Replication Data and Code for: Dynamic Age Tolerance to Visceral Adiposity: Redefining Cardiometabolic Screening Beyond the Universal 0.50 Threshold

Title: Dynamic Age Tolerance to Visceral Adiposity: Redefining Cardiometabolic Screening Beyond the Universal 0.50 Threshold This article is a large-scale epidemiological study at the intersection of clinical endocrinology and medical informatics that proposes a fundamental revision of cardiometabolic screening approaches. Methodology and Data: The research is based on a representative 12-year US cohort from the NHANES database (2011–2023, N = 13,569 patients). To ensure national representativeness, official analytical survey weights were rigorously applied5. To predict early insulin resistance (HOMA-IR ≥ 2.5), models based on the Extreme Gradient Boosting (XGBoost) algorithm were developed. To absolutely prevent overfitting and data leakage, a strict 10-fold cross-validation procedure was applied, featuring KNN-imputation performed exclusively within the training folds. The traditional "black-box" problem of ML algorithms was resolved using the Explainable AI (SHAP) methodology. Key Scientific Findings: Superiority of WHtR over BMI: Based on Out-of-Fold predictions, it was statistically proven that the Waist-to-Height Ratio (WHtR) significantly outperforms traditional Body Mass Index (BMI) as a marker of visceral adiposity in patients over 40 years of age. The maximum predictive advantage was recorded in perimenopausal women (40–59 years, DeLong’s p=0.0239). Dynamic Age Tolerance: Analysis of local explanations via SHAP Dependence Plots mathematically refuted the clinical utility of the universal cardiological standard (WHtR < 0.50) for all adults. The algorithm revealed a safe shift in the physiological risk threshold from 0.55 in youth (20–39 years) to 0.61–0.62 in the geriatric cohort (≥ 60 years), confirming a "dynamic age tolerance" to visceral fat. Evolution of Behavioral Triggers: SHAP Importance plots uncovered a profound age-related inversion of modifiable risk factors. While fast-food consumption acts as the primary trigger for early insulin resistance in youth, sleep deprivation and physical inactivity (Sedentary Minutes) become the dominant factors in older adults. Conclusion: The study demonstrates the urgent necessity of abandoning the "universal patient" concept and static cut-off values. The integration of Machine Learning and XAI enables a definitive transition to precision medicine, utilizing dynamic clinical thresholds strictly tailored to patient age and sex to prevent false-positive overdiagnosis and iatrogenic stress in geriatric practice