Brain age gap as a potential biomarker for schizophrenia: A multi-site structural MRI study.

Abstract

Gray matter atrophy in schizophrenia has been widely recognized; however, it remains controversial whether it reflects a neurodegenerative condition. Recent studies have suggested that the brain age gap (BAG) between the predicted and chronological ones may serve as a biomarker for early-stage neurodegeneration. Nevertheless, it is unknown its value for schizophrenia diagnosis and the potential meaning. We included structural MRI datasets from 8 independent sites in the current study, including 501 schizophrenia patients (SZ) and 512 healthy controls (HC). We first applied support vector regression (SVR) to train the age prediction model of the controls using the gray matter volume (GMV) and apply this model to predict the age of the SZ. Meta-analysis identified the SZ had significantly higher BAG than the HC (Cohen's d = 0.38, 95% confidence level = [0.19, 0.57]), and this trend was reliably repeated in each site. Furthermore, logistic regression demonstrated BAG can significantly discriminate the SZ from the HC (OR = 1.07, P = 7.14 × 10-8). Finally, the linear regression study demonstrated a significant negative correlation between the BAG and gray matter volume in both groups, especially at the subcortical regions and prefrontal cortex (P<0.05, corrected using the family-wise error method).Clinical Relevance- This multi-site study suggested that the brain age gap derived from machine learning can be taken as a potential biomarker for schizophrenia, which is significantly associated with brain gray matter atrophy.