Application of clinical nomograms to predicting overall survival and event-free survival in multiple myeloma patients: Visualization tools for prognostic stratification.

ABSTRACT

Background: This study aimed to develop reliable nomogram-based predictive models that could guide prognostic stratification and individualized treatments in patients with multiple myeloma (MM). Methods: Clinical information of 560 patients was extracted from the MM dataset of the MicroArray Quality Control (MAQC)-II project. The patients were divided into a development cohort (n = 350) and an internal validation cohort (n = 210) according to the therapeutic regimens received. Univariate and multivariate Cox regression analyses were performed to identify independent prognostic factors for nomogram construction. Nomogram performance was assessed using concordance indices, the area under the curve, calibration curves, and decision curve analysis. The nomograms were also validated in an external cohort of 56 patients newly diagnosed with MM at Nanjing Drum Tower Hospital from May 2016 to June 2019. Results: Lactate dehydrogenase (LDH), albumin, and cytogenetic abnormalities were incorporated into the nomogram to predict overall survival (OS), whereas LDH, ß2-microglobulin, and cytogenetic abnormalities were incorporated into the nomogram to predict event-free survival (EFS). The nomograms showed good predictive performances in the development, internal validation, and external validation cohorts. Additionally, we observed a superior prognostic predictive ability in nomograms compared to that of the International Staging System. According to the prognostic nomograms, risk stratification was applied to divide the patients into two risk groups. The OS and EFS rates of low-risk patients were significantly better than those of high-risk patients, suggesting a greater function of the nomogram models for risk stratification. Conclusion: Two simple-to-use prognostic models were established and validated. The proposed nomograms have potential clinical applications in predicting OS and EFS for patients with MM.