Normative modeling in neuroimaging across psychiatric and neurological disorders: a systematic review and the NORMA reporting checklist

Abstract Case-control comparisons of group means in psychiatric and neurological neuroimaging cannot quantify individual deviation from population norms. Normative modeling, akin to pediatric growth charts, locates individuals within a reference distribution to yield individualized deviation scores. We reviewed 122 normative-modeling studies across neuroimaging and neurophysiological measures. Structural MRI accounted for 71% of studies. Statistical methodology has shifted from Gaussian process regression toward Bayesian and distributional approaches. Reference cohort sizes spanned 51–77,157 (median 642). The 15 cohorts of 10,000 or more drew most often on UK Biobank (73%), whether directly or within multi-site aggregations, and 20% provided self-reported race or ethnicity. External validation was reported in 8% of applicable studies, and a single model accounted for 38% of pre-trained applications. Methodological refinement has outpaced the empirical foundations and reporting transparency required for clinical translation. To address this, we introduce NORMA, an 18-item reporting checklist for model construction, validation, and clinical application.