Acta Psychiatr Scand 2010: 121: 340–350  2009 John Wiley & Sons A/S

All rights reserved ACTA PSYCHIATRICA

DOI: 10.1111/j.1600-0447.2009.01479.x SCANDINAVICA

Evaluation of a cognitive behaviourally

oriented service for relapse prevention in
schizophrenia
Klingberg S, Wittorf A, Fischer A, Jakob-Deters K, Buchkremer G, S. Klingberg1, A. Wittorf1,
Wiedemann G. Evaluation of a cognitive behaviourally oriented service A. Fischer2, K. Jakob-Deters1,3,
for relapse prevention in schizophrenia. G. Buchkremer1, G. Wiedemann1,4
1
Department of Psychiatry and Psychotherapy,
Objective: There is little work demonstrating the effectiveness of University of Tuebingen, Tuebingen, 2Fachkrankenhaus
cognitive behaviourally oriented interventions in routine service fr Psychiatrie, Psychotherapie und Neurologie, Vinzenz
settings. This pragmatic trial is designed to test the impact of a group von Paul Hospital, Rottweil, 3Institute of Psychology and
treatment service on relapse rates under the conditions of routine Ergonomics, Technical University Berlin, Berlin and
4
health care. Department of Psychiatry and Psychotherapy,
Method: A total of 169 schizophrenia patients were randomly University of Frankfurt, Frankfurt, Germany
allocated either to a comprehensive cognitive behaviourally oriented
service (CBOS) or to treatment as usual (TAU). The primary outcome
is the time until the first relapse after discharge from hospital. Relapse
was defined as an increase in positive or negative symptoms as assessed
with the Positive and Negative Syndrome Scale. Survival analysis has Key words: schizophrenia; cognitive behavioural
therapy; group treatment; randomized controlled trial;
been conducted up to the 6-month assessment. follow-up
Results: The mean time to relapse after discharge from hospital in the
CBOS group was significantly longer than in the TAU group (log rank S. Klingberg, Universittsklinik fr Psychiatrie und Psy-
chotherapie, Osianderstr. 24, D-72116 Tbingen, Ger-
test, P = 0.033). This was due to less exacerbations regarding negative
many.
symptoms in the CBOS condition (log rank test, P = 0.014). The E-mail: stefan.klingberg@med.uni-tuebingen.de
number of social contacts was improved in the CBOS group only.
Conclusion: The CBOS intervention appears to be beneficial in
reducing early negative symptom exacerbations. Accepted for publication August 31, 2009

Significant outcomes
• Our cognitive behaviourally oriented service is useful in preventing early relapses.
• Cognitive behaviourally oriented strategies might be indicated for the treatment of negative
symptoms.
• Cognitive behaviourally oriented interventions are applicable in routine care settings.

Limitations
• The patients of the cognitive behaviourally oriented service received further group treatment after
discharge from hospital, which patients of the control group (treatment as usual) did not receive.
Thus, the study design does not allow any definite conclusion regarding active ingredients.
• As the raters were not blind, a potential bias in assessment might have contributed to the positive
results found.

number of successful clinical trials indicates that

Introduction
symptom reduction (1–3), relapse prevention (4–9)
A variety of psychological intervention strategies and improvement of social functioning (10, 11) can
for patients suffering from schizophrenic or other be achieved by using these interventions. It has been
psychotic disorders and their relatives has been established that the involvement of the family in the
developed in the last two decades. An increasing treatment decreases the risk of relapse (12–14).

340
 Cognitive behaviourally oriented service

There is evidence that studies applying more treatment conditions were largely provided on an
rigorous methodology show lower effect sizes (1). In in-patient basis.
addition, also negative results emerge. For example Our cognitive behaviourally oriented service
Garety et al. did not find that a cognitive beha- (CBOS) was manualized and consisted of five
vioural therapy (CBT) for the reduction of positive group components. i) A psychoeducational group
symptoms is efficacious in reducing relapses (15). therapy aimed at providing information about the
However, recently published evidence-based illness and the treatment, establishing a functional
treatment guidelines show that there is sufficient subjective illness concept, fostering the patient–
positive evidence to recommend psychological inter- therapist cooperation and improving crisis coping
vention, in particular cognitive behavioural therapy skills (16 · 60 min, two sessions per week). ii) A
and family intervention, for routine care (16–18). social–emotional skills training was designed to
In contrast to this recommendation, research on improve emotion perception and emotion expres-
the implementation of treatment programmes in sion and to enhance social skills by role-play
routine service settings is in its early stage and only training and homework assignment (16 · 60 min,
few studies are available (19–21). These studies two sessions per week). iii) A social group treatment
compared specialized psychological intervention addressed the patientsÕ living situation, occupation
programmes with standard care and found reduc- and leisure time. Patients engaged in discussions
tion of relapses (21), reduction of overall symp- about their personal situation (e.g. need for sup-
toms (19) and in particular negative symptoms ported housing or supported employment) and in
(20). role-play training of relevant skills (e.g. communi-
This pragmatic trial is designed to test the cation with the property owner or boss; 8 · 60min,
impact of a group treatment programme on relapse one session per week). iv) In addition, relatives
rates under the conditions of German mental received structured group sessions providing infor-
health care. The intervention programme of this mation about illness and treatment and aiming at
trial was designed to integrate strategies which relieving the burden of care (4 · 120 min, one
have shown efficacy regarding the reduction of session fortnightly). v) After in-patient treatment
relapse rates: psychoeducation (22), early symptom lasting approximately 8 weeks patients participated
management (6), stress management (7) and inclu- in a needs-based out-patient therapy group for
sion of relatives in the treatment (14). The combi- 6 months with six weekly and eight fortnightly
nation of these strategies draws on a previous trial sessions. The treatment addressed stress coping
(8) and is in line with the general definition of CBT. skills, crisis management skills, coping with day-to-
Jones et al. (2) for example defined CBT as a day problems and persistent positive and negative
treatment which Ôinvolves the recipient establishing symptoms depending on the needs of the patients.
links between their thoughts, feelings and actions Regular four weekly visits with a psychiatrist were
with respect to the target symptomÕ and aims at also part of the out-patient treatment. The compo-
Ôthe correction of the personÕs misperceptions, nents of our treatment package are believed to have
irrational beliefs and reasoning biases related to synergistic effects. The treatments are especially
the target symptomÕ. tailored to the patientsÕ stage of illness, their
increasing capabilities and their individual needs.
Further, CBOS patients got an individual support-
Aims of the study
ive treatment with one session per week during the
The present study aims at evaluating a combina- in-patient treatment phase.
tion of cognitive and behavioural strategies for Treatment as usual (TAU) was chosen as control
relapse prevention implemented in a service setting. intervention to address the question whether a
We hypothesized that a cognitive behaviourally systematic cognitive behaviourally oriented
oriented service would be superior to standard approach could improve the current German stan-
treatment in preventing relapses. dard treatment. In-patient TAU consisted of indi-
vidual supportive treatment (16 · 30 min, two
sessions per week), which focused on the wellbeing,
Material and methods the functional status, and on daily events. In
addition, patients participated in a group therapy
Interventions
comprising information, benefits advice, advocacy,
Both treatment conditions were conducted in two and emotional support (16 · 45 min, two sessions
centres, a university hospital and a state hospital of per week), and received support for the social
psychiatry and psychotherapy each responsible for situation provided by a social worker, if needed.
in-patient treatment of its catchment area. Both Out-patient TAU consisted of regular four weekly

341
Klingberg et al.

visits with a psychiatrist. The case load of therapists

Outcomes
was identical in both groups. Thus, in the CBOS
condition therapeutic resources were allocated Based on the standard deviation of 6 points
mainly in group therapy compared to the more observed in the original Kay et al. (25) study
individual supportive treatment in the TAU group. about the Positive and Negative Syndrome Scale
Therapists were residents ⁄ registrars in psychia- (PANSS), we defined relapse as an increase of 6
try or masters level clinical psychologists in both points on the standard PANSS positive and ⁄ or
study conditions. Therapists in the CBOS condi- negative syndrome subscale. The primary endpoint
tion had to be enrolled in a formal training in of the study was relapse in terms of an increase of
cognitive behavioural treatment for at least 1 year positive and ⁄ or negative symptoms for minimum
and to serve as co-therapist for 8 weeks. To of 7 days as in Hogarty et al. (7). This increase in
establish adherence to the treatment manual ther- symptoms was related to participantsÕ symptoms
apists had to complete sessionsÕ protocols. Patients rated at discharge from hospital on the PANSS.
and therapists filled in session rating questionnaires The primary endpoint was defined a priori at
after each session to assess the therapeutic rela- the time of the grant application and was analysed
tionship. Treatment sessions of the CBOS condi- as laid down in the study protocol. This paper
tion were video taped. Supervision was provided reports on the relapse data up to the 6-month
regularly by GW and SK. As no operationalized assessment.
adherence ratings for group treatments were appli- Clinical interviews and symptom ratings were
cable, we relied only on the number of sessions scheduled in monthly intervals after discharge from
participated by each patient. hospital to obtain continuous information about
Antipsychotic medication was offered to all the course of symptoms, relapses, the social situ-
patients according to established treatment guide- ation and the treatment (medication and side
lines. Patients received medication according to effects). The monthly assessments were conducted
their individual needs. Medication side effects were within a predefined timeframe (tx ± 1 week). In
measured by the Dosage Record and Treatment the case of missing assessments, relapses were
Emergent Symptom Scale (DOTES, 23). retrospectively recorded based on all available
information (e.g. reports of relatives or psychia-
trists and patientsÕ information). Further, we made
Participants any efforts to detect and promptly document
relapses independent of the scheduled appoint-
All patients living in the catchment areas of the two
ments (e.g. in the context of rehospitalization). In
psychiatric hospitals involved in this study were
each case, the onset of relapse had to be dated back
eligible. The inclusion criteria were:
to the precise date as reliable as possible. Four
i) schizophrenia or schizoaffective disorder clinical psychologists (masterÕs level) performed the
according DSM-IV assessed by a structured interviews. They had to have at least a professional
clinical interview (SCID I), experience of 1 year on a psychiatric ward as
ii) in-patient treatment with beginning stabiliza- clinical psychologist. To establish a sufficient inter-
tion, rater reliability raters were trained regarding the
iii) age ‡18 and £60 years, PANSS-ratings by the principal investigators (SK,
iv) verbal IQ ‡ 80 (less than 80 represents intel- GW). They discussed ratings of at least 10 patients
ligence below average taking error of mea- in detail. Inter-rater reliability was assessed using
surement into account) assessed by a another 10 videos of patient interviews. The
vocabulary test (24), intraclass correlation coefficient for the positive
v) no CNS disease as assessed applying routine syndrome was 0.91, for the negative syndrome
care examination, 0.86. The raters were independent from the treat-
vi) fluent German speaker, ment team. They were not allowed to be involved
vii) travel time to hospital £1 h, in any aspect of the treatment. However, ratings
viii) no comorbid substance dependence during the were not blind as the follow-up assessments cov-
last 6 months and, no ongoing substance ered much information about out-patient treat-
abuse during the last 4 weeks according to ment. PANSS-data were entered twice in order to
DSM-IV ⁄ SCID-criteria which would indicate avoid data entry errors.
the need for specialized treatment, Secondary endpoints of the study were rehospi-
ix) no admission to long-term in-patient rehabil- talization, measures of the social situation as well
itation planned, as questionnaire data regarding quality of life,
x) willingness to give written informed consent. illness- and self-concepts. In this paper, we will

342
 Cognitive behaviourally oriented service

report on rehospitalization, occupation, financial Sequence generation. A stratified block randomi-

support, living condition and social contact. zation was applied with study centre, medication
compliance (26) and gender as strata. Medication
compliance and gender were the stratification
Randomization
variables as defined in the study protocol. Medi-
All patients were admitted to hospital because of cation compliance is one of the most important
an acute episode of their psychotic disorder. We predictors of relapse. A balanced gender ratio was
considered patients to be able to give written in our opinion important for the group atmosphere
informed consent if they voluntarily agreed to be of the study treatment groups.
treated at an open ward. After giving written
informed consent, patients were allocated ran- Implementation. The Institute for Medical Infor-
domly to either CBOS or TAU. The intervals mation Processing (IMI), University of Tuebingen,
from admission to hospital to study inclusion performed randomization independently from the
(randomization) were comparable for both patient treatment institutions. The IMI generated and
groups [CBOS: n = 84, mean = 18.3 days, SD = stored the allocation sequence. The IMI was
16.9 days; TAU: n = 85, mean = 23.2 days, informed by fax about patient inclusions and
SD = 24.8 days; t (167) = 1.507, P = 0.134]. communicated the group allocation in return to
During this period of stabilization, patients got the recruiting psychologist (also by fax). To have
pharmacological treatment as well as individually comparable group sizes throughout the treatment
administered supportive contacts with their psy- phase of the study, a block randomization strategy
chiatrists and nurses. The moderate mean PANSS was applied. The block size was concealed up to
scores (Table 1) indicate that patients had reached the end of the recruitment phase. The block size
the beginning stabilization phase at the point of within each stratum was four.
randomization.

Table 1. Sample description

Total sample CBOS TAU P (statistics)

N 169 84 85
Age (years)
At study inclusion, M (SD) 33 (10) 33 (11) 33 (10) 0.976
At first hospitalization, M (SD) 25 (8) 25 (8) 26 (8) 0.820
Sex (female), % 52 51 53 0.878à
First hospitalization, % 32 34 31 0.742à
Number of previous hospitalizations, Mdn (range) 2 (26) 2 (17) 3 (26) 0.785§
Cumulated duration of hospitalization (weeks), Mdn (range) 9 (297) 8.5 (297) 10 (156) 0.928§
First degree relative with schizophrenia, % 20 24 16 0.306à
Suicide attempt, % 25 23 26 0.720à
Diagnosis (according to DSM-IV ⁄ SCID I)
Paranoid schizophrenia, % 61 63 59
Other schizophrenia subtypes, % 28 23 33
Schizoaffective psychosis, % 11 14 8 240à
Comorbidity axis I (according to DSM-IV ⁄ SCID I), % 14 12 15 0.655à
Personality disorder (according to DSM-IV ⁄ SCID II), % 30 30 29 0.598à
Global Assessment of Functioning Scale (GAF), M (SD) 43.49 (10.63) 42.80 (10.93) 44.16 (10.34) 0.405
Positive and Negative Syndrome Scale (PANSS)
Positive syndrome, M (SD) 2.12 (0.72) 2.24 (0.77) 2.00 (0.65) 0.028
Negative syndrome, M (SD) 2.30 (1.00) 2.28 (1.08) 2.32 (0.92) 0.768
General psychopathology, M (SD) 1.92 (0.46) 1.96 (0.51) 1.88 (0.40) 0.419§
Symptom Checklist SCL-90-R, GSI, M (SD) 1.04 (0.76) 0.97 (0.72) 1.11 (0.80) 0.277
Verbal-IQ (MWT-B), n = 140, M (SD) 105 (16) 105 (15) 105 (17) 0.921
Medication compliance (favourable), % 77 75 79 0.588à
Employment
Regular employment, % 24 25 24
Supported or no employment ⁄ housewife, % 76 75 76 0.859à
Relatives
Without contact ⁄ no relatives, % 34 29 39
With contact, % 66 71 61 0.246à

t-test for independent samples.

àv2-test.
§U-test has been computed because variances were inhomogeneous. All tests were computed two-tailed.

343
Klingberg et al.

social situation has been classified as Ôpre- and

Statistical methods and sample size calculation
postpositiveÕ (favourable at study entry and at
Power calculation. On the background of the 6-month assessment) Ôpre- and postnegativeÕ
2-year follow-up study of Buchkremer et al. (8), (unfavourable at study entry and at 6-month
we planned the study to detect a difference of 24% assessment), positive change (unfavourable at
regarding the relapse rate between the CBOS and study entry and favourable at 6-month assessment)
the TAU group. The reduction of the relapse rate and negative change (favourable at study entry and
about 20–25% (moderate effects) with an alpha of unfavourable at 6-month assessment). Changes
5% and a beta of 20% should lead to a significant within the groups will be analysed using two-
result. Relapse rates of 25% in the CBOS and of tailed McNemar v2-tests on dichotomized vari-
50% in the TAU group (8) result in an effect size ables. Parametrical statistics did not seem to be
h = 0.52. With h = 0.5 and the above-mentioned appropriate, as these variables were no interval
error probabilities a sample size of n = 49 is measurements. These analyses are to be interpreted
necessary for both the CBOS and the TAU group. as exploratory.
Thus, we planned to analyse 2 · 50 patients, who
were treated according to protocol. To compensate
for the relatively high rates of loss to follow-up and Results
withdrawal from treatment, we increased the
Patient selection
number of patients to be recruited.
The primary hypothesis according to the study The CONSORT-chart (Fig. 1) shows the patient
protocol is that patients in the CBOS group will flow through the study. The recruitment and
show a significantly longer time to relapse com- treatment phase started in July 1998. Recruitment
pared to the control group. Participants were ended in June 2001. The in-patient treatment phase
analysed in the treatment group to which they ended in December 2001. The out-patient treat-
were randomized irrespective of whether they ment phase was finished in December 2002. Anal-
adhered to their treatment [according to the inten- yses have not been started since the treatment
tion-to-treat principle (27)] using Kaplan–Meyer phase was completed to avoid any bias during the
survival analytic methods with one-tailed log rank treatment.
tests (a = 0.05) and interpreted as confirmatory A total of 375 patients had fit the inclusion
hypothesis test. We decided not to rely on imputa- criteria for the study, 169 ⁄ 375 (45%) had given
tion techniques like Ôlast observation carried for- agreement with study participation. The most
wardÕ or analysing drop-out as relapse as the frequent reason for refusal of study participation
survival analysis adequately accounts for missing was that patients did not see themselves as
(i.e. censored) data. The survival analysis includes suffering from any disorder or requiring any kind
censored cases based on their status (relapsed vs. of treatment (156 ⁄ 206). Randomization or the
non-relapsed) up to the point of withdrawal. need to fill out questionnaires was the reason for
Figure 1 shows that 16 CBOS and 19 TAU refusal in only 23 (11%) of the 206 cases. Twenty-
patients completely refused to be followed-up. seven (13%) of the 206 patients had other reasons
Further, five patients of each the CBOS and the for refusal, e.g. preference for treatment in other
TAU group discontinued to take part in the formats (e.g. individual treatment) or in other
follow-up assessment during the first 6 months institutions.
after discharge from hospital. These attrition rates
were comparable between the groups. Psychopathology of refusing patients. To compare
Secondary endpoints regarding otherwise participants (n = 169) and refusing patients
defined relapse rates and treatment according to (n = 206) with respect to symptom severity, we
protocol will be analysed applying the same obtained symptom ratings of 155 (75%) of the 206
statistical methods. refusing patients. The 51 (25%) refusing patients
Employment, financial situation, living situation did not agree to participate in a clinical interview.
and social contact have been dichotomized into We found that there was no significant difference
either a positive or a negative state. As positive between both groups regarding positive or negative
states have been defined regular employment, symptoms as assessed with PANSS (positive
financial independency, no supported housing, symptoms: participants mean = 14.5, SD = 4.8;
frequent social contacts. Supported or no employ- refusals mean = 14.8, SD = 5.6; F = 0.171;
ment, financial dependency of family or govern- P = 0.679. Negative symptoms: participants
ment, supported housing, and few social contacts mean = 16.4, SD = 6.7; refusals mean = 16.5,
have been defined as negative states. The course of SD = 7.7; F = 0.11; P = 0.746).

344
 Cognitive behaviourally oriented service

Screened: n =1652

Not included: n =1483

391 no schizophrenic or schizoaffective
disorder
886 fulfilled exclusion criterion

Inclusion
147 age >60, < 18 years
83 verbal intelligence < 80
35 CNS-disease
94 language problems
96 travel time to hospital > 1h
192 substance dependence
204 outpatient treatment not feasible
35 other

206 refusal of study participation

Randomised: n =169

CBOS: n =84 TAU: n =85

Treatment

61 of 84 inpatient treatment 58 of 85 inpatient treatment

according to protocol according to protocol
42 of 84 inpatient and outpatient
treatment according to
protocol

6-month assessment 6-month assessment

Assessment

63 of 84 cases uncensored 61 of 85 cases uncensored

21 of 84 cases censored 24 of 85 cases censored

Analyses sets Analyses sets

Data analysis

cross-sectional at 6-month assessment cross-sectional at 6-month assessment

63 of 84 main analysis 61 of 85 main analysis

41 of 84 per protocol 48 of 85 per protocol

Fig. 1. CONSORT flow chart.

the in-patient treatment phase. We considered

Baseline data
those patients as treatment takers who had partic-
Table 1 shows the sample description at random- ipated in at least 2 ⁄ 3 of the scheduled in-patient
ization. To control for the success of the random- sessions. In 54% of the CBOS cases (n = 45) one
ization, we analysed the differences between the or more relatives took part in the relatives support
study conditions. There was no significant differ- group. Nineteen (23%) of the 84 patients did not
ence with one exception. The mean item score of participate in the out-patient CBOS group treat-
positive symptoms (PANSS) was 0.24 points ment. After discharge from hospital, some patients
higher in the CBOS group. moved to another city. The other patients refused
to participate any longer. Forty-seven patients
(56%) at least partially participated (one-half of
Treatment
the scheduled sessions, i.e. >6 sessions) in the out-
Sixty-one (73%) of 84 patients received treatment patient group treatment. Forty-two (50%) of 84
according to protocol in the CBOS group during patients received two-thirds of the in-patient and

345
Klingberg et al.

one-half of the out-patient sessions (treatment hospital (15 and 111 days after discharge). The
according to protocol). In the TAU group 58 outcome of these patients was registered as unfa-
(68%) of 85 patients participated in a regular vourable in all outcome measures. No suicide was
standard treatment (treatment according to proto- observed in the CBOS group.
col), the other 27 (32%) left the hospital early Of 63 (75%) of 84 CBOS, and of 61 (72%) of 85
before a sufficient stabilization for discharge was TAU patients respectively continuous information
achieved. about the positive or negative symptoms during the
During the in-patient treatment phase, the 6 months after discharge from hospital was avail-
CBOS patients received a mean daily dose able (see Fig. 1). Sixteen CBOS and 19 TAU
(amount taken) of antipsychotic medication of patients completely refused to be followed-up.
490 mg chlorpromazine equivalents (CPE), the Further, five patients of each the CBOS and the
TAU patients of 527 mg CPE. The difference was TAU group discontinued to take part in the
not significant (U-test, P = 0.50). Six months after follow-up assessment after discharge from hospital.
discharge from hospital the CBOS patients had a These attrition rates were comparable between the
daily dose of 321 mg CPE, TAU patients of groups.
317 mg CPE. Again, the difference was not signif-
icant. Two patients of the CBOS and nine patients Primary endpoint main analysis (increase of positive
of the TAU condition did not take any antipsy- and ⁄ or negative symptoms). Eighteen of 61 (30%)
chotic medication at the point of 6-month follow- uncensored cases relapsed in the TAU group (with
up. We further calculated a repeated measurement 19 censored cases at the beginning of the analysis
anova across the mean daily doses for the months and five during the follow-up period) as opposed to
one to six after discharge from hospital. There was 10 of 63 (16%) uncensored cases in the CBOS
no significant main effect for the factor group (F = group (with 16 censored cases at the beginning of
0.664, P = 0.417) but a significant main effect the analysis and five during the follow-up period)
for the factor time (F = 10.347, P < 0.001). (see Fig. 2). The mean time to relapse after
However, the time · group interaction was not discharge from hospital was 157 days (SE =
significant (F = 1.792, P = 0.183). Thus, the 5.52) in the TAU group and 168 days (SE = 4.32)
antipsychotic dose was significantly but not differ- in the CBOS group. This difference is significant (log
entially reduced in both treatment conditions rank test: v2 = 3.38; P = 0.033). The sample risk
during the follow-up period. The proportion of
patients in the total sample receiving atypical
antipsychotics was 69% at baseline. This propor- 100
tion was not different between the study condi-
tions. Further, the proportion of patients receiving
benzodiazepines and antidepressant medication 90
was not significantly different between the treat-
ment conditions.
80
% without relapse

Side effects. There was no significant difference

between the study conditions regarding behavio-
70
ural, neurological, vegetative and cardiovascular
side effects as measured by the DOTES. The
objective weight increased in both groups (70.44– 60
79.69 kg; F = 42.6; P < 0.0001) without between
group differences (F = 0.034; P = 0.856) or inter- CBOS
action effects (F = 0.076; P = 0.784). 50 TAU
To control treatment outside of the study, we CBOS-censored
asked patients about visits with General Practitio- TAU-censored
ners, psychotherapists and social workers. There 40
0 30 60 90 120 150 180
was no significant difference between the study Days after discharge from hospital
conditions.
Fig. 2. Time to first relapse (primary endpoint) in cognitive
behaviourally oriented service (CBOS) and treatment as usual
Outcomes (TAU) groups (main analysis). Note: log rank test (one-tailed):
v2 = 3.38; P = 0.033; CBOS n = 84 (16 censored cases at the
Two patients of the TAU group committed suicide beginning of the analysis, five cases censored during the
during the first 6 months after discharge from interval); TAU n = 85 (19 + 5 censored cases).

346
 Cognitive behaviourally oriented service

difference is RD = 0.136 (CI: 0.014–0.259) in 100

favour of the CBOS group. The risk ratio is RR =
1.859 [CI (log units): 1.042–3.318]. The odds ratio is
odds ratio (OR) = 2.219 [CI (log units): 1.065– 90
4.620].
80

% without relapse
Negative symptoms (main analysis). Thirteen of
59 (22%) uncensored cases relapsed in the TAU
group (19 + 7 censored cases), five of 62 (8%) 70
cases in the CBOS group (16 + 6 censored cases).
The mean time to relapse is significantly different (see
Fig. 3; TAU = 161 days, SE = 5.23; CBOS = 60
176 days, SE = 2.64; log rank test: v2 = 4.89;
P = 0.014). CBOS
50 TAU
CBOS-censored
Positive symptoms (main analysis). Thirteen of 60 TAU-censored
(22%) uncensored cases relapsed in the TAU
40
group (19 + 6 censored cases), eight of 63 (13%) 0 30 60 90 120 150 180
cases in the CBOS group (16 + 5 censored cases). Days after discharge from hospital
The mean time to relapse is not significantly different
Fig. 4. Time to first exacerbation of positive symptoms in
(TAU = 162 days, SE = 5.08; CBOS = 171 cognitive behaviourally oriented service (CBOS) and treatment
days, SE = 3.72; log-rank test: v2 = 1.71; P = as usual (TAU) groups (main analysis). Note: log rank test
0.095l; Fig. 4). (one-tailed): v2 = 1.71; P = 0.095; CBOS n = 84 (16 cen-
sored cases at the beginning of the analysis, five cases censored
during the interval); TAU n = 85 (19 + 6 censored cases.
Rehospitalization (main analysis). Nineteen of 84
(23%) uncensored cases were rehospitalized in the
TAU group (one censored case at the beginning) as (TAU = 155 days, SE = 5.51; CBOS = 162
opposed to 16 of 84 (19%) cases in the CBOS days, SE = 4.93; log rank test: v2 = 0.38; P =
group (0 censored cases). The mean time to 0.270).
rehospitalization is not significantly different
Per-protocol (PP) analyses of the primary end-
100
point. The n = 41 and n = 48 cases of the
CBOS and TAU group which were treated accord-
ing to protocol were subjected to the following
90 survival analysis. The mean time to relapse is
significantly different (TAU = 162 days, SE =
5.99; CBOS = 175 days, SE = 4.01; log rank
80 test: v2 = 3.09; P = 0.039). The relapse rate was
% without relapse

25% (11 ⁄ 44) in the TAU group and 10% (4 ⁄ 40) in

the CBOS group.
70 The PP survival analysis regarding the mean
time to negative symptomatic relapse failed to be
significant (log rank test: v2 = 2.01, P = 0.078):
60
TAU relapse rate 18% (8 ⁄ 44), CBOS relapse rate
CBOS 8% (3 ⁄ 40). However, positive symptomatic
TAU relapses occurred significantly later in the CBOS
50
CBOS-censored
group (log rank test: v2 = 2.87, P = 0.045). The
TAU-censored
relapse rates were 20% (9 ⁄ 44) in the TAU and 8%
40 (3 ⁄ 40) in the CBOS group. The mean time to
0 30 60 90 120 150 180 rehospitalization is not significantly different as in
Days after discharge from hospital
the main analysis (TAU rehospitalization rate
Fig. 3. Time to first exacerbation of negative symptoms in 12 ⁄ 57; CBOS rehospitalization rate 6 ⁄ 42; log-
cognitive behaviourally oriented service (CBOS) and treatment rank test: v2 = 0.72, P = 0.198).
as usual (TAU) groups (main analysis). Note: log rank test
(one-tailed): v2 = 4.89; P = 0.014; CBOS n = 84 (16 cen-
sored cases at the beginning of the analysis, six cases censored Social outcome (main analysis). Two patients of
during the interval); TAU n = 85 (19 + 7 censored cases). the CBOS and one patient of the TAU group

347
Klingberg et al.

refused to give information about their social compared to 122 that refused or did not attend the
status. Thus, n = 61 ⁄ 62 (CBOS) and n = 60 screening. OÕDonnell et al. (32) found 32 of 88
(TAU) patients could be included in the following eligible patients refusing. Startup et al. (33) had
analyses. In the CBOS group, we found signifi- 100 patients refusing compared to 90 included.
cantly more improvements than deteriorations Thus, high refusal rates seem to be common in
regarding social contacts. In the TAU group routine care settings. The mismatch of patientsÕ
improvements and deteriorations were not signifi- perception of their needs and the offered treatment
cantly different. The living situation was charac- remains an important clinical problem in routine
terized by an increase in the need of support in care. However, we could show that refusing
both groups. The financial situation as well as the patients were comparable to those participating
work situation did not change in both groups (see in the study regarding their level of symptoms.
Table 2). Refusal was not related to severity of symptoms.
However, the refusal rate of 55% reduces the
generalizability of findings to the more cooperative
Discussion
patients.
This study aimed at providing evidence whether a Internal validity was addressed first by random-
comprehensive CBOS approach is effective in ization, which was conducted externally in order to
reducing relapse compared to a good standard prevent any bias in group allocation. Second, we
care. External validity was addressed by the controlled for major confounding factors. We
following characteristics. First, we conducted the found that the medication was comparable
treatment not only in a university hospital but also between groups. Complementary treatment in the
in a state hospital. Access to the study was community was conducted in comparable rates in
available for all patients living in the catchment both groups. Thus, there is no indication that
area of both hospitals and being admitted to in- differences in the out-patient treatment could
patient treatment. Consequently, patients did not explain differences in outcome. Third, the primary
have advantages by study participation in terms of outcome was based on clinical interviews con-
a general higher quality of care, more costly ducted by raters not involved in the treatment and
medication, or payment for participation. Second, trained in applying the PANSS. Fourth, we imple-
we carefully observed the process of patient selec- mented measures to establish treatment adherence
tion. The majority of patients refused because they as manualized treatment, supervision in regular
did not see themselves as suffering from a psychi- intervals, session protocols, and video tapes of the
atric disorder. The refusal rate for study partici- treatment sessions.
pation was 55%. Compared to efficacy studies this We had attrition rates of 25% (n = 21 lost to
refusal rate is considerably higher. Herz et al. (6) follow-up) for CBOS and 28% (n = 24 lost to
reported a refusal of about 0%, Kuipers et al. (28) follow-up) for TAU at the 6-month follow-up. The
of 8%. Other studies as Sensky et al. (29) or main reason for attrition was comparable to the
Hogarty et al. (30) did not report details about refusal of study participation. Durham et al. (31)
recruitment. On the other hand, studies conducted lost 14% of their group at 3-month follow-up.
in service settings have reported refusal rates even Startup reported a loss of 28% in the CBT and of
higher or comparable to the refusal rate of our 26% in the TAU group at the 6-month assessment.
study. Durham et al. (31) included 66 patients Kuipers et al. (28) reported an attrition rate of
Table 2. Dichotomized assessment of four social outcome domains pre- and post-treatment – number of cases in both treatment conditions

Treatment Total Pre- and Pre- and Negative Positive

condition (n) postpositive (n) postnegative (n) change (n) change (n) Pà

Employment CBOS 62 17 38 3 4 0.999

TAU 60 12 40 3 5 0.727
Financial situation CBOS 61 26 27 2 6 0.289
TAU 60 21 32 2 5 0.453
Living situation CBOS 62 18 35 8 1 0.039
TAU 60 11 38 10 1 0.012
Social contact CBOS 61 23 18 5 15 0.041
TAU 60 15 25 7 13 0.263

CBOS, cognitive behaviourally oriented service; TAU, treatment as usual.

Three patients (partially) were not available for this assessment.
àTwo-tailed McNemar v2-tests.
P-values < 0.05 are given in bold letters.

348
 Cognitive behaviourally oriented service

18% over 9 months. Thus, the attrition rate seems out-patient group treatment which TAU patients
to be comparable with the literature. When inter- did not receive. This implicates also a higher degree
preting the group differences, it is important to of therapeutic continuity. Thus, the study design
note that the attrition rate was comparable does not allow any definite conclusion regarding
between groups. active ingredients. Second, the out-patient CBOS
There is no unitary definition of relapse in the group therapy was continuing throughout the
literature. The chosen criterion based on the 6 months after discharge from hospital. Thus, it
expectation that the majority of patients should remains unanswered whether the CBOS treatment
not demonstrate medium or higher levels of effects would disappear after completing the treat-
positive symptoms at the time of study inclusion ment programme. Third, as the raters were not
as was in fact the case (see Table 1). Thus, we did blind, a potential bias in assessment might have
not differentiate between relapse and symptom contributed to the positive results found. Finally, it
exacerbation. is possible that attrition resulted in differences
The main finding of this study is that the CBOS between the groups, as withdrawal probably is not
condition succeeded in preventing early relapses. a random process.
During the first 6 months after discharge from
hospital, CBOS patients relapsed significantly later
Acknowledgements
(approximately 2 weeks) than TAU patients did
(primary endpoint). The relapse rate differed by This study was funded by the German Research Foundation
14% (TAU 30% vs. CBOS 16%). The OR of 2.219 (Deutsche Forschungsgemeinschaft, grants Wi1523, Kl1179) as
well as by the Medical Faculty, University of Tuebingen,
indicated that CBOS reduced the probability for Germany (grant 594).
relapse by approximately 50%. Moreover, the
confidence interval CI (log units) of 1.065–4.620
shows that this reduction is significant. Hogarty Declaration of interest
et al. (7) reported relapse rates of 8% with and None.
21% without Ôpersonal therapyÕ at the 6 month
follow-up. Herz found 9% within the ÔProgramme
for relapse preventionÕ and 20% in the TAU References
condition. Thus, the relapse rate in our study is 1. Wykes T, Steel C, Everitt B, Tarrier N. Cognitive behavior
higher but the treatment effect seems to be com- therapy for schizophrenia: effect sizes, clinical models, and
methodological rigor. Schizophr Bull 2008;34:523–537.
parable with other important studies addressing 2. Jones C, Cormac I, Silveira-da-Mota-Neto JI, Campbell C.
relapse. This is worth noting as Mojtabai et al. (34) Cognitive behaviour therapy for schizophrenia. Cochrane
found group interventions to be less effective than Database Syst Rev 2004;4:CD000524.
individual treatment approaches in their meta- 3. Zimmermann G, Favrod J, Trieu VH, Pomini V. The effect of
analysis. cognitive behavioral treatment on the positive symptoms
of schizophrenia spectrum disorders: a meta-analysis.
The reduction of the relapse rate in our study Schizophr Res 2005;77:1–9.
was mainly because of a low rate of exacerbations 4. Gumley A, OÕGrady M, McNay L, Reilly J, Power K, Norrie
of negative symptoms in the CBOS as compared to J. Early intervention for relapse in schizophrenia: results of
the TAU group. This result is in line with the a 12-month randomized controlled trial of cognitive
integrated treatment from the Danish OPUS trial, behavioural therapy. Psychol Med 2003;33:419–431.
5. Bach P, Hayes SC. The use of acceptance and commitment
which also found the most marked results on the therapy to prevent the rehospitalization of psychotic
negative dimension (20). Rector & Beck (35) patients: a randomized controlled trial. J Consult Clin
reviewed the effect sizes regarding negative symp- Psychol 2002;70:1129–1139.
toms in studies on CBT. They found an effect size 6. Herz MI, Lamberti S, Mintz J et al. A program for relapse
of 0.88 for CBT and 0.39 for supportive treatment prevention in schizophrenia. A controlled study. Arch Gen
Psychiatry 2000;57:277–283.
indicating a superiority of CBT. Thus, treatment 7. Hogarty GE, Kornblith SJ, Greenwald D et al. Three-year
effects on negative symptoms have been shown trials of personal therapy among schizophrenic patients
previously. The findings regarding social contacts living with or independent of family. I: Description of
lend further support for a superiority regarding study and effects on relapse rates. Am J Psychiatry
negative symptoms in the CBOS condition. CBOS 1997;154:1504–1513.
8. Buchkremer G, Klingberg S, Holle R, Schulze Mönking H,
patients had an increase of social contacts corre- Hornung WP. Psychoeducational psychotherapy for
sponding to less negative symptoms. schizophrenic patients and their key relatives or care-
There are four major limitations of our study. givers: results of a 2-year follow-up. Acta Psychiatr Scand
First, although the therapistsÕ case load during 1997;96:483–491.
the in-patient treatment phase was identical in 9. Bechdolf A, Knost B, Kuntermann C et al. A randomized
comparison of group cognitive-behavioural therapy and
both treatment conditions, CBOS patients received

349
Klingberg et al.

group psychoeducation in patients with schizophrenia. adherence and functioning in psychotic disorders: a meta-
Acta Psychiatr Scand 2004;110:21–28. analysis. Schizophr Res 2007;96:232–245.
10. Bustillo J, Lauriello J, Horan W, Keith S. The psychosocial 23. National Institute of Mental Health. 029 DOTES. Dosage
treatment of schizophrenia: an update. Am J Psychiatry Record and Treatment Emergent Symptom Scale. In: Guy
2001;158:163–175. W, ed. ECDEU assessment manual for psychopharma-
11. Patterson TL, Leeuwenkamp OR. Adjunctive psychosocial cology (rev. ed.). Rockville: National Institute of Mental
therapies for the treatment of schizophrenia. Schizophr Health, 1976:223–244.
Res 2008;100:108–119. 24. Lehrl S. Mehrfachwahl-Wortschatz-Intelligenztest (MWT-
12. Pharoah F, Mari J, Rathbone J, Wong W. Family interven- B) [A multiple choice vocabulary intelligence test (MWT-
tion for schizophrenia. Cochrane Database Syst Rev B)], 2nd edn. Nürnberg: Perimed-spitta, 1992.
2006;4:CD000088. 25. Kay SR, Fiszbein A, Opler LA. The positive and negative
13. Pilling S, Bebbington P, Kuipers E et al. Psychological syndrome scale (PANSS) for schizophrenia. Schizophr Bull
treatments in schizophrenia: I. Meta-analysis of family 1987;13:261–276.
intervention and cognitive behaviour therapy. Psychol 26. Kemp R, Kirov G, Everitt B, Hayward P, David A. Rando-
Med 2002;32:763–782. mised controlled trial of compliance therapy. 18-month
14. Pitschel-Walz G, Leucht S, Bäuml J, Kissling W, Engel R. follow-up. Br J Psychiatry 1998;172:413–419.
The effect of family interventions on relapse and rehospi- 27. International Conference on Harmonisation of Technical
talization in schizophrenia – a meta-analysis. Schizophr Requirements for Registration of Pharmaceuticals for
Bull 2001;27:73–92. Human Use. ICH Topic E9: Statistical principles for
15. Garety PA, Fowler DG, Freeman D et al. Cognitive- clinical trials, 1998. http://www.ich.org (last accessed on
behavioural therapy and family intervention for relapse October 23, 2009).
prevention and symptom reduction in psychosis: 28. Kuipers E, Garety PA, Fowler D et al. London-East Anglia
randomised controlled trial. Br J Psychiatry 2008;192:412– randomised controlled trial of cognitive-behavioral ther-
423. apy for psychosis. I. Effects of the treatment phase. Br J
16. National Institute for Health and Clinical Excellence. Psychiatry 1997;171:319–327.
Schizophrenia. Core interventions in the treatment and 29. Sensky T, Turkington D, Kingdon D et al. A randomized
management of schizophrenia in primary and secondary controlled trial of cognitive-behavioral therapy for persis-
care (update). National Clinical Practice Guideline Num- tent symptoms in schizophrenia resistant to medication.
ber 82. London: National Institute for Health and Clinical Arch Gen Psychiatry 2000;57:165–172.
Excellence, 2009. 30. Hogarty GE, Flesher S, Ulrich R et al. Cognitive
17. Deutsche Gesellschaft für Psychiatrie Psychotherapie und enhancement therapy for schizophrenia: effects of a 2-year
Nervenheilkunde. Behandlungsleitlinie Schizophrenie, randomized trial on cognition and behavior. Arch Gen
Redaktion: Gaebel W, Falkai P. Darmstadt: Steinkopff, Psychiatry 2004;61:866–876.
2006. 31. Durham RC, Guthrie M, Morton RV et al. Tayside-Fife
18. Lehman AF, Lieberman JA, Dixon LB, McGlashan TH, clinical trial of cognitive-behavioural therapy for medica-
Miller AL, Perkins DO, Kreyenbuhl J. American Psychiat- tion-resistant psychotic symptoms. Results to 3-month
ric Association. Steering Committee on Practice Guide- follow-up. Br J Psychiatry 2003;182:303–311.
lines. Am J Psychiatry 2004; 161(2 Suppl.): 1–56. 32. OÕDonnell C, Donohoe G, Sharkey L et al. Compliance
19. Turkington D, Kingdon D, Turner T. Effectiveness of a brief therapy: a randomised controlled trial in schizophrenia.
cognitive-behavioural therapy intervention in the treat- BMJ 2003;327:834.
ment of schizophrenia. Br J Psychiatry 2002;180:523–527. 33. Startup M, Jackson MC, Bendix S. North Wales random-
20. Thorup A, Petersen L, Jeppesen P et al. Integrated treatment ized controlled trial of cognitive behaviour therapy for
ameliorates negative symptoms in first episode psychosis – acute schizophrenia spectrum disorders: outcomes at 6 and
results from the Danish OPUS trial. Schizophr Res 12 months. Psychol Med 2004;34:413–422.
2005;79:95–105. 34. Mojtabai R, Nicholson RA, Carpenter BN. Role of psy-
21. Craig TK, Garety P, Power P et al. The Lambeth Early chosocial treatments in management of schizophrenia: a
Onset (LEO) Team: randomised controlled trial of the meta-analytic review of controlled outcome studies.
effectiveness of specialised care for early psychosis. BMJ Schizophr Bull 1998;24:569–587.
2004;329:1067–1072. 35. Rector NA, Beck AT. Cognitive behavioral therapy for
22. Lincoln TM, Wilhelm K, Nestoriuc Y. Effectiveness of schizophrenia: an empirical review. J Nerv Ment Dis
psychoeducation for relapse, symptoms, knowledge, 2001;189:278–287.

350