Introduction: Reliable, non-invasive biomarkers or rejection and under-immunosuppression will aid in personalized immunosuppression management. Combining Gene Expression Profiles (GEP) and dd-cfDNA in the blood improved diagnostic performance for rejection. We tested the hypothesis that combining Urine CXCL9 and CXCL10, both established biomarkers of rejection, with GEP and dd-cfDNA would further improve diagnostic performance for rejection in kidney transplant patients.
Methods: We studied 344 biopsy-paired samples taken from 214 pts enrolled in the CTOT-08 trial. Biopsy based clinical phenotypes included clinical acute rejection (cAR, N=26), subclinical acute rejection (subAR, N=60), acute dysfunction but no rejection (ADNR, N=47) and Normal surveillance biopsies (TX, N=211). Urine CXCL9/CXCL10 values were normalized by dividing by urine creatinine from the same sample. A multi-variable logistic regression model was built including the log_cfDNA values, TruGraf v1.3 microarray probability score (0-100), and the urine CXCL9/Cr, and CXCL10/Cr values. AUC values were validated by bootstrap resampling.
Results: Results are shown in Figure 1, which examined all rejections (cAR + subAR, n=86) compared to no rejection (ADNR + TX, n=258). The combined biomarker algorithm yielded an AUC = Area under the curve: 0.853 with a bootstrap resampling validated AUC: 0.844 suggesting minimal overfitting. This gave an NPV=87% and PPV=63%.In the surveillance biopsies only (subAR vs. TX), the AUC=0.865.In the for-cause biopsies (cAR vs. ADNR), the AUC=0.879.

Conclusion: Combining established biomarkers in the blood and urine using a logistic regression model yields excellent diagnostic performance for diagnosing or excluding rejection in patients with both surveillance and for-cause biopsies. Further validation is needed to establish the performance of this biomarker combination in additional cohorts.