MEASUREMENT OF AORTIC BLOOD-FLOW BY DOPPLER ECHOCARDIOGRAPHY - DAY TO DAY VARIABILITY IN NORMAL SUBJECTS AND APPLICABILITY IN CLINICAL RESEARCH

To assess the reliability of Doppler ultrasound for detecting serial changes in cardiac output in response to experimental interventions, the day to day variability of the minute distance of aortic flow was determined in seven normal subjects maintained in a tightly controlled environment with regard to diet and activities. Measurements were made at the same time on 5 to 6 sequential days from an apical window with use of both continuous wave and pulsed wave Doppler techniques. Two statistical measures of reliability were calculated, the intraclass coefficient of correlation (R), which varies between 0 (null reliability) and +1 (perfect reliability), and the 95% confidence interval for the error-free value of a single measurement. For sequential measurements of arterial pressure, 24 h urinary volume and sodium excretion and body weight, the intraclass coefficients of correlation ranged from 0.85 to 0.99, indicating low day to day variability consistent with tight environmental control. Continuous and pulsed wave modes were proved equally and highly reliable for measuring minute distance of aortic flow. However, continuous wave Doppler ultrasound provided acceptable signals more frequently than did the pulsed wave technique. For continuous wave Doppler ultrasound, R was 0.87 (p < 0.00001); the 95% confidence interval was +/- 1.81 m/min (or 11% of the mean of all measurements), which indicates that this method can be used in a single individual to detect a > 11% change in minute distance measured once before and after an intervention. In the case of n repeated measurements before and after an intervention, detectability of change in a single individual increases substantially because of a change of 11%/square-root n then is significant. The reliability measures can also be used to estimate the detectability of a change for a group of subjects; specifically, they allow determination of the sample size required to demonstrate a specified magnitude of change caused by an intervention. A procedure is presented for such a sample size determination for prospective studies of both parallel and crossover study design.