AFFILIATIONS
1 Department of Hematology, Institute of Hematology, and Center for High Altitude Medicine, West China Hospital, Sichuan University, Chengdu, Sichuan, China
2 Department of Hematology, Institute of Hematology, and Center for High Altitude Medicine, West China Hospital, Sichuan University, Chengdu, Sichuan, China
3 Department of Hematology, Institute of Hematology, and Center for High Altitude Medicine, West China Hospital, Sichuan University, Chengdu, Sichuan, China
4 Department of Hematology, Institute of Hematology, and Center for High Altitude Medicine, West China Hospital, Sichuan University, Chengdu, Sichuan, China
5 Department of Hematology, Institute of Hematology, and Center for High Altitude Medicine, West China Hospital, Sichuan University, Chengdu, Sichuan, China
6 Department of Hematology, Institute of Hematology, and Center for High Altitude Medicine, West China Hospital, Sichuan University, Chengdu, Sichuan, China
7 Department of Hematology, Institute of Hematology, and Center for High Altitude Medicine, West China Hospital, Sichuan University, Chengdu, Sichuan, China
8 Department of Hematology, Institute of Hematology, and Center for High Altitude Medicine, West China Hospital, Sichuan University, Chengdu, Sichuan, China
9 Hematology department, Mianyang central hospital,School of Medicine.University of Electronic Science and Technology of China, Mianyang, China, Mianyang, China
10 Hematology department, Mianyang central hospital,School of Medicine.University of Electronic Science and Technology of China, Mianyang, China, Mianyang, China
11 Department of Hematology, West China School of Medicine, Sichuan University, Sichuan University affiliated Chengdu Second People’s Hospital, Chengdu Second People’s Hospital, Chengdu, China., Chengdu, China
12 Department of Hematology, West China School of Medicine, Sichuan University, Sichuan University affiliated Chengdu Second People’s Hospital, Chengdu Second People’s Hospital, Chengdu, China., Chengdu, China
13 Department of Hematology, Institute of Hematology, and Center for High Altitude Medicine, West China Hospital, Sichuan University, Chengdu, Sichuan, China
14 Department of Hematology, Institute of Hematology, and Center for High Altitude Medicine, West China Hospital, Sichuan University, Chengdu, Sichuan, China
15 Department of Hematology, Institute of Hematology, and Center for High Altitude Medicine, West China Hospital, Sichuan University, Chengdu, Sichuan, China
16 Department of Hematology, Institute of Hematology, and Center for High Altitude Medicine, West China Hospital, Sichuan University, Chengdu, Sichuan, China
17 Department of Hematology, Institute of Hematology, and Center for High Altitude Medicine, West China Hospital, Sichuan University, Chengdu, Sichuan, China
18 Department of Hematology, Institute of Hematology, and Center for High Altitude Medicine, West China Hospital, Sichuan University, Chengdu, Sichuan, China
19 Department of Hematology, Institute of Hematology, and Center for High Altitude Medicine, West China Hospital, Sichuan University, Chengdu, Sichuan, China
20 Department of Hematology, Institute of Hematology, and Center for High Altitude Medicine, West China Hospital, Sichuan University, Chengdu, Sichuan, China
21 Department of Hematology, Institute of Hematology, and Center for High Altitude Medicine, West China Hospital, Sichuan University, Chengdu, Sichuan, China
ABSTRACT
Aims:
B-cell lymphoma–associated hemophagocytic lymphohistiocytosis (HLH) is a diagnostically and therapeutically challenging condition. We aimed to identify accessible biomarkers distinguishing malignant from non-malignant HLH, develop a machine learning–integrated prognostic model, and assess its potential to guide treatment intensity and lymphoma-directed therapy.
Methods:
This multicenter retrospective cohort included consecutive patients treated at West China Hospital from January 2019 to March 2026 with HLH-onset B-cell lymphoma and an EBV-negative non-malignant HLH comparator cohort from the same period. HLH was defined by HLH-2004 criteria, and lymphoma was confirmed histologically or by bone marrow flow cytometry with clinicopathologic assessment. Overall survival (OS) was measured from HLH admission to death or last follow-up.
Missing covariates were multiply imputed. Logistic regression was used to identify admission variables distinguishing malignant from non-malignant HLH and construct an interpretable diagnostic score. LASSO-penalized Cox regression derived the HLH Admission Risk Predictor for B-cell Lymphoma (HARP-B), with age and sex retained a priori. Performance was assessed by Harrell’s C-index, time-dependent AUC, calibration, decision-curve analysis, and internal and temporal validation. XGBoost using the same variables and the optimized HLH inflammatory index (OHI) served as benchmarks; ensemble predictions were explored. Treatment timing was assessed using restricted cubic splines, a maximal log-rank cutoff, landmark analysis, and time-dependent Cox models to address time-dependent bias.
Results:
Among 181 patients, 101 had HLH-onset B-cell lymphoma and 80 had EBV-negative non-malignant HLH. Eight lymphoma patients lacked outcome data, leaving 93 for survival analysis; median overall survival (OS) was 110 days (95% CI, 61–395). For etiologic discrimination, 79 EBV-negative lymphoma-associated HLH cases were compared with 80 EBV-negative controls. Older age, higher soluble CD25 (sCD25), a higher sCD25/ferritin ratio, and lower globulin, platelet count, hemoglobin, high-density lipoprotein cholesterol were independently associated with underlying B-cell lymphoma. An interpretable 8-variable diagnostic score incorporating age, sCD25, sCD25/ferritin ratio, platelet count, hemoglobin, HDL cholesterol, globulin achieved an AUC of 0.795 and stratified patients into low-, intermediate-, and high-risk groups for occult lymphoma.
The HARP-B Score, comprising interleukin-6, procalcitonin, total cholesterol, age, and sex, achieved a C-index of 0.696 and time-dependent AUCs of 0.713–0.817 at 30–365 days. Internal validation yielded an optimism-corrected C-index of 0.670, with acceptable calibration and potential clinical benefit on decision-curve analysis; performance was attenuated in the temporally split validation cohort. HARP-B showed better time-dependent discrimination than both XGBoost and the optimized HLH inflammatory index (OHI), supporting the value of combining interpretable survival modeling with nonlinear machine-learning benchmarking. Increasing HARP-B risk was associated with progressively worse OS across multiple stratification schemes (all log-rank P<0.0001).
Ninety patients received HLH-directed therapy and 67 subsequently received lymphoma-directed therapy. Initiation of lymphoma-directed therapy within 37 days of HLH onset was associated with significantly better OS than later treatment, and adding treatment timing improved prognostic discrimination beyond baseline HARP-B.
Conclusions:
Admission-phase clinical and laboratory features may facilitate early recognition of occult B-cell lymphoma in patients presenting with HLH. The HARP-B Score provides interpretable risk stratification using inflammatory, metabolic, and host-related variables and outperformed XGBoost and the optimized HLH inflammatory index in time-dependent discrimination. Exploratory analyses further suggested that lymphoma-directed therapy initiated within 37 days of HLH onset was associated with improved survival. These findings support early recognition, admission-based risk assessment, and timely malignant trigger control, while requiring external validation.