A clinical-laboratory nomogram integrating inflammatory, metabolic and proliferative markers to predict refractory disease or early relapse in diffuse large B-cell lymphoma

Background First-line R-CHOP immunochemotherapy controls diffuse large B-cell lymphoma (DLBCL) in most patients, but 30%–40% develop primary refractory disease or early relapse, both with poor survival. The composite endpoint of refractory disease or early relapse (R/ER) was predefined, as both reflect early treatment failure. The International Prognostic Index captures tumor biology only partially, and no bedside tool integrates routine inflammatory, metabolic, and proliferative markers. We developed and internally validated a routine-marker nomogram to predict R/ER after first-line immunochemotherapy and quantified predictor contributions. Materials and methods In this retrospective single-center study, 342 patients were randomly split 7:3 into training ( n = 239) and internal validation ( n = 103) cohorts. Predictors were screened by univariate analysis ( P < 0.05), followed by LASSO and multivariable logistic regression. The logistic model was compared with random forest (RF) and support vector machine (SVM) models by ROC-AUC, calibration, and decision curve analysis; hyperparameters were tuned by training-set cross-validation. SHAP analysis quantified predictor contributions on the log-odds scale. Results Elevated lactate dehydrogenase (LDH), β 2-microglobulin ( β 2-MG), neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), and Ki-67, with reduced lymphocyte-to-monocyte ratio (LMR), independently predicted R/ER (all P < 0.05). Logistic regression had the highest AUC (0.815, 95% CI 0.747–0.883 in training; 0.776, 0.660–0.891 in validation), with no significant between-model AUC differences (DeLong P = 0.68 vs. RF, P = 0.08 vs. SVM in training; all P > 0.05 in validation). Calibration slope and intercept were 0.97 and 0.03 in training and 0.93 and 0.07 in validation; net benefit at thresholds 0.2, 0.3, and 0.4 was 0.26, 0.19, and 0.13 in training and 0.22, 0.16, and 0.10 in validation. The logistic model was selected for parsimony and interpretability. SHAP ranked predictors as Ki-67, PLR, LDH, NLR, β 2-MG, and LMR, forming a six-factor nomogram with acceptable discrimination and calibration. Conclusion A six routine-marker nomogram showed acceptable discrimination and calibration for R/ER after first-line immunochemotherapy in this single-center cohort. As it was only internally validated and not compared with IPI, NCCN-IPI, or R-IPI, it remains exploratory and requires multicenter prospective external validation before clinical use.

Authors

Institutions

Publication Details

Journal
Frontiers in Medicine
Published
2026-11-16
DOI
https://doi.org/10.3389/fmed.2026.1995591
Primary Topic
Lymphoma Diagnosis and Treatment
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

A clinical-laboratory nomogram integrating inflammatory, metabolic and proliferative markers to predict refractory disease or early relapse in diffuse large B-cell lymphoma

Wei Chen, Shuang Li, Weikai Zhang, Quanzhong Yang
1 links
Frontiers in Medicine
Lymphoma Diagnosis and Treatment
article

A clinical-laboratory nomogram integrating inflammatory, metabolic and proliferative markers to predict refractory disease or early relapse in diffuse large B-cell lymphoma

Wei Chen, Shuang Li, Weikai Zhang, Quanzhong Yang
article en

Abstract

Background First-line R-CHOP immunochemotherapy controls diffuse large B-cell lymphoma (DLBCL) in most patients, but 30%–40% develop primary refractory disease or early relapse, both with poor survival. The composite endpoint of refractory disease or early relapse (R/ER) was predefined, as both reflect early treatment failure. The International Prognostic Index captures tumor biology only partially, and no bedside tool integrates routine inflammatory, metabolic, and proliferative markers. We developed and internally validated a routine-marker nomogram to predict R/ER after first-line immunochemotherapy and quantified predictor contributions. Materials and methods In this retrospective single-center study, 342 patients were randomly split 7:3 into training ( n = 239) and internal validation ( n = 103) cohorts. Predictors were screened by univariate analysis ( P < 0.05), followed by LASSO and multivariable logistic regression. The logistic model was compared with random forest (RF) and support vector machine (SVM) models by ROC-AUC, calibration, and decision curve analysis; hyperparameters were tuned by training-set cross-validation. SHAP analysis quantified predictor contributions on the log-odds scale. Results Elevated lactate dehydrogenase (LDH), β 2-microglobulin ( β 2-MG), neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), and Ki-67, with reduced lymphocyte-to-monocyte ratio (LMR), independently predicted R/ER (all P < 0.05). Logistic regression had the highest AUC (0.815, 95% CI 0.747–0.883 in training; 0.776, 0.660–0.891 in validation), with no significant between-model AUC differences (DeLong P = 0.68 vs. RF, P = 0.08 vs. SVM in training; all P > 0.05 in validation). Calibration slope and intercept were 0.97 and 0.03 in training and 0.93 and 0.07 in validation; net benefit at thresholds 0.2, 0.3, and 0.4 was 0.26, 0.19, and 0.13 in training and 0.22, 0.16, and 0.10 in validation. The logistic model was selected for parsimony and interpretability. SHAP ranked predictors as Ki-67, PLR, LDH, NLR, β 2-MG, and LMR, forming a six-factor nomogram with acceptable discrimination and calibration. Conclusion A six routine-marker nomogram showed acceptable discrimination and calibration for R/ER after first-line immunochemotherapy in this single-center cohort. As it was only internally validated and not compared with IPI, NCCN-IPI, or R-IPI, it remains exploratory and requires multicenter prospective external validation before clinical use.

Frontiers in MedicineVol. 13
Luoyang Polytechnic (CN), Luoyang Institute of Science and Technology (CN)
Good health and well-being
Lymphoma Diagnosis and Treatment
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon