Prognostic analysis and predictive model construction of bleeding events in allogeneic hematopoietic stem cell transplant patients

Abstract

Objective: To study hematopoietic stem cell transplantation-related bleeding prognosis and construct a bleeding prediction model. Methods: The clinical data of 555 patients with malignant hematologic diseases who underwent allogeneic hematopoietic stem cell transplantation between May 1(st) 2004, and April 1(st) 2012 was analyzed retrospectively, and a prediction model was constructed. Results: Of the 555 patients, a total of 302 (54.0% ) patients exhibited bleeding events of varying degrees, including 151 (27.0% ) with grade Ⅰ bleeding, 63 (11.0% ) with grade Ⅱ bleeding, 48 (9.0% ) with grade Ⅲ bleeding, and 40 (7.0% ) with grade Ⅳ bleeding. Multifactorial analysis showed that the overall mortality (HR=12.53, 95% CI 7.91-19.87, P<0.001) and non-recurrence mortality (HR=23.79, 95% CI 12.23-46.26, P<0.001) were higher in patients with higher bleeding grades (Ⅲ and Ⅳ bleeding) compared to those with lower bleeding grades. Additionally, the donor's underlying disease, graft-versus-host disease (GVHD) score, poor platelet reconstitution, and ineffective platelet transfusion were independently associated with bleeding risk. The bleeding model constructed using the above variables showed good accuracy (C-Index=0.934) , and its efficacy was significantly higher than previous bleeding models. Conclusion: Hematopoietic stem cell transplant patients are at increased risk of death after a bleeding event. The cross-validated bleeding risk prediction model is valuable for early intervention.