Understanding the Results of the PARAGON-HF Trial

Abstract

The angiotensin receptor–neprilysin inhibitor sacubitril/valsartan reduces the risk of hospitalisation for heart failure (HF) or death from cardiovascular (CV) causes in patients with HF with reduced ejection fraction (HFrEF), but not in those with preserved ejection fraction (HFpEF).1, 2 Contrary to the Prospective Comparison of ARNI [Angiotensin Receptor–Neprilysin Inhibitor] with ACEI [Angiotensin-Converting Enzyme Inhibitor] to Determine Impact on Global Mortality and Morbidity in Heart Failure trial (PARADIGM-HF) where all the pre-specified subgroups homogeneously showed a possible benefit, in the Prospective Comparison of ARNI with ARB Global Outcomes in Heart Failure With Preserved Ejection Fraction trial (PARAGON-HF) of the 12 pre-specified subgroups only two suggested a possible treatment benefit for patients with an ejection fraction between 45% and 57% (the lower range studied) and in women in whom HFpEF is more common than in men.2-4 PARAGON-HF is a study that was carefully planned and well conducted. It had an active control arm (with valsartan) despite no approved efficacy for any treatment in HFpEF. The included patients were properly characterised for having signs and symptoms of HF, New York Heart Association (NYHA) class II to IV, an ejection fraction of ≥45%, documented structural ventricular abnormality on echocardiography and elevated N-terminal pro-B-type natriuretic peptide (NT-proBNP) (ranging from 453 to 1625 pg/mL). Out of 4822 randomised patients, 4787 completed the study; 26 patients, equally distributed, were withdrawn for major good clinical practice violation.2 It follows that the results came as a surprise considering the high expectations in view of the known benefits of treatment with sacubitril/valsartan in patients with HFrEF. The scope of the present commentary is to provide possible explanations for the results of PARAGON-HF considering that, although the patients enrolled had almost identical entry criteria than those of PARADIGM-HF apart from a lower ejection fraction, the pathophysiologic milieu of the two studies was completely different. In no way this is a criticism to the construction and conduction of the trial that is one of the best, if not the best in this pathology. Our contribution would represent not only a commentary strictly related to PARAGON-HF but, more in general, an attempt to guide the reader to a critical approach when interpreting ‘negative trials’ avoiding a summary judgement of ‘drug inefficacy’. One should consider that ‘HFpEF’ could be simply a nosographic term, exciting from a conceptual point of view but too weak to really group under the same umbrella so intrinsically different diseases. The wording HFpEF includes multiple unselected patients with many different phenotypes. It has been proposed that heterogeneity across many aspects of HFpEF as well as of acute HF is a major factor influencing clinical trials.5, 6 Multiple pathophysiologic pathways can lead to HFpEF and to the lack of significant results. The same authors who advocate the ‘HFpEF syndrome’ distinguish between the systemic inflammatory syndrome and specific underlying aetiologies, including specific myocardial and pericardial disease. As regards patients' profile, compared with HFrEF, patients with HFpEF are generally older, more likely to be female, have higher body mass, higher rates of iron deficiency, hypertension, and atrial fibrillation but have less coronary artery or valvular diseases. Often, all these different conditions are concomitant and contribute in different ways to the symptoms of HF. It is therefore difficult that one treatment strategy applied to such a broad population of patients could improve outcome. This suggests that phenotyping patients and administering treatment specific for the underlying phenotype could be a more effective approach.6 For example, patients with HFpEF due to transthyretin amyloid cardiomyopathy, which can represent up to 12% of those with HFpEF, will not respond to sacubitril/valsartan but will respond to a specific treatment with tafamidis, which reduced all-cause mortality and CV hospitalisation in a study of only 173 treated patients.7 In the real world, and also in the PARAGON-HF population, it is likely that some cases of amyloid cardiomyopathy and, possibly, other conditions, such as hypertrophic cardiomyopathy, may account for the reduced responsiveness to sacubitril/valsartan, particularly in those patients with higher ejection fraction. HF trials have become increasingly global to achieve in due time the requisite number of patients and to compensate for the lower enrolment rates in many Western countries. Geographical heterogeneity might also influence the results of clinical trials, both in HFpEF as well as in acute HF.6, 8 The PARAGON-HF trial enrolled patients from 43 different countries. Although the interaction among different geographic regions on the primary endpoint was not statistically significant, sacubitril/valsartan was superior to valsartan alone in reducing the primary endpoint only in Western Europe [rate ratio 0.69, 95% confidence interval (CI) 0.53–0.89] and not in the other four geographic regions (pre-specified subgroups). All the evidence confirms that HFpEF is a separate syndrome from HFrEF9-11 (Figure 1). The two differ in terms of aetiology, pathophysiology, comorbidities, clinical and demographic characteristics and response to therapy.9 HFrEF is a cardiac syndrome, driven by myocardial cell loss and fibrosis which, in turn, as a result of reduced cardiac output evokes an important systemic (neuro-hormonal) reaction responsible for the symptoms and which constitutes a key target for therapy.12, 13 Sacubitril/valsartan exert an important and novel action on neuro-hormonal activation.14, 15 On one hand, the valsartan component inhibits the deleterious effect on the CV system of the increase of the renin–angiotensin system. On the other hand, the sacubitril component, through the inactivation of neprilysin, allows the recruitment of the favourable effects of atrial natriuretic peptides. Since neprilysin inhibition, unfortunately, may augment levels of the vasoconstrictor angiotensin II, decrease levels of the vasodilator angiotensin-(1-7) and has a key role in the metabolism of other molecules such as bradykinins, the association with valsartan [instead of angiotensin-converting enzyme (ACE) inhibitors] has a double rationale. The net result is a re-equilibrium of two opposed forces: the angiotensin II-induced vasoconstriction and water retention and the vasodilatation and diuresis induced by atrial natriuretic peptides.14 The predominance of the latter system induced by sacubitril/valsartan results in an amelioration of symptoms, reduced progression and, in some cases, even a regression of left ventricular remodelling and, eventually, improvement of outcome as demonstrated by the PARADIGM-HF trial.1, 16, 17 What is relevant for the present discussion is that in HFrEF the ‘primum movens’ that drives all changes is in the heart: a loss of the pumping capacity of the ventricle. Quite to the contrary, in HFpEF the ‘primum movens’ is not the heart, but the periphery.9, 10 HFpEF is considered to be a systemic syndrome driven by accumulated risk factors and comorbidities which, in vulnerable subjects, especially elderly women, cause symptoms of HF and structural changes in the ventricle, mainly a loss of compliance but with maintenance of the circumferential function of the ventricle, as described by ejection fraction (notably overall systolic function and myocardial contractile performance in particular can be reduced). The usual risk factors at the basis of HFpEF, apart from age and female gender, include hypertension, diabetes, metabolic syndromes, renal dysfunction, obesity. Consequently, HFpEF is often considered to be more an inflammatory than a neuroendocrine disease.18 The cumulative expression of these risk factors and comorbidities leads to a pro-inflammatory status which activates the endothelium of the heart resulting in an increase of nitric oxide and oxygen free radicals production which, in turn, cause fibrosis, cardiomyocyte malfunctioning and hypertrophy and, eventually, diastolic but not systolic dysfunction. It follows that the two syndromes are quite different: one, HFrEF, is due to cardiomyocyte dysfunction, starts from the heart and leads through an important neuro-hormonal activation to the periphery, whilst the other, HFpEF, is essentially due to endothelial dysfunction, starts from the periphery and leads to the heart (Figure 1). As mentioned above, the direct pharmacological effects of sacubitril/valsartan are on the neuroendocrine balance and cause a physiological diuresis and vasodilatation.13, 14 This occurs when one of the neuroendocrine component of the disease (mainly in the catecholamine and renin–angiotensin system) is strongly activated as it generally occurs when systolic dysfunction triggers a sustained reduction of cardiac output. Among patients with HFpEF, cardiac output is not generally dramatically reduced and the neuroendocrine system is more likely to be less activated. Indeed not only NT-proBNP (HFpEF has been sometime explained as a natriuretic peptide deficiency state) but also the neurohormones derived by the renin–angiotensin–aldosterone and sympathetic nervous systems are significantly less in HFpEF compared to HF with mid-range or reduced ejection fraction1, 2, 19 Thus, in patients with HFpEF, HF could be, in general, much less conditioned by the balance of the major neurohormonal systems than it is in HFrEF and therefore less sensitive to all known neuro-hormonal treatments. In line with this interpretation is the possible benefit found in the pre-specified subgroup with an ejection fraction in the lower part of the range studied, i.e. 45% to 57%, and the correlation between ejection fraction and risk reduction when PARADIGM-HF and PARAGON-HF are analysed together.20 It is plausible to anticipate that in these patients there was a subtle systolic dysfunction causing activation and disequilibrium of the neuroendocrine system. Under these conditions, it is likely that sacubitril/valsartan could exert its positive pharmacological effects and reduce the risk of hospitalisation. Several post-hoc analyses of other trials in HFpEF with different drugs targeting various components of the neuroendocrine activation also showed possible benefits in patients with left ventricular ejection fraction in the range of 40% to 55%. The last European Society of Cardiology guidelines decided to consider these patients separate from those with fully preserved ejection fraction and created a new group named HF with mid-range ejection fraction (HFmrEF).21 The possibility that this specific group could benefit from sacubitril/valsartan treatment was not a priori anticipated in the design of PARAGON-HF, probably because of the larger number of patients to be included. This would have better characterised the phenotype, which could respond favourably to sacubitril/valsartan. At present, it is not even known whether the improvement of quality of life, the changes in NYHA class from baseline to month 8, and the improvement in renal function found in PARAGON-HF occurred mainly in this subgroup or in the total population.1 Another third possibility to explain the results of PARAGON-HF relates to neprilysin. Neprilysin is a cell-membrane bound endopeptidase that cleaves a variety of active peptidases, being responsible for removal of, at least, 50% of circulating atrial natriuretic peptides (NPs).14 In so doing, neprilysin exerts an anti-fibrotic, anti-proliferative, myocardial relaxation, vasodilator, and diuretic effect. The enzyme can be released from the cell surface into the circulation. In patients with HFrEF, a linear significant relationship has been shown between circulating neprilysin with CV death or HF hospitalisation22 and this could contribute to explain the beneficial effects of neprilysin inhibition in PARADIGM-HF. Whether the same correlation exists for patients with HFpEF is not known, although it seems unlikely. Circulating neprilysin levels were initially measured in few patients with HFpEF and found not to be associated with NYHA functional class, 6-min walk test, or cardiac hospitalisation.23 However, to prove that circulating neprilysin is a biomarker in HFpEF, it is necessary to compare its circulating levels in HFpEF patients with those of control subjects. This was recently done by Lyle et al.24 who performed a case control study in 242 symptomatic HFpEF patients enrolled either in the Phosphodiesterase-5 Inhibition to Improve Clinical Status and Exercise Capacity in HFpEF trial (RELAX)25 and in the Nitrate's Effect on Activity Tolerance in HFpEF trial (NEAT-HFpEF)26, 27 and in 891 asymptomatic subjects without HF or diastolic dysfunction enrolled in the Prevalence of Asymptomatic Left Ventricular Dysfunction study.28 Overall, and surprisingly, neprilysin levels were lower in the HFpEF group compared with controls (3.5 mg/mL, 95% CI 2.5–4.8 vs. 8.5 mg/mL, 95% CI 7.2–10.0; P < 0.001). Although the proportion of neprilysin in the circulation and in the tissues is unknown, and much more research is necessary on this aspect, these data are of paramount importance for the results of PARAGON-HF as they question the rather simplistic assumption of circulating neprilysin being a biomarker to be inhibited in HFpEF. Paradoxically, in absence of a biological target to be inhibited by sacubitril, it may be possible that PARAGON-HF, at least in some patients, has compared two groups both just receiving valsartan. This, obviously, is an over simplistic explanation, the biology of circulating neprilysin being more complex. For instance, activation of the leptin–aldosterone–neprilysin axis appears to contribute importantly to the evolution and progression of HF in people with obesity.28 Furthermore, the tissue and circulating activity of neprilysin was not measured in the study of Lyle et al.24 It is possible that the activity of the enzyme rather than its levels are increased in HFpEF. To this end, Nougué et al.29 showed that switching 73 patients from ACE inhibitor to sacubitril/valsartan reduced circulating neprilysin activity whereas its circulating levels remained unchanged. It is known that the activity in the blood of any enzyme may vary because of the interference of several circulating factors.24 It is possible that circulating neprilysin is increased in patients with HFmrEF, i.e. those of the pre-specified PARAGON-HF subgroup with an ejection fraction between 45% and 57% who showed a treatment benefit. Equally, it is not known whether there are gender differences in the circulating levels of neprilysin which would explain the best results obtained in women. Finally, a fourth possibility may relate to the design of PARAGON-HF. Like in PARADIGM-HF, PARAGON-HF did not test the effects of sacubitril/valsartan against placebo but against an active comparator 160 mg twice daily of valsartan. On one hand, the authors have to be congratulated as both studies did not follow the classical add-on concept of all the other trials in HF. On the other hand, this design might have contributed to the smaller than anticipated difference among groups. The potential efficacy to reduce CV mortality and hospitalisation of angiotensin receptor blockers in patients with HFpEF shown in the Candesartan in HF: Assessment of Reduction in Mortality and morbidity (CHARM)-Preserved trial was 11%.30 The same 11% reduction was found in TOPCAT with spironolactone.31, 32 It follows that tested against a placebo sacubitril/valsartan could have resulted in a 20% risk reduction of the primary endpoint, especially for HF hospitalisation. As the majority (86%) of patients were receiving a renin–angiotensin inhibitor mainly for hypertension before enrolment, it was difficult to design a placebo-controlled trial. Probably, it would have been better to perform an ongoing randomisation based on which treatment was used to treat hypertension as it was done in the PARALLAX study.33 This is a study with triple randomisation depending on the ongoing hypertension therapy at randomisation: if patients were on ACE inhibitors there is a randomisation between sacubitril/valsartan and enalapril, if they were on angiotensin receptor blockers there is a randomisation with valsartan, and if they were not treated with renin–angiotensin system inhibitors the randomisation was sacubitril/valsartan vs. placebo.33 The running period with the majority of patients receiving a renin–angiotensin system inhibitor may also have influenced the results, excluding high-risk patients or those who could not tolerate the trial drugs. Finally, the number of patients enrolled in PARAGON-HF was significantly lower than that of PARADIGM-HF (4822 vs. 8442). One could hypothesize that the final power of the study could be insufficient. From a formal point of view, this is not the case. It was estimated that 4600 patients would provide 95% power for the primary analysis, assuming approximately 1847 primary events had accrued, if the true primary endpoint rate reduction were 22%, which approximately corresponds to a 30% reduction for total HF hospitalisation and 10% for CV death, under the assumed model. For a smaller reduction of 19% in the primary endpoint rate, corresponding to a 25% reduction for total HF hospitalisations and 10% for CV death, the study would have 84% power for the primary analysis. So, the 1993 events reached in the PARAGON-HF ensure a sufficient formal power. Other considerations even support the hypothesis that some methodological choices could have affected the results of the study, for instance the exclusion of unprogrammed emergent visits from the list of primary endpoints and the choice not to consider investigator-adjudicated events. Very few more or less events on one side or the other could have changed the fortunes of the trial as expressed by the P-value. Probably PARAGON-HF represents an opportunity for ‘moving to a world beyond ‘P < 0.05’,34 where the final interpretation of a trial is based on a large number of statistical and clinical considerations and not on a single conventionally dichotomous variable.35, 36 Given the global burden of HFpEF, its projected increase over time and the lack of disease-modifying therapeutics at this time, the results of PARAGON-HF are somehow disappointing. As overall the treatment effect is modest, especially in terms of CV death, there was a non-significant trend towards a reduction of hospitalisation for HF and an improvement of the secondary outcomes, mainly NYHA class, quality of life and renal function. There was also a positive interaction between two subgroups, mainly in patients with lower ejection fraction and in women. Although the subgroup analysis was pre-specified, it is significantly and biologically plausible and observed in prior studies, but should be interpreted with caution as PARAGON-HF is a non-significant trial. As always, it is difficult to understand the reasons for the unexpected results of a trial and further analysis are needed. At a glance, four hypothesis can be considered, none is exclusive. On the other hand, as often is the case, all of them may be implicated. This work was supported by a grant from Fondazione Anna Maria Sechi per il Cuore (FASC), Italy. Conflict of interest: R.F. has received speaker fees from CIPLA, Iamicon, Merck Serono, Novartis, Servier International; and research grants/contracts from Alfasigma, DOC Generici, Novartis, Pfizer, Servier International, SPA Società Prodotti Antibiotici. A.F. has nothing to disclose. C.R. has received research grants from Pfizer, and speaker fees from Novartis, Alnylam, Akcea.

Topics & Concepts