Wu · British journal of cancer 2008 · Systematic review and meta-analysis of observational studies · n=19 studies

Epidemiology of soy exposures and breast cancer risk.

Cited 416 times in the scientific literature.

Level 4 - case-series / case-control

Systematic review and meta-analysis composed predominantly of retrospective case-control studies.

PubMed 18182974 · doi:10.1038/sj.bjc.6604145 · record verified 2026-08-28

What was done

The authors reviewed and meta-analyzed observational studies examining dietary soy isoflavone exposure and breast cancer risk, focusing on studies with detailed exposure assessments and multivariable adjustment for confounders. The quantitative meta-analysis evaluated 8 studies (1 cohort, 7 case-control) in high-soy-consuming Asian populations and compared them with findings from 11 studies in low-soy-consuming Western populations.

What was found

In Asian populations, increasing soy intake was associated with a dose-dependent reduction in breast cancer risk compared with intake of ≤5 mg isoflavones per day: - Modest intake (~10 mg/day): OR 0.88 (95% CI 0.78–0.98) - High intake (≥20 mg/day): OR 0.71 (95% CI 0.60–0.85) In 11 Western studies, where average highest and lowest isoflavone intakes were only ~0.8 mg/day and ~0.15 mg/day respectively, soy intake showed no relationship with breast cancer risk.

Why it matters

This analysis highlights that potential protective associations between soy consumption and breast cancer appear limited to populations consuming meaningful quantities of dietary isoflavones (≥10–20 mg/day), explaining the lack of association in Western cohorts with negligible intake.

Limits

The primary evidence for risk reduction comes almost entirely from case-control studies (7 of 8 Asian studies), which are vulnerable to recall bias and selection bias. Western exposure levels were too low to evaluate whether high soy consumption confers similar effects in non-Asian populations. Total sample size and statistical heterogeneity were not reported in the abstract.

Cited by