Sixty patients with malignant melanoma were studied. The sex ratio was male 29: female 31. The patient age ranged from 17 to 86 years (mean age, 59.6 years) . The primary tumors were 28 acral lentigenous melanomas, 17 superficial spreading melanomas, 4 nodular melanomas, 3 lentigo maligna melanomas and 2 mucosal melanomas. The level of tumor invasion was identified as II (2 patients), III (6 patients), IV (32 patients) and V (6 patients) . Tumor thickness was distinguished into four groups: less than 0.76mm (4 patients), between 0.76 and 1.50mm (5 patients), between 1.51 and 4.0mm (15 patients) and more than 4.0mm (18 patients) . Using the Kaplan-Meier life table method coupled with the log-rank test, univariate survival analyses were performed for 8 factors (age, sex, treatment history, histologic type of melanoma, level of invasion, tumor thickness, metastases to regional lymph nodes and pathologic stage). These variables other than stage, which was essentially a dependent factor, were further analyzed by the multivariate analytical method of the Cox proportional hazards model in order to determine the primary independent risk factor predicting survival in melanoma patients. The Kaplan-Meier life table method demonstrated the following three factors that worsen patient survival significantly evidence of metastasis to the regional nodes (p<0.0001), advancement of pathologic stage (p=0.021), and advancement of level (p=0.038) . Other factors like patient age, sex, treatment history, tumor thickness or histologic type did not significantly correlate with the patient survival. When the above factors except for stage were applied to the Cox proportional hazards multivariate analysis model, we were able to acquire amodel that significantly fitted the real data (p=0.012) . This model demonstrated that the regional lymph nodes metastasis and level were the variables that significantly reflects patient survival (p=0.017, 0.029 respectively) . Stepwise selection was then performed by removing the least important variable at respective steps until the remaining variables were all significantly important at the p<0.05 level. The regional lymph nodes metastasis and level were the variables that remained at the final step of the selection. The relative death rate (hazard ratio) of the regional lymph nodes metastasis and level were 9.07 and 12.6 respectively. Although tumor thickness has been reported to be the best prognostic indicator, it dose not necessarily predict patient survival when all stages are taken together. The characteristics of our materials may have reached the evidence that tumor thickness was not a significant prognostic factor.
