BG01 Factors affecting survival of older patients with cutaneous malignant melanoma: a population-based study of 140 856 patients

Abstract Malignant melanoma (MM) is the fifth most common malignancy in the UK and the USA. MM has the highest mortality among all types of skin cancer. The population is becoming increasingly frail and comorbid, and the mean age of diagnosis of MM is 65 years. This study aims to explore the factors associated with survival in older patients and whether they differ from those in younger patients, using two different methods. Patients with cutaneous MM diagnosed during 2011–2021 were identified from the Surveillance, Epidemiology, and End Results (SEER) database, with 13 associated demographic and tumour-related variables extracted. They were divided into two groups: group 1 (< 70 years old) and group 2 (≥ 70 years old), as the European Society for Medical Oncology recommends using 70 years as the cutoff for older patients in oncology. Univariable and Multivariable Cox regression and an XGBoost machine learning model were used to study the factors associated with survival for both groups. SHapley Additive exPlanations (SHAP) was used to visually illustrate the most influential factors for survival. Kaplan–Meier overall survival was calculated for both groups. In total data for 140 856 patients with cutaneous MM were extracted from SEER. There were more male patients (54.9%) than female, and most were of White ethnicity (88.0%). The most common site was the trunk (33.0%) and the most common histology was melanoma not otherwise specified (46.3%). Stage I was the most common stage (48.7%), with most patients (94.6%) having no metastases. Only 12.2% of tumours were ulcerated, and only 45.2% had a mitotic rate < 1. The mean Breslow thickness and number of positive lymph nodes were 0.6 mm and 0.16, respectively. The 5-year rates of overall survival for group 1 and group 2 were 91.4% and 65.7%, respectively. Cox regression for both groups agreed on the same factors associated with worse survival, which were lower income; Black ethnicity; male sex; ulcerated lesions; a higher number of positive lymph nodes or Breslow index or mitotic rate; having a lesion on the head and neck; and having nodular, lentigo maligna or acral lentiginous melanoma, or melanoma not otherwise specified on histology. SHAP of the machine learning model showed that the top five factors influencing survival of both groups were Breslow thickness, income, stage, age and histology. Although group 1 had higher survival than group 2, the demographic and tumour-related factors affecting survival of both groups were similar. Some of these factors are not included in the current prognostication systems for MM. Further studies are needed to explore the effects of including these factors in prognostication systems, given the limitations of the SEER database, and to explore the effect of ageing on treatment of cutaneous MM.