Table 4.

Risk prediction in models of eGFR decline with and without baseline biomarker concentration and change over 6 mo in participants of the Examination of Cardiovascular Outcomes with Alogliptin versus Standard of Care (EXAMINE) trial

VariableC-StatisticGoodness of Fit Chi SquaredNet Reclassification Index (95% CI Bootstrap)Integrated Discrimination Index (95% CI Bootstrap)
Base modela0.935.13
Plus baseline biomarker
 uKIM-1/Cr0.935.14−0.04 (−0.11 to 0.00)−0.00 (−0.00 to 0.00)
 uNGAL/Cr0.933.550.0215 (0.00 to 0.07)0.00 (−0.00 to 0.01)
 Cystatin- C0.945.31−0.04 (−0.16 to 0.07)0.07 (0.03 to 0.13)
Base model0.973.87
Plus change in biomarker over 6 mo
 uKIM-1/Cr0.973.760.00 (−0.00 to 0.00)0.00 (−0.00 to 0.01)
 uNGAL/Cr0.973.760.00 (−0.00 to 0.00)0.00 (−0.01 to 0.03)
  • The base model included nine predictors: age, sex, race, ethnicity, history of hypertension, systolic BP, hemoglobin A1c, Chronic Kidney Disease Epidemiology Collaboration eGFR, and baseline proteinuria. All predictors with missing values were estimated through the Markov Chain Monte Carlo method with five-time imputation. Time from randomization to end point was fit using the Cox proportional hazards model with stratification per treatment and adjustment on the baseline variables listed above. Subjects with eGFR decline within 180 d (6 mo) were excluded for the models of biomarker change over 6 mo (n=32). 95% CI, 95% confidence interval; uKIM-1/Cr, urinary kidney injury molecule–1–to-creatinine; uNGAL/Cr, urinary neutrophil gelatinase-associated lipocalin–to-creatinine.

  • a The base model for the change in biomarkers included the nine traditional predictors plus the baseline biomarker concentrations.