Heart Failure with Preserved Ejection Fraction: In Perspective

Left Ventricular Ejection Fraction (LVEF):

The assessment of the fraction of the left ventricular volume ejected per beat (LVEF) was introduced as a measurement to “permit a more complete understanding of the activity of the left ventricle in normal subjects and in patients with a variety of cardiovascular abnormalities.”¹² Even in this early quantitative study, those with overt cardiac disease ejected a lower proportion of their left ventricular volume with each contraction. LVEF was shown to both pre and afterload dependent and as such, there were other invasive measurements that better reflected ventricular contractility.¹³ Perhaps because of convenience and ease of noninvasive assessment, despite this load dependent shortcoming, LVEF became the leading term clinicians used to characterize the left ventricular function of their patients.

Diastolic HF:

The repeated observations that despite the same signs and symptoms, a proportion of those with HF did not have major overt reductions in systolic function led to the recognition that HF could also be the consequences of abnormalities in diastole. Initially termed diastolic HF, supporting hemodynamic studies demonstrated high filling pressures without marked increases in ventricular chamber size.¹⁴ The hallmark leftward shifted ventricular pressure-volume relationship indicated that some intrinsic or extrinsic constraining force was impeding ventricular filling rather than emptying.¹⁵ Pericardial disorders, by limiting ventricular filling, also shift both right and left ventricular pressure volume curves upward or to the left producing a similar congestive presentation. However, when clearly attributed to the pericardium it is not mechanistically generally considered HF. Rather, diastolic HF implies that there is a myocardial process impairing filling. This can be either passively from increased ventricular stiffness and/or actively from slower relaxation. A variety of abnormalities in cardiac structure and/or function, including an increase in intrinsic myocardial stiffness of the myocytes and extracellular collagen matrix, infiltrative disorders, or abnormalities in the cardiac and systemic microvasculature are associated with abnormalities in filling. Ventricular relaxation was shown to be an active process with ischemia adversely impacting filling.¹⁶ As recognition increased about the occurrence of diastolic HF, the markedly concentrically hypertrophied left ventricle with small to normal cavity volumes (increased mass to volume ratio) and maintained systolic function became the expected trademark feature of diastolic HF.¹⁷ However, the magnitude of the problem was yet to be generally appreciated.

It was only with the more widespread use of non-invasive assessments of ventricular function by the 1980s that it become better appreciated that congestive HF was not synonymous with reduced ejection fraction. In an early report of echocardiograms from patients with HF, the investigators were clear to state “neither history, physical examination or even chest radiographic findings were able to discern patients with normal versus the abnormal reduced ejection fraction.”¹⁸ Despite this emerging recognition that the constellation of signs and symptoms justifying the diagnosis of HF could be present in patients with an apparently normal ejection fraction, the magnitude of the problem was not initially adequately appreciated. In the mid-1980s, two independent though similar reports from respected nuclear cardiology laboratories raised the profile of diastolic HF from the occasional case report to the justified not to be ignored important component of the HF population.^{19, 20} Both evaluated consecutively referred patients with a clinical diagnosis of HF. Of 188 patients reported by Dougherty, 67 (36%) had an ejection fraction <45 % and even this early study specifically showed that there was no relationship between the ejection fraction and functional class (NYHA).¹⁹ Soufer identified 58 patients with a clinical diagnosis of HF referred to their nuclear laboratory for a quantitation of ejection fraction and found a value greater than 45% in 42% of the referrals during a 3-month sampling period.²⁰

There were however still lingering doubts that those found to have signs and symptoms of HF with an apparently normal ejection fraction might have had transient systolic dysfunction, which was responsible for the clinical presentation. This concept was relegated to unlikely by an important study of paired measurements of ejection fraction, the first during the acute presentation in decompensated HF and the second following treatment and recovery from the pulmonary edema.²¹ The similar measurements of ejection fraction despite the marked difference in clinical status refocused attention to mechanisms producing elevated filling pressures with apparently adequate systolic function. With this acceptance of diastolic HF as an important entity, there was a greater emphasis on finding mechanisms and noninvasive indices of diastolic dysfunction.

Taxonomy Diastolic HF to HF with preserved Ejection Fraction (HFpEF)

In the 1980’s and 90’s, the term congestive HF was widely used to encompass all those with the clinical signs and symptoms of HF. The epidemiologic studies had yet to reveal the magnitude of the extent of those with no or minimal apparent abnormalities in ventricular ejection. Diastolic HF aptly identified this understudied, poorly understood cohort of patients.¹⁵ In this same era, major randomized clinical trials testing therapies with all-cause mortality as a primary outcome were just beginning to address patients with congestive HF. In fact, neither of the first two of these trials, the Veterans Administration Cooperative Study (V-HeFT) and Cooperative North Scandinavian Enalapril Survival Study (CONSENSUS) specifically excluded those with what would be considered diastolic HF.^{22, 23} The V-HeFT investigators subsequently compared the 13% of their patients with an ejection fraction (EF) over 45% (mean 54%), to the majority with low EF (mean 26%) and reported only a minor difference in peak exercise oxygen consumption but a clearly lower rate of death in those with the more normal EF.²⁴ This observation had important repercussions on the subsequent early mortality HF trials, which then used an arbitrary upper limit of EF above 35 or 40% as an exclusion to attempt to make sample size projections more practical.^25–27 Around that time based on encouraging surrogate outcome studies, oral positive inotropic agents appeared to be a promising therapeutic breakthrough. With the exception of a subgroup in the DIG trial,²⁸ those with diastolic HF were also understandably excluded from these mortality outcome trials.^29–31 As clinical trial experience accumulated, the rates of death from the completed trials with an upper EF exclusion served as the foundation for sample size calculations of future randomized controlled clinical trials (RCTs).

The introduction of angiotensin receptor blockers (ARB), as an approved antihypertensive therapy with theoretical additional properties considered beneficial for patients with HF spurred investments in major outcome trials.^{32, 33} Evaluation of Losartan in the Elderly (ELITE II) directly tested whether the ARB losartan would have a superior influence on survival compared to the angiotensin converting enzyme inhibitor (ACEi) captopril on all-cause mortality in HFrEF.³⁴ Despite the anticipated lower withdrawal rate in the ARB group, there was no advantage on rates of death.³⁴ The Candesartan in HF Assessment of Reduction in Mortality and morbidity (CHARM) was a broader program of three distinct but complementary populations of patients based on ACEi use and LVEF, each independently assessing the potential role of candesartan on the composite outcome of cardiovascular death and/or hospitalization for HF in its unique population.³⁵ In those with an EF ≤40%, CHARM ADDED addressed whether there was an additional benefit of adding the ARB to those on an ACEi (“ADDED), while CHARM ALTERNATIVE tested whether the ARB would improve prognosis in patients that had previously been considered intolerant of an ACEi.^{36, 37} The third component of this concomitant trial program selected those with the same clinical diagnosis of HF but with an EF >40%.³⁸ This level was specifically chosen because it identified a large group of patients with HF without prior outcome trial directed evidence for treatment. As such, placebo was the appropriate comparator to the ARB.

From a historic perspective, this 1997 designed trial reflected the HF clinical climate at the end of the last century.³⁹ Although by this time there was a general appreciation that a substantial proportion of patients with symptomatic HF did not necessarily have a low EF, except for a subgroup in the DIG trial, this higher EF group was not represented in prior outcome RCTs. Accordingly, international guidelines lacked specific evidence to base therapeutic recommendations.^{40, 41} Moreover, neither event rates nor extensive experiences in identifying sufficient number of these subjects were available from prior international trials. These key issues which had to be addressed in the CHARM protocol, engendered many lively pre-trial discussions amongst the study leadership.

The presumed structural and functional cardiac properties of those with diastolic HF were considered for eligibility criteria. However, aside from the impractical direct measure of ventricular filling pressure, the non-invasive indices of diastolic function did not offer adequate discrimination for identifying patients for this large outcome trial.⁴² The current trial’s objective was to determine whether the ARB would reduce rates of cardiovascular and hospitalization for HF in a population where there were no current evidenced-based recommendations. Therefore, the pragmatic approach of using the EF >40% for entry was adopted since this was measured in all patients, and was in fact the discriminating criteria between the well characterized patients in prior trials.³⁸ Adopting the specific cut point of EF >40% was based on addressing this therapeutic void rather than a mechanistic distinction. This was particularly practical in this specific investigative program since it addressed the unmet need and allowed the concomitant randomization of the broadest population of patients with HF regardless of EF. Based on how these patients were being identified for the trial, the investigators designated this component trial of those with EF >40% as CHARM-Preserved. The term was to distinguish from the well-studied lower EF groups and in no way meant to imply normal left ventricular structure or function.

Although this was clearly not the first use of the term HF with preserved ejection fraction,⁴³ the publication of a major outcome trial using that terminology was associated with an upsurge in its usage (Figure 1). As greater attention was focused on this segment of patients with HF, the acronym HFpEF was rapidly adopted.⁴⁴ Although there are many demographic and cardiac structural and functional differences between those considered to have HF based on predominantly diastolic or systolic disorders, EF continues to be used as the distinguishing feature.

There has been a recent expansion of the EF derived categories of patient with the clinical diagnosis of HF. HF with midrange EF 40 to 49% (HFmrEF) describes a group with sub normal EF who were for the most part not included in the in the prior outcome trials.^{4, 45} Others have proposed recovered EF (HFrecEF) to designate a group that now has EF ≤ 50% but was known to have a lower value previously.⁴⁶ The EF names continue with the term improved (HFiEF) for those that were previously <40% and have any apparently meaningful increase in EF even if still below the arbitrary 50% level for HFrecEF.⁴⁷ Although clinically ascertained, EF has an expected measurement error, these additions to nomenclature highlight that the clinical presentation of HF occurs across a spectrum of current and prior EF levels. Although diastolic HF remains an important term, the pragmatic use of EF is based on linkage to clinical outcome trial data. The importance of any of these multiple EF designations of patients with the clinical diagnosis of HF will await coupling of a benefit or even risk of a specific therapeutic intervention.

Assessment of HFpEF in Epidemiologic Studies

As a clinical syndrome, HF ascertainment in epidemiologic studies is challenging (Table 1). Ascertainment has typically relied on physician diagnostic codes with or without additional medical record abstraction and adjudication (Table 2). Diagnostic codes are readily and consistently available, but may be influenced by non-medical factors such a re-imbursement incentives. Additional adjudication helps mitigate such effects, improves consistency of the HF designation, and can more accurately assess whether HF is the cause of hospitalization or a co-existing condition. Several criteria exist for HF adjudication, including – but not limited to – the Framingham, Boston, Gothenburg, and ESC criteria.^{4, 48–51} All rely on some combination of symptoms, physical exam findings, radiographic findings, medical history, and response to therapy. Importantly, none include LVEF as a criteria. The most commonly employed criteria in epidemiologic studies of HFpEF has been the Framingham criteria, or a variant thereof.⁴⁸ While the Framingham criteria demonstrates good specificity for HF diagnosis, its sensitivity appears particularly limited in elderly persons and HF patients without acute decompensation⁵² – the pattern of HF patients frequently encountered in epidemiologic cohort studies (Table 1).

The designation of HF subgroups by LVEF is epidemiologic studies is relatively recent (Table 2). HFpEF is typically defined by first making the clinical assessment of HF as above, which is then coupled with an LVEF assessed as ≥45 or 50% (Table 2). I In some studies, LVEF information is based on abstracted cardiac imaging results (e.g. echocardiography, nuclear imaging, angiography, MRI) from the time of HF diagnosis. The close temporal linkage between HF event and LVEF assessment is an important strength of this approach, while missing LVEF data is a limitation as LVEF is often not assessed at the time of HF diagnosis (e.g. 11 to 37% of incident cases).^53–56 Another approach commonly employed in studies of HFpEF prevalence in longitudinal cohorts has been to use an LVEF assessed uniformly as part of the study protocol but performed a variable amount of time after that actual HF diagnosis.^{42, 57, 58} While the ability to classify all HF cases with respect to LVEF is a strength of this approach, the non-uniform lag between HF diagnosis and LVEF assessment is a limitation. Although there are concerns about longitudinal changes, LVEF is generally stable over the short to medium term unless there was an interval myocardial infarction.^{21, 59}

Population-Based Studies of HFpEF Prevalence

For approximately a decade following the initial descriptions of HFpEF in the early to mid-1980s, published studies on HFpEF were either case series or comparative clinical studies among prevalent HF patients.⁶⁰ These studies resulted in significant variability in estimates of HFpEF prevalence and prognostic relevance. In 1997, a report from the Helsinki Ageing Study of a random sample of 501 persons born in 1904, 1909, and 1914 demonstrated preserved LV systolic function – based on a fractional shorting ≥0.25 – in 51% of prevalent cases.⁶¹ Over the next 5 years, multiple diverse community based cohorts reported on HFpEF prevalence. These included the Framingham Heart Study (largely white general population),⁵⁷ the Strong Heart Study (Native Americans),⁴² and the Cardiovascular Health Study (persons in late life),^{58, 62} and each confirmed that HFpEF accounts for a large proportion (51 to 63%) of HF cases in the community. These findings are also concordant with data from consecutive hospitalized patients,⁶³ and from several large HF registries.^64–66 Importantly, HFpEF ascertainment in all of these studies was based solely on the combination of HF – assessed based largely on clinical signs and symptoms – and an LVEF threshold, without additional data regarding diastolic function or atrial or ventricular structure.

Population-Based Studies of HFpEF Incidence

Studies of incident HF in community-based cohorts suggest that HFpEF accounts for approximately 40–50% of incident HF overall. In a study of all incident HF cases in Olmsted County, MN in 1991, 43% of cases with LVEF assessed had an LVEF ≥50%.⁵³ Similarly, longitudinal follow-up of HF-free persons in the PREVEND cohort, the Framingham Heart Study, and the Cardiovascular Health Study demonstrated HFpEF accounts for 37–53% of incident HF.⁶⁷ Higher proportions of incident HFpEF were observed in studies with older participants. The proportion of incident HF due to HFpEF also appears to be increasing. An analysis of HF cases evaluated at the Mayo Clinic between 1987–2001 demonstrated an increasing proportion of HFpEF patients over time, particularly among HF patients evaluated from the community.⁵⁴ These findings are consistent with a broader analysis of all incident inpatient and outpatient HF cases in Olmsted County, MN occurring over the 10 year period of 2000–2010.⁶⁸ Of 2,762 incident HF cases, HFpEF accounted for 53% of incident HF cases overall. While a modest increase in the proportion of HFpEF cases was observed over time, overall incidence of HF decreased during the study period by 37%. However, the magnitude of this decline was less for HFpEF (28% decrease) compared to HFrEF (45% decrease), and women in particular demonstrated a much greater reduction in incident HFrEF compared to incident HFpEF. In contrast to these data on all HF, a recent analysis of hospitalizations for acute decompensated HF in 4 U.S. communities between 2005–2014 through the ARIC Community Surveillance study demonstrated rising rates of hospitalization for acute decompensated HF, driven primarily by increasing rates of hospitalization for HFpEF.⁶⁹

Demographics and Clinical Co-Morbidities in HFpEF

The clinical syndrome of HFpEF develops from a complex interaction of a number of risk factors that cause organ dysfunction and ultimately clinical symptoms (Figure 2). Since its initial description, older age and higher prevalence of women have been recognized as more frequent in HFpEF compared to HFrEF, and have been confirmed repeatedly in epidemiologic studies of prevalent and incident HFpEF. Recent data from Olmsted County suggest that persons developing HFpEF to be ~6 years older on average than those developing HFrEF,⁶⁸ and HFpEF accounts for a higher proportion of incident HF cases in older persons. Similarly, the prevalence of female gender is approximately 30–40% higher in prevalent HFpEF compared to HFrEF in epidemiologic cohort studies.^{42, 57, 62} Not only does HFpEF account for a higher proportion of incident HF in women compared to men, but the incidence of HFrEF appears particularly low in women and has fallen more rapidly over time.⁶⁸ There is less robust epidemiologic data regarding racial and ethnic differences in HFpEF epidemiology. Compared to whites, African Americans (AAs) have a 50% higher prevalence⁷⁰ and ~80% higher incidence of HF,^71–73 and worse outcomes once HF develops regardless of LVEF.^{74, 75} Over the past 20 years, the rates of HF hospitalization in AAs have remained more than twice that of whites and declines in age-standardized HF rates have been especially modest in AA women.⁷⁶ HFpEF accounts for up to 70% of prevalent HF in AAs.⁷⁷ However, important sex-based differences exist in HFpEF epidemiology among AAs. The incidence of HFrEF is particularly high among AA men while black women demonstrate higher rates of incident HFpEF compared to other race- and sex-based groups.⁶⁹

The high prevalence of co-morbid cardiovascular disease and cardiovascular risk factors in HFpEF is well recognized (Figure 2). The most prevalent cardiovascular disease in HFpEF is hypertension, which is present in the large majority of HFpEF cases across epidemiologic and registry studies. That HFpEF has often been considered an expression of advanced hypertensive heart disease speaks to its close association with hypertension. Several additional co-morbidities, in addition to being common in HFpEF, may also play a role in syndrome pathogenesis. Although most strongly associated with HFrEF,⁷⁸ co-morbid coronary artery disease is also common in HFpEF, with prevalence in epidemiologic and registry studies of 35–60%.⁷⁹ When systematically evaluated, the prevalence of epicardial CAD in HFpEF is even higher. A recent study prospectively performing coronary angiography in all HFpEF patients admitted with acute decompensation at a single center identified epicardial coronary disease in 64%. ⁸⁰ When present, CAD is associated with greater risk of progressive LVEF decline and mortality in HFpEF.⁸¹

Common HFpEF co-morbidities that may also influence the pathophysiology of the syndrome include atrial fibrillation (discussed in detail below), diabetes (prevalence of 20–40%), chronic kidney disease (varying prevalence depending on definition; ~20–30%), and obesity (prevalence of up to 50% ).⁷⁹ The prevalence of obesity may be even higher in HFpEF, because excess adiposity both lowers natriuretic peptide levels and makes physical examination findings of congestion less appreciable.⁸² For example, in a recent clinical trial that did not restrict enrollment to patients with elevated natriuretic peptide levels the prevalence of obesity was 75% and the prevalence of overweight or obesity was 95%.⁸³ The presence of these comorbid conditions is believed to lead to the structural and functional changes that culminate in the clinical syndrome of HFpEF (Figure 2), though the cellular mechanisms and mediators remain unclear.

While these co-morbidities are common among patients with HFpEF and identify those at higher risk for adverse outcomes, it remains unclear if the burden of co-morbidities is quantitatively higher in HFpEF compared to HFrEF. Multiple studies have suggested that certain co-morbid conditions tend to be more common in HFrEF (e.g. CAD) or HFpEF (e.g. hypertension, obesity) compared to the other, and at least some studies have demonstrated that the number of comorbid conditions is higher – on average – in HFpEF compared to HFrEF.⁸⁴ However, none of these co-morbidities discriminate patients with HFpEF from HFrEF. Hypertension and diabetes similarly do not appear to differentially predict incident HFpEF versus HFrEF,⁶⁷ although recent studies suggest that obesity may be associated with a greater risk of incident HFpEF.⁸⁵ Furthermore, the prevalence of obesity, diabetes, and hypertension also increase in late life more generally, and the extent to which they discriminate older persons with HFpEF from those without HF is unclear.⁸⁶ In contrast, the reported prevalence of cardiovascular co-morbidities, such as CAD and atrial fibrillation, are much more similar in HFpEF and HFrEF compared to HF-free persons.

Clinical Outcomes in HFpEF

Regardless of LVEF, HF is associated with greater mortality compared to persons free of HF. Epidemiologic studies have consistently demonstrated a heightened risk of all-cause and CV mortality in HFpEF compared to HF-free individuals.^{57, 61, 62} While several studies also observed similar mortality rates between individuals with HFpEF and HFrEF,^{54, 55, 87} this has not been a consistent finding and several other studies have reported lower rates in HFpEF compared to HFrEF.^{57, 61, 62} Epidemiologic studies suggest that incidence of CV mortality is lower, and non-CV mortality higher, in HFpEF compared to HFrEF,⁶⁸ findings that are concordant with observations from clinical trial samples.⁸⁸ Importantly, rates of hospitalization, hospitalization duration, and impairments in patient-reported outcomes such as quality of life appear similar between HFpEF and HFrEF.^{89, 90} Additionally, while clinical trials impose necessary exclusions to enrollment that may limit generalizability, HFpEF patients enrolled in clinical trials clearly demonstrate heightened risk for HF and all-cause mortality compared to patients in trials of hypertension and/or diabetes but without a diagnosis of HF.⁶

Cardiac Structure in HFpEF

Echocardiography is the most commonly employed imaging modality in epidemiologic, registry, and clinical trial studies of HFpEF, and many have employed core laboratories to mitigate acquisition and measurement variability as recommended by professional societies.⁹¹ Still, inter-laboratory measurement differences remain a potential source of between-study differences in reports of cardiac structure and function. While studies in select patients with HFpEF have led to the view of stiff, hypertrophied walls with increased fibrosis and a relatively small chamber,^{14, 92} investigations in broader HFpEF samples have established a much more heterogeneous cardiac phenotype (Figure 2).⁹³

Epidemiology

Across epidemiologic studies, perhaps the least bias study type, LV structure in HFpEF is characterized by normal mean LV cavity size with variable degrees of LV wall thickening. The Olmsted County cohort, one of the largest and most comprehensive epidemiologic evaluations of cardiac structure and function in HFpEF, demonstrated a prevalence of LVH of 42% among 244 HFpEF cases.⁹⁴ Concentric hypertrophy or remodeling was present in 53%, while eccentric hypertrophy was also noted in 16%. LV geometry was normal in nearly one-third of HFpEF cases (31%). Similar average LV mass index was observed among participants in the Cardiovascular Health Study with HFpEF who were more uniformly older, but ventricular size was somewhat larger with a lesser degree of concentric remodeling.⁹⁵ Perhaps the most comprehensive assessment to date of cardiac structure and function in a referral registry setting is the Northwestern HFpEF Registry, which demonstrated greater concentric remodeling and higher LVH prevalence than most epidemiologic studies.⁹⁶ The relatively higher proportion of African Americans in this study may contribute to these differences. Similar to the Olmsted County cohort, 12% of patients had an eccentric pattern of ventricular hypertrophy, highlighting the heterogeneity of ventricular remodeling characterizing the HFpEF syndrome.

Clinical Trials

Imaging sub-studies of large clinical trials provide the opportunity for detailed characterization of large numbers of HFpEF cases, although trial inclusion and exclusion criteria likely impact the observed phenotypes. For example, normal LV geometry was observed in only 14% of 875 HFpEF patients in the echocardiographic sub-study of the TOPCAT trial,⁹⁷ but was seen in nearly half (46%) of 745 HFpEF patients in the I-PRESERVE echocardiographic sub-study.⁹⁸ Despite this variability, LVH is consistently predictive of worse prognosis in HFpEF, including higher risk of death or HF hospitalization independent of clinical predictors and measures of diastolic function.^{98, 99}

HFpEF in Non-Western Regions of the World

Information is more limited in other regions of the world. Data from the multiregional cross-sectional Identification of patients with heart failure and PREserved systolic Function: an Epidemiological Regional (I-PREFER) study demonstrated that HFpEF also accounts for a significant proportion of HF in non-Western countries.¹⁰⁰ Notably, important regional variation was observed, with HFpEF accounting for a higher prevalence of HF in Latin American (69%) and North Africa (75%) compared to the Middle East (41%). HFpEF similarly accounted for 51% of HF in the Japanese Chronic Heart failure Analysis and Registry in the Tohoku district (CHART-1) study.¹⁰¹ Notably, while CHART-1 recruited patients in 2000, in the follow-up CHART-2 study (2006–2010) HFpEF accounted for 69% of HF,¹⁰² highlighting the increasing burden of HFpEF in that region. A recent report from the Asian Sudden Cardiac Death in Heart Failure (ASIAN-HF) registry highlighted the diversity of the HFpEF syndrome between, and within, regions.¹⁰³ For example, HFpEF patients in that study were nearly a decade younger with lower BMI than compared to reports from Western samples. Furthermore, appreciable differences existed in co-morbidity prevalence, cardiac remodeling, and outcomes between regions within Asia.

LV Diastolic Dysfunction

The most conspicuous and unifying hemodynamic finding of HFpEF is an elevation in LV filling pressures (LVFP). In advanced stages of HFpEF, LVFP are elevated at rest, but in patients with earlier stages of disease, LVFP become markedly elevated only during the stress of exercise.^{108, 111} High LVFP during exertion in HFpEF are directly correlated with heightened inspiratory drive, symptoms of dyspnea, alterations in gas exchange and pulmonary ventilation, and reductions in aerobic capacity (Figure 3).^{109, 110} Elevation in LVFP alter the Starling forces across the pulmonary capillaries, favoring filtration of water out of the vascular space and into the interstitium.^{118, 119} Over time, high LVFP may promote capillary stress failure¹²⁰ and vascular remodeling, particularly in the pulmonary veins.¹²¹ High LVFP, even when normal at rest and restricted to exercise, identifies patients at increased risk for both HF hospitalization and death.^{122, 123}

Elevation in LVFP in HFpEF is caused predominantly by diastolic dysfunction. As compared to controls, patients with HFpEF display an LV diastolic pressure-volume relationship that is shifted up and to the left, indicating an increase in passive chamber stiffness.¹⁰⁶ During dynamic exercise, LV diastolic stiffness acutely increases to an even greater extent in HFpEF.¹⁰⁷ Patients with HFpEF display prolonged relaxation,¹⁰⁶ and an inability to hasten relaxation as heart rate increases, as during exercise.¹⁰⁷ This, together with an impairment in systolic reserve (discussed below), impairs LV diastolic “suction”, so that left atrial hypertension becomes necessary to drive LV filling, in striking contrast to the normal heart where blood is “pulled” into the LV due to vigorous suction.¹²⁴ While prolonged relaxation does not usually affect LVFP when heart rate is normal at rest, this can significantly impact LVFP during exertion as cycle length shortens.¹⁰⁷ Indeed, the inability to enhance early LV diastolic velocities during exertion is directly correlated with the magnitude of LVFP increase in patients with HFpEF.¹⁰⁸

While these mechanisms underlying elevated LVFP in HFpEF have been described in select patients undergoing invasive hemodynamic assessment, the extent to which they generalize to the broader group of persons with the HFpEF syndrome identified in epidemiologic studies and therapeutic clinical trials is less clear. Diastolic dysfunction is most commonly assessed noninvasively in clinical practice, in clinical trials, and in population-based studies, and multiple potential indices of diastolic performance exist.¹²⁵

One important challenge in determining the prevalence of diastolic dysfunction in HFpEF is that non-invasive measures of diastolic function change with age. However, the extent to which these prominent age-related changes are pathologic versus benign is unclear. Indeed, using existing guideline-based cutpoints, at least 1 abnormal diastolic measure is noted in >90% of persons >65 years of age.¹²⁶ By using reference limits for abnormal diastolic measures derived from low risk elderly persons (>65 years old) without CV disease or risk factors, fewer elderly persons were classified as having diastolic abnormalities compared to guideline cutpoints, while risk prediction for incident HF or death was improved. Notably, even using these age-appropriate cutpoints, diastolic abnormalities were identified in 54% of elderly persons. These finding argue that age-appropriate reference limits need to be used in interpreting diastolic indices.

Detailed data on diastolic function in epidemiologic studies have been limited as assessment in most existing studies relied primarily on transmitral diastolic flow patterns, and specifically the ratio of early (E wave) to atrial (A wave) inflow velocity (E/A ratio).^{42, 58} The most comprehensive assessment to date has been performed by the Olmsted County investigators. Among 556 HF cases in Olmsted County, LVEF was ≥50% in 55% among whom Doppler evidence of diastolic dysfunction was present in 78% (mild 7%, moderate 63% severe 8%). Diastolic assessment was normal in 10% and indeterminate in the remainder.⁸⁷

More comprehensive data on LV diastolic measures is available from HFpEF clinical trials compared to epidemiology studies. Prevalence estimates in these studies for abnormal diastolic measures varies appreciably, possibly related to trial inclusion criteria, and are unlikely generalizable to persons in the community with HFpEF.⁹⁷ However, both community-based and clinical trial studies have consistently demonstrated an association between diastolic dysfunction severity, and worse individual diastolic measures, and adverse outcomes in HFpEF including mortality and HF hospitalization.⁹⁹

Extrinsic Restraint and LV Filling Pressures

Elevation in LVFP may also be driven in part by increases in extrinsic restraint on the heart, mediated by the right heart and pericardium.^{127, 128} This effect is amplified when venous return to the heart increases, as during exercise.¹⁰⁷ Patients with the obese phenotype of HFpEF display greater cardiac hypertrophy and plasma volume expansion, which increase total heart volume from within, and an increase in epicardial fat thickness, which increases external restraint from without.⁸² The net effect is an increase in diastolic ventricular interaction in obese HFpEF patients, whereby a greater proportion of the increase in intracavitary pressures is mediated by external restraint.⁸² In addition to mechanical effects,⁸² the increase in epicardial fat in patients with HFpEF may promote cardiac injury, inflammation, fibrosis, and coronary microvascular dysfunction.^{129, 130}

LV Systolic Dysfunction

The LVEF is, by definition, preserved in HFpEF, but LV systolic function is not necessarily normal. Impairments in myocardial and chamber-level function are present at rest, and these become much more dramatic during exercise.^{108, 131–133} This limits the ability of the heart to augment stroke volume, which substantially impairs the cardiac output response to exercise.^{105, 108–112, 116, 132, 134} In addition, LV systolic reserve deficits importantly compromise diastolic reserve, because the ability to enhance contractility plays a key role in determining the restoring forces that enhance early diastolic annular motion or recoil.¹⁰⁸

Impairments in LV systolic performance at rest can also be detected among broader samples of HFpEF patients based on transthoracic echocardiography. Recently, significant attention has focused on strain imaging, which quantitatively assesses myocardial deformation in multiple planes (longitudinal, circumferential), and appears to be a less load-dependent index of systolic function than LVEF.¹³⁵ Worse longitudinal strain has been associated with several common HF risk factors, including hypertension, diabetes, obesity, and CAD, and is predictive of incident HF in community-based studies.¹³⁶ Importantly, worse longitudinal strain despite preserved LVEF is robustly prognostic of worse outcomes in HFpEF. In the echocardiographic substudy of the TOPCAT trial, longitudinal strain was abnormal in over half of patients, and was associated with a heightened risk of CV death or HF hospitalization beyond clinical and other echocardiographic predictors.¹³⁷

Left Atrial Dysfunction

As LV diastolic filling pressures are elevated intermittently over time, there is secondary remodeling and dysfunction that develops in the left atrium (LA) in HFpEF. It has been hypothesized that preservation of LA function may be an important adaptation in HFpEF that “protects” the lungs and right heart,¹³⁸ because development of LA dysfunction is associated with worse exercise capacity, more profound pulmonary vascular disease, and increased risk of death.^138–140 Atrial fibrillation is very common in HFpEF and is further associated with LA remodeling and dysfunction.¹⁴¹ As discussed below HFpEF patients with LA dysfunction and atrial fibrillation are even more likely to also develop pulmonary hypertension and right HF (Figure 2).^{142, 143}

Pulmonary Hypertension

The prevalence of pulmonary hypertension (PH) in patients with HFpEF varies widely based on the sample being studied¹⁴⁴ and method of assessment, but its presence uniformly identifies patients at increased risk of death.¹⁴⁵ In addition to PH from “passive” LA hypertension, a substantial number of patients go on to develop pulmonary vascular disease, with elevations in pulmonary vascular resistance (PVR).^{120, 146, 147} As compared to patients with pure passive PH (i.e. caused by isolated LA hypertension), this cohort displays further increased risk of death.¹⁴⁶ Elevated PVR in HFpEF is related in part to vasoconstriction, since acute vasodilators can reduce PVR,¹⁴⁸ but recent data also indicate that there is structural remodeling in the lung vasculature that importantly contributes, though its mechanisms remain unresolved.¹²¹

Patients with HFpEF and pulmonary vascular disease display a unique pathophysiology whereby the inability to enhance RV ejection leads to augmented right heart distention and LV underfilling, despite the fact that LVFP are elevated.¹⁴⁷ Even among HFpEF patients that display normal PVR at rest, there is inadequate pulmonary vasodilation during exercise in HFpEF.^{108, 149} This is associated with adverse clinical outcomes¹⁴⁹ and impairment in right ventricular (RV) functional reserve.¹⁰⁸ Patients with HFpEF and greater lung congestion display more pulmonary vascular disease,¹¹⁸ possibly due to effects of vessel wall edema itself¹⁵⁰ or chronic structural remodeling related to increased hydrostatic pressures.¹²¹

Right Ventricular Dysfunction

Longstanding PH in HFpEF eventually causes RV dysfunction, which develops in roughly one-third of patients.^{142, 143} However, this is not mediated purely by afterload mismatch, as RV myocardial dysfunction is demonstrable even after accounting for the severity of PH and non-hemodynamic risk factors have been identified including ischemic heart disease, obesity and atrial fibrillation.^{142, 143, 151} The development of RV dysfunction identifies patients with HFpEF with markedly increased risk of death, and might be preventable through treatment of PH and risk factors.^{142, 143,151}

Coronary Microvascular Dysfunction

Given the combined systolic and diastolic myocardial reserve limitations in HFpEF, which are demonstrable in both ventricles, it is logical to consider abnormalities in cardiomyocyte energy availability or utilization as potential contributors. Indeed, one human study did reveal deficits in myocardial oxygen utilization in HFpEF which was correlated with hemodynamic severity.¹⁵² Numerous recent studies suggest that coronary microvascular dysfunction may plan an important role in the pathophysiology of HFpEF. Myocardial ischemia and injury are common in HFpEF, and are correlated with the ventricular functional abnormalities noted in human physiologic studies.¹⁵³ Ischemia may be caused by supply-demand mismatch due to high LVFP,¹⁵³ macrovascular (epicardial) disease¹⁵⁴ as well as coronary microvascular dysfunction.^{153, 155, 156}

Redfield and colleagues first demonstrated in an autopsy study that microvascular density is reduced in patients with HFpEF, and the severity of microvascular rarefaction is correlated with the magnitude of myocardial fibrosis.¹⁵⁷ In addition to anatomic causes, there is evidence for impaired coronary flow reserve in HFpEF, with is associated with markers of more advanced disease.¹⁵⁶ These data have led to the emerging hypothesis that much of HFpEF is related to coronary microvascular dysfunction.^{158, 159}

Large Vessel and Microvascular Dysfunction

In addition to cardiac and pulmonary vascular abnormalities, there is abnormal systemic vascular function in HFpEF.^{111, 116} Aortic and conduit vessel stiffness is increased in HFpEF, leading to excessive blood pressure variability and greater arterial afterload mismatch, particularly during exercise.^{111, 116, 160} This appears to be mediated in part by endothelial dysfunction, since abnormal flow-mediated dilation is present in HFpEF¹¹¹ and because acute provision of NO improves vascular function and central hemodynamics in tandem.^{111, 116}

Microvascular function is impaired in the periphery, evidenced by impaired flow mediated dilation during reactive hyperemia in HFpEF.¹⁶¹ The degree of impairment in systemic microvascular function is associated with abnormal regional vasodilation during exercise, pulmonary vasoconstriction, greater dyspnea and fatigue severity, reduced aerobic capacity, and increased risk for HF hospitalization.^161–164

Peripheral Abnormalities

Recent studies have also identified important roles peripheral to the heart and large arteries. Skeletal muscle composition is altered in HFpEF, with increased fatty infiltration and frank sarcopenia, even when total body weight is increased (sarcopenic obesity).^{111–113, 115} The ability to enhance oxygen extraction in skeletal muscle is impaired in HFpEF due to an inability to augment diffusional conductance for gas transfer within skeletal muscle, and an inability to improve oxygen utilization, likely due at least in part to mitochondrial dysfunction.^{114, 115, 117}

Cardiac Amyloidosis

Patients with infiltrative cardiomyopathies such as cardiac amyloid have pathophysiologically distinct causes for HF are generally not considered to be part of the “garden variety” HFpEF syndrome. While cardiac amyloidosis is generally considered rare, recent data suggest that the prevalence of wild type ATTR cardiac amyloidosis, characterized by deposition of misfolded transthyretin protein, may be present in 13–19% of persons with prevalent HFpEF. This has become particularly important to recognize as novel therapeutic agents such as Inotersen, Tafamidis, and Patisiran have demonstrated efficacy in treating TTR amyloidosis. Furthermore, the high sensitivity and specificity of ^99mtechnetium pyrophosphate SPECT imaging to detect cardiac TTR deposition now provides a non-invasive approach to screen appropriate patients.

Systemic Inflammation

One major paradigm relates cardiac and extracardiac reserve limitations to abnormalities in NO-cGMP signaling, which are believed to be ultimately related to low grade systemic inflammation.¹⁶⁵ Elevation in inflammatory biomarkers predict development of incident HFpEF,¹⁶⁶ and have been shown to be higher in patients with HFpEF as compared to HFrEF.^{167, 168}

Comorbidities such as obesity, hypertension and metabolic syndrome are extremely common in HFpEF are accordingly believed to cause low-grade systemic inflammation, which impairs endothelial formation of NO in the heart and elsewhere (Figure 2).¹⁶⁵ This paradigm is supported by data from human HFpEF LV biopsy specimens demonstrating reduced PKG activity and cGMP concentrations that were correlated with increased passive myocyte stiffness.¹⁶⁹ In myocardium of HFpEF patients there is microvascular endothelial activation, presumably related to underlying systemic inflammation, that is demonstrated by upregulation of E-selectin and intercellular adhesion molecule-1 expression levels and uncoupling of endothelial NO synthase associated with reduced myocardial nitrite/nitrate concentration, cGMP content, and PKG activity.¹⁷⁰

Other Cellular Mechanisms

Intracellular calcium also plays a key role in relaxation. A recent study of isolated myocardium from patients with HFpEF has revealed that contraction and relaxation are prolonged compared to controls, in tandem with elevated sarcomere calcium levels, even as intracellular sodium and calcium handling protein expression appear unaffected.¹⁷¹ In myocardial strip preparations from patients with LV hypertrophy undergoing cardiac surgery, relaxation reserve was inversely related to LV mass and associated with increased calcium load at high heart rates, with increased resting tone due to diastolic cross-bridge cycling.¹⁷²

Increases in diastolic LV stiffness in HFpEF have also been related to alterations in titin content and phosphorylation as well as fibrillar collagen.¹⁷³ Titin in particular appears to be a major player since it is a dominant cellular determinant of stiffness that can be dynamically modulated, either by altering phosphorylation status, or protein expression.^{174, 175} Other possible unifying cellular mechanisms include exaggerated cardiomyocyte senescence, lipotoxicity, or deranged autophagy.^176–178

Barriers to Improved Mechanistic Understanding

There are no well-established animal models of HFpEF, though a number recapitulate certain features of human HFpEF. Models based upon combinations of accelerated senescence, metabolic stress (western diet), obesity, pressure overload, neurohormonal stimulation, and even radiation exposure have been developed in recent years.^{170, 179–184} These models have demonstrated increases in mitochondrial ROS¹⁸⁰ and energy deficiency with decreased ATP production.¹⁷⁹ While each of these model systems generally produce hemodynamic and even systemic alterations that are relatively typical of “garden variety” HFpEF, the relevance of new discoveries from these models to the human condition will continue to be questioned until more data from direct human tissue can be obtained and analyzed.

Hypertension Treatment

The Framingham Heart Study established hypertension as one of the earliest ‘factors of risk’ for coronary heart disease²⁰², and subsequently also for incident HF ²⁰³. Several randomized clinical trials of antihypertensive agents have subsequently established the reduction in incidence of HF resulting from effective blood pressure control.²⁰⁴ The Systolic Hypertension in the Elderly (SHEP) trial specifically expanded evidence for this benefit to persons >60 years of age,²⁰⁵ while the Hypertension in the Very Elderly Trial (HYVET) expanded it to persons ≥80 years of age who are highest risk of HF development.²⁰⁶ More recently, the Systolic Blood Pressure Intervention Trial (SPRINT) demonstrated the further reduction in incident HF associated with targeting a systolic blood pressure of <120 mmHg compared to <140 mmHg among 9,361 non-diabetic persons with elevated cardiovascular risk.²⁰⁷ In that study, intensive compared to conventional BP treatment targets reduced incident HF by 37%, with separation of HF event curves between groups evident after 6 months. Importantly, among the 2,636 participants who were ≥ 75 years old at enrollment ²⁰⁸, intensive compared to conventional treatment targets was also associated with significant reduction in risk of incident HF (HR 0.62; 95% CI 0.40–0.95). The Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT) results also demonstrated the importance of the antihypertensive class employed. The doxazosin arm of that trial was terminated prematurely due to a 2-fold increase in risk of incident HF compared to chlorthalidone ²⁰⁹. Furthermore, risk of incident HF was lower among those randomized to chlorthalidone compared to lisinopril or amlodipine.²¹⁰ Among randomized trials addressing cardiovascular outcomes with BP lowering, a decrease in HF risk is one of the most prominent effects.

SGLT2 Inhibitors

In contrast to SPRINT, a more aggressive blood pressure treatment target in ACCORD was not associated with lower risk of incident HF among diabetic patients.²¹¹ In addition, limited data exist linking degree of glycemic control to HF risk, with several oral hypoglycemic agents associated with heightened risk of HF. However, SGLT2 inhibitors have demonstrated convincing reductions in incidence of HF among diabetic patients with varying degrees of baseline cardiovascular risk.^212–214

Lifestyle Modification

The beneficial impacts of lifestyle modification, including weight reduction, dietary consumption, and physical activity and cardiorespiratory fitness on HF risk have recently been thoroughly reviewed.²¹⁵ While there is limited randomized controlled trial data, robust epidemiologic data support the conclusion that optimization of these lifestyle factors is associated with reductions in important HF risk factors (e.g. hypertension, diabetes, atherosclerotic disease), and suggest that reductions in incident HF independent of these risk factors.