Sex-specific associations of episode type in bipolar I disorder with neuro-metabolic lateralization of the cortico-striatal-cerebellar (CSC) circuit.
Authors
Affiliations (4)
Affiliations (4)
- Department of Radiology, Affiliated People's Hospital of Jiangsu University, Zhenjiang, Jiangsu, 212002, China.
- Department of Neurology, Affiliated Drum Tower Hospital of Nanjing University Medical School, Nanjing, Jiangsu, 210008, China.
- Department of Radiology, Zhenjiang Mental Health Center, Zhenjiang, Jiangsu, 212002, China. [email protected].
- Department of Radiology, Affiliated People's Hospital of Jiangsu University, Zhenjiang, Jiangsu, 212002, China. [email protected].
Abstract
This study aims to examine whether neuro-metabolic lateralization of the cortico-striatal-cerebellar (CSC) circuit is associated with sex-specific mechanisms underlying episode types in bipolar I disorder (BD-I). A total of 109 patients in the acute phase of BD-I were included in the analysis. For comparisons of general clinical characteristics, CSC circuit structural volumes, metabolic features, and laterality indices (LI), analyses were conducted between the male group (n = 49) and the female group (n = 60). For predictive modeling, sex-stratified XGBoost models were built separately: the male cohort (n = 49) was subdivided into mania (n = 31) and depression (n = 18), and the female cohort (n = 60) was subdivided into mania (n = 17) and depression (n = 43). Structural MRI and MRS were utilized to obtain volumetric and metabolic features within the CSC circuit. Key sex-related features were identified through inter-group consistency analysis, Mann-Whitney U tests, and LASSO regression, followed by the calculation of laterality indices (LI). Sex-stratified XGBoost models were constructed to evaluate the predictive performance of CSC circuit features and LI for symptom onset, with SHAP analysis used to determine feature contributions. LI analysis demonstrated that males had significantly higher LI for ACC-sV, ACC-NAA, BG-Cho, and ACC-Glx (all P < 0.007), whereas females exhibited significantly higher LI for BG-sV (P < 0.007). In the sex-stratified XGBoost models, the LI-based models significantly outperformed the absolute-value-based models in predicting manic episodes in males (AUC = 0.803) and depressive episodes in females (AUC = 0.824) (Delong test: P < 0.05). However, the cross-sex prediction performance remained limited. Additionally, SHAP analysis indicated that ACC-Glx LI contributed most to manic episode prediction in males, while ACC-NAA LI and BG-Cho LI had the highest contributions to depressive episode prediction in females. In patients with BD-I, significant sex differences are evident in the structural and metabolic characteristics of the CSC circuit. Neuro-metabolic lateralization alterations in the ACC and BG regions may be involved in sex-specific mechanisms underlying different symptom onset types. Not applicable.