Comparisons of Noninvasive Methods Used to Assess Exercise Stroke Volume in Heart Failure with Preserved Ejection Fraction

Abstract

Introduction: Cardiopulmonary exercise testing (CPET) plays an important role in properly phenotyping signs and symptoms of heart failure with preserved ejection fraction (HFpEF). The prognostic value of CPET is strengthened when accompanied by cardiac hemodynamic measurements. Although recognized as the "gold" standard, cardiac catheterization is impractical for routine CPET. Thus, advancing the scientific/methodologic understanding of noninvasive techniques for exercise cardiac hemodynamic assessment is clinically impactful in HFpEF. This study tested the concurrent validity of noninvasive acetylene gas (C2H2) uptake, echocardiography (ECHO), and oxygen pulse (O2pulse) for measuring/predicting exercise stroke volume (SV) in HFpEF.

Methods: Eighteen white HFpEF and 18 age-/sex-matched healthy controls participated in upright CPET (ages, 69 ± 9 yr vs 63 ± 9 yr). At rest, 20 W, and peak exercise, SV was measured at steady-state via C2H2 rebreathe (SVACET) and ECHO (SVECHO), whereas O2pulse was derived (=V˙O2/HR).

Results: Resting relationships between SVACET and SVECHO, SVECHO and O2pulse, or SVACET and O2pulse were significant in HFpEF (R = 0.30, 0.36, 0.67), but not controls (R = 0.07, 0.01, 0.09), respectively. Resting relationships persisted to 20 W in HFpEF (R = 0.70, 0.53, 0.70) and controls (R = 0.05, 0.07, 0.21), respectively. Peak exercise relationships were significant in HFpEF (R = 0.62, 0.24, 0.64), but only for SVACET versus O2pulse in controls (R = 0.07, 0.04, 0.33), respectively. Standardized standard error of estimate between techniques was strongest in HFpEF at 20 W: SVACET versus SVECHO = 0.57 ± 0.22; SVECHO versus O2pulse = 0.71 ± 0.28; SVACET versus O2pulse = 0.56 ± 0.22.

Conclusions: Constituting a clinically impactful step towards construct validation testing, these data suggest SVACET, SVECHO, and O2pulse demonstrate moderate-to-strong concurrent validity for measuring/predicting exercise SV in HFpEF.

Figure 1

Ordinary least squares univariate linear regressions between noninvasive techniques for measuring or predicting stroke volume (SV) at rest in healthy controls (CTL) or patients with heart failure with preserved ejection fraction (HFpEF). Solid line represents the goodness of fit line with 95% confidence limits as dotted lines. Goodness of model fit expressed as coefficient of determination (R²) from Cohen (10); modest = 0.02, moderate = 0.15, and strong ≥ 0.25. For panels A to F, the variable set on the abscissa was considered the practical method, whereas the variable set on the ordinate was considered the quasi-criterion method. ECHO = echocardiography; C₂H₂ = acetylene uptake; O₂pulse = oxygen pulse. Panels G to I align vertically with panels A to C or D to F for CTL or HFpEF, respectively. Predicted (abscissa), represents predicted SV using the respective goodness of fit equation (y = mx + b) from univariate regressions in panels A to F, whereas residuals (ordinate), represent the difference between observed SV (via either ECHO or acetylene) and predicted SV. In panel G, for corresponding regression plots where ECHO was set as the criterion see the Figure in Supplemental Digital Content 1.

Figure 2

Ordinary least squares univariate linear regressions between noninvasive techniques for measuring or predicting stroke volume (SV) at matched 20 W exercise in healthy controls (CTL) or patients with heart failure with preserved ejection fraction (HFpEF). Solid line represents the goodness of fit line with 95% confidence limits as dotted lines. Goodness of model fit expressed as coefficient of determination (R²) from Cohen (10); modest = 0.02, moderate = 0.15, and strong ≥ 0.25. For panels A to F, the variable set on the abscissa was considered the practical method, whereas the variable set on the ordinate was considered the quasi-criterion method. ECHO = echocardiography; C₂H₂ = acetylene uptake; O₂pulse = oxygen pulse. Panels G to I align vertically with panels A to C or D to F for CTL or HFpEF, respectively. Predicted (abscissa), represents predicted SV using the respective goodness of fit equation (y = mx + b) from univariate regressions in panels A to F, whereas residuals (ordinate), represent the difference between observed SV (via either ECHO or acetylene) and predicted SV. In panel G, for corresponding regression plots where ECHO was set as the criterion see the Figure in Supplemental Digital Content 1.

Figure 3

Ordinary least squares univariate linear regressions between noninvasive techniques for measuring or predicting stroke volume (SV) at peak exercise in healthy controls (CTL) or patients with heart failure with preserved ejection fraction (HFpEF). Solid line represents the goodness of fit line with 95% confidence limits as dotted lines. Goodness of model fit expressed as coefficient of determination (R²) from Cohen (10); modest = 0.02, moderate = 0.15, and strong ≥ 0.25. For panels A to F, the variable set on the abscissa was considered the practical method, whereas the variable set on the ordinate was considered the quasi-criterion method. ECHO = echocardiography; C₂H₂ = acetylene uptake; O₂pulse = oxygen pulse. Panels G to I align vertically with panels A to C or D to F for CTL or HFpEF, respectively. Predicted (abscissa), represents predicted SV using the respective goodness of fit equation (y = mx + b) from univariate regressions in panels A to F, whereas residuals (ordinate), represent the difference between observed SV (via either ECHO or acetylene) and predicted SV. In panel G, for corresponding regression plots where ECHO was set as the criterion see the Figure in Supplemental Digital Content 1.