Background This study investigated the association between birth weight and diabetes in a Chinese population, and the effects of body mass index (BMI) and lifestyle factors in later life on this association. Methods Data from 49 118 participants aged >= 40 years with recalled birth weight from the Risk Evaluation of cAncers in Chinese diabeTic Individuals: a lONgitudinal (REACTION) study, a nationwide population-based cohort, were used. Diabetes diagnosis was based on oral glucose tolerance tests and HbA1c measurements. Logistic regression models were used to evaluate the association of birth weight and risk of diabetes in later life. Results Increased risk of diabetes was associated with lower or higher birth weight. Compared with individuals with a birth weight of 2500 to 3499 g, the odds ratios (ORs) and 95% confidence intervals (CIs) of diabetes for individuals with a birth weight of <2500, between 3500 and 3999, and >= 4000 g were 1.28 (1.11-1.47), 1.11 (1.04-1.19), and 1.20 (1.07-1.34), respectively. Significant associations were prominent in participants with a current BMI >= 24 kg/m(2), but not detected in those with a normal BMI (OR 1.20 [95% CI 0.96-1.49], 1.11 [95% CI 0.98-1.25], and 1.10 [95% CI 0.89-1.37], respectively). Moreover, there was no increased risk of diabetes in individuals with a low birth weight but with healthy dietary habits (OR 0.94; 95% CI 0.68-1.29) or ideal physical activity (OR 1.41; 95% CI 0.97-2.04). Conclusions A U-shaped association was observed between birth weight and the risk of diabetes. Healthy lifestyles (healthy dietary habits or ideal physical activity) may eliminate the negative effects of low birth weight in the development of diabetes, but not the effect of high birth weight.