In the manuscript, the events of nephropathy and retinopathy recorded for the LEADER trial (Marso et al. 2016) were erroneously imputed; the numbers reported in Table 1 for that trial refer to serious adverse events, and not to total events, as stated in Methods. The correct figures are 268 and 416 cases of nephropathy, and 106 and 92 cases of retinopathy in liraglutide and control arms, respectively. As a consequence, the overall risk (MH-OR [95% CI]) of nephropathy and retinopathy with GLP-1RA (vs all comparators) is 0.77 [0.67-0.88], p < 0.001, and 0.99 [0.80-1.21], p = 0.90, respectively. The analyses reported in Figs. 1 and 2 were also corrected and reported below, together with supplementary figures 3 and 5. Please note that liraglutide is associated with a significant reduction in the incidence of nephropathy, whereas for retinopathy the difference from placebo is no longer statistically significant.