Improving diagnostic performance of kidney allograft rejection with a model combining relative fraction and absolute copies of donor-derived cell-free DNA—Results from 5 independent cohorts

Abstract

Donor-derived cell-free DNA (dd-cfDNA) is a widely used biomarker in kidney transplantation, reported mainly as either a percentage of total cfDNA or copies per milliliter (cp/mL). We hypothesized that combining both metrics into a continuous composite score would reduce false classifications, thus improving the accuracy of rejection diagnosis. We analyzed 443 dd-cfDNA measurements in 361 patients from 5 independent kidney transplant cohorts with both percentage and cp/mL. A random discovery set of 25 biopsy-proven rejections and 26 nonrejections was used to derive a composite score (CM-Score), integrating dd-cfDNA (%) and cp/mL. Diagnostic performance was validated in 81 rejections and 282 nonrejections and compared to 11 published cohorts. For all evaluations, CM-Score was better than dd-cfDNA in cp/mL, which was better than dd-cfDNA in percent. The CM-Score retained a high NPV of 91% (89%-93%), while significantly improving PPV to 81% (72%-89%; P < .0001, prevalence: 25%), compared to published values (weighted average NPV: 90% [89%-92%]; PPV: 54% [50%-57%], N = 6536). Decision curve analysis yielded a significantly higher net benefit for the CM-Score (P < .003). The dd-cfDNA CM-Score is a robust and clinically meaningful tool improving diagnostic accuracy for ruling-out and ruling-in rejection using dd-cfDNA as a stand-alone metric with high potential to improve decision-making in kidney transplantation.