DISCUSSION
To the best of our knowledge, this is the first study to evaluate the accuracy of multiple echocardiographic algorithms to estimate PAPM and study diagnostic performance to identify PH in the specific setting of HF. All 4 DE algorithms demonstrated reasonable association with RHC and good agreement on Bland-Altman analysis, with generally lower bias seen in methods interrogating the TR signal. Of the 4 methods, the Chemla et al algorithm demonstrated comparable diagnostic performance with TRVmax, both when employing ROC and sensitivity analysis. However, none of the DE algorithms outperformed TRVmax.
The accuracy of DE to estimate pulmonary artery pressures has been a matter of debate. Earlier studies suggest that DE frequently over- or underestimates invasive pulmonary pressures and should not be relied upon.15 16 More recent studies, however, have emphasized results of Bland-Altman analyses that display low bias between echocardiographic PAPM and RHC, suggesting that Doppler estimates are highly accurate.17 Our data suggests that accuracy of DE estimates may also vary based on approach utilized. Minimal bias was observed in methods that incorporated TRVmax, corroborating an earlier study employing high-fidelity catheters that suggests that such an approach, despite being routinely used as an estimate of PAPS, provides the most accurate estimate of PAPM.18Higher systemic bias with RHC and lower precision reflected in wider limits of agreement employing both PI (Abbas et al7) and RVOTAT (Dabestani et al12) seen in this study may, at least in part, be attributable to smaller patient cohorts (n = 23 and 39 respectively) and less severe clinical presentations in the original studies. As seen in the Bland-Altman plots, a greater dispersion of points is observed at higher mean values of PAPM, suggesting that these methods may be less reliable in the setting of severe PH. The cohort examined by Abbas et al demonstrated a PAPM = 25 (range 10-57) mmHg and PAWPM =15 (range 2-38) mmHg, suggesting a milder hemodynamic presentation compared with the present cohort.7 Dabestani et al do not present corresponding values in their cohort, but suggest a PAPM range that is relatively lower than that in our study with lower PH cut-off (20mmHg).12 Additionally, the empirical algorithms presented using this method may demonstrate limited utility in the setting of severely elevated PAPM, as alluded to in certain comparative studies evaluating multiple echocardiographic approaches.13
Importantly, despite displaying relatively lower precision and agreement with invasive measurements, both the above-mentioned methods demonstrated good diagnostic ability to identify PH in our cohort. Uninterpretable TR signals are frequent in HF,19 have been reported in as many as 39% of subjects and may present a limitation to echocardiographic evaluation of PH.20 In our study, TRVmax could not be adequately assessed in 14% and VTI in 19% of patients, suggesting a potential diagnostic role for methods that do not necessitate TR jet interrogation.
Early identification of PH in HF has direct consequences on treatment and prognosis. Despite reasonable diagnostic ability demonstrated by all echocardiographic algorithms, only the approach postulated by Chemla et al8 demonstrated diagnostic ability comparable with recommended 2.8m/sec TRVmax cut-off in both ROC and sensitivity analysis. However, none of the methods outperformed TRVmax. This finding is contrary to a recent comparative report where the chosen DE algorithms showcased generally superior performance as compared with TRVmax.13The authors suggest in the abovementioned study that DE algorithms that consider estimates of right atrial pressure in addition to TRVmax demonstrate generally stronger correlation with invasive measurements and superior diagnostic performance when compared with TRVmax. This was substantiated by data from their study where right atrial pressure > 15mmHg estimated by echocardiography demonstrated highest odds ratio for invasively confirmed PH. In the setting of HF, echocardiographic estimates of right atrial pressure are frequently falsely elevated and sole reliance on the IVC to estimate RAP may be misleading.21 In our study, 30% of patients with echocardiographically estimated RAPM = 15mmHg demonstrated normal corresponding invasive pressures, suggesting that these estimates are frequently inaccurate and may not necessarily contribute to stronger performance of derived PAPM variables as suggested in certain derivation cohorts7 and the comparative study.13 Echocardiographic estimates of RAPM have been incorporated into empirical derivations of PAPM in all but one selected PAPMalgorithms in this study. This may play a role in the observed lower performance when compared with TRVmax alone, but needs to be further examined.
The use of fluid-filled catheters instead of high-fidelity manometer-tipped catheters for pressure measurement might introduce additional error and may be considered a limitation in this study. Retrospective analysis of echocardiographic data did not permit a closer inspection factors leading to lower feasibility of certain algorithms included in this comparative analysis. Finally, we did not employ agitated saline bubble contrast to strengthen TR jet signal as this is not part of routine protocol in our laboratory.