The corrected left ventricular ejection fraction: a potential new measure of ventricular function

Abstract

Left ventricular ejection fraction (LVEF) has a limited role in predicting outlook in heart diseases including heart failure. We quantified the independent geometric factors that determine LVEF using cardiac MRI and sought to provide an improved measure of ventricular function by adjusting for such independent variables. A mathematical model was used to analyse the independent effects of structural variables and myocardial shortening on LVEF. These results informed analysis of cardiac MRI data from 183 patients (53 idiopathic dilated cardiomyopathy (DCM), 36 amyloidosis, 55 hypertensives and 39 healthy controls). Left ventricular volumes, LVEF, wall thickness, internal dimensions and longitudinal and midwall fractional shortening were measured. The modelling demonstrated LVEF increased in a curvilinear manner with increasing mFS and longitudinal shortening and wall thickness but decreased with increasing internal diameter. Controls in the clinical cohort had a mean LVEF 64  ±  7%, hypertensives 66  ±  8%, amyloid 49 ±  16% and DCM 30  ±  11%. The mean end-diastolic wall thickness in controls was 8  ±  1 mm, DCM 8  ±  1 mm, hypertensives 11  ±  3 mm and amyloid 14  ±  3 mm, P < 0.0001). LVEF correlated with absolute wall thickening relative to ventricular size (R² = 0.766). A regression equation was derived from raw MRI data (R² = 0.856) and used to ‘correct’ LVEF (EF_c) by adjusting the wall thickness and ventricular size to the mean of the control group. Improved quantification of the effects of geometric changes and strain significantly enhances understanding the myocardial mechanics. The EF_c resulted in reclassification of a ‘ventricular function’ in some individuals and may provide an improved measure of myocardial performance especially in thick-walled, low-volume ventricles.

Appendix

Midwall fractional shortening

Midwall fractional shortening was estimated using the following established [26]:

$${\text{mFS}}\left( \% \right)\, = \,\left( {\left( {{\text{LVIDd}}\, + \,{\text{EDWT}}} \right){-}\left( {{\text{LVIDs}}\, + \,{\text{H}}} \right)} \right)/\left( {{\text{LVIDd}}\, + \,{\text{EDWT}}} \right)\, \times \,{1}00$$

where:

$${\text{H}}\, = \,\left( {\left( {{\text{LVIDd}}\, + \,{\text{EDWT}}} \right)^{{3}} {-}\left( {{\text{LVIDd}}} \right)^{{3}} \, + \,\left( {{\text{LVIDs}}} \right)^{{3}} } \right)^{{{1}/{3}}} {-}{\text{LVIDs}}$$

Regression equations

Multiple linear regression (n = 183) of MRI cohort. D = left ventricular end-diastolic diameter (mm), W = end-diastolic wall thickness (mm), L = end-diastolic left ventricular length (mm), ε_L = long axis shortening (%), ε_C = midwall (circumferential) fractional shortening (%). Four statistical models were produced (Eqs. 1, 2, 3, 4). The highest correlation coefficient was obtained using all five input variables (Eq. 1) and the lowest with three variables including longitudinal shortening (Eq. 3):

5 degrees of freedom (D, W, ε_L, ε_C, L)

$${\text{EF}}\% \, = \, - 0.{\text{423D}}\, + \,{2}.0{7} {\text{W}} - 0.{648}\varepsilon_{{\text{L}}} - {2}.0{7}\varepsilon_{{\text{C}}} - 0.{\text{174L}}\, + \,{38}.{9},{\text{ R}}^{{2}} \, = \,0.{856},{\text{ p}}\, < \,0.000{1}$$

(1)

4 degrees of freedom (D, W, ε_L, ε_C)

$${\text{EF}}\% \, = \, - 0.{\text{577D}}\, + \,{1}.{81} {\text{W}} - 0.{54}0\varepsilon_{{\text{L}}} - {2}.0{7}\varepsilon_{{\text{C}}} \, + \,{33}.{8},{\text{ R}}^{{2}} \, = \,0.{853},{\text{ p}}\, < \,0.000{1}$$

(2)

3 degrees of freedom (D, W, ε_L)

$${\text{EF}}\% \, = \, - 0.{\text{934D}}\, + \,{1}.{37} {\text{W}} - {2}.{17}\varepsilon_{{\text{L}}} \, + \,{7}0.0,{\text{ R}}^{{2}} \, = \,0.{747},{\text{ p}}\, < \,0.000{1}$$

(3)

3 degrees of freedom (D, W, ε_C)

$${\text{EF}}\% \, = \, - 0.{\text{629D}}\, + \,{1}.{61} {\text{W}} - {2}.{37}\varepsilon_{{\text{C}}} \, + \,{39}.{8},{\text{ R}}^{{2}} \, = \,0.{848},{\text{ p}}\, < \,0.000{1}$$

(4)