(P11-007-24) Identification and Categorization of Ultra-Processed Food Intake in Clinical Trials Using Dietary Records and Nutrition Data System for Research

Objectives: Rigorous methodology for categorization of ultra-processed foods (UPF) is vital to ensure standardization and comparability across populations and research studies. Specific guidance for application of the Nova classification system to data obtained from dietary records and analyzed using Nutrition Data System for Research (NDSR) has not been established. The objective of this study is to detail methods to determine UPF intake within clinical trials using dietary records and NDSR.

Methods: Individuals (n=165) enrolled in a behavioral weight loss intervention completed 7-day dietary records at months 0, 3, 6, and 12. Records were entered into NDSR by trained nutrition personnel and food-level data exported. All unique food items across all participants and time points were assigned to a Nova group (unprocessed or minimally processed, processed culinary ingredients, processed, or ultra-processed) based on food description. Absence of descriptive data regarding processing resulted in opting for the lesser degree of processing (e.g., culinary preparation). For a subset of foods from the first two enrolled cohorts, consensus was achieved after independent coding and discussion with a team of three PhD-trained registered dietitians. Nova groups for each food were compared to a dataset from a published trial. Interrater reliability was assessed by percent concordance between datasets and Cohen’s κ coefficient. Discrepancies were used to identify foods for disaggregation into component ingredients or that could be assigned to alternative groups for sensitivity analyses.

Results: All unique foods (n=3481) were assigned Nova groupings: unprocessed or minimally processed foods (18%), processed culinary ingredients (1%), processed foods (13%), and ultra-processed foods (68%). Of these, 1486 foods were in the external dataset, with 91% concordance (κ = 0.84) in Nova groupings. Mixed dishes and those missing adequate descriptors (n=512) were identified as targets for disaggregation and/or sensitivity analyses.

Conclusions: Despite good agreement, there remains discordance between trained researchers in the identification and categorization of foods according to the Nova classification system, particularly for complex foods and mixed dishes, suggesting clear, stepwise methodology is needed.

Funding Sources: American Heart Association; NIH/NIDDK