BACKGROUND: Early detection of schizophrenia risk is a critical goal in the field. Endophenotypes in children to relatives of affected individuals may contribute to this early detection. One of the lowest cost and longest theorized domains is motor development in children.

METHODS: A meta-analysis was conducted comparing individuals ≤21 years old with affected first-degree relatives (FDR) with (1) individuals from unaffected families (controls), or (2) individuals with FDR having other psychiatric disorders. Studies were classified by motor outcome and separate meta-analyses were performed across six correlated domains, with available N varying by domain.

RESULTS: Inclusion criteria were met by k = 23 independent studies with a total N = 18,582, and N across domains varying from 167 to 8619. The youth from affected families had delays in gross and fine motor development in infancy (k = 3, n = 167, Hedges'g = 0.644, confidence intervals (CI) = [0.328, 0.960], p < .001), walking milestones (k = 3, n = 608, g = 0.444, CI = [0.108, 0.780], p = .01), coordination (k = 8, n = 8619, g = 0.625, CI = [0.453, 0.797], p < .0001), and had more abnormal movements such as involuntary movements (k = 6, n = 8365, g = 0.291, CI = [0.041, 0.542], p = .02) compared with controls. However, not all effects survived correction for publication bias. Effects for neurological soft signs were small and not reliably different from zero (k = 4, n = 548, g = 0.238, CI = [-0.106, 0.583], p = .18). When comparing the FDR group to youth from families with other psychiatric disorders, the FDR group was distinguished by poorer gross and fine motor skills (k = 2, n = 275, g = 0.847, CI = [0.393, 1.300], p < .001).

CONCLUSIONS: Motor deficits during development likely represent an endophenotype for schizophrenia, although its specificity is limited in relation to other serious mental disorders. It holds promise as a low cost domain for early risk detection, although it will have to be combined with other indicators to achieve clinically usable prediction accuracy. Impaired coordination was the most robust result with a moderate effect size and lack of heterogeneity and publication bias.