Cognitive behavior therapy for schizophrenia:

A meta-analytical review of randomized

controlled trials
FREDDY SARIN, LENNART WALLIN, BIRGITTA WIDERLÖV
Nord J Psychiatry Downloaded from informahealthcare.com by Biblioteka Uniwersytetu Warszawskiego on 03/13/15

Sarin F, Wallin L, Widerlöv B. Cognitive behavior therapy for schizophrenia: A meta-analytical

review of randomized controlled trials. Nord J Psychiatry 2011;65:162–174.

Background: In the UK and in Sweden, cognitive behavior therapy (CBT) has been recom-
mended for schizophrenia. The two recent meta-analyses examined results soon after treatment
and not at follow-up. Aim: To determine the effectiveness of CBT in people with schizophrenia,
both after treatment and at follow-up, and to compare it with treatment as usual (TAU) and
other psychological treatments. Methods: The search was carried in the databases CENTRAL
(Cochrane Central Register of Controlled Trials), PsycINFO and PubMed (Medline). Inclusion
criteria were randomized controlled trials (RCTs) with low risk of bias. Two reviewers, working
independently, extracted data. The results were analyzed using risk ratio (RR), risk difference
(RD), mean difference (MD), or standardized mean difference (SMD). Outcome measures were
symptoms, use of medication, relapse and clinically important improvement. Results: When
For personal use only.

CBT was compared with other psychological treatments at follow-up, there was strong evidence
(with small treatment effect) that intervention has an effect with positive symptoms (P ⫽ 0.02),
negative symptoms (P ⫽ 0.03) and general symptoms (P ⫽ 0.003). After treatment, there was
a trend in favor of CBT, but not statistically significantly so. Conclusion: It appears that the
effect of CBT is delayed; it could be seen a few months after the treatment had terminated.
Therapies for patients with schizophrenia that were 20 sessions long or more had better out-
comes than those that were shorter.
• Cognitive therapy, Meta-analysis, Randomized controlled trial, Schizophrenia.

Freddy Sarin, M.Sc., Järvapsykiatrin, Psychiatry, Praktikertjänst AB, Rinkebysvängen 70 A,

SE-163 74 Spånga, Sweden, E-mail: freddy.sarin@ptj.se; Accepted 29 March 2011.

C ognitive behavior therapy (CBT) is a form of

psychotherapy that is based on the premise that
there is a relationship between thoughts, feelings and
over the years; these include reduction of symptoms
(positive, negative and general symptoms), reduction of
relapse, better social functioning and more insight. Fur-
behavior (1). CBT has been shown to be effective in the thermore, the range of patient populations targeted has
treatment of depression (2) and anxiety (3) and it has increased, with recent developments in CBT for schizo-
also been used to treat a vast range of other clinical dis- phrenia being the treatment of first-episode psychosis and
orders, with varied effectiveness. In the early 1990s, fol- treatment of schizophrenia together with co-morbidity
lowing an improvement in our understanding of the disorders such as abuse of substances (1). Reviews of
cognitive psychology of psychotic symptoms, interest studies on CBT have indicated that they are useful for
grew in the use of CBT for people with schizophrenia. the treatment of schizophrenia. Both the Schizophrenia
The intention is to help the individual normalize and Patient Outcome Report in the USA and the National
make sense of his/her psychotic experiences, and to Board of Health and Welfare in Sweden recommend
reduce the associated distress and negative effects on CBT for patients with schizophrenia (4, 5). The National
functioning. It also deals with learning to view thoughts Institute for Clinical Excellence (NICE) in the UK has
and interpretations as just hypotheses, not facts, and recommendations such as: “Offer cognitive behavioral
teaches the patient to search for evidence for and against therapy (CBT) to all people with schizophrenia. This can
their beliefs when these beliefs cause distress. A range be started either during the acute phase or later, includ-
of outcomes of CBT in schizophrenia has been described ing an inpatient setting” (p. 218) (1); “CBT should be

© 2011 Informa Healthcare DOI: 10.3109/08039488.2011.577188

 COGNITIVE BEHAVIOR THERAPY FOR SCHIZOPHRENIA

delivered on a one-to-one basis over at least 16 planned PRODROME or ONSET “DISORDERS”). CENTRAL:
sessions” (p. 218) (1); “Offer CBT to assist in promot- (SCHIZO* or PSYCHOTIC* or PSYCHOS*) AND
ing recovery in people with persistent positive and nega- (BEHAVIOR THERAPY or COGNITIVE THERAPY)
tive symptoms and for people in remission” (p. 219) (1). AND (EARLY DIAGNOSIS or DUAL DIAGNOSIS or
However, the authors of the two recent meta-analyses PRODROME).
(4, 6) concluded that the effect size is greatly influenced People with schizophrenia-like illness (schizophrenia,
by the methodological rigor, the trials with low risk of schizoaffective syndrome or delusion disorder) were
bias tending to produce the lower effect sizes. One meta- included. The diagnosis criteria that apply to this meta-
analysis—by Lynch et al. (6)—made a comparison with analysis are ICD-10 (The ICD-10 Classification of Mental
other psychological treatments, and the second—by Wykes and Behavior Disorders) (13) and DSM-III R or IV (Diag-
et al. (4)—considered control groups, i.e. they made no nostic and Statistical Manual of Mental Disorders) (14).
differentiation between treatment as usual (TAU) and other Irrespective of whether they had CBT, the individuals
Nord J Psychiatry Downloaded from informahealthcare.com by Biblioteka Uniwersytetu Warszawskiego on 03/13/15

forms of psychological treatment. Both of these meta- who participated in the studies always required standard
analyses investigated the sizes of effects after treatment and care, since this patient group must receive it because of
not at follow-up. One relatively large randomized controlled their illness (9).
study examined the effectiveness of CBT in schizophrenia This review included only RCTs with a low risk of
(7). It was found that CBT was no better than other psy- bias. Quasi-experimental studies were excluded. Only
chological treatments directly after the treatment, but published studies were included, and the participants had
showed significant advantages at follow-up. However, in to be over 18 years old. Small studies (with ⱕ10 partici-
another study such delayed effects were not observed (8). pants) were excluded from this meta-analysis.
Thus, the purpose of the present meta-analysis was Screening and selection of all the search results was
to investigate the effect of CBT treatment (reduction of conducted in three phases by two of the authors working
symptoms, reduction of relapse, clinically important independently (FS and LW). In phase 1, titles and
improvement and use of medication) both after treatment abstracts were read. Abstracts that were considered rele-
and at follow-up. CBT was then compared with two vant and those that were deemed questionable went on
For personal use only.

different control groups: TAU and other psychological to phase 2. In this phase, the authors read the full study
treatment. In addition, the studies included were restricted and performed even more close reading to decide
to randomized controlled trials (RCTs) with rater blind- whether the study should be included or excluded. In
ing for group allocation, which was also the case for the phase 3, the scientific method of the individual studies
other two meta-analyses. was judged. The assessment was based on a form from
the Cochrane Collaboration; it was a domain-based eval-
uation and was answered “yes” (low risk of bias), “no”
Methods (high risk of bias) or “?” (unclear whether there was any
Trials included bias). It helped in assessing the validity and minimized
The search was done in the databases CENTRAL the risk of systematic errors. Studies that did not have a
(Cochrane Central Register of Controlled Trials), low risk of bias were excluded from this meta-analysis.
PsycINFO and PubMed (Medline). The electronic The areas that were evaluated were adequacy of sequence
searches were carried out with the help of a librarian. generation, adequacy of allocation concealment, single
The key words were chosen on the basis of existing blinding, incomplete description of outcome data, free-
index words, apart from CENTRAL, which was a free dom from selective reporting, freedom from other bias,
text search. The database search was complemented with adequate description of treatment, use of a treatment pro-
manual searching in the reference list of the studies and tocol or manual, and whether there was assessment of
meta-analyses included (4, 6, 9–11). Only published stud- treatment quality (12, 15, 16).
ies were included (12). For the search, we chose an
inception date of 10 years before the publication of
Wykes et al. (4), which would have captured earlier stud-
Types of interventions
ies. The search was conducted until October 18, 2010.
EXPERIMENTAL GROUP
The key words were as follows. PubMed: (SCHIZO- COGNITIVE BEHAVIOR THERAPY (CBT): CBT for schizophrenia is
PHRENIA and DISORDERS WITH PSYCHOTIC considered well defined if the therapist helps patients to estab-
FEATURES and COGNITIVE THERAPY) AND lish connections between thoughts, feelings and behavior
(DIAGNOSIS, DUAL “PSYCHIATRY” or EARLY concerning their symptoms. The purpose of therapy must be
DIAGNOSIS). PsycINFO: (PSYCHOSIS or SCHIZOAF- to correct misunderstandings and dysfunctional beliefs of the
FECTIVE DISORDER) AND (COGNITIVE TECH- patient and to work on alternative ways of dealing with the
NIQUES or BEHAVIOR THERAPY or COGNITIVE symptoms. The goal is to reduce psychotic symptoms and
BEHAVIOR THERAPY) AND (DUAL DIAGNOSIS or modify negative emotions. Studies that do not meet these

NORD J PSYCHIATRY·VOL 65·NO 3·2011 163

 F SARIN ET AL.

criteria in CBT, or do not give clear information about it, were CLINICALLY IMPORTANT IMPROVEMENT
categorized as not being sufficiently defined and were excluded If efforts to treat the patient achieve a reduction in symp-
from this meta-analysis (9). There should also have been toms, then in practice it may be a slight improvement. If,
a description of how the therapy was completed, to avoid however, reduction of symptoms is 25% or greater than
variation and differences in quality (16). at baseline, this is interpreted as a clinically important
improvement (9, 24).
CONTROL GROUP
TREATMENT AS USUAL (TAU): TAU is the treatment the par- USE OF MEDICATION
ticipants would normally have received from mental health Intervention can lead to a reduction in the use of medica-
services if they had not been included in the experimental tion (anti-psychotics), and this is therefore of value as a
group. Although TAU was not standardized, it usually con- measure of the success of treatment. It is difficult to
compare different types of medications; therefore, the
Nord J Psychiatry Downloaded from informahealthcare.com by Biblioteka Uniwersytetu Warszawskiego on 03/13/15

sisted of a combination of case management and anti-psy-

chotic medication (17). studies must be adjusted in terms of equivalents of the
anti-psychotic chlorpromazine in milligrams per day
(mg/day) (25, 26).
Other psychological treatments
OTHER PSYCHOLOGICAL TREATMENTS, for example psychoedu- RELAPSE
cation and family intervention, are given in addition to People with schizophrenia may be falling ill in a psy-
standard treatment. Psychoeducation programs consist of chotic episode and a goal of treatment is to prevent
a series of efforts to increase knowledge about the disease, relapse. Relapse was defined as when the patient ended
medication, coping with their daily life, etc. The idea is up in hospital and became an inpatient (9) or when he or
that when patients learn more about their disease, relapse she needed a greater degree of care over a long period,
is reduced and adherence to treatment is improved. Fam- such as more visits to the outpatient clinic or home visits,
ily intervention involves transfer of knowledge, problem- meeting the doctor or key person for an extended period
For personal use only.

solving and learning skills, and reduces the negative of time (27).
aspects of the family environment (hostility and criti-
cism). The knowledge is conveyed not only to the patients
Data collection and analysis
but to their whole family (10). More examples of other
Data extracted from the studies included were done by
psychological treatments are befriending (7) and sup-
two of the authors (FS and LW) independently. The
portive counseling (8), which means that the therapist
information about age, sex, study design, CBT, control
uses unspecific techniques such as empathy, warmth and
group, analysis and outcome was extracted from all of
engagement. They talk about ordinary things such as the
these studies. When a disagreement arose between the
patient’s hobbies and interests.
authors, it was resolved through discussion. For studies
that did not present raw data, mean and standard devia-
Types of outcome measures tion were excluded. In studies in which raw data were
The outcome measures were estimated on two occasions, presented with confidence intervals, the authors recal-
the first being directly after treatment and the second culated them as standard deviations. Where data had
being at the time of follow-up. Follow-up is from 3 to been transformed, e.g. with logarithmic scales, they
15 months after treatment (9). were excluded from the meta-analysis (9, 28). After the
data had been extracted, they were put in to the statis-
SYMPTOMS tical program Review Manger 5 (29, 30). This meta-
Symptoms were defined as discomfort that a person expe- analysis used intention-to-treat (ITT) analysis, which
rienced as result of the disease. One aim of the treatment means that every participant who begins the treatment
is to reduce various symptoms, which can be depression, is considered part of the trial, whether they finish it or
anxiety, or other negative symptoms (affective flattening, not. An ITT analysis is often recommended as the least
alogia, avolition and anhedonia), or even positive symp- biased way to estimate intervention effects in random-
toms (delusion and/or auditory hallucination). The scales ized trials (31).
in this review were as follows: Beck Depression Inven-
tory (BDI-II) (18), Beck Anxiety Inventory (BAI) (19), DICHOTOMOUS DATA AND CONTINUOUS DATA
Comprehensive Psychopathological Rating Scale (CPRS) This meta-analysis used summary (pooled) intervention
(20), Positive and Negative Syndrome Scale for Schizo- effect estimates calculated as weighted average of the
phrenia (PANSS) (21), Psychotic Symptoms Rating Scale intervention effects estimated in the individual studies.
(PSYRATS) (22) and Schedule for Assessment of Nega- Outcome measures for clinically important improvement
tive Symptoms (SANS) (23). and relapse were dichotomous data and were analyzed

164 NORD J PSYCHIATRY·VOL 65·NO 3·2011

 COGNITIVE BEHAVIOR THERAPY FOR SCHIZOPHRENIA

with relative risk ratio (RR) or risk difference (RD) and Results
random effects. The outcomes for symptoms and medica- Descriptions of studies
tion use were extracted as continuous data. They were During the search, which was performed both manually
analyzed with mean difference (MD) or standardized and electronically, 386 references were identified, of
mean difference (SMD) and fixed effects. The method for which 101 full texts were read. After the selection pro-
SMD was Hedges’ adjusted g. For both dichotomous and cess, 31 articles describing 22 studies remained. Fourteen
continuous data, 95% confidence intervals were used (32). studies were carried out in the UK, two in the USA, two
The statistical analysis was done with the help of a in Canada, two in Australia, one in Germany and one in
statistician. the Netherlands. There were no studies from Scandina-
MD and SMD are two different approaches. MD is via. A more detailed description of the sample is given
used when measures of outcome in all studies are in Table 1 (35–63).
made on the same scale and SMD is used when the In total, 2469 people participated in this meta-analysis
Nord J Psychiatry Downloaded from informahealthcare.com by Biblioteka Uniwersytetu Warszawskiego on 03/13/15

studies all assess the same outcome but measure it in and with post-treatment the drop-out was 335 persons
a variety of ways; for example, with all studies mea- (14%). With follow-up, there were 373 drop-outs from
suring negative symptoms but using different psycho- 2231 people (17%). It could be deduced that of 2457
metric scales. participants, 1731 (70%) were men and 726 (30%) were
The probability test (P-value) was done with the women. Mean age was 36 years (standard deviation,
Z-test for both dichotomous and continuous data. sd ⫽ 11.36, n ⫽ 1787) and average duration of illness
When the P-value was less than 0.05, this was inter- was 10.08 years (sd ⫽ 10.05, n ⫽ 1073). Of the 22
preted as strong evidence that the intervention had an studies, 13 examined participants who were chronically
effect. When the P-value was greater than 0.05, this ill with residual positive and negative symptoms, six
was interpreted as no strong evidence that the inter- studies examined participants after their first episode of
vention had an effect, meaning that the intervention schizophrenia and three studies examined participants
may have had an effect but that there was no strong who were acutely ill or in a sub-acute psychotic condi-
evidence. With dichotomous data, when the RR is tion. Eighteen studies provided individual CBT and four
For personal use only.

smaller than one (⬍1) it favors the experimental offered group-based CBT.
group, and when it is larger than one (⬎1) it favors
the control group. Effect size was used with the SMD Effects of interventions
and the estimate of the treatment effect was 0.2 for SYMPTOMS
small effects, 0.4 for moderate effects and 0.8 for large When CBT was compared with other psychological treat-
effects. Another variant would be ⬍0.41 for small ments at follow-up, there was strong evidence that inter-
effects, 0.40–0.70 for moderate effects and ⬎0.70 for vention had an effect on positive, negative and general
large effects (33). symptoms. After treatment, there was a trend in favor of
When it came to continuous data from summary CBT, but it was not statistically significant (Table 2).
information on the individual studies, a rough check was Generally speaking, for all results dealing with symp-
available for detecting skew, i.e. that the data were not toms, both after treatment and at follow-up, CBT was
normally distributed. This was done by taking the found to be of advantage.
observed mean, subtracting the smallest value of the When CBT was compared with TAU during follow-
scale and dividing this by 2 (32, 34). A ratio of less up, there was strong evidence that intervention had an
than 2 suggested skew (34). If the ratio was less than 1, effect with general symptoms as outcome, but this effect
there was strong evidence of a skewed distribution (32). was marginal. After treatment, there was a trend in
For continuous mean change data (changes score), it is favor of CBT, but this did not reach statistical signifi-
impossible to know if data are skewed because the cance (Table 3). Positive and negative symptoms as out-
absence of individual patient data (9). comes showed a trend in favor of CBT, but this was
not statistically significant. In the case of depression and
HETEROGENEITY anxiety, the result was not in favor of CBT, but this
An assessment of the variability (heterogeneity) between was of marginal significance and there were few studies
the studies was done. The statistical heterogeneity was in the analysis.
checked with chi-squared and I2. The P-value for chi-
squared may not be less than 0.10 and a rough guide CLINICALLY IMPORTANT IMPROVEMENT
to interpretation of I2 is as follows: 0–40% may not be Two studies reported clinically important improvement
important, 30–60% may represent moderate heterogene- after treatment with CBT relative to TAU. The results
ity, 50–90% may represent substantial heterogeneity showed that there was no strong evidence that CBT had
and 75–100% may represent considerable heterogeneity treatment effects (3 RCTs, n ⫽ 298, RD ⫽ ⫺0.11,
(9, 32). CI ⫽ ⫺0.29 to 0.07, P ⫽ 0.22). For follow-up, there

NORD J PSYCHIATRY·VOL 65·NO 3·2011 165

 Nord J Psychiatry Downloaded from informahealthcare.com by Biblioteka Uniwersytetu Warszawskiego on 03/13/15
For personal use only.

166
Table 1. Summary of the studies that were included.
Study n Design Experimental group Control group Outcome
F SARIN ET AL.

(Baker et al., 2006) (35) 130 RCT (blinded) Individual CBT (10 sessions) TAU BPRS, BDI-II
(Barrowclough et al., 2006) (36) 113 RCT (blinded) Group based CBT TAU PANSS, RELAPSE
(18 sessions)
(Bechdolf et al., 2004) (37, 38) 88 RCT (blinded) Group based CBT Psychoeducation PANSS, RELAPSE, MEDICATION USE
(16 sessions)
(Cather et al., 2005) (39) 30 RCT (blinded) Individual CBT (16 sessions) Psychoeducation PANSS, PSYRATS
(Durham et al., 2003) (40) 66 RCT (blinded) Individual CBT (20 sessions) 1) TAU, 2) Supportive PANSS, PSYRATS, MEDICATION USE, CLINICALLY
psychotherapy IMPORTANT IMPROVEMENT
(Fowler et al., 2009) (41) 77 RCT (blinded) Individual CBT (m ⫽ 12 TAU PANSS, BDI-II, BAI
session)
(Garety et al., 2008) (42) 301 RCT (blinded) Individual CBT (20 sessions) 1) TAU, 2) Family RELAPSE , PANSS, BAI, BDI-II
intervention
(Gumley et al., 2003) (43) 144 RCT (blinded) Individual CBT (md ⫽ 10 TAU MEDICATION USE, RELAPSE
sessions)
(Haddock et al., 1999) (44) 36 RCT (blinded) Individual CBT (m ⫽ 10,2 Supportive counseling BPRS
sessions)
(Haddock et al., 2009) (45) 77 RCT (blinded) Individual CBT (25 session) Social activity therapy PANSS, PSYRATS
(Jackson et al., 2008) (46) 62 RCT (blinded) Individual CBT (20 sessions) Befriending RELAPSE, MEDICATION USE, SANS
(Jackson et al., 2009) (47) 66 RCT (blinded) Individual CBT (26 session) TAU CLINICALLY IMPORTANT IMPROVEMENT
(Jolley et al., 2003) (48) 21 RCT (blinded) Individual CBT (18 sessions) TAU RELAPSE
(Lecomte et al., 2008) (49) 129 RCT (blinded) Group-based CBT Symptom management BPRS
(24 session)
(Lewis et al., 2002) (50–52) 315 RCT (blinded) Individual CBT (20 sessions) 1) TAU; 2) Supportive PANSS, PSYRATS, RELAPSE
counseling
(Penn et al., 2009) (53) 65 RCT (blinded) Group-based CBT Supportive therapy PANSS, PSYRATS, BDI-II
(12 session)
(Rector et al., 2003) (54) 50 RCT (blinded) Individual CBT (20 sessions) ETAU (Enriched PANSS, BDI-II, CLINICALLY IMPORTANT IMPROVEMENT
treatment as usual)
(Sensky et al., 2000) (7, 55) 90 RCT (blinded) Individual CBT (20 sessions) Befriending CPRS, MEDICATION USE, CLINICALLY IMPORTANT
IMPROVEMENT
(Tarrier et al., 1999) (8, 56–58) 87 RCT (blinded) Individual CBT (20 sessions) 1) TAU; 2) Supportive SANS, MEDICATION USE, RELAPSE
counseling
(Trower et al., 2004) (59) 38 RCT (blinded) Individual CBT (md ⫽ 10 TAU MEDICATION USE
sessions)
(Turkington et al., 2002) 422 RCT (blinded) Individual CBT (6 sessions) TAU CPRS (change score), PSYRATS (change score)
(60–62)
(Valmaggia et al., 2005) (63) 62 RCT (blinded) Individual CBT (16 sessions) Supportive counseling PANSS,
Total 2469

n, no. of participants; RCT, randomized controlled trial; CBT, cognitive behavioral therapy; m, mean; md, median; BAI, Beck Anxiety Inventory; BDI-II, Beck Depression Inventory; BPRS, Brief Psychiatric
Rating Scale; CPRS, Comprehensive Psychopathological Rating Scale; PANSS, Positive and Negative Symptoms of Schizophrenia; PSYRATS, Psychotic Symptom Rating Scales; SANS, Schedule for
the Assessment of Negative Symptoms in Schizophrenia.

NORD J PSYCHIATRY·VOL 65·NO 3·2011

 Nord J Psychiatry Downloaded from informahealthcare.com by Biblioteka Uniwersytetu Warszawskiego on 03/13/15
For personal use only.

Table 2. Cognitive behavioral therapy (CBT) compared with other psychological treatments (OPTs); outcomes are symptoms, and a negative sign for effect size means that CBT
was favored.
Symptom Control group Studies Statistical method Effect size P-value Heterogeneity (I2) Measurements

NORD J PSYCHIATRY·VOL 65·NO 3·2011

Positive symptom PANSS pos
Post-treatment OPT 7 RCTs; n ⫽ 504 MD ⫺0.55 (CI ⫺ 1.37 to 0.27) 0.19 0%
Follow-up OPT 7 RCTs; n ⫽ 483 MD ⫺1.01 (CI ⫺ 1.85 to ⫺0.18) 0.02‡ 0%
Negative symptom PANSS neg, SANS
Post-treatment OPT 8 RCTs; n ⫽ 461 SMD ⫺0.10 (CI ⫺ 0.28 to 0.09) 0.31 24%
Follow-up OPT 8 RCTs; n ⫽ 452 SMD ⫺0.20 (CI ⫺ 0.39 to ⫺0.02) 0.03‡ 23%
General symptom BPRS, CPRS, PANSS-gen
Post-treatment OPT 9 RCTs; n ⫽ 571 SMD ⫺0.07 (CI ⫺ 0.23 to 0.10) 0.43 33%
Follow-up OPT 8 RCTs; n ⫽ 537 SMD ⫺0.21 (CI ⫺ 0.38 to ⫺0.04) 0.02‡ 52%
PANSS total
Post-treatment OPT 5 RCTs; n ⫽ 357 MD ⫺2.09 (CI ⫺ 4.96 to 0.78) 0.15 0%
Follow-up OPT 5 RCTs; n ⫽ 359 MD ⫺4.80 (CI ⫺ 8.01 to ⫺1.59) 0.003‡ 0%
Hallucination PSYRATS
Post-treatment OPT 5 RCTs; n ⫽ 268 MD ⫺1.04 (CI ⫺ 3.41 to 1.33) 0.39 0%
Follow-up OPT 4 RCTs; n ⫽ 239 MD ⫺1.58 (CI ⫺ 4.23 to 1.07) 0.24 0%
Delusion PSYRATS
Post-treatment OPT 5 RCTs; n ⫽ 322 MD ⫺0.63 (CI ⫺ 2.12 to 0.86) 0.41 64%
Follow-up OPT 4 RCTs; n ⫽ 305 MD ⫺1.01 (CI ⫺ 2.54 to 0.53) 0.20 0%
Depression BDI-II
Post-treatment OPT 3 RCTs; n ⫽ 150 MD ⫺1.03 (CI ⫺ 4.31 to 2.24) 0.54 0%
Follow-up OPT 3 RCTs; n ⫽ 139 MD ⫺3.25 (CI ⫺ 6.77 to 0.27) 0.07 0%
Anxiety BAI
Post-treatment OPT 1 RCT; n ⫽ 41 MD ⫺0.79 (CI ⫺ 9.30 to 7.72) 0.86 Not applicable
Follow-up OPT 1 RCT; n ⫽ 40 MD ⫺0.59 (CI ⫺ 9.10 to 7.92) 0.89 Not applicable

RCT, randomized controlled trial; MD, mean difference; SMD, standardized mean difference (Hedges’ adjusted g); CI, 95% confidence interval. Other psychological treatments, psychoeducation;
befriending; supportive counseling; supportive psychotherapy; etc.; TAU, treatment as usual. SANS, Schedule for the Assessment of Negative Symptoms; BPRS, Brief Psychiatric Rating Scale; CPRS,
Comprehensive Psychopathological Rating Scale; PANSS, Positive and Negative Syndrome Scale (pos, positive scale; neg, negative scale; gen, general scale; total, total scale); PSYRATS, Psychotic
symptom rating scale; BDI-II, Beck Depression Inventory; BAI, Beck Anxiety Inventory.
‡The result showed strong evidence that CBT had a treatment effect.

167
COGNITIVE BEHAVIOR THERAPY FOR SCHIZOPHRENIA
 F SARIN ET AL.

Table 3. Cognitive behavioral therapy (CBT) compared with treatment as usual (TAU); outcomes are symptoms, and a negative sign
for effect size means that CBT was favored.
Control Statistical Heterogeneity
Symptom group Studies method Effect size P-value (I2) Measurements

Positive symptoms PANSS pos

Post-treatment TAU 3 RCTs; n ⫽ 417 MD ⫺0.62 (CI ⫺ 1.86 to 0.61) 0.32 0%
Follow-up TAU 3 RCTs; n ⫽ 415 MD ⫺0.67 (CI ⫺ 1.70 to 0.36) 0.20 88%
Negative symptoms PANSS neg, SANS
Post-treatment TAU 3 RCTs; n ⫽ 328 SMD ⫺0.10 (CI ⫺ 0.32 to 0.11) 0.35 0%
Follow-up TAU 3 RCTs; n ⫽ 318 SMD ⫺0.21 (CI ⫺ 0.43 to 0.01) 0.07 22%
General symptom BPRS, CPRS, PANSS-gen
Post-treatment TAU 4 RCTs; n ⫽ 476 SMD ⫺0.10 (CI ⫺ 0.29 to 0.08) 0.26 10%
Follow-up TAU 4 RCTs; n ⫽ 469 SMD ⫺0.14 (CI ⫺ 0.32 to 0.05) 0.14 12%
Nord J Psychiatry Downloaded from informahealthcare.com by Biblioteka Uniwersytetu Warszawskiego on 03/13/15

PANSS total
Post-treatment TAU 4 RCTs; n ⫽ 529 MD ⫺1.25 (CI ⫺ 3.75 to 1.25) 0.33 0%
Follow-up TAU 4 RCTs; n ⫽ 452 MD ⫺2.87 (CI ⫺ 5.67 to ⫺0.08) 0.04‡ 24%
Hallucination PSYRATS
Post-treatment TAU 2 RCTs; n ⫽ 130 MD ⫺1.94 (CI ⫺ 5.98 to 2.11) 0.35 0%
Follow-up TAU 2 RCTs; n ⫽ 116 MD ⫺0.91 (CI ⫺ 4.78 to 2.95) 0.64 0%
Delusion PSYRATS
Post-treatment TAU 2 RCTs; n ⫽ 171 MD 0.16 (CI ⫺ 1.95 to 2.27) 0.88 0%
Follow-up TAU 2 RCTs; n ⫽ 174 MD ⫺1.57 (CI ⫺ 3.56 to 0.42) 0.12 0%
Depression BDI-II
Post-treatment TAU 3 RCTs; n ⫽ 356 MD 0.92 (CI ⫺ 1.78 to 3.61) 0.51 69%
Follow-up TAU 2 RCTs; n ⫽ 274 MD 1.22 (CI ⫺ 1.48 to 3.92) 0.37 80%
Anxiety BAI
Post-treatment TAU 2 RCT; n ⫽ 226 MD 0.67 (CI ⫺ 2.52 to 3.87) 0.68 0%
1 RCT; n ⫽ 152 1.41 (CI ⫺ 2.92 to 5.74)
For personal use only.

Follow-up TAU MD 0.52 Not applicable

For explanation of abbreviations, see footnote to Table 2.

was one study and the result showed no strong evidence was not advantageous (2 RCTs, n ⫽ 73, MD ⫽ 147.76,
that CBT had an effect (1 RCT, n ⫽ 33, RD ⫽ ⫺0.17, CI ⫽ ⫺267.90 to 563.41, P ⫽ 0.49). CBT was 147.76
CI ⫽ ⫺0.46 to 0.13, P ⫽ 0.26). mg/day chlorpromazine equivalents more than with TAU.
Two studies reported clinically important improvement At follow-up, there were four studies and the result
after treatment with CBT relative to any other psycho- showed that CBT was not advantageous, but not as
logical treatment. The results showed that there was no clearly as immediately after treatment (4 RCTs, n ⫽ 258,
strong evidence that CBT had treatment effects (2 RCTs, MD ⫽ 4.36, CI ⫽ ⫺71.53 to 80.25, P ⫽ 0.91). The
n ⫽ 95, RD ⫽ ⫺0.11, CI ⫽ ⫺0.27 to 0.06, P ⫽ 0.21). numbers in the individual studies after treatment and at
No studies with follow-up were found. follow-up were suspected of being skewed.

MEDICATION USE RELAPSE

Concerning the outcome measure medication use, there There was one study with the outcome measure relapse,
were four studies that reported medication use after treat- which compared CBT compared with TAU. The result
ment with CBT and other psychological treatments. showed that CBT was not advantageous (1 RCT,
The result gave strong evidence that CBT had an effect n ⫽ 189, RR ⫽ 1.48, CI ⫽ 1.07 to 2.04, P ⫽ 0.02). At
(4 RCTs, n ⫽ 287, MD ⫽ ⫺174.92, CI ⫺ 250.51 follow-up, there were five studies and the results did not
to ⫺99.32, P ⬍ 0.00001). CBT was 174.92 mg/day chlo- show any strong evidence that CBT had treatment effects
rpromazine equivalents less than with other psychological in prevention of relapse (5 RCTs, n ⫽ 520, RR ⫽ 0.93,
treatments. It should be pointed out that the results in the CI ⫽ 0.71 to 1.22, P ⫽ 0.59).
individual studies were statistically heterogeneous One study had relapse as outcome measure and com-
(I2 ⫽ 83%). At follow-up, there were five studies and the pared the results of CBT and other psychological treat-
result showed no strong evidence that CBT had an effect ments after the treatment. The results showed that CBT
(5 RCTs, n ⫽ 321, MD ⫽ ⫺0.48, CI ⫽ ⫺82.57 to was not advantageous (1 RCT, n ⫽ 49, RR ⫽ 1.34,
81.61). The numbers in the individual studies after treat- CI ⫽ 0.49 to 3.66, P ⫽ 0.56). For follow-up, four stud-
ment and at follow-up were suspected of being skewed. ies were identified and the results were similar to the
There were two studies that compared post-treatment results after treatment (4 RCTs, n ⫽ 379, RR ⫽ 1.15,
results of CBT and TAU. The results showed that CBT CI ⫽ 0.82 to 1.61, P ⫽ 0.41).

168 NORD J PSYCHIATRY·VOL 65·NO 3·2011

 COGNITIVE BEHAVIOR THERAPY FOR SCHIZOPHRENIA

COMPARISON BETWEEN LONGER AND SHORTER THERAPIES negative and general symptoms. After treatment, there
The therapies that had more sessions showed better was a trend in favor of CBT but this was not statisti-
results than those with fewer sessions. This applied to cally significant. It appears that the effect of CBT is
the results in general but was seen most clearly when delayed and can be seen few months after the treatment
CBT was compared with other psychological treatments has ended. This could not be seen by the two previous
for the negative and general symptoms. After treatment meta-analyses, Lynch et al. (6) and Wykes et al. (4),
in patients with negative symptoms, those with 20 ses- because they analyzed the result after treatment and not
sions or more (ⱖ20) had better treatment results (5 RCTs, at follow-up. Thus, it seems understandable that the
n ⫽ 250, SMD ⫽ ⫺0.32, CI ⫽ ⫺0.57 to ⫺0.07, effects of CBT immediately after treatment would be
P ⫽ 0.01) than those who had 16 sessions or less (ⱕ16) similar to those of other psychological treatments, since
(3 RCTs, n ⫽ 211, SMD ⫽ 0.17, CI ⫽ ⫺0.10 to 0.45, it has been shown, for instance, that befriending has pos-
P ⫽ 0.21). With general symptoms, at follow-up those itive effects on delusions (66). Thus, in therapy (both
Nord J Psychiatry Downloaded from informahealthcare.com by Biblioteka Uniwersytetu Warszawskiego on 03/13/15

with ⱖ20 sessions had better treatment results (5 RCTs, CBT and the others), the therapeutic relationship is
n ⫽ 326, SMD ⫽ ⫺0.30, CI ⫽ ⫺0.52 to ⫺0.08, always a very important therapeutic ingredient and can
P ⫽ 0.007) than those who had ⱕ 16 sessions (3 RCTs, be assessed at the end of treatment, while specific thera-
n ⫽ 211, SMD ⫽ ⫺0.07, CI ⫽ ⫺0.34 to 0.21, P ⫽ 0.63). peutic effects in CBT are delayed in this patient group—
Notably, there were few studies, and the lower end of the as strong beliefs take time to change. A delayed effect is
CI should be observed. also common in the drug treatment of psychosis, i.e.
after starting medication, improvements usually come
The results compared with other meta-analyses after a few weeks or months (67). It takes time for the
It is difficult to make a comparison with other meta- patients to change their thoughts and behavior so that
analyses based on their design, selection criteria, outcome they feel better.
measures and group comparisons because these were dif- When CBT was compared with TAU, the outcome
ferent. However, it is important that an attempt should was in favor of CBT but there was no strong evidence
for this, i.e. it was not statistically significant. One expla-
For personal use only.

be made. A comparison was done of six meta-analyses

(4, 6, 9, 11, 64, 65) (Table 4). As in other meta-analyses, nation may be that the TAU provides case management
CBT had beneficial effects on positive, negative and gen- in which patients receive a good active treatment. The
eral symptoms (but with small treatment effect), and it TAU condition is therefore not as “clean” as it was some
seemed that the more recently the meta-analysis was per- years ago, when it was not accepted practice to discuss
formed, the less its effect size. The most recent meta- symptoms with patients.
analysis took into account whether the individual studies This group of patients is in need of medication, but
were single-blinded, and whether randomization was done, the negative aspect of this is the side effects. Thus, to
which the older ones did not do. achieve the lowest effective dose is important. The result
of medicine use showed evidence in favor of CBT after
treatment but not at follow-up. This means that the CBT
Discussion was 174.92 mg/day chlorpromazine equivalents less than
The studies included involved individuals with schizo- with other psychological treatment. One idea is that CBT
phrenia, and they followed the diagnostic criteria from may help to stabilize the patient, so that it would lead to
ICD-10, DSM-III and DSM-IV. Participants were mainly reduced medication and avoid unnecessary side effects.
studied in open psychiatric care, but also as inpatients. This could not be seen in the other meta-analyses
Results from this meta-analysis are relevant to individu- because they did not have medicine use as a measure.
als who are diagnosed with schizophrenia and who One must be cautious in drawing conclusions, as the
receive CBT. After our selection criteria, there remained individual studies were statistically heterogeneous. When
31 papers describing 22 studies. In total, 2469 individu- CBT was compared with TAU, the result was not statis-
als were included in this meta-analysis and the majority tically significant.
were men. It seems that the longer therapies, 20 sessions or
The two previous meta-analyses came to different more, had better treatment results than the shorter thera-
conclusions. The meta-analysis by Wykes et al. (4) pies of 16 sessions or less. One explanation is that the
showed a beneficial effect on positive symptoms, and the patient group is seriously ill and the therapy needs to be
meta-analysis by Lynch et al. (6) concluded that CBT longer to have a better effect. This means that therapies
is no better than other psychological treatments for the for patients with schizophrenia should be at least 20 ses-
treatment of schizophrenia. In this meta-analysis, when sions long or more and that it is necessary to have
CBT was compared with other psychological treatments follow-up sessions since change often continues after
at follow-up, there was strong evidence (with small treat- completion of therapy. “Survival analyses, conducted
ment effect) that intervention had an effect on positive, using data from more than 6,000 patients attending therapy

NORD J PSYCHIATRY·VOL 65·NO 3·2011 169

 Nord J Psychiatry Downloaded from informahealthcare.com by Biblioteka Uniwersytetu Warszawskiego on 03/13/15
For personal use only.

Table 4. The results of this meta-analysis compared with those of other meta-analyses.

170
Statistical
Meta-analyses Control group Studies method Effect size Measurements
F SARIN ET AL.

Positive symptoms
(Rector & Beck, 2001) (65) PSE, PAS, SAPS,
Post-treatment TAU 5 RCTs; n ⫽ 341 Cohen’s d 1.31 ⫾ 0.71
Follow-up TAU 3 RCTs; n ⫽ 214 Cohen’s d 1.48 ⫾ 0.95
(Jones et al., 2004) (9) PANSS pos
Post-treatment TAU 2 RCTs; n ⫽ 167 MD ⫺0.99† (CI ⫺ 1.93 to ⫺0.04)
(Zimmerman et al., BPRS, CPRS, MADS, PANSS, PAS, PSE,
2005) (11) PSYRATS, SAPS
Post-treatment Control group 14 studies; n ⫽ 1484 Hedges’ g 0.35 (CI 0.23 to 0.47)
Not blinded Control group 7 studies; n ⫽ 422 Hedges’ g 0.54 (CI 0.29 to 0.79)
Blinded Control group 7 RCTs; n ⫽ 1062 Hedges’ g 0.29 (CI 0.15 to 0.43)
Follow-up Control group 11 studies; n ⫽ 540 Hedges’ g 0.40 (CI 0.24 to 0.57)
Post-treatment Other psychological treatment 7 studies; n ⫽ 350 Hedges’ g 0.42 (CI 0.21 to 0.64)
Post-treatment TAU 8 studies; n ⫽ 702 Hedges’ g 0.30 (CI 0.15 to 0.46)
(Wykes et al., 2008) (4) PANSS, BPRS, PSYRATS
Post-treatment Control group 32 RCTs; n ⫽ 1918 Glass Δ 0.372 (CI 0.228 to 0.516)
High CTAM Control group 9 RCTs; n ⫽ 631 Glass Δ 0.222 (CI 0.016 to 0.427)
Low CTAM Control group 20 RCTs; n ⫽ 794 Glass Δ 0.487 (CI 0.311 to 0.664)
(Lynch et al., 2009) (6) MAN pos, SAPS, BPRS pos, PANSS pos,
PSYRATS pos
Post-treatments Other psychological treatment 8 CTs; n ⫽ 511 Hedges’ g ⫺0.19† (CI ⫺ 0.37 to ⫺0.02)
Not blinded Other psychological treatment 2 CTs; n ⫽ 77 Hedges’ g ⫺0.87† (Not reported)
Blinded Other psychological treatment 6 CTs; n ⫽ 434 Hedges’ g ⫺0.08† (Not reported)
This meta-analysis PANSS pos
Post-treatment TAU 3 RCTs; n ⫽ 417 Hedges’ g ⫺0.10† (CI ⫺ 0.29 to 0.09)
Follow-up TAU 3 RCTs; n ⫽ 415 Hedges’ g ⫺0.12† (CI ⫺ 0.31 to 0.08)
Post-treatment Other psychological treatment 7 RCTs; n ⫽ 504 Hedges’ g ⫺0.11† (CI ⫺ 0.29 to 0.06)
Follow-up Other psychological treatment 7 RCTs; n ⫽ 483 Hedges’ g ⫺0.21† (CI ⫺ 0.39 to ⫺0.03)
Negative symptom
(Rector & Beck, 2001) (65) SANS
Post-treatment TAU 3 RCTs; n ⫽ 214 Cohen’s d 1.08 ⫾ 0.83
Follow-up TAU 3 RCTs; n ⫽ 214 Cohen’s d 1.19 ⫾ 0.95
(Jones et al., 2004) (9) PANSS neg
Post-treatment TAU 2 RCTs; n ⫽ 167 MD ⫺2.30† (CI ⫺ 3.78 to ⫺0.82)
(Wykes et al., 2008) (4) PANSS neg
Post-treatment Control group 23 RCTs; n ⫽ 1268 Glass Δ 0.437 (CI 0.171 to 0.704)
High CTAM Control group 9 RCTs; n ⫽ 631 Glass Δ 0.206 (CI ⫺ 0.104 to 0.516)
Low CTAM Control group 14 RCTs; n ⫽ 637 Glass Δ 0.610 (CI 0.200 to 1.020)
(Lynch et al., 2009) (6) MAN neg, SANS, PANSS neg
Post-treatments Other psychological treatment 7 CTs; n ⫽ 411 Hedges’ g ⫺0.02† (CI ⫺ 0.22 to 0.18)

NORD J PSYCHIATRY·VOL 65·NO 3·2011

 Nord J Psychiatry Downloaded from informahealthcare.com by Biblioteka Uniwersytetu Warszawskiego on 03/13/15
For personal use only.

Not blinded Other psychological treatment 2 CTs; n ⫽ 77 Hedges’ g ⫺0.26† (Not reported)
Blinded Other psychological treatment 5 CTs; n ⫽ 334 Hedges’ g 0.04‡ (Not reported)
This meta-analysis PANSS neg, SANS
Post-treatment TAU 3 RCTs; n ⫽ 328 Hedges’ g ⫺0.10† (CI ⫺ 0.32 to 0.11)
Follow-up TAU 3 RCTs; n ⫽ 318 Hedges’ g ⫺0.21† (CI ⫺ 0.43 to 0.01)
Post-treatment Other psychological treatment 8 RCTs; n ⫽ 461 Hedges’ g ⫺0.10† (CI ⫺ 0.28 to 0.09)
Follow-up Other psychological treatment 8 RCTs; n ⫽ 452 Hedges’ g ⫺0.20† (CI ⫺ 0.39 to ⫺0.02)
General symptom
(Gould et al., 2001) (64) BPRS, PAS, CPRS, MADS
Post-treatment Control group 7 CTs; n ⫽ 340 Glass Δ 0.65 (CI 0.56 to 0.71)

NORD J PSYCHIATRY·VOL 65·NO 3·2011

Follow-up Control group 4 CTs; n ⫽ 277 Glass Δ 0.93 (Not reported)
(Jones et al., 2004) (9)
Post-treatment TAU 2 RCTs; n ⫽ 167 MD ⫺2.59† (CI ⫺ 4.92 to ⫺0.28) PANSS gen
Post-treatment TAU 1 RCT; n ⫽ 52 MD ⫺1.70† (CI ⫺ 5.40 to 2.00) BPRS
Follow-up TAU 1 RCT; n ⫽ 47 MD ⫺4.72† (CI ⫺ 9.19 to ⫺0.25) BPRS
This meta-analysis BPRS, CPRS, PANSS gen
Post-treatment TAU 4 RCTs; n ⫽ 476 Hedges’ g ⫺0.10† (CI ⫺ 0.29 to 0.08)
Follow-up TAU 4 RCTs; n ⫽ 469 Hedges’ g ⫺0.14† (CI ⫺ 0.32 to 0.05)
Post-treatment Other psychological treatment 9 RCTs; n ⫽ 571 Hedges’ g ⫺0.07† (CI ⫺ 0.23 to 0.10)
Follow-up Other psychological treatment 8 RCTs; n ⫽ 537 Hedges’ g ⫺0.21† (CI ⫺ 0.38 to ⫺0.04)
Depression
(Jones et al., 2004) (9) BDI-II
Post-treatment TAU 1 RCT; n ⫽ 42 MD ⫺0.10† (CI ⫺ 6.28 to 6.08)
Follow-up TAU 1 RCT; n ⫽ 34 MD ⫺0.70† (CI ⫺ 7.93 to 6.53)
This meta-analysis BDI-II
Post-treatment TAU 3 RCTs; n ⫽ 356 MD 0.92‡ (CI ⫺ 1.78 to 3.61)
Follow-up TAU 2 RCTs; n ⫽ 274 MD 1.22‡ (CI ⫺ 1.48 to 3.92)

Outcomes are symptoms. In this meta-analysis, dealing with positive symptoms, mean difference (MD) was changed to Hedges’ g (standardized mean difference, SMD) to facilitate comparison with other
meta-analyses.
RCTs, randomized controlled trials; CT, controlled trial; CI, 95% confidence interval.
Other psychological treatment, psychoeducation, befriending, supportive counseling, supportive psychotherapy, etc. TAU, treatment as usual.
PSE, Present State Examination; PAS, Psychiatric Assessment Scale; SAPS, Scale for the Assessment of Positive Symptoms; SANS, Schedule for the Assessment of Negative Symptoms; BPRS, Brief
Psychiatric Rating Scale; CPRS, Comprehensive Psychiatric Rating Scale; MADS, Maudsley Assessment of Delusions Scale; PANSS, Positive and Negative Syndrome Scale; Man, Manchester/Krawieka
scale; BDI-II, Beck Depression Inventory; pos, positive scale; neg, negative scale; gen, general scale.
CTAM, Clinical Trial Assessment Measure; high CTAM score means better trial quality, low CTAM score means less trial quality.
†A negative sign means in favor of CBT; ‡not in favor of CBT.

171
COGNITIVE BEHAVIOR THERAPY FOR SCHIZOPHRENIA
 F SARIN ET AL.

at a variety of US clinical services, replicated the dose- Glass Δ ⫽ 0.378, CI ⫽ 0.154, 0.602). This means that
response curve and illustrated that the session at which CBT affects not only psychotic symptoms but also other
50% and 75% of patients reached clinically significant outcomes such as functioning. For the future, we believe
recovery ranged from session 11 to 21, and sessions 25 that the main unsolved problems regarding knowledge of
to 45 respectively” (p. 134) (68). CBT in schizophrenia are to find reliable outcome mea-
The lack of effect of CBT on relapse rates is consis- sures from a clinical point of view that could also be a
tent with other meta-analyses (6, 10). It is not known consensus for the whole world. The authors are aware
how relapse should be prevented so that CBT may that this is not an easy task. The choice of outcome in
become a preferable treatment, but CBT is not hazardous CBT for psychoses (CHOICE) (69) is perhaps a promis-
and does not create relapse in patients. Another problem ing measure to obtain reliable results and to find evi-
is that relapse has been defined and treated differently dence of the effects of CBT on individuals with
over the years. Previously, relapse was treated in hospi- schizophrenia.
Nord J Psychiatry Downloaded from informahealthcare.com by Biblioteka Uniwersytetu Warszawskiego on 03/13/15

tal but now it is often treated in open psychiatric care, Another issue is that impairments in cognitive func-
with more visits to the doctor and home visits by key tions are a core feature in schizophrenia, and to achieve
workers (27). This means that studies measuring relapse benefit from CBT, the patient must be able to concentrate
have varied somewhat depending on what year they were and to think about their own thinking (metacognition). It
done, and the exact nature of treatment may have influ- would be interesting to assess cognitive functioning in
enced the outcome. patients before the start of CBT, and to see whether the
When our results are compared with the other meta- ones with best cognitive functioning show more treat-
analyses, they are similar to the two most recent ones, ment effects at the end of treatment, and at follow-up,
i.e. in favor of CBT but showing a small treatment compared with patients who show more impairments in
effect, which is not surprising because of the severity of cognitive function. Furthermore, it is perhaps time to
the disorder. When meta-analyses take care to exclude determine the cumulative effects of CBT and cognitive
studies where the rater is not blinded to the group allo- remediation therapy, for example, and/or to determine
cation, it reduces the size of the effect. “For instance, which treatment might be more effective for which
For personal use only.

poorer quality masking of allocation of treatments has patients. For those who show more impairment in cogni-
been shown to be associated with up to 40% increased tive function, perhaps they would benefit more if they
estimate of benefit in circulatory and digestive diseases, first received cognitive remediation therapy before they
mental health obstetrics, and childbirth” (p. 524) (4). were offered CBT therapy.
The quality of the conclusions depends on which
studies are included and how these have been combined. Acknowledgements—This project was supported by grants from
In this meta-analysis, we were careful to include well Praktikertjänst AB.
designed studies that had a low risk of bias. One limita-
tion of this survey was that no contact was made with
the authors of the studies, which might mean that the Declarations of interest: Freddy Sarin works as cognitive
studies excluded may have been more accurate than what behavior therapist for individuals with serious mental
appears in the published report. Other limitations of this illness. The authors alone are responsible for the content
meta-analysis were lack of comparison of individual and writing of the paper.
CBTs with group CBT, and of studies that had chosen
early psychosis with those where patients were more
chronic. This could not be done because there were too References
few studies, particularly in the case of early onset of 1. NICE. Schizophrenia: Core intervention in the treatment and man-
psychosis and group therapies. One consequence of this agement of schizophrenia in primary and secondary care (update).
is that effect size might be higher in the treatment of London: National Institute for Clinical Excellence; 2009.
2. SBU. Behandling av depressionssjukdomar—en systematisk litter-
early onset of psychosis than in patients who have been aturöversikt. Stockholm: The Swedish Council on Technology
ill for many years and have no job. Assessment in Health Care; 2004.
Another problem is that different studies described 3. SBU. Behandling av ångestsyndrom. Stockholm: The Swedish
Council on Technology Assessment in Health Care; 2005.
several outcomes and were presented in various ways. 4. Wykes T, Steel C, Everitt B, Tarrier N. Cognitive behavior ther-
The range of different types of treatment has led to dif- apy for schizophrenia: Effect sizes, clinical model, and methodo-
ficulties in making comparisons and in measuring the logical rigor. Schizophr Bull 2008;34:523–37.
5. NBH. Effekter av psykosociala insatser för personer med schizof-
effectiveness of CBT. This meta-analysis focused on reni eller bipolär sjukdom. En sammanställning av systematiska
symptom reduction, relapse, clinically important improve- översikter. The National Board of Health and Welfare; 2006.
ment and medication use and not on functioning, which 6. Lynch D, Laws KR, Mckenna PJ. Cognitive behavioural therapy
for major psychiatric disorder: Does it really work? A meta-
was done by Wykes et al. (4) and they demonstrated that analytical of well-controlled trials. Psychol Med 2009;40:
the CBT had a positive effect (15 studies, n ⫽ 867, 9–24.

172 NORD J PSYCHIATRY·VOL 65·NO 3·2011

 COGNITIVE BEHAVIOR THERAPY FOR SCHIZOPHRENIA

7. Sensky T, Turkington D, Kingdon D, Scott JL, Scott J, Siddle R, 31. Higgins J, Deeks J, Altman DG. Chapter 16: Special topics in statis-
et al. A randomized controlled trial of cognitive–behavioral ther- tics. In Higgins J, Green S, editors. Cochrane handbook for system-
apy for persistent symptoms in schizophrenia resistant to medica- atic reviews of intervention. Chichester: John Wiley & Sons; 2008.
tion. Arch Gen Psychiatry 2000;57:165–72. p. 481–529.
8. Tarrier N, Wittkowski A, Kinney C, McCarthy E, Morris J, 32. Deeks J, Higgins J, Altman DG. Chapter 9: Analysing data and
Humphreys L. Durability of the effects of cognitive–behavioural undertaking meta-analyses. In Higgins J, Green S, editors. Cochrane
therapy in the treatment of chronic schizophrenia: 12-month handbook of systematic reviews of intervention. Chichester: John
follow-up. Br J Psychiatry 1999;174:500–4. Wiley & Sons; 2008. p. 243–96.
9. Jones C, Cormac I, Silveira da Mota Neto JI, Campbell C. Cogni- 33. Schünemann H, Oxman A, Vist G, Higgins J, Deeks J, Glasziou P,
tive behaviour therapy for schizophrenia. Cochrane Database of et al. Chapter 12: Interpreting results and drawing conclusions. In
Systematic Reviews 2004, Issue 4. Art. No.:CD000524.2004. Higgins J, Green S, editors. Cochrane handbook of systematic
10. Pilling S, Bebbington P, Kuipers E, Garety PA, Geddes J, Orbach G, reviews of intervention. Chichester: John Wiley & Sons; 2008.
et al. Psychological treatments in schizophrenia: I. Meta-analysis of p. 359–87.
family intervention and cognitive behaviour therapy. Psychol Med 34. Altman DG, Bland JM. Detecting skewness from summary informa-
2002;32:763–82. tion. BMJ 1996;313:1200.
11. Zimmermann G, Favrod J, Trieu VH, Pomini V. The effect of cog- 35. Baker A, Bucci S, Lewin TJ, Kay-Lambkin F, Constable PM, Carr
Nord J Psychiatry Downloaded from informahealthcare.com by Biblioteka Uniwersytetu Warszawskiego on 03/13/15

nitive behavioral treatment on the positive symptoms of schizophre- VJ. Cognitive–behavioural therapy for substance use disorder in
nia spectrum disorders: A meta-analysis. Schizophr Res people with psychotic disorders. Br J Psychiatry 2006;188:439–48.
2005;77:1–9. 36. Barrowclough C, Haddock G, Lobban F, Jones S, Siddle R, Roberts
12. Lefebvre C, Manheimer E, Glanville J. Chapter 6: Searching for C, et al. Group cognitive–behavioural therapy for schizophrenia.
studies. In Higgins J, Green S, editors. Cochrane handbook of sys- Br J Psychiatry 2006;189:527–32.
tematic reviews of intervention. Chichester: John Wiley & Sons; 37. Bechdolf A, Knost B, Kuntermann C, Schiller S, Klosterkötter J,
2008. p. 95–150. Hambrecht M, et al. A randomized comparison of group cognitive–
13. ICD-10. The ICD-10 classification of mental and behavioural disor- behavioural therapy and group psychoeducation in patients with
ders. Clinical descriptions and diagnostic guidelines. Geneva: World schizophrenia. Acta Psychiatr Scand 2004;110:21–8.
Health Organization (WHO); 1992. 38. Bechdolf A, Köhn B, Knost B, Pukrop R, Klosterkötter J. A rand-
14. APA. Diagnostic and statistical manual of mental disorder. 4 ed. omized comparison of group cognitive–behavioural therapy and
International version: American Psychiatric Association; 1994. group psychoeduction in acute patients with schizophrenia: Out-
15. Higgins J, Altman DG. Chapter 8: Assessing risk of bias in included come at 24 months. Acta Psychiatr Scand 2005;112:173–9.
studies. In Higgins J, Green S, editors. Cochrane handbook of sys- 39. Cather C, Penn D, Otto MW, Yovel I, Mueser KT, Goff DC. A
tematic reviews of intervention. Chichester: John Wiley & Sons; pilot study of functional cognitive behavioral therapy (fCBT) for
2008. p. 187–235. schizophrenia. Schizophr Res 2005;74:201–9.
16. Tarrier N, Wykes T. Is there evidence that cognitive behaviour ther- 40. Durham RC, Guthrie M, Morton RV, Reid DA, Trevling LR,
For personal use only.

apy is an effective treatment for schizophrenia? A cautious or cau- Fowler D, et al. Tayside-Fife clinical trial of cognitive–behavioural
tionary tale? Behav Res Ther 2004;42:1377–401. therapy for medication-resistant psychotic symptoms. Br J Psychia-
17. Malm U, editor. Case Management—evidensbaserad integrerad psy- try 2003;182:303–11.
kiatri. Lund: Studentlitteratur; 2002. 41. Fowler D, Hodgekins J, Painter M, Reilly T, Crane C, Macmillan I,
18. Beck AT, Steer RA, Garbin MG. Psychometric properties of the et al. Cognitive behaviour therapy for improving social recovery in
beck depression inventory: Twenty five year of evaluation. Clin psychosis: A report from the ISREP MRC trial platform study
Psychol Rev 1988;8:77–100. (improving social recovery in early psychosis). Psychol Med
19. Beck AT, Epstein N, Brown G, Steer RA. An inventory for measur- 2009;39:1627–36.
ing clinical anxiety: Psychometric properties. J Consult Clin 42. Garety PA, Fowler DG, Freeman D, Bebbington P, Dunn G,
Psychol 1988;56:893–7. Kuipers E. Cognitive–behavioural therapy and family intervention
20. Åsberg M, Montgomery S, Perris C, Shalling D, Sedvall G. A com- for relapse prevention and symptom reduction in psychosis: Ran-
prehensive psychopathological rating scale. Acta Psychiatr Scand domised controlled trial. Br J Psychiatry 2008;192:412–23.
Suppl 1978;278:5–27. 43. Gumley A, O’Grady M, McNay J, Reilly J, Power K, Norrie J.
21. Kay SR, Fiszbein A, Opler LA. The positive and negative syndrome Early intervention for relapse in schizophrenia: Results of a
scale (PANSS) for schizophrenia. Schizophr Bull 1987;13: 12-month randomized controlled trial of cognitive behavioural
261–76. therapy. Psychol Med 2003;33:419–31.
22. Haddock G, McCarron J, Tarrier N, Faragher EB. Scales to measure 44. Haddock G, Tarrier N, Morrison AP, Hopkins R, Drake R, Lewis S.
dimensions of hallucinations and delusions: The psychotic symptom A pilot study evaluating the effectiveness of individual inpatient
rating scales (PSYRATS). Psychol Med 1999;29:879–89. cognitive–behavioural therapy in early psychosis. Soc Psychiatry
23. Andreasen N. The schedule for assessment of negative symptoms in Psychiatr Epidemiol 1999;34:254–8.
schizophrenia (SANS). Iowa City 1983. 45. Haddock G, Barrowclough C, Shaw JJ, Dunn G, Novaco RW,
24. Lipsey MW, Wilson WB. The efficacy of psychological, educational Tarrier N. Cognitive–behavioural therapy v. social activity therapy
and behavioural treatment. Am Psychol 1993;48:1181–209. for people with psychosis and history of violence: Randomised
25. Woods SW. Chlorpromazine equivalent doses for the newer atypical controlled trial. Br J Psychiatry 2009;194:152–7.
antipsychotics. J Clin Psychiatry 2003;64:663–7. 46. Jackson HJ, McGorry PD, Killackey E, Bendall S, Allott K,
26. Atkins M, Burgess A, Bottomley C, Riccio M. Chlorpromazine Dudgeon P, et al. Acute-phase and 1-year follow-up results of a ran-
equivalents: A consensus of opinion for both clinical and research domized controlled trial of CBT versus Befriending for first-episode
applications. Psychiatr Bull 1997;21:224–6. psychosis: The ACE project. Psychol Med 2008;38:725–35.
27. Bebbington P, Craig T, Garety PA, Fowler D, Dunn G, Colbert S, 47. Jackson C, Trower P, Reid I, Smith J, Hall M, Townend M, et al.
et al. Remission and relapse in psychosis: Operational definitions Improving psychological adjustment following a first episode of
based on case-note data. Psychol Med 2006;36:1551–62. psychosis: A randomised controlled trial of cognitive therapy to
28. Higgins J, Deeks J. Chapter 7: Selecting studies and collecting data. reduce post psychotic trauma symptoms. Behav Res Ther
In Higgins J, Green S, editors. Cochrane handbook of systematic 2009;47:454–62.
reviews of intervention. Chichester: John Wiley & Sons; 2008. 48. Jolley S, Garety PA, Craig T, Dunn G, White J, Aitken M. Cogni-
p. 151–85. tive therapy in early psychosis: A pilot randomized controlled trial.
29. RevMan. Review Manger (RevMan) 5.0 ed. Copenhagen: The Behav Cogn Psychother 2003;31:473–8.
Nordic Cochrane Centre, The Cochrane Collaboration; 2008. 49. Lecomte T, Leclerc C, Corbière M, Wykes T, Wallace CJ, Spidel A.
30. Deeks J, Higgins J. Statistical algorithms in review manager 5: Group cognitive behavior therapy or social skills training for indi-
On behalf of the Statistical Methods Group of the Cochrane viduals with a recent onset of psychosis? Results of a randomized
Collaboration; 2007. controlled trial. J Nerv Ment Dis 2008;196:866–75.

NORD J PSYCHIATRY·VOL 65·NO 3·2011 173

 F SARIN ET AL.

50. Haddock G, Lewis S, Bentall R, Dunn G, Drake R, Tarrier N. 61. Turkington D, Kingdon D, Rathod S, Hammond K, Pelton J,
Influence of age on outcome of psychological treatments in Mehta R. Outcomes of an effectiveness trial of cognitive–behav-
first-episode psychosis. Br J Psychiatry 2006;188:250–4. ioural intervention by mental health nurses in schizophrenia.
51. Lewis S, Tarrier N, Haddock G, Bentall R, Kinderman P, Kingdon Br J Psychiatry 2006;189:36–40.
D, et al. Randomised controlled trial of cognitive–behavioural ther- 62. Turkington D, Kingdon D, Turner T. Effectiveness of a brief
apy in early schizophrenia: Acute-phase outcomes. Br J Psychiatry cognitive–behavioural therapy intervention in the treatment of
Suppl 2002;181:91–7. schizophrenia. Br J Psychiatry 2002;180:523–7.
52. Tarrier N, Lewis S, Haddock G, Bentall R, Drake R, Kinderman P, 63. Valmaggia LR, Van Der Gaag M, Tarrier N, Pijnenborg M, Sloof
et al. Cognitive–behavioural therapy in first-episode and early schiz- CJ. Cognitive–behavioural therapy for refractory psychotic symp-
ophrenia. Br J Psychiatry 2004;184:231–9. toms of schizophrenia resistant to atypical antipsychotic medication.
53. Penn D, Meyer P, Evans E, Wirth RJ, Cai K, Burchinal M. A rand- Br J Psychiatry 2005;186:324–30.
omized controlled trial of group cognitive–behavioral therapy vs. 64. Gould RA, Mueser KT, Bolton EE, Mays VK, Goff DC. Cognitive
enhanced supportive therapy for auditory hallucinations. Schizophr therapy for psychosis in schizophrenia: An effect size analysis.
Res 2009;109:52–9. Schizophr Res 2001;48:335–42.
54. Rector NA, Seeman MV, Segal ZV. Cognitive therapy for schizo- 65. Rector NA, Beck AT. Cognitive behavioral therapy for
phrenia: A preliminary randomized controlled trial. Schizophr Res schizophrenia: An empirical review. J Nerv Ment Dis 2001;189:
Nord J Psychiatry Downloaded from informahealthcare.com by Biblioteka Uniwersytetu Warszawskiego on 03/13/15

2003;63:1–11. 278–87.
55. Turkington D, Sensky T, Scott J, Barnes TRE, Nur U, Siddle R, 66. Samarasekara N, Kingdon D, Siddle R, O’Carroll M, Scott JL,
et al. A randomized controlled trial of cognitive–behavior therapy Sensky T, et al. Befriending patients with medication-resistant
for persistent symptoms in schizophrenia: A five-year follow-up. schizophrenia: Can psychotic symptoms predict treatment response?
Schizophr Res 2008;98:1–7. Psychol Psychoter 2007;80:97–106.
56. Tarrier N, Kinney C, McCarthy E, Humphreys L, Wittkowski A, 67. Fass. FASS. Stockholm: Läkemedelsindustriföreningen.www.fass.
Morris J. Two-Year follow-up of cognitive–behavioral therapy and se; 2009.
supportive counseling in the treatment of persistent symptoms in 68. Newnham EA, Page AC. Bridging the gap between best evidence
chronic schizophrenia. J Consult Clin Psychol 2000;68:917–22. and best practice in mental health. Clin Psychol Rev 2010;30:
57. Tarrier N, Kinney C, McCarthy E, Wittkowski A, Yusupoff L, Gle- 127–42.
dhill A. Are some types of psychotic symptoms more responsive to 69. Greenwood KE, Sweeney A, Williams S, Garety P, Kuipers E,
cognitive–behaviour therapy. Behav Cogn Psychother 2001;29:45–55. Scott J, et al. CHoice of Outcome In Cbt for psychosEs (CHOICE):
58. Tarrier N, Yusupoff L, Kinney C, McCarthy E, Gledhill A, The development of a new service user-led outcome measure of
Haddock G, et al. Randomised controlled trial of intensive cognitive CBT for psychosis. Schizophr Bull 2010;36:126–35.
behaviour therapy for patients with chronic schizophrenia. BMJ
1998;317:303–7. Freddy Sarin, M.Sc., Järvapsykiatrin, Psychiatry, Praktikertjänst AB,
59. Trower P, Birchwood M, Meaden A, Byrne S, Nelson A, Ross K. Rinkebysvängen 70 A, SE-163 74 Spånga, Sweden.
For personal use only.

Cognitive therapy for command hallucinations: Randomised Lennart Wallin, B.Sc., 22 Socialservices, SE-691 83 Karlskoga,
controlled trial. Br J Psychiatry 2004;184:312–20. Sweden.
60. Naeem F, Kingdon D, Turkington D. Cognitive behaviour therapy Birgitta Widerlöv, Dr. Med. Sci., Department of Neuroscience,
for schizophrenia in patients with mild to moderate substance Psychiatry, Uppsala University Hospital, Uppsala University,
misuse problems. Cogn Behav Ther 2005;34:207–15. SE-751 85 Uppsala, Sweden.

174 NORD J PSYCHIATRY·VOL 65·NO 3·2011