Background: Optimal surveillance for MEL recurrence is elusive. While consensus guidelines agree on surveillance imaging for high-risk MEL, there is no consensus regarding optimal modality for surveillance and routine imaging surveillance is not recommended for stage IA-IIA MEL. We examined the utility of surveillance in stage I-III MEL. Methods: Patients (pts) at the University of Pittsburgh's Melanoma Program 1991-2011 were queried using a clinical database. Eligible pts had stage I-III MEL and underwent routine surveillance with clinical examination, chest x-rays (CXR). Minimum follow-up was 9 months (mos). CXR positivity was determined by primary review of attending radiologist report. Pt documentation was queried for information pertaining to relapse, details of advanced imaging, pathology and treatment. Primary endpoints were the incidence of pulmonary (pulm) and extra-pulm metastases. Results: 324 pts with 2,700 CXRs were identified, of whom 114 (35%) had stage I, 63 (20%) stage II, and 147 (45%) stage III MEL. Median duration of screening was 46 mos. During this period, pulm mets were identified in 11%, 25% and 26% of stage I, II and III MEL respectively. Of these, CXR was initially diagnostic in 85%, 81%, and 92% respectively. Incidence of extra-pulm mets was 8%, 24% and 29% for stage I-III MEL, in whom diagnosis was made on advanced imaging ordered to evaluate examination findings and/or pt symptoms. Conclusions: Incidence of pulm and extra-pulm mets in MEL increases by stage, and CXR reliably detects pulm mets, although a higher incidence of non-pulm mets in stage II-III pts diminishes the value of CXR for relapses overall. Our data suggests that routine CXR surveillance detects few relapses in high-risk stage II-III pts.