Biomarker implementation: Evaluation of the decision-making impact of CXCL10 testing in a pediatric cohort.

Abstract

Children are at high risk for subclinical rejection, and kidney biopsy is currently used for surveillance. Our objective was to test how novel rejection biomarkers such as urinary CXCL10 may influence clinical decision-making to indicate need for a biopsy. A minimum dataset for standard decision-making to indicate a biopsy was established by an expert panel and used to design clinical vignettes for use in a survey. Pediatric nephrologists were recruited to review the vignettes and A) estimate rejection risk and B) decide whether to biopsy; first without and then with urinary CXCL10/Cr level. Accuracy of biopsy decisions was then tested against the biopsy results. IRA was assessed by Fleiss Kappa (κ) for binary choice and ICC for probabilities. Eleven pediatric nephrologists reviewed 15 vignettes each. ICC of probability assessment for rejection improved from poor (0.28, P<.01) to fair (0.48, P<.01) with addition of CXCL10/Cr data. It did not, however, improve the IRA for decision to biopsy (K=0.48 and K=0.43, for the comparison). Change in clinician estimated probability of rejection with additional CXCL10/Cr data was correlated with CXCL10/Cr level (r2 =0.7756, P<.0001). Decision accuracy went from 8/15 (53.3%) cases to 11/15 (73.3%) with CXCL10/Cr, although improvement did not achieve statistical significance. Using CXCL10/Cr alone would have been accurate in 12/15 cases (80%). There is high variability in decision-making on biopsy indication. Urinary CXCL10/Cr improves probability estimates for risk of rejection. Training may be needed to assist nephrologists in better integrate biomarker information into clinical decision-making.