Cadmium is a toxic heavy metal that has been implicated in breast cancer etiology, albeit with inconsistent results.

To investigate the shape of the relation between cadmium exposure and breast cancer incidence and mortality in cohort studies.

Following a literature search through April 14, 2020, we carried out a systematic review and dose-response meta-analysis to investigate the shape of the relation between cadmium exposure (assessed either through diet or urine excretion) and disease incidence and mortality.

For inclusion, a study had to report incidence or mortality for breast cancer according to baseline cadmium exposure category; be a prospective cohort, case-cohort or nested case-control study with a minimum one-year follow-up, and reporting effect estimates for all exposure categories.

Studies were evaluated using the ROBINS-E risk of bias tool. The effects in humans were assessed quantitatively using one-stage dose-response meta-analysis in a random effects meta-analytical model.

We identified 10 studies eligible for inclusion in the dose-response meta-analysis, six based on cadmium dietary intake, and four on urinary excretion levels. We found a marginal and imprecise positive relation between dietary cadmium intake and breast cancer, and no association when urinary cadmium excretion was used for exposure assessment. Compared to no exposure, at 20 µg/day of cadmium intake the summary risk ratio was 1.12 (95% confidence interval 0.80–1.56), while at 2 µg/g creatinine of cadmium excretion the summary risk ratio was 0.89 (95% confidence interval 0.38–2.14). Analysis restricted to post-menopausal women showed no association between either dietary or urinary cadmium and subsequent breast cancer incidence and mortality.

Overall, we found scant evidence of a positive association between cadmium and breast cancer. Available data were too limited to carry out stratified analyses according to age, smoking and hormone receptor status. Therefore, possible associations between cadmium exposure and breast cancer in selected subgroups cannot be entirely ruled out.