Validation of the boscore model for predicting invasive mold disease in Chinese patients with hematological malignancies undergoing chemotherapy: A single-center retrospective study

Abstract Background: Patients with hematological malignancies (HM) are at high risk of invasive mold disease (IMD) due to chemotherapy-induced immunosuppression, with high morbidity and mortality. The BOSCORE model, a risk score predicting 60-day IMD probability in HM patients, has not been validated in Chinese populations. This study aimed to evaluate its performance in distinguishing IMD risk among Chinese HM patients receiving chemotherapy. Methods: We conducted a single-center retrospective cohort study of consecutive HM patients treated with chemotherapy at our institution (2015-2018). Inclusion criteria: diagnosis of acute leukemia (AL) or lymphoma; receipt of chemotherapy; ≥60 days of follow-up. Using the BOSCORE model, 7 risk factors were assessed: history of IMD, prednisone-equivalent ≥0.5mg/kg within 30 days, uncontrolled underlying malignancy, high-risk chemotherapy, absolute neutrophil count (ANC) <100/µl for >10 days, absolute lymphocyte count (ALC) <50/µl, and cytomegalovirus (CMV) infection. Patients were stratified into high-risk (BOSCORE probability ≥5%) and low-risk (<5%) groups. The primary endpoint was confirmed/suspected IMD within 60 days. Kaplan-Meier analysis and Cox regression were used for statistical evaluation. Results: A total of 461 patients were included (median age 47 years; 245 males, 216 females), including 325 (70.5%) with AL and 136 (29.5%) with lymphoma. Over 60 days, 41 (8.9%) developed IMD: 40 (12.3%) in AL and 1 (0.7%) in lymphoma. Per BOSCORE stratification: 374 (81.1%) were low-risk, 87 (18.9%) high-risk. High-risk patients had a significantly higher IMD incidence than low-risk patients (20.7% vs 6.1%, p<0.001). Among AL patients, 25.9% were high-risk (vs 2.2% in lymphoma, p<0.001), consistent with deeper chemotherapy-induced myelosuppression in AL. Kaplan-Meier analysis showed a rapid decline in IMD-free survival in the high-risk group (p<0.001 overall; p=0.004 in AL subgroup). Multivariate Cox regression identified ANC <100/µl for >10 days as the only independent IMD predictor (HR=2.67, p=0.0086). Conclusion: The BOSCORE model effectively distinguishes high- vs low-risk IMD in Chinese HM patients, with high-risk patients showing significantly higher IMD incidence. It performs particularly well in AL patients with chemotherapy-induced profound neutropenia. Validation in multi-center prospective studies is needed to confirm its utility, with potential subgroup analyses for AL and lymphoma to optimize risk stratification and targeted antifungal prophylaxis.