#3743 MULTIDIMENSIONAL RISK ASSESSMENT OF KIDNEY ALLOGRAFT REJECTION USING DONOR DERIVED CELL-FREE DNA

Abstract

Abstract Background and Aims Post-transplantation patient care requires development and validation of non-invasive biomarkers to improve allograft monitoring and prevention from unnecessary and costly biopsies. Reports have suggested the association of donor derived cell-free DNA (dd-cfDNA) with allograft rejection. However, there is no proof of its added value on standard of care in large, unselected and deep phenotyped cohorts. Method We enrolled 1134 kidney transplant recipients having concomitant evaluation of allograft histology, anti-HLA DSA and functional parameters between April 2013 and June 2018 in the derivation cohort, representing 1415 biopsies. Dd-cfDNA was measured in plasma at the time of the biopsy. Diagnoses were performed using Banff 2019 criteria. 171 AMR, 34 TCMR and 17 mixed rejections occurred. Parameters associated with rejection were assessed using uni- and multivariable logistic regression. We then developed a risk model using the variables that were independently associated with kidney rejection. The validation cohort comprised 1929 evaluations including 499 evaluations in one Belgian center and 1430 evaluations in nine North American centers. Results Higher levels of dd-cfDNA were observed for AMR and TCMR or both compared to other diagnoses (Fig. 1A). Dd-cfDNA incrementally increased with Banff acute lesions without significant increase for chronic lesions. In multivariable analysis, the variables independently associated with rejection were anti-HLA DSA (P<0.0001), dd-cfDNA (P<0.0001), eGFR (P<0.033), proteinuria (P = 0.016), and previous history of rejection (P<0.0001). Dd-cfDNA remained independently associated with kidney allograft rejection in validation cohorts from Belgium (P = 0.0006) and North America (P<0.0001). Discrimination of the model without dd-cfDNA was 0.777 and 0.821 with its inclusion, showing its added value (Fig. 1B). The good discrimination performances of the model with dd-cfDNA were also confirmed in the validation cohorts from Belgium (AUC: 0.815) and North America (AUC: 0.826). Conclusion We here demonstrate the independent and added value of dd-cfDNA in addition to conventional features to predict rejection. This first integrative system shows improved performance for patient monitoring and could help physicians in decision-making process.