Aims/hypothesis We examined the association between serum C-reactive protein (CRP) and incident diabetes in a prospective study, and added these data to a literature-based meta-analysis to explore potential sources of heterogeneity between studies. Methods We analysed a case-control study nested within the European Prospective Investigation of Cancer (EPIC)-Norfolk cohort, including 293 incident diabetes cases and 708 controls. We combined 16 published studies on CRP and incident diabetes in a random-effect meta-analysis. Results In the EPIC-Norfolk cohort, serum CRP was associated with a higher risk of diabetes after adjusting for age, sex, BMI, family history of diabetes, smoking and physical activity (OR 1.49, comparing the extreme thirds of CRP distribution [95% CI 1.03-2.15], p=0.03). However, the association was completely attenuated after further adjustment for WHR, serum γ-glutamyltransferase and serum adiponectin (OR 1.00; 95% CI 0.66-1.51, p=1.0). In a meta-analysis of 16 published studies with 3,920 incident diabetes cases and 24,914 controls, the RR was 1.72 (95% CI 1.54-1.92), comparing the extreme thirds of CRP distribution, with substantial heterogeneity between studies (I 2 =52.8%, p=0.007). Conclusions/interpretation Initial evidence of association between CRP and incident diabetes was confounded by central adiposity, markers of liver dysfunction and adiponectin in the primary analysis. Despite an overall positive association in the meta-analysis, considerable heterogeneity existed between studies. The degree of adjustment for central adiposity and baseline glycaemia explained some of this heterogeneity and suggests that CRP may not be an independent risk factor for type 2 diabetes.