A simplified prediction model for end-stage kidney disease in patients with diabetes

Abstract

This study aimed to develop a simplified model for predicting end-stage kidney disease (ESKD) in patients with diabetes. The cohort included 2549 individuals who were followed up at Kyushu University Hospital (Japan) between January 1, 2008 and December 31, 2018. The outcome was a composite of ESKD, defined as an eGFR < 15 mL min^-1 [1.73 m]^-2, dialysis, or renal transplantation. The mean follow-up was 5.6 [Formula: see text] 3.7 years, and ESKD occurred in 176 (6.2%) individuals. Both a machine learning random forest model and a Cox proportional hazard model selected eGFR, proteinuria, hemoglobin A1c, serum albumin levels, and serum bilirubin levels in a descending order as the most important predictors among 20 baseline variables. A model using eGFR, proteinuria and hemoglobin A1c showed a relatively good performance in discrimination (C-statistic: 0.842) and calibration (Nam and D'Agostino [Formula: see text]² statistic: 22.4). Adding serum albumin and bilirubin levels to the model further improved it, and a model using 5 variables showed the best performance in the predictive ability (C-statistic: 0.895, [Formula: see text]² statistic: 7.7). The accuracy of this model was validated in an external cohort (n = 5153). This novel simplified prediction model may be clinically useful for predicting ESKD in patients with diabetes.

Figure 1

Observed vs predicted probabilities of end-stage kidney disease (ESKD) events at a 5-year risk in the development cohort and the external validation cohort. The predicted (white bar) and observed (black bar) event probabilities represent the mean predicted probability calculated from 5-year risk equations and the mean observed probability from the patients divided into deciles of the predicted probability, respectively. (a) Model 3 and (b) Model 5 in the development cohort. (c) Model 5 in the validation cohort and (d) Model 5 in the patients with chronic kidney disease (CKD) (eGFR < 60 mL min⁻¹ [1.73 m]⁻² and/or positive proteinuria) (n = 1350) of the validation cohort. Nam and D’Agostino $\chi$² statistics were 22.4 and 7.7 for Models 3 and 5 in the development cohort, and 36.1 and 23.1 for Models 5 in the external validation cohort and Model 5 in the patients with CKD of the external cohort, respectively.

Figure 2

Nomogram for end-stage kidney disease (ESKD)-free event probabilities of individuals with diabetes. To use the nomogram. Locate an individual’s value on each variable axis, and draw a line upward to obtain the point for each variable. Then, locate the sum of these points on the total points axis, and draw a line downward to the event-free axis to obtain the 5-year ESKD-free probability. eGFR estimated glomerular filtration rate, HbA1c hemoglobin A1c.