Completion lymph node dissection (CLND) is routinely performed after metastatic melanoma is detected at sentinel lymph node (SLN) biopsy. Nonsentinel lymph node (NSLN) involvement is found in less than one-third of the cases. Possible predictors of NSLN involvement are examined. A retrospective review of 70 patients with a positive SLN biopsy for melanoma and drainage to one lymphatic basin was performed. The size of metastatic deposits was defined as macrometastases (>2 mm), micrometastases (< or =2 mm), a cluster of cells (10-30 grouped cells) in the subcapsular space or interfollicular zone, or isolated melanoma cells (1-20 or more individual cells) in subcapsular sinuses. Tumor stage, ulceration, SLN tumor burden, mitoses, number of positive SLNs, and total number of lymph nodes removed were examined as predictors of NSLN involvement after CLND. Two additional models based on SLN tumor burden and the number of nodes biopsied were designed. Nineteen patients (24.3%) were found to have NSLN metastases after CLND. Tumor stage, ulceration, SLN tumor burden, mitoses, number of positive SLN, and number of lymph nodes removed were not statistically significant. Residual disease at CLND stratified by SLN tumor burden was: isolated melanoma cells, 0; cluster of cells, 8 (38.1%); < or =2 mm, 5 (20.8%); and >2 mm, 6 (27.3%). A comparison of the means for the models was not predictive of NSLN involvement. None of the risk factors or models examined could predict nonsentinel lymph node involvement with melanoma. The SLN sample and minimal SLN metastatic disease when defined as isolated clusters of cells warrant further study as a potential indicator against CLND after positive SLN.