Abstract / Summary
Background. Meal scores are commonly interpreted as describing both the glycemic burden of a meal and the metabolic risk of the person eating it. Those claims require different analyses. Objective. To compare seven meal scoring rules after separating associations among meals eaten by the same participant from associations among participants. Methods. I reconstructed 120-minute glucose responses for 1,330 isolated meals from 42 adults in CGMacros. Seven outcome-independent rules were calculated from recorded macronutrients: the Bite Balance Score, a fixed-glycemic-index net-carbohydrate proxy, carbohydrate:fiber ratio, a four-component balance score, a macro inflammatory proxy, fiber density, and raw carbohydrate. Within-person estimates used participant-centered scores, participant fixed effects, and participant-clustered standard errors. Between-person analyses correlated each participant's mean score with mean positive incremental area under the curve (iAUC). Benjamini-Hochberg correction was applied within each family of seven tests. Results. Six rules were associated with positive iAUC within participants after correction. Per within-person SD, estimates ranged from -972 mg/dL x min for the Bite Balance Score to 860 mg/dL x min for the macro inflammatory proxy. Fiber density was null in the full dataset. The same six rules were associated with peak rise. No participant-level iAUC correlation survived correction; correlations ranged from rho = -.230 to .210, and every bootstrap interval crossed zero. In a joint component-feature model, meal pressure remained associated with iAUC, while early buffering, delayed risk, and postmeal activity did not. Twenty-three records reported fiber above total carbohydrate. Removing them made fiber density significant and produced a nominal BBS participant-level correlation, though the latter did not survive correction. Interpretation. Several carbohydrate-linked rules tracked meal-to-meal glucose variation within participants. Their averages did not reliably rank participants in this cohort, and nutrient-record errors changed two conclusions. Meal scoring and person-level risk assessment should remain separate tasks.