When Ultra-Processed Foods are Eaten: The Missing Dimension in Exposure Measurement for Obesity Research

PURPOSE OF REVIEW: Almost every study linking ultra-processed foods (UPFs) to obesity measures exposure as the proportion of energy or weight derived from Nova group 4 foods, averaged across days. This scalar is commonly treated as though it were the exposure itself, yet it embodies an untested assumption: that the timing and context of intake carry no information beyond total amount. This review examines what is lost when occasion-level intake is collapsed to a daily average and how that loss may affect epidemiological inference. RECENT FINDINGS: Controlled circadian-misalignment protocols and time-restricted eating trials show that the metabolic response to a fixed quantity of food varies with its timing. Postprandial studies further show that meal sequence, co-ingestion and eating-occasion structure modify glycaemic and appetite responses at equal daily totals. These findings expose a mismatch between the timescale at which relevant processes operate and the daily scale at which UPF exposure is conventionally measured. Measurement-error and causal-inference frameworks show how such temporal collapse could attenuate or otherwise distort effect estimates, reduce transportability and obscure the intervention to which an association refers. The daily group 4 fraction is a modelling choice, not a complete representation of the diet. It discards temporal and contextual structure that may explain heterogeneity across UPF-obesity cohorts and determine the causal contrast represented by their estimates. This review proposes a time-resolved, context-aware construct and specifies the studies needed to test whether it adds information beyond the conventional scalar. The controlled timing literature manipulated total energy, macronutrients or glucose loads rather than Nova group 4 foods, so its extension to UPFs remains a testable inference rather than an established finding. The immediate priority is therefore not to replace the scalar, but to analyse cohorts whose dietary records retain item-level time stamps and determine whether the discarded structure predicts outcomes beyond total amount.