Although magnetic resonance imaging has been proposed for the diagnosis of deep venous thrombosis (DVT), its role in diagnostic strategy remains to be defined. We compared prospectively magnetic resonance angiography (MRA) with two-dimensional time-of-flight with contrast venography (CV) and colour duplex sonography (CDS) in 25 patients with DVT of the pelvis confirmed by CV. All patients were examined by CV (gold standard) and MRA and 17 by CDS. These studies were compared for DVT diagnosis in the pelvis and inferior vena cava and analysis of thrombotic spread. MRA was positive in 25 patients whose DVT was diagnosed by CV (100% sensitivity). MRA sensitivity and negative predictive value were 100%, specificity 98.5% and positive predictive value 97.5% for the diagnosis of thrombosis at each anatomic level. There were discrepancies between MRA and CV (2 false-positive results for 2 venous segments) and between CDS and CV (2 false-positive and 3 false-negative results). CV was uninterpretable for 8.8% of segments and CDS was often technically limited to the pelvic level, whereas all venous segments explored were analysable in MRA. MRA gave excellent results for positive diagnosis and DVT spread. MRA is a potentially valuable technique for assessing iliofemorocaval venous thrombosis.