A Network-based Analysis of Ingredients Lists in Nutrient Profile Systems

Abstract

It is well-established that diseases can be prevented or mitigated through dietary intervention, yet proper nutrition is one aspect consumers struggle to manage. Recent studies have shown that there are barriers to understanding the nutrition facts component of a food label which can be linked to dietary choices. In this work, we demonstrate reproducibility and replicability of a network-based method for automating the analysis of ingredients on a food product label en masse using the Open Food Facts Database and the USDA FoodData Central Branded Foods database in February 2020. Our results, which analyze the co-occurrence of 72,754 ingredients across show some consensus in labeling across FALCPA-regulated ingredients in food product labels across databases but highlight potential areas for discrepancy in consumer understanding and labeling practices for terms not subject to strict regulations. The key findings or contributions of this work include the provision of a reproducible method for quantifying the ingredients of packaged food in the United States across two nutritional profiling systems, and have identified 17 total ingredients that appear in the top 20 most co-occurring ingredients for both databases examined. We compare how of 8 FALCPA-regulated ingredients are represented in ingredients lists versus a common, but non-FALCPA regulated ingredient (corn), to demonstrate how one could examine differences between ingredient labeling between products. These findings suggest more research is needed in developing information systems to increase information available for consumers.