New Algorithm for Estimating Left Ventricular Filling Pressure by Echocardiography

Evaluation of whether dyspnea has a cardiac cause is essential. Guidelines from 2016 were reported to result in a high incidence of indeterminate left ventricular (LV) filling pressure. We sought to validate a new algorithm for the estimation of LV filling pressure (LVFP) in a multicenter study, with the objective of decreasing the yield of indeterminate filling pressure and increasing accuracy.

In an observational study, echocardiography was performed in 951 patients referred for cardiac catheterization. Echocardiographic measurements included mitral inflow, pulmonary vein and tissue Doppler mitral annulus velocities, tricuspid regurgitation velocity, assessment of mean right atrial pressure, biplane LV and left atrial volumes, and LV and left atrial strain. A stepwise approach was applied in a new algorithm for estimation of LVFP, whereby pressure >15 mm Hg was considered abnormally elevated. The first step included mitral annulus early diastolic velocity (e'), the ratio of mitral early flow velocity to e', and pulmonary artery systolic pressure. With concordant findings in all 3 variables, conclusions about LVFP could be reached. In case of discordant or incomplete variables, left atrial reservoir strain, left atrial maximum volume index, isovolumic relaxation time, and pulmonary vein flow were analyzed in a second step. In the presence of ≥1 abnormal measurement in the second step, the conclusion of elevated LVFP could be reached.

Only 2 patients had indeterminate LVFP as per the new algorithm versus 38 applying 2016 guidelines (P<0.0001). In 949 patients, sensitivity was 86% and specificity was 86%, with accuracy of 86%. Accuracy was higher than the 2016 algorithm in all patients (P<0.0001), and in patients with ejection fraction ≥50% (P<0.0001), whereas accuracy was similar in patients with ejection fraction <50%. In 663 patients with natriuretic peptides data, net reclassification improvement for echocardiography over natriuretic peptides was 1.1 (P<0.0001), and integrated discrimination improvement was 0.3 (P<0.0001).

The new algorithm increases the feasibility of estimating LVFP and has good accuracy with incremental value when natriuretic peptides are considered.