e21059 Background: Traditionally, larger tumor size and lymph node (LN) involvement have been considered independent predictors of mortality in melanoma. We aimed to characterize the interaction between tumor size and LN involvement in melanoma specific mortality. In particular, we evaluated whether very small tumor size represented a particularly aggressive disease variant compared with larger LN-positive melanoma. Methods: Using Surveillance, Epidemiology and End Results registry data, we identified 57,223 patients (aged 18-85 years) diagnosed between 1998 and 2012 with histologically confirmed nonmetastatic melanoma treated with surgery. Primary study variables were tumor size, LN involvement, and their corresponding interaction term. Kaplan-Meier methods, adjusted Cox proportional hazards models with interaction terms were performed. Potential confounders included age, sex, year of diagnosis, marital status and number of LN dissected. Results: Median follow-up was 48 months. In multivariable analysis, there was significant interaction between tumor size and LN involvement ( P < 0.0001) using the likelihood ratio test and Wald test. In the absence of LN involvement (n = 54,922), the hazard ratio (HR) increased monotonically with increasing tumor size. Among patients with LN involvement (n = 2,301), using the smallest tumors as the reference group, hazard ratio of cancer specific mortality decreased unexpectedly in 0.01-1.00mm tumors (HR 0.61, p = 0.022) and 1.01-2.00mm tumors (HR 0.58, p = 0.007), and reached to a similar level in tumors sized 2.01-4.00mm (HR 0.81, p = 0.280), and subsequently increased in tumors sized larger than 4.00mm (HR 1.57, p = 0.016). Conclusions: In LN positive melanoma, very small tumors may predict for higher mortality compared with larger tumors. These results should be validated in future database studies. Table. Effect of tumor size in LN positive and LN negative melanoma. [Table: see text]
Outcome of multidisciplinary treatment of peripheral primitive neuroectodermal tumor