Soy and prostate cancer incidence: an association, not therapy.
An updated observational meta-analysis separated soy foods, isoflavones and advanced disease rather than treating them as one outcome.

Illustrative food image, not a photograph of the study intervention.
The prevention question is separate
Soy research can become confusing when cancer incidence and cancer treatment appear under the same headline. Applegate and colleagues’ 2018 review addressed associations between soy-related exposures and prostate cancer risk. It included 30 articles examining soy foods, dietary isoflavones and circulating isoflavones, with analyses concerning primary and advanced prostate cancer. This was a synthesis of observational evidence, not an experiment showing that adding soy changes the course of diagnosed disease. Its findings therefore belong in a discussion of differences in observed risk, not a claim that tofu treats a tumor or replaces prostate cancer care.
Applegate CC, Rowles JL, Ranard KM, Jeon S, Erdman JW.. Nutrients, 2018. PMID 29300347 · DOI
The overall signal favored soy foods
The review reported a statistically significant inverse association between total soy food intake and prostate cancer risk. Unfermented soy food intake also showed an inverse association. Those results describe patterns across the underlying populations; they do not identify what would happen to a particular person assigned to start eating soy. Combining studies can sharpen a summary, but it does not turn the source studies into randomized trials. The authors themselves concluded that further studies were needed before soy consumption could be supported as a prophylactic dietary approach. That reservation is part of the finding, not an objection added afterward.
Applegate CC, Rowles JL, Ranard KM, Jeon S, Erdman JW.. Nutrients, 2018. PMID 29300347 · DOI
Food and isoflavones are different exposures
The analyses did not tell one uniform story for every soy-related measure. Genistein and daidzein intakes were inversely associated with risk, but total isoflavone intake and circulating isoflavone measures were not associated with risk. Fermented soy food intake likewise did not show an association. These distinctions are a reason to name the exposure precisely rather than use “soy” as shorthand for foods, individual compounds and blood measurements at once. A favorable food association is not evidence that an isolated isoflavone product reproduces it. Nor does a null pooled association establish that a food category is harmful.
Applegate CC, Rowles JL, Ranard KM, Jeon S, Erdman JW.. Nutrients, 2018. PMID 29300347 · DOI
Advanced disease needs its own evidence
Neither soy food intake nor circulating isoflavones was associated with advanced prostate cancer risk in the review, and the authors emphasized that very few studies addressed those relationships. Sparse evidence limits what can be concluded: the result is not proof that soy prevents advanced disease, but neither is it a definitive demonstration of no possible relationship. Most importantly, risk of advanced disease is not the same endpoint as response to treatment, recurrence after therapy or survival. The clear editorial boundary is to report the incidence association while leaving those different clinical questions unanswered by this particular meta-analysis. The same caution applies when contrasting exposure categories: different pooled results do not demonstrate that switching from one soy food to another would change an individual’s cancer risk by a predictable amount.
Applegate CC, Rowles JL, Ranard KM, Jeon S, Erdman JW.. Nutrients, 2018. PMID 29300347 · DOI