Real-world hematological data from 714 DLBCL patients treated with R‑CHOP: A multicenter retrospective cohort study for predicting severe neutropenia

Dataset Description 1. Background and Research Hypothesis This dataset supports a multicenter retrospective cohort study investigating cumulative myelosuppression in diffuse large Bcell lymphoma (DLBCL) patients receiving RCHOP chemotherapy. Current risk models focus on baseline factors or firstcycle events, but a critical gap exists in predicting severe neutropenia in later cycles (e.g., cycle 4), which often leads to dose delays and increased infection risk. Hypothesis: Integrating baseline patient characteristics with early dynamic responses from cycle 1 (hematological nadirs and inflammatory markers) enables development of an accurate, explainable AI model to prospectively identify patients at high risk of grade 4 neutropenia at cycle 4. 2. Data Content and Variables This dataset contains deidentified clinical and laboratory data from 714 adult patients with newly diagnosed DLBCL who received firstline RCHOPlike chemotherapy at two Chinese tertiary centers (2018–2025). The data include 34 variables per patient, organized as: Demographics & Baseline: Age, gender, diabetes, B symptoms (fever, night sweats, weight loss), Ann Arbor stage, ECOG, bone marrow involvement, bulky disease (≥5cm), EBVDNA status. Prognostic Indices: IPI score and risk category (Low/Lowintermediate/Highintermediate/High). Baseline Labs: LDH, beta2 microglobulin. Cycle 1 Nadir (Predictors): RBC, WBC, platelet, hemoglobin, ANC; inflammatory markers (CRP, PCT, albumin, ferritin, AST, ALT, IL6, IL10). Cycle 1 Outcomes: Grade 4 neutropenia (ANC