Translational cancer research · 2026
The obesity paradox effect of triglyceride-glucose-body mass index (TyG-BMI) on all-cause mortality in patients with malignant tumors: a machine learning-driven cohort study.
СтатьяНазвание и аннотация: Перевод в очереди
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.
Wang Mingxing, Li Yahao, Chen Jian, Dong Wanhui, Wang Yuru, Bo Siyi, Wu Pantong, Wang Zhiying, Zhong Zhixian, Zhong Yi
Аннотация
BACKGROUND: The prognostic significance of the triglyceride-glucose (TyG) index and its adiposity-related derivatives-incorporating body mass index (TyG-BMI), waist circumference (TyG-WC), and waist-to-height ratio (TyG-WHtR)-remains underexplored in patients with malignancies. This study aimed to systematically evaluate their predictive value for all-cause, cancer-specific, and cardiovascular disease (CVD) mortality utilizing traditional survey-weighted Cox regression and advanced machine learning (ML) algorithms.
METHODS: A total of 2,151 cancer patients from the National Health and Nutrition Examination Survey (NHANES) (1999-2018) cohort were enrolled, with a median follow-up of 89 months. Multivariable survey-weighted Cox proportional hazards models and restricted cubic splines (RCS) were employed to evaluate linear and non-linear epidemiological associations. For prognostic benchmarking, survival ML algorithms-specifically, random survival forests (RSF) and gradient boosting survival analysis (GBSA)-were applied. Model discrimination was quantified using Harrell's concordance index (C-index), while permutation feature importance and SHAP analysis were utilized to provide individual-level interpretability.
RESULTS: In unadjusted models, TyG, TyG-WC, and TyG-WHtR were positively associated with mortality; however, these associations were markedly attenuated following multivariable adjustment for confounders. Only TyG-BMI maintained a robust non-linear plateau association with all-cause mortality [approximate nadir at 259.4; Q2 vs. Q1 hazard ratio (HR) 0.753; Q3 vs. Q1 HR 0.729], reflecting an "obesity paradox" pattern that warrants cautious causal interpretation. TyG-WC exhibited an inverted L-shaped curve (inflection point at 850.2). Crucially, in traditional Cox regression, none of the TyG indices independently predicted cancer-specific mortality. Survival ML benchmarking significantly enhanced predictive performance: the TyG index paired with the GBSA model emerged as the optimal predictive tool for all-cause mortality [C-index =0.771; 95% confidence interval (CI), 0.741-0.801], whereas TyG-BMI combined with the RSF model achieved the highest predictive accuracy for cardiovascular mortality (C-index =0.840; 95% CI, 0.802-0.875).
CONCLUSIONS: TyG-BMI demonstrates a significant non-linear relationship with all-cause mortality in malignancy survivors, indicating that composite metabolic-adiposity profiles can serve as biomarkers for systemic resilience. For predictive stratification, ML identified the TyG index and TyG-BMI as the optimal predictors for all-cause and cardiovascular mortality, respectively. Crucially, their prognostic utility is strictly confined to non-cancer competing outcomes, offering novel perspectives for individualized systemic risk management.
METHODS: A total of 2,151 cancer patients from the National Health and Nutrition Examination Survey (NHANES) (1999-2018) cohort were enrolled, with a median follow-up of 89 months. Multivariable survey-weighted Cox proportional hazards models and restricted cubic splines (RCS) were employed to evaluate linear and non-linear epidemiological associations. For prognostic benchmarking, survival ML algorithms-specifically, random survival forests (RSF) and gradient boosting survival analysis (GBSA)-were applied. Model discrimination was quantified using Harrell's concordance index (C-index), while permutation feature importance and SHAP analysis were utilized to provide individual-level interpretability.
RESULTS: In unadjusted models, TyG, TyG-WC, and TyG-WHtR were positively associated with mortality; however, these associations were markedly attenuated following multivariable adjustment for confounders. Only TyG-BMI maintained a robust non-linear plateau association with all-cause mortality [approximate nadir at 259.4; Q2 vs. Q1 hazard ratio (HR) 0.753; Q3 vs. Q1 HR 0.729], reflecting an "obesity paradox" pattern that warrants cautious causal interpretation. TyG-WC exhibited an inverted L-shaped curve (inflection point at 850.2). Crucially, in traditional Cox regression, none of the TyG indices independently predicted cancer-specific mortality. Survival ML benchmarking significantly enhanced predictive performance: the TyG index paired with the GBSA model emerged as the optimal predictive tool for all-cause mortality [C-index =0.771; 95% confidence interval (CI), 0.741-0.801], whereas TyG-BMI combined with the RSF model achieved the highest predictive accuracy for cardiovascular mortality (C-index =0.840; 95% CI, 0.802-0.875).
CONCLUSIONS: TyG-BMI demonstrates a significant non-linear relationship with all-cause mortality in malignancy survivors, indicating that composite metabolic-adiposity profiles can serve as biomarkers for systemic resilience. For predictive stratification, ML identified the TyG index and TyG-BMI as the optimal predictors for all-cause and cardiovascular mortality, respectively. Crucially, their prognostic utility is strictly confined to non-cancer competing outcomes, offering novel perspectives for individualized systemic risk management.