Testing the test: The reliability of echocardiography in the sequential assessment of valvular regurgitation

Background and Objective Substantial variability in serial echocardiographic qualitative assessment of valvular regurgitation may exist. Reader variability is generally well understood, but acquisition variability (portions of variability caused by equipment, sonographers, physiologic changes) has been less frequently assessed, particularly in combination with reader variability. We attempted to determine the relative contributions of acquisition and reader variability as components of total test-retest variability for aortic (AR) and mitral (MR) regurgitation. Methods Outpatient echocardiographic study was done at 2 clinical sites. Twenty-three predominantly obese middle-aged females had 3 echocardiograms, 2 performed 14 +/- 3 days apart and the third performed within I to 2 hours of the second. Triplets of echocardiograms were evaluated for change in grade of AR and MR. Medical history, anthropometrics, and blood pressures were obtained. Results Average intrareader variability (percentage of reads for which there is within-reader disagreement) was 5.6% for AR and 16.7% for MR. The average total test-retest variability (percentage of reads for which there is disagreement between visits) was 29.0% for AR and 24.6% for MR. The acquisition variability for AR was 23.4% +/- 7.7%; for MR, it was 7.9% +/- 10.2%. A significant predictor of change for AR/MR was the initial grade. Change in diastolic blood pressure was positively associated with change in AR and MR. Conclusions Intrareader agreement was substantial for AR and MR. Components of total test-retest variability found were reader, biological (change in diastolic blood pressure), and regression to the mean. Recommendations for clinical practice include monitoring blood pressure changes and understanding the confidence limits of the clinical test. Test-retest variability and its components should be considered in echocardiography and other diagnostic testing.