Sigma Nutrition Radio #615: Why Processing Doesn't Determine the Healthfulness of Plant-Based Diets – Alysha Thompson, PhD
12 snips
Aug 4, 2026 Alysha Thompson, research fellow in nutritional epidemiology specializing in plant-based diets, walks through a large cohort study she led. She explains how diet scores were built and the challenges of applying NOVA processing categories to recalls. The conversation covers how processing interacts with plant-based pattern classifications, key food groups driving results, and implications for using processing-focused frameworks in nutrition research.
AI Snips
Chapters
Transcript
Episode notes
How The UK Biobank Dietary Sample Was Constructed
- The analysis used UK Biobank participants aged 40–70 who completed at least two Oxford WebQ 24-hour recalls and were followed for mortality, T2D, CVD and cancer.
- Follow-up ranged ~8.3–10.5 years with Cox models estimating hazard ratios.
Crossing PDI With NOVA Creates Four Diet Scores
- They combined the Harvard plant-based diet index (PDI) with NOVA processing categories to create four scores: high-UPF HPDI, low-UPF HPDI, high-UPF UPDI, low-UPF UPDI.
- Scores rewarded plant foods and penalised animal foods, then were split by NOVA4 versus NOVA1–3.
NOVA Classification Is Tricky With Recall Data
- Applying NOVA to 24-hour recalls required subjective judgement because recalls lack brand, ingredient, and preparation detail needed to classify some items as UPF.
- Some items (e.g., fruits) couldn't be classified as UPF, forcing additional model adjustments.