Produce and breast cancer: repeated diets, different tumor associations
Two long-running cohorts linked higher produce intake with lower incidence, with differences across tumor subtypes.

Illustrative food image, not a photograph of the study intervention.
A long view of reported diets
The analysis followed 182,145 women in the Nurses' Health Study and Nurses' Health Study II, initially aged 27 to 59 years. The investigators documented 10,911 invasive breast cancer cases. Instead of using just one dietary report, they cumulatively averaged fruit and vegetable intake across repeated, validated questionnaires. Follow-up covered 1980–2012 in one cohort and 1991–2013 in the other. This makes the study a long-term analysis of dietary patterns and subsequent incidence. It remains a prospective observational study: participants were not randomly allocated to different produce intakes, even though the statistical models adjusted for known risk factors.
Farvid MS, Chen WY, Rosner BA, Tamimi RM et al.. International journal of cancer, 2019. PMID 29978479 · DOI
The overall comparison has a denominator
Women reporting more than 5.5 servings of total fruits and vegetables per day had a hazard ratio of 0.89 compared with those reporting 2.5 servings or fewer. The 95% confidence interval was 0.83 to 0.96, and the trend P value was 0.006. These are relative incidence comparisons between reported intake groups, not percentage-point changes in an individual's absolute risk. The abstract particularly highlights cruciferous and yellow/orange vegetables within the inverse association. It does not establish that a single vegetable accounts for the total-produce result, or that one recipe reproduces the exposure observed over decades of follow-up.
Farvid MS, Chen WY, Rosner BA, Tamimi RM et al.. International journal of cancer, 2019. PMID 29978479 · DOI
Receptor status changes the question
The association for total vegetables was particularly apparent for estrogen receptor negative tumors. For each two additional daily servings modeled continuously, the hazard ratio was 0.84, with a 95% confidence interval of 0.77 to 0.93. The reported heterogeneity P value was 0.02. This is a different comparison from the highest-versus-lowest total-produce categories: it concerns vegetables, a continuous intake increment, and a receptor-defined outcome. Keeping those labels attached prevents an easy but misleading blend of numbers. The finding concerns incidence of these tumors in the cohorts, not the effectiveness of vegetables as treatment after a breast cancer diagnosis.
Farvid MS, Chen WY, Rosner BA, Tamimi RM et al.. International journal of cancer, 2019. PMID 29978479 · DOI
Molecular subtypes do not move together
For each two additional daily servings of total fruits and vegetables, the reported hazard ratios were 0.79 for HER2-enriched tumors, 0.84 for basal-like tumors, and 0.94 for luminal A tumors. Their respective confidence intervals were 0.67–0.93, 0.72–0.97, and 0.89–0.99. The authors did not report the same inverse association for luminal B tumors; the heterogeneity P value was 0.03. These differences are part of the result, not a reason to advertise a universal anticancer effect. Repeated dietary questionnaires and adjustment strengthen the description, but they do not make these cohort associations randomized evidence of prevention.
Farvid MS, Chen WY, Rosner BA, Tamimi RM et al.. International journal of cancer, 2019. PMID 29978479 · DOI