Introduction: Corticosteroid-refractory graft-versus-host disease (GVHD) remains a serious complication of hematopoietic stem cell transplantation (HSCT) with high morbidity and mortality rates. Unfortunately, no standard therapy exists for this setting. Ruxolitinib (ruxo), an oral selective Janus-associated kinase (JAK) inhibitor, achieved good results for corticoresistant acute and chronic GVHD in preclinical and clinical studies, with 80% overall response rates. Recent studies showed an increased risk of infections in patients treated with ruxo, especially Epstein-Barr virus (EBV) and cytomegalovirus (CMV) reactivations. Patients and methods: In order to assess the efficacy and the safety of ruxo, we reviewed here the outcome of 57 patients who received ruxo for corticosteroid-refractory GVHD (in case of digestive tract involvement or in the absence of available ongoing clinical trials) or as a steroids-sparing medication. EBV and CMV reactivation risks were also assessed in the 137 consecutive patients received HSCT between January 1, 2012 and December 31, 2017 and who presented acute or chonic GVHD (57 of whom received ruxo). For this purpose, each reactivation were analyzed separately as a competing risk with death in a cause specific Cox model of survival after the first GVHD occurrence and the onset of ruxo therapy was coded as time dependent covariate. Results: The median age of 57 patients with ruxo was 55 years (range, 49 to 61). Indication for HSCT was acute myeloid leukemia for 25% patients, lymphoma for 30%, acute lymphoblastic leukemia for 11%, myeloproliferative neoplasm for 14% and myelodysplastic syndrome for 7%. Only 9% of this patients received ruxo before HSCT. Unrelated donor was used for 60% patients and main source of hematopoietic stem cells was peripheral stem cells (93%). T cell depletion with polyclonal anti-thymocyte globulin was performed for 89% of patients. Conditioning with high doses cyclophosphamide was used for 21% patients. A lymphopenia <1G/L persisted at day 100 for 42/51 patients evaluated. In these 57 patients, ruxo was given for 62 episode of GVHD (acute: 21 [which 95% for an acute grade III-IV GVHD]; chronic: 36; overlap syndrome: 5). Response rates to ruxo were 48% for acute GVHD, 60% for overlap syndrom and 58% for chronic GVHD. The overall GVHD related-death rate was 33% (19% for acute GVHD and 14% for chronic GVHD). Our median follow-up was 30 months (range, 9 to 42) after HSCT. Main non-infectious adverse events were cytopenias (17/57) and hepatic cytolysis (6/57) leading to discontinuation or tapering of ruxo in 12 patients. EBV reactivation (> 4 log or increasing viral charge of 0.5 log) occurred in 19 patients after ruxo with a 6-weeks cumulative incidence (6WCuI) of 22% (95CI [95% confidence interval]: 15-34). It was the first reactivation in 13 patients. CMV reactivation (> 3 log) occurred in 8 patients after ruxo with a 6WCuI of 4% (95CI: 7-25). It was the first reactivation in 3 patients (Figure 1). The distribution of first reactivation before ruxo and in the remaining 80 patients is shown in Table 1. Thus, 6WCuI of first EBV and CMV reactivation after the first episode of GVHD was 24% (95CI: 19-31) and 20% (95CI: 15-25) respectively. Finally, onset of ruxo coded as time dependent covariates retained a significant adverse prognostic value for the competing risks of death and first episode of EBV reactivation (HR [Hazard Ratio]: 2,657, p<0,05) as well as first episode of CMV reactivation (HR: 1,747, p<0,05) after first episode of GVHD. Discussion /Conclusion: Ruxo initiation coded as a time-dependent covariate was significantly associated with the overall risk of viral reactivation after the first episode of GVHD and the viral reactivations incidences after ruxo were similar with the incidence of reactivation at the onset of GVHD. Thus, it might be linked to the immuno-compromised state induced by both HSCT and GVHD. Furthermore, as we used ruxo for serious gastro intestinal GVHD, we could have selected patients with a more severe GVHD, requiring multiple immunosuppressive therapy, worsening immune reconstitution. To our knowledge it is the first study to assess the competing risk of CMV and EBV reactivation during ruxo treatment for GVHD. Given its effectiveness in corticoresistant GVHD, ruxo use must not be limited by the fear of viral reactivation at the light of our data, conditioned upon a close monitoring of viral loads in the first weeks. Disclosures No relevant conflicts of interest to declare.