Latent Profiles of Codex-Mapped Additive Functional Classes and Their Association with Nutri-Score in a Sampled Turkish Retail Food Inventory

Food additives recur in combinations across packaged foods, but their association with nutritional quality is uncertain. We identified latent profiles of Codex-mapped functional classes from label-declared additives in a sampled Turkish retail inventory and related them to Nutri-Score grade. Of the 6515 products, 3383 with a retained functional-class indicator entered latent class analysis. Five profiles were retained using a prespecified multi-criterion procedure without numerical cutoffs; the fifth, gas-related profile was mapping-sensitive and not interpreted. Investigator-computed grades (2023 algorithm) were analyzed in 2942 products across 56 food subgroups with two-step ordinal models. The raising and texture profile was associated with poorer grades across the scale (odds ratio, 1.99; 95% confidence interval, 1.51–2.63); the oxidation and color-retention profile was inversely associated at the grade-E boundary (0.49, 0.37–0.63); and the coating and structure profile was displaced from the favorable end (imprecise estimate). In profile-only models, these directions held under a structurally reduced specification and after restoring nonsweetener classes of excluded sweeteners and polyols; magnitudes and cutpoint-specific forms, particularly for the raising and texture profile, were specification-sensitive. The associations are subgroup-conditional descriptive contrasts in a sampled, non-national inventory; the profiles describe Codex-mapped declarations with mapping-induced dependence, not verified functions, additive intake, or health effects.