Introduction
What does the evidence actually show when we look at plant-based dietary patterns through the lens of food processing? And does classifying foods as “ultra-processed” meaningfully change how we interpret the relationship between plant-based diets and long-term health outcomes?
A recent large prospective cohort analysis examined both healthful and unhealthful plant-based dietary patterns, and importantly, stratified the contributing foods based on whether they were classified as ultra-processed or non-ultra-processed.
In this episode, the study’s lead author Dr Alysha Thompson, discusses the many intesresting aspects to the analysis. These include how the dietary indices were constructed, the challenges of applying NOVA classification to self-reported dietary data, the role of residual confounding in observational nutrition research, and the extent to which we can or cannot infer independent effects of processing from these findings.
We also consider the broader implications for the use of processing-based classifications in nutrition science and whether this adds meaningful explanatory value beyond traditional assessments of diet quality.
- [03:02] Interview start
- [05:08] The key research question
- [07:52] Study design
- [17:25] Building diet scores
- [21:11] NOVA classification challenges
- [25:04] Key findings
- [38:30] NOVA debate and stance
- [43:44] Adjustments and mediators
- [48:11] Future research directions
- [54:37] Key ideas (premium-only)
Episode Resources
- Join the Sigma newsletter for free
- Subscribe to Sigma Nutrition Premium
- Enroll in the next cohort of our Applied Nutrition Literacy course
- Paper: Plant-Based Diets, Ultra-Processed Foods, and Risks of Mortality and Major Chronic Diseases: A Prospective Cohort Study
Guest Information
Alysha Thompson, PhD is a Research Fellow at the Co-Centre for Sustainable Food Systems within the Institute for Global Food Security at Queen’s University Belfast. She holds a PhD in Nutritional Epidemiology, with doctoral research examining plant-based dietary patterns and chronic disease risk. Her research uses large-scale epidemiological datasets, including the UK Biobank, to investigate how diet quality, food processing and bioactive compounds such as flavonoids relate to physical and psychological health outcomes, with the aim of informing evidence-based dietary recommendations.
Study Notes
Overview
The episode centres on Plant-Based Diets, Ultra-Processed Foods, and Risks of Mortality and Major Chronic Diseases: A Prospective Cohort Study. The study used UK Biobank data to ask whether the ultra-processed food content of a plant-based dietary pattern altered its associations with mortality, type 2 diabetes, cardiovascular disease and cancer.
- Core finding: Greater adherence to a healthful plant-based dietary pattern was associated with lower risks of all-cause mortality and type 2 diabetes whether the relevant foods were classified as high- or low-UPF.
- Equally important finding: Greater adherence to an unhealthful plant-based dietary pattern was generally associated with higher risk, again across both processing strata.
- Best-supported interpretation: In these particular indices, the healthfulness of the dietary pattern showed more consistent associations with outcomes than its binary UPF classification.
- What the study does not establish: It does not show that processing is harmless, that all UPFs are equivalent, or that two otherwise identical diets have the same effects when one contains more UPF.
Useful Terminology
- Plant-Based Diet Index (PDI): A graded research score that ranks dietary patterns according to their relative emphasis on plant and animal foods. It is not a vegetarian-versus-omnivorous classification.
- Healthful PDI (hPDI): A score that gives higher values to greater intake of plant-food groups considered healthful, such as wholegrains, fruit, vegetables, nuts, legumes, tea and coffee, while reverse-scoring less healthful plant foods and animal foods.
- Unhealthful PDI (uPDI): A score that gives higher values to greater intake of less healthful plant-food groups, including refined grains, potatoes, sugar-sweetened beverages, and sweets and desserts, while reverse-scoring healthful plant foods and animal foods.
- NOVA classification: A system that classifies foods according to the nature, extent and purpose of processing. In this analysis, NOVA Group 4 defined ultra-processed foods, while Groups 1–3 were treated as non-UPF.
- Oxford WebQ: A web-based 24-hour dietary questionnaire designed for repeated use in large cohorts. Validation work supports its use for ranking several aspects of intake, but it remains vulnerable to self-report error and incomplete product detail.
- Hazard ratio (HR): A relative comparison of event rates over follow-up. An HR below 1 indicates a lower observed rate in the comparison group; an HR above 1 indicates a higher observed rate. It is not a direct measure of absolute risk.
- Residual confounding: Distortion that remains after statistical adjustment because a confounder was unmeasured, measured imprecisely, modelled inadequately, or not recognised.
- Mediator: A variable that lies on a possible pathway from exposure to outcome. Body mass index may partly mediate diet-related effects while also acting as a confounder, depending on the causal question and timing.
The Research Question
- Previous prospective studies generally distinguish between healthful and unhealthful plant-based dietary patterns.
- UPF research introduces a second dimension: foods with similar positions in a plant-based diet score may differ in processing classification.
- The investigators therefore asked whether the UPF content of healthful and unhealthful plant-based patterns modified their associations with major chronic-disease outcomes.
- The question was deliberately narrower than either “Are plant-based diets healthy?” or “Are ultra-processed foods harmful?”
How the Four Dietary Indices Were Constructed
The investigators adapted the plant-based diet indices developed by Satija and colleagues and crossed diet quality with processing classification.
Dimension 1: Plant-Based Diet Quality
- The underlying framework contained 17 food groups classified as healthful plant foods, less