Risk factors and predictive models for early death in patients with advanced melanoma: A population-based study

• Published in: Medicine (Baltimore) Vol. 102; no. 40; p. e35380
• Main Authors: Li, Siru, Yin, Cunli, Yang, Xi, Lu, Yingchun, Wang, ChunYu, Liu, Bin
• Format: Journal Article
• Language: English
• Published: Hagerstown, MD Lippincott Williams & Wilkins 06.10.2023
• Subjects: Observational Study; early death; predictive model; nomogram; surveillance; advanced melanoma; Epidemiology and End Results (SEER)
• DOI: 10.1097/MD.0000000000035380
• ISSN: 0025-7974; 1536-5964

The prognosis for advanced melanoma (AM) is extremely poor. Some patients are already in an advanced stage at the time of their first diagnosis and face a significant risk of early death. This study predicted all-cause early death and cancer-specific early death in patients with AM by identifying independent risk factors, building 2 separate nomogram models, and validating the efficiency of the models. A total of 2138 patients diagnosed with AM from 2010 to 2015 were registered in the Surveillance, Epidemiology and End Results (SEER) database and randomly assigned to a training cohort and a validation cohort. Logistic regression models were used to identify the associated independent risk factors. These factors have also been used to build nomograms for early deaths. Next, we validated the model's predictive power by examining subject operating characteristic curves, then applied calibration curves to assess the accuracy of the models, and finally, tested the net benefit of interventions based on decision curve analysis. The results of the logistic regression model showed that marital status, primary site, histological type, N stage, surgery, chemotherapy, bone, liver, lung and brain metastases were significant independent risk factors for early death. These identified factors contributed to the creation of 2 nomograms, which predict the risk of all-cause early death and cancer-specific early death in patients with AM. In the all-cause early death model, the area under the curve was 0.751 and 0.759 for the training and validation groups, respectively, whereas in the cancer-specific early death model, the area under the curve was 0.740 and 0.757 for the training and validation groups. Calibration curves indicated a high degree of agreement between the predicted and observed probabilities, and the decision curve analysis demonstrated a high value for the model in terms of its applicability in clinical settings. These nomograms have practical applications in predicting the risk of early death in patients with AM, helping oncologists to intervene early and develop more personalized treatment strategies.