The reproducibility of a variety of Doppler indexes of diastolic function in an epidemiologic setting and in atrial fibrillation have not been reported. This study examined the reproducibility of left ventricular inflow in subjects in sinus rhythm (n = 80) and atrial fibrillation (n = 12), randomly selected from the original cohort of the Framingham Heart Study. The following Doppler indexes were assessed for all subjects: peak and integral of early (E) diastolic inflow velocity, acceleration slope and time, deceleration slope and time, and pressure half-time. For subjects in sinus rhythm, the following parameters also were measured: the peak and integral of late (A) diastolic inflow velocity, ratios of peak velocities and integrals E/A, and atrial filling fraction. Intraobserver and interobserver variability were evaluated by statistical methods including Student's t test of the systematic differences (bias), percent bias, correlation coefficients, measurement precision, and percent precision. In subjects in sinus rhythm, although the interobserver bias was statistically significant for most of the parameters, it was < 10% for all but 1 parameter (acceleration time). For the peak and integral measures, the intra- and interobserver correlations were > or = 0.89, with intra- and interobserver percent precision measures within 2.2 to 13.0% of the corresponding mean values. The acceleration, deceleration and pressure half-time measures had somewhat lower correlations (interobserver correlations ranging from 0.59 to 0.96), with percent precision measures further from the corresponding means (interobserver percent precision ranging from 10.1 to 19.5%).(ABSTRACT TRUNCATED AT 250 WORDS)