Download New generation antipsychotics versus low

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.