Metacognition itself is broadly defined as the capacity to monitor and regulate one’s own cognitive processes, and it plays a central role in the pathogenesis and maintenance of psychotic symptoms. Within this framework, cognitive insight is operationalized as the ability to reflect upon and critically evaluate one’s own beliefs and interpretations, and to remain open to corrective feedback. The Beck Cognitive Insight Scale (BCIS) is the most widely used instrument to assess cognitive insight in participants with schizophrenia spectrum disorders, comprising two key subscales: Self-Reflectiveness (SR), which measures the tendency to acknowledge the fallibility of one’s beliefs and consider alternative interpretations, and Self-Certainty (SC), which reflects the degree of certainty in one’s judgments and resistance to correction. A composite index (CI) is derived from these subscales, providing an overall measure of cognitive insight, with higher values indicating better insight. Clinically, this outcome is highly significant, as lower SR has been linked to greater symptom severity four years later, and metacognitive functioning generally is associated with poorer functional outcomes in schizophrenia spectrum disorders.

The current research addressed the persistent inconsistency in MCT efficacy by investigating whether baseline cognitive insight moderates treatment response. Previous systematic reviews and meta-analyses have reported mixed results for MCT on cognitive insight dimensions, often citing low-to-medium effect sizes, but these estimates were frequently interpreted cautiously due to methodological limitations, including the pooling of cognitive and clinical insight measures across studies of varying designs. Furthermore, some specific meta-analyses reported small-to-medium effects on SR and SC, alongside a large effect on the CI, yet these findings were complicated by substantial heterogeneity. The researchers hypothesized that baseline cognitive insight would moderate treatment response, predicting that participants with the most impaired cognitive insight at baseline would derive the greatest benefit from MCT. A critical element of this hypothesis was the prediction that a non-negligible proportion of participants would already score in the high cognitive insight category at baseline, suggesting that a ceiling effect might be partially obscuring treatment effects at the group level in previous trials.

To test this hypothesis, the study utilized harmonized individual participant data from the PERMEPSY database, which integrated retrospective data from 13 international studies involving 614 participants diagnosed with schizophrenia spectrum disorders. This approach was chosen specifically because it allows for continuous moderation analyses, preserving the full distribution of baseline scores, thereby overcoming the limitations of aggregate-level meta-analytic methods in detecting interaction effects. Cognitive insight was assessed using the BCIS, which includes the SR, SC, and the CI derived from the arithmetic difference between SR and SC. The statistical analysis employed ANCOVA models, incorporating a continuous Group × Baseline interaction term, complemented by stratified ANOVAs to visualize the interaction patterns and estimate within-stratum effect sizes. This rigorous methodology, which included sensitivity analyses using multilevel linear mixed models to account for potential non-independence of participants within the contributing studies, ensured the robustness of the findings.

The results of the analyses provided compelling evidence for the moderating role of baseline cognitive insight. ANCOVA models revealed significant Group × Baseline interactions across all three BCIS dimensions, indicating that baseline cognitive insight significantly moderated treatment response to MCT. Furthermore, stratified ANOVAs yielded convergent results for both SR and SC. The effect sizes observed followed a consistent and clear gradient: larger improvements in cognitive insight were consistently observed among participants who presented with lower levels of cognitive insight at baseline. Notably, the study found that 26.5% of the participants already scored in the high-insight category on the CI at baseline. This finding strongly suggests that ceiling effects may have partially obscured the true treatment effects of MCT when assessed at the group level in previous trials, which is a crucial methodological observation.

The consistent gradient pattern was further detailed through stratified analyses. For SR and CI, where higher scores reflect greater insight improvement, the largest significant effects were found in the tertile 1 group (those with the lowest baseline insight), while tertiles 2 and 3 showed negligible or non-significant effects. For SC, where more negative values indicate greater reduction in overconfidence, effect sizes became increasingly negative from tertile 1 to tertile 3, with significant effects appearing in tertiles 2 and 3. When age was included as an additional covariate in the secondary ANCOVA models, the Group × Baseline interaction remained significant across all three BCIS scores, and age did not emerge as a significant predictor in any of the models, confirming that the observed moderation was not merely a function of age differences between the groups. A sensitivity analysis using multilevel linear mixed models also confirmed that the moderating effect of baseline cognitive insight was robust to accounting for between-study variability.

These findings suggest that MCT does appear to improve cognitive insight, but its efficacy is highly dependent on the participant’s starting point. The evidence strongly indicates that MCT is most beneficial for participants with greater baseline impairment in their metacognitive functioning. This suggests that baseline heterogeneity is a key determinant of MCT efficacy and provides a potential explanation for the previously inconsistent results reported in the literature. The study underscores the importance of moving beyond aggregate-level analyses and incorporating individual participant data to accurately capture complex interaction effects in psychological interventions for psychosis.

The study carries significant implications for both clinical practice and future research. Clinically, the gradient of efficacy suggests that participants presenting with the most pronounced metacognitive difficulties—those with low self-reflectiveness and high self-certainty—may derive the greatest benefit from MCT when improving metacognitive functioning is a therapeutic goal. However, the authors caution that using baseline cognitive insight to identify participants should be embedded within a recovery-oriented approach, mindful of the potential for stigma. From a research perspective, these results advocate for the systematic inclusion of baseline cognitive insight measures and the consideration of minimum impairment thresholds when cognitive insight is a primary outcome in future MCT trials. While the absolute magnitude of the between-group differences in BCIS scores was modest, the consistent moderation pattern provides a clear direction for refining the application and interpretation of MCT in the treatment of psychosis.

Subject of Research: The study demonstrates that an individual's baseline level of cognitive insight significantly moderates the treatment response to Metacognitive Training (MCT) in patients with schizophrenia spectrum disorders.

Article Title: Baseline cognitive insight moderates the efficacy of metacognitive training for psychosis: an individual participant data analysis

Article References: Goncalves, A., Ochoa, S., Moritz, S., Gawęda, Ł., Acuña, V., König, C., Lamarca, M., & Berna, F. (2026). Baseline cognitive insight moderates the efficacy of metacognitive training for psychosis: an individual participant data analysis. Frontiers in Psychology, 17, 1934181. https://doi.org/10.3389/fpsyg.2026.1934181

Image Credits: AI Generated

DOI: 10.3389/fpsyg.2026.1934181

Keywords: Metacognitive Training, Psychosis, Cognitive Insight, Schizophrenia, Individual Participant Data, Treatment Efficacy, Metacognition, Clinical Trials, BCIS, Psychiatry, Baseline, cognitive