Risk factors analysis and nomogram development for myelosuppression in diffuse large B-cell lymphoma patients undergoing first-line chemotherapy: a dual-centre retrospective cohort study

Abstract

Objective: The primary objective of this research was to examine the characteristics of myelosuppression following first-line chemotherapy in patients suffering from diffuse large B-cell lymphoma (DLBCL). Furthermore, the study aimed to identify and analyze the risk factors impacting myelosuppression after chemotherapy and to construct a predictive model for evaluating the risk of myelosuppression.

Methods: This retrospective cohort study was conducted across two medical centers. The study included 243 patients with DLBCL treated at the Anning First People's Hospital Affiliated with Kunming University of Science and Technology from January 2022 to December 2023 as the development cohort, and 107 patients treated at the Third Affiliated Hospital of Kunming Medical University from January 2024 to May 2024 as the validation cohort. The study investigated the incidence of myelosuppression in all patients, identified independent factors influencing this condition through logistic regression analysis, and constructed and validated a nomogram. Finally, the model's performance was evaluated using both internal and external validation cohorts.

Results: The research rigorously incorporated a cohort of 243 DLBCL patients, with myelosuppression observed in 93 individuals (38.27%). Multifactorial analysis revealed that the chemotherapy cycle, age, Ann Arbor stage, surgical history, and neutrophil levels were independently correlated with myelosuppression following initial chemotherapy in DLBCL patients. A nomogram was developed based on the multifactorial analysis. The receiver operating characteristic (ROC) analysis revealed myelosuppression in the nomogram of both the development set (area under the curve (AUC = 0.834, 95% CI [0.785-0.884]) and the validation set (AUC = 0.861, 95% CI [0.791-0.931])), indicating clear differentiation. Further calibration curve analysis and decision curve analysis (DCA) revealed strong calibration and clinical utility of the column-line graph model.

Conclusion: Patients with DLBCL are at an increased risk and frequency of myelosuppression following first-line chemotherapy. The development of a highly accurate prediction model for myelosuppression in this patient population facilitates individualized treatment strategies. Future studies should focus on expanding the sample size and developing and validating the model in additional types of cancer.

Keywords: Forecasting; Myelosuppression; Nomograms; Non-Hodgkin lymphoma; Risk.

Figure 1. Experimental roadmap of this study.

This experimental roadmap indicates the inclusion and exclusion for patients and the workflow of the method present in this study.

Figure 2. A nomogram for predicting myelosuppression in DLBCL patients following the first-line chemotherapy.

A nomogram was developed to predict severe myelosuppression in DLBCL patients undergoing first-line chemotherapy, based on five predictors. To utilize the nomogram, each variable classification is plotted upwards on the “Points” axis to determine its corresponding points. The total points are then calculated on the “Total Points” axis, and a line is drawn downwards to the “Myelosuppression” axis to ascertain the probability of myelosuppression.

Figure 3. Receiver operating characteristic curves (ROC) for the development and validation sets.

The 95% CI and P-value for the area under the ROC curve for the development set (A) and validation set (B) were 0.834 (95% CI [0.785–0.884]) and 0.861 (95% CI [0.791–0.931]), respectively. No significant difference was observed between the two AUCs (DeLong test, P > 0.05).

Figure 4. Calibration curves for development set and validation set.

(A) Calibration curves for the development set. (B) Calibration curve for validation set. The y-axis represents the actual probability of myelosuppression occurring. x-axis represents the predicted probability of myelosuppression occurring. Diagonal: The black dashed line represents a perfect prediction using the ideal model. The gray dashed line represents the target parameter and the black solid line represents the performance of the model. The closer to the diagonal thick gray line represents a better prediction. Use bootstrap resampling (times = 1,000).

Figure 5. Clinical decision curve analysis (DCA) for the development set and validation set column line plots.

(A) DCA of column-line plots of predicted risk of myelosuppression in the development set; (B) DCA of column-line plots of predicted risk of myelosuppression in the validation set. The black dashed line is the predictive model; the solid gray line is all patients who were intervened, and the solid black horizontal line is all patients who were not intervened. The graph depicts the expected net benefit per patient in predicting the risk of myelosuppression formation relative to the nomogram. The net benefit increases as the model curve is extended. (Use bootstrap resampling (times = 1,000).)