Supplementary Material for: A Pathological Scoring System to Predict Renal Outcome in Diabetic Nephropathy

Background: With the association between diabetic nephropathy (DN) and renal outcome being increasingly clear, we aimed at creating a new DN pathological scoring system that could predict the renal outcome. Methods: We studied 205 patients with DN confirmed by renal biopsy, sometime between March 1985 and January 2010, who met the inclusion criteria. Renal biopsy included clinical parameters and Tervaert classifications. Hazard ratios (HRs) for death-censored end-stage renal disease (ESRD) were estimated by adjusted Cox proportional-hazards regression. The overall pathological risk score (D-score) was calculated by summing the products of beta coefficient and bootstrap-inclusion fractions, its predictive utility evaluated by Kaplan-Meier methods and c-statistics for a 10-year risk of ESRD. Results: The D-scores of glomerular classes 1, 2A, 2B, 3, and 4 were, respectively, 0, 3, 4, 6, and 6. Those of interstitial fibrosis and tubular atrophy classes 0, 1, 2, and 3 were 0, 7, 9, and 11, and those of interstitial inflammation classes 0, 1, and 2 were 0, 3, and 4, respectively. The D-score of hyalinosis class 2 was 3 and that of arteriosclerosis class 2 was 1. So, a patient's D-score could be 0-25. HRs for ESRD in patients with D-score ≤14, 15-18, 19-21, and 22-25 were, respectively, 1.00 (reference) 16.21 (95% confidence interval (CI), 1.86-140.90), 19.78 (95% CI, 2.15-182.40), and 45.46 (95% CI, 4.63-446.68) after adjusting for clinical factors. The c-statistics suggested a better predictive ability for a 10-year renal death with models that included the D-score. Conclusion: Prediction of DN patients' renal outcome was better with the D-score than without it. Patients with a D-score ≤14 had excellent renal prognosis.