Soy and breast cancer risk: a null cohort beside a modest pooled association
A large Chinese cohort and a prospective meta-analysis gave different-looking estimates that should be reported together.

Illustrative food image, not a photograph of the study intervention.
This is incidence, not survival
This study examined whether soy intake was associated with developing breast cancer, not what happens after a diagnosis. The investigators analyzed the China Kadoorie Biobank and also conducted a dose-response meta-analysis of prospective studies. The cohort included more than 300,000 women aged 30 to 79 from ten Chinese regions, enrolled during 2004–2008 and followed for breast cancer events through December 2016. That gives a distinct question from survivor-cohort reports. It does not assess recurrence, treatment response, or whether a person with breast cancer should change therapy. Its findings should remain labeled as incidence evidence throughout the discussion.
Wei Y, Lv J, Guo Y, Bian Z et al.. European journal of epidemiology, 2020. PMID 31754945 · DOI
The large cohort was null
During ten years of follow-up, 2,289 women developed breast cancer. Mean soy isoflavone intake was 9.4 mg/day, with a standard deviation of 5.4 mg/day. Comparing the fourth intake quartile, 19.1 mg/day, with the first, 4.5 mg/day, the multivariable-adjusted relative risk was 1.00 (95% CI 0.81–1.22). The point estimate was exactly the null value, while the interval allowed associations in either direction. The authors concluded that moderate soy intake was not associated with breast cancer risk in this cohort. Omitting that result would make the paper sound more uniformly favorable than it actually was.
Wei Y, Lv J, Guo Y, Bian Z et al.. European journal of epidemiology, 2020. PMID 31754945 · DOI
Pooling produced a smaller-risk association
The meta-analysis of prospective studies found that each 10 mg/day higher soy isoflavone intake was associated with a 3% lower breast cancer risk, with a 95% confidence interval of 1% to 5% lower risk. This was a pooled relative association, not a three-percentage-point reduction in absolute risk. The paper began from inconsistent epidemiological evidence and differences in soy intake across studies. Reporting the pooled estimate alongside the cohort's null result respects that context. It also avoids pretending that a dose-response analysis is equivalent to assigning a soy food and observing a controlled prevention effect.
Wei Y, Lv J, Guo Y, Bian Z et al.. European journal of epidemiology, 2020. PMID 31754945 · DOI
Do not translate isoflavones into a prescription
The exposure was expressed in soy isoflavones, but that does not make the study a trial of isoflavone capsules or establish a supplement recommendation. Nor does the abstract provide a reliable conversion from its incremental estimate to the tofu portion in a home-cooked meal. A tofu-containing dinner is a food pairing, not the tested preventive dose. The useful conclusion is restrained: moderate intake was not associated with breast cancer incidence in this large cohort, while pooled prospective data suggested a modest inverse association. Neither component establishes that soy treats breast cancer, and neither addresses survival after a diagnosis.
Wei Y, Lv J, Guo Y, Bian Z et al.. European journal of epidemiology, 2020. PMID 31754945 · DOI