Impact of the use of food ingredients and additives on the estimation of ultra-processed foods and beverages.

INTRODUCTION

Increasing consumption of ultra-processed foods (UPF), defined by the NOVA classification, has been associated with obesity and other health outcomes. However, some authors have criticized the UPF definition because it is somewhat subjective. Most studies identify UPF using food descriptions; nevertheless, NOVA developers described a list of ingredients, including substances not commonly used for cooking and "cosmetic additives" that could be used to identify UPF. Assessing the impact of the use of different UPF definitions is particularly relevant with respect to children's diet, because several dietary policies target this age group. Thus, our study compared the frequency of UPF among foods and beverages and their share in the diet of Chilean preschoolers using three different methods of identifying UPF.

METHODS

We used cross-sectional 24-h dietary recall data from 962 preschoolers enrolled in the Food and Environment Chilean Cohort (FECHIC) in 2016. All foods and beverages consumed were classified according to NOVA, considering their description (classic method), the presence of ingredients markers of UPF (ingredient marker method), and the presence of markers plus all cosmetic additives (food additive method). We also estimated the caloric share and quintiles of UPF consumption using the three methods. We used kappa coefficients, consistency-of-agreement intra-class correlation (CA-ICC), absolute agreement intra-class correlation (AA-ICC), and weighted kappa coefficients for assessing agreement between methods.

RESULTS

The proportion of UPF products were 65% in the "classic," 67% in the "ingredient marker," and 73% in the "food additive" method, and kappa coefficients between methods varied from 0.79 to 0. 91. The caloric share of UPF was 47, 52, and 58% with "classic," "ingredient marker," and "food additive" methods, respectively. Consistency-of-agreement was higher than the absolute agreement between the methods (CA-ICC = 0.81; AA-ICC = 0.74). For quintiles of UPF consumption, we found weighted kappa of 0.65 as measure of agreement between "classic" and "ingredient marker," and 0.51 between "classic" and "food additive" methods.

CONCLUSION

Searching for all possible markers of UPF in the list of ingredients increased the proportion of food products identified as UPF compared to the classic method. These differences affected the estimated caloric share of UPF in Chilean preschoolers' diets.