Non-invasive alloimmune risk stratification of long-term liver transplant recipients.

Management of long-term immunosuppression following liver transplantation (LT) remains empirical. Surveillance liver biopsies in combination with transcriptional profiling could overcome this challenge by identifying recipients with active alloimmune-mediated liver damage despite normal liver tests, but has low applicability. Our aim was to investigate the utility of non-invasive tools in stratifying stable long-term LT recipients according to their immunological risk and need for immunosuppression. We conducted a cross-sectional multicentre study of 190 adult liver recipients assessed to determine their eligibility to participate in an immunosuppression withdrawal trial. Patients had stable liver allograft function and had been transplanted for non-autoimmune non-replicative viral liver disease >3 years before inclusion. We performed histological, immunogenetic and serological studies and measured the intrahepatic transcript levels of an 11-gene classifier highly specific for T cell-mediated rejection (TCMR). 36% patients harboured clinically-silent fibro-inflammatory liver lesions, which were scored as moderate-to-severe in 22%. The severity of liver allograft damage was positively associated with TCMR-related transcripts, class II donor-specific antibodies (DSA), ALT/AST and liver stiffness measurements (LSM), and negatively correlated with serum creatinine and tacrolimus trough levels. Liver biopsies were stratified according to their TCMR transcript levels using a cut-off derived from biopsies with clinically-significant TCMR. Two multivariable prediction models, integrating ALT/LSM or ALT/class II DSA, respectively, had a high discriminative capacity in classifying patients with or without alloimmune damage. The latter model had a good performance in an independent cohort of 156 liver biopsies obtained from paediatric liver recipients with similar inclusion/exclusion criteria. ALT, class II DSA and LSM are valuable tools to non-invasively identify stable LT re