Evaluation of the Mexican warning label nutrient profile on food products marketed in Mexico in 2016 and 2017: A cross-sectional analysis

Abstract

Background Different nutrient profiles (NPs) have been developed in Latin America to assess the nutritional quality of packaged food products. Recently, the Mexican NP was developed as part of the new warning label regulation implemented in 2020, considering 5 warning octagons (calories, sugar, sodium, saturated fats, and trans fats) and 2 warning rectangles (caffeine and non-nutritive sweeteners). The objective of this cross-sectional study was to evaluate the Mexican NP and other NPs proposed or used in Latin America against the Pan American Health Organization (PAHO) model. Methods and findings Nutrition content data of 38,872 packaged food products available in the Mexican market were collected in 2016 and 2017. The evaluation of the Mexican NP, including its 3 implementation phases of increasing stringency (2020, 2023, and 2025), was conducted by comparing the percentage of products classified as “healthy” (without warnings) or “less healthy” (with 1 or more warnings), as well as the number and type of warnings assigned to food products, against the PAHO NP. Using the calibration method, we compared the classifications produced by the PAHO model against those produced by the NP models of Ecuador, Chile (3 phases), Peru (2 phases), Uruguay, and Brazil. Kappa coefficients and Pearson correlations were estimated, and proportion tests were performed. We found that the 3 implementation phases of the Mexican NP had near to perfect agreement in the classification of healthy foods (Mexico NP models: 19.1% to 23.8%; PAHO model: 19.7%) and a strong correlation (>91.9%) with the PAHO model. Other NPs with high agreement with the PAHO model were the Ecuador (89.8%), Uruguay (82.5%), Chile Phase 3 (82.3%), and Peru Phase 2 (84.2%) NPs. In contrast, the Peru Phase 1, Brazil, and Chile Phase 1 NP models had the highest percentage of foods classified as healthy (49.2%, 47.1%, and 46.5%, respectively) and the lowest agreement with the PAHO model (69.9%, 69.3%, and 73%, respectively). Study limitations include that warnings considered by the Mexican NP models were evaluated as if all the warnings were octagon seals, while 2 out of the 7 were rectangular warnings (caffeine and non-nutritive sweeteners), and that our data are limited by the quality of the information reported in the list of ingredients and the nutrition facts table of the products. Conclusions The 3 implementation phases of the Mexican NP were useful to identify healthy food products. In contrast, the Peru Phase 1, Brazil, and Chile Phase 1 NP models may have limited usefulness for the classification of foods according to the content of ingredients of concern. The results of this study may inform countries seeking to adapt and evaluate existing NP models for use in population-specific applications.