A meta-analysis comparing cognitive function between individuals at clinical high-risk for psychosis and individuals at family high-risk for psychosis

Abstract

Background: Cognitive impairment is detectable before psychosis onset, yet no quantitative synthesis has directly contrasted neurocognition between clinical high risk (CHR) and familial high risk (FHR). We aimed to clarify domain-level differences to inform assessment and early intervention. Methods: Following PRISMA/MOOSE, we searched PubMed, Embase, and PsycINFO (1996–May 2025) for peer-reviewed studies reporting standardized neurocognitive outcomes in both CHR and FHR groups. Outcomes were organized by MATRICS domains. Effect sizes were Hedges’ g (coded CHR − FHR; negative = poorer CHR) synthesized with random-effects. When multiple outcomes occurred within a study/domain, a within-study fixed-effects composite (r = 0.50) was used. Heterogeneity (Q, I²), small-study effects (Egger’s test), and prespecified meta-regression (when k ≥ 10) were performed. Results: Fourteen studies were included (CHR n = 1,160; FHR n = 813). Processing speed: CHR underperformed FHR (pooled g = − 0.290, 95% CI: −0.521 to − 0.059, p = 0.014;); task-level analyses showed significant effects for TMT-A and Stroop baselines. Attention, working memory, executive function, and visual learning showed no robust between-group differences at the domain level (all p > 0.10; moderate–high heterogeneity). Verbal learning: subgroups indicated CHR deficits in list-learning—immediate (g = − 0.558, 95% CI − 1.099 to − 0.017, p = 0.043) and list-learning—delayed (g = − 0.296, 95% CI − 0.526 to − 0.067, p = 0.011). Wechsler Memory Scale (WMS) Logical Memory (immediate/delayed) was under-represented (k < 3). Meta-regression (processing speed): region moderated effects (Non-Asian vs. Asian: β = 0.486, 95% CI 0.051–0.921, p = 0.0287; R²_analog ≈ 24%); age difference, publication year, and study quality were not significant. Egger’s tests did not indicate small-study effects (all p > 0.05). Conclusions: Compared with FHR, CHR shows reliable impairments in processing speed and verbal list-learning, while other domains do not differ robustly. Findings refine the cognitive phenotype of CHR beyond familial liability and highlight processing speed and list-learning as pragmatic markers for risk stratification and monitoring. Clinical trial number: not applicable.