Background: Whole blood viscosity (WBV) has been linked to cardiometabolic risk, yet its relationship with short-term postprandial triglyceride (TG) response remains unclear. We evaluated this question within a fully de-identified, statistically reconstructed synthetic cohort (<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(n = 1{,}500\)</EquationSource> </InlineEquation>), with a primary methodological objective: to demonstrate a rigorously leakage-controlled machine-learning framework for assessing candidate physiological associations. Methods: A strictly leakage-controlled pipeline was implemented, with fold-specific preprocessing and probability calibration confined to training data. WBV was estimated using the de Simone formulation. Model development employed stratified <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(5\times 5\)</EquationSource> </InlineEquation> nested cross-validation, repeated <InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(5\times 10\)</EquationSource> </InlineEquation> resampling, 1,000-bootstrap uncertainty estimation, calibration assessment (sigmoid and isotonic), threshold-sensitivity analysis across <InlineEquation ID="IEq4"> <EquationSource Format="TEX">\(TG_{4h}\)</EquationSource> </InlineEquation> percentiles, and SHAP-based interpretability. Results: Within the synthetic reconstruction, WBV showed negligible association with postprandial response across correlation testing (<InlineEquation ID="IEq5"> <EquationSource Format="TEX">\(r\approx 0\)</EquationSource> </InlineEquation>), multivariable modeling, robustness analyses, and explainability assessments. In contrast, fasting triglycerides (TG<InlineEquation ID="IEq6"> <EquationSource Format="TEX">\(_{0h}\)</EquationSource> </InlineEquation>) exhibited stable monotonic effects and clear phenotype discrimination. Under the primary 75th-percentile <InlineEquation ID="IEq7"> <EquationSource Format="TEX">\(TG_{4h}\)</EquationSource> </InlineEquation> definition, the L2-penalized logistic regression model achieved stable discrimination (nested AUROC <InlineEquation ID="IEq8"> <EquationSource Format="TEX">\(= 0.9141\)</EquationSource> </InlineEquation>; Brier <InlineEquation ID="IEq9"> <EquationSource Format="TEX">\(=0.0886\)</EquationSource> </InlineEquation>) with bootstrap AUROC <InlineEquation ID="IEq10"> <EquationSource Format="TEX">\(=0.914\)</EquationSource> </InlineEquation> (95% CI [0.8957, 0.9314]) and consistent calibration-aware performance. Conclusions: Within this synthetic framework, WBV did not provide reproducible predictive or attributional value for short-term postprandial TG response. These findings represent methodological evidence under modeled assumptions rather than definitive physiological conclusions. The study illustrates how leakage-controlled, calibration-aware ML workflows can evaluate candidate metabolic associations in privacy-preserving settings; external validation in real-world cohorts remains necessary.

A leakage-controlled machine learning framework for postprandial triglyceride phenotyping using synthetic clinical data

期刊论文
出版模式：

开放获取
发表日期：

2026年4月16日

查看全文

Nattakitti Piyavechvirat,
Yi-Jheng Huang,
Qazi Mazhar Ul Haq

摘要

Background:

Whole blood viscosity (WBV) has been linked to cardiometabolic risk, yet its relationship with short-term postprandial triglyceride (TG) response remains unclear. We evaluated this question within a fully de-identified, statistically reconstructed synthetic cohort ( \(n = 1{,}500\) ), with a primary methodological objective: to demonstrate a rigorously leakage-controlled machine-learning framework for assessing candidate physiological associations.

Methods:

A strictly leakage-controlled pipeline was implemented, with fold-specific preprocessing and probability calibration confined to training data. WBV was estimated using the de Simone formulation. Model development employed stratified \(5\times 5\) nested cross-validation, repeated \(5\times 10\) resampling, 1,000-bootstrap uncertainty estimation, calibration assessment (sigmoid and isotonic), threshold-sensitivity analysis across \(TG_{4h}\) percentiles, and SHAP-based interpretability.

Results:

Within the synthetic reconstruction, WBV showed negligible association with postprandial response across correlation testing ( \(r\approx 0\) ), multivariable modeling, robustness analyses, and explainability assessments. In contrast, fasting triglycerides (TG \(_{0h}\) ) exhibited stable monotonic effects and clear phenotype discrimination. Under the primary 75th-percentile \(TG_{4h}\) definition, the L2-penalized logistic regression model achieved stable discrimination (nested AUROC \(= 0.9141\) ; Brier \(=0.0886\) ) with bootstrap AUROC \(=0.914\) (95% CI [0.8957, 0.9314]) and consistent calibration-aware performance.

Conclusions:

Within this synthetic framework, WBV did not provide reproducible predictive or attributional value for short-term postprandial TG response. These findings represent methodological evidence under modeled assumptions rather than definitive physiological conclusions. The study illustrates how leakage-controlled, calibration-aware ML workflows can evaluate candidate metabolic associations in privacy-preserving settings; external validation in real-world cohorts remains necessary.