Towards More Practical Phenotyping in Heart Failure with Preserved Ejection Fraction

This article refers to ‘Cardiovascular-kidney-metabolic overlap in heart failure with preserved ejection fraction: Cardiac structure and function, clinical outcomes, and response to sacubitril/valsartan in PARAGON-HF’ by M.C.H. Lassen et al., published in this issue on pages 1762–1774. With initial therapeutic failures, there was eventual recognition that traditionally defined heart failure (HF) with preserved ejection fraction (HFpEF) represented a heterogeneous syndrome. Contributing to this heterogeneity, HFpEF trials have variability defined a ‘preserved’ ejection fraction (EF), frequently including subsets with mild HF with reduced EF (HFrEF) with an EF between 40 to 50%. Even with stricter definitions of a preserved EF (>50%), several pathophysiological phenotypes could be identified related to obesity,1 atrial myopathy,2 pulmonary vascular disease,3 or ventricular-vascular stiffening,4 often with substantial phenotypic overlap.5 This hypothesis-driven discovery approach resulted in phenotype-specific trials most successfully in obesity-related HFpEF where semaglutide induced weight loss improved quality of life (QOL), functional status and natriuretic peptides with encouraging signals of benefit on HF hospitalization (HFH).6 Although sodium–glucose cotransporter 2 inhibitors (SGLT2i) demonstrate universal efficacy across all HFpEF phenotypes in the prevention of HFH, there appears to be divergent phenotypic effects in obesity-related HFpEF with greater weight loss7 and enhanced functional status benefits.8 In parallel to this hypothesis-driven investigative approach, there has been extensive exploration of machine learning approaches to agnostically identify unique HFpEF phenotypes. Although helpful to discover novel associations, these complex groupings lack direct clinical applicability since multivariable clustering frameworks are not easily applied to individual patients in clinical practice. The ultimate goal of any attempted phenotypic classification is to provide clinical guidance to alter therapy. In the era of shared decision-making, being able to describe simple HFpEF disease groups that predict future HFH risk and magnitude of anticipated benefit from existing therapies is a major unmet need. In this issue of the Journal, Lassen et al.9 provide an important advance in this regard through a thoughtful secondary analysis of the PARAGON-HF trial testing sacubitril/valsartan in HFpEF. The authors studied the clinical implications of having increased cardiovascular-kidney-metabolic (CKM) comorbidities in HFpEF. Each patient's burden of cardiovascular, kidney or metabolic comorbidity was determined based on cumulative presence of (i) atherosclerotic disease, (ii) chronic kidney disease, and (iii) type 2 diabetes. As a result, HFpEF patients could have either none, one, two or three independent CKM comorbidities on top of their myocardial dysfunction related to HFpEF. Among 4796 HFpEF patients, only about one out of every six HFpEF patients (16%) had isolated heart failure related to HFpEF without a known CKM comorbidity. In contrast, about five out of six HFpEF patients (84%) had at least one CKM comorbidity with 35%, 33% and 16% of the HFpEF cohort having one, two or all three CKM comorbidities, respectively. Greater CKM burden was associated with worse obesity, natriuretic peptide elevation, clinical congestion and QOL reflecting more advanced HFpEF with longer HF duration. Myocardial remodelling was worse with greater CKM burden with higher left heart filling pressures, left ventricular remodelling and biventricular dysfunction. Increasing CKM comorbidity burden in HFpEF was associated with higher risk of HFH, cardiovascular death, all-cause mortality, adverse kidney outcomes and longitudinal worsening QOL. When all three CKM comorbidities were present, the risk for HFH and death was nearly double those with only one CKM comorbidity. The authors then attempted to quantify the treatment implications of higher CKM predicted risk. If one rationally interprets PARAGON-HF without strict frequentist obsession over p < 0.05, sacubitril/valsartan was associated with a 15% relative risk reduction in total HFH, with no effect on cardiovascular death. Since there was no treatment interaction by CKM burden, the authors applied this 15% total HFH reduction to estimate number needed to treat for 1 year to prevent one HFH with increasing CKM burden. Since higher CKM burden was associated with higher absolute risk of HFH, a 15% reduction in relative risk translated into larger absolute benefit as CKM burden increased. The authors estimated that 125 patients would need treatment for 1 year with sacubitril/valsartan to prevent one HFH with isolated HFpEF without any CKM comorbidity. In contrast, the number of patients to treat for 1 year to prevent one HFH progressively decreased with increasing CKM comorbidity (72, 50 and 33 with one, two and three CKM comorbidities, respectively). The authors are to be congratulated on an elegant analysis that advances our understanding of the impact of CKM comorbidities in HFpEF, and their predictive ability for HFH. The breakdown of CKM comorbidity burden is nearly identical to those reported from the DELIVER HFpEF trial increasing confidence in the veracity of the results,10 while also adding novel insight into potential mechanisms with more advanced global myocardial remodelling observed with greater CKM burden. As with any study, there are a few caveats to consider. Although the primary analysis considered each individual CKM comorbidity to be of equal importance, there was modest heterogeneity of risk depending on individual comorbidity present. Atherosclerosis was associated with a lower EF likely related to the impact of coronary disease and prior myocardial infarction in causing a mildly reduced EF phenotype, a subgroup that likely responds more favourably to sacubitril/valsartan than those with more typical HFpEF with a truly preserved EF.11 Type 2 diabetes was also strongly associated with greater obesity, a complex collinear association that is impossible to retrospectively untangle statistically. Given that no diabetes-specific therapy has demonstrated benefit in HFpEF patients, with SGLT2i and semaglutide exerting benefit regardless of diabetes status with greater functional benefits in the presence of obesity,6-8 it is possible that clinical stratification could be improved by incorporating obesity in the CKM staging system instead of diabetes. Due to the PARAGON-HF natriuretic peptide inclusion criteria, the current risk estimates are also most directly applicable to less obese HFpEF cohorts with elevated natriuretic peptides, with exclusion of the large clinically relevant cohort of obese HFpEF with lower natriuretic peptides.1, 12 Excess adiposity is probably one of the greatest risk factors for kidney disease, atherosclerosis and type 2 diabetes in the modern era and how obesity fits into this CKM model deserves further evaluation in broader populations more representative of obesity-related HFpEF. The derived numbers needed to treat to prevent one HFH across CKM burden are also inherently dependent on the relative risk reduction estimate applied. Although the authors applied the overall PARAGON-HF trial treatment effect estimate for simplicity, there was a credible subgroup finding where the reduction in the primary endpoint (cardiovascular death and total HFH) was greater in the mild HFrEF patients enrolled,11 with no effect in those with more precisely defined HFpEF having a clearly preserved EF. Using the overall trial estimate dilutes this heterogeneity of effect, and numbers needed to treat to prevent one HFH with sacubitril/valsartan are likely even lower in HF patients with a mildly reduced EF with increasing CKM burden, with potentially no effect in those with true HFpEF regardless of CKM burden. These minor issues aside, this elegant analysis reaffirms the importance of comorbidities in the pathogenesis of HFpEF-related HFH.13 As opposed to HFrEF where isolated myocardial dysfunction is the primary problem driving neurohormonal sodium retention that can lead to HFH, the link between HFpEF and HFH is potentially more extracardiac, with traditional neurohormonal antagonists consistently ineffective at preventing HFH. In contrast, SGLT2i and perhaps semaglutide may prevent HFH in HFpEF through multiple mechanisms targeting extracardiac abnormalities including reduction in adiposity, preservation of renal reserve, improved vascular load, and other systemic effects that may favourably enhance the ability of HFpEF hearts to mitigate/prevent episodes of clinical volume overload. The burden of extracardiac comorbidities in HFpEF may therefore be important in the pathogenesis of subsequent HFH risk.13 In contrast, HFpEF patients with isolated myocardial dysfunction demonstrate lower risk of volume overload-related HFH with their disease dominated primarily by haemodynamically mediated exercise intolerance and poor QOL.14 This phenotypic distinction is clinically important since currently established HFpEF therapies are initiated with the goal of either improving QOL or preventing HFH, both of which may be independent goals. Semaglutide clearly improves QOL in obese HFpEF,6 while SGLT2i demonstrate heterogeneous effects on QOL with greater benefits in obese HFpEF,8 with sacubitril/valsartan demonstrating limited and inconsistent impact on QOL.15 As in clinical trials, a subset of patients may not experience a notable QOL improvement following initiation of therapy, but these same patients may still value therapy that can prevent future HFH even if symptoms are not improved. Simple classification schemes such as the CKM model proposed may help individual patients understand their expected absolute risk for HFH, and therefore anticipated HFH prevention from therapy to guide individual decisions about cost-effectiveness and value. In practice, breaking down each HFpEF patient's comorbidity burden by the proposed CKM framework may hold promise for shared decision-making to guide therapies that reduce HFH in HFpEF, while also optimally targeting each individual comorbidity contributing to the HFpEF systemic syndrome (Table 1). Given the universal presence of haemodynamic left HF in HFpEF, our early focus was excessively cardiac-centric. It is now becoming increasingly clear that systemic comorbidities are important mediators of the eventual risk of HFH in HFpEF beyond myocardial dysfunction, with therapies targeting extracardiac problems demonstrating the greatest benefit to mitigate HFH risk. ↓ CKD progression (DAPA-CKD, EMPA-KIDNEY) Potentially ↓ death (DAPA-CKD but not EMPA-KIDNEY) Better glycaemic control and weight loss Possible ↓ death in type 2 DM (EMPA-REG OUTCOME, borderline in CANVAS; but not in VERTIS-CV or DECLARE-TIMI 58) Atherosclerosis with BMI ≥27 kg/m2: ↓ HFH, MI and death (SELECT) Type 2 DM with CKD: ↓ death and CKD progression (FLOW) Better glycaemic control and weight loss Possible ↓ HFH (STEP-HFpEF, STEP-HFpEF DM) Possible ↓ death in type 2 DM (PIONEER-6, but not SUSTAIN-6) Persistent low-normal EF after MI in a HFpEF patient: ↓ HFH (PARAGON-HF) Revascularization if indicated Secondary prevention for atherosclerosis to prevent MI Type 2 DM with CKD: ↓ CKD progression and possible ↓ HFH with finerenone (FIDELIO-DKD, FIGARO-DKD) CKD with proteinuria: ↓ CKD progression with ARB/ACEi and possible ↓ HFH (RENAAL) Diet and exercise training Consideration of bariatric surgery Conflict of interest: Y.N.V.R. is supported by grants from NHLBI (K23HL164901), Sleep Number, Bayer, Merck, United Pharmaceuticals and the Early Wood Career Development Award from Mayo Clinic. V.S. has nothing to disclose.