Abstract This study aimed to investigate the relationship between age at diabetes diagnosis (ADD) and the development and progression of diabetic retinopathy (DR). This study combined analysis of the National Health and Nutrition Examination Survey (NHANES) data with Mendelian randomization (MR). We employed regression models, propensity score matching, generalized additive model smoothing splines (GAM), random forest algorithms, and other analytical techniques. Cross-sectional analysis of NHANES data (n = 877) showed that the DR group had a significantly younger ADD compared to the DR-free group (P