Sex-specific XGBoost machine learning models incorporating visceral fat area and clinical biomarkers achieved an AUC of 0.84 in males and 0.75 in females for predicting primary aldosteronism subtypes.
Observational (n=276)
Yes
Can machine learning models incorporating visceral fat and biochemical markers accurately predict subtypes of primary aldosteronism?
Background: Primary aldosteronism (PA) subtypes exhibit significant sex-specific differences, particularly regarding cardiovascular risk and the metabolic syndrome.This study investigated these differences between subtypes and developed a machine learning model for subtype prediction.Design and Methods: This retrospective study analyzed clinical and imaging data from 276 PA patients, subtyped by adrenal venous sampling.Visceral and cardiac adipose deposition were quantified, least absolute shrinkage and selection operator (LASSO) regression identified optimal predictive parameters for model development.Results: Unilateral PA (UPA) presented with prominent hypertension and hypokalemia, whereas bilateral PA (BPA) showed more visceral fat deposition, with notable sex differences.(1) In males, the BPA group had a higher body mass index, abdominal fat, and larger epicardial adipose tissue (EAT) volume along with a greater E/e' ratio compared to UPA. (2) Across both subtypes, males demonstrated more abdominal fat than females.(3) In females, BPA had higher triglycerides, serum calcium than UPA.(4) In males, an XGBoost model using visceral fat area (VFA), serum potassium, systolic blood pressure, and plasma aldosterone concentration (PAC) achieved an AUC of 0.840.05.An XGBoost model in females based on serum potassium, lipid profile, PAC after saline infusion test, and CT nodule characteristics yielded an AUC of 0.750.02.Conclusion: PA subtypes exhibit prominent sex-specific cardiometabolic profiles.In males, BPA correlates with significantly greater ectopic fat deposition and elevated cardiovascular risks.Combining VFA with key biochemical markers demonstrates superior diagnostic efficacy for PA subtyping in the male group.
Luo et al. (2026) conducted an observational in Primary aldosteronism (n=276). Sex-specific XGBoost machine learning models was evaluated on Area under the receiver operating characteristic curve (AUC) for predicting primary aldosteronism subtypes in males. Sex-specific XGBoost machine learning models incorporating visceral fat area and clinical biomarkers achieved an AUC of 0.84 in males and 0.75 in females for predicting primary aldosteronism subtypes.