Echocardiography in multicenter clinical trials: experience from the Treatment of Mild Hypertension Study

Echocardiography is becoming commonplace in clinical trials relating to hypertension for assessing changes in left ventricular mass (LVM). Factors affecting variability need to be considered in the design and management of such studies. These include (1) standardization of recordings and readings, (2) quality of recordings, (3) choice of sonographer vs. cardiologist readers, and (4) reader and temporal variability in measurements. The Treatment of Mild Hypertension Study (TOMHS) provides data and experience concerning these issues. TOMHS was a randomized trial of 902 participants, men and women, conducted in four clinical centers comparing six treatments for mild hypertension. M-mode echocardiograms were recorded at baseline and follow-up by centrally trained sonographers for assessment of LVM. Initial study design specified that each tracing be read by two sonographers with a study cardiologist adjudicating major differences in measurements. Poor agreement between sonographer readings prompted a change in design requiring one of two study cardiologists to read all tracings, with the same cardiologist reading all serial studies for a participant. Common tracings were read to assess interreader variability. The percentage of unreadable tracings varied by center (0.4-14.2%) and increased over follow-up. Reader agreement between the two cardiologists for LVM was greater (r = 0.83) than between sonographers (r = 0.68) or between sonographers and cardiologists (r = 0.64). The agreement between sonographer and cardiologist readings varied by center (range of r = 0.54-0.81), the highest correlations being in the two centers with physician echocardiographers. The intraclass correlation across visits for LVM was 0.66, being similar for all visit pairs. The standard deviation of the difference between baseline and follow-up LVM was 46 g using cardiologist readings, 51 g for readings made by the same sonographer, and 68 g for readings made by different sonographers, a difference that is large enough to appreciably affect sample size and power for studies. High-quality echocardiographic data can be obtained in multicenter trials, but this requires the continued training of sonographers and readers and ongoing monitoring of quality of tracings and measurements.