Hypothesis: A composite biomarker that integrates continuous glucose monitoring (CGM)‑derived glycemic variability with the magnitude and speed of ketone and long‑chain acylcarnitine suppression after a standardized mixed‑macronutrient meal combined with moderate exercise will accurately reflect clamp‑measured metabolic flexibility (ΔRQ) in healthy non‑diabetic individuals and will forecast the magnitude of improvement in insulin sensitivity following early time‑restricted eating (TRE).

Mechanistic rationale: Early TRE aligns food intake with the circadian peak of hepatic insulin sensitivity and the trough of gluconeogenic gene expression, thereby enhancing the liver’s ability to switch from glucose production to utilization. Exercise‑induced irisin secretion amplifies PGC‑1α–driven mitochondrial biogenesis in skeletal muscle, increasing the capacity for fatty acid oxidation and the rapid inhibition of lipolysis when glucose becomes available. The suppression of circulating β‑hydroxybutyrate and long‑chain acylcarnitines (C16, C18) after a carbohydrate‑rich challenge reflects the acute inhibition of adipose tissue lipolysis and the shift toward glucose oxidation, processes governed by malonyl‑CoA–mediated CPT1 inhibition. CGM captures the ensuing glucose excursion and variability, which together with metabolite suppression provide a dynamic readout of the interplay between hepatic glucose output, muscle fatty acid uptake, and mitochondrial flexibility.

Testable protocol: Recruit 30 healthy adults (age 20‑35, BMI 18‑25) in a randomized crossover design. Each participant completes two 2‑week interventions separated by a 4‑week washout: (1) early TRE (eating window 08:00‑14:00) and (2) a control habitual diet. Before and after each intervention, subjects undergo a hyperinsulinemic‑euglycemic clamp with indirect calorimetry to measure ΔRQ (difference between fasting and insulin‑stimulated respiratory exchange ratio). On a separate visit, they consume a standardized meal (50 % carbohydrate, 30 % fat, 20 % protein) followed by 30 minutes of moderate‑intensity cycling (60 % VO₂max). CGM records glucose at 5‑minute intervals for 3 hours; venous blood is drawn at baseline, 30, 60, and 90 minutes for β‑hydroxybutyrate and long‑chain acylcarnitine quantification. Compute: (i) CGM‑derived glycemic variability index (mean absolute glucose change + % time outside 70‑140 mg/dL); (ii) metabolite suppression index (area under the curve of ketone and acylcarnitine concentrations normalized to baseline). Combine the two indices via a weighted linear model (weights derived from partial least squares regression).

Analysis: Test whether the composite index predicts clamp‑measured ΔRQ (primary outcome) better than CGM alone using hierarchical linear modeling (ΔRQ ~ CGM + metabolite suppression + interaction). A significant improvement in model fit (ΔAIC < ‑2, p < 0.05 for added metabolite term) supports the hypothesis. Secondary outcomes: correlate changes in the composite index with changes in fasting insulin, HOMA‑IR, and Matsuda index after early TRE.

Falsifiability: If the metabolite suppression index does not significantly increase the predictive power of CGM for ΔRQ, or if changes in the composite index fail to correlate with improvements in insulin sensitivity after early TRE, the hypothesis is refuted. This framework directly addresses the translational gap noted in human TRE studies by providing a feasible, home‑compatible biomarker pair that captures the substrate‑switching dimension missed by glucose‑only metrics.