A tool to predict survival outcomes and guide adjuvant immunotherapy recommendations for patients with stage II melanoma.

e21556 Background: Patients diagnosed with AJCC stage II melanoma have a 10-year melanoma specific survival (MSS) of 84%, ranging from 88% for stage IIA to 75% for stage IIC. The 12%-25% rate of melanoma mortality correlates with the rate of recurrence for these patients. Current prognostic tools are based on the AJCC 8 th Edition (AJCC-8), which gives MSS based on tumor thickness and ulceration status but does not consider any other patient or tumor characteristics. The rate of recurrence-free survival (RFS) may be more important than MSS in deciding whether a patient should be offered adjuvant immunotherapy. We have therefore developed a risk prediction tool to assist in this process. Methods: Data were extracted from a large Australian melanoma treatment center research database for patients diagnosed with stage II melanoma (n = 3243). Parameters included: age, sex, tumor thickness, mitotic rate, ulceration status, lymphovascular invasion, presence of tumor infiltrating lymphocytes, regression, sentinel node status, presence of satellites, body site, recurrence (including time to event) and date of last follow-up. Multivariable Cox regression analyses were performed to develop models for survival prediction. These were then externally validated using a dataset from a large US melanoma treatment center (n = 703). Discrimination and calibration of each model were assessed using the C-statistics and calibration plots respectively. Results: The table shows the C-statistics for the Australian RFS model and the AJCC-8 staging, along with validations. These demonstrated statistically significant discrimination gains by using the Australian model over the AJCC model, which ranged from 8.3% to 12.2%. The Australian model was well calibrated. Conclusions: There was good discrimination of the RFS model for individual patients over both 5 and 10 years, which held true on external validation. This model offers considerable improvement in discriminative accuracy for predicting RFS compared to using the AJCC-8 staging and therefore may be clinically useful to guide adjuvant immunotherapy recommendations. An online tool will be made available at www.melanomarisk.org.au.[Table: see text]