NOVA-Defined UPF Intake: Half of Energy from Ultra-Processed Foods in CD and Healthy Adults

Baseline dietary recalls in a Nutrients analysis found that NOVA-defined ultra-processed foods (group 4) contributed about half of total energy in two trial cohorts: adults with mild-to-moderately active Crohn’s disease and healthy adults.

Investigators used baseline, non-consecutive 24-hour recalls collected with ASA24 and summarized NOVA-classified energy contributions as the exposure measure. The paper also reports that the NOVA group 4 share of energy was associated, in adjusted models, with multiple diet-quality and dietary inflammatory indices derived from the same recall data. Overall, the article describes high ultra-processed food energy contributions across both cohorts alongside modeled links to established index scores.

The analysis included 64 participants with Crohn’s disease and 33 healthy adults, with a mean of about three recalls per participant; median values were used to summarize intake across repeated recalls. For NOVA group 4, the authors report a median of 51.8% of energy (865.3 kcal) in the Crohn’s disease cohort and 51.3% of energy (1146.0 kcal) in the healthy cohort. They further report that, after Bonferroni correction, between-group differences in the proportion of energy derived from NOVA categories—including NOVA group 4—were not statistically significant. In that framing, the paper presents the two cohorts as having broadly similar ultra-processed food energy contributions at baseline.

To generate these estimates, investigators applied a NOVA workflow to ASA24 outputs that emphasized ingredient-level classification when foods were mixed or non-specific. As described, clearly identifiable ready-to-eat items were classified at the food-code level, while homemade or mixed dishes were disaggregated into standardized recipe ingredients linked to reference ingredient codes so energy could be apportioned at the ingredient level. The authors describe their coding as semi-conservative for ambiguous items and note that two investigators coded foods independently, then resolved discrepancies by consensus and performed internal consistency checks using R. For downstream analyses, the study used each participant’s median intake across available recalls and expressed nutrient measures on an energy-normalized basis (per 1000 kcal), outlining a structured approach to defining NOVA exposure from recall data.

In multivariable linear regression models adjusted for age, BMI, and disease status, higher NOVA group 4 percent energy was reported to associate with lower HEI-2015 and aMED scores and a higher Dietary Inflammatory Index (DII). Specifically, each 1% absolute increase in NOVA group 4 energy was associated with a 0.38-point decrease in HEI-2015 (95% CI −0.51 to −0.25, p<0.001) and a 0.035-point decrease in aMED (95% CI −0.05 to −0.02, p<0.001). The corresponding association with DII was a 0.027-point increase per 1% (95% CI 0.004 to 0.05, p=0.02), and the authors note this DII result did not meet their Bonferroni-corrected significance threshold (α=0.0167). Within these models, the reported pattern was that a higher ultra-processed food share tracked with lower diet-quality indices and a higher inflammatory index.

The discussion highlights methodological constraints the authors attribute to applying NOVA in ASA24-based recall datasets, including limited brand- or source-specific detail that can necessitate assumptions about commonly available products, and subjectivity that can arise when classifying mixed or home-prepared dishes. They also note typical limitations of self-reported dietary recalls (including potential recall bias and measurement error) and mention the relatively small healthy cohort as a sample-size consideration for between-group contrasts.

Looking ahead, the authors suggest larger studies and longitudinal and intervention-based research to further evaluate dietary-pattern differences and better understand how NOVA-defined ultra-processed food exposure relates to clinical outcomes in Crohn’s disease. In that forward-looking framing, the paper positions this NOVA-plus-index approach as a basis for subsequent longitudinal or intervention-focused research questions.

Key Takeaways:

• In two baseline cohorts (Crohn’s disease and healthy adults), ultra-processed foods (NOVA group 4) contributed roughly half of total energy (~51% in both groups), and between-group differences in the proportion of energy from NOVA categories were not significant (after Bonferroni correction).
• Adjusted models linked higher NOVA group 4 percent energy to lower HEI-2015 and aMED scores; the reported DII association did not meet the Bonferroni-corrected threshold.
• The paper describes ingredient-level NOVA coding of ASA24 recalls with dual independent coding and consensus adjudication, alongside noted constraints such as brand ambiguity, mixed-dish classification challenges, and sample-size considerations.