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

Background Metacognitive Training (MCT) for psychosis is designed to target metacognitive processes, yet evidence for its effects on cognitive insight remains inconclusive, owing primarily to the limited methodological quality of available trials. A potential explanation lies in baseline heterogeneity, with a substantial proportion of participants presenting high baseline cognitive insight, thereby limiting observable change. Methods This study used harmonized individual participant data from the PERMEPSY database, including 614 participants with schizophrenia spectrum disorders from 13 international studies. Cognitive insight was assessed using the Beck Cognitive Insight Scale (BCIS), including Self-Reflectiveness (SR), Self-Certainty (SC), and a composite index (CI). Baseline moderation effects were examined using ANCOVA models with a continuous Group × Baseline interaction term, complemented by stratified ANOVAs to visualize the interaction pattern and estimate within-stratum effect sizes. Results ANCOVA models revealed significant Group × Baseline interactions across all BCIS dimensions, indicating that baseline cognitive insight moderated treatment response. Stratified ANOVAs yielded convergent results for SR and SC. Effect sizes followed a consistent gradient, with larger improvements observed among participants with lower baseline insight. Notably, 26.5% of participants already scored in the high-insight category at baseline, suggesting that ceiling effects may have partially obscured treatment effects at the group level in previous trials. Conclusion MCT appears to improve cognitive insight, particularly among participants with greater baseline impairment. These findings suggest that baseline heterogeneity is a key determinant of MCT efficacy and may account for previously inconsistent results.