A sex-specific machine learning model combining computed tomography-derived visceral fat measurements with biochemical markers differentiated subtypes of primary aldosteronism with greater accuracy in men than in women, according to a research abstract presented at the 2026 AACE Annual Meeting in Las Vegas. 

Researchers, led by Yu Luo, MM, of Tongji Hospital in Wuhan, China,  evaluated 276 patients with primary aldosteronism confirmed by saline infusion testing and subtyped by adrenal venous sampling across multiple tertiary centers. The final analysis included 235 patients with successful adrenal venous sampling, comprising 111 men and 124 women. The study compared patients with aldosterone-producing adenoma (APA) vs idiopathic hyperaldosteronism (IHA), the two most common subtypes of primary aldosteronism. 

Using an automated AI-assisted segmentation pipeline applied to chest-abdomen computed tomography (CT) scans, researchers measured visceral fat area, renal sinus fat volume, and epicardial adipose tissue volume. These imaging biomarkers were incorporated into five machine learning algorithms, trained and internally validated against adrenal venous sampling, with separate models developed for men and women. 

Marked sex-specific differences emerged. Among men, IHA was associated with significantly greater visceral and ectopic fat deposition than APA. Visceral fat area normalized to body mass index was higher in men with IHA vs APA (6.50 vs 4.88 cm²/kg/m²). Men with IHA also had larger renal sinus fat volume, lower renal sinus fat CT attenuation, and greater epicardial adipose tissue volume compared with men with APA (145 vs 98 cm³). Renal sinus fat volume correlated positively with both visceral fat area and epicardial adipose tissue volume. 

Cardiac functional differences were also observed in men. Men with IHA demonstrated more impaired diastolic function than those with APA, reflected by higher A-wave velocity and elevated E/e′ ratio. Across both subtypes, men also exhibited greater visceral and cardiac adipose deposition and worse renal function than women, including higher creatinine and uric acid levels. 

In contrast, women showed fewer differences in adipose tissue between subtypes. Female patients with IHA had higher triglyceride levels and serum calcium concentrations, along with lower serum sodium levels, compared with women with APA. No significant differences in visceral fat-related parameters were observed between female subgroups. 

The highest-performing model was an XGBoost algorithm developed for men that combined visceral fat area with serum potassium, systolic blood pressure, and baseline and post-saline aldosterone levels. The model achieved an area under the curve of 0.838, with 80% sensitivity and 74% specificity. Performance exceeded that of the corresponding models developed for women, which achieved an area under the curve of 0.750, and the overall cohort, which achieved an area under the curve of 0.766. 

No funding sources or conflicts of interest were reported. 

(Editor's Note: These findings are from a conference presentation on an abstract and should be considered preliminary.)