We sought to evaluate the accuracy of detecting asymptomatic pulmonary metastases by surveillance chest X-rays (CXRs) in melanoma patients with a positive sentinel node biopsy. Sentinel node-positive patients treated at the Sydney Melanoma Unit between 1994 and 2003 were prospectively enrolled onto a monitoring schedule of 6 monthly CXRs for 5 years, then annual CXRs for another 5 years. The reference standard for pulmonary metastasis was a positive histopathology diagnosis from a lung biopsy. A total of 108 patients were followed for a median of 52.5 months. A total of 21% (23 of 108) developed pulmonary metastases, which were detected in 48% (11 of 23) by surveillance CXR (sensitivity, 48%; 95% confidence interval [95% CI], .27-.68), leading to resection in 13% (3 of 23). CXRs were abnormal in 19 additional patients but not due to recurrence (specificity, 78%; 95% CI, .77-.79). Additional metastatic disease was apparent in 18% of CXR-detected versus 76% of non-CXR-detected patients (p< .05), but median time to diagnosis of pulmonary metastases was 24 months (95% CI, 12-41) versus 16 months (95% CI, 10-30, p= .30 log rank) and median survival of 42 months (95% CI, 24-84) versus 36 months (95% CI, 18-46, p= .53 log rank) were not significantly different. The 6 to 12 monthly surveillance CXRs detected only half of pulmonary metastases, infrequently identified patients for potentially curative surgery, and did not lead to earlier detection of pulmonary metastases. Further, they may cause unnecessary patient anxiety, given the high rate of false-positive findings.