Clinical features and prognostic model for adenovirus infection following allogeneic hematopoietic stem cell transplantation: Insights from a Chinese cohort

Abstract

Abstract Introduction Adenovirus (AdV) infection is a potentially life-threatening complication following allogeneic hematopoietic stem cell transplantation (allo-HSCT), especially under profound immunosuppression. AdV can involve multiple organs—most notably the lungs, gastrointestinal tract, and liver—and contributes significantly to nonrelapse mortality. While previous studies have addressed AdV pneumonia, a comprehensive analysis of AdV infections, viremia, and organ-specific diseases across a full spectrum of posttransplant manifestations is lacking, particularly in Chinese populations, hindering early risk identification and intervention. Therefore, we aimed to develop and validate a prognostic model for AdV-related mortality after allo-HSCT based on a Chinese cohort, using clinical variables to enable early identification of high-risk patients and support timely intervention decisions. Methods We retrospectively analyzed consecutive allo-HSCT recipients diagnosed with AdV infection from April 2012 to June 2024 at Peking University Institute of Hematology. Patients (n=323) were randomized into training (60%, n=195) and validation cohorts (40%, n=128). Baseline characteristics were comparable (all P > 0.05). Prognostic variables were selected via LASSO-penalized Cox regression, and significant predictors (P < 0.05) were integrated into a weighted scoring model derived from β-coefficients. Model discrimination, calibration, and clinical utility were evaluated using time-dependent AUC, calibration plots, and decision curve analyses (R 4.2.2; SPSS 25). Results Among 3,560 allo-HSCT recipients, the cumulative incidence of AdV infection was 9.1% (n=323). The median age of these patients was 35 years (IQR 24–62), and the median interval from transplant to AdV onset was 168 days (range 20–1205). Overall mortality rates following AdV onset were 15.8% (n = 51) at 180 days, 17.0% (n = 55) at 360 days, and 19.2% (n = 62) by the end of follow-up (2570 days). A total of 323 patients were included in the study and randomly divided into training (n = 195) and test (n = 128) cohorts at a 60:40 ratio. No significant differences in baseline demographics or clinical characteristics were observed between the training and validation cohorts (all P > 0.05), confirming the suitability of the model for development and validation. In the univariate Cox regression analysis, delayed viral clearance (>21 days) (HR = 7.20, 95% CI: 4.02–12.90; P < 0.001), early AdV onset (<258 days) (HR = 2.68, 95% CI: 1.13–6.34; P = 0.025), week-2 absolute lymphocyte count (ALC) < 0.56×10⁹/L (HR = 2.36, 95% CI: 1.21–4.61; P = 0.012), gastrointestinal involvement (HR = 2.27, 95% CI: 1.28–4.02; P = 0.005), CMV coinfection (HR = 2.66, 95% CI: 1.48–4.76; P = 0.001), EBV coinfection (HR = 2.00, 95% CI: 1.08–3.70; P = 0.027), and graft failure (HR = 2.39, 95% CI: 1.35–4.23; P = 0.003) significantly predicted increased mortality. Conversely, a higher infused CD34⁺ cell dose (≥2.88 × 10⁶/kg) showed a protective effect (HR = 0.46, 95% CI: 0.25–0.86; P = 0.015). LASSO regression identified three independent prognostic factors confirmed by multivariable Cox regression: early onset (HR=2.68, 95% CI: 1.13–6.34; P =0.025), delayed clearance (HR = 4.88, 95% CI: 2.66 – 8.94; P < 0.001) and low week-2 ALC (HR = 2.36, 95% CI: 1.21 – 4.61; P = 0.012). Based on these results, a weighted scoring model was constructed, as follows: Y = 2.5a + 2b + 5c, where variables a, b, and c indicate the presence of early AdV onset, delayed viral clearance, and low week-2 lymphocyte count, respectively. The model demonstrated strong discriminatory capacity; time-dependent AUCs at 180, 270, and 360 days were 0.888, 0.878, and 0.864 (training), and 0.889, 0.847, and 0.834 (validation). Corresponding C-indices were 86.3%, 84.7%, and 82.0% (training); 83.3%, 79.0%, and 77.8% (validation). Calibration curves demonstrated strong consistency between predicted and observed mortality. Decision curve analysis demonstrated favorable clinical net benefit across a range of threshold probabilities. A nomogram was established to facilitate individualized clinical risk prediction. Conclusions In this study, we established and validated a prognostic model for adenovirus-related mortality after allo-HSCT. The model demonstrated strong discrimination and clinical utility, enabling early identification of high-risk patients and informing timely therapeutic decisions in transplant settings.