A systematic review and meta-analysis of 46 studies finds that higher ultra-processed food (UPF) consumption is associated with a 9% greater risk of cardiovascular disease and a 16% greater risk of hypertension, with particularly strong signals for coronary heart and cerebrovascular diseases. Critically, the review exposes deep methodological flaws across the field: only 2 of 46 studies used directed acyclic graphs to guide confounder selection, 43 applied suboptimal covariate adjustment, and 42 were statistically overfitted by adjusting for potential mediators rather than true confounders — a mistake that biases effect estimates downward.
The significance here is double-edged. On one hand, the UPF-CVD association survives even poorly designed research, suggesting the underlying biological signal — plausibly mediated through excess sodium, emulsifiers, advanced glycation end-products, and displacement of whole foods — is robust enough to withstand analytical noise. On the other, the finding that methodological limitations likely *attenuate* observed risks means the true hazard of UPF consumption may be meaningfully larger than the pooled estimates suggest. This reframes the public health calculus considerably.
The field's reliance on inconsistent NOVA classification further complicates cross-study comparisons. As a preprint not yet peer-reviewed, these conclusions warrant cautious interpretation. Still, this analysis is confirmatory yet consequential — it strengthens the case for UPF reduction while issuing an overdue methodological corrective to an increasingly influential research area.
