BACKGROUND A tumor-negative sentinel lymph node (SLN) does not preclude recurrence of melanoma. We hypothesized that certain patient-related and tumor factors are predictive of a worse outcome in these patients. METHODS Disease-free survival (DFS), overall survival (OS), and recurrence patterns were retrospectively analyzed in 773 patients who underwent lymphatic mapping and SLN biopsy for primary cutaneous melanoma at our institution between 1995 and 2002, and who had tumor-negative SLNs by standard pathological analysis. Patient sex, age, tumor site and thickness, ulceration status, Clark level, and histology were evaluated for their influence on outcome by univariate and multivariate Cox regression analysis and classification and regression tree analysis. RESULTS DFS and OS at 5 years were 88% and 93%, respectively. Sixty-nine (8.9%) of 773 patients developed recurrence. Three-year OS was lower in patients with distant recurrence (17.1%) than in those with local/regional recurrence (55.5%). By multivariate analysis, primary tumor thickness (P < .0001), site on head/neck versus trunk (P = .0093) versus extremity (P = .0042), and ulceration status (P = .0024) were independently significant for DFS; primary tumor thickness (P = .0106) and ulceration status (P = .0001) also were independently significant for OS. Classification and regression tree analysis demonstrated DFS was shortest in patients who had ulcerated tumors >2 mm. CONCLUSIONS Melanoma will recur in approximately 9% of patients with tumor-negative SLNs. Patients with thick, ulcerated melanomas on the head or neck have the highest risk for recurrence. This group should be followed closely for recurrence and considered for adjuvant therapy.