[Duplex sonography: accuracy, reproducibility and possibilities for error. Checking quantitative flow measurements against an in vitro model].

Using a pulsatile model of tubes, duplex-sonographic measurements of flow were compared with actual flow. The correlation coefficient between actual and duplex-sonographically obtained flow was 0.975, mean percentage deviation being -18.1% (P less than 0.0001). The obvious scatter (margin of error -69.2 to +50%) is to be explained by the simulation closely imitating a real situation: the transducer was hand-held and tubes were placed in muscle or fat tissue. Accuracy was increased by multiple measurements, especially when taking into account only the maximal value of any series of measurements. Vessel diameter was easy to measure accurately (mean error less than 0.01% [-18.7 to +9.4%]). Repeated measurements of flow velocity had only a small scatter (coefficient of variance 0.064). In muscle or fat the Doppler signal was attenuated and the error the greater the deeper the level at which measurements were made. Low flow (common in patients with obstructive vascular disease) can also cause faulty results.