www.elsevier.com/locate/schres

Schizophrenia Researc

The effect of cognitive behavioral treatment on the positive

symptoms of schizophrenia spectrum disorders: A meta-analysis

G. Zimmermanna,c,T, J. Favrodb, V.H. Trieub, V. Pominib

aResearch Unit in Cognitive-behavioral Therapy, Institute for Psychotherapy, Department of Psychiatry, University of Lausanne, SwitzerlandbRehabilitation Unit, Department of Psychiatry, University of Lausanne, Switzerland

cDepartment of Psychology, University of Lausanne, Switzerland

Received 19 January 2005; received in revised form 24 February 2005; accepted 28 February 2005

Available online 5 April 2005

Abstract

Background: Despite the effectiveness of anti-psychotic pharmacotherapy, residual hallucinations and delusions do not

completely resolve in some medicated patients. Additional cognitive behavioral therapy (CBT) seems to improve the

management of positive symptoms. Despite promising results, the efficacy of CBT is still unclear. The present study addresses

this issue taking into account a number of newly published controlled studies.

Method: Fourteen studies including 1484 patients, published between 1990 and 2004 were identified and a meta-analysis of

their results performed.

Results: Compared to other adjunctive measures, CBT showed significant reduction in positive symptoms and there was a

higher benefit of CBT for patients suffering an acute psychotic episode versus the chronic condition (effect size of 0.57 vs.

0.27).

Discussion: CBT is a promising adjunctive treatment for positive symptoms in schizophrenia spectrum disorders. However, a

number of potentially modifying variables have not yet been examined, such as therapeutic alliance and neuropsychological

deficits.

D 2005 Elsevier B.V. All rights reserved.

Keywords: Cognitive behavioral therapy; Schizophrenia; Positive symptoms; Meta-analysis

0920-9964/$ - see front matter D 2005 Elsevier B.V. All rights reserved.

doi:10.1016/j.schres.2005.02.018

T Corresponding author. URTC - Institut Universitaire de Psy-

chotherapie, DUPA - Universite de Lausanne, Les Cedres- Site de

Cery, CH-1008 Prilly-Lausanne, Switzerland. Tel.: +41 21 643 63

15; fax: +41 21 643 65 93.

E-mail address: Gregoire.Zimmermann@inst.hospvd.ch

(G. Zimmermann).

1. Introduction

Despite the proven efficacy of anti-psychotic

drugs, a substantial proportion of patients (10–

60%) experience medication-resistant positive symp-

toms (Lindenmayer, 2000). Adjunctive psychological

treatments that can improve outcome are clearly

needed. Based on earlier case reports (Beck, 1952),

h 77 (2005) 1–9

G. Zimmermann et al. / Schizophrenia Research 77 (2005) 1–92

specialized cognitive behavioral approaches have

been developed to decrease patient’s distress asso-

ciated with hallucinations and delusions (Haddock et

al., 1998) and these may be useful as adjunctive

treatments.

Three recent meta-analyses have confirmed that

cognitive behavioral therapy (CBT) produces prom-

ising clinical effect on patients with schizophrenia

spectrum disorders (Gould et al., 2001; Pilling et al.,

2002; Rector and Beck, 2001). In the first empirical

review, Gould et al. (2001) indicated that the mean

effect size of CBT in reducing psychotic symptoms

was 0.65 (95% CI=0.56–0.71) across seven con-

trolled trials including 340 patients. Moreover,

follow-up analyses in four studies showed that

patients continued to improve over time (Mean

ES=0.93). However, Gould and colleagues con-

ducted statistical analysis on effect sizes without

taking sample size into account. However, effect

sizes based on large samples should play a bigger

role in any statistical analysis than those based on

smaller samples (Lipsey and Wilson, 2001; Shadish

and Haddock, 1994). In a second meta-analysis of

seven randomized clinical trials (RCTs) involving

383 subjects, group contrast analysis in six of the

seven studies demonstrated a large effect size in

favor of CBT (Mean weighted ES=0.91, SD=0.14)

(Rector and Beck, 2001). However, results of Beck’s

study rely only on six studies (including 239

patients) and among them were two with methodo-

logical limitations (absence of blind rating) that

report very large effect sizes (N0.75) in favor of

CBT (Drury et al., 1996; Pinto et al., 1999). Finally,

in a more recent meta-analysis of eight RCTs

including 393 patients, Pilling et al. (2002) con-

cluded on the basis of four suitable trials (273

patients) that CBT produce clear improvements in

mental state (Pooled fixed effect odds ratio=0.27;

95% CI=0.15–0.49) compared to all other treat-

ments. This very rigorous meta-analysis provides

promising results but unfortunately did not focus on

the improvement in positive symptoms. Furthermore,

Pilling et al. (2002) did not include three eligible

trials. The first trial was probably excluded because

of the absence of randomization (Garety et al., 1994).

Surprisingly, two other studies were not included

although continuous mental state data were available

(Pinto et al., 1999; Tarrier et al., 1993).

Despite these encouraging results, the issue about

CBTefficacy in the treatment of patients suffering from

psychotic disorders remains an open question (Tur-

kington et al., 2003). Furthermore, the Cochrane

library (Cormac et al., 2003) stated that CBT may

be of interest, but that other approaches, perhaps more

widely available, are just as effective. Recently,

numerous controlled trials have been completed and

thus, it is useful to further examine the effects of CBT

now that more patients have been included in trials.

The majority of the trials of CBT have been

conducted with chronic out-patients suffering from

persistent residual psychotic symptoms. However,

results of three studies have now been published in

which CBT was used with sample of in-patients

suffering from an acute psychotic episode (Drury et

al., 1996; Lewis et al., 2002; Startup et al., 2004).

Furthermore, a number of trials have included in

addition to TAU non-specific interventions as a se-

parate treatment limb to control for common effects of

psychological treatment. Thus, it would be interesting

to examine the effects of CBT related to psychotic

status of patients (chronic condition, acute episode)

and type of control conditions (treatment as usual,

non-specific intervention, waiting-list). To address

these issues, we conducted a meta-analysis to measure

the efficacy of CBT in the treatment of positive

symptoms in schizophrenia spectrum disorders.

2. Method

2.1. Literature searches

Comprehensive searches of the psychological

literature in PsycInfo, Medline, CINHAL and Francis

databases were conducted as well as studies men-

tioned in books. We obtained article abstracts by cross

referencing the following key search words: schizo-

phrenia, psychosis, cognitive behavioral therapy, CBT

and controlled trial. We examined the reference

sections of all retrieved studies, as well as in a set

of standard books on CBT.

2.2. Selection of studies

The search produced fifty-three studies. These

identified studies were included in our meta-analysis

G. Zimmermann et al. / Schizophrenia Research 77 (2005) 1–9 3

if they met the following criteria: (1) at least one CBT

group compared with a control group (waiting-list,

treatment as usual or another therapeutic treatment),

(2) studies participants meeting DSM-III R or DSM-

IV criteria for schizophrenia, schizoaffective disorder

or delusional disorder (American Psychiatric Associ-

ation, 1987, 1994), (3) at least one validated outcome

measure of positive symptoms (hallucinations and/or

delusions), (4) study results that provided sufficient

information to compute common effect size statistics

(supplemented, when necessary, by obtaining unpub-

lished information from the authors), (5) studies

completed between 1990 and 2004.

Since Marshall et al. (2000) suggested that the use

of an unpublished scale may be a source of bias in

schizophrenia trials, we only included data from

validated scales (e.g., Brief Psychiatric Rating Scale,

Comprehensive Psychopathological Rating Scale,

Maudsley Assessment of Delusions Schedule, Pos-

itive and Negative Syndrome Scale, Present State

Examination, Psychiatric Assessment Scale, Psychotic

Symptoms Rating Scales, Scale for the Assessment of

Positive Symptoms). Table 1 specifies which criteria

were used to rate positive symptoms.

2.3. Meta-analysis procedure

The key point in meta-analysis is to compute effect

size statistics capable of representing quantitative

results of a group of studies in a form that allows

comparison and synthesis (Cohen, 1988; Lipsey and

Wilson, 2001). In the current paper, the assessment of

changes in positive psychotic symptoms was the focus.

We thus calculated effect sizes (ES) on measures of

positive symptoms. Between-groups comparisons at

post-treatment, early follow-up (12 months and less)

and late follow-up (more than a year) on variables

of interest were computed using Hedge’s g (Hedges

and Ollkin, 1985). For between-subjects designs, ES

was defined as Mt�Mc /SDpooled, where SDpooled ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiNt�1ð ÞSD2

t þ Nc�1ð ÞSD2c

� �= Nt�1ð Þþ Nc�1ð ÞÞð

q.Mt

is the mean of post-treatment experimental group,Mc is

the mean of post-treatment control group and SDpooled

the pooled within-group standard deviation. As it is not

unusual for studies to fail to report the means and

standard deviations needed to compute ES, a number of

different method were used to estimate ES from other

statistics that are reported (for a review, see Lipsey

and Wilson, 2001). Given that g is a slightly biased

estimate of ES particularly in small sample study

(Hedges and Ollkin, 1985), unbiased g-statistics

(unbiased g =[1� (3H (4N�9))]�g) were computed

and used in further analyses. By convention, ES were

coded such that a positive sign indicates that the CBT

group improves more than the control group. In cases

where a study used more than one validated measure

to examine positive symptoms, results of multiple

measures were standardized and averaged in order to

result in a single ES for any study. Moreover, in

studies that had more than one control condition,

comparisons of CBT and these separate conditions

were treated as individual studies.

Next, we aggregated effect sizes of the studies

following the procedures of Hedges and Ollkin (1985)

for weighting each study’s effect size by the inverse of

it’s variance, therefore giving more weight to studies

with smaller variance. To test the null hypothesis that

the population effect size was zero, 95% confidence

interval (CI) of the mean effect sizes were computed

for the population parameter. The null hypothesis was

rejected at significance level of .05 when zero was not

contained within the CI. In order to facilitate the

interpretation and to examine the practical importance

of ES, we used the widely known Binomial Effect

Size Display (BESD) developed by Rosenthal and

Rubin (1982). We used the homogeneity Q statistic to

test the assumption that all studies included in this

meta-analysis shared a common population effect size

(Hedges and Ollkin, 1985). The primary analysis used

a fixed effect model (FEM), considering that the

unique source of sampling error was the variation

resulting from the sampling of persons into studies

(Hedges, 1994; Hedges and Vevea, 1998). However,

to account for heterogeneity between treatments, a

randomized effect model (REM) was also employed

(Shadish and Haddock, 1994).

As the selective publication of positive findings

controlled trials is an important concern in meta-

analytic reviews of the literature (Lipsey and Wilson,

1993), the Fail-safe N (FSN) for the average effect

sizes was computed (Orwin, 1983). The FSN repre-

sent the number of unpublished (or unobtained)

studies with null effect sizes that would have to be

included to reduce the mean ES to a specified trivial

level.

Table 1

Characteristics of included trials

Studies N Design Patient

status

Interventions Outcome measuresa Blinded

assessment

Follow-up

Tarrier et al. (1993) 27 RCT Chronic Individual CBT

(6 weeks)

PSE, BPRS (Positive symptoms

elicited by the PSE and severity

rated on the BPRS)

no 6 months

Problem-solving

Waiting-list

Garety et al. (1994) 20 CT Chronic Individual CBT

(6 months)

MADS (Delusional

preoccupation, action and distress

subscales scores)

no –

Waiting-list

Drury et al.

(1996, 2000)

40 RCT Acute Individual and group

CBT (12 weeks)

PAS (Positive symptoms subscale

score)

no 9 months,

5 years

Recreation

and SPT

Kuipers et al.

(1997, 1998)

60 RCT Chronic Individual CBT

(9 months)

MADS (Delusional conviction

and distress subscales scores)

no 18 months

TAU

Tarrier et al.

(1998, 1999, 2000)

87 RCT Chronic Individual CBT

(10 weeks)

PSE, BPRS (Positive

symptoms elicited by the PSE and

severity rated on the BPRS)

yes 12 months,

2 years

Individual SPT

TAU

Pinto et al. (1999) 41 RCT Chronic Individual CBT

(12 weeks)

SAPS (Total score) no –

Individual SPT

Sensky et al. (2000) 90 RCT Chronic Individual CBT

(9 months)

CPRS (Schizophrenia

change score)

yes 9 months

Individual SPT

Lewis et al. (2002)/

Tarrier et al. (2004)

309 RCT Acute Individual CBT

(5 weeks+booster

sessions up to

3 months)

PANSS, PSYRATS (PANSS

positive subscale score and

PSYRATS delusions and

hallucinations subscales scores)

yes 18 months

Individual SPT

TAU

Turkington

et al. (2002)

422 RCT Chronic Individual CBT

(5 months)

CPRS (Schizophrenia change

score)

yes –

TAU

Durham et al. (2003) 66 RCT Chronic Individual CBT

(9 months)

PSYRATS (Hallucination

and delusion subscales scores)

yes 3 months

Individual SPT

TAU

Gumley et al. (2003) 144 RCT Early signs of

relapse

Individual CBT

(12 weeks)

PANSS (Positive subscale score) no 12 months

TAU

Rector et al. (2003) 50 RCT Chronic Individual CBT

(6 months)

PANSS (Positive subscale

score)

yes 6 months

TAU

Startup et al. (2004) 90 RCT Acute Individual CBT

(6 months)

SAPS (Positive symptoms

subscale score)

no 12 months

TAU

Trower et al. (2004) 38 RCT Chronic Individual CBT

(6 months)

PANSS, PSYRATS (PANSS

positive subscale score and

PSYRATS total score)

yes 12 months

TAU

TAU is treatment as usual, SPT supportive psychotherapy, RCT randomized clinical trial and CT controlled trial.a Names of the positive symptoms outcome instruments: BPRS=Brief Psychiatric Rating Scale, CPRS=Comprehensive Psychopathological

Rating Scale,MADS=MaudsleyAssessment ofDelusions Schedule, PANSS=Positive andNegative SyndromeScale, PAS=PsychiatricAssessment

Scale, PSE=Present State Examination, PSYRATS=Psychotic Symptoms rating Scales, SAPS=Scale for the Assessment of Positive Symptoms.

G. Zimmermann et al. / Schizophrenia Research 77 (2005) 1–94

G. Zimmermann et al. / Schizophrenia Research 77 (2005) 1–9 5

3. Results

Fifteen studies involving 1504 patients (68.1% of

male, mean age of 36.02 years, SD=4.53) met the

criteria for inclusion in this analysis. Sufficient data to

compute effect size were available from twelve

published articles. For the remaining three studies

(Haddock et al., 1999; Trower et al., 2004; Turkington

et al., 2002), authors were contacted by e-mail.

Because appropriate data were not available from

the authors, one study was excluded. Table 1 provides

an overview of descriptive information about the 14

included studies involving 1484 patients.

3.1. Post-treatment

Of theses 1484 participants, a total of 1001 patients

took part in the post-treatment analyses, including 515

who received CBT and 486 a comparison treatment.

Fig. 1 plots the individual studies effect sizes that

were included in the post-treatment analysis.

Se

Turkington et a

Durham, et al. 2003 (CBT vs SPT)

Durham, et al. 2003 (CBT vs TAU)

Tarrier, et al. 1993 (CBT vs problem solving)

-0.5 0

Effect si

Fig. 1. Effect sizes for each of the studies included in the post-treatment

The fixed effect model (FEM) mean weighted ES

across all studies was 0.35 (95% CI: 0.23–0.47), the

random effect model (REM) mean weighted ES,

0.37 (95% CI: 0.23–0.52). These ES are modestly

lower than that found for the difference between

bona fide and placebo treatments (Lipsey and

Wilson, 1993) and can be classified in the small to

moderate range of effect sizes (Cohen, 1988). Over-

all, these ES imply that a typical patient in the CBT

group improved more than 64% of the control

patients and that CBT increases the success rate of

reducing positive symptoms from 41% to 59%.

Moreover, homogeneity test was not significant

(Q =21.41, p N .05) suggesting that moderating var-

iables do not play an important role in predicting the

treatment outcome. The findings can consequently be

considered as relatively robust. The FSN analysis

indicated that 14 studies with null effect sizes would

be necessary to bring the present meta-analysis

overall ES value to the 0.20 level (small effect).

Therefore, the bfile drawer problemQ is unlikely to

Garety, et al. 1994

Drury, et al. 1996

Kuipers, et al. 1997

Tarrier, et al. 1998 (CBT vs SPT)

Tarrier, et al. 1998 (CBT vs TAU)

Pinto, et al. 1999

nsky, et al. 2000

Lewis, et al. 2002 (CBT vs SPT)

Lewis, et al. 2002 (CBT vs TAU)

l. 2002

Rector, et al. 2003

Startup, et al. 2004

Trower, et al. 2004

Tarrier, et al. 1993 (CBT vs waiting list)

Combined (17)

0.5 1

ze

group contrast analysis (unblinded trials are colored in light grey).

Table 2

Post-treatment group contrast analysis details

Analysis Studies Patients Fixed effect model (FEM) Random effect model (REM) Q statistics

N N Effect sizes

(95% CI)

BESD Control vs.

CBT success rates

Effect sizes

(95% CI)

BESD Control vs.

CBT success rates

Chronic condition 10 798 0.27 (0.11–0.42) .433–.567 0.27 (0.11–0.42) .433–.567 8.72, p N .05

Acute episode 3 203 0.57 (0.31–0.83) .363–.637 0.57 (0.31–0.83) .363–.637 1.50, p N .05

Q is a homogeneity statistic ( p N .05 indicates homogeneity).

G. Zimmermann et al. / Schizophrenia Research 77 (2005) 1–96

influence the overall result, even though only

published studies were used.

As methodological heterogeneity between studies

may result in different conclusions about treatment

efficacy (Kazdin, 1994), we compared studies with

and without blind outcome assessment (see Fig. 1).

Results indicated that the FEM and REM mean

weighted ES were 0.54 (95% CI: 0.29–0.79) across

studies without blind assessment and 0.29 (95% CI:

0.15–0.43) across studies with blind assessment.

However, an independent sample t-test revealed that

the difference between unbiased ES from blinded and

unblinded trials is non-significant (t(15)=�0.91,

p N .05).

Table 2 lists the FEM and REM mean weighted

ES for studies conducted with stabilized patients

suffering from chronic positive psychotic symptoms

and patients in acute psychotic episodes, as well as

the number of studies contributing to the ES, its 95%

CI, the BSED and the Q homogeneity statistic. Table

3 makes a distinction between comparison treatments

Table 3

Post-treatment group contrast analysis details according to different types

Analysis Studies Patients Fixed effect model (FEM)

N N Effect sizes

(95% CI)

BESD Co

CBT succ

CBT vs. non-specific treatment

Chronic condition 5 246 0.32 (0.06–0.57) .421–.579

Acute episode 2 104 0.69 (0.30–1.09) .337–.663

All conditions 7 350 0.42 (0.21–0.64) .397–.603

CBT vs. treatment as usual

Chronic condition 6 569 0.26 (0.09–0.43) .436–.565

Acute episode 2 133 0.48 (0.13–0.82) .383–.617

All conditions 8 702 0.30 (0.15–0.46) .423–.577

CBT vs. waiting-list

Chronic condition 2 47 0.55 (0.04–1.14) .367–.633

Q is a homogeneity statistic ( p N .05 indicates homogeneity).

categorized as non-specific treatment, treatment as

usual and waiting-list and provides identical statistics

than Table 2. As can be seen in Table 2, mean

weight ES were larger for patients in the acute phase

than for patients in the chronic phase of the disorder.

In the acute state, these ES were in the moderate to

large range of effects sizes, whereas for patients in

the chronic state they were in the small to moderate

range. Moreover, results reported in Table 3 indi-

cated that mean weight ES in favor of CBT were in

the moderate range when compared to waiting-list

and in the small to moderate range when compared

to non-specific treatment and treatment as usual.

However, two-way ANOVA using unbiased ES as

the dependent variable and patient status (chronic vs.

acute) and control treatment format (non-specific,

treatment as usual, waiting-list) as independent

variables, did not show any significant group effects

(patient status: F(1, 16)=1.08, p N .05; control treat-

ment format: F(2, 15)= .30, p N .05) and interaction

effects (F(2, 15)= .76, p N .05).

of control condition

Random effect model (REM) Q statistics

ntrol vs.

ess rates

Effect sizes

(95% CI)

BESD Control vs.

CBT success rates

0.31 (0.02–0.61) .423–.577 5.99, p N .05

0.69 (0.30–1.09) .337–.663 0.85, p N .05

0.43 (0.18–0.68) .395–.605 9.32, p N .05

0.30 (0.07–0.53) .426–.574 9.52, p N .05

0.48 (0.13–0.82) .383–.617 0.01, p N .05

0.32 (0.15–0.50) .421–.579 10.75, p N .05

0.55 (0.04–1.14) .367–.633 0.64, p N .05

G. Zimmermann et al. / Schizophrenia Research 77 (2005) 1–9 7

3.2. Follow-up

Of the 14 studies included, 11 studies reported

follow-up data. A total 540 patients took part in the

early follow-up analyses (256 with CBT and 284 with

comparison treatment), and a total 353 patients took

part in the late follow-up analyses (127 with CBT and

226 with comparison treatment).

At early follow-up (3–12 months), the FEM and

REM mean weight ES were 0.40 (95% CI: 0.24–

0.57), indicating that the CBT efficacy was even

slightly better compared with immediately post-treat-

ment data. The homogeneity test was not statistically

significant (Q =12.80, p N .05). Furthermore, results at

late follow-up (N12 months) suggested that this gain

was essentially maintained over an extended period

(N12 months) after the end of treatment (ES=0.33,

95% CI: 0.14–0.51 / Q =2.77, p N .05).

4. Discussion

This review provides effect size analyses for

controlled studies of cognitive behavioral treatments

of positive symptoms in schizophrenia spectrum

disorders published during the last 15 years. This

meta-analysis using a sample of 14 outcome published

studies supports the general conclusion that CBT is a

promising approach for adjunctive treatment of

positive symptoms in patients with schizophrenia.

Moreover, the therapeutic effects are preserved at

follow-up, suggesting that the CBT has long-term

effects (Garety et al., 2000).

The global mean weighted effect size of 0.37 is

considered as modest, which is not surprising,

according to Tarrier and Wykes (2004), given the

severity of the disorder. This effect size dropped

even to 0.29 when we considered only blinded trials,

indicating that methodological flaws could make it

premature to identify psychological treatment as

efficacious. The effect size in the current study is

two and half times lower than the mean weighted

between-group effect size reported in the study of

Rector and Beck (2001). However, results of Beck’s

study rely only on six studies (including 239

patients) among which two report very large effect

sizes (N0.75) in favor of CBT (Drury et al., 1996;

Pinto et al., 1999). Furthermore, our results show

that CBT has a better effect on patients in an acute

psychotic episode (mean weighed ES=0.57) than on

stabilized chronic patients suffering from persistent

psychotic symptoms (mean weighted ES=0.27).

However, this difference was not statistically sig-

nificant, and the inclusion of the Drury et al. (1996)

study presenting some methodological limitations

might have conducted to an overestimation of effect

size in acutely ill patients. Moreover, a replication of

Drury’s study with a greater methodological rigor

failed to show the superiority of CBT over suppor-

tive counseling on acutely psychotic in-patients

(Haddock et al., 1999).

The comparison group designed to control non-

specific effects of therapy exposure differ greatly

between studies, ranging from waiting-list (Tarrier et

al., 1993; Garety et al., 1994) to manual-based and

supervised supportive counseling (Tarrier et al., 1998;

Lewis et al., 2002). Further effect sizes analyses

indicate, as expected, that CBT is more superior to

waiting-list than to non-specific intervention (mostly

defined as supportive therapy) and treatment as usual.

Interestingly, effects of CBT are slightly less pro-

nounced when it is compared with treatment as usual

rather than with non-specific treatment, such as

supportive counseling. However, differences between

comparison groups were non-significant.

Several limitations of this study are acknowledged.

First, a number of studies have included non-specific

treatment in addition to standard psychiatric care.

While other treatments could be considered as placebo

treatment, these non-specific interventions are also

heterogeneous between studies (supportive counsel-

ing, befriending, problem solving). Secondly, we did

not completely investigate the methodological quality

of each trial, and methodological differences have

been shown to interfere with the detection of a true

treatment effect (Kazdin, 1994; Kazdin and Bass,

1989). In this study, unblinded trials were associated

with an increased estimate of 86% of effect size.

Consequently, stringent controlled blind trials are

recommended for future randomized clinical trials

evaluating CBT. Finally, a relatively small number of

studies available for the meta-analysis made exami-

nation of many variables of interest statistically

underpowered.

Despite the importance of research in psychother-

apy indicating that the quality of the therapeutic

G. Zimmermann et al. / Schizophrenia Research 77 (2005) 1–98

relationship is one of the strongest predictor of a

successful outcome, there is growing bias in psycho-

therapy towards bempirically supported treatmentsQdependent on the medical model. Further research is

needed to address some unexamined potentially

important variables (for example alliance, competence

of therapist, treatment dosage). Finally, recent studies

suggest that schizophrenic patients present with a

deficit in source monitoring (Keefe et al., 2002) and

that memory distortions are associated with psychotic

symptoms (Clancy et al., 2002). Thus, combining

CBT with training for specific neuropsychological

deficits associated with auditory hallucinations or

delusions may also be important for future research

trials.

Acknowledgements

We thank Kerry Ross and Douglas Turkington for

providing data.

References

American Psychiatric Association, 1987. Diagnostic and Statistical

Manual of Mental Disorders: DSM-III-R, 3rd revised ed.

American Psychiatric Association, Washington, DC.

American Psychiatric Association, 1994. Diagnostic and Statistical

Manual of Mental Disorders: DSM-IV, 4th ed. American

Psychiatric Association, Washington, DC.

Beck, A.T., 1952. Successful outpatient psychotheraphy of a

chronic schizophrenic with a delusion based on borrowed guilt.

Psychiatry 15, 305–312.

Clancy, S.A., McNally, R.J., Schacter, D.L., Lenzenweger,

M.F., Pitman, R.K., 2002. Memory distortion in people re-

porting abduction by aliens. J. Abnorm. Psychology 111 (3),

455–461.

Cohen, J., 1988. Statistical Power Analysis for the Behavioral

Sciences, 2nd ed. Academic Press, New York.

Cormac, I., Campbell, C., Silveira Da Mota Neto, J., 2003.

Cognitive behaviour therapy for schizophrenia. Cochrane Data-

base Syst. Rev. 3, 14.

Drury, V., Birchwood, M., Cochrane, R., MacMillan, F., 1996.

Cognitive therapy and recovery from acute psychosis: a

controlled trial. I. Impact on psychotic symptoms. Br. J.

Psychiatry 169, 593–601.

Drury, V., Birchwood, M., Cochrane, R., 2000. Cognitive therapy

and recovery from acute psychosis: a controlled trial. 3. Five-

year follow-up. Br. J. Psychiatry 177, 8–14.

Durham, R.C., Guthrie, M., Morton, R.V., Reid, D.A., Treliving,

L.R., Fowler, D., Macdonald, R.R., 2003. Tayside–Fife clinical

trial of cognitive-behavioural therapy for medication-resistant

psychotic symptoms. Results to 3-month follow-up. Br. J.

Psychiatry 182, 303–311.

Garety, P.A., Kuipers, L., Fowler, D., Chamberlain, F., Dunn, G.,

1994. Cognitive behavioural therapy for drug-resistant psycho-

sis. Br. J. Med. Psychol. 67, 259–271.

Garety, P.A., Fowler, D., Kuipers, E., 2000. Cognitive-behavioral

therapy for medication-resistant symptoms. Schizophr. Bull. 26

(1), 73–86.

Gould, R.A., Mueser, K.T., Bolton, E., Mays, V., Goff, D., 2001.

Cognitive therapy for psychosis in schizophrenia: an effect size

analysis. Schizophr. Res. 48 (2-3), 335–342.

Gumley, A., O’Grady, M., McNay, L., Reilly, J., Power, K., Norrie,

J., 2003. Early intervention for relapse in schizophrenia: results

of a 12-month randomized controlled trial of cognitive

behavioural therapy. Psychol. Med. 33 (3), 419–431.

Haddock, G., Tarrier, N., Spaulding, W., Yusupoff, L., Kinney, C.,

McCarthy, E., 1998. Individual cognitive-behavior therapy in

the treatment of hallucinations and delusions: a review. Clin.

Psychol. Rev. 18 (7), 821–838.

Haddock, G., Tarrier, N., Morrisson, A.P., Hopkins, R., Drake,

R., Lewis, S., 1999. A pilot study evaluating the effective-

ness of individual inpatient cognitive-behavioural therapy in

early psychosis. Soc. Psychiatry Psychiatr. Epidemiol. 34,

254–258.

Hedges, L.V., 1994. Statistical considerations. In: Cooper, H.,

Hedges, L.V. (Eds.), The Handbook of Research Synthesis.

Russel Sage Foundation, New York, pp. 29–38.

Hedges, L.V., Ollkin, I., 1985. Statistical Methods for Meta-

analysis. Academic Press, New York.

Hedges, L.V., Vevea, J.L., 1998. Fixed- and random effects models

in meta-analysis. Psychol. Methods 3, 486–504.

Kazdin, A.E., 1994. Methodology, design and evaluation in

psychotherapy research. In: Bergin, A.E., Garfield, S.L.

(Eds.), Handbook of Psychotherapy and Behavior Change, 4th

ed. John Wiley & Sons, New York, pp. 19–71.

Kazdin, A.E., Bass, D., 1989. Power to detect differences between

alternative treatments in comparative psychotherapy outcome

research. J. Consult. Clin. Psychol. 51 (1), 138–147.

Keefe, R.S.E., Arnold, M.C., Bayen, U.J., McEvoy, J.P., Wilson,

W.H., 2002. Source-monitoring deficits for self-generated

stimuli ion schizophrenia: multinomial modeling of data from

three sources. Schizophr. Res. 57, 51–67.

Kuipers, E., Garety, P., Fowler, D., Dunn, G., Bebbington, P.,

Freeman, D., Hadley, C., 1997. London–East Anglia rando-

mised controlled trial of cognitive-behavioural therapy for

psychosis: I. Effects of the treatment phase. Br. J. Psychiatry

171, 319–327.

Kuipers, E., Fowler, D., Garety, P., Chisholm, D., Freeman, D.,

Dunn, G., Bebbington, P., Hadley, C., 1998. London–East

Anglia randomised controlled trial of cognitive-behavioural

therapy for psychosis: III. Follow-up and economic evaluation

at 18 months. Br. J. Psychiatry 173, 61–68.

Lewis, S., Tarrier, N., Haddock, G., Bentall, R., Kinderman, P.,

Kingdon, D., Siddle, R., Drake, R., Everitt, J., Leadley, K.,

Benn, A., Grazebrook, K., Haley, C., Akhtar, S., Davies, L.,

Palmer, S., Faragher, B., Dunn, G., 2002. Randomised

G. Zimmermann et al. / Schizophrenia Research 77 (2005) 1–9 9

controlled trial of cognitive-behavioural therapy in early

schizophrenia: acute-phase outcomes. Br. J. Psychiatry 43

(Supplementum 43), s91–s97.

Lindenmayer, J.-P., 2000. Treatment refractory schizophrenia.

Psychiatr. Q. 71 (4), 373–384.

Lipsey, M.W., Wilson, D.B., 1993. The efficacy of psychological,

educational, and behavioral treatment. Confirmation from meta-

analysis. Am. Psychol. 48 (12), 1181–1209.

Lipsey, M.W., Wilson, D.B., 2001. Practical Meta-analysis. Sage,

Thousand Oaks, CA.

Marshall, M., Lockwood, A., Bradley, C., Adams, C., Joy, C.,

Fenton, M., 2000. Unpublished rating scales: a major source of

bias in randomised controlled trials of treatments for schizo-

phrenia. Br. J. Psychiatry 176, 249–252.

Orwin, R.G., 1983. A fail-safe N for effect size in meta-analysis.

J. Educ. Stat. 8, 157–159.

Pilling, S., Bebbington, P., Kuipers, E., Garety, P., Geddes, J.,

Orbach, G., Morgan, C., 2002. Psychological treatments in

schizophrenia: I. Meta-analysis of family intervention and

cognitive behaviour therapy. Psychol. Med. 32 (5), 763–782.

Pinto, A., Pia, S.L., Mennella, R., Giorgio, D., DeSimone, L., 1999.

Cognitive-behavioral therapy and clozapine for clients with

treatment-refractory schizophrenia. Psychiatr. Serv. 50 (7),

901–904.

Rector, N.A., Beck, A.T., 2001. Cognitive behavioral therapy for

schizophrenia: an empirical review. J. Nerv. Ment. Dis. 189 (5),

278–287.

Rector, N.A., Seeman, M.V., Segal, Z.V., 2003. Cognitive therapy

for schizophrenia: a preliminary randomized controlled trial.

Schizophr. Res. 63 (1-2), 1–11.

Rosenthal, R., Rubin, D.B., 1982. A simple, general purpose

display of magnitude of experimental effect. J. Educ. Psychol.

74 (2), 166–169.

Sensky, T., Turkington, D., Kingdon, D., Scott, J.L., Scott, J.,

Siddle, R., O’Carroll, M., Barnes, T.R., 2000. A randomized

controlled trial of cognitive-behavioral therapy for persistent

symptoms in schizophrenia resistant to medication. Arch. Gen.

Psychiatry 57 (2), 165–172.

Shadish, W.R., Haddock, C.K., 1994. Combining estimates of

effect size. In: Cooper, H., Hedges, L.V. (Eds.), The Handbook

of Research Synthesis. Russel Sage Foundation, New York,

pp. 261–281.

Startup, M., Jackson, M.C., Bendix, S., 2004. North Wales

randomized controlled trial of cognitive behaviour therapy for

acute schizophrenia spectrum disorders: outcomes at 6 and 12

months. Psychol. Med. 34, 413–422.

Tarrier, N., Wykes, T., 2004. Is there evidence that cognitive

behaviour therapy is an effective treatment for schizophre-

nia? A cautious or cautionary tale? Behav. Res. Ther. 42,

1377–1401.

Tarrier, N., Beckett, R., Harwood, S., Baker, A., et al., 1993. A trial

of two cognitive-behavioural methods of treating drug-resistant

residual psychotic symptoms in schizophrenic patients: I.

Outcome. Br. J. Psychiatry 162, 524–532.

Tarrier, N., Yusupoff, L., Kinney, C., McCarthy, E., Gledhill, A.,

Haddock, G., Morris, J., 1998. Randomised controlled trial of

intensive cognitive behaviour therapy for patients with chronic

schizophrenia. Br. Med. J. 317, 303–307.

Tarrier, N., Wittkowski, A., Kinney, C., McCarthy, E., Morris, J.,

Humphreys, L., 1999. Durability of the effects of cognitive-

behavioural therapy in the treatment of chronic schizophrenia:

12 month follow-up. Br. J. Psychiatry 174, 500–504.

Tarrier, N., Kinney, C., McCarthy, E., Humphreys, L., Wittkowski,

A., Morris, J., 2000. Two-year follow-up of cognitive-behav-

ioral therapy and supportive counseling in the treatment of

persistent symptoms in chronic schizophrenia. J. Consult. Clin.

Psychol. 68 (5), 917–922.

Tarrier, N., Lewis, S., Haddock, G., Bental, R., Drake, R.,

Kinderman, P., Kingdon, D., Siddle, R., Everitt, B., Leadley,

K., Benn, A., Grazebrook, K., Haley, C., Akhtar, S., Davies, L.,

Palmer, S., Dunn, G., 2004. Cognitive-behavioural therapy in

first episode and early schizophrenia. 18-month follow-up of a

randomised controlled trial. Br. J. Psychiatry 184, 231–239.

Trower, P., Birchwood, M., Meaden, A., Byrne, S., Nelson, A.,

Ross, K., 2004. Cognitive therapy for command hallucinations:

randomised controlled trial. Br. J. Psychiatry 184, 312–320.

Turkington, D., Kingdon, D., Turner, T., 2002. Effectiveness of a

brief cognitive-behavioural therapy intervention in the treatment

of schizophrenia. Br. J. Psychiatry 180, 523–527.

Turkington, D., McKenna, P.J., Cannon, M., McKenzie, K., Sims,

A., 2003. Is cognitive-behavioural therapy a worthwhile treat-

ment for psychosis? Br. J. Psychiatry 182, 477–479.