The study is timely as U.S. federal and state regulators look to define “ultra-processed foods” as the basis for future policies.

Four Classifications for Food Processing

The researchers compared NOVA, the World Health Organization’s International Agency for Research on Cancer (WHO’s IARC’s), the International Food Information Council (IFIC), and the University of North Carolina’s (UNC’s) frameworks for classifying processed foods. The systems were compared using dietary data from the 2017–2018 National Health and Nutrition Examination Survey (NHANES), representing a total of 4,605 food and beverage items consumed by 4,392 U.S. adults.

After categorizing all food and beverage items in the dataset according to the levels defined by each classification system, the categories were harmonized to enable cross-system comparisons: 1) minimally processed foods, 2) processed foods, and 3) highly processed/formulated foods. Specifically, foods classified as “ultra-processed” (NOVA), “highly processed” (IARC), “ready-to-eat” processed foods and prepared foods or meals (IFIC), and “highly processed” ingredients and standalone “highly processed” foods (UNC) were all grouped into the highly processed/formulated foods category for the purpose of the analysis.

Variability and Agreement Among Systems

The study found notable differences in how systems classified foods. IARC and UNC categorized the greatest proportion of items as highly processed or formulated, followed by NOVA and IFIC. The percentage of food and beverage items that were categorized under the same processing level by the two coders assigned the task ranged from 68 to 86 percent, with IARC demonstrating the highest inter-rater reliability and NOVA demonstrating the least. (Recognizing that initial coding could result in discrepancies, the researchers utilized expert adjudication to resolve discordant classifications across all four systems).

The highest cross-system food classification agreement existed between IARC and UNC (78.9 percent), while the lowest agreement was between IARC and IFIC (59.2 percent). The NOVA showed moderate agreement with all other classification systems (63.8–66.8 percent).

Although cross-system categorization agreement across the four systems was low, food groups such as grain products and meat, poultry, fish, and mixtures consistently dominated the highly processed/formulated category.

Consistent Nutrition and Health Associations

Despite differences among classification criteria, the systems identified some similar patterns regarding health outcomes, participant demographics, and nutrition.

Across all four systems, greater consumption of highly processed or formulated was consistently associated with higher body mass index (BMI) and elevated levels of high-sensitivity C-reactive protein, a marker of inflammation that indicates cardiovascular health risk (and consistent with previous research). For three of four systems, higher intake was related to greater waist circumference and lower HDL cholesterol. In contrast, associations with blood pressure, fasting glucose, triglycerides, LDL cholesterol, and total cholesterol were not consistent across systems.

The analysis showed significant associations between greater intake of highly processed/formulated meat products and higher BMI. Interestingly, despite previous research linking sugar-sweetened beverages to adverse metabolic and cardiovascular outcomes, the researchers did not observe consistent significant associations for this food group across classification systems.

Nutritionally, greater intake of highly processed/formulated foods was consistently associated with higher carbohydrate and total sugar intake, lower protein and fiber intake, and lower levels of several micronutrients.

Across systems, higher consumers of processed/formulated foods were consistently younger, more likely to be Non-Hispanic White or Black (and less likely to be Asian or Hispanic), and tended to have lower education levels.

Limitations of “Ultra-Processed” and Other Food Processing Classifications

The researchers hypothesized that the moderate inter-rater reliability observed in the study is likely attributable to the inability of current food processing categorization systems to consider nutritional content, ingredients, preparation methods, or the purpose of certain additives or processing steps. The lack of such information limits the ability to distinguish between foods that may appear similar, but in reality, differ meaningfully in formulation or processing intent.

For example, the widely used NOVA system, known for defining the term “ultra-processed foods” as it is most commonly used, faces criticism for overlooking the nutritional composition and potential benefits of certain foods. The researchers illustrated, under NOVA, “plant-based meat and dairy alternatives, though classified as ultra-processed, may offer advantages over traditional animal products, such as lower saturated fat content… [Additionally], some studies have linked [processed] foods like breakfast cereals, yogurt, and ice creams to lower risks of adverse health events, highlighting the complexity of assessing health impacts based on processing levels.”

The researchers also pointed out that safety data on many additives remain limited, and their role in health is insufficiently understood.

The Need for Standardization as Processing Remains an Important Factor

The authors emphasized that discrepancies in how food processing is defined may complicate nutrition research and policymaking.

Importantly, however, despite differing definitions and classification criteria, the consistent associations between higher intake of highly processed/formulated foods and demographics, nutrition, and health risks suggest that processed food intake should continue to be considered in future nutrition research.

Overall, the researchers concluded that a standardized approach to classifying food processing is needed to improve consistency in nutrition research, support more accurate dietary assessment, and inform evidence-based policy decisions.