A multivariate analysis of prognostic factors for melanoma patients with lesions greater than or equal to 3.65 mm in thickness. The importance of revealing alternative Cox models.

Fourteen prognostic factors were examined in 79 patients with clinical Stage I melanoma greater than or equal to 3.65 mm in thickness. All nine patients with melanoma of the hands or feet died of melanoma. A Cox proportional hazards (multivariate) analysis of the remaining 70 patients showed that a combination of the following four variables best predicted bony or visceral metastases: 1) a nearly absent or minimal lymphocyte response at the base of the tumor, 2) histologic type other than superficial spreading melanoma, 3) location on the trunk, and 4) positive nodes or no initial node dissection. Ulceration and/or ulceration width were not useful in predicting outcome either singly or in combination with other variables. Patients with negative lymph nodes and primary tumors of the trunk, hands, and feet did not do better than patients with positive nodes at those sites. Conversely, non of 16 patients with negative lymph nodes and extremity melanomas (excluding the hands and feet) or head and neck melanomas developed visceral or bony metastases (i.e., five-year disease-free survival rate 100%).