Defining HFpEF: where do we draw the line?

This editorial refers to ‘Influence of ejection fraction on outcomes and efficacy of spironolactone in patients with heart failure and preserved ejection fraction’†, by S.D. Solomon et al., on page 455. Most people agree that there are two different kinds of heart failure (HF).1,2 There is one kind where ventricular function during systole and diastole is severely impaired and the left ventricle is dilated. This is referred to as HF with reduced ejection fraction (HFrEF). A number of therapies targeting neurohormonal stimulation and remodelling have been shown to improve outcome in these patients.1,2 There is another type of HF where function during diastole is markedly impaired, function during systole is modestly impaired, and the ventricle is generally not dilated.1–7 This type is referred to as HF with preserved ejection fraction (HFpEF). Therapies that improve outcome in people with HFrEF have not been found to help people with HFpEF, further supporting the idea that these two kinds of HF are fundamentally different.1,2 What remains unknown is what EF partition value should be used to distinguish the two varieties of HF.1 EF poorly gauges contractile function and should more accurately be considered as an index of ventricular–arterial coupling.5 Nonetheless, EF provides a convenient method to distinguish HF patients given the universal availability of echocardiography.2 Prior studies have variably defined HFpEF using EF values ranging from 40% to 55%, but it is likely that there is significant mechanistic heterogeneity within this range. Identification of the optimal EF partition value that allows us to categorize HF patients into more homogenous groups that display similar pathophysiologies, natural histories, and responses to treatment would be an important accomplishment. In this issue of the journal, Solomon and colleagues provide intriguing new data that might help us better distinguish HFpEF and HFrEF moving forward.8 The authors analysed data from the TOPCAT trial, which demonstrated that the mineralocorticoid receptor antagonist spironolactone did not improve the composite primary endpoint of cardiovascular death, HF hospitalization, or aborted cardiac arrest in people with HF and EF ≥45%.9 The authors compared baseline characteristics, clinical outcomes, and the efficacy of spironolactone as a function of EF at the time of entry into the trial. Baseline EF values ranged from 44% to 85%, with a clear-cut digit preference observed (interpreters rounding to integers of 5). As compared with subjects with lower EF, those with higher EF were older, more likely to be women, less likely to have had a prior myocardial infarction, and more likely to display co-morbidities such as hypertension and diabetes.8 The incidence of both the primary endpoint and cardiovascular death was substantially higher in subjects with lower EF, but HF symptom severity [New York Heart Association (NYHA) class] and hospitalization rates were similar. Notably, the authors found that the efficacy of spironolactone also varied with EF, being more effective to reduce the primary endpoint in people with HF and lower EF, with no benefit in the higher EF range (EF × treatment P = 0.046). This interaction was driven mostly by subjects enrolled in the Americas, who have previously been found to display higher event rates than participants from Russia and Georgia.10 The authors conclude that patient characteristics and outcome vary substantially by EF, even within this range that has historically been considered to be ‘preserved’, and that the potential efficacy of spironolactone was greatest in the lower EF range.8 The authors are to be congratulated on a clever and elegant analysis that greatly enhances our understanding of how even minor differences in EF may reflect more fundamental differences in underlying pathophysiology (Figure 1).8 The older age, greater female predominance, higher body mass, more hypertension, and less myocardial infarction in the participants with higher EF as compared with lower EF all mirror what has been previously observed when comparing people with HFpEF with those with HFrEF.2,11 It would have been interesting to compare other echocardiographic indices in the different EF groups, including left ventricular and atrial volumes, diastolic dysfunction indices, and pulmonary artery pressure estimates—to see if these also mirror differences (and similarities) observed between HFpEF and HFrEF. The increased mortality observed in the lower EF range in this TOPCAT analysis mirrors observations from a large meta-analysis comparing survival in HFpEF and HFrEF.11 Spironolactone treatment seemed to improve outcome more in the lower EF range and, while this result must be considered hypothesis-generating given the post-hoc nature of the analysis and neutral result of the overall trial,9 it is again consistent with prior trials in HFrEF that have clearly shown improved outcomes with spironolactone.12 Differing subtypes of heart failure with relative preservation in ejection fraction. In male patients with coronary disease and prior myocardial infarction, there may be modest depression in ejection fraction often accompanied by mild chamber dilatation and higher risk of cardiovascular death. These patients appear to be favourably responsive to renin-angiotensin-aldosterone system inhibitors such as spironolactone. In contrast, women with comorbidities associated with the metabolic syndrome typically do not develop chamber dilatation, and there is no reduction in global ejection fraction, despite the presence of systolic dysfunction. This phenotype is more typical of ‘garden variety’ heart failure with preserved ejection fraction as seen in the community and does not appear to be responsive to renin-angiotensin-aldosterone system inhibition. While cardiovascular death rates are lower in this more typical heart failure with preserved ejection fraction subtype, symptoms and morbidity are equivalent to the lower ejection fraction subgroup. These results strongly support what has long been suspected—that people with mildly depressed EF are probably best categorized and treated as if they had HFrEF (Figure 1). Solomon and colleagues observed that risk increased below an EF of 50–55%, and this is also where spironolactone tended to show greater effectiveness, supporting this partition value.8 Of course, there are considerable limitations associated with using EF alone to segregate HF: there is variability in EF itself, variability in EF measurement, and shortcomings in the utility of EF to relate to other important biological differences between patient types. However, EF is firmly entrenched at this point in clinical practice, and these data certainly support 50% as the better choice than 40% or even 45%. That finding alone has impact and value. The nil effect of spironolactone in people with typical ‘garden variety’ HFpEF (EF ≥60%, older, hypertensive, female, obese)8 inches the door closer to being shut on renin–angiotensin–aldosterone (RAAS) inhibitors in this syndrome.9,13–15 Benefits from RAAS antagonists in HFrEF seem to relate to how effectively they elicit reverse remodelling,16 and perhaps the lack of chamber dilatation in HFpEF makes this group less responsive to agents that reduce ventricular size.1 Notably, CHARM-preserved was the RAAS trial in HFpEF that was closest to being a positive study13 and, compared with the other trials testing these agents, it enrolled patients with the lowest EF (≥40%) and included more men and more people with coronary disease.9,13–15 These data fit perfectly with the group differences and differential response to spironolactone observed in the lower EF patients in the current analysis.8 Cardiovascular mortality was higher with decreasing EF, even in this cohort with relatively preserved values.8 However, it is extremely important to recognize that HF severity in terms of symptoms and hospitalization burden was similar. People suffering with HFpEF complain of debilitating symptoms of dyspnoea and fatigue that are caused by elevation in filling pressures during even modest exercise.17 Many of these patients are not admitted to hospital, and do not have high natriuretic peptide levels, so they would not have been eligible for enrolment in TOPCAT.9 Regulatory bodies continue to embrace survival and hospitalization as necessary endpoints for trials seeking to guide an indication, but HF symptoms, exercise capacity, and quality of life may be even more relevant to our patients. The myopic view that death and hospitalization are the only endpoints of relevance in this enormous and rapidly expanding population is antiquated and needs to change right now. TOPCAT was ‘another neutral trial’ in HFpEF,9 and the observations from Solomon and colleagues that spironolactone worked even less well in what we usually consider to be more typical HFpEF8 could easily engender scepticism and despair. This would be a mistake. We forget that it took us many years to crack the code and identify the central importance of neurohormonal activation in HFrEF. Based upon the trials to date, it does not appear that RAAS activation is as central in HFpEF. That does not mean that there are not primary unifying abnormalities that can be targeted (Figure 1). It simply means that we need to work harder to find them. These data from TOPCAT are sending us a clear and unequivocal message: it's time to roll up our sleeves and get back to work. Conflict of interest: none declared.