Brain structural changes in lifelong premature ejaculation patients: insights from multivariate and neurotransmitter analyses

Abstract Background Lifelong premature ejaculation (PE) is one of the most commonly reported male sexual dysfunctions, previous studies of lifelong PE-related brain structures used univariate analysis and showed inconsistent results. Aim To explore the structural abnormalities and the potential neurotransmitter profile alterations in lifelong PE patients. Methods Based on T1-weighted anatomical data derived from 47 lifelong PE patients and 34 healthy controls (HCs), gray matter volume (GMV) maps were decomposed into independent components (ICs) by source-based morphometry technique to investigate the abnormal brain structure between groups. Shapley Additive exPlanations (SHAP) analysis was applied to assess the importance of the identified imaging features. Then, spatial association analyses were applied to explore possible relationships between GMV abnormalities and neurotransmitter profiles. Outcomes Study outcomes included GMV by two sample t test between groups, spatial correlation between identified components and neurotransmitter profiles, and the importance of the identified imaging features in SHAP analysis. Results GMV abnormalities were mainly located in the default mode network, prefrontal cortex and temporal lobe in lifelong PE. The SHAP analysis revealed that GMV components of IC22, IC25, and IC1 constitute the most critical features for classification prediction. Moreover, GMV abnormalities were correlated with the serotoninergic, dopaminergic, gamma-aminobutric acid, glutamatergic, and noradrenergic systems. Clinical implications These findings help to better understanding the underlying pathological mechanisms of lifelong PE and provide a basis for formulating targeted treatment strategies in the future. Strengths & limitations This study demonstrated the structure-neurotransmitter associations of lifelong PE by combining multivariate analysis with neurotransmitter profiles. Moreover, SHAP interpretability analysis was used to ensure model accountability in machine learning. The sample size of our study was relatively small. To enhance generalizability, larger-scale cohorts and independent validation datasets will be collected in the future. Conclusions These findings provided further evidence of GMV abnormalities in lifelong PE relevant to self-referential information processing and ejaculatory inhibition controlling.