Different Kinetics and Risk Factors for Isolated Extramedullary Relapse after Allogeneic Hematopoietic Stem Cell Transplantation in Children with Acute Leukemia

•Isolated extramedullary relapse (iEMR) is a frequent and significant relapse event.•Compared with systemic relapse, iEMR manifests clinically after systemic relapse.•Central nervous system iEMR is the most common subtype of iEMR in both acute lymphoblastic leukemia and acute myelogenous leukemia.•Risk factors, pathophysiology, and outcomes of iEMR differ from those of systemic relapse.•Current evidence shows the use of total body irradiation in HSCT conditioning may be more effective at preventing post-HSCT relapse than chemotherapy alone-based conditioning. Relapse after allogeneic hematopoietic stem cell transplantation (allo-HSCT) remains the most frequent cause of post-transplantation mortality. Isolated extramedullary (EM) relapse (iEMR) after HSCT is relatively rare and not well characterized, particularly in pediatric patients. We retrospectively analyzed 1527 consecutive pediatric patients with acute leukemia after allo-HSCT to study the incidence, risk factors, and outcome of iEMR compared with systemic relapse. The 5-year cumulative incidence of systemic relapse (either bone marrow [BM] only or BM combined with EMR) was 24.8%, and that of iEMR was 5.5%. The onset of relapse after allo-HSCT was significantly longer in EM sites than in BM sites (7.19 and 5.58 months, respectively; P = .013). Complete response (CR) 2+/active disease at transplantation (hazard ratio [HR], 3.1; P < .001) and prior EM disease (HR, 2.3; P = .007) were independent risk factors for iEMR. Chronic graft-versus-host disease reduced the risk of systemic relapse (HR, 0.5; P = .043) but did not protect against iEMR. The prognosis of patients who developed iEMR remained poor but was slightly better than that of patients who developed systemic relapse (3-year overall survival, 16.5% versus 15.3%; P = .089). Patients experiencing their first systemic relapse continued to have further systemic relapse, but only a minority progressed to iEMR, whereas those experiencing their iEMR at first relapse developed further systemic relapse and iEMR at approximately similar frequencies. A second iEMR was more common after a first iEMR than after a first systemic relapse (58.8% versus 13.0%; P = .001) and was associated with poor outcome. iEMR has a poor prognosis, particularly after a second relapse, and effective strategies are needed to improve outcomes.