Aims/hypothesis: We aimed to examine the association between breast-feeding and maternal risk of type 2 diabetes and to investigate whether this association is mediated by anthropometric and biochemical factors. Methods: A case–cohort study nested within the European Prospective Investigation into Cancer and Nutrition (EPIC)-Potsdam Study between 1994 and 2005 including 1,262 childbearing women (1,059 in a random sub-cohort and 203 incident cases) mainly aged between 35 and 64 years at baseline was applied. Self-reported lifetime duration of breast-feeding was assessed by questionnaire. Blood samples were used for biomarker measurement (HDL-cholesterol, triacylglycerols, C-reactive protein, fetuin-A, γ-glutamyltransferase, adiponectin). A systematic literature search and meta-analysis was conducted of prospective cohort studies investigating breast-feeding and risk of type 2 diabetes. Results: The HR for each additional 6 months of breast-feeding was 0.73 (95% CI 0.56, 0.94) in EPIC-Potsdam. Meta-analysis of three previous prospective studies and the current study revealed an inverse association between breast-feeding duration and risk of diabetes (pooled HR for lifetime breast-feeding duration of 6–11 months compared with no breast-feeding 0.89; 95% CI 0.82, 0.97). Adjustment for BMI and waist circumference attenuated the association (HR per six additional months in EPIC-Potsdam 0.80; 95% CI 0.61, 1.04). Further controlling for potentially mediating biomarkers largely explained this association (HR 0.89; 95% CI 0.68, 1.16). Conclusions/interpretation: Longer duration of breast-feeding may be related to a lower risk of diabetes. This potentially protective effect seems to be reflected by a more favourable metabolic profile; however, the role of body weight as a mediator or confounder remains uncertain.