Left Ventricular Ejection Fraction in Heart Failure: Crazy, Stupid Love—and Maybe, Redemption

The measurement of left ventricular ejection fraction (LVEF) was first proposed by Stuart Bartle, a young cardiovascular researcher at University of Virginia in the early 1960s. ¹ Based on images derived from contrast cineangiography, he coined the term ejected fraction to describe the ratio of stroke volume to LV end‐diastolic volume (LVEDV). At the time, no one thought that the calculation of LVEF represented a major intellectual achievement, either conceptually or physiologically. As I can personally affirm, ² Bartle himself dismissed the usefulness of the concept entirely, as evidenced by his subsequent decision to spurn cardiology and embrace psychiatry as his primary career path, becoming a highly distinguished psychotherapist and humanitarian. ²

EF Does Not Reflect Systolic Function

The circulation takes great pains to maintain cardiac output and stroke volume. Therefore, if stroke volume is held constant, the ratio of stroke volume to LVEDV (the LVEF) is essentially the inverse of the LVEDV. ³ In the heyday of contrast ventriculography, the measurement of LVEDV was based on exceedingly tedious planimetry. When noninvasive imaging emerged as a tool for the assessment of LV structure and function, a single dimension on M‐mode echocardiography provided the basis for the most common estimate of LVEDV, but a single beam–based estimate of LVEDV was based on many unverifiable assumptions about the shape of the left ventricle, especially in an era when segmental myocardial infarction represented the principal driver of LV dysfunction. Furthermore, depending on the angle of the transducer, the measurement of LV end‐diastolic dimension could vary wildly, and because the value was cubed to yield volume, the errors in estimating LVEDV could be magnified exponentially.

How could this dilemma be solved? For more than a decade, radionuclide ventriculography allowed for the easy collection of the number of tracer counts in end systole and end diastole, and, thus, allowed for the estimation of LVEF, without the need for a validated measure of LVEDV. ⁴ The dominance of radionuclide ventriculography (before the clinical availability of 2‐dimensional Doppler echocardiography) solidified the impression that the measurement of LVEF had meaning.

Our Weird Crazy Love Affair With EF

The crystallization of LVEF as a metric of importance was accelerated when numerous epidemiological studies reported that LVEF had prognostic significance across a broad range of cardiovascular disorders. ⁵ It is highly likely that LVEDV would have performed as well or even better than LVEF in those studies. Yet, once these studies were performed, it was not feasible to go back and suggest that the researchers should have focused their attention on volume, rather than chamber emptying during systole.

To complicate matters further, most cardiologists believed (erroneously) that LVEF represented systolic function. However, the primary action of cardiomyocytes in systole is to shorten, and, thus, the most physiological metric of systolic function is LV internal shortening, typically measured as strain, not LVEF. ³ There is an inverse relationship between LV strain and LVEDV, but only if the left ventricle remodels and enlarges in proportion to the decline in contractile function. That assumption was typically true in ischemic or nonischemic dilated cardiomyopathy. However, in most patients with heart failure with preserved ejection fraction (HFpEF), systolic function is meaningfully impaired with only modest LV dilation, and, in cardiac amyloidosis, strain is dramatically impaired in a left ventricle that is not dilated at all. In cardiac amyloidosis, cardiomyocyte contractile dysfunction is horrifically diminished, and yet, the LVEF can lie within the range of normal values, because LVEDV is not increased. LVEF does not measure systolic function. ³

The Measurement of EF Continues to Break Our Hearts

Even if cardiologists decided to embrace LVEDV rather than LVEF, it would not solve the most important issue, ie, the measurement of either metric is highly variable and not easily reproduced. Magnetic resonance imaging, radionuclide ventriculography, and 2‐dimensional Doppler echocardiography all yield different results for LVEF in the same patient. Intrapatient variability of repeat measures of LVEF using the same method within short periods exceeds 7%, as an absolute difference in either direction. Hence, in a patient with an EF of 35%, repeat imaging could yield values of 28% to 42%, with dramatically different clinical implications. Furthermore, if a patient has atrial fibrillation, the echocardiographer seeking to measure LVEF has an unlimited opportunity for mischief. Ideally, the measurement of LVEF in patients with atrial fibrillation should be based on the average of 15 beats, but no one does this in clinical practice. Given these uncertainties, some echocardiographers provide a range, usually with a distinct digit preference, selecting numbers ending in a 0 or 5. ⁶ Others simply supply a descriptive term (eg, mildly impaired), a test result that is highly subjective, exceptionally uninformative, and clinically meaningless.

Our Tortured History of Defining HF With a Reduced, Midrange, or Preserved EF

Sadly, our reliance on LVEF became more firmly entrenched when various threshold values for LVEF were incorporated as an inclusion criterion in the design of large‐scale trials of drugs for the treatment of HF. Trials in the 1980s and 1990s evaluating angiotensin‐converting enzyme inhibitors, β‐blockers, and mineralocorticoid receptor antagonists typically required patients to have HF with an LVEF ≤25%, ≤30%, or ≤35%. ⁷ In the PARADIGM‐HF (Prospective Comparison of ARNI With an ACE‐I to Determine Impact on Global Mortality and Morbidity in Heart Failure) trial with sacubitril/valsartan, the trial initially specified an LVEF ≤40% and then amended the protocol to patients with an LVEF ≤35%. In these landmark trials, neurohormonal antagonists reduced cardiovascular death and all‐cause mortality in patients with an LVEF ≤35%. There are only sparse data on the survival effects of neurohormonal antagonists in patients with an LVEF of 36% to 40%.

The Nonsensical Invention of HFpEF and HF With Midrange EF

In 2003, the steering committee of the CHARM (Candesartan in Heart Failure Assessment of Reduction in Mortality and Morbidity) trial made the fateful decision to evaluate the effects of candesartan in patients with HF across the entire spectrum of LVEF. They divided patients into 2 distinct trials: one that focused on patients with an LVEF ≤40% and the other on patients with an LVEF >40%. ⁸ The former were referred to as having HF with reduced EF (HFrEF) and the latter were designated as having HFpEF. The CHARM investigators used the term preserved (instead of normal) because it was understood that an LVEF of 41% to 56% was still reduced, as compared with the normal range, typically >55% to 57%. The choice of an LVEF of 40% as a decision line was arbitrary; it was not based on any clinical, pathophysiological, or therapeutic evidence. Nevertheless, for the next 20 years, the term HFpEF remained entrenched in the cardiology literature.

Some HFpEF trials focused on patients with an LVEF ≥50%, and, as a result, little information was gathered in patients with an LVEF of 41% to 49%. Recognizing that patients with an LVEF of 41% to 49% were not being enrolled in clinical trials, the 2016 iteration of the European Society of Cardiology HF guidelines suggested that patients with an LVEF of 41% to 49% deserved further study, but, unfortunately, to reinforce this intent, they assigned a new designation to these patients (ie, they were referred to as having HF with midrange EF [HFmrEF]). ⁶ , ⁹ Suddenly, physicians throughout the world believed that the experts had created a new distinctive phenotype of HF, and thousands of papers were published on HFmrEF.

The concept that patients with an LVEF of 41% to 49% had a distinct cardiovascular disorder was absurd, since this exceptionally narrow range was inconsistent with the variability inherent in the repeated measurement of LVEF. An LVEF of 44% assessed on a Monday could be 37% on Wednesday and 51% on Friday, leading a physician to diagnose HFrEF, HFmrEF, and HFpEF in the same patient within a 7‐day period. Would treatments started for HFrEF on Wednesday need to be stopped 2 days later? The possibility for variance was heightened if EF was measured using 2 different modalities, and if the patients had atrial fibrillation, the echocardiographer had liberty to report a wide range of values for LVEF, encompassing (potentially) 3 different phenotypes of HF in the same individual at the same time. This hypothetical scenario might seem far‐fetched, but the Stockholm‐CELOSIA study demonstrates that it is not.

The Stockholm‐CELOSIA Study

In this issue of the Journal of the American Heart Association (JAHA), Christersson et al ¹⁰ report their analysis of the Stockholm‐CELOSIA database, a registry of patients with HF, chronic kidney disease, or diabetes treated between January 2015 and December 2020. All 9716 patients with new‐onset HF who had a measurement of LVEF before and after the diagnosis, at least 15 days apart, were identified in order to discern the stability of a diagnosis of HFrEF, HFmrEF, or HFpEF over time. The investigators found the project to be challenging, since, in many instances, only a qualitative description (rather than a numerical value) was provided. Furthermore, realizing the extraordinary intrapatient variability, repeat measurements within 14 days were excluded, although such information would certainly have been informative and entertaining, and the timing of repeat measurements of EF was not standardized. Approximately 35% of patients had an interval clinical event between the 2 LVEF measurements, including events that would have decreased LVEF (eg, myocardial infarction), increased LVEF (eg, heart transplantation), or markedly augmented the variability of LVEF (eg, atrial fibrillation).

Christersson et al ¹⁰ report several striking findings. First, the distribution of values for LVEF followed a normal distribution, with a modal value of 40% to 45%. There was no evidence for 2 or 3 distinct subgroups that might justify the designations of HFrEF, HFmrEF, and HFpEF. Interestingly, the modal value of 40% to 45% lies precisely in the range of patients who are considered to have HFmrEF, indicating that the measurement of EF typically enhanced uncertainty, rather than provided clarity. Second, the SD of the within‐person variance was 7.4%, indicating that about two‐thirds of the repeat measurements would fall into a range of 7 units below and 7 units above the recorded value. Third, given the exceptionally narrow definition of HFmrEF, there was a 75% to 80% likelihood that a patient with an EF of 41% to 49% would have an LVEF <41% or >49% on repeated measurement, typically performed within a year. The likelihood of nonreproducibility would likely have been higher if the measurement of LVEF had been repeated more frequently. Therefore, if a physician wanted the diagnosis of HFmrEF to disappear, they only needed to repeat the LVEF—voila! It was gone.

Importantly, the original categorization of the HF phenotype was most likely to be confirmed in men with an initial diagnosis of HFrEF (40% likelihood of not transitioning to a different category) and in women with an initial diagnosis of HFpEF (60–70% likelihood of not transitioning to a different category). The transition of a patient from HFrEF to HFpEF occurred in 25% of patients, presumably related to an improvement in EF following prolonged use of neurohormonal antagonists. In marked contrast, the transition of a patient from HFpEF to HFrEF was <10%, strongly supporting the concept that the minimally dilated left ventricle in patients with HFpEF did not subsequently remodel and enlarge (ie, HFrEF is not an eventual outcome of patients with HFpEF).

Redemption of EF as a Classification Tool for HF

Given the observations from the Stockholm‐CELOSIA study, should the classification of HFmrEF be discarded? Such a conclusion would seem to be obvious, but, in reality, the difficulty with HFmrEF lies in what not it represents but with how it is defined. Despite the silliness of LVEF, it seems unrealistic to expect that cardiologists will discard it. So how do we put the measurement of LVEF to its best use?

Fundamentally, as noted earlier, there are 2 principal processes that can lead to HF (Figure). ⁷ First, the destruction of cardiomyocytes can be accompanied by proportional enlargement of the left ventricle through a process of remodeling. The LV dilatation leads to a shift in the LV end‐diastolic pressure–volume relationship downwards and to the right. The increase in LV capacitance allows the left ventricle to: (1) fill with a modest increase in LV end‐diastolic pressure and (2) maintain stroke volume even when cardiomyocyte shortening is impaired. Because the left ventricle markedly dilates, the LVEF (the inverse of LV volume) is meaningfully depressed. The phenotype of these patients is a dilated left ventricle with an LVEF ≤35%. These patients respond dramatically to neurohormonal antagonists, with a reduction in both all‐cause mortality and HF hospitalizations.

Figure . Reconceptualization of heart failure (HF) with a reduced, mildly reduced, and normal ejection fraction (EF).

Pathophysiological features and therapeutic responses in patients with HF with reduced EF (HFrEF), HF with mildly reduced EF (HFmrEF), and HF with a normal ejection fraction (HFnEF). The term HF with preserved EF has been discarded. The category of HFmrEF has been expanded and now encompasses patients with a left ventricular EF >35% to <60%. LV indicates left ventricular; and SGLT2, sodium‐glucose cotransporter 2. Adapted from Packer ⁷ with permission.

Second, HF may result from a marked decrease in LV distensibility and increase in LV stiffness. The left ventricle cannot dilate to accommodate increased venous inflows, and when venous return is markedly enhanced by sympathetic blood redistribution during exercise, LV end‐diastolic pressures increase dramatically and disproportionately. ⁷ , ¹¹ , ¹² Because of the absence of LV dilatation, the LV end‐diastolic pressure–volume relationship is shifted upwards and to the left, possibly as a result of LV infiltration or the result of pericardial constraint (Figure). LV remodeling is not relevant in these patients, and the phenotype mimics a form of cardiac contracture (a nondilated left ventricle and left atrium), with an LVEF ≥60%, typically seen in elderly women with hypertension who have minimal symptoms at rest but marked exercise intolerance. ¹³ , ¹⁴ The proper designation of these patients is HF with normal EF. These patients do not respond favorably to neurohormonal antagonists or sodium‐glucose cotransporter 2 inhibitors. ¹⁵ , ¹⁶ Other interventions (eg, pericardiectomy) are being explored. ¹⁷

Between an LVEF of >35% and <60% lies a substantial proportion of the patients with HF in the community. These patients are typically women with significant visceral adiposity, epicardial adipose tissue expansion, and underlying myocardial inflammation and fibrosis. ¹³ , ¹⁴ , ¹⁸ There is mild to moderate contractile function, and the left ventricle is enlarged, but only mildly so. Therefore, the LVEF is below normal (ie, <60%) but not markedly depressed. The LV end‐diastolic pressure–volume relationship is shifted modestly, downwards and to the right. These patients have a mild form of HFrEF, and they respond favorably to neurohormonal antagonists, but principally with a reduction in the risk of HF hospitalization, and not cardiovascular death. ¹⁸ This phenotype is HFmrEF, but it is defined as >35% and <60%, rather than 41% to 49%.

Conclusions

Having embraced LVEF for irrational reasons for decades, it is time that cardiologists put the concept to good use. It is time to retire and discard the artificial nonphysiological cut points of 40% and 50%. The category of HFmrEF should be greatly expanded to encompass all patients with a mild form of HFrEF, ranging from >35% and <60%, who share a common pathophysiology and who retain a nondramatic responsiveness to neurohormonal antagonists. It is time to abandon completely the nonuseful heterogenous term HFpEF, a term invented 20 years ago as an artificial contrivance.

In an ideal world, we might be tempted to define HFrEF, HFmrEF, and HF with normal EF, not by values for LVEF but by values for LVEDV. However, in the current era, that is not a realistic goal, especially since values for LVEDV need to be normalized for age and sex. It is better to allow cardiologists to continue their irrational embrace of LVEF but teach them how to use the information in a way that is based on pathophysiology and the results of large‐scale trials and promotes optimal care.

Disclosures

During the past 3 years, M.P. reports personal fees for consulting from 89bio, Abbvie, Actavis, Altimmune, Alnylam, Amarin, Amgen, Ardelyx, AstraZeneca, Attralus, Biopeutics, Boehringer Ingelheim, Caladrius, Casana, CSL Behring, Cytokinetics, Imara, Lilly, Medtronic, Moderna, Novartis, Pharmacocosmos, Reata, Relypsa, and Salamandra.