Development and Validation of a Model to Predict Recurrence in Patients With Stage IIIA Melanoma

Patients with stage IIIA cutaneous melanoma have relatively favorable prognoses (5-and 10-year overall survival probabilities of 93% and 88%, respectively), yet they are considered for systemic adjuvant therapy. Identifying the subset of patients at highest risk of recurrence could limit adjuvant treatment-related toxicity and direct therapy to those most likely to benefit.

Patients aged ≥18 years with stage IIIA melanoma diagnosed between 2001 and 2016 at 36 centers, primarily in Europe and North America, were included in the study. The full cohort was split in a 3:1 ratio into development (n=1,386) and validation (n=448) cohorts. A random forest procedure was applied to the development cohort to select variables for inclusion. The selected variables were incorporated into a Fine-Gray model to predict recurrence-free survival and graphically represented as a nomogram. Model performance was assessed using the C-index and inspection of the calibration plot. The model's clinical utility was assessed using decision curve analysis and compared with alternative strategies.

Eight variables were included in the model and were associated with the incidence of recurrence: mitotic rate (P<.001), number of positive sentinel lymph nodes (SLNs) per nodal basin (P<.001), diameter of the largest focus in the SLN (P=.001), ulceration (P<.001), age (P=.022), number of metastatic foci in the SLN (P=.019), lymphovascular invasion (P<.001), and extranodal extension (P=.005). In the validation cohort, the model had a C-index of 80.9% and appeared to be well calibrated. Decision curve analysis demonstrated a significant net benefit for adjuvant therapy guided by the nomogram compared with alternative strategies, including treating all patients with adjuvant therapy, treating no patients, or treating only patients with stage IIIA disease and >1.0-mm nodal tumor burden.

Our prediction model can improve patient selection for adjuvant therapy compared with other available strategies.