GRAND DEBATE | HFmrEF is similar to HFrEF – or not?
M. Metra (Brescia, IT) and P. Ponikowski (Wroclaw, PL) chaired this second Grand Debate of this year’s conference, about controversies in the definition and management of heart failure (HF). The first part dealt with the new HF category based on ejection fraction (EF), HFmrEF, which first appeared in the ESC HF guidelines in 2016. By way of positioning of the topic, Piotr Ponikowski defended the concept of acknowledging a midrange EF category (EF 40-49%), because he chaired the guideline committee that came up with it. He indeed thinks that this category represents patients with a different type of disease. Currently, the guideline recommends to treat HFmrEF patients just like HFpEF. This means screening/management of comorbidities and diuretics to control symptoms/signs, as no evidence-based disease-modifying therapy exists. Real-world experiences of management of these patients is being collected in the ESC HF registry.
The concept was first described by Carolyn Lam and Scott Solomon in 2014, as the middle child in HF: typically the one who gets the least attention. One of the aims of putting it in the guidelines, was also to generate more research on it, to expand our understanding of characteristics, pathophysiology and treatment of these patients. That seems to have worked, as several studies evaluating patients with EF between 40 and 49% have been published since publication of the guidelines.
The debaters, Lars Lund (Stockholm, SE) and Scott Solomon (Boston, US), only a few weeks ago, together published a study on HFmrEF (CHARM, Lund et al., Eur J Heart Fail 2018), describing the characteristics, outcomes and effect of candesartan across the entire EF spectrum. Thus, they must agree to some extent. But, for the sake of this debate, they took positions on opposite sites, with Lund defending the PRO position on the statement that HFmrEF is similar to HFrEF.
A question central to this debate is whether EF is appropriate for categorization of HF and if so, is HFmrEF a distinct category? One could object to EF as a means for categorization by saying that EF does not reflect pathophysiology. But then what does? Etiology? Neurohormonal activity? Remodeling? Cluster-phenotyping? That is not so easy. Left ventricular strain may be a better marker, but it has not been validated, and there are no treatment or even trial criteria. Lund predicted that Solomon might point out that measurement of EF is variable by 7%, but that is also true for the cut-off point that is used to diagnose HFrEF. Thus, he thinks the answers to his questions is yes, also because EF is well-known in medicine, accepted for diagnosis and prognosis, it determines therapy, and it continues to be used in trials.
When looking at the baseline characteristics of CHARM, quite some similarities are seen between HFmrEF and HFrEF, for instance the average age, and the proportion of women, systolic BP, % edema, %ischemic cause of HF and % of patients with AF. All of these traits in the HFmrEF category resemble the numbers seen in HFrEF more than those in HFpEF.
A contemporary paradigm on the development of HFpEF and HFrEF also supports the proposition debated today; in HFpEF, but not HFmrEF, comorbidities and multiorgan remodeling may drive HF development, or concomitantly exert harmful effects. In HFrEF, disease generally originates from acute coronary thrombosis or myopathy and/or neuroendocrine activation, resulting in direct myocardial injury.
Lund predicted that Solomon may say: but the outcomes do differ! Well, yes, they are milder. And NYHA class is lower, as is NT-proBNP. Lund thinks that all of these observations were to be expected, as HFmrEF is the same, but milder. The EF is higher and the prognosis is better, but HFmrEF can be considered the same syndrome. Moreover, in CHARM, candesartan was effective in HFrEF and HFmrEF, but not in HFpEF, and the same has been shown for beta-blockers
Not all agree with introduction of this new category, and some critical viewpoints have been published. Still, about 20% of patients fit into this category, and EF does determine treatment response. In any case, for now it is all we have we can work with. So, Lund sees it as an opportunity, but prospective RCTs are needed.
Scott Solomon, defending the contra argument, acted as one of the critics. But not before saying that he dislikes debates, as they force you to take extreme positions. To defend his position, he wanted to look at the evidence, and what it means to the patient. He noted that HFmrEF is more like HFpEF with respect to the things that matter; characteristics are intermediate, but outcomes are more like HFpEF.
First of all; the category is an arbitrary one. Prior to 1985 only ‘heart failure’ existed, without EF being part of the diagnosis. At a certain point the category with reduced EF was based on clinical trial benefit in this range, although the EF cut-off values varied among trials. ‘Preserved’ was then coined for anything that wasn’t reduced; no evidence-based therapies have been found for this group of patients. And now there is another category, the borders of which can be considered arbitrary.
When he and Lam wrote the article on HFmrEF as a middle child, the aim was to spurt debate, not for the ESC HF Guideline committee to come up with new category. And a word of caution is needed; Only EF is not good enough to phenotype patients. So will categorizing a patient more precisely help us to better care for our patients with HF? Can we even be certain that we can accurately, reliably and consistently place patients into this ‘phenotypic’ construct?
An analysis of Swedish registry data (Koh et al., Eur J Heart Fail 2017) showed that with regard to some outcomes, HFmrEF resembles HFrEF, and with regard to others, it resembles HFpEF. Overall, the article stated that ‘HFmrEF was an intermediate phenotype between HFpEF and HFrEF’.
Similar observations were made in the ESC HF Long-term registry (Chioncel et al., Eur J Heart Fail 2017). But this study saw death and HFH event rates in HFrEF that were substantially higher than in HFmrEF and HFpEF. Moreover, in two Japanese cohorts, outcomes in HFmrEF patients could be superimposed on those seen in HFpEF (Sato et al., Eur J Prev Cardiol and Tsuji et al., Eur J Heart Fail, CHART-2 study). The CHART-2 study also showed that a shift between categories was common, in all directions. That raises the question whether that is a biological phenomenon, or a measurement error.
The CHARM trial, which he wrote together with Lund and others, was the first and only HF trial unbiased with respect to EF, and it showed that patients in the HFmrEF category were intermediate in virtually all aspects. Interestingly, NYHA class was more similar between HFmrEF and HFpEF. He also showed the outcomes published in CHARM, and pointed out that HFmrEF are unquestionably distinct from HFrEF outcomes. Thus, he concludes that it is hard to argue that HFmrEF is the same as HFrEF. The CHARM authors ‘recognize, along with others, that EF is not an optimal classifier in HF, that cut-offs are arbitrary and that other tools to identify disease-specific phenotypes may emerge as more important than EF’.
As had already become clear, Lund noted in his rebuttal that he and Solomon agree on many things. In defense of EF, he said that it is the only thing we have now. When it was developed as a measure of HF, it was a strong representation of what we were looking for, namely contractibility. Indeed, future, better measures may emerge, but today it is EF.
Indeed much data supports that HFmrEF Is intermediate, but he thinks it might be misleading to call it that. 20-30% of HFrEF patients improve into the midrange category. And a proportion worsens. But HFmrEF is fundamentally different from patients with HFpEF. He agrees that unequivocally the prognosis is different from HFrEF, but that does not make it two distinct diseases.
Solomon also noted that they agreed on more than they disagree on, they did, after all, write that recent article together. His main focus is ‘have our ‘arbitrary’ constructs hurt us by denying our HF patients beneficial therapies?’ For that, we have to look beyond the concept of EF.
Maybe novel methods including machine learning that take into account more than one measure will provide the answer. That will improve HF care as compared with using a measure that is antiquated and difficult to measure accurately.