A large prospective cohort from NutriNet-Brasil shows for the first time that people consuming more ultra-processed foods, sugar-sweetened beverages, and trans fats develop depressive symptoms 6–21% more often. Conversely, fruits, vegetables, legumes, and whole grains lower the risk by 4–8%. The association holds even after adjustment for BMI, age, income, and physical activity.
The NutriNet-Brasil team, led by Felice Jacka (Food & Mood Centre, Deakin University) and Maria Laura Louzada (a co-author of the NOVA classification system), analyzed 18,033 Brazilian adults free of depression at baseline. Diet was assessed via two web-based 24-hour recalls (Nova24h), with foods categorized using the NOVA system into four classes — from unprocessed to ultra-processed. Depressive symptoms were tracked using the PHQ-9 ≥ 9 threshold (a validated clinical cutoff) over 31 months of follow-up.
The team applied Cox proportional hazard models adjusted for age, sex, income, education, BMI, physical activity, and smoking. The work is part of the GLAD project (Global burden of disease Lifestyle And mental Disorder Taskforce), which integrates diet as a risk factor in the global burden of mental illness — a first at the WHO level.
Over 31 months, 3,590 participants developed depressive symptoms (PHQ-9 ≥ 9) — roughly 19.9% of the cohort. After adjustment, the data revealed a clear two-way effect:
| Food group | Change in depression risk | Direction |
|---|---|---|
| Trans fatty acids | +21% | Increases |
| Sodium (salt) | +15% | Increases |
| Ultra-processed food (UPF) | +12% | Increases |
| Sugar-sweetened beverages | +10% | Increases |
| Processed meat (sausages, bacon) | +6% | Increases |
| Fruits, vegetables, legumes | −4 to −6% | Decreases |
| Whole grains, nuts, seeds | −5 to −7% | Decreases |
| Dietary fiber, calcium | −6 to −8% | Decreases |
The authors emphasize this is a prospective association: diet was assessed BEFORE symptoms appeared, not retrospectively. That rules out the classic objection that “depressed people just eat worse.” The signal held across all sensitivity analyses: excluding the first year of follow-up, splitting men and women, and varying the PHQ-9 cutoff.
Key signal. It’s not just specific foods that raise risk — the broader ultra-processed dietary pattern drives it. The reverse holds too: every additional 10 g of dietary fiber per day cuts depression risk by about 6%. Food works in both directions.
This isn’t the first study linking diet and mood, but it’s one of the strongest by design. Earlier meta-analyses (including the 633,000-person review we covered) often relied on cross-sectional data. This one is prospective, uses a validated scale, and applies a clear NOVA-based UPF definition. Practical takeaways:
Effect sizes are modest: even the +21% from trans fats is a relative risk; the absolute difference between top and bottom quartiles is around 2–3% over 2.5 years. Diet isn’t the only depression driver — genetics, sleep, social support, and chronic stress weigh in just as heavily. The Brazilian cohort also skews younger and more educated than the general population, so extrapolation to elderly diabetic patients requires caution.
The authors use PHQ-9 as a depression screen rather than a clinical diagnosis. That means a portion of the “cases” are subclinical symptoms, not psychiatrist-confirmed disorders.
⚠ This is a preprint. The paper was posted to PsyArXiv on 23 April 2026 and has not yet been peer-reviewed. Version v1 was published in February — the authors revised the statistical analyses and expanded sensitivity checks. Final numbers may shift slightly through peer review.
NutriAI Pro analyzes a photo of your dish and instantly tells you whether it’s ultra-processed, plus how much added sugar, trans fat, and sodium it contains. Free — 5 analyses per day.
Open @botnutraibot →