Cross-sectional associations of lipid accumulation and immune-nutritional indices with ultrasonography-detected gallstone disease: an explainable machine learning analysis.

BACKGROUND: Recent studies have linked individual lipid accumulation and triglyceride-glucose indices to the prevalence of gallstones, but direct comparisons of metabolic, inflammatory, and immune-nutritional indices in hospital-based health examination cohorts with ultrasonography-defined gallstones remain limited. We therefore examined whether routinely available composite indices were associated with prevalent gallstone disease and evaluated their cross-sectional discrimination performance. METHODS: We conducted a single-center, hospital-based cross-sectional study of adults undergoing routine health examinations at the Health Examination Center of the Second Affiliated Hospital of Shandong First Medical University during 2024-2025. The primary inferential analyses included multivariable logistic regression models evaluating the associations between each candidate index and ultrasonography-detected gallstone disease. Quartile analyses, restricted cubic splines, ROC analyses, calibration, decision curve analysis, and explainable machine learning analyses were treated as exploratory. Stratified 5-fold cross-validation was used to assess internal model stability. RESULTS: Among 1,376 participants [mean (SD) age, 47.27 (13.35) years; 743 men (54.00%)], 127 (9.23%) had ultrasonography-detected gallstone disease. In Model 3, TyG-WC was nominally associated with gallstone disease per 1-SD increment (OR, 1.33; 95% CI, 1.05-1.69; p = 0.017) and in Q4 versus Q1 (OR, 3.10; 95% CI, 1.38-6.96; p = 0.006), although the quartile-specific CIs were wide. The CALLY index (OR per 1-SD, 0.62; 95% CI, 0.46-0.84; p = 0.002), NLR (OR, 1.30; 95% CI, 1.09-1.55; p = 0.004), and PNI (OR, 0.74; 95% CI, 0.61-0.91; p = 0.004) remained significant after within-family false discovery rate (FDR) correction, whereas TyG-WC and CTI did not. TyG-WC had limited single-marker discrimination (AUC, 0.66; 95% CI, 0.61-0.71). Penalized logistic regression showed only modest internally validated discrimination (5-fold cross-validated AUC, 0.74; 95% CI, 0.69-0.78), with low precision-recall performance. CONCLUSION: Several routinely derived metabolic and immune-nutritional indices were associated with prevalent ultrasonography-detected gallstone disease, but effect estimates were imprecise for some comparisons and discrimination was modest. The findings are hypothesis-generating; they do not establish incremental predictive value beyond conventional risk factors or support stand-alone clinical risk stratification without external and longitudinal validation.