Acta diabetologica · 2026
Unmasking covert metabolic syndrome in normal-weight women: regional adiposity and muscle quality outperform BMI in a temporally validated machine learning triage framework.
СтатьяНазвание и аннотация: Перевод пока не подготовлен
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Palamarchuk Olga S, Moroz Oleh O, Feketa Volodymyr P
Аннотация
AIMS: Traditional screening based on body mass index (BMI) fails to identify the metabolically obese normal-weight (MONW) phenotype. We aimed to develop and temporally validate an interpretable machine learning framework utilizing non-invasive surrogate markers of central adiposity (waist-to-height ratio, WHtR) and relative functional muscle capacity (relative handgrip strength, RHGS) for cross-sectional triage of prevalent metabolic syndrome (MetS) in women.
METHODS: In 2,418 women (NHANES 2011-2014), we excluded pregnant women and applied independent K-nearest neighbors imputation (k = 5) without data leakage. Using XGBoost, we evaluated baseline M1 (Age + BMI), intermediate M2 (Age + BMI + RHGS), and proposed M3 (Age + WHtR + RHGS). Following TRIPOD Type 3 temporal validation, we trained models on Cycle 2011-2012 (N = 1,167) and tested them on Cycle 2013-2014 (N = 1,251). SHapley Additive exPlanations (SHAP) extracted exploratory, model-dependent decision reference points.
RESULTS: In the unstratified cohort, M3 achieved comparable discrimination to M2 (AUC 0.782 vs. 0.769, DeLong p = 0.117). However, in normal-weight women (n = 371), M3 demonstrated marked superiority over M1 (AUC 0.855 [95% CI 0.800-0.910] vs. 0.757 [95% CI 0.678-0.835], p < 0.001) and M2 (AUC 0.801, p = 0.015), with near-perfect calibration (Brier score = 0.068), high rule-out reliability (NPV = 98.9%), and a negative likelihood ratio (LR-) of 0.12 (95% CI 0.06-0.22). SHAP dependence analysis illustrated non-additive muscle-fat interactions, where lower strength (< 0.891 kg/kg) amplified risk primarily when WHtR exceeded 0.529.
CONCLUSIONS: Replacing BMI with indices of visceral adiposity and relative strength markedly improves MetS triage in normal-weight women. This temporally validated framework provides a non-invasive pre-laboratory triage tool to prioritize high-risk MONW individuals for comprehensive clinical evaluation.
METHODS: In 2,418 women (NHANES 2011-2014), we excluded pregnant women and applied independent K-nearest neighbors imputation (k = 5) without data leakage. Using XGBoost, we evaluated baseline M1 (Age + BMI), intermediate M2 (Age + BMI + RHGS), and proposed M3 (Age + WHtR + RHGS). Following TRIPOD Type 3 temporal validation, we trained models on Cycle 2011-2012 (N = 1,167) and tested them on Cycle 2013-2014 (N = 1,251). SHapley Additive exPlanations (SHAP) extracted exploratory, model-dependent decision reference points.
RESULTS: In the unstratified cohort, M3 achieved comparable discrimination to M2 (AUC 0.782 vs. 0.769, DeLong p = 0.117). However, in normal-weight women (n = 371), M3 demonstrated marked superiority over M1 (AUC 0.855 [95% CI 0.800-0.910] vs. 0.757 [95% CI 0.678-0.835], p < 0.001) and M2 (AUC 0.801, p = 0.015), with near-perfect calibration (Brier score = 0.068), high rule-out reliability (NPV = 98.9%), and a negative likelihood ratio (LR-) of 0.12 (95% CI 0.06-0.22). SHAP dependence analysis illustrated non-additive muscle-fat interactions, where lower strength (< 0.891 kg/kg) amplified risk primarily when WHtR exceeded 0.529.
CONCLUSIONS: Replacing BMI with indices of visceral adiposity and relative strength markedly improves MetS triage in normal-weight women. This temporally validated framework provides a non-invasive pre-laboratory triage tool to prioritize high-risk MONW individuals for comprehensive clinical evaluation.