چه کسی می ماند، چه کسی سود می برد؟ پیش بینی افت و تغییر در رفتاردرمانی شناختی برای روان پریشی

This study investigates the predictors of outcome in a secondary analysis of dropout and completer data from a randomized controlled effectiveness trial comparing CBTp to a wait-list group (Lincoln et al., 2012). Eighty patients with DSM-IV psychotic disorders seeking outpatient treatment were included. Predictors were assessed at baseline. Symptom outcome was assessed at post-treatment and at 1-year follow-up. The predictor×group interactions indicate that a longer duration of disorder predicted less improvement in negative symptoms in the CBTp but not in the wait-list group whereas jumping-to-conclusions was associated with poorer outcome only in the wait-list group. There were no CBTp specific predictors of improvement in positive symptoms. However, in the combined sample (immediate CBTp+the delayed CBTp group) baseline variables predicted significant amounts of positive and negative symptom variance at post-therapy and 1-year follow-up after controlling for pre-treatment symptoms. Lack of insight and low social functioning were the main predictors of drop-out, contributing to a prediction accuracy of 87%. The findings indicate that higher baseline symptom severity, poorer functioning, neurocognitive deficits, reasoning biases and comorbidity pose no barrier to improvement during CBTp. However, in line with previous predictor-research, the findings imply that patients need to receive treatment earlier.

Cognitive behavioral therapy for psychosis (CBTp) has been demonstrated to be effective for psychotic disorders (Wykes et al., 2008) and has been incorporated into several national guidelines (Gaebel et al., 2009 and NICE, 2009). Nevertheless, a number of patients discontinue therapy (on average 16% according to a meta-analysis by Lincoln et al., 2008) and among those who continue, approximately half do not show reliable symptom improvement (Jones et al., 2004 and Wykes et al., 2008). Knowing who is likely to benefit from CBTp would provide a better basis for an evidence-based allocation of patients to treatment. Furthermore, knowing about who is unlikely to benefit helps us to understand where CBTp needs to be adapted in order to serve specific groups more effectively. Several studies have attempted to identify baseline predictors of improvement in CBTp. In regard to socio-demographic variables, these studies have found that younger patients benefit more in terms of positive symptoms (Thomas et al., 2011 and Morrison et al., 2012) and that women benefit more than men in overall psychopathology (Drury et al., 1996 and Brabban et al., 2009). Furthermore, higher level of education was shown to predict better outcome in negative symptoms (Allott et al., 2011). A clinical baseline variable relevant to outcome is a shorter duration of treated or untreated psychosis, which has been found to be associated with a shorter recovery time (Drury et al., 1996), greater symptom improvement during CBTp (Tarrier et al., 1998, Thomas et al., 2011 and Morrison et al., 2012), and less symptomatology at post-assessment (Morrison et al., 2004). Also, lower baseline symptomatology overall was shown to be related to more symptomatic improvement during CBTp (Tarrier et al., 1998), in particular less pronounced negative symptoms were related to greater symptom improvement (Thomas et al., 2011) and outcome (Allott et al., 2011). In contrast, there is some indication that more severe positive symptoms were a positive predictor of symptom improvement (Morrison et al., 2004 and Dunn et al., 2006). No study found baseline depression to be related to outcome. Higher insight into the disorder predicted overall symptom improvement in two studies (Garety et al., 1997 and Naem et al., 2008). Interestingly, Garety et al. (1997) also found that among patients with delusions acknowledging the possibility of being mistaken was a predictor of better outcome, although this was strongly associated with insight. Similarly, Brabban et al. (2009) found lower delusion conviction to be associated with overall symptom reduction in a subgroup of patients with delusions who had received CBTp. On a similar note, cognitive insight, in terms of self-reflectiveness and self-certainty was found to be predictive of favorable outcome (Perivoliotis et al., 2010 and Premkumar et al., 2011). Furthermore, higher baseline occupational functioning predicted lower levels of positive symptoms at 1 year follow-up (Allott et al., 2011). With regard to neurocognitive variables, Penades et al. (2010) found better baseline memory performance to predict symptom improvement following treatment. However, most studies (Garety et al., 1997, DeVille et al., 2011 and Premkumar et al., 2011) failed to find predictive value of memory, executive functioning, attention, or verbal fluency on outcome of CBTp. One problem in drawing valid conclusions from the previous research is that studies have focused on different domains and time-points of outcome. Moreover, most studies are inconsistent in whether they investigate unspecific predictors of change or those specific to CBTp or even merely predict symptom levels at post-therapy without controlling for baseline symptoms. Nevertheless, previous findings highlight the positive predictive value of a shorter duration of psychosis and better insight on outcome. They also indicate that more pronounced negative symptoms at baseline are associated with less favorable outcome, whereas more severe baseline positive symptoms seem to be positively related to symptom improvement. The majority of studies do not find neurocognitive functioning to be a predictor of outcome, while there are singular findings indicating that patients with higher education, younger age, and female gender might benefit more from CBTp. Surprisingly, some predictors that are likely to be specifically relevant to CBTp have not received sufficient attention. Psychotic symptomatology is associated with a range of reasoning biases, such as jumping-to-conclusions, difficulties in theory of mind and attribution biases and, consequently, CBTp has a strong focus on increasing peoples׳ ability to question their beliefs and to take more time to weigh the evidence before drawing conclusions (Kuipers et al., 2006). This also involves learning to take peoples׳ cognitive and emotional perspective. On a transdiagnostic level, there is some indication that people with stronger cognitive resources (in the sense of fewer dysfunctional attitudes) benefit more from cognitive approaches (e.g. Sotsky et al., 1991). Garety et al.׳s (1997) finding that less pronounced delusion–conviction and cognitive flexibility were associated with better outcome seems to support this for patients with psychosis. It would therefore be interesting to test whether lower levels of reasoning biases predict better outcome. Second, psychosis generally goes along with a range of comorbid disorders, in particular anxiety disorders and depression (Fenton, 2001). In clinical practice, CBTp also targets these disorders. Due to the high efficacy of cognitive behavioral interventions for anxiety disorders and depression (Butler et al., 2006), patients with comorbid Axis I disorders might benefit more from therapy than those for whom the sole focus lies on psychotic symptoms. In contrast, Axis II disorders are likely to complicate and prolong the therapy and have been found to be a negative predictor of outcome in treatment of depression and anxiety (Reich, 2003). With regard to outcome, most of the studies have focused on global symptomatology, positive symptoms (as the prime target of CBTp) or functioning. To our knowledge, no study has attempted to predict improvement in negative symptoms although it is agreed that negative symptoms constitute a distinct and important therapeutic domain (Kirkpatrick et al., 2006). Finally, although predictors of dropout related to psychosocial treatments for schizophrenia in general have been investigated, finding age, gender, duration of disorder and treatment-related variables to be associated with dropout (Villeneuve et al., 2010), only one study, by Perivoliotis et al. (2010), has focused specifically on drop-out during CBTp. The aim of this study is therefore to extend the research on baseline predictors of short- and long-term improvement in positive and negative symptoms and dropout in a large and clinically heterogeneous sample of patients who received CBTp. The study is a secondary analysis of dropout and completer data from a randomized controlled effectiveness trial of CBTp (Lincoln et al., 2012) that found significant improvement in positive symptoms and overall psychopathology but not in negative symptoms in the CBTp group compared to a waitlist group. Over and above the predictors investigated in previous studies, we will analyze the impact of social cognition and reasoning which we expect to have unique relevance to CBTp, as well as the impact of comorbidity.

3. Results 3.1. Sample characteristics and therapy delivery Fifty-nine patients fulfilled DSM-IV-criteria for schizophrenia, 12 for schizo-affective disorder, five for delusional disorder and four for acute psychotic disorder. Comorbid Axis I disorder was diagnosed in 45 patients, comorbid Axis II disorder in 15 patients. Sixteen percent of the sample fulfilled criteria for two or more comorbid disorders. Among the Axis I disorders, anxiety disorders were the most prevalent (n=25), followed by affective disorders (n=18) and substance abuse or dependency (n=12). Among the personality disorders, borderline personality disorder was diagnosed as the most frequent (n=4) followed by avoidant personality disorder (n=3). Eighteen patients were acutely psychotic, 56 were partly remitted and/or episodic, and six patients were classified as fully remitted (DSM-IV single episode, full remission). The mean age of the sample was 33.1 (S.D.=10.6) and 35 patients were female. The mean duration of psychosis was 10.4 years (S.D.=8.5). The mean number of previous episodes requiring hospital admission was 4.5 (S.D.=6.9). All but four patients were on antipsychotic medication and the majority was reliably taking them as prescribed (n=48). The mean GAF score was 45.3 (S.D.=12.8), the mean scores for the PANSS positive, negative and general subscales were 14.9 (S.D.=4.5), 14.2 (S.D.=4.6) and 34.0 (S.D.=7.5). Patients who completed therapy received 28.9 (S.D.=7.4) therapy sessions in 38.0 (S.D.=15.8) weeks. The average waiting time was 19.2 weeks (S.D.=7.9). The mean follow-up period was 53.3 weeks (S.D.=40.2). Across both groups, patients received additional 7.6 sessions (S.D.=10.7) between post-therapy and 1-year follow-up. 3.2. Differential predictors of improvement in positive and negative symptoms by post-assessment between the CBTp group versus the WL group PANSS positive symptoms at post-assessment were significantly related to PANSS positive symptoms at baseline (β=0.58, p<0.01), as was group status (β=0.28, p<0.01), indicating the significant effect of the treatment on positive symptoms. As can be seen in Table 2 there were no significant interaction effects, indicating that none of the 17 variables investigated predicted improvement in one group more relative to the other. However, there were two trend effects towards significant group×predictor interactions in regard to cognitive set shifting and memory. Post-hoc analyses by group indicated that in the WL group impaired set-shifting (β=0.17, p=0.14) and impaired memory performance (β=−0.18, p=0.13) tended to predict higher levels of positive symptoms at outcome while in the CBTp group these variables tended to predict lower levels of positive symptoms at outcome (β=−0.11, p=0.56 and β=0.18, p=0.35, respectively). Table 2. Linear Regressions of the Group x Predictor Interaction on Positive and Negative Symptoms Controlling for Baseline Symptoms. Predictor Negative Symptoms Positive Symptoms β p β p Socio-demographic variables Female gender 0.017 0.856 −0.132 0.187 Age 0.118 0.906 0.049 0.629 Years of education −0.011 0.901 0.131 0.194 Clinical variables Duration of disorder 0.050 0.614 0.251 0.015 Comorbid Axis I disorder 0.010 0.914 0.002 0.981 Comorbid Axis II disorder 0.043 0.654 −0.192 0.052 Symptom severity Negative/positive symptoms (PANSS)a −0.093 0.289 0.019 0.859 Delusion conviction (PDI) 0.019 0.845 0.072 0.481 Lack of Insight (PANSS G-12) −0.093 0.326 −0.161 0.108 Depression (CDSS) 0.057 0.554 −0.129 0.176 Social functioning Role Functioning Scale 0.069 0.469 0.002 0.988 Social skills −0.063 0.523 −0.023 0.821 Neurocognitive variables Cognitive flexibility (TMT-B) −0.184 0.075 0.101 0.363 Memory (WMS) 0.187 0.061 −0.129 0.191 Reasoning biases Jumping-to-conclusions −0.054 0.580 0.217 0.039 Internal attributions of neg. events 0.036 0.731 −0.186 0.110 Theory of mind 0.088 0.341 0.087 0.375 Note: Baseline symptoms were entered in block one (with β=0.59, p<0.001 for positive symptoms and β=0.56, p<0.001 for negative symptoms). Group status was entered in block two along with the predictor and the group×predictor interaction (with β׳s ranging from 0.27 to 0.36, all p׳s<0.01, for positive symptoms and from −0.04 to −0.12, all p׳s>0.20 for negative symptoms). a Baseline negative symptoms were entered as a predictor of positive symptoms at outcome and baseline positive symptoms were entered as a predictor of negative symptoms at outcome. Table options PANSS negative symptoms at post-assessment were significantly associated with PANSS negative scores at baseline (β=0.56, p<0.01) while group status was not (β=−0.06, p=0.58) indicating the absence of a therapy effect on negative symptoms. As can be seen in Table 2 only two of the 17 group×predictor interactions were significant. This was the case for duration of disorder and for JTC, with post-hoc analyses by group indicating that a longer duration of disorder tended to predict higher levels of negative symptoms in the CBTp (β=0.28, p=0.07) and lower levels in the WL group (β=−0.22, p=0.10) and that a stronger JTC-bias tended to predict negative symptoms in the WL group (β=−0.23, p=0.09) but not in the CBTp group (β=0.20, p=0.20). Also, there was a trend interaction for comorbid Axis II disorders with post-hoc analyses by group indicating that comorbidity predicted more negative symptoms in the WL (β=0.33, p<0.01) but not in the CBTp group (β=0.04, p=0.82). 3.3. Predictors of improvement from pre-treatment to post-treatment and follow-up in the complete sample PANSS positive symptoms at post-therapy were significantly related to PANSS positive symptoms at baseline (β=0.33, p<0.01), but not to cohort (β=0.15, p=0.22) indicating that there was no difference in outcome between the immediate (previously the CBTp group) and the delayed therapy groups (previously the WL group). Over and beyond these variables, higher levels of depression (β=0.40, p<0.01), more negative symptoms (β=0.51, p<0.01), poorer social skills (β=−0.39, p<0.01), poorer role functioning (β=−0.48, p<0.01) and poorer ToM ability (β=−0.25, p<0.05) significantly predicted positive symptoms at post-treatment. Together, these five variables accounted for a significant amount of variance in post-therapy positive symptoms over and above the variance explained by pre-therapy positive symptoms (change in R2=0.40, d.f.=4,53; p<0.01). PANSS positive symptoms at follow-up were not related to PANSS positive symptoms at baseline (β=0.06, p=0.62) or to cohort (β=−0.01, p=0.93). Over and beyond these variables, the predictors age (β=0.28, p<0.05), years of education (β=0.32, p<0.05), depression (β=0.28, p<0.05), negative symptoms (β=0.31, p<0.05) and externalizing bias (β=0.36, p<0.01), were significantly related to positive symptoms at follow-up. Together, these five variables accounted for a significant amount of variance in follow-up positive symptoms over and above the variance explained by pre-therapy positive symptoms (change in R2=0.30; d.f.=5,48; p<0.01). PANSS negative symptoms at post-assessment were associated with PANSS negative symptoms at baseline (β=0.32, p<0.01) while cohort was not (β=0.13, p=0.31). Over and beyond these variables, lower functioning (β=−0.30, p<0.05), higher delusion conviction (β=0.46, p<0.01) and more positive symptoms (β=0.55, p<0.01) were significantly related to negative symptoms at outcome. Together, these two variables accounted for a significant amount of variance in follow-up negative symptoms over and above the variance explained by pre-therapy negative symptoms (change in R2=0.33; d.f.=3,63; p<0.01). Negative symptoms at follow-up were associated with PANSS negative symptoms before treatment (β=0.27, p<0.05) while cohort was not (β=−0.13, p=0.29). Over and beyond these variables, the predictors age (β=0.25, p<0.05) and years of education (β=0.29, p<0.05) were significantly related to negative symptoms at follow-up. Together, these two variables accounted for a significant amount of variance in follow-up negative symptoms over and above the variance explained by pre-therapy negative symptoms (change in R2=0.14; d.f.=2,64; p<0.01). 3.4. Predictors of dropout Patients who dropped out (n=12) had been hospitalized less often (2.8, S.D.=1.7) than completers (4.8, S.D.=7.5; t(74.3)=2.1, p<0.05), had more lack of insight (2.8, S.D.=1.1 versus 1.8, S.D.=1.0; t (77)=2.0, p<0.01), lower social functioning (4.5, S.D.=3.2 versus 7.0, S.D.=2.9; t(78)=−2.7, p<0.01) and more negative symptoms (16.2, S.D.=2.8 versus 14.1, S.D.=4.7; t(24.3)=2.1, p<0.05). They also showed a trend towards more positive symptoms (17.4, S.D.=5.9 versus 13.8, S.D.=3.8; t(12.7)=1.1, p=0.06) and less Axis II disorders (O versus 15, χ2=3.3, p=0.07). None of the other predictors reached significance. The results of logistic regression of the four significant predictors on dropout are presented in Table 3. As can be seen, the full model was statistically significant in predicting variance in dropout. However, no variable reached significance as a single predictor within the model. Furthermore, the baseline model already predicted dropout accurately in 85% based on the assumption that no patient would drop out. The regression model predicted seven of the dropouts correctly. However, it also incorrectly predicted three completers to be dropouts resulting in a total prediction accuracy of 87%. Table 3. Results of the logistic regression of predictors on dropout versus non-dropout. Included B(SE) 95% CI for odds ratio Lower Odds ratio Upper Constant −0.68 (2.1) No of previous hospitalizations −0.09 (0.09) 0.78 0.92 1.1 Negative symptoms −0.04 (0.10) 0.79 0.96 1.2 Insight 0.57 (0.32) 0.96 1.8 3.4 Social functioning −0.38 (0.16) 0.66 0.78 1.0 Note: R2=0.14 (Cox and Snell), 0.25 (Nagelkerke). Model χ2(4)=12.2, p=0.016.