Vegan diets cut total cancer risk by 23%, major meta-analysis finds

Cancer claimed 9.7 million lives in 2022, and researchers have long suspected that what people eat is one of the most modifiable levers available. A new meta-analysis published in the European Journal of Epidemiology now offers the most comprehensive picture yet of how vegetarian and vegan diets relate to cancer incidence, pooling data from 17 publications across seven major prospective cohort studies and more than 472,000 participants followed for up to 20 years.

Vegan diets associated with 23% lower total cancer risk

The headline finding is striking even with appropriate caution: people following vegan diets were 23% less likely to develop cancer overall compared to non-vegetarians (relative risk 0.77, 95% CI 0.70 to 0.85). Vegetarians showed a 13% reduction (RR 0.87, 0.84 to 0.91). These are associations from observational data, not a controlled trial, but they held up across sensitivity analyses and the study populations were large and geographically diverse, spanning the Oxford Vegetarian Study, EPIC-Oxford, the UK Biobank, the Netherlands Cohort, and both Adventist Health Studies.

The reductions were not uniform across cancer types, and that specificity is part of what makes the findings worth taking seriously. Vegetarians showed a 45% lower risk of stomach cancer, a 23% reduction for pancreatic cancer, a 21% reduction each for bladder cancer and melanoma, and a 24% reduction for non-Hodgkin's lymphoma. Colorectal cancer risk was 14% lower, and breast cancer risk was 8% lower overall, rising to 19% for postmenopausal breast cancer specifically. For vegans, breast cancer risk was 20% lower, and a supplementary analysis from the Adventist Health Study II found a 24% reduction in breast cancer and a 42% reduction in prostate cancer risk at age 65.

Who benefits most from plant-based diets for cancer prevention

The findings carry real public health weight. Cancer is a multifactorial challenge shaped by social and behavioral determinants that cut across income, geography, and lifestyle, but diet is one of the few levers individuals can actually adjust. The populations in these studies were largely healthy adults followed over the long term, making the results reasonably applicable to general audiences thinking about prevention.

Body weight appears to be a meaningful part of the story. When researchers adjusted for BMI, the total cancer risk reduction for vegetarians fell from 13% to 8%, and for vegans from 23% to 18%. The authors calculated that differences in baseline adiposity explained 22% of the observed reduction in vegetarians and 42% in vegans. In other words, staying leaner accounts for a substantial share of the benefit, though it does not explain all of it.

Other proposed mechanisms include reduced intake of red and processed meat, which is an established risk factor for colorectal cancer and has been associated with liver, lung, breast, and endometrial cancers as well. Plant-based eaters also tend to consume more fiber, whole grains, and antioxidant-rich produce. At a hormonal level, vegans show lower circulating estradiol and higher sex hormone-binding globulin, patterns consistent with reduced breast cancer risk. Lower insulin-like growth factor 1 levels in vegans may also help explain the prostate cancer signal, since dairy products have been linked to higher IGF-1 and increased prostate cancer risk.

How seven prospective cohort studies were pooled and graded

The researchers searched PubMed and Embase through July 2025 and restricted their analysis to prospective cohort studies, excluding retrospective case-control and cross-sectional designs that are more vulnerable to recall bias and reverse causation. They used DerSimonian and Laird random-effects models and assessed study quality with a modified Newcastle Ottawa Scale, with mean quality scoring 7.05 out of 8.

Evidence strength was graded using World Cancer Research Fund criteria. Vegetarian diets reached a "probable" grade for total, colorectal, colon, total breast, and postmenopausal breast cancers. Stomach, pancreatic, bladder, melanoma, and non-Hodgkin's lymphoma associations were graded "limited-suggestive," as was the evidence for vegan diets and both total and breast cancer. The distinction matters: "probable" in this framework reflects a consistent body of evidence with a plausible biological mechanism, while "limited-suggestive" signals a real but less certain pattern.

To probe whether unmeasured confounders could be driving the results, the authors calculated E-values, a statistical tool that estimates how strong an unmeasured confounder would need to be to explain away an observed association. The E-values suggested that any such confounder would need to be unusually powerful, lending some additional confidence that the findings are not simply an artifact of healthier-than-average lifestyles.

Key limitations that keep this evidence emerging

Several caveats deserve equal weight. All seven cohort studies assessed diet only at baseline, meaning any dietary changes during follow-up went unmeasured, likely producing conservative estimates but also obscuring the actual dietary exposure. Residual confounding is inherent to observational research: people who adopt vegan diets also tend to exercise more, smoke less, and avoid alcohol, and even careful statistical adjustment cannot fully untangle these factors. This is a recognized challenge in minimizing biases in epidemiological methods.

The number of studies available, particularly for vegan diets, remains small enough to limit statistical power for individual cancer types. Subgroup analyses by sex, geographic region, and follow-up duration were constrained by the same small study count. The protocol was also not pre-registered, which is now considered standard practice for systematic reviews. These are real gaps, and they explain why the evidence sits at "emerging" rather than established.

What comes next in plant-based diet and cancer research

The authors call for prospective studies with repeated dietary assessments and longer follow-up, as well as better measurement of potential mediators like BMI, inflammatory markers, and hormonal profiles. Integrating population, behavioral, and clinical data into future cohort designs would help isolate the specific dietary components driving risk reduction, rather than attributing effects to plant-based eating as a single category.

For now, the cumulative picture supports what nutrition researchers have been suggesting for years: shifting toward more plant-based eating is likely to reduce cancer risk, with vegans showing the largest observed benefit. The mechanisms are plausible, the effect sizes are meaningful, and the evidence grade for several cancer types has crossed into "probable" territory. Whether that is enough to change individual behavior is a different question, but the data are increasingly hard to ignore.

Evidence rating

EMERGING — 86/100

Caveats:

• All included studies relied on a single baseline dietary assessment; diet changes during follow-up were not captured, meaning the true associations may be stronger than reported.
• Only seven prospective cohort studies were available, limiting statistical power for many individual cancer sites and precluding robust subgroup analyses for vegans.
• The associations are observational — vegetarians and vegans may differ from meat-eaters in many health behaviors beyond diet, and residual confounding cannot be ruled out despite adjustments for smoking, alcohol, and physical activity.
• BMI accounted for an estimated 22–42% of the total cancer risk reduction, meaning a meaningful portion of the benefit may reflect weight differences rather than diet composition alone.
• The protocol was not pre-registered before analyses were conducted.
• Funding source was not disclosed by the authors.

Systematic review and meta-analysis of seven prospective cohort studies (up to 655,849 participants per analysis). Findings are associational, not causal — vegetarians and vegans differ from non-vegetarians across multiple lifestyle domains, and residual confounding cannot be excluded despite multivariable adjustment. Most cancer-site analyses rest on four to seven studies, limiting sensitivity analyses and publication-bias assessment; results for several sites were not robust to removal of single studies. Not yet replicated in a fully independent meta-analysis incorporating the same breadth of cancer outcomes.

References

1. Dagfinn Aune, Sabrina Schlesinger, Jakub G. Sobiecki (2026). Vegetarian and vegan diets and cancer incidence: a systematic review and meta-analysis of prospective studies. European Journal of Epidemiology. https://doi.org/10.1007/s10654-026-01380-8
2. Pursuing Collaborative Advantage to Solve Wicked Problems in Public Health. frontiersin.org. https://www.frontiersin.org/articles/10.3389/fpubh.2022.868725/full
3. Sara Mesquita, Lília Perfeito, Daniela Paolotti, Joana Gonçalves-Sá (2025). Epidemiological methods in transition: Minimizing biases in classical and digital approaches. PLOS Digital Health. https://doi.org/10.1371/journal.pdig.0000670
4. Francisco Guillen-Grima, Maria Morales-Suarez-Varela, Agustin Llopis-Gonzalez (2025). Clinical Epidemiology Facing New Challenges: Multimorbidity, Digital Surveillance, and Adolescent Mental Health in the Post-Pandemic Era. Journal of Clinical Medicine. https://doi.org/10.3390/jcm14217498