Cerebral small vessel disease burden and latent MRI phenotypes in young-to-middle-aged inpatients with schizophrenia-spectrum disorders: a retrospective cross-sectional study

Cerebral small vessel disease (CSVD) markers are increasingly reported in schizophrenia-spectrum disorders (SSD), but their age-related distribution and phenotypic heterogeneity within young-to-middle-aged clinical populations remain unclear. This study aimed to characterise age gradients of CSVD markers, latent imaging phenotypes, and their clinical and biomarker correlates in a real-world inpatient SSD cohort. Consecutive SSD inpatients aged 18–55 years who underwent brain magnetic resonance imaging at a tertiary mental hospital during 2023–2024 were included ( N = 1,043). CSVD markers were visually rated using Standards for Reporting Vascular Changes on Neuroimaging (STRIVE)-based criteria. A 3-marker CSVD score summarised confluent white matter hyperintensities (WMH), moderate-to-severe basal ganglia perivascular spaces, and lacunes. Latent class analysis was used to derive imaging phenotypes, and classification-error-corrected three-step regression examined demographic, clinical, vascular-metabolic, and circulating biomarker correlates. Composite CSVD burden was present in 11.70% of participants, increasing from 3.11% at ages 18–29 years to 37.18% at ages 50–55 years. The selected three-class LCA solution yielded three exploratory imaging profiles: Low-lesion (52.92%), WMH-predominant (29.43%), and Mixed-marker high-burden (17.64%). The WMH-predominant phenotype peaked in the midlife age bands, whereas the Mixed-marker high-burden phenotype increased progressively with age and was most frequent in the oldest group. Age was the strongest correlate of membership in both non-Low-lesion phenotypes. Female sex was associated with the WMH-predominant phenotype, whereas no circulating biomarker associations remained significant after FDR correction. CSVD markers in young-to-middle-aged SSD inpatients showed pronounced age gradients and were organised into three exploratory latent imaging profiles. This phenotype-based framework may help characterise heterogeneous cerebrovascular involvement in SSD and inform future longitudinal, comparative, and mechanistic studies. Not applicable.