Our objective was to assess the inter-observer and intra-observer agreement in the interpretation of digital subtraction venography (DSV) in patients with suspected deep vein thrombosis of the upper extremity (DVTUE). Prospectively, 62 consecutive DSV studies in 54 patients with clinically suspected DVTUE were included. Hard copies were presented without demographic data or original report. All venograms were read twice, at 3-month intervals, by an interventional vascular radiologist (observer 1) and an experienced general radiologist (observer 2). Consensus reading took place in the presence of a third experienced interventional radiologist. Inter-observer and intra-observer agreement were assessed using kappa statistics. Initial reading in 62 venograms showed an inter-observer agreement of 71% (kappa 0.48). The inter-observer agreement of the second reading was 83% (kappa 0.71). The agreement with the consensus report ranged from 76 to 94%. The intra-observer agreement for the first and second observer was 94% (kappa 0.89) and 76% (kappa 0.56), respectively (p<0.01). Digital subtraction venography has moderate to excellent intra- and inter-observer agreement, suggesting that digital subtraction venography is reliable for the diagnosis of DVTUE.