A novel predictive model for posttransplant lymphoproliferative disorder in the letermovir prophylaxis era

Abstract Introduction Posttransplant lymphoproliferative disorder (PTLD) is a complication following allogeneic hematopoietic stem cell transplantation. Its pathogenesis is primarily associated with Epstein–Barr virus (EBV) infection, with a cumulative incidence of approximately 1–5%. Patients who develop PTLD have a poor prognosis. Letermovir is a prophylactic drug against cytomegalovirus (CMV) infection that was introduced in recent years for use after allogeneic hematopoietic stem cell transplantation. It effectively reduces the incidence of posttransplant complications such as CMV viremia, CMV pneumonia, and CMV retinitis, thereby improving patient outcomes. However, reports suggest that using letermovir after allogeneic hematopoietic stem cell transplantation may increase the risk of EBV infection. Although the impact of letermovir on EBV infection remains controversial, preliminary results from our center indicate that its use may increase the risk of posttransplant PTLD. In the pre-letermovir era, Fujimoto A et al. established a scoring system based on risk factors, including antithymocyte globulin (ATG) dosage, donor‒recipient relationship, and aplastic anemia, to predict the occurrence of PTLD and developed a relatively accurate model. However, with the broader application of letermovir in clinical practice, there is a need for more precise predictive models to rapidly identify patients at high risk for PTLD. Methods A total of 2017 consecutive patients who underwent allogeneic hematopoietic stem cell transplantation at Peking University People’s Hospital from January 2020 to May 2025 were included in this study. Patients with genetic metabolic disorders, bone marrow fibrosis or early death within 28 days were excluded. Transplant protocols, including conditioning, stem cell mobilization, and acute graft-versus-host-disease (aGVHD) prophylaxis, were consistent with our previous studies. Patients were divided into a study cohort and a validation cohort according to the time of transplantation. Our aim was to develop and validate a grading scale using these cohorts to predict the risk of PTLD within the first year after HSCT. Using Fine‒Gray competing risk regression and machine learning, a dynamic prediction model for PTLD was developed. Internal validation by bootstrapping and external validation were used to assess model discrimination (C-index) and calibration. Clinical utility was evaluated through decision curve analysis, with final implementation as a risk-stratification nomogram. Results A total of1010 patients were included in the study cohort, and the remaining patients were included in the validation cohort. PTLD was diagnosed in 47 patients in the study cohort and in 53 patients in the validation cohort. A competing risk model and multivariate analysis revealed that the application of letermovir, an EBV viral load greater than 1×10^4/L, an EBV infection duration longer than 2 weeks, and an ATG dosage greater than 7.5 mg/kg were independent risk factors for PTLD in patients who underwent allo-HSCT. Based on the above results, the following scoring system was established: use of letermovir (1 point), an EBV viral load greater than 1×10^4/L (1 point), a duration of EBV infection greater than 2 weeks (1 point), and a dosage of ATG greater than 7.5 mg/kg (1 point). Risk scores were classified into groups: low risk (0 points), intermediate risk (1–2 points), and high risk (3–4 points). In the validation cohort, the probabilities for PTLD development for patients in the low-, intermediate-, and high-risk groups were 1.34%, 5.09%, and 8.97%, respectively. The score model showed acceptable discrimination, with a C statistic of 0.65 and a relatively good moderate calibration curve (HL test, p = 0.74). Conclusions Our study demonstrates a novel comprehensive model for predicting PTLD in patients undergoing allogeneic hematopoietic stem cell transplantation, in the new context of the wide range of letermovir use. This model may help clinicians quickly identify patients at high risk of PTLD and take corresponding prevention and treatment measures.