Prognostic factors and nomogram development for adult diffuse large B-cell lymphoma: a single-center retrospective cohort study.

BACKGROUND

Diffuse large B-cell lymphoma (DLBCL) exhibits substantial prognostic heterogeneity despite standardized treatment approaches. Integrating clinical, metabolic, and molecular biomarkers into a unified prediction model may enhance risk stratification beyond conventional indices.

OBJECTIVE

To identify independent prognostic factors and develop an interanlly validated nomogram for predicting mortality in adult DLBCL patients.

METHODS

This retrospective cohort study enrolled 150 consecutive adult DLBCL patients diagnosed between January 2016 and December 2022 at a Cancer Hospital Affiliated to Shanxi Medical University. Primary endpoint was all-cause mortality assessed through August 2025. Clinical characteristics, serum biomarkers (lactate dehydrogenase [LDH], β2-microglobulin [β2-MG]), ¹⁸F-FDG PET/CT metabolic parameters (SUVmax, metabolic tumor volume [MTV], total lesion glycolysis [TLG]), and immunohistochemical markers (Bcl-2, Bcl-6, C-MYC) were analyzed. Notably, all six prognostic variables represent standard-of-care assessments routinely performed in contemporary hematology practice. Multivariable logistic regression with supplementary Cox proportional hazards regression identified independent predictors, which were integrated into a nomogram. Model performance was evaluated using discrimination, calibration metrics, and decision curve analysis.

RESULTS

After median follow-up of 54 months, 61 patients (40.7%) died. Among deaths, 54 (88.5%) were attributable to disease progression, 4 (6.6%) to treatment-related complications, and 3 (4.9%) to unrelated causes. Multivariable analysis identified six independent prognostic factors: bone marrow invasion (OR = 3.54; 95%CI:1.34–9.36), elevated LDH (OR = 3.13; 95%CI:1.20–8.15), elevated β2-MG (OR = 3.86; 95%CI:1.15–12.92), high total MTV (OR = 1.22 per 10 mL increase; 95%CI:1.02–1.46), Bcl-2 positivity (OR = 11.45; 95%CI:3.98–32.93), and C-MYC positivity (OR = 8.94; 95%CI:3.47–23.03). The integrated nomogram demonstrated excellent discrimination (AUC = 0.902; 95%CI:0.855–0.949) with 72.1% sensitivity and 92.1% specificity. Cox regression confirmed concordant findings (C-index = 0.847; 95%CI:0.795–0.899). The Brier score was 0.128, with calibration slope of 0.94 and intercept of 0.02. Calibration analysis confirmed strong agreement between predicted and observed outcomes (Hosmer-Lemeshow χ²=10.14, P = 0.255). Decision curve analysis demonstrated superior net benefit compared with IPI and NCCN-IPI across clinically relevant threshold probabilities.

CONCLUSIONS

Integration of clinical, metabolic, and molecular biomarkers yields superior prognostic stratification in DLBCL. This internally validated nomogram, composed entirely of routinely available clinical assessments and developed in a real-world patient population, provides a hypothesis-generating tool for identifying high-risk patients that warrants external validation before clinical implementation.