This study confirms that LV Ees and Ea are the main determinants of LVEF. Not surprisingly, we also corroborated the strong relationship between LVEF and both VAC and LVeff, which are mathematically coupled to Ees and Ea. Thus, LVEF is mainly determined by the coupling between the LV systolic function and the arterial system and less so to changes in preload and heart rate. As such, LVEF represents more of a global index of cardiovascular performance rather than an absolute measure of LV contractility. We also propose an afterload correction to LVEF that significantly improved the performance over the standard LVEF calculation for estimating LV contractility during changes in loading conditions, while based on easily available parameters.
Ideally, an index of contractility index should be sensitive to changes in myocardial contractile state but indifferent to loading conditions [6]. Although LVEF has been traditionally used as a clinical measure of LV systolic performance, its dependence on preload and especially afterload changes have been well documented experimentally over five decades ago by Krayenbuhl et al. [23], later corroborated in isolated canine ventricles by Kass et al. [24] and theoretically described by Robotham [6]. However, the recognition that LVEF does not only depend solely upon myocardial contractility but also upon other determinants of the ventricular function has been increasingly recognized in critically ill patients with the growing use of echocardiography in ICU to assess LV function [3, 9, 25]. For example, the afterload influence in LVEF has been particularly evident and documented in septic patients [8, 9, 26]. Impaired intrinsic contractility and vasoplegia are the hallmark of the hemodynamic disorders in septic shock [27, 28]. Under these conditions, a normal LVEF may not necessarily indicate a normal LV contractility, but rather reflect the profoundly impaired LV afterload [9, 29]. Similarly, a low LVEF within the context of a decreased LV afterload would support the diagnosis of severely depressed LV systolic function. On the other hand, a reduction in LVEF after correcting arterial hypotension with vasopressors need not reflect a worsening of LV systolic function, but only the unmasking of existing LV dysfunction by normalization of LV afterload [10]. Only if LVEF still remains low after restoring arterial pressure and afterload, then an impairment in LV contractility is likely. This reasoning could explain the large variability of LV dysfunction reported in previous clinical studies and the lack of association between LVEF and mortality in septic shock [8, 30]. As loading conditions can be significantly altered by vasopressors or fluid administration, differences on the timing and the amount of hemodynamic resuscitation could have influence the reported value of LVEF regardless the intrinsic contractile state. Accordingly, considering the multiple factors that can influence LVEF in septic shock patients, where perturbations of loading conditions are particularly frequent, the interpretation of LVEF as an index of LV contractility should be done cautiously and after considering the potential impact of these factors [25].
While afterload changes largely affected LVEF, preload variations, even when abrupt and of large amplitude, such as during bleeding and fluid loading, failed to greatly influence LVEF. This was particularly evident even if CO significantly increased during fluid loading (from 7.89 ± 1.7 l/min to 9.11 ± 2.42 l/min; p = 0.01), while LVEF showed only a modest decrease (from 54 ± 13% to 47 ± 10%; p = 0.02). This is presumably related to the fact that LVEF normalized changes in SV by EDV, while a reduction in LVEF after fluid infusion would be associated with an increase with EDV without a subsequent improvement in SV (non-preload dependency). Therefore, a decrease in LVEF after fluid administration could be indicative of a lack of preload responsiveness, while an unchanged or an increased LVEF would denote a significant increase in SV with volume administration (fluid responsiveness).
Additionally, LVEF is a function of both SV and EDV. Patients with sustained lusitropy but preserved SV will paradoxically have a reduced LVEF, owing to the increased EDV, but have better cardiovascular reserve. This assumption supports the multiple studies reporting a decreased LVEF in survivors of severe sepsis as compared to non-survivors [31]. Non-survivors presumably are unable to relax appropriately during diastole, while both survivors and non-survivors have sepsis-induced impaired systolic function. This hypothesis is supported by the association found between diastolic disfunction and mortality in septic patients [32, 33]. As we did not also study the impact of sepsis in our porcine model, we cannot speculate further on this hypothesis. Moreover, the effects of our interventions on diastolic function, though measurable, were small and not relevant, because they minimally altered EDV. However, in the setting of altered diastolic function, as may occur in sepsis and LV hypertrophy, these effects may become relevant. Further focused investigation of these issues would be needed to address these concerns.
Our in vivo experimental study with an intact cardiovascular system confirms previous observations about the influence of the cardiac loading conditions on LVEF [23, 24, 34,35,36], but more importantly, it also provides the empirical demonstration about the real nature of LVEF. Since LVEF is primarily governed by the influence of Ees and Ea, it mainly reflects the balance between LV contractile function and the arterial system. Therefore, LVEF should be considered as a ventriculo-arterial coupling index and a variable mostly related to LV systolic mechanical efficiency, but not a pure measure of LV systolic performance [6, 37]. Furthermore, even if LVEF is associated with a unique VAC and LV mechanical efficiency state, the same LVEF level can be obtained with different values of Ees and Ea. Thus, the clinical utility of measures of LVEF under conditions of varying cardiovascular states is unclear and may be misleading.
Thirty years ago Robotham stated that LVEF “reflects the integrated system’s ability to cope, under the conditions at the time of measurement, with abnormalities in preload, afterload and/or contractility”, and therefore, “ejection fraction becomes a measure, not of ventricular performance, but of the integrated system’s performance in dealing with a pathological process” [6]. Although his measures were less accurate than ours, our data agree with his fundamental observation. Accordingly, an abnormal LVEF represents the inability of the cardiovascular system to modulate the current contractile and loading conditions for sustaining normal homeostasis. On the contrary, a normal LVEF does not necessarily mean that LV contractility is preserved, but it rather represents a composite variable expressing the cardiovascular system’s response to the current loading and contractile conditions. Practically speaking, a low LVEF should always be considered as a sign of impaired cardiovascular function, reflecting an unfavorable LVeff and VAC, but it does not inform about the underlying mechanisms leading to its low value, which could be an impaired contractility, increased afterload or both. Furthermore, a normal LVEF, particularly in critically ill patients, should never be interpreted as evidence of normal LV systolic function, especially in the presence of hypotension and systemic hypoperfusion. Relevant to this last point, increased LVEF due to a predominant reduction on LV afterload, even in the presence of an impaired LV contractility, has been associated with a higher mortality [8, 38, 39].
While standard LVEF remains of clinical interest, since it gives a global picture of LV pump function for the given contractility and loading conditions, we propose reporting an Ea-adjusted LVEF to significantly improve the ability of LVEF to tracking changes in Ees during acute perturbations of LV afterload, without affecting the performance during changes in contractility. Moreover, we have recently demonstrated that radial, femoral and aortic MAP values are interchangeably as valid surrogates for the estimation of Pes, so LVEF can be adjusted using a peripheral estimation of Ea regardless the arterial pressure measurement site [12]. Since the additional information required to calculate LVEFEA can be obtained using currently available measures of arterial pressure and ventricular function, its application for estimating LV contractility could be easily implemented. This afterload-corrected LVEF could be particularly helpful for identifying those patients with depressed LV contractility in the setting of altered loading conditions, as seen during septic shock. Furthermore, identification of depressed contractility by the Ea-corrected LVEF in the context of a persistent tissue hypoperfusion could theoretically better define the population potentially eligible for inotropic therapy. However, the usefulness and superiority of these surrogates in the clinical practice needs further clinical validation.
Our results require a few comments. First, we did not use echocardiography for calculating LVEF, which is the standard in clinical practice. Instead, we used the data obtained from the LV PV analysis by the classical conductance catheter technique. This method has been demonstrated to provide the most comprehensive description of the LV function [40], whereas estimation of LVEF by standard echocardiography relies on geometrical assumptions and is prone to technical inaccuracies and inter-/intra-observer variability. Thus, echocardiographic estimates of LVEF should show even greater variance than our findings, underscoring further the lack of sensitivity of LVEF estimates to track LV contractility. However, our results are not limited to the use of this conductance catheter methodology, since they should remain valid to any estimation of LVEF, such as that obtained from the 3D echocardiography or global ejection fraction from transpulmonary thermodilution. However, as our LVEF estimation was based on the volumetric data from the conductance catheter, these clinical estimates of LVEF would still rely on the accuracy of their measurements and their assumptions about the LV geometry. Finally, our study was performed on pigs receiving anesthesia, so our results should be interpreted with caution when extrapolating to human cardiovascular physiology in awake individuals.