. The effects of atypical antipsychot-ic drugs on neurocognitive impairment inschizophrenia: a review and meta-analysis.Schizophr Bull 1999;25:201-22.15. Tollefson GD, Sanger TM, Lu Y, ThiemeME. Depressive signs and symptoms inschizophrenia: a prospective blinded trial ofolanzapine and haloperidol. Arch Gen Psy-chiatry 1998;55:250-8. [Erratum, Arch GenPsychiatry 1998;55:1052.]16. Csernansky JG, Mahmoud R, BrennerR. A comparison of risperidone and halo-peridol for the prevention of relapse in pa-tients with schizophrenia. N Engl J Med2002;346:16-22. [Erratum, N Engl J Med2002;346:1424.]17. Allison DB, Mentore JL, Heo M, et al.Antipsychotic-induced weight gain: a com-prehensive research synthesis. Am J Psychi-atry 1999;156:1686-96.18. Henderson DC, Cagliero E, CopelandPM, et al. Glucose metabolism in patientswith schizophrenia treated with atypical an-tipsychotic agents: a frequently sampled in-travenous glucose tolerance test and mini-mal model analysis. Arch Gen Psychiatry2005;62:19-28.19. Koro CE, Fedder DO, L’Italien GJ, et al.An assessment of the independent effectsof olanzapine and risperidone exposureon the risk of hyperlipidemia in schizophren-ic patients. Arch Gen Psychiatry 2002;59:1021-6.20. Atypical antipsychotics — generatingevidence to inform policy and practice.London: IMS Health, 2002. (Accessed August26, 2005, at http://research.imshealth.com/research/research_schizophrenia.htm.)21. Harrington C, Gregorian R, Gemmen E,et al. Access and utilization of new anti-depressant and antipsychotic medications.Falls Church, Va.: Lewin Group, 2000. (Ac-cessed August 26, 2005, at http://aspe.hhs.gov/search/health/reports/Psychmedaccess/index.htm#TOC.)22. Tunis SR, Stryer DB, Clancy CM. Practi-cal clinical trials: increasing the value ofclinical research for decision making inclinical and health policy. JAMA 2003;290:1624-32.

23. Lebowitz BD, Vitiello B, Norquist GS.Approaches to multisite clinical trials: theNational Institute of Mental Health perspec-tive. Schizophr Bull 2003;29:7-13.24. Stroup TS, McEvoy JP, Swartz MS, et al.The National Institute of Mental HealthClinical Antipsychotic Trials of InterventionEffectiveness (CATIE) project: schizophre-nia trial design and protocol development.Schizophr Bull 2003;29:15-31.25. Swartz MS, Perkins DO, Stroup TS,McEvoy JP, Nieri JM, Haak DC. Assessingclinical and functional outcomes in theClinical Antipsychotic Trials of InterventionEffectiveness (CATIE) schizophrenia trial.Schizophr Bull 2003;29:33-43.26. Keefe RS, Mohs RC, Bilder RM, et al.Neurocognitive assessment in the ClinicalAntipsychotic Trials of Intervention Effec-tiveness (CATIE) project schizophrenia trial:development, methodology, and rationale.Schizophr Bull 2003;29:45-55.27. Rosenheck R, Doyle J, Leslie D, FontanaA. Changing environments and alternativeperspectives in evaluating the cost-effective-ness of new antipsychotic drugs. SchizophrBull 2003;29:81-93.28. Davis SM, Koch GG, Davis CE, LaVangeLM. Statistical approaches to effectivenessmeasurement and outcome-driven re-ran-domizations in the Clinical AntipsychoticTrials of Intervention Effectiveness (CATIE)studies. Schizophr Bull 2003;29:73-80.29. Cox DR. Regression models and life-tables. J R Stat Soc [B] 1972;34:187-220.30. Hochberg Y. A sharper Bonferroni pro-cedure for multiple tests of significance. Bio-metrika 1988;75:800-2.31. Rosenheck RA. Open forum: effective-ness versus efficacy of second-generationantipsychotics: haloperidol without anti-cholinergics as a comparator. Psychiatr Serv2005;56:85-92.32. Citrome L, Volavka J. Optimal dosing ofatypical antipsychotics in adults: a review ofthe current evidence. Harv Rev Psychiatry2002;10:280-91.33. Davis JM, Chen N. Dose response anddose equivalence of antipsychotics. J ClinPsychopharmacol 2004;24:192-208.34. Intercontinental Medical Systems Na-tional Disease and Therapeutic Index. Ply-mouth Meeting, Pa.: IMS Health, January2001-December 2004.35. Lehman AF, Lieberman JA, Dixon LB,et al. Practice guideline for the treatment ofpatients with schizophrenia, second edition.Am J Psychiatry 2004;161:Suppl:1-56.Copyright © 2005 Massachusetts Medical Society.



Articles

www.thelancet.com Vol 379 February 25, 2012 721

Lancet 2012; 379: 721–28

Published OnlineJanuary 20, 2012DOI:10.1016/S0140-6736(11)61516-X

See Comment page 690

Department of Psychiatry, University of Oxford, Warneford Hospital, Oxford, UK (R F McKnight BMBCh, K Budge MSc, S Stockton BA, G M Goodwin FMedSci, Prof J R Geddes MD); and University Department of Psychiatry, Solaris, Sainte-Marguerite Hospital, Marseille, France (M Adida PhD)

Correspondence to:Prof John R Geddes, Department of Psychiatry, University of Oxford, Warneford Hospital, Oxford OX3 7JX, [email protected]

Lithium toxicity profi le: a systematic review and meta-analysisRebecca F McKnight, Marc Adida, Katie Budge, Sarah Stockton, Guy M Goodwin, John R Geddes

SummaryBackground Lithium is a widely used and eff ective treatment for mood disorders. There has been concern about its safety but no adequate synthesis of the evidence for adverse eff ects. We aimed to undertake a clinically informative, systematic toxicity profi le of lithium.

Methods We undertook a systematic review and meta-analysis of randomised controlled trials and observational studies. We searched electronic databases, specialist journals, reference lists, textbooks, and conference abstracts. We used a hierarchy of evidence which considered randomised controlled trials, cohort studies, case-control studies, and case reports that included patients with mood disorders given lithium. Outcome measures were renal, thyroid, and parathyroid function; weight change; skin disorders; hair disorders; and teratogenicity.

Findings We screened 5988 abstracts for eligibility and included 385 studies in the analysis. On average, glomerular fi ltration rate was reduced by –6·22 mL/min (95% CI –14·65 to 2·20, p=0·148) and urinary concentrating ability by 15% of normal maximum (weighted mean diff erence –158·43 mOsm/kg, 95% CI –229·78 to –87·07, p<0·0001). Lithium might increase risk of renal failure, but the absolute risk was small (18 of 3369 [0·5%] patients received renal replacement therapy). The prevalence of clinical hypothyroidism was increased in patients taking lithium compared with those given placebo (odds ratio [OR] 5·78, 95% CI 2·00–16·67; p=0·001), and thyroid stimulating hormone was increased on average by 4·00 iU/mL (95% CI 3·90–4·10, p<0·0001). Lithium treatment was associated with increased blood calcium (+0·09 mmol/L, 95% CI 0·02–0·17, p=0·009), and parathyroid hormone (+7·32 pg/mL, 3·42–11·23, p<0·0001). Patients receiving lithium gained more weight than did those receiving placebo (OR 1·89, 1·27–2·82, p=0·002), but not those receiving olanzapine (0·32, 0·21–0·49, p<0·0001). We recorded no signifi cant increased risk of congenital malformations, alopecia, or skin disorders.

Interpretation Lithium is associated with increased risk of reduced urinary concentrating ability, hypothyroidism, hyperparathyroidism, and weight gain. There is little evidence for a clinically signifi cant reduction in renal function in most patients, and the risk of end-stage renal failure is low. The risk of congenital malformations is uncertain; the balance of risks should be considered before lithium is withdrawn during pregnancy. Because of the consistent fi nding of a high prevalence of hyperparathyroidism, calcium concentrations should be checked before and during treatment.

Funding National Institute for Health Research Programme Grant for Applied Research.

IntroductionLithium is the most eff ective long-term therapy for bipolar disorder, protecting against both depression and mania and reducing the risk of suicide and short-term mortality.1–3 Although effi cacious, lithium has some clinical disadvantages: it has a narrow therapeutic index requiring routine monitoring of serum concentrations and endocrine and renal function; a slow onset of action in acute mania; and acute eff ects of thirst, unpleasant taste, and tremor. Because lithium has always been an unpatented, cheap drug, it is not commercially promoted and the potential for adverse eff ects has been a substantial deterrent to use. Alternative drugs for bipolar disorder have increasingly been proposed, licensed, and adopted into clinical practice even when evidence for effi cacy is modest and often limited to one pole of bipolar illness.

Particular concerns have been the eff ect of lithium on renal function and the risk from teratogenicity. Lithium commonly induces a clinically evident nephrogenic diabetes insipidus,4,5 which would be explained by actions

on tubular renal function, but of more concern is the speculative description of a specifi c lithium nephropathy6,7 or disease of the renal glomerulus. However, the extent of reduction of glomerular renal function in a typical patient and the true long-term increase in risk of renal failure in well monitored patients remain poorly quantifi ed. The risk of congenital malformations is generally thought to be high. One study reported a 400-fold increased risk of Ebstein’s anomaly,8 and present clinical practice recom-mendations have been to avoid lithium in pregnancy when possible. An up-to-date estimate of the true risks of lithium together with a systematic assessment of the associated renal problems has not been available.

Evidence has confi rmed the important therapeutic benefi ts of lithium relative to some of the alternative drugs that have replaced it, which might lead to wider use of lithium.1 Clinicians and patients therefore need accurate evidence of harms and benefi ts. We report a systematic review and meta-analysis of studies investigating the association between lithium and all

Articles

722 www.thelancet.com Vol 379 February 25, 2012

reported major adverse eff ects, to provide a clinically informative systematic toxicity profi le for lithium.

MethodsSearch strategy and selection criteriaWe searched Medline (1966–2010), Medline In-Process and other non-indexed citations (from 1966 to October, 2010), Embase (1980–2010), the Cumulative Index to Nursing and Allied Health Literature (1982–2010), PsycINFO (1806–2010), the Cochrane Library database (inception–2010), Biosis Previews (1926–2010), TOXNET database (inception–2010; webappendixwebappendix), and archives of the journals Lithium, Lithium Therapy Monographs, and Teratology (search terms listed in webappendix). All relevant references were checked for additional and unpublished citations. Major textbooks of mood dis orders and conference abstracts were hand-searched. We contacted pharma-ceutical companies that market lithium, relevant clinicians, and authors of trials with incompletely reported data. All studies were assessed for meeting inclusion criteria, and those used for analysis were reviewed by a second researcher.

Studies were included in the review if they investigated one or more of the adverse events of interest. Randomised controlled trials (RCTs) comparing lithium with placebo, no treatment, or other drug therapies in patients with depression or bipolar disorder were considered most reliable if they included safety data for adverse eff ects, followed by prospective cohort studies comparing patients given lithium with those not given lithium, and then case-control studies. In the absence of controlled studies, we included uncontrolled prospective studies following up patients with depression or bipolar disorder given lithium and, fi nally, individual case reports. For each outcome, all studies meeting inclusion criteria were assessed and tabulated, but only the highest available form of evidence was included in the formal analysis. When only poor quality data from a higher level of evidence were available, we routinely included the next level down. For adverse events that often occur after months or even years of treatment, observational studies are often more informative than are RCTs.

OutcomesThe main outcomes investigated were: renal function (glomerular fi ltration rate [GFR, normal >90 mL/min], renal concentrating ability [maximum urinary concen-trating ability, normal 800–1200 mOsm/kg]); thyroid function (thyroid stimulating hormone [TSH, normal 0·5–5·7 IU/mL], subclinical hypothyroidism [raised TSH with normal thyroxine] or clinical hypothyroidism [raised TSH and low thyroxine], or hyperthyroidism [depressed TSH and high thyroxine]);9 parathyroid function (total calcium [normal 2·1–2·8 mmol/L] and parathyroid hormone [PTH, normal 10–70 pg/mL]); bodyweight (clinically signifi cant change in bodyweight [>7% total

weight in kg]); hair disorders; skin disorders; and teratogenicity (risk of major congenital and cardiac malformations in infants exposed to lithium in utero).

We judged study quality by assessing design aspects likely to introduce bias—ie, method of randomisation and concealment of treatment allocation, blinding, length of follow-up, reporting withdrawals and dropouts, and method of analysis for RCTs; and likelihood of measure-ment bias, handling of confounding, and loss to follow-up for observational studies. Authors were contacted when published reports did not contain adequate details.

Statistical analysisWhen appropriate, data from individual trials were pooled by meta-analysis with STATA (version 11.1). Both Mantel-Haenszel fi xed and DerSimonian and Laird random eff ects models were used to assess the degree to which results were robust to the choice of statistical model. Non-standard units were converted to standard international units. Continuous data were combined to produce weighted mean diff erences (WMDs, for common measures) and standardised WMDs (for heterogeneous measures). Dichotomous and categorical data were combined to produce odds ratios (ORs) and absolute risk diff erences. Heterogeneity between study-specifi c estimates was investigated and, when important heterogeneity was expected or identifi ed, sources for such variation were sought with meta-regression. We undertook sensitivity analyses to investigate the eff ect of exclusion of studies of inferior quality or with highly discrepant results.

Role of the funding sourceThe sponsor of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. RFMcK and JRG had full access to all the data, and JRG had fi nal responsibility for the decision to submit for publication.

ResultsThe search process identifi ed 5988 records, 385 of which fi tted the inclusion criteria and were included for analysis (fi gure 1). The quality of evidence available varied between outcomes. High quality evidence was sparse: we identifi ed only one systematic review (which was excluded from analysis because the data was used from the original studies included within it) and 22 RCTs. Most studies were case-control, uncontrolled cohort, or cross-sectional studies (n=197) or case reports (166). When cohort studies were reported in several publications, we analysed only the most complete set of data to avoid double counting cases. Studies published in English, French, and German were included; no studies in other languages met inclusion criteria.

30 studies investigated the eff ect of lithium GFR or maximum urinary concentrating ability, or both: nine case-control studies and 21 uncontrolled-cohort studies (webappendix). The uncontrolled cohorts could not be used for quantitative analysis because of the absence of

See Online for webappendix

Articles


data for within-patient change. The data were not ade-quate to analyse the eff ect of age or concomitant drugs (including diuretics). Overall, however, the results showed a small (0–5 mL/min) reduction in GFR over a mean observation time of 1 year (webappendix). Meta-analysis of case-control studies (cases=372, controls=307) showed that the GFR of patients taking lithium was lower than that of matched controls (fi gure 2). Maximum urinary concentrating ability was reduced by about 15% in patients taking lithium compared with controls (fi gure 3).

Data for the most clinically important outcome, renal failure, were scarce. The only substantial cohort study of patients on a lithium register reported 18 of 3369 (0·5%) as being treated with renal replacement therapy, compared with 0·2% of the Swedish general population.10,11

We identifi ed 77 studies that reported the eff ects of lithium on thyroid function (webappendix): four RCTs,

16 case-control studies, 15 cohort studies, 20 cross-sectional reports, and 22 case reports. Because the RCTs collected heterogeneous data and the cohorts were uncontrolled, we used the case-control studies for analysis. Many studies before 1980 reported measures that were incompatible with more recent studies; these studies are shown in the webappendix but could not be used for analysis.

Eight studies compared the prevalence of subclinical or clinical hypothyroidism in patients given lithium (n=1402) for a mean of 70·1 months (SD 2·6) with the prevalence in controls (n=1032). Meta-analysis showed more hypothyroidism in patients given lithium than in controls (fi gure 4). The relative risk increased when only cases of clinical hypothyroidism were included (OR 6·05, 95% CI 2·72–13·37, p<0·0001; heterogeneity χ²=7·09 [df=5], p=0·21).

30 studies assessed renal function21 Co‡

9 CC‡

385 studies included in qualitative analysis

1645 full-text articles assessed for eligibility 1260 full-text articles excluded†

4343 references excluded*5988 records screened

77 studies assessedthyroid function

4 RCT15 Co‡16 CC‡20 CS22 CR

60 studies assessedparathyroid function

4 Co‡14 CC‡

6 CS‡36 CR

24 studies assessedhair

2 RCT 3 CC 5 CS 14 CR

77 studies assessedskin

2 RCT‡1 Co3 CC3 CS

68 CR

55 studies assessedweight14 RCT‡23 Co‡

9 CC9 CS

62 studies assessedteratogenicity

7 Co7 CC

48 CR

698 additional records identified through other sources

5290 references identified throughelectronic database searching(duplicates excluded)

Figure 1: Study selectionCo=cohort study. CC=case-control study. RCT=randomised controlled trial. CS=cross-sectional study. CR=case report. *Included animal studies, non-biological science studies, and human studies of lithium not reporting adverse events. †Included patients with diagnoses other than mood disorders; reviews, editorials, and commentaries; types of study not in inclusion criteria; outcome measures other than those in inclusion criteria; and RCTs in which data collected for adverse event did not include any of our outcome measures. ‡Data included in quantitative analysis.

WMD (95% CI)

Hullin (1979)

Bendz (1985)

Hetmar (1987a)

Bendz (1996)

Coskunol (1997)

Turan (2002)

Overall (χ2=11·65 [df 4], I2=57·1%, p=0·040)

–20·20 (–41·72 to 1·32)

–4·00 (–20·69 to 12·69)

–12·60 (–22·34 to –2·86)

–9·00 (–12·08 to –5·92)

2·60 (–36·09 to 41·29)

24·94 (3·29 to 46·59)

–6·22 (–14·65 to 2·20); p=0·148

Weight (%)

10·80

15·00

24·52

34·79

4·20

10·70

100·00

–30 –20 –10 0 10 20GFR (mL/min)

Figure 2: Meta-analysis of case-control studies comparing glomerular fi ltration rate (GFR) in patients given lithium versus controlWeights are from random-eff ects analysis. The webappendix provides the references for the included studies. WMD=weighted mean diff erence.

Articles


15 uncontrolled cohort studies (n=1085) measured a change in TSH over a mean of 18·5 months (SD 1·4); meta-analysis was not possible because of insuffi cient data (webappendix). Meta-analysis of the case-control studies (cases=645, controls=377) showed an increase in TSH concentrations in patients given lithium compared with controls (WMD 4·00 iU/mL, 95% CI 3·90–4·10, p<0·0001; heterogeneity χ²=1868·59 [df 10], p<0·0001).

Four case-control studies reported possible increased thyroid function (webappendix). Meta-analysis showed no evidence of a diff erence between those taking lithium (n=178) and controls (n=181; OR 1·46, 95% CI 0·23–9·35, p=0·69; heterogeneity χ²=1·34 [df 2], p=0·51). Data from the RCTs accorded with that from the observational studies: a meta-analysis of lithium versus placebo trials reported that 4% of patients given lithium developed hypothyroidism compared with none given placebo (webappendix).1

60 studies (no RCTs) reported the eff ect of lithium on parathyroid function, and results were consistent. We identifi ed four cohort studies, 14 case-control studies, 36 case reports, and six cross-sectional studies (web-appendix). Calcium and PTH were increased by 10% compared with normal values in patients given lithium (n=730) compared with controls (n=699; fi gures 5, 6).

Weight change was included in 14 RCTs comparing lithium with placebo or other drug treatment (webappendix). Clinically signifi cant weight gain (>7%) was more frequent in patients receiving lithium than in

those receiving placebo (OR 1·89, 95% CI 1·27–2·82, p=0·002; hetero geneity χ²=2·28 [df 4], p=0·69; webappendix). Weight gain was lower with lithium than with olanzapine (n=285; OR 0·32, 95% CI 0·21–0·49, p<0·0001; heterogeneity χ²=0·72 [df 1], p=0·39; webappendix).

24 publications reported an adverse eff ect of lithium on hair, 14 of which were case reports (webappendix). One RCT of lithium (n=91) versus placebo (n=94) for 12 months reported hair loss in seven of 91 (8%) patients in the lithium group compared with six of 94 (6%) in the placebo group,12 whereas another reported hair loss in one of 32 (3%) patients given lithium versus none of 28 given placebo.13

We identifi ed little high quality evidence supporting the association between lithium and skin disorders. 77 publications met inclusion criteria, 68 of which were case reports (webappendix). Two RCTs reported skin disorders within one combined analysis (webappendix). Meta-analysis showed no signifi cant diff erence in the prevalence of skin disorders between patients given lithium and those given placebo (OR 1·28, 95% CI 0·49–3·36, p=0·62; heterogeneity χ²=0·29 [df 1], p=0·59).14,15

We identifi ed 62 studies of the teratogenic potential of lithium: seven cohort studies, seven case-control studies, and 48 case reports (webappendix). Six case-control studies (n=264) measured the association between

WMD (95% CI)

Hullin (1979)

Bendz (1985)

Bendz (1996)

Coskunol (1997)

Overall (χ2=16·07 [df 3], I2=81·3%, p=0·001)

–70·00 (–171·27 to 31·27)

–68·00 (–162·06 to 26·06)

–211·00 (–254·76 to –167·24)

–229·00 (–269·41 to –188·59)

–158·43 (–229·78 to –87·07); p<0·0001

–300 –200 –100 0 100Umax (mOsm/kg)

Figure 3: Meta-analysis of case-control studies comparing maximum urinary concentrating ability (Umax) in patients given lithium versus controlWeights are from random-eff ects analysis. The webappendix provides the references for the included studies. WMD=weighted mean diff erence.

OR (95% CI) Events

Lithium Control

McLarty (1975)

Linstedt (1977)

Cho (1979)

Bocchetta (1991)

Deodhar (1999)

Kupka (2002)

Ozpoyraz (2002)

Van Melick (2010)

Overall (χ2=14·73 [df 7], I2=52·5%, p=0·040)

1/17

10/53

7/195

25/129

4/121

39/226

7/49

28/79

121/869

3·18 (0·12 to 83·76)

34·52 (1·97 to 603·91)

1·12 (0·23 to 5·52)

2·28 (0·50 to 10·45)

1·88 (0·10 to 36·00)

106·39 (6·50 to 1741·96)

16·41 (0·91 to 296·12)

7·23 (2·80 to 18·68)

5·78 (2·00 to 16·67); p=0·001

0/17

0/71

2/62

2/21

0/24

0/252

0/46

6/85

10/578

0·1 0·2 0·5 1 2 5 10 50 100OR (95% CI)

Figure 4: Meta-analysis of case-control studies comparing clinical hypothyroidism in patients given lithium versus control Weights are from random-eff ects analysis. The webappendix provides the references for the included studies. OR=odds ratio.

Articles


Ebstein’s anomaly and lithium. The odds of exposure to lithium in cases of Ebstein’s anomaly did not diff er signifi cantly from controls; however, estimates are unstable because of the low number of events (Peto OR 0·27, 95% CI 0·004–18·17, p=0·54; heterogeneity χ²=0·00 [df 1], p=0·96; Mantel-Haenszel OR 2·0, 95% CI 0·20–20·6, p=0·54; heterogeneity χ²=1·98 [df 1], p=0·16).

A case-control study of 10 698 infants born with any major congenital abnormality and 21 546 healthy controls showed no signifi cant association between lithium and congenital abnormalities (Peto OR 2·62, 95% CI 0·74–9·20, p=0·132; webappendix).16 The number of infants exposed to lithium was low in cases (six of 10 698) and controls (fi ve of 21 546).

DiscussionThe objective of this review was to synthesise what is known about the harmful eff ects of lithium. Findings from our study have shown that lithium is associated with

increased risk of reduced urinary concentrating ability, hypothyroidism, hyperparathyroidism, and weight gain. We recorded no signifi cant increased risk of congenital malformations, alopecia, or skin disorders, and little evidence for a clinically signifi cant reduction in renal function in most patients.

The main limitations of this study are the quality and quantity of the primary evidence. High-quality data from long-term randomised or controlled cohort studies were sparse, and the sample size of most included observational studies was quite small. Although included studies reported doses and concentrations of lithium that are consistent with modern use, and data mainly represent the eff ects of lithium within the generally accepted therapeutic range rather than at concentrations of toxicity, dose information was incompletely reported and any potential eff ect of dose could not be specifi cally addressed in the meta-analysis. This review cannot, therefore, establish the relative safety of low doses or

WMD (95% CI)

Christiansen (1976)

Christiansen (1978)

Toffaletti (1979)

Davis (1981)

Franks (1982)

McIntosh (1987)

Mallette (1989a)

Mallette (1989b)

Nordenstrom (1994)

Komatsu (1995)

Bendz (1996)

Haden (1997)

El Khoury (2002)

Overall (χ2=2299·48 [df 12], I2=99·5%, p<0·0001)

0·03 (0·01 to 0·05)

0·28 (0·28 to 0·28)

0·11 (0·09 to 0·13)

–0·01 (–0·04 to 0·02)

0·18 (0·15 to 0·21)

0·06 (0·02 to 0·10)

0·10 (0·00 to 0·20)

–0·01 (–0·13 to 0·11)

0·10 (0·08 to 0·12)

0·09 (0·03 to 0·15)

0·16 (0·15 to 0·17)

0·03 (0·00 to 0·06)

0·09 (0·00 to 0·18)

0·09 (0·02 to 0·17); p=0·009

Weight (%)

7·99

8·04

7·98

7·91

7·94

7·89

6·95

6·50

8·01

7·60

8·02

7·96

7·21

100·00

–0·2 –0·1 0 0·1 0·2 0·3Calcium (mmol/L)

Figure 5: Meta-analysis of case-control studies comparing calcium in patients given lithium versus controlWeights are from random-eff ects analysis. The webappendix provides the references for the included studies. WMD=weighted mean diff erence.

WMD (95% CI)

Christiansen (1976)

Christiansen (1978)

Franks (1982)

McIntosh (1987)

Mallette (1989a)

Mallette (1989b)

Nordenstrom (1994)

Komatsu (1995)

Haden (1997)

El Khoury (2002)

Overall (χ2=83·53 [df 9], I2=89·2%, p<0·0001)

8·00 (5·54 to 10·46)

8·00 (5·55 to 10·45)

9·75 (–18·53 to 38·03)

1·04 (0·10 to 1·98)

–0·20 (–7·80 to 7·40)

14·50 (5·92 to 23·08)

7·00 (–0·62 to 14·62)

2·70 (–4·22 to 9·62)

17·20 (11·58 to 22·82)

9·80 (–1·60 to 21·20)

7·32 (3·42 to 11·23); p<0·0001

14·00

14·01

1·69

14·70

9·49

8·64

9·48

10·12

11·34

6·55

100·00

Weight (%)

–20 –10 0 10 20 30 40Parathyroid hormone (pg/mL)

Figure 6: Meta-analysis of case-control studies comparing parathyroid hormone in patients given lithium versus controlWeights are from random-eff ects analysis. The webappendix provides the references for the included studies. WMD=weighted mean diff erence.

Articles


dosing on alternative days. Further more, most studies excluded patients with a history of lithium toxicity or did not provide appropriate information to separate out these individuals or link their clinical presentation to number of episodes of toxicity or dosing regimens.

The studies were published over 60 years from 1950, and were highly variable in design (webappendix) and execution (data not shown). Diagnostic criteria, standard treatments, methods, and accuracy of measure ment of physiological parameters have changed during that period. Moreover, because most cohort studies and RCTs did not use a patient group that was new to lithium or did not provide this information, length of follow-up was usually poorly defi ned so the average interval between fi rst starting lithium and the onset of adverse events is unknown or approximate.

Many of the important cohort studies had a high dropout rate with little explanation of the cause of withdrawal. Although we made every eff ort to include studies report ing the same parameter investigated with a similar meth od ology, diff erences could be attributable to unidentifi ed confounders.

We could not identify or obtain any unpublished data; therefore, there is a risk of publication bias. Nonetheless, we were able to locate a reasonable amount of evidence that allows cautious conclusions to be drawn about the

safety of lithium. The panel shows our recommendations for clinical practice.

Although GFR is impaired by lithium treatment, impairment is not clinically signifi cant in most patients. A maximum reduction in GFR of 5 mL/min represents only 5% of the minimum normal GFR. The patho-physiological mechanism underlying the eff ects of lithium on glomerular function is not understood.

Progressive reductions in glomerular function can lead to end-stage renal failure, and lithium is thought to play a direct part in this process. In the 1970s, chronic tubulointerstitial nephropathy was described in patients with lithium-related end-stage renal failure, but this pathology is non-specifi c and not reliably linked to lithium.6,7 The risk of end-stage renal failure might be increased compared with healthy controls but the absolute risk seems to be low (0·5%). The incidence of chronic kidney disease is rising, especially in ageing populations, with an excess in women and an association with hypertension and diabetes. Chronic kidney disease can lead to end-stage renal failure in 2% of cases. Identifi cation of the potential causal eff ect of lithium is diffi cult because of the confounding eff ects of diabetes and cardiovascular disease, which might lead to end-stage renal failure; but these disorders are also increased in patients with bipolar disorder compared with the general population.17 Large-scale epidemiological studies are needed that control for confounders (including age and sex) and model the eff ects of lithium dose, concomitant drugs (eg, angiotensin-converting-enzyme inhibitors, diuretics), treatment length, and repeated episodes of toxicity. Present clinical recommendations include recording of renal function before start of lithium therapy, and henceforth monitoring at intervals as short as 6 weeks. Because the absolute risk of end-stage renal failure is so low, yearly testing is probably suffi cient in the absence of clinical reasons to monitor more frequently.

Tubular renal function, expressed as urinary concen-trating ability, is reduced by about 15%. Unlike its poorly understood glomerular eff ects, the probable mechanism is known and relates to lithium’s inhibition of a G-protein-coupled pathway that is activated by anti diuretic hormone to increase aquaporin channels in the collecting ducts.18 Diff erential recovery of activation of these aquaporin channels accounts for the variable rate of recovery from lithium-induced diabetes insipidus on lithium with-drawal.19 Polyuria can limit acceptability in patients, but concentrating ability is often fully reversible on cessation of therapy.20

The rate of hypothyroidism is increased about six-fold in patients receiving lithium therapy. Whether the widespread practice of treating hypothyroidism in patients given lithium should be mandatory is unclear. Most such patients are asymptomatic and the diagnosis is purely biochemical. There is no evidence as to whether stopping lithium tends to lead to a recovery of thyroid function when function is very abnormal. In small studies, withdrawal of lithium has

Panel: Summary of recommended monitoring of lithium therapy in clinical practice

Before start of lithium therapy• The risk of major adverse events (as summarised in this

Article) should be discussed with the patient• A serum calcium should be added to baseline blood tests*• Uncertainty about risk of congenital malformations to

women of childbearing age should be explained*

During lithium therapy• Renal, parathyroid, and thyroid function (at least GFR,

TSH, calcium) should be repeated, at a minimum interval of every 12 months*, more frequently if an abnormal result is found or the patient has a family history of endocrine disease

• Blood tests should all be repeated immediately if there is a change in mood state (eg, mania)

• Occurrence of adverse eff ects (including skin and hair disorders) should be routinely recorded*

• Women who would like to conceive or have become pregnant while receiving lithium should be advised that the increased risk of congenital malformations is uncertain; patient and clinician should discuss the balance of risks between harm to the baby and maternal mood instability before making a decision to stop lithium therapy*

GFR=glomerular fi ltration rate. TSH=thyroid-stimulating hormone.*Changes to present therapy that we recommend; previous standard practice refers to UK guidelines.

Articles


led to normalisation of increases in T4 and decreased TSH.21 However, mood symptoms can be harder to treat when patients are in the low normal range of thyroid function,22 so treatment might be warranted on psychiatric grounds. Lithium is concentrated by the thyroid gland and has four poten tially negative eff ects on thyroid function: inhibition of iodine uptake, inhibition of iodotyrosine coupling, alteration in thyroglobulin structure, and inhibition of thyroxine secretion.23,24 TSH concentrations tend to be increased in response to the inhibitory eff ect on thyroxine availability.

Primary hyperparathyroidism was quite frequent in patients receiving lithium: an absolute risk of 10% (vs 0·1% of the general population25) is probably attributable to lithium’s inactivation of the calcium-sensing receptor and interference with intracellular second messenger signalling. This eff ect leads to an increased release of parathyroid hormone, which raises calcium concen trations in blood.26

Thyroid and parathyroid abnormalities occur in about 25% of patients receiving lithium therapy, and clinical monitoring should refl ect this fi nding. Guidelines for bipolar disorder make no mention of monitoring of calcium, which seems to be an important omission in view of the high absolute risk of hyperparathyroidism.27 Baseline blood tests before lithium is given should include TSH and calcium, and should be monitored every year or more frequently if clinical symptoms are reported. We recorded no evidence for a toxic eff ect of hypercalcaemia on renal function in patients given lithium, but it could contribute in view of the known risk of a decrease in renal function in long-term hyper-calcaemia.25 More research is needed to clarify the relation between lithium, calcium, and the kidney.

Several factors might explain the association between lithium and weight gain: its insulin-like properties in increasing cellular glucose uptake, increased thirst, direct stimulation of the hypothalamic appetite centre, and the induction of hypothyroidism. Lithium might also aff ect relevant neurotransmitter receptor function, although the eff ect is less than, for example, olanzapine, which inhibits histaminergic and serotonergic receptors in the brain.

The evidence that exposure to lithium is teratogenic is quite weak, and our fi ndings accord with the notion that the risk has been overestimated.28 Thus, the risk estimates were not signifi cant, although the upper confi dence limit is consistent with a clinically signifi cant increase in risk of congenital malformations. Therefore, uncertainty about the risk of harm remains. The present clinical recommendation is to avoid lithium in pregnancy. Our review suggests a sounder approach would be to explain the uncertainty of risk to women of childbearing age, considering the balance of risks between harm to the baby and maternal mood instability before continu-ation or stopping of lithium therapy.

We recorded no good evidence that lithium therapy increases the risk of skin and hair disorders, although

exacerbations of psoriasis, non-specifi c maculo papular rashes, acne, and alopecia have all been described anecdotally.

Lithium is dangerous in overdose, or under circum-stances that predispose to sodium or volume depletion.29 Case series of acute lithium toxicity report patients with clinical signs of toxicity at concentrations of 1·5 mEq/L or greater, and most patients who become toxic do so when ill (diarrhoea, vomiting, heart failure, renal failure, or surgery) or secondary to a drug interaction (non-steroidal anti-infl ammatory drugs, angiotensin-converting-enzyme inhibitors). The present guidance to monitor serum lithium concentrations every 3 months is mainly aimed at avoiding drift into the toxic range, but evidence to support this approach is scarce. Because few patients spon taneously develop toxic eff ects without a precipitating illness, yearly monitoring plus moni-toring of one-off lithium concentrations in high-risk circum stances might be more clinically relevant and cost eff ective.

In conclusion, clinical practice guidelines have long recommended lithium as a fi rst-line long-term treatment for bipolar disorder but its use has decreased, partly because of safety concerns. Evidence confi rming its effi cacy has led to suggestions that lithium should again be more widely used. This review provides a compre-hensive synthesis of the evidence of harm that should inform clinical decisions and draw attention to key questions in urgent need of further clarifi cation.

ContributorsRFMcK located references, extracted data, assisted with analyses and results interpretation, and drafted the report. KB assisted with locating references and data recording. MA initiated the review and helped to design the methodology, located references, and assessed quality. SS ran the electronic database searches. GMG initiated the review and helped to design the methodology. JRG initiated the review, helped to design the methodology, assessed the quality of studies, and did the analyses. All authors helped to interpret fi ndings and write the fi nal report. JRG is guarantor.

Confl icts of interestWe declare that we have no confl icts of interest.

AcknowledgmentsThis Article presents independent research commissioned by the National Institute for Health Research (NIHR) under Research Programme Grant for Applied Research: RP-PG-0108-10087. The views expressed in this publication are those of the authors and not necessarily those of the National Health Service, the NIHR, or the Department of Health. The review was designed, conducted, analysed, and interpreted by the authors entirely independently of the funding sources.

References1 Geddes JR, Burgess S, Hawton K, Jamison K, Goodwin GM.

Long-term lithium therapy for bipolar disorder: systematic review and meta-analysis of randomized controlled trialls. Am J Psychiatry 2004; 161: 217–22.

2 The BALANCE investigators and collaborator. Lithium plus valproate combination therapy versus monotherapy for relapse prevention in bipolar I disorder (BALANCE): a randomised open-label trial. Lancet 2010; 375: 385–95.

3 Cipriani A, Pretty H, Hawton K, Geddes JR. Lithium in the prevention of suicidal behavior and all-cause mortality in patients with mood disorders: a systematic review of randomized trials. Am J Psychiatry 2005; 162: 1805–19.

4 Schou M. Lithium studies.1. Toxicity. Acta Pharmacol Toxicol (Copenh) 1958; 15: 70–84.

Articles


5 Paul R, Minay J, Cardwell C, Fogarty D, Kelly C. Review: meta-analysis of the eff ects of lithium usage on serum creatinine levels. J Psychopharmacol 2010; 24: 1425–31.

6 Aurell M, Svalander C, Wallin L, Alling C. Renal-function and biopsy fi ndings in patients on long-term lithium treatment. Kidney Int 1981; 20: 663–70.

7 Hestbech J, Hansen HE, Amdisen A, Olsen, S. Chronic renal lesions following long-term treatment with lithium. Kidney Int 1977; 12: 205–13.

8 Nora JJ, Nora AH, Toews WH. Lithium, Ebsteins anomaly, and other congenital heart defects. Lancet 1974; 304: 594–95.

9 Razvi S, Weaver JU, Pearce SHS. Subclinical thyroid disorders: signifi cance and clinical impact. J Clin Pathol 2010; 63: 379–86.

10 Bendz H, Schon S, Attman PO, Aurell M. Renal failure occurs in chronic lithium treatment but is uncommon. Kidney Int 2010; 77: 219–24.

11 Coresh J, Astor BC, Greene T, Eknoyan G, Levey AS. Prevalence of chronic kidney disease and decreased kidney function in the adult US population: third National Health and Nutrition Examination Survey. Am J Kidney Dis 2003; 41: 1–12.

12 Bowden CL, Calabrese JR, McElroy SL, et al, for the Divalproex Maintenance Study Group. A randomized, placebo-controlled 12-month trial of divalproex and lithium in treatment of outpatients with bipolar I disorder. Arch Gen Psychiatry 2000; 57: 481–89.

13 Calabrese JR, Shelton MD, Rapport DJ, et al. A 20-month, double-blind, maintenance trial of lithium versus divalproex in rapid-cycling bipolar disorder. Am J Psychiatry 2005; 162: 2152–61.

14 Bowden CL, Calabrese JR, Sachs G, et al, for the Lamictal 606 Study Group. A placebo-controlled 18-month trial of lamotrigine and lithium maintenance treatment in recently manic or hypomanic patients with bipolar I disorder. Arch Gen Psychiatry 2003; 60: 392–400.

15 Calabrese JR, Bowden CL, Sachs G, et al, for the Lamictal 605 Study Group. A placebo-controlled 18-month trial of lamotrigine and lithium maintenance treatment in recently depressed patients with bipolar I disorder. J Clin Psychiatry 2003; 64: 1013–24.

16 Czeizel A, Racz J. Evaluation of drug intake during pregnancy in the Hungarian case-control surveillance of congenital-anomalies. Teratology 1990; 42: 505–12.

17 Zhang Q, Rothenbacher D. Prevalence of chronic kidney disease in population-based studies: systematic Review. BMC Public Health 2008; 8: 117–25.

18 Marples D, Christensen S, Christensen EI, Ottosen PD, Nielsen S. Lithium-induced down-regulation of aquaporin-2 water channel expression in rat-kidney medulla. J Clin Invest 1995; 95: 1838–45.

19 Blount MA, Sim JH, Zhou R, et al. Expression of transporters involved in urine concentration recovers diff erently after cessation of lithium treatment. Am J Physiol Renal Physiol 2010; 298: F601–08.

20 Bendz H. Kidney function in a selected lithium population. A prospective, controlled, lithium-withdrawal study. Acta Psychiatr Scand 1985; 72: 451–63.

21 Souza FGM, Mander AJ, Foggo M, Dick H, Shearing CH, Goodwin GM. The eff ects of lithium discontinuation and the non-eff ect of oral inositol upon thyroid-hormones and cortisol in patients with bipolar aff ective-disorder. J Aff ect Dis 1991; 22: 165–70.

22 Haggerty JJ, Prange AJ. Borderline hypothyroidism and depression. Ann Rev Med 1995; 46: 37–46.

23 Berens SC, Bernstei RS, Robbins J, Wolff J. Antithyroid eff ects of lithium. J Clin Invest 1970; 49: 1357–67.

24 Burrow GN, Burke WR, Himmelho JM, Spencer RP, Hershman JM. Eff ect of lithium on thyroid function. J Clin Endocrinol Metab 1971; 32: 647–52.

25 Palmer M, Bergstrom R, Akerstrom G, Adami HO, Jakobsson S, Ljunghall S. Survival and renal-function in untreated hypercalcemia population-based cohort study with 14 years of follow-up. Lancet 1987; 1: 59–62.

26 Szalat A, Mazeh H, Freund HR. Lithium-associated hyperparathyroidism: report of four cases and review of the literature. Eur J Endocrinol 2009; 160: 317–23.

27 National Institute for Health and Clinical Excellence. Bipolar disorder: the management of bipolar disorder in adults, children and adolescents, in primary and secondary care. London: National Institute for Health and Clinical Excellence, 2006: 1–31.

28 Cohen LS, Friedman JM, Jeff erson JW, Johnson EM, Weiner MLA. Reevaluation of risk of in-utero exposure to lithium. JAMA 1994; 271: 146–50.

29 Okusa MD, Crystal LJT. Clinical manifestations and management of acute lithium intoxication. Am J Med 1994; 97: 383–89.

792

·

September 17, 1998

The New England Journal of Medicine

A COMPARISON OF FOUR TREATMENTS FOR GENERALIZED CONVULSIVE STATUS EPILEPTICUS

D

AVID

M. T

REIMAN

, M.D., P

ATTI

D. M

EYERS

, M.P.A., N

ANCY

Y. W

ALTON

, P

H

.D., J

OSEPH

F. C

OLLINS

, S

C

.D., C

INDY

C

OLLING

, R.P

H

., M.S., A. J

AMES

R

OWAN

, M.D., A

DRIAN

H

ANDFORTH

, M.D., E

DWARD

F

AUGHT

, M.D., V

INCENT

P. C

ALABRESE

, M.D., B

ASIM

M. U

THMAN

, M.D., R. E

UGENE

R

AMSAY

, M.D.,

AND

M

EENAL

B. M

AMDANI

, M.D.,

FOR

THE

V

ETERANS

A

FFAIRS

S

TATUS

E

PILEPTICUS

C

OOPERATIVE

S

TUDY

G

ROUP

*

A

BSTRACT

Background and Methods

Although generalizedconvulsive status epilepticus is a life-threateningemergency, the best initial drug treatment is uncer-tain. We conducted a five-year randomized, double-blind, multicenter trial of four intravenous regimens:diazepam (0.15 mg per kilogram of body weight) fol-lowed by phenytoin (18 mg per kilogram), lorazepam(0.1 mg per kilogram), phenobarbital (15 mg per kil-ogram), and phenytoin (18 mg per kilogram). Pa-tients were classified as having either overt general-ized status epilepticus (defined as easily visiblegeneralized convulsions) or subtle status epilepticus(indicated by coma and ictal discharges on the elec-troencephalogram, with or without subtle convulsivemovements such as rhythmic muscle twitches ortonic eye deviation). Treatment was considered suc-cessful when all motor and electroencephalographicseizure activity ceased within 20 minutes after thebeginning of the drug infusion and there was no re-turn of seizure activity during the next 40 minutes.Analyses were performed with data on only the 518patients with verified generalized convulsive statusepilepticus as well as with data on all 570 patientswho were enrolled.

Results

Three hundred eighty-four patients had averified diagnosis of overt generalized convulsive sta-tus epilepticus. In this group, lorazepam was suc-cessful in 64.9 percent of those assigned to receive it,phenobarbital in 58.2 percent, diazepam and pheny-toin in 55.8 percent, and phenytoin in 43.6 percent(P=0.02 for the overall comparison among the fourgroups). Lorazepam was significantly superior tophenytoin in a pairwise comparison (P=0.002).Among the 134 patients with a verified diagnosis ofsubtle generalized convulsive status epilepticus, nosignificant differences among the treatments weredetected (range of success rates, 7.7 to 24.2 per-cent). In an intention-to-treat analysis, the differenc-es among treatment groups were not significant, ei-ther among the patients with overt status epilepticus(P=0.12) or among those with subtle status epilepti-cus (P=0.91). There were no differences among thetreatments with respect to recurrence during the 12-hour study period, the incidence of adverse reac-tions, or the outcome at 30 days.

Conclusions

As initial intravenous treatment forovert generalized convulsive status epilepticus, lora-zepam is more effective than phenytoin. Althoughlorazepam is no more efficacious than phenobarbitalor diazepam and phenytoin, it is easier to use. (N EnglJ Med 1998;339:792-8.)

©1998, Massachusetts Medical Society.

From the Neurology Services of the Veterans Affairs Medical Centers inWest Los Angeles, Calif. (D.M.T., P.D.M., N.Y.W., A.H.), Bronx, N.Y.(A.J.R.), Birmingham, Ala. (E.F.), Richmond, Va. (V.P.C.), Gainesville, Fla.(B.M.U.), and Miami (R.E.R.), and the Hines Veterans Affairs MedicalCenter, Chicago (M.B.M.); the Departments of Neurology of the Univer-sity of California at Los Angeles School of Medicine, Los Angeles (D.M.T.,N.Y.W.), Mount Sinai College of Medicine, New York (A.J.R.), the Univer-sity of Alabama School of Medicine, Birmingham (E.F.), the Medical Col-lege of Virginia, Richmond (V.P.C.), the University of Florida School ofMedicine, Gainesville (B.M.U.), the University of Miami School of Medi-cine, Miami (R.E.R.), and the Loyola University School of Medicine, Chi-cago (M.B.M.); the Veterans Affairs Cooperative Studies Program Coordi-nating Center, Perry Point, Md. (J.F.C.); and the Veterans AffairsCooperative Studies Program Clinical Research Pharmacy CoordinatingCenter, Albuquerque, N.M. (C.C.). Address reprint requests to Dr. Trei-man at the Department of Neurology, University of Medicine and Dentist-ry of New Jersey–Robert Wood Johnson Medical School, 97 Paterson St.,New Brunswick, NJ 08901-0019.

Other authors were Pratap Yagnik, M.D. (Neurology Service, VeteransAffairs Medical Center, and Department of Neurology, Medical College ofPennsylvania — both in Philadelphia); John C. Jones, M.D. (NeurologyService, Veterans Affairs Medical Center, and Department of Neurology,University of Wisconsin School of Medicine — both in Madison); Eliza-beth Barry, M.D. (Neurology Service, Veterans Affairs Medical Center, andDepartment of Neurology, University of Maryland School of Medicine —both in Baltimore); Jane G. Boggs, M.D. (Neurology Service, Veterans Af-fairs Medical Center, and Department of Neurology, Medical College ofVirginia — both in Richmond); and Andres M. Kanner, M.D. (NeurologyService, Veterans Affairs Medical Center, and Department of Neurology,University of Wisconsin School of Medicine — both in Madison).

*Other members of the study group are listed in the Appendix.

TATUS epilepticus is a life-threatening emer-gency that affects 65,000

1

to 150,000

2

peo-ple in the United States each year. General-ized convulsive status epilepticus is the most

common and most dangerous type. Phenobarbital,

3-5

phenytoin,

6-14

diazepam plus phen-ytoin,

15,16

and lorazepam

17-28

have been advocated forthe initial treatment of generalized convulsive statusepilepticus, and each is used by a substantial numberof physicians.

3

There are few data from controlledtrials, however, to document the efficacy of thesetreatments, and they have not been directly com-pared. We therefore undertook this study to com-pare the efficacy of standard doses of these four drugsin the treatment of generalized convulsive status ep-ilepticus.

METHODS

Study Design

In a double-blind study conducted at 16 Veterans Affairs med-ical centers and 6 affiliated university hospitals between July 1,1990, and June 30, 1995, patients with generalized convulsivestatus epilepticus were randomly assigned to receive intravenous

S

The New England Journal of Medicine Downloaded from nejm.org at LINKOPING UNIVERSITY on October 3, 2013. For personal use only. No other uses without permission.



Volume 339 Number 12

·

793

treatment with lorazepam, phenobarbital, phenytoin, or diaz-epam followed by phenytoin.

Overt generalized convulsive status epilepticus was defined asrecurrent convulsions without complete recovery between sei-zures, and subtle generalized convulsive status epilepticus as thestage of generalized convulsive status when the patient is in con-tinuous coma but only subtle motor convulsions are seen.

29

Pa-tients were classified as having one of these two types of status ep-ilepticus according to the following operational definitions. Overtgeneralized convulsive status epilepticus was considered presentwhen there were two or more generalized convulsions, withoutfull recovery of consciousness between seizures, or continuousconvulsive activity for more than 10 minutes (treatment after 10minutes of continuous seizure activity was considered essential toprotect against neuronal and systemic damage from ongoing sei-zure activity). Subtle generalized convulsive status epilepticus wasconsidered present when the patient had coma and ictal dischargeson the electroencephalogram,

30

with or without subtle convulsivemovements (rhythmic twitching of the arms, legs, trunk, or facialmuscles; tonic eye deviation; or nystagmoid eye jerking). If the in-vestigator required an electroencephalogram to diagnose general-ized convulsive status epilepticus, the patient was considered tohave subtle generalized convulsive status epilepticus.

The key criterion for study entry was evidence of overt or subtlegeneralized convulsive status epilepticus at the time of evaluation,regardless of prior drug treatment. Patients who had receivedtreatment and whose seizures had stopped were not eligible for in-clusion. Other exclusion criteria included status epilepticus of atype other than generalized convulsive, an age of less than 18years, pregnancy, a neurologic emergency requiring immediatesurgical intervention, and the presence of a specific contraindica-tion to therapy with hydantoin, benzodiazepine, or barbituratedrugs. If patients with repeated episodes of generalized convulsivestatus epilepticus were inadvertently enrolled more than once,only the first episode was included in the analysis.

A blood sample was obtained before treatment for hematologicand serum-chemistry tests and for screening for antiepilepticdrugs. Intravenous access was established with normal saline. Theorder of the treatments was determined by random assignment.Separate randomization schemes were used at each site for eachtype of status epilepticus. Treatment kits were placed at three orfour locations within each participating center, and for each pa-tient the lowest-numbered kit at the location nearest to the pa-

tients was used. Electroencephalographic recording was started assoon as possible after the initiation of the protocol, but treatmentwas never delayed until the electroencephalogram could be ob-tained unless it was necessary to confirm the diagnosis. Bloodpressure, heart rate, respiratory rate, level of consciousness, andseizure activity were recorded every 5 minutes for the first 20minutes after the drug infusion began and then every 10 minutesfor the next 40 minutes. Seizure activity and level of conscious-ness were recorded every hour thereafter until the completion ofthe 12-hour study period. Blood was obtained before the initialinfusion, at the completion of the infusion, and 2 hours and 12hours after the start of the infusion for the measurement of anti-convulsant-drug concentrations.

Treatment was considered successful if all clinical and electricalevidence of seizure activity stopped within 20 minutes after thestart of the infusion and did not recur during the period from 20to 60 minutes after the start of treatment. Electrical seizure ac-tivity included any of the five ictal patterns described previously.

30

Informed Consent

From the institutional review board at each participating hos-pital, we obtained approval for the study and permission to waivethe requirement for informed consent until after the initial ran-domized treatment. The rationale for the waiver was that the fourstudy treatments were the four best treatments then available, andthe patient was assumed to want the best available treatment. Be-cause there were insufficient data to guide the selection of onetreatment over another, any of the four treatments could be con-sidered the best. Informed consent for continued participation inthe study was obtained from the patient or a family member orother surrogate after initial treatment. Most patients were notcompetent to give informed consent during the 12-hour studyperiod; some never recovered sufficiently to be able to provideconsent. In such cases, consent was obtained from a surrogate, ifone was available. State laws and institutional review board regu-lations regarding surrogate consent were adhered to at all sites.Consent was written, unless oral consent was permitted by a spe-cific institution.

Drug Treatment

Phenytoin (Dilantin, Parke-Davis, Morris Plains, N.J.), pheno-barbital (Spectrum Chemical, Gardena, Calif.), and diazepam

*To convert drug concentrations to millimoles per liter, multiply by the following: lorazepam, 3.11;phenobarbital, 4.31; diazepam, 3.51; and phenytoin, 3.96.

T

ABLE

1.

D

RUG

D

OSES

, R

ATES

OF

A

DMINISTRATION

,

AND

C

OMPOSITION

OF

D

RUG

-T

REATMENT

K

ITS

.

V

ARIABLE

L

ORAZEPAM

P

HENOBARBITAL

D

IAZEPAM

AND

P

HENYTOIN

P

HENYTOIN

Dose (mg/kg) 0.1 15 0.15 and 18 18

Maximal rate of administration (mg/min)

2 100 5 and 50 50

Drug-solution concentration (mg/ml)

*

4 100 5 and 50 50

Contents of first treatment boxTubexVial AVial BVial CVial DVial E

LorazepamDummyDummyDummyDummyDummy

DummyPhenobarbitalPhenobarbitalPhenobarbitalDummyDummy

DummyDiazepamPhenytoinPhenytoinPhenytoinPhenytoin

Dummy PhenytoinPhenytoinPhenytoinPhenytoinDummy

Active drug in second treatment box Phenytoin Phenytoin Lorazepam Lorazepam

Active drug in third treatment box Phenobarbital Lorazepam Phenobarbital Phenobarbital



794

·

September 17, 1998


(Valium, Hoffmann–LaRoche, Nutley, N.J.) were packaged inidentical vials at appropriate concentrations so each of the drugscould be administered at a rate of 1 ml per minute to produce themaximal rates of drug infusion shown in Table 1. Lorazepam (Ati-van, Wyeth–Ayerst, Philadelphia) was administered by means ofTubex injection at a maximal rate of 0.5 ml per minute.

Identical-appearing drug-treatment kits were prepared for eachdrug regimen, with each kit containing a first, a second, and athird treatment box. The first treatment box consisted of oneTubex syringe and five vials, labeled A through E. A nomogram,based on the patient’s weight, was used to determine the volumeof solution administered from the Tubex and from each vial toproduce the desired dose without compromising the blinded na-ture of the study. The Tubex solution and the solution from vialA were injected simultaneously. Tubexes and vials with activedrug contained propylene glycol, as did dummy Tubexes; dummyvials contained normal saline. The second and third treatmentboxes were provided to allow subsequent treatment, if necessary,without revealing the identity of the study drug. Active drugs inthe second and third boxes for each treatment regimen are listedin Table 1.

Central Review

We used a central procedure for review of electroencepha-lograms and data verification to ensure consistency among thehospitals in the diagnosis and classification of generalized convul-sive status epilepticus and the determination of treatment success.A committee consisting of the study chairman, project director,and electroencephalographer reviewed all clinical data and elec-troencephalographic recordings obtained during the 12-hourstudy period. Differences of opinion between the central com-mittee and investigators at the study sites were resolved by discus-sion. The review committee remained blinded to the identity ofthe treatment drug in each case until the review of all cases wascompleted.

Statistical Analysis

The study was designed to analyze the patients with overt andsubtle generalized convulsive status epilepticus separately, becausewe anticipated a significant difference in outcome in the twogroups. Analyses were performed both on an intention-to-treatbasis, with all enrolled patients included, and with only patientsincluded who had a verified diagnosis of generalized convulsivestatus epilepticus. The intention-to-treat analyses included pa-tients who were mistakenly assigned to the wrong status group(e.g., overt instead of subtle status epilepticus), had other typesof status epilepticus, or did not actually have status epilepticus. Inthe intention-to-treat analysis, patients in the last group were clas-sified as having been successfully treated. For the analyses restrict-ed to patients with verified diagnoses, patients were assigned totheir correct status group (overt or subtle), as determined by thecentral review. Patients who did not have generalized convulsivestatus epilepticus were excluded from the verified-diagnosis analy-ses. Chi-square techniques were used to analyze rates of treat-ment success, recurrence, and adverse events. The alpha level wasset at 0.05 for analyses of all four treatments, and a two-tailed al-pha level of 0.01 was used to determine the significance of differ-ences between any two regimens.

31

Cochran–Mantel–Haenszelstatistics

32

were used for the post hoc combined status-groupanalyses.

RESULTS

We attempted to enroll all eligible patients at eachparticipating hospital. We screened 1705 patientsduring the study period; 570 were enrolled (395with overt status epilepticus and 175 with subtle sta-tus epilepticus). Of the 1135 who were not enrolled,113 were eligible but were not included because the

study team was not called. The other 1022 were ex-cluded for one or more of the following reasons: ab-sence of generalized convulsive status epilepticus(868 patients), previous enrollment in the study (41),contraindication to phenytoin therapy (48), con-traindication to barbiturate therapy (21), contraindi-cation to benzodiazepine therapy (13), age of lessthan 18 years (1), pregnancy (1), presence of a neu-rosurgical emergency (20), and other reasons (236).Fifty-two of the 570 enrolled patients did not havegeneralized convulsive status epilepticus at the timeof randomization. In addition, 18 patients classifiedas having subtle status epilepticus at randomizationactually had overt generalized convulsive status epi-lepticus. As a result, 384 patients with verified overtstatus epilepticus and 134 with subtle status epilep-ticus (total, 518) were included in all the analysesperformed in the verified-diagnosis group; all 570enrolled patients were included in the intention-to-treat analysis. Data on efficacy could not be ob-tained for five patients with overt status epilepticus.

Table 2 summarizes the characteristics of the pa-tients with verified diagnoses. Although there wereno significant differences among the four drug-treat-ment groups in any of the characteristics we exam-ined (data not shown), patients with overt and sub-tle status epilepticus differed significantly with respectto age, race, current use of anticonvulsants, historyof seizures, cause of status epilepticus, functionalstatus before the current episode, whether or notthey were veterans, the part of the hospital wheretreatment took place (emergency room, ward, or in-

*Plus–minus values are means ±SD.

†Some patients had more than one causal factor.

T

ABLE

2.

B

ASE

-L

INE

C

HARACTERISTICS

OF

THE

518 P

ATIENTS

WITH

V

ERIFIED

D

IAGNOSES

, A

CCORDING

TO

T

YPE

OF

G

ENERALIZED

C

ONVULSIVE

S

TATUS

E

PILEPTICUS

.*

C

HARACTERISTIC

O

VERT

(N=384)

S

UBTLE

(N=134)

Age (yr) 58.6±15.6 62.0±15.1

Veteran (%) 70.1 80.6

Male sex (%) 82.3 85.1

Not previously treated for current episode (%) 51.3 51.5

History of acute seizures (%) 54.2 25.4

History of epilepsy (%) 42.4 12.7

History of status epilepticus (%) 12.8 4.5

Median duration of status epilepticus at enrollment (hr)

2.8 5.8

Causal factors (%)†Remote neurologic cause Acute neurologic cause Life-threatening medical condition Cardiopulmonary arrest Toxic effects of therapeutic or recreational

drug Alcohol withdrawal

69.527.332.06.36.3

6.5

34.337.356.738.15.2

0.7




Volume 339 Number 12

·

795

tensive care unit), history of status epilepticus, histo-ry of excessive alcohol use, and duration of status ep-ilepticus.

Table 3 gives the mean doses, serum concentra-tions after infusion, and length of time necessary tocomplete infusion for each of the four initial treat-ments. Lorazepam required the least time to infuse(P<0.001 in paired comparisons); the treatmentsthat included phenytoin took the longest (P<0.001).

Figure 1 presents the results of the four treatmentregimens with respect to efficacy. In the analysis ofthe 518 patients with verified diagnoses (Fig. 1A),the first treatment regimen was successful in 55.5percent of patients with overt status epilepticus, butin only 14.9 percent of those with subtle status epi-lepticus. Among the patients with overt status epi-lepticus, chi-square analysis showed a significant dif-ference overall (P=0.02) in the frequency of successamong the four treatments, but no differences weredetected in the group with subtle status epilepticus(P=0.18). Lorazepam was effective significantly moreoften than phenytoin (P=0.002) in patients withovert status epilepticus. Other pairwise comparisonsdid not show significant differences between individ-ual treatments.

When the two groups were combined in a posthoc analysis, lorazepam was successful as the firsttreatment in 52.2 percent of the patients to whomit was administered, phenobarbital in 49.2 percent,diazepam followed by phenytoin in 43.1 percent,and phenytoin alone in 36.8 percent. Chi-squareanalysis revealed a significant difference (P=0.008)in the frequency of success among treatments. Inpaired comparisons, lorazepam was effective moreoften than phenytoin (P=0.001). The difference inefficacy between phenobarbital and phenytoin ap-proached significance (P=0.02). The results of theintention-to-treat analysis (Fig. 1B) were similar, but

*Values are means ±SD. To convert concentrations to micromoles perliter, multiply by the following: lorazepam, 3.11; phenobarbital, 4.31; di-azepam, 3.51; and phenytoin, 3.96.

†P<0.001 for the differences among the drug regimens.

T

ABLE

3.

D

OSES

, S

ERUM

D

RUG

C

ONCENTRATIONS

AFTER

F

IRST

D

RUG

I

NFUSION

,

AND

L

ENGTH

OF

D

RUG

I

NFUSION

.*

R

EGIMEN

D

OSE

S

ERUM

C

ONCENTRATION

L

ENGTH

OF

I

NFUSION

†

mg/kg µg/ml min

Lorazepam 0.10±0.01 0.231±0.299 4.7±7.2

Phenobarbital 14.96±2.53 31.2±37.2 16.6±11.5

Diazepam and phenytoin 0.15±0.02 and

15.08±4.84

0.245±0.307and

31.8±19.2

42.0±38.1

Phenytoin 16.02±3.21 30.0±13.6 33.0±20.1

Figure 1. Rate of Successful Initial Treatment with Each of theFour Regimens.Open bars indicate patients with overt generalized convulsivestatus epilepticus, and solid bars those with subtle status epi-lepticus. Panel A shows the results of our analysis of patientswith a verified diagnosis of generalized convulsive status epi-lepticus. Among the patients with overt status epilepticus, dif-ferences in the frequency of success among the treatmentswere significant (P=0.02); in pairwise comparisons, lorazepamwas effective significantly more often than phenytoin (P=0.002).Panel B shows the results of the intention-to-treat analysis. Thedifferences in the frequency of success among the treatmentgroups were not significant. The success rate for each treat-ment is indicated above the bars.

0

70

All Enrolled Patients

Lorazepam

Phenobarbital

Diazepam andG

Phenytoin

100G46

92G41

99G47

104G41

Phenytoin

10

20

30

40

50

60

NO. OF PATIENTS

OvertGSubtle

Su

cces

sfu

l Tre

atm

ent

(%)

26.1

67.0

24.4

63.0

23.4

59.6

19.5

51.0

0

70

Patients with Verified Diagnoses

Lorazepam

Phenobarbital

Diazepam andG

Phenytoin

97G39

91G33

95G36

101G26

Phenytoin

10

20

30

40

50

60

NO. OF PATIENTS

OvertGSubtle

Su

cces

sfu

l Tre

atm

ent

(%)

17.9

64.9

24.2

58.2

8.3

55.8

7.7

43.6

SubtleOvert

B

A



796 · September 17, 1998


differences between groups were not significant foreither the patients with overt status epilepticus (P=0.12) or those with subtle status epilepticus (P=0.91). In the verified-diagnosis analysis, status epi-lepticus recurred during the 12-hour study period in11 percent (24 of 213) of patients with successfullytreated overt status epilepticus and 20 percent (4 of20) of successfully treated patients with subtle cases.There were no significant differences in the rates ofrecurrence among the four treatments for either thepatients with a verified diagnosis of overt general-ized convulsive status epilepticus or those with veri-fied subtle status epilepticus.

The commonly reported side effects of treatmentare shown in Table 4. There were no significant dif-ferences among the four treatments in either group(those with overt or subtle generalized convulsivestatus epilepticus). However, hypotension requiringtreatment occurred more often in the patients withsubtle status epilepticus than in those with overtstatus epilepticus (P<0.001). Sixty-seven of the pa-tients with overt status epilepticus (17 percent) re-gained full consciousness before the end of the 12-hour study period, with no significant differencesamong the four treatment groups (P=0.59). Noneof the patients with subtle status epilepticus com-pletely regained consciousness during the 12-hourstudy period.

Outcomes 30 days after treatment were signifi-cantly worse (P<0.001) for patients with subtle sta-tus epilepticus. At 30 days, 50.1 percent of patientswith overt status epilepticus had been dischargedfrom the hospital, as compared with 8.8 percent ofpatients with subtle status epilepticus; 22.9 percentof those with overt status epilepticus were still in thehospital, as compared with 26.5 percent of thosewith subtle status epilepticus; mortality rates were

27.0 percent and 64.7 percent, respectively. Therewere no significant differences in outcome at 30 daysamong the four treatments for either the patientswith overt status epilepticus or those with subtle sta-tus epilepticus. In both the overt and subtle statusepilepticus groups, however, patients successfully treat-ed with the first study drug had a better prognosisthan did those in whom the first treatment was notsuccessful (for patients with overt status epilepticus,P<0.001; for those with subtle status epilepticus,P=0.01). Among patients with overt and subtle sta-tus epilepticus, mortality was twice as high for pa-tients whose status was not controlled with the firstdrug as for those in whom the first treatment wassuccessful. No other follow-up data were collected.

DISCUSSION

Our results show that lorazepam is more likelythan phenytoin to be successful when used as theinitial intravenous treatment for overt generalizedconvulsive status epilepticus. Although this study wassponsored by the Department of Veterans Affairs, 30percent of the patients with overt status epilepticuswere not veterans, and 18 percent were female, sug-gesting that our results are widely applicable to thetreatment of generalized convulsive status epilepti-cus in adults.

The lower overall efficacy rates reported here, ascompared with data from earlier studies,10-14,16-26 prob-ably result from our including only patients withgeneralized convulsive status epilepticus and our useof a stringent definition of treatment success: cessa-tion of all clinical and electrical evidence of seizureactivity within 20 minutes, with no recurrence dur-ing the next 40 minutes. The use of a period longerthan 20 minutes in the definition of treatment suc-cess was discussed during the development of thisprotocol, but it was rejected as exposing patients tounnecessary risk. It is desirable that medications forthis condition be given in a short intravenous infu-sion and enter the brain rapidly during infusion.Lorazepam, the most effective drug in the pairedcomparisons, required the least time to administer.It could be argued that the 20-minute time limitconstituted a disadvantage for phenytoin, because ofits mandatory slow infusion rate. On the other hand,phenobarbital, which enters the brain slowly,33 wasnot significantly less effective than lorazepam inpatients with overt generalized convulsive status ep-ilepticus. Thus, it appears that the 20-minute crite-rion for success does not in itself explain the differ-ences we found.

We found no differences among the treatments inthe frequency of recurrence of overt or subtle statusepilepticus during the 12-hour study period, sug-gesting that any of the four treatments, if successful,can protect equally well against recurrence. We alsofound no differences among the treatments in the

TABLE 4. FREQUENCY OF COMMON DRUG-RELATED SIDE EFFECTS AMONG THE 518 PATIENTS WITH VERIFIED DIAGNOSES.

TYPE OF GENERALIZED CONVULSIVE

STATUS EPILEPTICUS AND SIDE EFFECT LORAZEPAM PHENOBARBITAL

DIAZEPAM AND

PHENYTOIN PHENYTOIN

Overt

No. of patients 97 91 95 101Hypoventilation (%) 10.3 13.2 16.8 9.9Hypotension (%) 25.8 34.1 31.6 27.0Cardiac-rhythm

disturbance (%)7.2 3.3 2.1 6.9

Subtle

No. of patients 39 33 36 26Hypoventilation (%) 12.8 15.2 2.9 7.7Hypotension (%) 59.0 48.5 58.3 57.7Cardiac-rhythm

disturbance (%)7.7 9.1 5.6 0.0




Volume 339 Number 12 · 797

incidence of hypotension requiring treatment, respi-ratory depression, or cardiac-rhythm disturbances inpatients with either overt or subtle generalized con-vulsive status epilepticus. Thus, the risk of these ad-verse events appears similar with any of the four reg-imens if the drugs are administered at a safe rate topatients who have been appropriately screened forcontraindications. Hypotension requiring treatmentoccurred more often in the patients with subtle gen-eralized convulsive status epilepticus; this differenceprobably reflects the fact that patients with subtlestatus epilepticus were sicker than those with overtstatus epilepticus. Life-threatening medical condi-tions and cardiopulmonary arrest associated withgeneralized convulsive status epilepticus were morecommon among the patients with subtle status epi-lepticus (Table 2). The longer duration of status ep-ilepticus in the patients with subtle cases (Table 2)may also have contributed to their greater suscepti-bility to drug-induced hypotension. Even patientswith overt status epilepticus had a long delay be-tween onset and treatment. Many of these episodesoccurred at night, and paramedics were not calleduntil after several seizures had occurred. For suchpatients, we considered status epilepticus to have be-gun at the time of the first seizure.

The rapidity of recovery of consciousness aftertreatment of status epilepticus is another clinicallyimportant factor when choosing a treatment regi-men. The small number of patients who recoveredfully within 12 hours suggests, however, that rapid,complete recovery may not be a realistic goal whentreating generalized convulsive status epilepticus.The condition causing the episode and the effects ofrepetitive seizure activity also contribute to the im-pairment of consciousness. Identifying significantdifferences in the rates at which patients recoverfrom drug-induced impairment of consciousness isdifficult, because such a comparison would have tobe made in a group of patients in whom the causa-tive factors and duration and intensity of seizuresbefore treatment were identical.

Overall, the patients with subtle generalized con-vulsive status epilepticus did much worse than thosewith overt generalized convulsive status epilepticus.The first drug treatment was successful in less than15 percent of the patients with subtle status epilep-ticus, and their outcome was poor. Sixty-five percentof the patients with subtle status epilepticus diedwithin 30 days after the episode, as compared with27 percent of the patients with overt status epilepti-cus. Death in a patient with generalized convulsivestatus epilepticus is probably attributable largely tothe underlying cause and duration of the episode.Nonetheless, successful treatment was significantlyassociated with improved outcomes in both patientswith overt episodes and those with subtle episodes,although it is not clear whether the success of treat-

ment was the cause or the effect of the better prog-nosis, or a combination of both.

Because even the best treatments were successfulin only about two thirds of the patients with overtstatus epilepticus and one fourth of the patients withsubtle status epilepticus, our study underscores theneed for better methods of treating generalized con-vulsive status epilepticus and its underlying causes.Until new therapies become available, however, werecommend lorazepam for the initial intravenoustreatment of generalized convulsive status epilepti-cus. Although lorazepam was no more efficaciousthan phenobarbital or than diazepam and phenyto-in, it is easier to use.

Supported by the Department of Veterans Affairs Medical ResearchService Cooperative Studies Program (CSP 265). Lorazepam and dummylorazepam Tubexes used in the study were donated by Wyeth–Ayerst Lab-oratories.

Drs. Barry, Faught, Ramsay, Treiman, and Uthman serve as consultantsto Parke-Davis. They and Drs. Boggs, Calabrese, Kanner, and Rowan alsoserve on the Parke-Davis speakers bureau. Dr. Treiman has also served as aconsultant to Hoffmann–LaRoche and Wyeth–Ayerst.

We are indebted to the members of the Data Monitoring Commit-tee (Timothy Pedley, M.D., Ilo Leppik, M.D., Eric Lothman, M.D.,Ph.D. [deceased], Roger D. Porter, M.D., and Gerald VanBelle,Ph.D.) for their careful oversight of the progress of the trial; and tothe Veterans Affairs Cooperative Studies Program Office staff(Daniel Deykin, M.D., Janet Gold, and Ping Huang, Ph.D.) fortheir support.

APPENDIX

Additional study participants included other investigators at the studysites: S.D. Collins, S. Dane, S.L. Fish, D.A. Hosford, R.I. Kuzniecky, M.Muxfeldt, H. Price, D. Rosenbaum, M.P. Remler, R.L. Ruff, P.A. Rutecki,M.V. Sowa, M.L. Tomyanovich, and B.J. Wilder; the study coordinators:D. Correll, A. Cugley, P. Encomienda, B. Gallo, K. Hall-Behrman, V.Hamilton, E. Hwang, P. Jarres, A. Mann, J. McWhorter, D. Meija, C. Mix-on, M. Moroney, D. Mueller, R. Nemire, K. Ohara, R. Palovcik, R. Reed,M. Shove, L. Tuchman, L. Williams, and H.L. Yoon; staff of the studychairman’s office: J. Bejaune, K. Fagan, and C. Stone; staff at the VeteransAffairs Cooperative Studies Program Coordinating Center: B. Calvert, R.Horney, D. Preston, and M. Rhoads; staff at the Veterans Affairs Cooper-ative Studies Program Clinical Research Pharmacy Coordinating Center: J.Peterson; and the staff of the central laboratory: E. Esteban, S. Gunawan,Q. Jiang, and V. Pham.

REFERENCES

1. Hauser WA. Status epilepticus: epidemiologic considerations. Neurolo-gy 1990;40:Suppl 2:9-13.2. DeLorenzo RJ, Pellock JM, Towne AR, Boggs JG. Epidemiology of status epilepticus. J Clin Neurophysiol 1995;12:316-25.3. Goldberg MA, McIntyre HB. Barbiturates in the treatment of status ep-ilepticus. In: Delgado-Escueta AV, Wasterlain CG, Treiman DM, Porter RJ, eds. Status epilepticus: mechanisms of brain damage and treatment. Vol. 34 of Advances in neurology. New York: Raven Press, 1983:499-503.4. Shaner DM, McCurdy SA, Herring MO, Gabor AJ. Treatment of status epilepticus: a prospective comparison of diazepam and phenytoin versus phenobarbital and optional phenytoin. Neurology 1988;38:202-7.5. Crawford TO, Mitchell WG, Fishman LS, Snodgrass SR. Very-high-dose phenobarbital for refractory status epilepticus in children. Neurology 1988;38:1035-40.6. Murphy JT, Schwab RS. Diphenylhydantoin (Dilantin) sodium used parenterally in control of convulsions: a five-year report. JAMA 1956;160:385-8.7. McWilliam PKA. Intravenous phenytoin sodium in continuous convul-sions in children. Lancet 1958;2:1147-9.8. Carter CH. Use of parenteral diphenylhydantoin (Dilantin) sodium in control of status epilepticus. Arch Neurol Psychiatry 1958;79:136-7.



798 · September 17, 1998


9. Wallis W, Kutt H, McDowell F. Intravenous diphenylhydantoin in treat-ment of acute repetitive seizures. Neurology 1968;18:513-25.10. Wilder BJ, Ramsay RE, Willmore LJ, Feussner GR, Perchalski RJ, Shu-mate JB Jr. Efficacy of intravenous phenytoin in the treatment of status ep-ilepticus: kinetics of central nervous system penetration. Ann Neurol 1977;1:511-8.11. Cranford RE, Leppik IE, Patrick B, Anderson CB, Kostick B. Intrave-nous phenytoin in acute treatment of seizures. Neurology 1979;29:1474-9.12. Wilder BJ. Efficacy of phenytoin in treatment of status epilepticus. In: Delgado-Escueta AV, Wasterlain CG, Treiman DM, Porter RJ, eds. Status epilepticus: mechanisms of brain damage and treatment. Vol. 34 of Ad-vances in neurology. New York: Raven Press, 1983:441-6.13. Leppik IE, Patrick BK, Cranford RE. Treatment of acute seizures and status epilepticus with intravenous phenytoin. In: Delgado-Escueta AV, Wasterlain CG, Treiman DM, Porter RJ, eds. Status epilepticus: mecha-nisms of brain damage and treatment. Vol. 34 of Advances in neurology. New York: Raven Press, 1983:447-51.14. von Albert H-H. A new phenytoin infusion concentrate for status ep-ilepticus. In: Delgado-Escueta AV, Wasterlain CG, Treiman DM, Porter RJ, eds. Status epilepticus: mechanisms of brain damage and treatment. Vol. 34 of Advances in neurology. New York: Raven Press, 1983:453-6.15. Sutherland JM, Tait H. The epilepsies: modern diagnosis and treat-ment. Edinburgh, Scotland: Livingstone, 1969.16. Delgado-Escueta AV, Enrile-Bacsal F. Combination therapy for status epilepticus: intravenous diazepam and phenytoin. In: Delgado-Escueta AV, Wasterlain CG, Treiman DM, Porter RJ, eds. Status epilepticus: mecha-nisms of brain damage and treatment. Vol. 34 of Advances in neurology. New York: Raven Press, 1983:477-85.17. Waltregny A, Dargent J. Preliminary study of parenteral lorazepam in status epilepticus. Acta Neurol Belg 1975;75:219-29.18. Amand G, Evrard P. Le lorazepam injectable dans les états de mal épi-leptiques. Rev Electroencephalogr Neurophysiol Clin 1976;6:532-3.19. DeOliverira RSP. Treatment of convulsive seizures with a new benzo-diazepine, lorazepam. Rev Bras Clin Ter 1978;7:295-8.

20. Walker JE, Homan RW, Vasko MR, Crawford IL, Bell RD, Tasker WG. Lorazepam in status epilepticus. Ann Neurol 1979;6:207-13.21. Griffith PA, Karp HR. Lorazepam in therapy for status epilepticus. Ann Neurol 1980;7:493.22. Sorel L, Mechler L, Harmant J. Comparative trial of intravenous lorazepam and clonazepam in status epilepticus. Clin Ther 1981;4:326-36.23. Leppik IE, Derivan AT, Homan RW, Walker J, Ramsay RE, Patrick B. Double-blind study of lorazepam and diazepam in status epilepticus. JAMA 1983;249:1452-4.24. Levy RJ, Krall RL. Treatment of status epilepticus with lorazepam. Arch Neurol 1984;41:605-11.25. Lacey DJ, Singer WD, Horwitz SJ, Gilmore H. Lorazepam therapyof status epilepticus in children and adolescents. J Pediatr 1986;108:771-4.26. Crawford TO, Mitchell WG, Snodgrass SR. Lorazepam in childhood status epilepticus and serial seizures: effectiveness and tachyphylaxis. Neu-rology 1987;37:190-5.27. Gabor AJ. Lorazepam versus phenobarbital: candidates for drug of choice for treatment of status epilepticus. J Epilepsy 1990;3:3-6.28. Working Group on Status Epilepticus. Treatment of convulsive status epilepticus: recommendations of the Epilepsy Foundation of America’s Working Group on Status Epilepticus. JAMA 1993;270:854-9.29. Treiman DM. Generalized convulsive status epilepticus in the adult. Epilepsia 1993;34:Suppl 1:S2-S11.30. Treiman DM, Walton NY, Kendrick C. A progressive sequence of elec-troencephalographic changes during generalized convulsive status epilepti-cus. Epilepsy Res 1990;5:49-60.31. Everitt BS. The analysis of contingency tables. London: Chapman & Hall, 1977.32. SAS/STAT user’s guide, version 6. 4th ed. Cary, N.C.: SAS Institute, 1989.33. Ramsay RE, Hammond EJ, Perchalski RJ, Wilder BJ. Brain uptakeof phenytoin, phenobarbital, and diazepam. Arch Neurol 1979;36:535-9.



RESEARCH ARTICLE

Benzodiazepines, Z-drugs and the risk of hip

fracture: A systematic review and meta-

analysis

Karen Donnelly1,2, Robert Bracchi1, Jonathan Hewitt2, Philip A. Routledge1,

Ben Carter2,3,4*

1 Pharmacology, Therapeutics and Toxicology, Cardiff University School of Medicine, Academic Centre,

University Hospital Llandough, Cardiff, United Kingdom, 2 Institute of Primary Care and Public Health, Cardiff

University, School of Medicine, Neuadd Meirionnydd, Cardiff, United Kingdom, 3 Department of Biostatistics

and Health Informatics, Institute of Psychiatry, Psychology and Neuroscience, King’s College London,

London, United Kingdom, 4 Cochrane Skin Group, School of Medicine, The University of Nottingham,

Nottingham, United Kingdom

* [email protected]

Abstract

Background

Hip fractures in the older person lead to an increased risk of mortality, poorer quality of life

and increased morbidity. Benzodiazepine (BNZ) use is associated with increased hip fracture

rate, consequently Z-drugs are fast becoming the physician’s hypnotic prescription of choice

yet data on their use is limited. We compared the risk of hip fracture associated with Z-drugs

and BNZ medications, respectively, and examined if this risk varied with longer-term use.

Methods and findings

We carried out a systematic review of the literature and meta-analysis. MEDLINE and SCO-

PUS were searched to identify studies involving BNZ or Z-drugs and the risk of hip fracture

up to May 2015. Each included study was quality-assessed. A pooled relative risk of hip frac-

ture was calculated using the generic inverse variance method, with a random effects

model, with the length of hypnotic usage as a subgroup. Both BNZ, and Z-drug use respec-

tively, were significantly associated with an increased risk of hip fracture (RR = 1.52, 95% CI

1.37–1.68; and RR = 1.90, 95% CI 1.68–2.13). Short-term use of BNZ and Z-drugs respec-

tively, was also associated with the greatest risk of hip fracture (RR = 2.40, 95% CI 1.88–

3.05 and RR = 2.39, 95% CI 1.74–3.29).

Conclusions

There is strong evidence that both BNZ and Z-drugs are associated with an increased risk

of hip fracture in the older person, and there is little difference between their respective

risks. Patients newly prescribed these medicines are at the greatest risk of hip fracture. Cli-

nicians and policy makers need to consider the increased risk of fallings and hip fracture par-

ticularly amongst new users of these medications.

PLOS ONE | https://doi.org/10.1371/journal.pone.0174730 April 27, 2017 1 / 14

a1111111111

a1111111111

a1111111111

a1111111111

a1111111111

OPENACCESS

Citation: Donnelly K, Bracchi R, Hewitt J,

Routledge PA, Carter B (2017) Benzodiazepines, Z-

drugs and the risk of hip fracture: A systematic

review and meta-analysis. PLoS ONE 12(4):

e0174730. https://doi.org/10.1371/journal.

pone.0174730

Editor: Tuan Van Nguyen, Garvan Institute of

Medical Research, AUSTRALIA

Received: October 19, 2016

Accepted: March 14, 2017

Published: April 27, 2017

Copyright: © 2017 Donnelly et al. This is an open

access article distributed under the terms of the

Creative Commons Attribution License, which

permits unrestricted use, distribution, and

reproduction in any medium, provided the original

author and source are credited.

Data Availability Statement: All data included in

this study can be found on public databases.

Funding: We acknowledge the support of the

National Institute for Health Research (NIHR)

Biomedical Research Centre at South London and

Maudsley NHS Foundation Trust and King’s

College London (BC).

Competing interests: The authors have declared

that no competing interests exist.

https://doi.org/10.1371/journal.pone.0174730

http://crossmark.crossref.org/dialog/?doi=10.1371/journal.pone.0174730&domain=pdf&date_stamp=2017-04-27








http://creativecommons.org/licenses/by/4.0/

Introduction

Hip fractures are associated with repeat fractures and substantial morbidity and mortality. It

has been reported that one third of older people die within the year following hip fracture [1,

2]. Many who survive face significantly reduced capacity to carry out activities of daily living

and one third require adjustment to institutionalised living [3]. Post-hip fracture, one year all-

cause-mortality has remained constant in the UK from 2000 to 2010, with an increased risk

of 3.5 for males and 2.4 for females [4]. In the US, one year all-cause mortality is estimated at

24% [5].

The Center for Disease Control estimated that at least 300,000 older people are hospitalised

in the US per annum due to hip fracture [6] In Europe the total number of hip fractures is pre-

dicted to reach 4.5 million by the year 2025 [7]. Hip fracture carries a major economic burden.

Sahota and co-workers reported median cost per UK hip fracture as £9,429 [€10,896] for care

home residents and £14,435 [€16,681] for those requiring transition post-hip fracture into a

care home [8]. The annual costs of hip fracture are approximately £2 billion in the UK [2] and

$10.3–15.2 billion in the US respectively [5, 9, 10].

Indeed there is a well-established association between psychotropic medications and falls

[11, 12]. Benzodiazepines (BNZ) have long been associated with fracture risk [13–16]. The

national institute for clinical excellence (NICE) reported that an estimated 10–30% of chronic

users are BNZ dependent with evidence of tolerance or drug seeking behaviour and that 50%

of all users experience symptoms of withdrawal [17]. BNZ are subdivided according to their

elimination time (half-life): short (t½ = 1–12 hours), intermediate (t½ = 12 to 40 hours) or long

(t½ = 40–250 hours)[18].

Zolpidem, zaleplon and zopiclone are classed as non-benzodiazepine hypnotics (often

referred to as a group as “Z-drugs"). Z-drugs were designed with a shorter elimination time

(t½ = 1–7 hours) to act as a more clinically attractive alternative to BNZ [19]. They aim to

achieve hypnotic effect without the undesired side effects of BNZ, such as dependence or seda-

tion on the following day. Nevertheless, the pharmacological action of Z-drugs are similar to

BNZ as it involved the benzodiazepine receptor site which are associated with gamma-amino-

butyric acid (GABA) receptors. Siriwardena et al. 2006 reported that general practitioners

believed Z-drugs are safer and more effective than BNZ [20]. In 2014, it was estimated that

26.1% of the adult UK population have ever taken a BNZ or Z-drug[21]. In 2004 and 2008, the

annual prevalence of BNZ use has been estimated at 4% of the Canadian population and 5.2%

of the US population respectively [22]. In Europe, interventions amongst general practitioners

have been successful in decreasing BNZ prescriptions [23]. Despite this, concerns have been

reported that prescriptions of Z-drugs such as zopiclone have increased over the past decade

[22, 24–26].

The current literature base illustrates that there is an increased risk of hip fracture following

BNZ use [27, 28] and zolpidem use [29] respectively. Furthermore, NICE has warned that

there is a lack of compelling evidence to distinguish between the safety of short-acting BNZ

and Z-drugs[17]. However, importantly clinical perceptions differ [20] and there is debate sur-

rounding the suitable length of hypnotic prescription. In the United States, the Compendium

of Therapeutic Choices (CTC) has recommended a short course of hypnotics combined with

good sleep hygiene, for the treatment of insomnia [30]. In Canada, the Physicians’ Desk Refer-

ence has recommended short term BNZ prescription only (two to four weeks) [31–34]. In the

UK the British National Formulary (BNF) has recommended that BNZ should be used to treat

insomnia “only when it is severe, disabling, or causing the patient extreme distress” [35]. The

BNF also states that “The use of benzodiazepines to treat short-term ‘mild’ anxiety is inappro-

priate”, and that they are indicated for the short-term relief (two to four weeks only) of anxiety

Benzodiazepines, Z-drugs and the risk of hip fracture: A systematic review and meta-analysis


https://www.medicinescomplete.com/mc/bnf/current/PHP2119-zaleplon.htm

https://www.medicinescomplete.com/mc/bnf/current/PHP2124-zopiclone.htm


that is severe, disabling, or causing the patient unacceptable distress, occurring alone or in

association with insomnia or short-term psychosomatic, organic, or psychotic illness[35].

However others, have suggested significant balance impairment can occur following a single

dose [36] and Wagner et al. determined BNZ use to be most dangerous within the first 15 days

compared continuous use (RR = 1.74 1.07–2.82) [37].

The objectives of this study were to investigate the association between (1) BNZ use and hip

fracture risk and (2) Z-drug use and hip fracture risk. We aimed to study this relationship

according to length of use.

Methods

Inclusion and exclusion criteria

This study followed the PRISMA checklist and flowchart [38] (S1 File). A protocol was devel-

oped and agreed prior to the review (S2 File). Studies were included if all of the following crite-

ria applied (i) designed as a randomised controlled trial, cohort or case-control study (ii)

reported outcome was hip fracture (ICD-10: S72) or fragility fracture (within which outcome

�70% of fractures were hip fractures) (iii) included patients were prescribed either BNZ or Z-

drug, or were matched as a non-exposed control population (iv) the study population were

aged at least 50 years old and with a mean age over 65. Clinically, clonazepam is frequently

used as an anti-epileptic medication rather than as a hypnotic and were therefore excluded

[39]

Search strategy

Two databases (i) Medline via Ovid (S3 File) (EMBASE, Ovid, Psych INFO) and (ii) Scopus

were systematically searched on 11th May, 2015. Searches were independently carried out by

two reviewers (KD, BC), and disagreements were resolved by discussion. The search was lim-

ited to studies published in the English language.

Definition of exposure

Exposure was categorised into two main subgroups: exposure to BNZ v non-exposure; and

exposure to Z-drugs v non-exposure. BNZ exposure was defined as patients prescribed diaze-

pam, lorazepam, chlordiazepoxide, oxazepam, temazepam, nitrazepam, loprazolam or cloba-

zam [40]. Z-drug exposure was defined as those prescribed zaleplon, zolpidem or zopiclone

[35]. Length of usage was defined from the first prescription date, provided there was at least

one hypnotic free month. Short term use was defined as those prescribed medication for up to

14 days, medium term use 15 days to 30 days, and long term use was longer than one month,

mixed use was a combination of medium and long-term users.

Study selection & quality assessment

Database results were reviewed by examining titles and abstracts. Full articles were examined

for methodological quality using the Newcastle-Ottawa quality assessment tool. Studies were

assessed to be of good, fair or poor quality [41]. Quality assessment was carried out indepen-

dently by two reviewers (KD and BC), disagreements were resolved by discussion.

Data extraction and characteristics of included studies

The outcome was the proportion of participants that had a hip fracture (ICD-10:S72). The

characteristics of the eligible studies included: author, year, location, environment, study

design, age profile, length of exposure, length of follow up, total participants and adjustment




for confounders with particular attention to dosage. Data was separately extracted by two

reviewers (KD, BC) and discrepancies were resolved following discussion.

Data synthesis

The risk of hip fracture in those exposed to a BNZ (or Z-drugs), compared to patients not tak-

ing these medications. The measure of effect was the adjusted relative risk (RR) with the associ-

ated 95% confidence interval (95% CI). Included comparisons were studies of: people using

BNZ compared to those not exposed; and people using Z-drugs compared to those not expo-

sure. Non-randomised study designs were described narratively and only pooled into a meta-

analysis if their context, population, medication (including delivery) were considered clinically

similar [42]. A pooled relative risk using generic inverse variance methods with a random

effects model was used in RevMan 5.3 [43, 44]. The shortest time point was included in the

pooled analysis from studies that reported at multiple time-points [37, 45–47]. Funnel plots to

explore reporting bias were examined for all pooled meta-analyses, or subgroup of greater

than ten studies.

Subgroups analysis and assessment for heterogeneity. Statistical heterogeneity was sum-

marised using an I2 statistic. Where I2 was reported higher than 75%, subgroups were explored

to explain the heterogeneity [43]. Subgroups used to explore heterogeneity were: length of use;

case mix of patients (insomniac-only studies or not); and type of study design (population

based or non-population based studies).

Results

There were 219 studies identified. After screening of abstracts and titles, 44 articles met the

inclusion criteria which led to 22 being quality assessed and 18 included studies (Fig 1).

Study characteristics & quality assessment

The included studies were published between 1995 and 11th of May 2015. No randomised

controlled trials were identified; twelve case-control studies and 10 cohort studies were identi-

fied. Four studies were excluded on the basis of quality [two fair studies [48, 49] and two poor

studies [50, 51]]. Overall eighteen studies were included; nine case control studies [14–16, 46,

52–56] and nine cohort studies [37, 45, 47, 57–62]. Further details can be found in Table 1.

Clinical differences for definition of drug exposure were compared across all studies. Studies

were compared for differences in the context of their setting including: of location, design,

fracture type, mean age, sample size, length of drug exposure and adjustment for confounders

with particular attention to dose. The included sample sizes ranged from 500 to 906,422 partic-

ipants. The mean age of participants in the included studies ranged from 72.0 to 84.3 years.

Further details can be found in S4 File.

Effect of BNZ compared to non-exposure

Eighteen studies were included [13–16, 37, 45, 47, 52–62]. There was an associated increase in

hip fracture risk with BNZ use (RR = 1.52, 95% CI 1.37–1.68, P<0.001, I2 = 67%; Fig 2). Severe

heterogeneity was explained by the varying length of usage; therefore the risk of fracture was

dependent on the length of use.

Short term use carried a 140% increased risk of hip fracture (RR = 2.40, 95% CI 1.88–3.05,

P<0.001, I2 = 27%). Medium term use carried 53% increased risk (RR = 1.53, 95% CI 1.22–

1.92, P<0.001, I2 = 0%) and long term use carried 20% increased risk (RR = 1.20, 95% CI

1.08–1.34, P<0.001, I2 = 0%). The mixed length of use subgroup carried a 52% increased risk




(RR = 1.52, 95% CI 1.35–1.71, P<0.001, I2 = 59%), but given the heterogeneous nature of this

group, this finding should be interpreted cautiously. We explored possible reporting bias

within and between subgroup using funnel plots, and conclude there was no reason to suspect

this bias after accounting for the length of use excluding mixed use studies (S5 File).

Population based studies

Population based studies were presented separately, with a range of results [13, 45, 62] (Fig 3).

Severe heterogeneity was exhibited, due to their clinical diversity and magnitude of effect.

Thus, no pooling was carried out. However, all population based studies demonstrated an

increased risk of hip fracture following BNZ and following Z-drugs [13, 45, 62].

Effect of Z-drugs compared to non-exposure

There were 6 studies included [46, 47, 52, 55, 56, 62]. There was an associated increase in hip

fracture risk with Z-drug use (RR = 1.90, 95% CI 1.68–2.13, P<0.001, I2 = 26%) (Fig 4). Short

term use carried a 139% increased risk of hip fracture (RR = 2.39, 95% CI 1.74–3.29, P<0.001,

I2 = 26%). Mixed use carried an 80% increased risk (RR = 1.80, 95%CI 1.60–2.02, P = 0.001,

I2 = 0%).

Fig 1. PRISMA flowchart: Study selection for systematic review and meta analysis.

https://doi.org/10.1371/journal.pone.0174730.g001





Subgroup analyses. The following subgroup analyses were was to explore the heterogene-

ity and found to explain the heterogeneity: duration of medication; case mix of patients

(insomniac-only studies or not); and the type of study design (population based or non-popu-

lation based studies).

Discussion

This is the first meta-analysis to have compared the association between hypnotic medications

and hip fracture, according the length of usage. We found an increase in the association

between both BNZ and Z-drug use, and hip fracture. There appeared little difference in the

findings between BNZ and Z-drugs. Our findings reinforce the evidence base that has

highlighted the increased risk of hip fracture following BNZ or zopiclone use [27–29].

Importantly this study has outlined the potential dangers of all Z-drugs in relation to hip

fractures, not solely zopiclone.

Furthermore, we found that the risk of fracture depended on the length of time people used

their medication, since newly prescribed users of BNZ and Z-drugs were at the greatest risk of

hip fracture, as previously postulated in the literature [36, 37]. This suggests the need for fur-

ther supportive multifactorial intervention to prevent falls, these may include: education about

the risks; strength and balance training; home assessment; vision assessment and referral; or

Table 1. Quality assessment using the Newcastle Ottawa scale.

Selection (S) Comparability (C) Exposure /Outcome E/O Sub Total assessment

1 2 3 4 1a 1b 1 2 3 S+ C& E/O& Conclusion

Case Control Studies:

(Berry et al. 2013) * * * * * * * No * Good Good Good Good

(Chang et al. 2008) * * * * * * * * * Good Good Good Good

(Coutinho et al. 2008) * * * * * * No * * Good Good Good Good

(Golden et al. 2010) * * * * * * * * * Good Good Good Good

(Hoffmann et al. 2006) No * No * * * * No * Fair Good Good Fair

(Jensen et al. 1991) * * * * * * No Unclear No Good Good Poor Poor

(Kang et al. 2012) * * * * * * * No * Good Good Good Good

(Lichtenstein et al. 1994) * * * No * * * No No Good Good Poor Poor

(Perreault et al. 2008) * * * * * * * * * Good Good Good Good

(Pierfitte et al. 2001) * * * * * * * * * Good Good Good Good

(Wang et al. 2001) * * * * * * * * * Good Good Good Good

(Zint et al. 2010) * * No * * * * * * Good Good Good Good

Selection (S) Comparability (C) Outcome (O)

1 2 3 4 1a 1b 1 2 3

Cohort Studies:

(Bakken et al. 2014) * * * * * * * * * Good Good Good Good

(Chan et al. 2010) * * * No * * * * * Good Good Good Good

(Cummings et al. 1995) * * * * * * * * * Good Good Good Good

(Ensrud et al. 2003) * * * * * * * * * Good Good Good Good

(Finkle et. al 2011) * * * * * * * * * Good Good Good Good

(Guo et al. 1998) * * * * * * * * * Good Good Good Good

(Huybrechts et al. 2011) No No * * * * * * No Fair Good Good Fair

(Kragh et al. 2011) * * * No * * * * * Good Good Good Good

(Thorell et al. 2014) * * * No * * * * * Good Good Good Good

(Wagner et al. 2004) * * * No * * * * * Good Good Good Good

+Domain scored: 0–1 (Poor); 2 (Fair); 3+ (Good);&Domain scored: 0 (Poor); 1 (Fair); 2+ (Good).

* Domain acceptable

https://doi.org/10.1371/journal.pone.0174730.t001



https://doi.org/10.1371/journal.pone.0174730.t001


Fig 2. The adjusted relative risk of hip fracture in participants who used BNZ, compared to people who did

not, by the length of use.






Fig 3. The adjusted relative risk of hip fracture in population-based studies of participants who used BNZ or a Z-drug,

compared to people who did not.


Fig 4. The adjusted relative risk of hip fracture in participants who used a Z-drug, compared to people who did not, by the length of

use.







physiotherapy [63]. These interventions should be considered at the beginning of hypnotic

prescription to reduce the risk of fall and subsequent hip fracture.

Biological mechanisms

BNZ and Z-drugs both induce sedation by enhancing the effect of the neurotransmitter GABA

within the central nervous system. Consequently they may cause drowsiness, delayed reaction

times and impair balance. We suggest that new users (up to 14 days) may be unaccustomed to

potentiated levels of GABA prior to prescription. As such their risk of fall and subsequent hip

fracture may be higher than patients with medium or long term hypnotic prescription.

NICE currently recommends that hypnotics should be prescribed to patients with severe

insomnia, at the lowest dose for the shortest period of time [17]. However, Berry et al. 2013,

Bakken et al. 2014, Chang et al. 2008 and Zint et al. 2010 suggested significant harm with

even short-term prescriptions. Our study has highlighted the immediate risk of hip fracture

amongst new BNZ or Z-drugs users, which is not addressed in the current NICE guidance [36,

37, 64]. This work raises debate of the risk-benefit of hypnotics and anxiolytics, and the need

to explore the relative effectiveness of other (often non-pharmacological) approaches to these

conditions.

Strengths & limitations

There was a consistent direction of effect across all included studies and clear findings across

all eighteen studies of an increased risk. No studies were included that directly compared BNZ

versus Z-drugs for this outcome, so no direct comparison could be made between the two

exposures. All studies included were non-randomised, and there was heterogeneity in some of

the meta-analyses, so in these findings need to be taken with caution. The majority of studies

measured dispensing or prescription data, and thus could not confirm patient adherence.

Studies adjusted for a variety of covariates, and factors (S4 File), but none of the studies address

non-registered drug use or concomitant alcohol. Many patients take BNZ and Z-drugs as and

when needed, so the exposure of these drugs may vary throughout the studies. The cause of

the fall may be in some cases due to other causes (e.g. the insomnia itself) rather than the medi-

cine given to treat it.

Recommendations for future research

Adequately powered RCTs exploring a head-to-head comparison of short acting BNZ v Z-

drugs and risk of hip fracture would be useful to investigate which medicine (if any) has the

better safety profile. Also adequately powered RCTs investigating interventions for newly pre-

scribed users are needed to minimise the ‘new user’ effect that we have found, for example the

use of a patient information sheet highlighting what expect, and what routine tasks to avoid,

when newly prescribed BNZ or Z-drugs. In addition, research into alternative non-pharmaco-

logical interventions is needed to determine the best use of hypnotics within the wider context

of clinical practice.

Recommendations for future clinical practice

Patients require clearer information about the risks associated with hypnotics. Attention

needs to be drawn to the increased risk of falling when prescribing BNZ and Z-drugs to new

patients. Clinical guidance and education needed to reduce prescribers’ perceptions con-

cerning the relative risk benefit of Z-drugs compared with BNZs. Finally, the length of hyp-

notic prescription needs to be carefully re-evaluated for each individual within the broader




clinical context. Long term prescription carries well documented risks- dependence, falls

and cognitive impairment. This study reinforces the need to carefully evaluate the indica-

tions for BNZ or Z drug prescription in older persons and to consider if multifactorial inter-

vention might be necessary, as outlined in the NICE clinical guideline “Falls in older people:

assessing risk and prevention “[63].

Conclusions

This review indicated a similar risk profile of hip fracture for individuals receiving Z-drugs or

receiving BNZs. We have highlighted the greatest risk appears to be in patients newly pre-

scribed hypnotic/anxiolytic agents. Alternative non-drug options need to be considered for

even short-term use (e.g. in treating short-term insomnia) in order to reduce the risk of fall

and subsequent hip fracture, particularly in the older person.

Supporting information

S1 File. PRISMA checklist.

(PDF)

S2 File. Systematic review protocol for “Benzodiazepines, Z-drugs and the risk of hip frac-

ture: A systematic review and meta-analysis”.

(DOCX)

S3 File. Appendix 1—Medline via Ovid search strategy.

(DOCX)

S4 File. Characteristics of included studies: (a)- Case control studies; (b)—Cohort studies.

(DOCX)

S5 File. Sensitivity analysis funnel-plot, showing the three subgroup.

(DOCX)

Author Contributions

Conceptualization: KD RB PR.

Formal analysis: KD BC.

Investigation: KD BC.

Methodology: BC.

Project administration: KD.

Software: KD BC.

Supervision: RB PR BC.

Validation: KD BC.

Visualization: KD.

Writing – original draft: KD BC.

Writing – review & editing: KD BC JH RB PR.



http://www.plosone.org/article/fetchSingleRepresentation.action?uri=info:doi/10.1371/journal.pone.0174730.s001






References1. Boddaert J, Raux M, Khiami F, Riou B. Hip fractures: Epidemiology and risk factors. Annales francaises

de medecine d’urgence. 2015; 5(2):119–25.

2. National Institute of Clinical Excellence. Costing Report: Hip Fracture. London: 2011.

3. Leibson C, Tosteson ANA, Gabriel S, Ransom J, Melton L. Mortality, disability, and nursing home use

for persons with and without hip fracture: A population- based study. Journal Of The American Geriat-

rics Society. 2002; 50(10):1644–50. PMID: 12366617

4. Klop C, Welsing P, Cooper C, Harvey NC, Elders P, Bijlsma J, et al. Mortality in British hip fracture

patients, 2000–2010: A population-based retrospective cohort study. Bone. 2014; 66:171–7. https://doi.

org/10.1016/j.bone.2014.06.011 PMID: 24933345

5. Huddleston J, Whitford K. Medical care of elderly patients with hip fractures. Mayo Clinical Proceedings.

2001; 3(76):295.

6. Centers for Disease Control and Prevention. Hip Fractures Among Older Adults 2016 [cited 2017 26

Jan 2017]. https://www.cdc.gov/homeandrecreationalsafety/falls/adulthipfx.html.

7. Kanis JA, Borgstrom F, Compston J, Dreinhofer K, Nolte E, Jonsson L, et al. SCOPE: a scorecard for

osteoporosis in Europe. Archives of Osteoporosis. 2013; 8(1):1–63.

8. Sahota O, Morgan N, Moran CG. The direct cost of acute hip fracture care in care home residents in the

UK. Osteoporosis International. 2012; 3(23):917–20. Epub May 8 2011.

9. LaVelle D. Fractures of hip. In: Campbell’s Operative Orthopaedics. 2003. Epub 10th, Canale ST.

10. Dy C, McCollister K, Lubarsky D, Lane J. An economic evaluation of a systems-based strategy to expe-

dite surgical treatment of hip fractures. Journal of Bone and Joint Surgery 2011 93(14):1326–34.

https://doi.org/10.2106/JBJS.I.01132 PMID: 21792499

11. Glass J, Lanctot K, Herrmann N, Sproule B, Busto U. Sedative hypnotics in older people with insomnia:

meta-analysis of risks and benefits. The BMJ. 2005; 331(1169).

12. Neutel C, Perry S, Maxwell C. Medication use and risk of falls Pharmacoepidemiology and Drug Safety.

2002;(11):97–104.

13. Bolton JM, Metge C, Lix L, Prior H, Sareen J, Leslie WD. Fracture risk from psychotropic medications: a

population-based analysis. Journal of Clinical Psychopharmacology. 2008; 28(4):384–91. https://doi.

org/10.1097/JCP.0b013e31817d5943 PMID: 18626264

14. Chang CM, Wu ECH, Chang IS, Lin KM. Benzodiazepine and risk of hip fractures in older people: A

nested case-control study in Taiwan. American Journal of Geriatric Psychiatry. 2008; 16(8):686–92.

https://doi.org/10.1097/JGP.0b013e31817c6a99 PMID: 18669947

15. Coutinho ES, Fletcher A, Bloch KV, Rodrigues LC. Risk factors for falls with severe fracture in elderly

people living in a middle-income country: a case control study. BMC Geriatrics. 2008; 8:21. https://doi.

org/10.1186/1471-2318-8-21 PMID: 18727832

16. Golden AG, Ma Q, Nair V, Florez HJ, Roos BA. Risk for fractures with centrally acting muscle relaxants:

an analysis of a national Medicare Advantage claims database. Annals of Pharmacotherapy. 2010; 44

(9):1369–75. https://doi.org/10.1345/aph.1P210 PMID: 20606016

17. National Institute for Health and Clinical Excellence. Guidance on the use of zaleplon, zolpidem and

zopiclone for the short-term management of insomnia. London: NICE; 2004.

18. Griffin C, Kaye A, Bueno F, Kaye A. Benzodiazepine Pharmacology and Central Nervous System–

Mediated Effects. The Ochsner Journal. 2013; 13(2):214–23. PMID: 23789008

19. Gunja N. The Clinical and Forensic Toxicology of Z-drugs. Journal of Medical Toxicology. 2013; 9

(12):155–62.

20. Siriwardena A, Qureshi Z, Gibson S, Collier S, Latham M. GPs’ attitudes to benzodiazepine and ’ Z-

drug’ prescribing: a barrier to implementation of evidence and guidance on hypnotics. British Journal Of

General Practice. 2006; 56(533):964–7. PMID: 17132386

21. Kapil V, Green J, Le Lait C, Wood D, Dargan P. Misuse of benzodiazepines and Z-drugs in the UK. The

British Journal of Psychiatry. 2014; 5(205):407–8.

22. Olfson M, King M, Schoenbaum M. Benzodiazepine Use in the United States. JAMA Psychiatry. 2015;

2(72):136–42.

23. Alves-dos-Reis T, Papoila A, Gusmão R. Changes in prescribing patterns of benzodiazepines after

training of general practitioners. European Psychiatry. 2016; 33:S86.

24. Alessi-Severini S, Boulton J, Enns M, Dahl M, Collins D, Chateau D, et al. Use of benzodiazepines and

related drugs in Manitoba:a population-based study. Canadian Medical Association Journal Open.

2014; 2(4):208–16.



http://www.ncbi.nlm.nih.gov/pubmed/12366617

https://doi.org/10.1016/j.bone.2014.06.011

https://doi.org/10.1016/j.bone.2014.06.011


https://www.cdc.gov/homeandrecreationalsafety/falls/adulthipfx.html

https://doi.org/10.2106/JBJS.I.01132


https://doi.org/10.1097/JCP.0b013e31817d5943

https://doi.org/10.1097/JCP.0b013e31817d5943


https://doi.org/10.1097/JGP.0b013e31817c6a99


https://doi.org/10.1186/1471-2318-8-21

https://doi.org/10.1186/1471-2318-8-21


https://doi.org/10.1345/aph.1P210





25. De Wilde S, Carey I, Harris T, Richards N, Victor C, Hilton S, et al. Trends in potentially inappropriate

prescribing amongst older UK primary care patients. Pharmacoepidemiology and Drug Safety. 2007;

16:658–67. https://doi.org/10.1002/pds.1306 PMID: 16906628

26. Smolders M, Laurant M, van Rijswijk E, Mulder J, Braspenning J, Verhaak P, et al. The impact of co-

morbidity on GPs’ pharmacological treatment decisions for patients with an anxiety disorder. Family

Practice. 2007; 24(6):538–46. https://doi.org/10.1093/fampra/cmm062 PMID: 18003604

27. Xing D, Ma XL, Ma JX, Wang J, Yang Y, Chen Y. Association between use of benzodiazepines and risk

of fractures: A meta-analysis. Osteoporosis International. 2014; 25(1):105–20. https://doi.org/10.1007/

s00198-013-2446-y PMID: 24013517

28. Khong TP, de Vries F, Goldenberg J, Klungel O, Robinson N, Ibanez L, et al. Potential Impact of Benzo-

diazepine Use on the Rate of Hip Fractures in Five Large European Countries and the United States.

Calcified Tissue International. 2012; 91(1):24–31. https://doi.org/10.1007/s00223-012-9603-8 PMID:

22566242

29. Park S, Ryu J, Lee D, Shin D, Yun J, Lee J. Zolpidem use and risk of fractures: a systematic review and

meta-analysis. Osteoporosis International. 2016; 27(10):2935–44. https://doi.org/10.1007/s00198-016-

3605-8 PMID: 27105645

30. Canadian Pharmacists Association. Psychiatric Disorders. 2014. In: Compendium of Therapeutic

Choices (CTC) [Internet]. 7. www.pharmacists.ca/cpha-ca/assets/File/CTC7_Sample%20Chapter_

Insomnia.pdf.

31. Physicians’ Desk Reference. Ativan Tablets (lorazepam)—Drug Summary 2016 [07/09/2016]. http://

www.pdr.net/drug-summary/Ativan-Tablets-lorazepam-2135.1869.

32. Physicians’ Desk Reference. Xanax (alprazolam)—Drug Summary 2016 [07/09/2016]. http://www.pdr.

net/drug-summary/Xanax-alprazolam-1873.31.

33. Physicians’ Desk Reference. Valium (diazepam)—Drug Summary 2016 [17/09/2016]. http://www.pdr.

net/drug-summary/Valium-diazepam-2100.1196.

34. Physicians’ Desk Reference. Librium (chlordiazepoxide hydrochloride)—Drug Summary 2016 [07/09/

2016]. http://www.pdr.net/drug-summary/Librium-chlordiazepoxide-hydrochloride-2717.

35. Joint Formulary Commitee. Hypnotics and Anxiolytics. 2017 [cited 26 Jan 2017]. In: British National For-

mulary [Internet]. London: BMJ Group and Pharmaceutical Press, [cited 26 Jan 2017]. https://www.

medicinescomplete.com/mc/bnf/current/PHP78112-hypnotics-and-anxiolytics.htm?q=

onlywhenitissevere%2Cdisabling%2Corcausingthepatientextremedistress&t=search&ss=text&tot=

1&p=1-_hit.

36. Mets MAJ, Volkerts ER, Olivier B, Verster JC. Effect of hypnotic drugs on body balance and standing

steadiness. Sleep Medicine Reviews. 2010; 14(4):259–67. https://doi.org/10.1016/j.smrv.2009.10.008

PMID: 20171127

37. Wagner AK, Zhang F, Soumerai SB, Walker AM, Gurwitz JH, Glynn RJ, et al. Benzodiazepine use and

hip fractures in the elderly: who is at greatest risk? Archives of Internal Medicine. 2004; 164(14):1567–

72. https://doi.org/10.1001/archinte.164.14.1567 PMID: 15277291

38. Moher D, Liberati A, Tetzlaff J, Altman DG, The PG. Preferred Reporting Items for Systematic Reviews

and Meta-Analyses: The PRISMA Statement. PLOS Medicine. 2009; 6(7):e1000097. https://doi.org/10.

1371/journal.pmed.1000097 PMID: 19621072

39. Joint Formularly Committee. Clonazepam. 2017 [cited 07 Jan 2017]. In: British National Formularly

[Internet]. London: BMJ Group and Pharmaceutical Press, [cited 07 Jan 2017]. https://www.

medicinescomplete.com/mc/bnf/current/PHP3010-clonazepam.htm?q=clonazepam&t=search&ss=

text&tot=32&p=1-_hit.

40. Joint Formulary Committee. Benzodiazepines. 2017 [cited 07 Jan 2017]. In: British National Formularly

[Internet]. London: BMJ Group and Pharmaceutical Press, [cited 07 Jan 2017]. https://www.

medicinescomplete.com/mc/bnf/current/_882653840.htm?q=benzodiazepines&t=search&ss=

text&tot=29&p=17-_hit.

41. Likis F, Andrews J, Fonnesbeck C, Hartmann K, Jerome R, Potter S, et al. Methods. 2014 [cited 17 Jan

2017]. In: Smoking Cessation Interventions in Pregnancy and Postpartum Care Evidence Report/Tech-

nology Assessment No214 [Internet]. US: Rockvill, MD: Agency for Healthcare Research and Quality,

[cited 17 Jan 2017]. www.effectivehealthcare.ahrq.gov/reports/final.cfm.

42. Reeves B, Deeks J, Higgins J, Wells G. Chapter 13: Including Non-Randomized Studies. 2008. In:

Cochrane Handbook for Systematic Reviews of Interventions [Internet]. The Cochrane Collaboration.

www.handbook.cochrane.org.

43. Deeks J, Higgins J, Altman D. Chapter 9: Analysing data and undertaking meta-analyses. 2011. In:

Cochrane Handbook for Systematic Reviews of Interventions [Internet]. The Cochrane Collaboration.

www.handbook.cochrane.org.



https://doi.org/10.1002/pds.1306


https://doi.org/10.1093/fampra/cmm062


https://doi.org/10.1007/s00198-013-2446-y

https://doi.org/10.1007/s00198-013-2446-y


https://doi.org/10.1007/s00223-012-9603-8


https://doi.org/10.1007/s00198-016-3605-8

https://doi.org/10.1007/s00198-016-3605-8


http://www.pharmacists.ca/cpha-ca/assets/File/CTC7_Sample%20Chapter_Insomnia.pdf

http://www.pharmacists.ca/cpha-ca/assets/File/CTC7_Sample%20Chapter_Insomnia.pdf

http://www.pdr.net/drug-summary/Ativan-Tablets-lorazepam-2135.1869

http://www.pdr.net/drug-summary/Ativan-Tablets-lorazepam-2135.1869

http://www.pdr.net/drug-summary/Xanax-alprazolam-1873.31

http://www.pdr.net/drug-summary/Xanax-alprazolam-1873.31

http://www.pdr.net/drug-summary/Valium-diazepam-2100.1196

http://www.pdr.net/drug-summary/Valium-diazepam-2100.1196

http://www.pdr.net/drug-summary/Librium-chlordiazepoxide-hydrochloride-2717

https://www.medicinescomplete.com/mc/bnf/current/PHP78112-hypnotics-and-anxiolytics.htm?q=onlywhenitissevere%2Cdisabling%2Corcausingthepatientextremedistress&t=search&ss=text&tot=1&p=1-_hit




https://doi.org/10.1016/j.smrv.2009.10.008


https://doi.org/10.1001/archinte.164.14.1567


https://doi.org/10.1371/journal.pmed.1000097

https://doi.org/10.1371/journal.pmed.1000097


https://www.medicinescomplete.com/mc/bnf/current/PHP3010-clonazepam.htm?q=clonazepam&t=search&ss=text&tot=32&p=1-_hit



https://www.medicinescomplete.com/mc/bnf/current/_882653840.htm?q=benzodiazepines&t=search&ss=text&tot=29&p=17-_hit



http://www.effectivehealthcare.ahrq.gov/reports/final.cfm

http://www.handbook.cochrane.org

http://www.handbook.cochrane.org


44. Higgins J, Green S. Cochrane Handbook for Systematic Reviews of Interventions. The Cochrane Col-

laboration 2011. http://handbook.cochrane.org.

45. Bakken MS, Engeland A, Engesaeter LB, Ranhoff AH, Hunskaar S, Ruths S. Risk of hip fracture among

older people using anxiolytic and hypnotic drugs: a nationwide prospective cohort study. European

Journal of Clinical Pharmacology. 2014; 70(7):873–80. https://doi.org/10.1007/s00228-014-1684-z

PMID: 24810612

46. Berry SD, Lee Y, Cai S, Dore DD. Nonbenzodiazepine sleep medication use and hip fractures in nursing

home residents. JAMA Internal Medicine. 2013; 173(9):754–61. https://doi.org/10.1001/

jamainternmed.2013.3795 PMID: 23460413

47. Finkle WD, Der JS, Greenland S, Adams JL, Ridgeway G, Blaschke T, et al. Risk of fractures requiring

hospitalization after an initial prescription for zolpidem, alprazolam, lorazepam, or diazepam in older

adults. Journal of the American Geriatrics Society. 2011; 59(10):1883–90. https://doi.org/10.1111/j.

1532-5415.2011.03591.x PMID: 22091502

48. Hoffmann F, Glaeske G. New use of benzodiazepines and the risk of hip fracture: A case-crossover

study. Zeitschrift fur Gerontologie und Geriatrie. 2006; 39(2):143–8. https://doi.org/10.1007/s00391-

006-0337-y PMID: 16622636

49. Huybrechts KF, Rothman KJ, Silliman RA, Brookhart MA, Schneeweiss S. Risk of death and hospital

admission for major medical events after initiation of psychotropic medications in older adults admitted

to nursing homes. Canadian Medical Association Journal. 2011; 183(7):E411–9. https://doi.org/10.

1503/cmaj.101406 PMID: 21444611

50. Jensen J, Nielsen LH, Lyhne N, Hallas J, Brosen K, Gram LF. Drugs and femoral neck fracture: a case-

control study. Journal of Internal Medicine. 1991; 229(1):29–33. PMID: 1995760

51. Lichtenstein MJ, Griffin MR, Cornell JE, Malcolm E, Ray WA. Risk factors for hip fractures occurring in

the hospital. American Journal of Epidemiology. 1994; 140(9):830–8. PMID: 7977293

52. Kang DY, Park S, Rhee CW, Kim YJ, Choi NK, Lee J, et al. Zolpidem use and risk of fracture in elderly

insomnia patients. Journal of Preventive Medicine & Public Health / Yebang Uihakhoe Chi. 2012; 45

(4):219–26.

53. Perreault S, Dragomir A, Blais L, Moride Y, Rossignol M, Ste-Marie LG, et al. Population-based

study of the effectiveness of bone-specific drugs in reducing the risk of osteoporotic fracture. Phar-

macoepidemiology and Drug Safety. 2008; 17(3):248–59. https://doi.org/10.1002/pds.1551 PMID:

18213734

54. Pierfitte C, Macouillard G, Thicoipe M, Chaslerie A, Pehourcq F, Aissou M, et al. Benzodiazepines

and hip fractures in elderly people: case-control study. BMJ. 2001; 322(7288):704–8. PMID:

11264208

55. Zint K, Haefeli WE, Glynn RJ, Mogun H, Avorn J, Sturmer T. Impact of drug interactions, dosage, and

duration of therapy on the risk of hip fracture associated with benzodiazepine use in older adults. Phar-

macoepidemiology and Drug Safety. 2010; 19(12):1248–55. https://doi.org/10.1002/pds.2031 PMID:

20931664

56. Wang PS, Bohn RL, Glynn RJ, Mogun H, Avorn J. Zolpidem use and hip fractures in older people. Jour-

nal of the American Geriatrics Society. 2001; 49(12):1685–90. PMID: 11844004

57. Chan ALF, Lin SJ. Trends of benzodiazepine prescribing and the risk of hip fracture in elderly patients in

Taiwan: A population-based study. International Journal of Psychiatry in Clinical Practice. 2010; 14

(1):47–52. https://doi.org/10.3109/13651500903434461 PMID: 24917232

58. Cummings SR, Nevitt MC, Browner WS, Stone K, Fox KM, Ensrud KE, et al. Risk factors for hip fracture

in white women. Study of Osteoporotic Fractures Research Group. New England Journal of Medicine.

1995; 332(12):767–73. https://doi.org/10.1056/NEJM199503233321202 PMID: 7862179

59. Ensrud KE, Blackwell T, Mangione CM, Bowman PJ, Bauer DC, Schwartz A, et al. Central nervous sys-

tem active medications and risk for fractures in older women. Archives of Internal Medicine. 2003; 163

(8):949–57. https://doi.org/10.1001/archinte.163.8.949 PMID: 12719205

60. Guo Z, Wills P, Viitanen M, Fastbom J, Winblad B. Cognitive impairment, drug use, and the risk of hip

fracture in persons over 75 years old: a community-based prospective study. American Journal of Epi-

demiology. 1998; 148(9):887–92. PMID: 9801019

61. Kragh A, Elmståhl S, Atroshi I. Older adults’ medication use 6 months before and after hip fracture: A

population-based cohort study. Journal of the American Geriatrics Society. 2011; 59(5):863–8. https://

doi.org/10.1111/j.1532-5415.2011.03372.x PMID: 21517788

62. Thorell K, Ranstad K, Midlov P, Borgquist L, Halling A. Is use of fall risk- increasing drugs in an elderly

population associated with an increased risk of hip fracture, after adjustment for multimorbidity level: a

cohort study. BMC Geriatrics. 2014; 14:131. https://doi.org/10.1186/1471-2318-14-131 PMID:

25475854



http://handbook.cochrane.org/

https://doi.org/10.1007/s00228-014-1684-z


https://doi.org/10.1001/jamainternmed.2013.3795

https://doi.org/10.1001/jamainternmed.2013.3795


https://doi.org/10.1111/j.1532-5415.2011.03591.x

https://doi.org/10.1111/j.1532-5415.2011.03591.x


https://doi.org/10.1007/s00391-006-0337-y

https://doi.org/10.1007/s00391-006-0337-y


https://doi.org/10.1503/cmaj.101406

https://doi.org/10.1503/cmaj.101406










https://doi.org/10.3109/13651500903434461


https://doi.org/10.1056/NEJM199503233321202


https://doi.org/10.1001/archinte.163.8.949



https://doi.org/10.1111/j.1532-5415.2011.03372.x

https://doi.org/10.1111/j.1532-5415.2011.03372.x


https://doi.org/10.1186/1471-2318-14-131



63. National Institute for Health and Clinical Excellence. Falls in older people: assessing risk and preven-

tion. London: NICE; 2013.

64. Gustavsen I, Bramness JG, Skurtveit S, Engeland A, Neutel I, Mørland J. Road traffic accident risk

related to prescriptions of the hypnotics zopiclone, zolpidem, flunitrazepam and nitrazepam. Sleep Med-

icine. 2008; 9(8):818–22. http://dx.doi.org/10.1016/j.sleep.2007.11.011. PMID: 18226959



http://dx.doi.org/10.1016/j.sleep.2007.11.011



For personal use. Only reproduce with permission from The Lancet Publishing Group.

Summary

Background The triptans, selective serotonin 5-HT1B/1D

agonists, are very effective acute migraine drugs with a well-developed scientific rationale. Seven different triptans willsoon be clinically available, making evidence-based selectionguidelines necessary. Triptan trials have similar designs,facilitating meta-analysis; this will provide a foundation forusing triptans in clinical practice.

Method We asked pharmaceutical companies and theprincipal investigators of company-independent trials forraw patient data of all double-blind, randomised, controlled,clinical trials of oral triptans in migraine. We calculatedsummary estimates across studies for important efficacyand tolerability parameters, and separately summariseddirect comparator trials.

Results 53 clinical trials (12 unpublished) involving 24 089patients, met the criteria for inclusion. Mean results for 100 mg sumatriptan were 59% (95% CI 57–60) for 2 hheadache response (improvement from moderate or severeto mild or no pain); 29% (27–30) for 2 h pain free(improvement to no pain); 20% (18–21) for sustained painfree (pain free by 2 h and no headache recurrence or use ofrescue medication 2–24 h post dose); and 67% (63–70) forconsistency (response in at least two of three treatedattacks); placebo-subtracted proportions for patients with atleast one adverse event (AE) were 13% (8–18), for at leastone central nervous system AE 6% (3–9), and for at least onechest AE 1·9% (1·0–2·7). Compared with these data, 10 mgrizatriptan showed better efficacy and consistency, andsimilar tolerability; 80 mg eletriptan showed better efficacy,similar consistency, but lower tolerability; 12·5 mgalmotriptan showed similar efficacy at 2 h but better otherresults; 2·5 mg naratriptan and 20 mg eletriptan showedlower efficacy and (the first two) better tolerability; 2·5 mgand 5 mg zolmitriptan, 40 mg eletriptan, and 5 mg rizatriptanshowed very similar results. The results of the 22 trials thatdirectly compared triptans show the same overall pattern. Wereceived no data on frovatriptan, but publicly available datasuggest lower efficacy.

Interpretation At marketed doses, all oral triptans wereeffective and well tolerated. 10 mg rizatriptan, 80 mgeletriptan, and 12·5 mg almotriptan provide the highestlikelihood of consistent success.

Lancet 2001; 358: 1668–75

IntroductionMigraine is a common, chronic, multifactorialneurovascular disorder, typically characterised byrecurrent disabling attacks of severe headache,autonomic nervous system dysfunction and, in up to athird of patients, neurological aura symptoms.1,2 Ergotderivatives used to be the only specific treatments formigraine attacks, although they had many limitations.3,4

Improved understanding of the neurobiology ofmigraine and 5-HT (5-hydroxytryptamine serotonin)receptors have resulted in a new class of selective 5-HT1B/1D agonists, known as the triptans.5 They have threemain mechanisms of action: cranial vasoconstriction,peripheral trigeminal inhibition, and inhibition oftransmission through second order neurons of thetrigeminocervical complex.6 The relative importance ofeach of these mechanisms remains uncertain.7,8 Bycontrast with ergots, triptans have selectivepharmacology, simple and consistent pharmacokinetics,evidence-based prescribing instructions, well establishedefficacy, modest side-effects, and a well establishedsafety record; they are, however, also contraindicated inthe presence of cardiovascular disease.9 Despite thehigher price, triptans were preferred over ergots in mostpatients.3,4

Given that seven different triptans will soon beclinically available, physicians need evidence-basedguidelines to select the triptans with the highestlikelihood of success. Direct active comparator trialswere available for only a few triptans and it is unlikelythat they will ever all be compared. Although suchstudies were deemed the gold standard for comparingdrugs, there were also some important caveats,complicating their interpretation.10 The triptan trialswere very similar in study methods and populations,facilitating meta-analysis to summarise the efficacy andtolerability of the different triptans across studies.11,12

Previous triptan meta-analyses were based on summarydata from published trials only, and only analysed alimited number of agents, doses, and outcome andadverse event variables.13,14

Although oral absorption of many drugs is delayedduring migraine attacks,15 most patients prefer oralformulations;16 they account for more than 80% of alltriptan prescriptions (H Mansbach, GlaxoSmithKline,personal communication). We shall thereforeconcentrate on the oral formulations. Sumatriptan isalso available in parenteral formulations; these arediscussed elsewhere.10

MethodsClinical assessment in acute migraine trialsTypically, patients were instructed to treat a migraineheadache when pain is moderate or severe on a 4-pointpain severity scale (0=no pain; 1=mild; 2=moderate;3=severe pain) and within 6–8 h of onset.11 The primaryendpoint in most studies was the proportion of patientswith a headache response (ie, improvement to mild orno pain 2 h post-dose). More recently, the proportion ofpatients who become pain free 2 h post-dose has

ARTICLES

1668 THE LANCET • Vol 358 • November 17, 2001

Oral triptans (serotonin 5-HT1B/1D agonists) in acute migrainetreatment: a meta-analysis of 53 trials

Michel D Ferrari , Krista I Roon, Richard B Lipton, Peter J Goadsby

Department of Neurology, Leiden University Medical Centre, 2300 RC Leiden, Albinusdreef 2, 2333 ZA Leiden, Netherlands (M D Ferrari MD, K I Roon MD); Departments of Neurology,Epidemiology and Social Medicine, Albert Einstein College ofMedicine and the Montefiore Headache Unit, New York, andInnovative Medical Research Inc, Stamford, CT, USA (R B Lipton MD);Institute of Neurology, The National Hospital for Neurology andNeurosurgery, Queen Square, London, UK (P J Goadsby MD)

Correspondence to: Dr Michel D Ferrari(e-mail: [email protected])


become the preferred and clinically most relevantprimary endpoint.10,16,17 The headache may sometimesreturn within 24 h of initial relief (headache recurrenceor relapse) requiring re-dosing.1,18–20 This is inconvenientand may lead to medication overuse.21,22 Simplecomparison of recurrence rates (as proportion ofresponders) without accounting for differences in initialrelief rates and use of rescue medications might bemisleading.1,10,23 We therefore recommend use ofsustained pain-free: the proportion of patients who werepain free by 2 h post-dose and who do not have arecurrence of moderate or severe headache and who donot use any rescue headache medication 2–24 h post-dose.1,10,23 It represents the ideal efficacy endpoint (ie,patients who require only a single dose to abort theirattack by 2 h and for at least 24 h) but also the mostdifficult one to achieve.16 Note that, with the definitionused here, recurrence of only mild headache notprompting the use of rescue medication, will not berecorded; we, however, do not consider this a clinicallysignificant recurrence. Patients also highly value aconsistent effect over recurrent attacks (intrapatientconsistency):16,17 the proportion of patients with response(or painfree) in at least two or three of three activelytreated attacks in placebo-controlled trials. Finally,tolerability and safety were mainly assessed by reportingof adverse events.

Meta-analysis of oral triptan trialsAfter a systematic review of published English trials, wesent a standard letter to all six pharmaceuticalcompanies that market triptans. The letter explained theobjectives and exact procedures of the study and askedfor raw patient data of all randomised controlled trials(both published and unpublished) that used their drug.Five companies provided all the requested data.Vanguard (now Vernalis) declined to disclose any dataon frovatriptan; data were thus extracted from congressabstracts. Where possible, we crosschecked all data withpublished or presented data. In addition, we approachedthe principal investigators of triptan trials that were notcompany-sponsored with the same request. Thecompanies received the results (not the interpretations)of the analyses, for their drug only, 2 months before theplanned submission of the manuscript, and were askedto check them for accuracy; there were no comments.The database was closed on Nov 1, 2000.

Studies and data includedStudies had to meet the following inclusion criteria:randomised, double-blind, controlled (placebo oractive) clinical trial; treatment of moderate or severemigraine attacks within 8 h of onset in migraine patients(18–65 years of age) defined according to theInternational Headache Society criteria;24 treatment withan oral triptan at a recommended clinical dose; andmeasurement of the headache on the 4-point painscale.11 We assessed in total 76 clinical trials: 53 met theeligibility criteria and 23 studies were excluded (see web tables 1 and 2 on The Lancet’s website:www.thelancet.com); the most common reasons forexclusion were lack of a control group, use of non-recommended drug doses, or selected study populations(eg, adolescents).

We combined data from placebo-controlled trials,with or without an active comparator, in the meta-analysis (per patient, only the first study attack). Datafrom direct active comparator trials were also analysedseparately. For rizatriptan, the results of both traditional

tablets and soluble wafers were combined, as the studydesigns and results were identical.

Patients classified as having at least one adverse event(AE; any AE) typically had mild and short-livedtingling, paraesthesias, warm sensations in the head,neck, chest, and limbs, or less frequently, dizziness,flushing, and neck pain or stiffness. Central nervoussystem AE refers to the proportion of patients with atleast one central nervous system AE (aesthenia,abnormal dreams, agitation, aphasia, ataxia, confusion,dizziness, somnolence, speech disorder, thinkingabnormally, tremor, vertigo, and other focalneurological symptoms). Chest AE refers to theproportion of patients with at least one chest AE (chestpressure, chest pain, radiating pain in arm, other chestfeelings, heavy arms, shortness of breath, palpitations,and anxiety).

Statistical analysisWe assessed differences in all endpoints betweentriptans and placebo with random effect models.25 Thesemodels incorporate potential heterogeneity of theendpoints among different studies by assuming thateach study estimates a unique endpoint.26 We assessedthe homogeneity of observed endpoints with the �2 test.27

None of the endpoints showed homogeneity for alltriptans. When between-studies variance is zero, thestudy is homogeneous for that triptan dose andendpoint; a random effect model is identical to a fixedeffect model. Therefore, we used random effect modelsfor all endpoints.

Although study design and eligibility criteria wereremarkably similar across the triptan trials, even smalldifferences may affect comparisons of treatment effectsacross studies. To control for these differences, at leastpartly, the placebo response may be subtracted from theactive response (placebo-subtracted proportion ortherapeutic gain). These metrics measure theincremental benefit of active drug over placebo; theimplicit assumption is that these benefits were additiveand the limitations recognised. Similarly, subtractingthe placebo AE rate from the active drug AE rate canhelp to correct the differences in the methods ofcollection and definitions of AEs among studies(therapeutic harm). These approaches may facilitateacross-trial comparisons13 and have been used in othertherapeutic areas including pain.28 A similar outcomewhen using absolute proportions and when usingplacebo-subtracted proportions increases the validity ofthe results. We will therefore present data both ways.We also calculated the active drug:placebo ratios,another strategy to control for differences across studies,but will not show the data as these provided similarresults.

ResultsSumatriptan is the first and most widely prescribedtriptan: most European countries use 100 mg as theprimary oral dose, whereas North America and someother countries use 50 mg.10,29 We selected the 100 mgdose as the single reference dose for a number ofreasons.10

Figure 1A shows the mean absolute and placebo-subtracted rates and 95% CI of the headache responseat 2 h. Compared with 100 mg sumatriptan (mean 59%[95% CI 57–60]), 10 mg rizatriptan and 80 mgeletriptan showed higher, and 2·5 mg naratriptan, 20 mg eletriptan, and 2·5 mg frovatriptan (data from abstracts only) lower response rates. 2·5 mg

ARTICLES

THE LANCET • Vol 358 • November 17, 2001 1669


zolmitriptan had a slightly higher response rate than 100mg sumatriptan (p<0·05), whereas the difference in ratefor 50 mg sumatriptan, 5 mg zolmitriptan, and 5 mgrizatriptan was not significant. There were nodifferences for the other doses and drugs. Placebo-subtracted values showed wider CI and overlap betweenmost triptans (mean for sumatriptan 100 mg=29%[95% CI 26–34]). A significant positive differencepersisted for 80 mg eletriptan (42% [95% CI 36–48])and a negative difference for 2·5 mg frovatriptan (17% [95% CI 13–20]).

Figure 1B shows the pain-free rates for each triptan.Compared with 100 mg sumatriptan (29% [95% CI27–30]), 25 mg sumatriptan, 2·5 mg naratriptan, and 20 mg eletriptan showed lower mean absolute pain-freerates, whereas 80 mg eletriptan, 12·5 mg almotriptan,and 10 mg rizatriptan showed higher values. The othertriptans and doses did not differ from 100 mgsumatriptan. Placebo-subtracted values (100 mgsumatriptan: 19% [95% CI 17–22]) were significantlyhigher for 10 mg rizatriptan and 80 mg eletriptan.

Compared with 100 mg sumatriptan (30% [95% CI27–33]), recurrence rates were lower for 40 and 80 mgeletriptan and higher for 5 and 10 mg rizatriptan (figure2A). 2·5 mg naratriptan had a lower recurrence rate, butthis is based on 4 h rather than on 2 h response rates andis therefore not directly comparable. Other recurrencerates overlap. Isolated comparison of recurrence ratesmight be misleading so we have compared sustainedpain-free rates (figure 2B). These were calculated, post-

hoc, for those trials with all available relevant data.Compared with 100 mg sumatriptan (20% [95% CI18–21]), sustained pain-free rates were higher for 10 mgrizatriptan, 80 mg eletriptan, and 12·5 mg almotriptan,and lower for 20 mg eletriptan. 25 mg sumatriptan and2·5 mg naratriptan tended to show lower values,whereas no differences were reported for the othertriptans. Because the interpretation of recurrence after aresponse following placebo is unclear no placebo-subtracted sustained pain-free rates have beencalculated.

Placebo-controlled intrapatient consistency of efficacyover multiple attacks was investigated in only a fewstudies. No such studies were available for 25 and 50 mgsumatriptan, 2·5 and 5 mg zolmitriptan, and 5 mgrizatriptan. All drugs (except 10 mg rizatriptan) weretested in a parallel-group design, treating threeconsecutive attacks with either active drug or placebo(figure 3). These studies showed that consistent lack ofresponse is rare: response in at least one of three treatedattacks occurs in 79–89% of patients (placebo: about50%) and freedom from pain in 51–59% (placebo:18%). Response in at least two of three treated attacksoccurs in 47–72% of patients (placebo: 17–33%) andfreedom from pain in 14–42% (placebo: 3–13%);highest consistency rates were for 100 mg sumatriptanand 12·5 mg almotriptan (but here placebo rates werealso highest); lowest rates were for 2·5 mg naratriptanand 25 mg sumatriptan. Response in all three attacksoccurs in 16–47% of patients (placebo: up to 9%) and

ARTICLES


Placebo-subtracted Absolute (%)

25 mg

50 mg

100 mg

2·5 mg

5 mg

2·5 mg

10 mg

5 mg

20 mg

40 mg

80 mg

12·5 mg

2·5 mg

Sumatriptan

Zolmitriptan

Naratriptan

Rizatriptan

Eletriptan

Almotriptan

Frovatriptan

Response at 2 h

200 40 60 80

A

Placebo-subtracted Absolute (%)

25 mg

50 mg

100 mg

2·5 mg

5 mg

2·5 mg

5 mg

10 mg

20 mg

40 mg

80 mg

12·5 mg

Sumatriptan

Zolmitriptan

Naratriptan

Rizatriptan

Eletriptan

Almotriptan

Pain free at 2 h

100 20 30 40 50

B

Figure 1: Absolute and placebo subtracted efficacy results at 2 h A: rates of headache response; B: rates of pain-free. Mean and 95% CIs given for each triptan. Grey shaded regions are the 95% CIs for 100 mgsumatriptan.


ARTICLES


freedom from pain in 1–17% (placebo: <2%); highestconsistency rates were for 100 mg sumatriptan and 12·5mg almotriptan (with highest placebo rates).

The consistency of 10 mg rizatriptan was assessed in adouble-blind, crossover design over four attacks, withplacebo in one attack interspersed at random in four offive patient groups; the fifth group received 10 mgrizatriptan for four attacks.30 The different design of thisstudy complicates a comparison with the otherconsistency rates, although it seems unlikely that itwould have increased consistency. Consistency ratesover three attacks were the highest of all triptans:response (and pain-free) rates were 96% (77%) in at least one of three, 86% (48%) in at least two of three and 60% (20%) in all three actively treatedattacks.31 In the subgroup of 125 patients who treatedthree consecutive attacks wih rizatriptan, without priorexposure to placebo, the results were very similar:response (and pain-free) rates were 87% (42%) in atleast two of three attacks and 50% (16%) in all threeattacks.

In figure 4 values greater than zero indicate that AEoccurred in more patients for active drug than forplacebo; values with narrow 95% CIs that cross the zeroline indicate placebo-like incidences. 100 mgsumatriptan had a mean placebo-subtracted rate of anyAEs of 13% (95% CI 8–18). Rates for other triptansoverlap, except for lower values for 2·5 mg naratriptanand 12·5 mg almotriptan; these rates also do not differfrom placebo. A similar pattern emerged when only AEs

were included which were (blindly) considered by thetrial investigator as drug-related (data not shown). Forcentral nervous system AEs compared with 100 mgsumatriptan (6% [95% CI 3–9]), 80 mg eletriptanshowed higher and 12·5 mg almotriptan lower values.For chest AEs compared with 100 mg sumatriptan(1·9% [95% CI 1·0–2·7]), 12·5 mg almotriptan showeda lower value. All other incidences overlap.

Comparison of the results for placebo andsumatriptan were discussed in detail elsewhere;10 thesedata serve as internal standards to check formethodological differences among the studiesconducted by the different companies. The placeborates proved remarkably consistent across mostcompanies except for very high efficacy and low AErates in the almotriptan studies, and very low efficacyand high AE rates in the eletriptan studies. Thesumatriptan efficacy rates were very consistent acrosscompanies except for low pain-free and sustained pain-free rates in the comparator studies versus eletriptan.The sumatriptan AE rates vary markedly; they were notably low in the comparator study versusalmotriptan.

Webtable 3 summarises all 22 eligible trials thatcompared one triptan with another, or with ergotamine;they are reviewed in detail elsewhere.10 The mainefficacy and AE differences (and 95% CIs) between thetwo indicated compounds were listed; the primary studyendpoints and appropriate statistics were indicated withgrey boxes. Differences were generally small, which is to

Recurrence of headache 2–24 h

50403020100

Sumatriptan

25 mg

50 mg

100 mg

Zolmitriptan2·5 mg

5 mg

Naratriptan 2·5 mg

Rizatriptan5 mg

10 mg

Eletriptan

20 mg

40 mg

80 mg

Almotriptan 12·5 mg

2–24 h

4–24 h

ASustained pain free

5 mg

Naratriptan 2·5 mg

25 mg

B

Sumatriptan 50 mg

100 mg

2·5 mgZolmitriptan

Rizatriptan5 mg

10 mg

Eletriptan

20 mg

40 mg

80 mg

Almotriptan 12·5 mg

3020100

Figure 2: Recurrence from 2–24h and sustained pain-free rateA: headache recurrence; B: sustained pain-free rates. Mean and 95% CI values given for each triptan. Grey shaded region is the 95% CI for 100 mgsumatriptan. For naratriptan the recurrence rate is given for 4–24 h post-dose (as presented in the original publications) and for 2–24 h post-dose (afterrecalculating the data).


be expected when comparing active compounds, but theoverall pattern is very similar to that in the meta-analysis.

DiscussionWe used two complementary approaches for comparingthe efficacy and tolerability of the oral triptans: a large

meta-analysis of all the eligible, high-quality,randomised, placebo-controlled clinical trials and aseparate analysis of all direct comparative studies. Bothapproaches give very similar results. Our meta-analysisused studies of a fundamentally similar design so thatsummary estimates of the efficacy and tolerability of thefull range of compounds could be derived. The use of

ARTICLES


100%

50%

0%S25 S50 S100 N2·5 R10 E20 E40 E80 A12·5

Response in one of three attacks100%

50%

0%S25 S50 S100 N2·5 R10 E20 E40 E80 A12·5

Pain free in one of three attacks

100%

50%

0%S25 S50 S100 N2·5 R10 E20 E40 E80 A12·5

Response in at least two of three attacks100%

50%

0%S25 S50 S100 N2·5 R10 E20 E40 E80 A12·5

Pain free in at least two of three attacks

100%

50%

0%S25 S50 S100 N2·5 R10 E20 E40 E80 A12·5

Response in all three attacks100%

50%

0%S25 S50 S100 N2·5 R10 E20 E40 E80 A12·5

Pain free in all three attacks

Figure 3: Intra-individual consistency2 h headache response and pain free in at least one of three attacks, at least two of three, and all three attacks for each triptan. Data are presented asgroup result and 95% CI. For each drug the white bar indicates the consistency rate for placebo. For rizatriptan this could not be calculated due to thedifferent design. S=sumatriptan. N=naratriptan. R=rizatriptan. E=eletriptan. A=almotriptan.


placebo-subtracted measures allows partial adjustmentfor the methodological differences among studies thatcould affect the results. The great strength of randomisedhead-to-head comparator trials is their internal validity.However, factors such as patient selection, study size,and encapsulation of a drug may limit the generalisibilityof the results into clinical practice.10 Furthermore, it isunlikely that all triptans will ever all be compared. Theremarkable similarity of the results from the meta-analysis of the placebo-controlled trials (both for theabsolute and placebo-subtracted rates, and the activeplacebo ratios) and from the head-to-head studies drug:reinforces the validity of the conclusions.

Safety of drugs can only be reliably assessed afterlarge-scale and long-term clinical exposure. Althoughless so than with the ergots,3,4 the main concern with alltriptans is their potential for coronary vasoconstriction.32

This has been exacerbated by the occurrence of chestsymptoms that sometimes resemble pectoral angina;33

the usual underlying mechanism, however, is notmyocardial ischaemia.1,2 When patients were warnedabout these events, they rarely cause problems.1,33 Arecent long-term post-marketing review concluded thattriptans were very safe as long as they were not used inpatients with cardiovascular disease or major riskfactors.9 Since there were no clinically importantdifferences in coronary vasoconstriction effects, notriptan is demonstrably safer than the others.

Differences in total AE rates must be interpretedcautiously since they reflect proportions of patients withat least one AE, irrespective of their number, nature, orintensity; trivial and significant AEs were thus pooled.In addition, in the almotriptan studies AE rates forplacebo and sumatriptan are remarkably low. Thisfinding could indicate different methods of collectingand defining AEs, a study population with a higherthreshold for reporting AEs, or both.

All oral triptans were more effective than placebo.

ARTICLES


Any AE: placebo subtracted

25 mg50 mg

100 mg

2·5 mg5mg

2·5 mg

5mg10 mg

20 mg40 mg80 mg

12·5 mg

Sumatriptan

Zolmitriptan

Naratriptan

Rizatriptan

Eletriptan

Almotriptan

403020100–10–20

A

403020100–10–20CNS AEs: placebo subtracted

25 mg50 mg

100 mg

2·5 mg5mg

2·5 mg

5 mg10 mg

20 mg40 mg80 mg

12·5 mg

Sumatriptan

Zolmitriptan

Naratriptan

Rizatriptan

Eletriptan

Almotriptan

B

Chest AEs: placebo subtracted

25 mg50 mg

100 mg

2·5 mg5 mg

2·5 mg

5 mg10 mg

20 mg40 mg80 mg

12·5 mg

Sumatriptan

Zolmitriptan

Naratriptan

Rizatriptan

Eletriptan

Almotriptan

7·552·52·7

0–2·5–5C

Figure 4: Placebo subtracted AE dataA: any AE, B: CNS AE, C: chest AE. Mean and 95% CI given for each triptan. Grey shaded region is the 95% CI for 100 mg sumatriptan.


Consistent lack of response is rare, as 79–89% ofpatients respond in at least one of three treated attacks.Differences among the triptans were small but wereclinically relevant for the individual patient. Comparedwith 100 mg sumatriptan, 12·5 mg almotriptan was 24%better for pain-free, 30% better for sustained pain-free,and 57% for adverse events. 80 mg eletriptan was 10% better for response and 25% better for sustainedpain-free, whereas 10 mg rizatriptan was 17% better forresponse, 38% better for pain-free, and 25% better forsustained pain-free. When including consistency over allthree attacks, the percentages for rizatriptan were evenhigher (67% for response and 58% for pain-free).

The table above compares the main efficacy andtolerability measures for the oral triptans versussumatriptan. Three compounds showed favourableresults: 10 mg rizatriptan, 80 mg eletriptan, and 12·5 mg almotriptan. In the almotriptan trials, placeboefficacy was high and AE rates for placebo andsumatriptan were very low. This suggests that thepatients in these studies were more therapy-responsiveand had a higher threshold to report AEs; however,almotriptan retained its tolerability advantage in a head-to-head study with 100 mg sumatriptan (webtable 3). Inthe direct comparator trials versus eletriptan,sumatriptan (but not eletriptan) was encapsulated (formasking purposes) and significantly underperformed forfreedom from pain compared with other trials. In apharmacokinetic study, the early absorption ofencapsulated sumatriptan was delayed compared withthat of normal sumatriptan, but the open label 2 hresponses were equivalent.34 For the other compounds,differences were minor and sometimes favour 100 mgsumatriptan. 50 mg sumatriptan, 2·5 mg and 5 mgzolmitriptan, 5 mg rizatriptan, and 40 mg eletriptanhave efficacy and tolerability profiles very similar to 100mg sumatriptan. 25 mg sumatriptan, 2·5 mgnaratriptan, and 20 mg eletriptan have inferior efficacy,but better tolerability.

Data for frovatriptan were not received nor published.Based on congress abstracts, headache response(41%; placebo 21%) and pain free (12%; placebo 3%)were well below those of the other triptans. Recurrenceand AE rates do not significantly differ from 100 mgsumatriptan; therefore, a claim for better cardiovascularsafety is unsustainable and potentially hazardous.10

Patients’ characteristics and preferences vary, andindividual responses to a triptan cannot be predicted.Finding the best therapy may involve trial and error: ifthe first triptan fails one may successfully switch toanother. Physicians thus need more than one triptan in

their repertoire to best treat patients with migraine. 10 mg rizatriptan (especially when consistent and rapidfreedom from pain is desired), 80 mg eletriptan(especially when high efficacy and low recurrence werefavoured over tolerability), and 12·5 mg almotriptan(especially when high tolerability and good efficacy werefavoured) offer the highest likelihood of success. 100 mgand 50 mg sumatriptan provide good efficacy andtolerability and by far the longest clinical experience.Sumatriptan also, and uniquely, offers non-oralformulations, allowing tailor-made treatments; the 6 mgsubcutaneous formulation is the most effective acutemigraine treatment, but is also associated with moreintense AEs and the need for self-injection.35 2·5 mgnaratriptan offers very good tolerability coupled to aslower onset of improvement; this can be useful inpatients with mild or moderate migraine. 2·5 mg and 5 mg zolmitriptan were good alternatives in manypatients; they offer no specific advantages nor flaws.Frovatriptan cannot be fully judged in view of the lack ofdata but does not seem to offer any particularadvantage.

ContributorsM D Ferrari initiated, coordinated, and supervised the study, andhelped in designing the meta-analysis, data analysis, interpretation ofthe results, and writing the paper. K I Roon coordinated the contactswith the principal investigators of the trials and the manufacturers of theseven triptans, collected raw patient data, and checked the validity ofthe data. She did the primary data analyses and statistics, contributed todata interpretation, and the writing of the paper. R B Lipton and P J Goadsby contributed to designing the meta-analysis, data analysis,interpretation of the results, and writing of the paper.

Conflict-of-interest statementM D Ferrari’s salary is fully covered by LUMC. He has received:consultancy and industry support from Allergan, Almirall, AstraZeneca,Boehringer, Glaxo Wellcome, Merck, Sharpe & Dohme, Pfizer,SmithKlineBeecham,Vanguard Medica; grant support from GlaxoWellcome, Merck, Sharpe & Dohme; and independent support fromNWO, Asclepiade, Migraine Trust, Dutch Brain Trust, Biomed EC,Dutch Heart Foundation, and the Gisela Thier Foundation.P J Goadsby’s salary is fully covered by Wellcome Trust. He hasreceived: consultancy and industry support from Allergan, Almirall,AstraZeneca, Boehringer, Glaxo Wellcome, Merck, Sharpe & Dohme,Pfizer, SmithKlineBeecham,Vanguard Medica; grant support fromWellcome Trust, Migraine Trust, Brain Research Trust/ION; a clinicalstaff salary from SKB, AstraZeneca, Glaxo Wellcome, and Pfizer,AstraZeneca, SKB. R B Lipton has received financial support in theform of research grants as a principal investigator, consulting fees, andlecture honoraria from several triptan manufacturers (Astra-Zeneca,Elan/Vanguard, Glaxo Wellcome, Merck, Sharpe & Dohme, Pfizer).Most of his funds have been paid by his employer.

AcknowledgmentsWe thank the following pharmaceutical companies and people forproviding data and for patiently answering all related questions:GlaxoWellcome (sumatriptan and naratriptan): Diane J Boswell, Scott B McNeal, Gayla P Putman, Jane Saiers, Reijo Salonen;AstraZeneca (zolmitriptan): David Lee, James Sawyer, Andrew A M Stone; Merck Sharp & Dohme (rizatriptan): Christopher Lines, Eric Sandquist; Pfizer (eletriptan): Scott Haugie,Neville Jackson, Phil Poole, John A Schalhoub; Almirall (almotriptan):Xavier Carabarrocas, Pau Ferrer, Jose Palacios. We thank Ton de Craen (Department of Clinical Epidemiology, LUMC) forstatistical and methodological advice.

P J Goadsby is supported by the Migraine Trust and the WellcomeTrust, and is a Wellcome Senior Research Fellow.

References1 Ferrari MD. Migraine. Lancet 1998; 351: 1043–51.2 Lance JW, Goadsby PJ. Mechanism and management of headache.

London: Butterworth-Heinemann, 1998.3 Quality standards subcommittee of the American Academy of

Neurology: practice parameter—appropriate use of ergotaminetartrate and dihydroergotamine in the treatment of migraine andstatus migrainosus. Neurology 1995; 45: 585–87.

4 Tfelt-Hansen P, Saxena PR, Dahlof C, et al. Ergotamine in the

ARTICLES


Initial Sustained Consistency Tolerability2 h relief pain-free

Sumatriptan 50 mg = = =/– =Sumatriptan 25 mg – =/– – +Zolmitriptan 2·5 mg = = = =Zolmitriptan 5 mg = = = =Naratriptan 2·5 mg – – – ++Rizatriptan 5 mg = = = =Rizatriptan 10 mg + + ++ =Eletriptan 20 mg – – – =Eletriptan 40 mg =/+ =/+ = =Eletriptan 80 mg +(+) + = –Almotriptan 12·5 mg = + + ++

Based on the results of the present meta-analysis and the direct comparatortrials. =indicates no difference when compared with sumatriptan. + indicatesbetter when compared with sumatriptan. – indicates inferior when comparedwith sumatriptan.

Comparison of the main efficacy and tolerability measures forthe oral triptans versus 100 mg sumatriptan


acute treatment of migraine—a review and European consensus.Brain 2000; 123: 9–18.

5 Humphrey PPA, Feniuk W, Perren MJ, et al. Serotonin andmigraine. Ann N Y Acad Sci 1990; 600: 587–98.

6 Goadsby PJ. The pharmacology of headache. Prog Neurobiol 2000;62: 509–25.

7 Humphrey PPA, Goadsby PJ. Controversies in headache. The modeof action of sumatriptan is vascular? A debate. Cephalalgia 1994; 14:401–10.

8 Goadsby PJ. 5-HT1B/1D agonists in migraine: comparativepharmacology and its therapeutic implications. CNS Drugs 1998; 10:271–86.

9 Welch KMA, Mathew NT, Stone P, et al. Tolerability ofsumatriptan: clinical trials and post-marketing experience.Cephalalgia 2000; 20: 687–94.

10 Ferrari MD, Goadsby PJ, Roon KI, Lipton RB. Triptans (5-HT1B/1D

agonists) in migraine: methods and detailed results of a meta-analysis of 53 trials. Cephalalgia (in press).

11 Pilgrim AJ. Methodology of clinical trials of sumatriptan in migraineand cluster headache. Eur Neurol 1991; 31: 295–99.

12 Tfelt-Hansen P, Block G, Dahlof C, et al. Guidelines for controlled trials of drugs in migraine: 2nd edn. Cephalalgia 2000;20: 765–86.

13 Tfelt-Hansen P. Efficacy and adverse events of subcutaneous, oral,and intranasal sumatriptan used for migraine treatment: a systematicreview based on number needed to treat. Cephalalgia 1998; 18:532–38.

14 Tfelt-Hansen P. A comparative review of pharmacology,pharmacokinetics and efficacy of triptans in migraine. Drugs 2000;6: 1259–87.

15 Volans GN. Absorption of effervescent aspirin during migraine.BMJ 1974; 2: 265–69.

16 Lipton RB, Stewart WF. Acute migraine therapy: do doctorsunderstand what patients with migraine want from therapy?Headache 1999; 39: 20–26.

17 Davies GM, Santanello NC, Lipton RB. Determinants of patientsatisfaction with migraine therapy. Cephalalgia 2000; 20: 554–60.

18 Visser WH, Jaspers NM, de Vriend RH, Ferrari MD. Risk factorsfor headache recurrence after sumatriptan: a study in 366 migrainepatients. Cephalalgia 1996; 16: 264–69.

19 Ferrari MD, James MH, Bates D, et al. Oral sumatriptan: effect of asecond dose, and incidence and treatment of headache recurrence.Cephalalgia 1994; 14: 330–38.

20 Teall J, Tuchman M, Cutler N, et al. Rizatriptan (MAXALT) forthe acute treatment of migraine and migraine recurrence:

a placebo-controlled, outpatient study. Headache 1998; 38: 281–87.21 Kaube H, May A, Diener HC, Pfaffenrath V. Sumatriptan misuse in

daily chronic headache. BMJ 1994; 308: 1573–74.22 Limmroth V, Kazarawa S, Fritsche G, Diener HC. Headache after

frequent use of new serotonin agonists zolmitriptan and naratriptan.Lancet 1999; 353: 378.

23 Ferrari MD. How to assess and compare drugs in the managementof migraine: success rates in terms of response and recurrence.Cephalalgia 1999; 19: 2–8.

24 Headache classification committee of the International HeadacheSociety: classification and diagnostic criteria for headache disorders, cranial neuralgias and facial pain. Cephalalgia 1988;8: 1–96.

25 DerSimonian R, Laird N. Meta-analysis in clinical trials. ControlClin Trials 1986; 7: 177–88.

26 Raudenbush SW. Random effect models. In: Cooper H, Hedges LV,eds. The handbook of research synthesis. New York: Russell SageFoundation, 1994: 301–21.

27 Whitehead A, Whitehead JA. General parametric approach to themeta-analysis of randomized clinical trials. Stat Med 1991; 10:1665–77.

28 McQuay HJ, Moore RA. Using numerical results from systematicreviews in clinical practice. Ann Intern Med 1997; 126: 712–20.

29 Pfaffenrath V, Cunin G, Sjonell G, Prendergast S. Efficacy andsafety of sumatriptan tablets (25 mg, 50 mg, and 100 mg) in theacute treatment of migraine; defining the optimum doses of oralsumatriptan. Headache 1998; 38: 184–90.

30 Kramer MS, Matzura-Wolfe D, Polis A, et al. A placebo-controlledcrossover study of rizatriptan in the treatment of multiple migraineattacks. Neurology 1998; 51: 773–81.

31 Dahlof CGH, Lipton RB, McCarroll KA, et al. Within-patientconsistency of response of rizatriptan for treating migraine.Neurology 2000; 55: 1511–16.

32 Fuseau E, Petricoul O, Sabin A, et al. Effects of encapsulation onabsorption of sumatriptan tablets: data from healthy volunteers andpatients during a migraine. Clin Ther 2001; 23: 242–51.

33 MaassenVanDenBrink A, Reekers M, Bax WA, Ferrari MD, Saxena PR. Coronary side-effect potential of current andprospective antimigraine drugs. Circulation 1998; 98: 25–30.

34 Ottervanger JP, Valkenburg HA, Grobbee DE, Stricker BHC.Characteristics and determinants of sumatriptan-associated chestpain. Arch Neurol 1997; 54: 1387–92.

35 Ferrari MD. The subcutaneous sumatriptan international studygroup: treatment of migraine attacks with sumatriptan. N Engl JMed 1991; 325: 316–21.

ARTICLES


More tables and references are available at www.thelancet.com

Documents

Cronograma 2020 Farmacología II - Martes 17-21 hs€¦ · Cronograma 2020 Farmacología II - Martes 17-21 hs Laboratorio de Estadística Aplicada a las Ciencias de la Salud (LEACS)