Lack of robust meta‐analytic evidence to favour cognitive behavioural therapy for prevention of psychosis

Abstract

While achievements in detection and prognostic assessment of young people at clinical high risk for psychosis (CHR-P) have been recently consolidated, the efficacy of preventive interventions remains unclear1. Cognitive behavioural therapy (CBT) is the currently recommended preventive intervention, but the most updated network meta-analysis2 found no robust evidence to favour it (and any of the other indicated interventions) compared to the control condition (i.e., needs-based interventions). A subsequent independent pairwise meta-analysis by the Cochrane group3 confirmed these findings, concluding that there was "no convincing unbiased, high-quality evidence" that any type of intervention is more effective than needs-based interventions (another pairwise meta-analysis was subsequently published4, but used older data). A further umbrella review showed no evidence that CBT impacts other clinical outcomes such as acceptability of treatments, severity of attenuated positive/negative psychotic symptoms, depression, symptom-related distress, social functioning, general functioning, and quality of life5. These studies highlighted that uncertainty of evidence is high and that caution is required in recommending CBT for the prevention of psychosis in CHR-P individuals. In contrast with these cautionary warnings, a recent pairwise meta-analysis6 concluded that "robust and sound evidence supports cognitive behavioural therapy in reducing transition" to psychosis and in decreasing the severity of attenuated psychotic symptoms. First, no new large-scale randomized controlled trials of CBT have been published since the previous network/Cochrane meta-analyses2, 3, which could justify different conclusions. Only a small, single-site trial (N=58) of CBT has been published meanwhile7. This trial has several weaknesses relating to the measurement of outcomes, incorrect interpretation of Kaplan-Meier outputs, selective reporting, and failure to adhere to CONSORT guidance (e.g., failure to report trial registration)8. Using the Clinical Trials Assessment Measure, the recent meta-analysis6 assigned to this CBT trial the highest methodological quality (97/100) of all ran­domized controlled trials ever conducted in CHR-P individuals. It is difficult to understand how a trial that was never registered, with inaccuracies in psychometric classification and basic mistakes in statistical reporting rates so highly, casting doubts on the validity of the quality assessment conducted in the meta-analysis6. Second, while the protocol of this recent meta-analysis6 stated that unpublished literature was considered for inclusion, the authors did exclude the large CBT PREVENT trial (N=216), although its preliminary findings – showing no statistical significant effect of CBT in preventing psychosis – were presented at a major international conference and included in the previous network meta-analysis3. The fact that a large CBT trial has been excluded means that the findings of the new meta-analysis6 may be affected by publication bias. Indeed, the authors of the meta-analysis acknowledged that only one missing trial would be needed to render their end-of-treatment results non-significant6. To empirically test this, we have updated that meta-analysis by removing the low-quality small trial7 and adding the large PREVENT trial. The updated risk ratio for CBT vs. control interventions to prevent transition to psychosis at 12 months was 0.631 (95% CI: 0.388-1.028, p=0.064), which shows no significant meta-analytic evidence that CBT can robustly prevent transition to psychosis. Third, the authors' conclusion that CBT can robustly improve attenuated psychotic symptoms conflicts with the very small effect size, approaching the non-significance level (standardized mean difference = –0.15; 95% CI: –0.28 to –0.01)6, which is unlikely to be associated with clinically meaningful benefits in the real-world. Finally, the meta-analysis in question may be affected by reporting biases, which increased the likelihood of the results being significant in favour of CBT. For example, additional transitions to psychosis beyond those originally reported were included as "the most accurate data on transition rates"6. These data have never been acknowledged as primary outcomes in the original publications, and operationalization of primary outcomes is not clearly specified a priori in the corresponding meta-analytic protocol. Based on the considerations above, we conclude that the lack of robust meta-analytic evidence to favour CBT to prevent psychosis, as appraised by the most recent network meta-analysis2 and the Cochrane meta-analysis3, still stands. These meta-analyses, which emphasized methodological biases and the inconsistency of the current evidence, may have caused disappointment and frustration and the production of some over-optimistic literature favouring CBT. It has been claimed that unfavourable meta-analytic evidence needs to be contextualized, because pr