Participants

Participants were in-patients and out-patients, recruited from mental health services from five National Health Service trusts in the North West of England between 2000 and 2004. Potential participants were identified from mental health staff and from case-note search. Case notes were screened to see whether potential participants met initial inclusion criteria (see below for further details) and those meeting the criteria were interviewed using the Positive and Negative Syndrome Scale ^{Reference Kay, Opler and Lindenmayer7} (PANSS) to ascertain whether psychotic symptoms were present. Those who met all criteria were asked whether they would be willing to take part in the trial and those giving consent were subject to the full assessment described below. The inclusion criteria were as follows.

1. (a) Diagnosis of DSM–IV ⁸ schizophrenia or schizoaffective disorder ascertained from multiple sources and case-note checklist. ^{Reference Lewis, Tarrier, Haddock, Bentall, Kinderman and Kingdon9}

2. (b) History of violent behaviour defined as follows: either

   1. (i) at least two recorded incidents of violence to property or others in 12 months prior to trial entry, or

   2. (ii) at least three recorded incidents of verbal aggression threatening violence to property or others in 12 months prior to trial entry, or

   3. (iii) for participants in secure care, one recorded violent index offence plus an indication that the individual was at risk of further violence without the protective effect of the treatment setting. This was rated by the responsible medical officer or care coordinator using the Security, Dependency, Treatment and Political Needs Assessment Scale. ^{Reference Shaw, Davies and Morey10}

3. (c) Experiencing persistent hallucinations and/or delusions scoring four or more on the relevant PANSS sub-scales (P1 and P3).

4. (d) Receiving antipsychotic medication (dose between 400 mg and 1000 mg chlorpromazine or equivalent) ascertained from detailed case-note review.

Eligible, consenting potential participants were assessed by masked raters at baseline, after 6 weeks (baseline 2), at end of treatment (6 months) and at follow-up (12 months). Masking was maintained by ensuring therapists and assessors were housed in separate accommodation, therapy files were kept separately from data and clinical staff was repeatedly instructed not to disclose any knowledge of therapy group to assessors.

Primary outcome: assessment of aggression and violence

Data on acts of aggression and violence were collected at four time points from case notes by masked assessors:

1. (a) from 3 months prior to inclusion in trial to baseline 1

2. (b) from baseline 1 to baseline 2 (6 weeks)

3. (c) from baseline 2 to the end of treatment (6 months)

4. (d) from end of treatment to follow-up (6 months).

Data were collected on eight aggression and violence variables taken from the Ward Anger Rating Scale (WARS; see below, details available from R.W.N. on request) and were recorded if a description of an event was written in any part of the case notes during the target period. From these data, the following three variables were computed: total number of any incidents across each of the four time points by group; number of verbal incidents across each time period by group; and the number of acts of physical violence carried out over the four time periods by group. In addition, a dichotomous variable of ‘no violence’ v. ‘presence of violence’ over the intervention and follow-up period was recorded for each individual.

Secondary outcomes

Treatment

Participants were randomly allocated to group by personnel independent of the trial using a computer-generated randomisation procedure stratified for gender, presence of substance misuse, severity of anger, presence or absence of actual violence in the last 12 months and facility (out-patient v. in-patient). Participants were allocated to either:

1. (a) a CBT programme which included motivational strategies to aid engagement, CBT strategies to reduce the severity and distress of psychotic symptoms and CBT strategies to reduce the severity of anger linked to aggression and violence; or

2. (b) a SAT programme aimed at helping participants to identify activities they enjoyed and helping them to carry these out.

The rationale for the integrated CBT was based on two main factors. First, it was envisaged that this group of individuals might show considerable problems in engaging with therapy and hence the therapy was enhanced with motivational interviewing strategies to improve the potential for engagement in CBT. Second, it has been demonstrated that both psychotic symptoms and anger are implicated in the occurrence of violence, hence strategies to reduce the severity of both were included. The SAT treatment was designed to control for non-specific elements of therapy such as having a one-to-one relationship with an empathic therapist. Both treatments consisted of 25 sessions carried out by therapists who had been trained in the protocol and received ongoing supervision. Therapists who carried out the CBT programme met the British Association of Behavioural and Cognitive Psychotherapies' Minimum Training Standards for the practice of CBT and had prior experience of applying CBT for people with psychosis. Stringent procedures to manage treatment fidelity were employed as follows.

1. (a) A therapy manual for each treatment was developed during pilot work. ^{Reference Haddock, Lowens, Brosnan, Barrowclough and Novaco3}

2. (b) Individual and group supervision was provided at least fortnightly.

3. (c) Where possible, sessions were tape-recorded and samples of these were listened to by supervisors to ensure therapists were carrying out the treatment according to the protocol. Tapes from the CBT treatment were also assessed by supervisors using the Cognitive Therapy Scale for Psychosis ^{Reference Haddock, Devane, Bradshaw, McGovern, Tarrier and Kinderman19} to ensure therapists were carrying out treatment to a high standard and met criteria for CBT. Similarly, SAT tapes were also rated on the same scale to ensure that the therapists were not carrying out CBT techniques and to ensure that the SAT met non-specific therapeutic quality standards which are measured on the scale (e.g. understanding, empathy, feedback).

4. (d) An independently randomly selected sample of 20 audio-tapes of intervention sessions (10 CBT, 10 SAT) from the pool of 264 available therapy tapes were assessed by a clinical psychologist independent of the trial and masked to the therapy status of the tapes in order to identify the nature of the treatment being carried out (i.e. either CBT or SAT). All tapes were accurately identified. Treatment was carried out during a 6-month therapy envelope. Participants were assessed by masked researchers at baseline 1, baseline 2, 6- and 12-month follow-up and were assessed on the outcomes described above.

Statistical analysis

Data were analysed using SPSS for Windows version 14. All analyses were conducted on an intention-to-treat basis. Chi-squared analysis was used to assess differences between groups on occurrence of violence and change in risk management. Non-parametric analysis was carried out on violence frequency data and staff-rated anger. Group differences on other quantitative outcomes were examined in a repeated measures analysis using the SPSS mixed model REML procedure (SPSS version 12), contrasting baseline and baseline 2 scores with end of treatment (time 1) and follow-up (time 2) scores with secondary outcome variables of symptoms, functioning and anger. The significance tests of interest were for the end of treatment (time 1)× treatment group and follow-up (time 2)×treatment group interactions. Where scores from assessment measures deviated significantly from a normal distribution, log transformed scores were attempted and where distributions remained skewed or there was significant kurtosis, non-parametric statistics were employed.

Therapy participation

Of the 77 participants, 3 decided not to take part in therapy (2 CBT, 1 SAT) and 1 was transferred to a facility out of region making visits not possible. One person died of natural causes before the end of treatment and is not included further in the analysis. Of the rest, the mean number of sessions attended was 17.0 (s.d.=6.8) sessions for CBT and 17.4 (s.d.=6.6) sessions for SAT. The mean number of minutes of therapy received was 787.7 (s.d.=356.2) for CBT and 894.3 (s.d.=413.3) for SAT. There were no significant differences between the groups on the number of sessions (F(1,75)=0.89, P=0.77) or minutes of therapy received (F(1,75)=1.45, P=0.23). Of the 73 participants who took part in therapy, all but 8 received at least 10 sessions of therapy. The eight (four from CBT, four from SAT) received between three and nine sessions. Overall, 38 people received CBT and 39 received SAT. There was little difference between the groups in terms of the age of the participants (mean age: CBT=35.7 years, s.d.=12.5; SAT=33.9 years, s.d.=9.7).

Primary outcome: case-note identification of aggression and violence

There was little difference between the treatment groups pre-trial entry (3 months prior to baseline 1) or between baseline 1 and 2 (pre-treatment intervention) on numbers of people who were aggressive or violent or on total incidents of aggression and violence.

Overall over the treatment period, there were no significant differences between the groups on the number of people who had no incidents of violence or aggression v. the number who had one or more incidents (χ²=2.351, d.f.=1, P=0.125). However, the CBT group had a significantly lower number of incidents during the treatment period (P=561.0, P=0.039). Over the follow-up period, there was no significant difference between the number of people in each group that carried out aggressive or violent acts (χ²=0.160, d.f.=1, P=0.689) or on total number of incidents (Mann–Whitney U=697.0, P=0.594). The total number of incidents of verbal or physical aggression recorded from case notes over the intervention and follow-up period by group is shown in Table 2.

Table 2

Total number of incidents for each group by case note identification over the treatment and follow-up period

Incidents, n	CBT (n=38)	SAT (n=39)
During treatment
Verbal aggression	31	103
Physical aggression	2	46
Follow-up
Verbal aggression	34	80
Physical aggression	5	22

CBT, cognitive–behavioural therapy; SAT, social activity therapy

There were no significant differences between the groups over the treatment period on the numbers of people who engaged in verbally aggressive acts (χ²=1.597, d.f.=1, P=0.256) or on the number of incidents of verbal aggression (U=611.0, P=0.150). This was similar for the follow-up period on number of people (χ²=0.687, d.f.=1, P=0.490) and number of incidents (U=555.5, P=0.765).

Over the treatment period, there were no significant differences between the groups on the number of people who were physically aggressive (χ²=2.160, d.f.=1, P=0.142) or the total number of incidents involving physical aggression (U=653.5, P=0.172). However, over the follow-up period, there were significantly more people who had been physically aggressive in the SAT group compared with the CBT group (χ²=8.081, d.f.=1, P=0.004) and there was a significantly lower number of physically aggressive incidents in the CBT group (U=466.0, P=0.028).

Secondary outcomes

Limitations and strengths

There are a number of limitations that should be considered when interpreting the results from this study. First, the sample size was relatively small which may have limited the power of the study to detect treatment differences. Estimates of power based on a review of previous trials of CBT in schizophrenia recommend at least 70 in each group to show differences between groups. ^{Reference Jones, Cormac, Silveira da Mota Neto and Campbell1} Similarly, the multiple testing in relation to the secondary outcomes may have increased the likelihood of the positive findings happening as a result of chance, particularly in relation to the positive psychotic symptom outcome. Furthermore, the sample was quite heterogeneous with a mix of in-patients, out-patients and those residing in high-security environments. It is possible that the interventions may have had different impacts in different environments. However, an exploratory analysis of the in-patient participants in the sample on violence, symptoms and anger showed no differences to the pattern of findings found for the whole sample. In addition, this study failed to address issues relating to personality in relation to outcome or in the treatment protocol. This may be necessary in future studies as personality factors may be important predictors of violence for some individuals and may be influential in the degree to which an individual engages with therapy. ^{Reference Hodgkins, Hiscoke and Freese23}

Despite these limitations, strengths of the study included:

1. (a) a robust treatment protocol and careful attention to therapist training issues, supervision and treatment fidelity;

2. (b) masked, independent assessors of outcome who were trained to a high standard of reliability and validity on the outcome measures;

3. (c) an intervention that was applied in routine settings by therapists who worked within the institutions;

4. (d) a follow-up period to examine maintenance of any gains; and

5. (e) outcomes that focused not just on symptoms but on issues to do with anger, functioning, violence and risk.

These are important outcomes particularly for those people residing in in-patient or secure environments where risk and prediction of future risk are especially pertinent. The positive findings regarding CBT are promising and suggest optimism for further exploration of psychological treatments for people with psychosis and problems with aggression and violence. However, some further modifications and fine tuning of the treatment may be necessary. In the light of these preliminary findings, further evaluation of the approach is warranted.