* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download New generation antipsychotics versus low
Survey
Document related concepts
Neuropharmacology wikipedia , lookup
Psychedelic therapy wikipedia , lookup
Polysubstance dependence wikipedia , lookup
Drug interaction wikipedia , lookup
Pharmacognosy wikipedia , lookup
Adherence (medicine) wikipedia , lookup
Clinical trial wikipedia , lookup
Pharmaceutical industry wikipedia , lookup
Neuropsychopharmacology wikipedia , lookup
Prescription costs wikipedia , lookup
Pharmacogenomics wikipedia , lookup
Theralizumab wikipedia , lookup
Psychopharmacology wikipedia , lookup
Transcript
ARTICLES Articles New generation antipsychotics versus low-potency conventional antipsychotics: a systematic review and meta-analysis Stefan Leucht, Kristian Wahlbeck, Johannes Hamann, Werner Kissling Summary Background The clearest advantage of new generation, atypical antipsychotics is a reduced risk of extrapyramidal side-effects (EPS), compared with conventional compounds. These findings might have been biased by the use of the highpotency antipsychotic haloperidol as a comparator in most of the trials. We aimed to establish whether the new drugs induce fewer EPS than low-potency conventional antipsychotics. Methods We did a meta-analysis of all randomised controlled trials in which new generation antipsychotics had been compared with low-potency (equivalent or less potent than chlorpromazine) conventional drugs. We included studies that met quality criteria A or B in the Cochrane Collaboration Handbook, and assessed quality with the Jadad scale. The primary outcome of interest was the number of patients who had at least one EPS. We used risk differences and 95% CIs as measures of effect size. Findings We identified 31 studies with a total of 2320 participants. Of the new generation drugs, only clozapine was associated with significantly fewer EPS (RD=–0·15, 95% CI –0·26 to –0·4, p=0·008) and higher efficacy than lowpotency conventional drugs. Reduced frequency of EPS seen with olanzapine was of borderline significance (–0·15, –0·31 to –0·01, p=0·07). Only one inconclusive trial of amisulpride, quetiapine, and risperidone and no investigations of ziprasidone and sertindole were identified, but some evidence indicates that zotepine and remoxipride do not lead to fewer EPS than low-potency antipsychotics. Mean doses less than 600 mg/day of chlorpromazine or its equivalent had no higher risk of EPS than new generation drugs. As a group, new generation drugs were moderately more efficacious than lowpotency antipsychotics, largely irrespective of the comparator doses used. Interpretation Optimum doses of low-potency conventional antipsychotics might not induce more EPS than new generation drugs. Potential advantages in efficacy of the new generation drugs should be a factor in clinical treatment decisions to use these rather than conventional drugs. Lancet 2003; 361: 1581–89 Klinik und Poliklinik für Psychiatrie und Psychotherapie der Technischen Universität München, Klinikum rechts der Isar, München, Germany (S Leucht MD, J Hamann MD, W Kissling MD); and The Zucker Hillside Hospital, Glen Oaks, New York, 11004, USA (S Leucht MD); STAKES (National Research and Development Centre for Welfare and Health), Helsinki, Finland (Prof K Wahlbeck MD) Correspondence to: Dr Stefan Leucht, Klinik und Poliklinik für Psychiatrie und Psychotherapie der Technischen Universität München, Klinikum rechts der Isar, Ismaningerstr 22, 81675 München, Germany (e-mail: [email protected]) THE LANCET • Vol 361 • May 10, 2003 • www.thelancet.com Introduction Although new generation antipsychotics are increasingly replacing conventional agents such as chlorpromazine and haloperidol in some countries, many issues about these compounds need to be clarified. Of all the new generation drugs, only clozapine has proven better than low-potency conventional drugs in patients with schizophrenia that is resistant to treatment.1 Whether the new antipsychotics have an effect on primary negative symptoms or only on secondary negative symptoms is debatable.2 Although results of a meta-analysis showed that use of the new drugs led to a modest, but significant, reduction of schizophrenic relapses, the role of improved compliance in the analysis was unclear.3 According to the reviews of the Cochrane schizophrenia group,4,5 the main advantage of the new drugs is a low risk of extrapyramidal sideeffects (EPS). Reduction of EPS is important because they have been associated with non-compliance, and can be disabling and stigmatising for patients. However, this finding was biased by the widespread use of high doses of comparator drugs, mainly haloperidol, which have a high risk of EPS. Low-potency conventional antipsychotics such as chlorpromazine are frequently used worldwide,6 especially in poorer countries, and are also known to induce fewer EPS than haloperidol. Therefore, we did a systematic review and meta-analysis of all prospective, randomised controlled trials, in which new antipsychotics had been compared with low-potency conventional compounds. The main aim of our investigation was to establish whether new generation antipsychotics induce fewer EPS than low-potency conventional drugs. Our secondary aims were to compare efficacy and drop-out rates of both generations of drugs, and to assess dose effects. Methods Identification of trials Randomised controlled trials were identified that compared new generation antipsychotics (amisulpride, clozapine, olanzapine, quetiapine, remoxipride, risperidone, sertindole, ziprasidone, and zotepine) with low-potency conventional drugs (chlorpromazine, chlorprothixene, levomepromazine, melperone, mesoridazine, methotrimeprazine, perazine, pipamperone, promethazine, prothipendyl, thioridazine). Chlorpromazine is a standard low-potency conventional antipsychotic. These drugs are labelled low-potency because high doses are needed to block dopamine substantially.7 In sufficiently high doses, low-potency drugs are not, in principle, less effective than high-potency antipsychotics such as haloperidol. All antipsychotics less potent than, or of equivalent potency to, chlorpromazine were classified as low-potency.8 We included remoxipride even though it was withdrawn from the market several years ago, because we did not expect to find many studies of these drugs, and therefore wanted to include all scientifically interesting data. 1581 For personal use. Only reproduce with permission from The Lancet Publishing Group. ARTICLES NGA NGA dose Cochrane Jadad n (mg/day)* criteria score Duration Comparator (weeks) Comparator dose Patients’ characteristics Response criterion Amisulpride16 B/3 400–1000 B 3 30 4 Perazine 400–1000 Global judgment Clozapine18 Clozapine19 310 B 800 B (300–900) 543 (–900) A 4 4 223 25 6 6 Chlorpromazine Chlorpromazine 40 12 Chlorpromazine 360 1330 (600–1600) 1163 (–1800) 5 3 40 6 Chlorpromazine 196 (75–600) Clozapine22 155 B (50–300) 600 (–1600) A 4 50 6 Chlorpromazine 600 Acute schizophrenia, schizophreniform, or schizoaffective disorder (ICD-9) Schizophrenia (NDCS) Acute schizophrenia (NDCS) Treatment resistant schizophrenia (DSM-IV) Acute schizophrenia (NDCS) Schizophrenia (NDCS) Clozapine23 400 B 3 58 8 Chlorpromazine 693 Schizophrenia (DSM-III) Clozapine17 50–600 B 2 37 8 Chlorpromazine 100–600 NI B 3 40 6 Thioridazine NI Schizophrenia (DSM-III Chinese criteria) Schizophrenia (NDCS) Clozapine25 75–450 B 4 50 7 Chlorpromazine 150–900 Clozapine26 300 B 4 64 6 Chlorpromazine Clozapine27 Clozapine20 Clozapine Clozapine 21 24 292 B 2 164 52 Chlorpromazine Acute schizophrenia (NDCS) 300 Acute schizophrenia (NDCS) 600 Geriatric schizophrenia (DSM-IV) 1200 Treatment resistant (1000–1800) + schizophrenia (DSM-III) benztropine 6 mg/day 795 Schizophrenia (DSM-II) (300–1800) 606 (–1050) Acute schizophrenia and schizoaffective disorder (DSM-II) 319 Schizophrenia (NDCS) 32 14 (5–20) A 3 30 6 Chlorpromazine 388 (200–800) Schizophrenia DSM-IV Olanzapine33 16 (5–20) A 3 41 6 Chlorpromazine 465 (200–800) Schizophrenia DSM-IV Olanzapine 12 (5–20) A 3 39 6 Chlorpromazine 232 (200–800) Schizophrenia DSM-IV Olanzapine34 25 B 4 84 8 Chlorpromazine (+ benztropine 4 mg/day) 1200 Quetiapine35 407 (–750) B 3 201 6 Chlorpromazine 384 (–750) Remoxipride36 150–600 (361) B 4 41 4 Remoxipride37 404 (150–600) B 3 144 6 Remoxipride38 238 (75–375) B 3 61 6 Remoxipride39 200 B 3 18 6 Risperidone40 7·4 (–12) B 3 42 4 B 4 41 4 B 5 106 8 B 4 40 4 Zotepine44 500 (100–600) 241 (150 or 300) 240 (50–450) (–450) B 2 112 4 Zotepine12 (300) B 2 23 5 300 B 4 42 12 Chlorpromazine 28 600 (500–900) A 5 268 6 Chlorpromazine + benzatropine 6 mg/day Clozapine29 417 (150–900) 279 (125–525) B 4 151 4–8 Chlorpromazine B 3 15 4 Chlorpromazine Clozapine Clozapine30 Clozapine31 Olanzapine 41 Zotepine 42 Zotepine Zotepine43 Treatment resistant schizophrenia (DSM-III) Global judgment CGI 20% BPRS reduction NI Global judgment (own scale) Not used (only means) Not used (only mean BPRS) Not used (only mean CGI) CGI (no data) Improved according to BPRS CGI (no data) 20% BPRS reduction CGI (only means) Improved according to NOSIE NI 40% PANSS reduction 40% PANSS reduction 40% PANSS reduction 20% BPRS reduction + CGI<3 (or BPRS<35) 50% BPRS reduction Schizophrenia with acute exacerbation (DSM-III-R) Chlorpromazine 555 Schizophrenia with 50% BPRS (300–1200) acute exacerbation reduction (DSM-III) Thioridazine 378 (150–600) Schizophrenia or 50% BPRS schizophreniform reduction disorder (DSM-III-R) Thioridazine 440 (150–750) Acute schizophrenia 50% BPRS or schizophreniform reduction disorder (DSM-III) Thioridazine 133 Elderly patients with 50% BPRS schizophrenia or reduction paranoid disorder (ICD-9) Methotrimeprazine 100 (–150) Schizophrenia with acute 20% PANSS exacerbation (DSM-III-R) reduction Perazine 250 (150–900) Schizophrenia with CGI (only p value) positive symptoms (ICD-9) Chlorpromazine 532 Acute schizophrenia 20% BPRS (300 or 600) (DSM-III-R) reduction Perazine 350 (75–675) Acute schizophrenia Not used (only (ICD-9) mean CGI) Chlorpromazine (–450) Schizophrenia (NDCS) Moderately improved (subjective) Thioridazine (300) Schizophrenia (NDCS) Improvement scale (no data available) GIR=Global improvement rating. n=sample size. NI=not indicated. NDCS=no diagnostic criteria specified. CGI=clinical global impression. BPRS=brief psychiatric rating scale. NOSIE=nurses, observational scale of inpatient evaluation. PANSS=positive and negative symptoms scale. DSM-II to IV=various editions of the diagnostic and statistical manual of diseases. ICD-9=international classification of diseases 9th edition. *Mean and/or range or upper limit of range are given dependent on data available. Table 1: Characteristics of included studies 1582 THE LANCET • Vol 361 • May 10, 2003 • www.thelancet.com For personal use. Only reproduce with permission from The Lancet Publishing Group. ARTICLES At least one EPS N Amisulpride 1 n 30 Clozapine 11 758 Olanzapine 4 194 Quetiapine 1 201 Remoxipride 3 223 Risperidone 1 42 Zotepine 4 299 No clinically significant response Antiparkinsonian medication Drop-outs because of adverse events N RD (95% CI) p N n RD (95% CI) p N n 0 (–0·29 to 0·29) –0·15 (–0·26 to –0·04) –0·15 (–0·31 to 0·01) 0·03 (–0·07 to 0·13) 0·02 (–0·07 to 0·10) –0·10 (–0·30 to 0·11) 0·01 (–0·07 to 0·09) 1·0 1 30 0·07 (–0·24 to 0·37) –0·15 (–0·27 to –0·03) –0·22 (–0·42 to –0·02) –0·13 (–0·27 to 0·00) 0·05 (–0·13 to 0·24) –0·29 (–0·56 to –0·01) –0·17 (–0·39 to 0·06) 0·7 1 30 0·13 0·4 (–0·18 to 0·45) 125 –0·26 0·06 (–0·54 to 0·01) ·· ·· ·· ·· 0·008 7 685 0·07 4 194 0·6 1 201 0·7 4 264 0·4 1 42 0·8 2 218 0·02 3 0·03 NI 0·05 1 0·6 4 0·04 1 0·14 4 RD (95% CI) 201 –0·05 (–0·14 to 0·04) 264 0·06 (–0·14 to 0·27) 42 0·10 (–0·14 to 0·33) 299 0·02 (–0·05 to 0·09) p 1 11 4 0·3 1 0·5 4 0·4 1 0·5 3 n RD (95% CI) p 30 0·00 (–0·18 to 0·18) 760 –0·01 (–0·05 to 0·03) 194 –0·03 (–0·10 to 0·04) 201 –0·05 (–0·12 to 0·02) 264 –0·05 (–0·12 to 0·02) 42 –0·10 (–0·24 to 0·05) 187 –0·02 (–0·11 to 0·06) 1·0 0·7 0·4 0·14 0·13 0·2 0·6 N=number of trials included in the analysis. n=total number of patients included in the analysis. RD=risk difference for comparison with low-potency conventional antipsychotics. NI=not indicated. ··=no data. Table 2: New generation antipsychotics versus conventional low-potency antipsychotics Exclusion of remoxipride in a sensitivity analysis did not change the overall results. Investigations were included only if they met quality criterion A (adequate randomisation) or B (usually studies stated to be randomised without a precise explanation of the randomisation method) as described in the Cochrane Collaboration Handbook.9 We used the Jadad scale10 (range 0–5) to assess the quality of randomisation, doubleblinding, and description of drop-outs. We searched the Cochrane schizophrenia group’s register of randomised schizophrenia trials (up to March, 2002), using combinations of old and new generation drug names. The register is compiled by methodical searches of biological abstracts (1985–2002), CINAHL (1982–2002), Cochrane library (issue 1, 2002), dissertation abstracts (1861–2002), EMBASE (1980–2002), LILACS (1982–2002), MEDLINE (1966–2002), PSYNDEX (1977–1995), PsycINFO (1876–2002), RUSSMED (1988–2002), and Sociofile (1973–2002). Relevant journals and conference proceedings were searched by hand.4 Furthermore, we searched the reference lists of Cochrane reviews of new generation antipsychotics,5 other important reviews,2,11,12 and a review on remoxipride.13 We contacted manufacturers of new generation antipsychotics and the first authors of primary research articles to identify recently published reports, and to request further information on the trials we had identified. All data were extracted independently by two reviewers (SL, KW). Only dichotomous variables were analysed, because continuous data from rating scales on EPS in trials of new generation antipsychotics were often skewed. Furthermore, most trial samples were small, so judgment of normal distribution was difficult. The primary outcome of interest was the number of patients who had at least one EPS. As a measure of efficacy, we used the number of patients with no clinically significant improvement as defined by the research workers of the original studies. Additionally, we analysed the number of patients who received antiparkinsonian medication at least once, and the ones who dropped out of studies because of side-effects. We could not analyse the number of patients with specific EPS since only a few trials resulted in usable data for specific EPS, and it was often unclear whether investigators had failed to report on a specific EPS or whether the side-effect was absent. Furthermore, different names were used for specific EPS, so classification was sometimes difficult. Thus, these results are not shown, but only one of 38 tests gave a p value less than 0·05. THE LANCET • Vol 361 • May 10, 2003 • www.thelancet.com Statistical analysis We combined data from all eligible studies in a metaanalysis using risk differences, and used 95% CIs as a measure of effect size. The risk difference—ie, the risk of an unfavourable outcome in experimental treatmentallocated participants, minus the corresponding risk in controls—was used because we expected low frequencies of EPS in both groups. The risk difference gives more meaningful results than relative risks and odds ratios in such situations. To combine the results of the individual studies, we used a random effects model.14 Heterogeneity—ie, significant differences between the results of trials—was assessed by 2 test of heterogeneity. The Mantel-Haenszel 2 test was used for calculation of two-tailed significances of outcomes. The numberneeded-to-treat (NNT) was calculated for significant results. NNT indicates the number of patients who must be treated to prevent one bad outcome, and is the inverse of the risk difference. Studies with negative or non-significant results are less likely to be published than studies with significant results. The possibility of such a publication bias, which can affect the results of a meta-analysis, was examined using the funnel-plot method.15 In this method, the effect sizes of individual studies are plotted against their sample sizes and if all studies that have been done are published, a symmetrical figure resembling a funnel should result. To identify potential dose effects, studies were plotted according to dose of low-potency conventional drug. Results are presented as (weighted) risk differences along with their 95% CIs. Values less than 0 indicate effects favouring the new antipsychotics; 95% CIs that do not cross the y-axis are significant (p<0·05, two-tailed). Calculations were done with review manager 4.1, the meta-analytic software used by the Cochrane Collaboration. We did post-hoc meta-regression analyses using a random effects model (MetaWin version 2.0) to rule out study quality measured in Jadad scores, or association of duration of wash-out phase or publication year with the two main outcomes. Role of the funding source There was no funding source for this meta-analysis. Results A search of the Cochrane schizophrenia group yielded 109 citations of which 60 were excluded. Four were not relevant, 29 did not use a low-potency antipsychotic comparator, 14 were not randomised (case series or 1583 For personal use. Only reproduce with permission from The Lancet Publishing Group. ARTICLES Mean comparator LPA doses <600 mg/day chlorpromazine equivalent Study LPA n/N Blin40 2/21 4/21 0·10 (0·30 to 0·11) Phanjoo39 0/9 1/9 0·11 (0·37 to 0·15) 133 Singer21 7/20 6/20 0·05 (0·24 to 0·34) 196 HGDV (personal communication) Wetzel41 1/27 2/12 0·13 (0·35 to 0·09) 232 3/20 5/21 0·09 (0·33 to 0·15) 250 Chiu26 4/33 0/31 0·12 (0·00 to 0·24) 300 Dieterle44 11/20 9/20 0·10 (0·21 to 0·41) 350 Guiriguis25 1/22 2/28 0·03 (0·16 to 0·10) 375 Peuskens35 16/101 13/100 0·03 (0·07 to 0·13) 384 1/20 4/10 0·35 (0·67 to 0·03) 388 16/73 10/71 0·08 (0·05 to 0·20) 404 McCreadie38 1/30 1/31 0·00 (0·09 to 0·09) 440 Nishizono44 28/60 24/52 0·01 (0·18 to 0·19) 450 Fischer-C18 16/110 19/113 0·02 (0·12 to 0·07) 460 Loza33 13/27 5/14 0·12 (0·19 to 0·44) 465 5/53 4/53 0·02 (0·09 to 0·12) 532 Kostakoglu32 Keks37 Cooper44 2 =15·08 125/646 Heterogeneity 15 (p=0·45) Overall effect: p=0·7 Mean comparator LPA doses 600 mg/day Risk difference (95% CI) NGA n/N Risk difference (95% CI) Mean comparator dose (mg/day) 100 0·01 (0·03 to 0·04) 109/606 Howanitz27 4/24 4/18 0·06 (0·30 to 0·19) 600 Leon22 9/25 17/25 0·32 (0·58 to –0·06) 600 Gelenberg30 0/7 4/8 0·50 (0·87 to 0·13) 606 Xu23 4/30 16/28 0·44 (0·66 to 0·22) 693 Claghorn29 9/75 19/76 0·13 (0·25 to 0·01) 795 Hong20 2/21 7/19 0·27 (0·52 to 0·02) 1163 12/42 21/42 0·21 (0·42 to 0·01) 1200 0/13 5/12 0·42 (0·70 to 0·13) 1333 40/237 93/228 0·26 (0·37 to 0·16) Conley34 Shopsin19 Heterogeneity 72=12·30 (p=0·09) Overall effect: p<0·0001 Favours NGA Favours LPA Figure 1: Risk difference between groups for EPS NGA=new generation antipsychotic. LPA=low-potency conventional antipsychotic. n=number of patients with at least one EPS. N=number of patients in group. controlled clinical trials), nine were reviews with no usable data, three investigated healthy volunteers, and one study was still in the planning phase. Thus, 49 references for 27 studies were identified (only the principal reference of each study is indicated). We identified 3 studies from Cochrane reviews16,17 (HGDV Study Group, personal 1584 communication) and one from the meta-analysis by Butler and others.12 Table 1 lists the main characteristics of the 31 trials (2320 patients in total) that were included. Clozapine was most widely assessed (15 investigations), followed by zotepine (five), olanzapine and remoxipride (four each), THE LANCET • Vol 361 • May 10, 2003 • www.thelancet.com For personal use. Only reproduce with permission from The Lancet Publishing Group. ARTICLES Mean comparator LPA doses <600 mg/day chlorpromazine equivalent Study NGA n/N LPA n/N Blin40 4/21 10/21 Phanjoo39 3/9 6/9 HGDV (personal communication) 10/27 Chiu26 Risk difference (95% CI) Mean comparator dose (mg/day) 0·29 (0·56 to 0·01) 100 0·33 (0·77 to 0·10) 133 11/12 0·55 (0·79 to 0·31) 232 25/33 20/31 0·11 (0·11 to 0·34) 300 Peuskens35 35/101 48/100 0·13 (0·27 to 0·00) 384 Kostakoglu32 13/20 9/10 0·25 (0·53 to 0·03) 388 Keks37 58/73 57/71 0·01 (0·14 to 0·12) 404 McCreadie38 17/30 13/31 0·15 (0·10 to 0·40) 440 Nishizono44 45/60 42/52 0·06 (0·21 to 0·10) 450 Fischer-C18 33/110 46/113 0·11 (0·23 to 0·02) 460 Loza33 24/27 14/14 0·11 (0·26 to 0·04) 465 Cooper42 14/53 29/53 0·28 (0·46 to 0·10) 532 Chouinard36 14/20 9/21 0·27 (0·02 to 0·56) 555 2 Heterogeneity 12 =37·33 (p<0·001) Overall effect: p=0·02 Mean comparator LPA doses 600 mg/day Risk difference (95% CI) 295/584 * 314/538 0·11 (0·21 to 0·02) 0·06 (0·14 to 0·02)* 0·08 (0·35 to 0·19) 600 0·45 (0·88 to 0·01) 606 Leon22 8/25 Gelenberg30 3/7 7/8 Hong20 15/21 19/19 0·29 (0·49 to 0·08) 1163 Conley34 39/42 42/42 0·07 (0·16 to 0·02) Kane28 88/126 137/142 3/13 3/12 Shopsin19 Heterogeneity 52=14·74 (p=0·01) Overall effect: p=0·004 156/234 10/25 1200 0·27 (0·35 to 0·18) 1200 0·02 (0·35 to 0·32) 1333 0·18 (0·30 to 0·06) 218/248 Favours NGA Favours LPA Figure 2: Efficacy analysis: number of patients with no clinically significant response *Pooled RD after studies with comparator doses lower than 300 mg/day of chlorpromazine equivalent were excluded. NGA=new generation antipsychotic. LPA=low-potency conventional antipsychotic. n=number of patients with at least one EPS. N=number of patients in group. and quetiapine, risperidone, and amisulpride (one each). No randomised controlled trials comparing sertindole or ziprasidone with a low-potency antipsychotic were identified. Chlorpromazine was the comparator drug in most trials (n=22), but several others used thioridazine,12,25,38–40 methotrimeprazine,41 and perazine.16,42,44 THE LANCET • Vol 361 • May 10, 2003 • www.thelancet.com The mean daily dose of chlorpromazine equivalents ranged from 100 mg41 to 1330 mg;20 the median of the mean daily dose was 440 mg. Six investigations used an adequate randomisation method (Cochrane quality score A); all others were said to be randomised, but the exact method was not explained (Cochrane quality score B). 1585 For personal use. Only reproduce with permission from The Lancet Publishing Group. ARTICLES Risk difference (standard error) 0·00 0·04 0·08 0·12 0·16 0·20 1·00 0·5 0 0·5 1·5 Risk difference Figure 3: Funnel-plot of the outcome p for funnel-plot asymmetry=0·05.15 Only three were single-blind (patients masked) (HGDV Study Group, personal communication),32,33 all others were double-blind. All but five trials12,17,22,31,44 gave specific numbers and reasons for drop-outs. The mean Jadad score was 3·4 (median 4). The median of the mean wash-out phase was 4 days (range 0–14). With the exception of a 1-year study,32 all were short-term with a duration between 4 and 12 weeks (median 6 weeks). Most investigations were small with a median of 41 participants (range 15–268). Most patients had schizophrenia, but 18 had schizoaffective disorders, 19 had schizophreniform disorders, 13 had delusional disorders, and 69 had an unclear diagnosis. However, the reports were published from 1974 to 2000, during which time diagnostic criteria varied, and several studies used only clinical judgment for diagnosis (table 1). Table 2 shows the pooled results of the new generation antipsychotics (figures showing effect sizes of single studies can be obtained from the authors upon request). The overall results of trials that did not report dichotomous data are described here. In only one small study with 30 participants16 was amisulpride compared with perazine (a phenothiazine that is biochemically similar to chlorpromazine), and no substantial difference was found in any outcome analysed. 11 of 15 studies of clozapine provided data for the number of patients who had at least one EPS. Importantly, fewer patients treated with clozapine had EPS, than those treated with chlorpromazine (NNT 7, 95% CI 4–25, p=0·008). However, results of individual studies were heterogeneous, which might be explained by a dose effect. Ranking of studies according to mean daily dose of chlorpromazine equivalents suggested that the benefit of the drug was lost when low chlorpromazine doses were used. Of the four studies that did not include dichotomous data, two groups28,31 noted significantly fewer patients with EPS with clozapine EPS, using rating scales, whereas other researchers17,24 did not. More patients treated with clozapine achieved a clinically significant improvement (NNT 7, 95% CI 4–33, p=0·02) than did those on chlorpromazine. A dose effect was seen when reports were sorted according to dose ratio of clozapine to chlorpromazine. In studies with low ratios (ie, higher doses of chlorpromazine than clozapine) patients benefited more from clozapine than chlorpromazine. There were several trials of patients resistant to treatment, in which high doses of conventional drugs might have been used. In these groups, there might have been a better 1586 treatment response to clozapine, not because of comparative drug dosing, but because of resistance to the comparator. Of the eight studies17,21,23–25,27,29,31 for which dichotomous response data were not available, only two29,31 showed significant effects of clozapine with respect to overall efficacy at endpoint. In three small trials, fewer participants treated with clozapine used antiparkinsonian medication (p=0·06) than conventional drugs. The risk difference for drop-outs because of adverse events did not significantly differ between groups (table 2). In four trials of olanzapine (n=194), more patients on chlorpromazine than on olanzapine had at least one EPS, but this result was of borderline significance (p=0·07). It should be noted that in the study by Conley and others,34 all patients in the chlorpromazine group were also given benzatropine prophylactically. Despite this prophylaxis, more patients in the chlorpromazine group had EPS, so we can assume that the benefit of olanzapine would have been greater if benztropine had not been given. Furthermore, in a fixed-effects rather than random-effects model (which might also have been appropriate since there was no significant heterogeneity) the difference was significant (p=0·02). More patients in the olanzapine group improved compared with chlorpromazine (NNT 5, 95% CI 2–50, p=0·03). No study provided data on the use at least once of antiparkinsonian medication and no significant difference in terms of drop-outs due to adverse events was noted. In only one trial (n=201) was quetiapine compared with chlorpromazine.35 Quetiapine did not seem to be tolerated better than chlorpromazine. However, the low mean dose of chlorpromazine, 384 mg/day, should be taken into account. Furthermore, a greater number of participants taking quetiapine were rated as improved compared with chlorpromazine (p=0·05). Remoxipride was compared with chlorpromazine36 and thioridazine37-39 in four trials (n=264); there was no pronounced difference between drugs with respect to the outcome indices. In one study,36 dichotomous data were not shown for EPS, but there was no difference in antiparkinsonian medication used between groups. In only one small trial (n=42) was risperidone compared with methotrimeprazine.41 No significant difference in any tolerability outcome was seen. Patients given risperidone improved significantly compared with those given methotrimeprazine (NNT 3, 95% CI 2–100, p=0·04). However, methotrimeprazine was used at very low doses (mean 100 mg/day maximum 150 mg/day). Zotepine was compared with chlorpromazine12,42,44 and perazine41,43 in five trials (n=322). Zotepine did not prove better than chlorpromazine in any outcome parameter analysed. For one trial,12 no information could be obtained, with the exception of global efficacy, which was similar to thioridazine. Two groups41,43 did not report dichotomous response data, but recorded no significant differences in mean brief psychiatric rating scale scores at endpoint. In figures 1 and 2, the results of all trials are plotted according to the mean comparator dose of chlorpromazine equivalents. The data suggest that with mean chlorpromazine doses of less than 600 mg/day—a general cut-off for moderate doses8,11 derived from graphical presentation of the results—there were no differences in risk of EPS. Indeed, the pooled risk difference for the 16 studies (n=1252) in which doses lower than 600 mg of chlorpromazine per day were given was 0·01 (95% CI –0·03 to 0·04, p=0·7). However, in eight studies (n=465) with mean doses 600 mg or higher, the risk difference was –0·26 (95% CI –0·37 to –0·16; THE LANCET • Vol 361 • May 10, 2003 • www.thelancet.com For personal use. Only reproduce with permission from The Lancet Publishing Group. ARTICLES NNT 4, 95% CI 3–6, p<0·0001). Additionally, a funnelplot (figure 3) suggests that trials that showed no benefit of the new drugs compared with conventional drugs, with respect of EPS, might not have been identified (p=0·05).15 Funnel plots for other outcomes did not suggest publication bias. However, no clear overall dose-effect was found in an analysis of the group of patients without significant improvement. In a heterogeneous dataset in which some patients were given less than (RD –0·11, 95% CI –0·12 to –0·02; NNT 9, 95% CI 5–50), and some 600 mg or more (RD –0·18, 95% CI –0·30 to –0·06; NNT 6, 95% CI 3–17) of chlorpromazine equivalent per day, significantly more patients given new generation antipsychotics improved compared with those given conventional drugs (figure 2). Figure 2 also shows that the reduced efficacy of conventional drugs at a dose of less than 600 mg/day could partly result from three investigations with mean doses of less than 300 mg/day.18,39,40 Chlorpromazine doses less than 300 mg/day are often no more efficacious than placebo for patients with schizophrenia.46 Indeed, after exclusion of these three trials, the new drugs proved no more efficacious than conventional ones (n=1023, RR –0·06, 95% CI –0·14 to 0·02, p=0·15). With metaregression, no significant associations were noted between Jadad scores, washout phase, publication year, and the two main outcomes (all p>0·1). With the same method, dose of conventional drug was associated with EPS (coefficient=–0·0003, p=0·0007), but not with non-response (coefficient=0·0000, p=0·9). A similar result was obtained with an ANOVA model in which the studies with doses less than 600 mg/day or more were compared with those 600 mg/day (at least one EPS, p=0·0007; no response, p=0·39). Finally, we confirmed the results by a sensitivity analysis, in which comparators other than chlorpromazine were excluded. No significant differences between groups in terms of at least one EPS were recorded in ten studies (n=906, RD 0·01, 95% CI –0·04 to 0·06, p=0·8), but in eight studies in which 600 mg/day or more of the drug was used (n=465, RD –0·26, 95% CI –0·37 to –0·16, p<0·0001; NNT 4, 95% CI 3–6). A significant benefit of the new drugs in terms of response to treatment was seen in nine studies that used less than 600 mg/day of the drug (n=457, RD –0·13, 95% CI –0·24 to –0·01, p=0·03; NNT 8, 95% CI 4–100) and in six studies that used more than this amount (n=482, RD –0·18, 95% CI –0·30 to –0·06, p=0·004; NNT 6, 95% CI 3–17). Discussion The results of our review indicate that there is a reasonable amount of evidence to suggest that clozapine produces fewer EPS than conventional drugs. Less robust evidence points to a similar result for olanzapine. For amisulpride, quetiapine, and risperidone only one, inconclusive, trial was identified. Further evidence shows that zotepine and remoxipride do not lead to fewer EPS than chlorpromazine or similar drugs. The data indicate a dose effect because, for doses less than 600 mg/day chlorpromazine or its equivalent, patients on new generation antipsychotics were not at significantly lower risk of EPS than those on low-potency conventional drugs. Thus, at mean doses less than this threshold amount, conventional drugs might not be associated with more EPS than new generation antipsychotics. This result is important because a low risk of EPS is thought to be the main advantage of an atypical antipsychotic. Our finding that the new generation drugs had no lower risk of EPS (at mean doses less than 600 mg/day THE LANCET • Vol 361 • May 10, 2003 • www.thelancet.com chlorpromazine equivalent) than conventional drugs is clinically relevant only if low-potency antipsychotics are sufficiently effective in this dose range. In an important review, Baldessarini and others46 concluded that no additional benefit of efficacy is obtained by using doses higher than 500–600 mg/day chlorpromazine or its equivalent. Bollini and colleagues47 noted that doses higher than 375 mg/day of chlorpromazine equivalents did not result in increased efficacy. Other investigators suggested that the optimum dose of chlorpromazine equivalents is between 540 mg/day and 940 mg/day.45 The differing conclusions of these reviews of conventional drugs45,47 might be caused by methodological difficulties. The search strategies used were not clearly indicated, so that different trials might have been included, metaanalytic procedures were not applied, and both reviews included low-potency as well as high-potency conventional antipsychotics. Some specific EPS (eg, dystonia) might be considered worse than others (eg, tremor), but for the reasons indicated, it was not possible to adequately analyse specific EPS. Also, scale-derived continuous data could be more sensitive than global EPS rates (eg, quetiapine induced less akathisia according to a scale).35 Interpretation of the NNT was restricted to the trial populations, and might not be indicative of NNTs outside these patients. Many patients in the trials were chronically ill, and such patients are known to be less sensitive with respect to EPS. In a long-term investigation, patients who had not received antipsychotics previously, and had only one clinical episode, had fewer EPS with clozapine (mean dose 292 mg/day), compared with chlorpromazine (mean 319 mg/day).31 This trial lasted for 1 year, by contrast with the reports included in our review. Unfortunately, we were not able to extract the mean doses of the acute phase or any other usable data for meta-analysis. Compared with high-potency drugs, low-potency conventional antipsychotics are associated with more autonomic side-effects. Such side-effects are less consistently monitored in drug trials than EPS, and they would therefore have been difficult to summarise in a meta-analysis. Several new generation drugs—especially clozapine—are sedatives, and need slow titration to avoid hypotension. Furthermore, autonomic side-effects are dose related, and thus not problematic up to a threshold dose of chlorpromazine. Finally, there were no significant differences between groups with respect to dropouts arising from adverse events. However, this is a broad measure of tolerability. Although we reviewed a large number of patients, individual studies were usually old and small. Investigations with few patients have results that are prone to fluctuation. We think that small study size often led to heterogeneity that could not be accounted for solely by design differences between individual trials, and especially not by the mean doses used. With respect to efficacy, some of this heterogeneity might result from different response criteria used by the investigators. We avoided using mean values derived from scales, which might have been more uniform, because in small samples such results are frequently skewed and cannot be used for metaanalysis. Furthermore, although our mean Jadad quality score (3·4) was higher than that noted by Thornley and Adams49 in a large sample of schizophrenia trials (2·5), we share their concern about diagnostic criteria that change with time, short duration of wash-out phases, and absent descriptions of allocation and masking methods.48 The data do not indicate a clear dose effect in terms of 1587 For personal use. Only reproduce with permission from The Lancet Publishing Group. ARTICLES effectiveness, and show that the new drugs are better than conventional ones for all doses, whether less or more than 600 mg/day chlorpromazine. Thus, we were not able to replicate the correlation between comparator dose and advantage of the new generation drugs suggested by Geddes and colleagues,9 who focused on the high-potency antipsychotic, haloperidol. Similar benefits of new generation antipsychotics compared with high-potency conventional antipsychotics have been described at best as moderate.1,11,49 They can still be meaningful, however, since schizophrenia is usually a chronic disorder. An NNT of 9 means that for every nine patients who are given a new antipsychotic instead of chlorpromazine, one has a better outcome; or if 1000 patients are treated there will be 110 more responders than with chlorpromazine. Exclusion of studies that used potentially ineffective comparator doses of less than 300 mg/day chlorpromazine equivalents45 showed no benefit of the new generation antipsychotics compared with conventional drugs. However, this analysis was post hoc. Because every new generation antipsychotic has a specific receptor-binding profile, the pooled analysis of all new generation antipsychotics should be considered preliminary and be followed up by detailed differentiated analyses. With one exception,34 only clozapine has been compared with chlorpromazine doses greater than 600 mg/day. Some of the other new drugs might have failed to show lower risk of EPS because low doses of chlorpromazine were used in those trials; without trial data, the cause cannot be correctly attributed to either a drug or a dose effect. For all these reasons, we do not consider our data to be a firm basis for treatment recommendations, but instead the starting point for generating a hypothesis. In consideration of the optimum dose of chlorpromazine and similar drugs, well designed trials to find appropriate doses of low-potency antipsychotics are still justified, because chlorpromazine will continue to be a preferred antipsychotic worldwide.6 These studies might show that some conventional low-potency drugs, used in appropriate doses, have properties similar to those of atypical drugs. This finding would be very important for patients with schizophrenia in settings where new generation drugs are not generally affordable. Additionally, further trials on the newer atypical antipsychotics including aripiprazole, sertindole, and ziprasidone are needed to assess whether they are really associated with lower risk of EPS than low-potency conventional compounds. data, John Kane for his comments on the manuscript, and Mark Fenton for his assistance in the literature search. S Leucht is a contributing editor for the Cochrane Schizophrenia Group. This meta-analysis received no funding. References 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 Contributors S Leucht contributed to protocol design, quality assessment of single studies, data extraction, statistical analysis, and writing of the report. K Wahlbeck helped with protocol design, quality assessment of single studies, data extraction, statistical analysis, and revision of the report. J Hamann contributed to quality assessment of single studies, verification of data extraction, and revision of the report. W Kissling contributed to protocol design, interpretation of data, and revision of the report. 20 21 22 Conflict of interest statement S Leucht and W Kissling received lecture honoraria and/or travel grants to attend scientific meetings from Janssen-Cilag, Eli Lilly, Lundbeck, Pfizer, Sanofi-Synthélabo, and Zeneca. They also received financial support for a randomised trial from Eli Lilly, and for a meta-analysis on amisulpride from Sanofi-Synthélabo. W Kissling received lecture honoraria from Novartis and BMS. K Wahlbeck received lecture honoraria and travel grants from Eli Lilly, Janssen-Cilag, Novartis, BMS, and Pharmacia. J Hamann received financial support to attend a scientific meeting from Janssen-Cilag. Acknowledgments We thank Martin Dossenbach of Eli Lilly, Richard Owen of AstraZeneca, Olivier Blin and Masahisa Nishizono for providing us with unpublished 1588 23 24 25 26 Wahlbeck K, Cheine M, Essali A. Clozapine for schizophrenia (Cochrane review). Cochrane Database Syst Rev 2003; 3: CDD51334. Leucht S, Pitschel-Walz G, Engel R, Kissling W. Amisulpride–an unusual atypical antipsychotic: a meta-analysis of randomized controlled trials. Am J Psychiatry 2002; 159: 180–90. Leucht S, Barnes T, Kissling W, Engel R, Correll C, Kane J. Relapse prevention for schizophrenia with new generation antipsychotics: a systematic review and explorative meta-analysis of randomized controlled trials. Am J Psychiatry 2003; in press. Adams CE, Coutinho E, Duggan L, Gilbody S, Leucht S, Wahlbeck K. Cochrane Schizophrenia Group. Cochrane Database Syst Rev 2003; 4: CDD51334. The Cochrane Collaboration. The Cochrane Library, issue 4. Oxford: Update Software; 2002. WHO. Essential drugs. WHO Drug Information 1999; 13: 249–62. Seeman P, Lee T. Antipsychotic drugs: direct correlation between clinical potency and presynaptic action on dopamine neurons. Science 1975; 188: 1217–19. American Psychiatric Association. Practice guideline for the treatment of patients with schizophrenia. Am J Psychiatry 1997; 154 (suppl): 1–63. Clarke M, Oxman AD. Cochrane collaboration handbook version 4.0 (updated July 2000). The Cochrane Library. Oxford: Update Software; 2000. Jadad A, Moore A, Carroll D, et al. Assessing the quality of reports of randomised clinical trials: Is blinding necessary? Control Clin Trials 1996; 17: 1–12. Geddes J, Freemantle N, Harrison P, Bebbington P. Atypical antipsychotics in the treatment of schizophrenia: systematic overview and meta-regression analysis. BMJ 2000; 321: 1371–76. Butler A, Wighton A, Welch C, Tweed J, Byrom BD, Reynolds C. The efficacy of zotepine in schizophrenia: a meta-analysis of BPRS and improvement scale scores. Int J Psych Clin Pract 2000; 4: 19–27. Wadworth AN, Heel RC. Remoxipride: a review of its pharmacodynamic and pharmacokinetic properties, and therapeutic potential in schizophrenia. Drugs 1990; 40: 863–79. Der-Simonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986; 7: 177–88. Egger M, Davey SG, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. BMJ 1997; 315: 629–34. Rüther E, Blanke J. Therapievergleich von Aminosultoprid (DAN 2163) und Perazin bei schizophrenen Patienten. In: Helmchen H, Hippius H, Tölle R, ed. Therapie mit Neuroleptika—Perazin. Stuttgart, New York: Georg Thieme Verlag, 1988: 65–70. Potter WZ, Ko GN, Zhang LD, Yan W. Clozapine in China: a review and preview of US/PRC collaboration. Psychopharmacology 1989; 99 (suppl): 87–91. Fischer-Cornelssen K, Ferner U, Steiner H. Multifocal psychopharmaceutic testing “Multihospital trial” (Multifokale Psychopharmakaprüfung). Arzneimittelforschung 1974; 24: 1706–24. Shopsin B, Klein H. Clozapine, chlorpromazine and placebo in newly hospitalized, acutely schizophrenic patients. Arch Gen Psychiatry 1979; 36: 657–64. Hong CJ, Chen JY, Chiu HJ, Sim CB. A double blind comparative study of clozapine versus chlorpromazine on Chinese patients with treatment refractory schizophrenia. Int Clin Psychopharmacol 1997; 12: 123–30. Singer K, Law SK. A double-blind comparison of clozapine (Leponex) and chlorpromazine in schizophrenia of acute symptomatology. J Int Med Res 1974; 2: 433–35. Leon CA, Estrada H. Efectos terapeuticos de la clozapina sobre los sintomas de psicosis. Revista Colombiana de Psiquiatria 1974; 3: 309–18. Xu WR, Bao-Long Z, Qui C, Mei-Fang G. A double-blind study of clozapine and chlorpromazine treatment in the schizophrenics. Chin J Nerv Ment Dis 1985; 11: 222–24. Liu BL, Chen YY, Yang DS. Effects of thioridazine on schizophrenics and clinical utility of plasma levels. Chin J Neurol Psychiatry 1994; 27: 364–67. Guirguis E, Voineskos G, Gray J, Schlieman E. Clozapine (Leponex) vs chlorpromazine (Largatil) in acute schizophrenia (a double-blind controlled study). Curr Ther Res Clin Exp 1977; 21: 707–19. Chiu E, Burrows G, Stevenson J. Double-blind comparison of clozapine with chlorpromazine in acute schizophrenic illness. Aust N Z J Psychiatry 1976; 10: 343–47. THE LANCET • Vol 361 • May 10, 2003 • www.thelancet.com For personal use. Only reproduce with permission from The Lancet Publishing Group. ARTICLES 27 Howanitz E, Pardo M, Smelson DA, Engelhart C, Eisenstein N, Losonczy MF. The efficacy and safety of clozapine versus chlorpromazine in geriatric schizophrenia. J Clin Psychiatry 1999; 60: 41–44. 28 Kane JM, Honigfeld G, Singer J, Meltzer H, and the Clozaril Collaborative study group. Clozapine for the treatment-resistant schizophrenic: a double-blind comparison with chlorpromazine. Arch Gen Psychiatry 1988; 45: 789–96. 29 Claghorn J, Honigfeld G, Abuzzahab FS. The risks and benefits of clozapine versus chlorpromazine. J Clin Psychopharmacol 1987; 7: 377–84. 30 Gelenberg AJ, Doller JC. Clozapine versus chlorpromazine for the treatment of schizophrenia: preliminary results from a double-blind study. J Clin Psychiatry 1979; 40: 238–40. 31 Lieberman JA, Philipp M, Kong L, Gu H, Koch G. Efficacy and safety of clozapine versus chlorpromazine in first episode psychosis: results of a 52-week randomized double-blind trial. Schizophr Res 2001; 49 (suppl): 236. 32 Kostakoglu E, Alptekin K, Binnur Kivircik B, Dossenbach M, Tunca Z, Gogus A. Olanzapine versus chlorpromazine—6 weeks treatment of acute schizophrenia. Eur Neuropsychopharmacol 2001; 11 (suppl 3): S369. 33 Loza N, El-Dosoky AM, Okasha TA, et al. Olanzapine compared to chlorpromazine in acute schizophrenia. Eur Neuropsychopharmacol 1999; 9 (suppl): S291. 34 Conley RR, Tamminga CA, Bartko JJ, et al. Olanzapine compared with chlorpromazine in treatment-resistant schizophrenia. Am J Psychiatry 1998; 155: 914–20. 35 Peuskens J, Link CGG. A comparison of quetiapine and chlorpromazine in the treatment of schizophrenia. Acta Psychiatr Scand 1997; 96: 265–73. 36 Chouinard G. A placebo-controlled clinical trial of remoxipride and chlorpromazine in newly admitted schizophrenic patients with acute exacerbation. Acta Psychiatr Scand Suppl 1990; 358: 111–19. 37 Keks N, McGrath J, Lambert T, et al. The Australian multicentre double-blind comparative study of remoxipride and thioridazine in schizophrenia. Acta Psychiatr Scand 1994; 90: 358–65. 38 McCreadie RG, Todd N, Livingston M, et al. A double blind comparative study of remoxipride and thioridazine in the acute phase of schizophrenia. Acta Psychiatr Scand 1988; 78: 49–56. 39 Phanjoo AL, Link C. Remoxipride versus thioridazine in elderly psychotic patients. Acta Psychiatr Scand Suppl 1990; 358: 181–85. 40 Blin O, Azorin JM, Bouhours P. Antipsychotic properties of risperidone, haloperidol, and methotrimeprazine in schizophrenic patients. J Clin Psychopharmacol 1996; 16: 38–44. 41 Wetzel H, von Bardeleben U, Holsboer F, Benkert O. Zotepine versus perazine in patients with paranoid schizophrenia: a double-blind controlled trial of its effectiveness. Fortschr Neurol Psychiatr 1991; 59 (suppl 1): 23–29. 42 Cooper SJ, Tweed J, Raniwalla J, Butler A, Welch C. A placebo controlled comparison of zotepine versus chlorpromazine in patients with acute exacerbation of schizophrenia. Acta Psychiatr Scand 2000; 101: 218–25. 43 Dieterle DM, Müller-Spahn F, Ackenheil M. Wirksamkeit und Verträglichkeit von Zotepin im Doppelblindvergleich mit Perazin bei schizophrenen Patienten. Fortschr Neurol Psychiatr 1991; 59 (suppl 1): 18–22. 44 Nishizono M. A comparative trial zotepine, chlorpromazine and haloperidol in schizophrenic patients. Neuropsychopharmacol 1994; 10 (suppl): S30. 45 Davis JM, Barter JT, Kane JM. Antipsychotic drugs. In: Kaplan HJ, Saddock BJ, eds. Comprehensive textbook of psychiatry. 5th edn. Baltimore: Williams and Wilkins, 1989: 1591–626. 46 Baldessarini RJ, Cohen BM, Teicher MH. Significance of neuroleptic dose and plasma level in the pharmacological treatment of psychoses. Arch Gen Psychiatry 1988; 45: 79–91. 47 Bollini P. Antipsychotic drugs: is more worse? A meta analysis of the published randomised control trials. Psychol Med 1994; 24: 307–16. 48 Thornley B, Adams C. Content and quality of 2000 controlled trials in schizophrenia over 50 years. BMJ 1998; 317: 1181–84. 49 Kennedy E, Song F, Hunter R, Clarke A, Gilbody S. Risperidone versus typical antipsychotic medication for schizophrenia (Cochrane Review). Cochrane Database Syst Rev 2003; 4: CD51334. Uses of error Diagnosis of viral disease Pedro F C Vasconcelos When the 1997 dengue epidemic occurred in Belém, Brazil, I saw several patients with fever and other common symptoms of dengue including skin rash. I recall one patient, who was the daughter of a colleague with the same clinical symptoms as the others. After the physical examination, I concluded that she also had dengue fever. I only requested blood tests to confirm the case as dengue, at the request of the relatives. To my surprise, the serology, including a convalescent sample, and the attempts at viral isolation were negative. Due to my diagnosis, and pressure from her family, she had been discharged without further tests. Subsequently, I was obliged to request them and they showed an exuberant serological conversion for cytomegalovirus. During a rubella epidemic that happened in a small city near Belém several years ago, a doctor telephoned to our laboratory to tell us the occurrence of several cases in the city. The picture was typical with fever, skin rash, arthralgia, and lymphadenopathy. The colleague was worried because several patients had also developed jaundice and hemorrhages, and had died. The colleague was puzzled because he didn’t know that fatal cases of rubella occurred or were accompanied by jaundice and haemorrhages. When the samples were examined in our laboratory, it was observed that the cases diagnosed as rubella were in fact caused by Mayaro virus, an Alphavirus related to chikungunya and Semliki Forest viruses, and that the cases presenting with jaundice and haemorrhage were in fact yellow fever. This was one of the first simultaneous epidemics of these two viruses in the Brazilian Amazon region, both transmitted by mosquitoes in the forest. Since then, I adopted a more conservative position for the clinical diagnosis of diseases presenting with fever and skin rash. The most prudent course is to request examinations for the viruses most prevalent in the area and wait for the laboratory results. Some results can be unexpected. Department of Arbovirus, Instituto Evandro Chagas, FUNASA, Ministry of Health, 66090-000, Belém, PA, Brazil (P F C Vasconcelos PhD) THE LANCET • Vol 361 • May 10, 2003 • www.thelancet.com 1589 For personal use. Only reproduce with permission from The Lancet Publishing Group.