Pathogenetic markers of diabetic kidney disease (DKD) progression to ESRD are lacking. We characterized the prognostic value of histologic findings in DKD for time to ESRD in native kidney specimens from biopsies performed from 1995 to 2011 with diabetic glomerulosclerosis as the only glomerular disease diagnosis (=109). Biopsy specimens were analyzed according to standard methods, including determination of diabetic nephropathy class, as defined by the Renal Pathology Society. Clinical data were extracted from electronic medical records. We used competing risk models, with death as the competing risk, to estimate subdistribution hazard ratios (HRs) for ESRD. All multivariable models included age, sex, black race, baseline eGFR, and baseline proteinuria. Pathologic characteristics achieving <0.1 were added into successively complex models. ESRD occurred in 56% of patients, and 26% of patients died before reaching ESRD. In univariate analyses, diabetic nephropathy class was not statistically significant in predicting time to ESRD. The final multivariable model (=106) showed a borderline association between mild mesangial expansion and decreased risk for ESRD (subdistribution HR, 0.64; 95% confidence interval, 0.40 to 1.00). Poor prognostic factors in the final model included segmental sclerosis and extracapillary hypercellularity (subdistribution HR, 2.04; 95% confidence interval, 1.36 to 3.05; and subdistribution HR, 2.21; 95% confidence interval, 1.19 to 4.11, respectively). In conclusion, we identified segmental sclerosis and extracapillary hypercellularity as novel, poor prognostic indicators of time from DKD to ESRD. Whether these indicators represent a distinct pathogenetic phenotype of DKD will require a large study with a broad spectrum of disease severity.