OBJECTIVE: To determine sensitivity and specificity of colour-assisted compression ultrasound (CUS) in the diagnosis of deep venous thrombosis (DVT) isolated to the calf veins. DESIGN: Prospective comparison of CUS with contrast venography in patients undergoing both procedures, with blinded evaluation of results. SETTING: Alfred Hospital, Melbourne (an urban tertiary referral hospital), between November 1990 and May 1992. SUBJECTS: Patients presenting for contrast venography with signs or symptoms of lower limb DVT. OUTCOME MEASURES: Presence of DVT; technical adequacy of examination. RESULTS: 92 of 402 patients (104 limbs) were examined. DVT was detected by venography in 43 limbs and was isolated to the calf veins in 19. It was diagnosed by CUS for eight of these 19. Calf CUS was technically inadequate in 26 limbs. Sensitivity and specificity of CUS for isolated calf DVT were 67% (95% confidence interval [CI], 40%-94%) and 93% (95% CI, 83%-100%), respectively, when the CUS was adequate. The low sensitivity of CUS was related to small size of the thrombi, inadequate studies because of previous DVT or very swollen limbs and technical errors. CONCLUSION: CUS is accurate for detecting larger isolated calf DVT when the calf study is adequate. Including the calf veins as well as the femoropopliteal veins in the initial CUS examination for symptomatic suspected lower limb DVT may reduce the need for follow-up CUS.