The utility of sectioning at multiple levels in the histopathologic analysis of sentinel lymph nodes (SLNs) for melanoma and the correlation of metastasis size with risk of subsequent metastasis were investigated. Metastatic melanoma was identified in SLNs from 91 of 475 (19%) melanoma patients with SLN sampling at the Massachusetts General Hospital between 2004 and 2008. All SLNs were evaluated by a 9-slide protocol: sets of MART-1, hematoxylin and eosin, and S100 stains at 3 distinct levels separated by 80 μm. The location and size of the tumor deposits were evaluated in the context of subsequent metastasis and overall survival. Of the 91 patients with positive sentinel nodes, all 9 protocol slides were available for review in 61 (67%). Eleven of 61 patients had no tumor present in the first set of levels; 2 of these patients died of metastatic melanoma. Patients in whom 11 or more tumor cells were detected in the sentinel node had a greater chance of developing subsequent metastases when compared with patients in whom 10 or fewer tumor cells were detected (P=0.05). Of those with either metastases >2 mm in diameter or extracapsular extension, 50% developed metastases beyond the SLN basin. Eliminating 1 of the 3 levels in the SLN detection protocol would have led to a false-negative diagnosis in 18% of patients.