Biomarker Panels for Discriminating Risk of CKD Progression in Children.

Abstract

BACKGROUND We have previously studied biomarkers of tubular health (EGF), injury (KIM-1), dysfunction (alpha-1 microglobulin), and inflammation (TNFR-1, TNFR-2, MCP-1, YKL-40, suPAR), and demonstrated that plasma KIM-1, TNFR-1, TNFR-2 and urine KIM-1, EGF, MCP-1, urine alpha-1 microglobulin are each independently associated with CKD progression in children. In this study, we used bootstrapped survival trees to identify a combination of biomarkers to predict CKD progression in children. METHODS The CKiD Cohort Study prospectively enrolled children 6 months to 16 years old with an eGFR of 30-90 ml/min/1.73m2. We measured biomarkers in stored plasma and urine collected 5 months after study enrollment. The primary outcome of CKD progression was a composite of 50% eGFR decline or kidney failure. We constructed a regression tree-based model for predicting the time to the composite event, using a panel of clinically relevant biomarkers with empirically derived thresholds, in addition to conventional risk factors. RESULTS Of the 599 children included, the median age was 12 years [IQR, 8 - 15], 371 (62%) were male, baseline urine protein to creatinine ratio was 0.33 [IQR: 0.12 - 0.95] mg/mg, and baseline eGFR was 53 [IQR, 40 - 66] ml/min/1.73m2. Overall, 205 (34%) children reached the primary outcome of CKD. A single regression tree-based model using the most informative predictors with data driven biomarker thresholds suggested a final set of 4 prognosis groups. In the final model, urine albumin/creatinine was the variable with the highest importance, and along with urine EGF/creatinine identified the highest risk group of 24 children, 100% of whom developed CKD progression at a median time of 1.3 years [95% CI: 1.0, 1.7]. When the regression tree-derived risk group classifications were added to prediction models including the clinical risk factors, the C-statistic increased from 0.76 [95%CI: 0.71 - 0.80] to 0.85 [95%CI: 0.81 - 0.88]. CONCLUSIONS Using regression tree-based methods, we identified a biomarker panel of urine albumin/creatinine, urine EGF/creatinine, plasma KIM-1, and eGFR which significantly improved discrimination for CKD progression.