Development and prospective validation of a prognostic model for diffuse alveolar hemorrhage after allogeneic hematopoietic stem cell transplantation Free

Introduction Diffuse alveolar hemorrhage (DAH) is a life-threatening pulmonary complication that can occur after allogeneic hematopoietic stem cell transplantation (allo-HSCT) without an explicit etiology, and there is no standard treatment. Despite rapid administration of prophylactic antibiotics, systemic glucocorticoid therapy and supportive care, the poor prognosis of DAH patients has not improved. To date, the mortality rate of DAH remains as high as 80–100%. Early therapy is critical for eliciting better patient outcomes. However, predictors of early mortality in patients who develop DAH after allo-HSCT have not been well investigated. In this study, we used the largest nationwide cohort of DAH patients after allo-HSCT to establish a prognostic model capable of identifying patients at increased risk of early death.

Methods This is an observational, retro-prospective study looking at the early mortality of DAH following allo-HSCT. From July 2013 to December 2021, 160 adult patients with a diagnosis of DAH from Peking University Institute of Hematology were retrospectively included to develop the prognostic model. Validation was performed using the prospective cohort. A total of 83 patients from January 2022 to June 2025 composed the prospective validation cohort. The outcome of interest was 28-day mortality. We assessed 26 widely available clinical parameters (demographic, transplant, clinical, and laboratory factors) as candidate predictors and used multivariable logistic regression to develop our prediction model. A scoring system to predict the prognosis of DAH after allo-HSCT was also established, and scores were assigned to the prognostic factors based on the regression coefficient.

Results In the entire cohort, the overall incidence of DAH after allo-HSCT was 2.5%. The median age was 35 (IQR: 24–62) years. The median time from bone marrow transplantation to the diagnosis of DAH was 168 (20–1205) days. On the day of DAH diagnosis, 92 (37.9%) patients had grade I–II aGVHD, and 27 (11.1%) patients had grade III–IV aGVHD. A total of 29 (11.9%) patients had extensive chronic GVHD. The baseline clinical characteristics of patients in the validation cohort were different from those of patients in the derivation cohort; there were more patients with peripheral blood as a stem cell source (P = 0.001), there were fewer patients with pneumonia (P = 0.015), and CRP levels were lower (P = 0.001).

The prognosis of patients with DAH after transplantation was extremely poor. Overall, 195 patients (80.2%), 219 patients (90.1%), and 224patients (92.2%) died within 28 days, 60 days, and 100 days after the diagnosis of DAH, respectively. The 28-day mortality rates were 82.1% and 78.0% in the derivation cohort and validation cohort, respectively. In total, 26 candidate predictor variables were collected upon DAH diagnosis. Using multivariable logistic regression methods, respiratory failure (OR, 2.823; 95% CI, 1.495-4.530), patient age (OR, 2.381; 95% CI, 1.312-4.238), cGVHD (OR, 2.176; 95% CI, 1.492-5.189), and a platelet count less than 20*10⁹/L (OR, 2.376; 95% CI, 1.351-5.013) were identified as independent prognostic factors for 28-day mortality in patients with DAH. A risk score model termed RACE (including respiratory failure, age, cGVHD, and low platelet count) was constructed according to the regression coefficients. Patients were stratified into a low-risk group (0–2 points), an intermediate-risk group (3–4 points) and a high-risk group (5–6 points). The RACE model showed an AUC of 0.868 (95% CI 0.828–0.909) in the internal validation cohort and 0.785 (95% CI 0.693–0.917) in the prospective validation cohort. The calibration plots demonstrated good agreement between the RACE-predicted probabilities and the actual observations on both internal and external validation. Decision curve analysis demonstrated the clinical usefulness of the RACE model.

Conclusions The RACE score is the first scoring system capable of predicting the 28-day survival of patients with DAH after allo-HSCT. It showed good performance in identifying allo-HSCT patients at increased risk of 28-day mortality after DAH diagnosis. This predictive model may contribute to the early identification of high-risk patients and facilitate future studies on developing individualized and novel interventions for patients within different risk groups.