Background Reports of the association between breastfeeding and ovarian cancer are inconsistent, and few studies have been able to explore possible heterogeneity by histological subtype.

Methods We used data from the Million Women Study, a prospective study of UK women. Participants were recruited in 1996–2001 at breast-screening invitation, and completed a questionnaire on various exposures, including breastfeeding. Follow-up was via routinely-collected cancer registry data. Using Cox regression models, we estimated adjusted relative risks of ovarian cancer amongst women who breastfed compared to women who had not, and by average duration of breastfeeding per child. All analyses were restricted to parous women, stratified by parity (1, 2, 3+), and adjusted for ten potential confounding factors. We used a competing risks approach to explore variation by histological subtype. Analyses were conducted using STATA-13.

Results After excluding nulliparous women and those with bilateral oophorectomy or unknown breastfeeding, the study population included 813,035 women. There were 5026 incident cases of ovarian cancer, after a mean follow-up of 12.9 years. 553,624 (68%) of the women reported breastfeeding (median duration: 2 months per child; IQR: 1–5 months). Overall, the relative risk of ovarian cancer was not significantly different amongst women who breastfed compared to women who had not (RR: 0.94, 95% CI: 0.89–1.00, p = 0.07). However, there was strong evidence that this varied by duration of breastfeeding. Women who breastfed for <6 months per child had no significant change in risk (RR: 0.97, 95% CI: 0.91–1.03), while women who breastfed for ≥6 months/child had an estimated 16% reduction in the relative risk of ovarian cancer (RR: 0.84, 95% CI: 0.76–0.92), compared to women who never breastfed (p-het = 0.0009). On average, there was a 12% reduction in the relative risk of ovarian cancer per 6-month increase in breastfeeding per child (RR: 0.88, 95% CI: 0.83–0.93, p-trend < 0.001). Amongst the four main histological subtypes, there was only evidence of a significant reduction in risk of serous tumours (n = 2251, RR: 0.87 per 6-months/child, 95% CI: 0.79–0.95), but not of mucinous (n = 533, RR: 0.91, 95% CI: 0.76–1.08), endometrioid (n = 417, RR:0.90, 95% CI: 0.74–1.10) or clear cell tumours (n = 233, RR: 1.09, 95% CI: 0.88–1.36), although there was not statistical evidence of heterogeneity by subtype (p-het = 0.4).

Conclusion Our results provide strong evidence of a reduced risk of ovarian cancer amongst women who breastfed for ≥6 months/child. Our results are in keeping with those from other epidemiological studies, but emphasise the importance of longer durations of breastfeeding. We found no evidence of significant heterogeneity by histological subtype.

On behalf of the Million Women Study Collaborators.