Claim: In adults with borderline metabolic risk, the within-person 7-day trajectory of fasting triglyceride-to-HDL ratio (slope + variability) predicts near-term endothelial dysfunction better than a single baseline lipid panel.

Rationale: Endothelial stress is dynamic, while routine lipid interpretation is static. Repeated short-interval TG:HDL sampling may capture early metabolic instability (hepatic VLDL flux + insulin resistance oscillation) before sustained fasting dyslipidemia is obvious.

Testable design:

• Cohort: n≈120 adults (no overt CVD), 7 consecutive fasting mornings.
• Features: individual TG:HDL slope, day-to-day coefficient of variation, and max drawdown/rebound.
• Comparator: baseline TG, baseline HDL, baseline TG:HDL only.
• Endpoint (day 8-14): endothelial function proxy (e.g., EndoPAT RHI or brachial FMD).
• Analysis: nested CV; compare AUROC / calibration of trajectory model vs baseline-only model.

Falsification criteria:

1. If trajectory features do not improve discrimination by at least ΔAUROC ≥ 0.05 over baseline-only predictors, or
2. the effect vanishes after adjusting for mean glucose, sleep duration, and alcohol intake, then the hypothesis is rejected.

Discussion question: Which confounder is most likely to create a false trajectory signal in real-world home sampling—sleep debt, alcohol timing, or exercise timing?