August 30th, 2014
Will There Be a PARADIGM Shift in Treatment of Heart Failure?
John McMurray, MbChB MD and Milton Packer, MD
CardioExchange’s Harlan M. Krumholz interviews John J.V. McMurray and Milton Packer, the two lead authors of the PARADIGM-HF trial. The industry-funded randomized trial, comparing angiotensin–neprilysin inhibition with enalapril, is published in The New England Journal of Medicine.
Read CardioExchange’s news coverage of the trial, including a summary of the findings.
Krumholz: On a 1 to 10 scale, how likely is it, given what we know so far, that in 5 years this will be the main way we treat patients with heart failure and depressed systolic function in countries that can afford the medication?
Packer: I would say 9 to 10 that angiotensin-receptor neprilysin inhibition will be the cornerstone of treatment for patients with chronic heart failure and a reduced ejection fraction, replacing the current use of ACE inhibitors and angiotensin-receptor blockers.
McMurray: Cost is key. If this is affordable, it will be used as commonly as beta-blockers are now — 90%. This will be a class I, LOE A guideline recommendation.
Krumholz: What was most challenging about the trial?
Packer: The most challenging aspect of the trial was to design it in a way that, as a single trial standing on its own, it would be compelling enough to convince physicians to switch from a class of drugs that they have been using comfortably for 25 years.
McMurray: There were a few bumps along the road, but nothing too major. Early termination and rapid publication were quite challenging!
Krumholz: Given the relatively young age of your cohort (64 years) and that heart failure is most common in older populations, how should clinicians think about whether this trial is relevant to their patients, many of whom are not well represented in this study?
Packer: The trial recruited a very highly representative group of patients. As noted by Dr. Jessup in her NEJM editorial, the patients in this trial were quite comparable to those in other trials that changed clinical practice. Patients who were very elderly in this trial responded just as well as younger patients. Furthermore, many very elderly patients who have heart failure have a preserved ejection fraction, and these are being studied in a new trial with the acronym PARAGON.
McMurray: We had a sizeable subset of older patients, and the treatment benefit was consistent across subgroups.
Krumholz: Briefly explain for our members how you think this drug produces the observed effect.
Packer: This drug simultaneously decreases the influence of a maladaptive system (the renin-angiotensin system) while enhancing the influence of endogenous peptides that have favorable adaptive effects on the circulation. No other drug for heart failure has ever addressed the deficient response to adaptive factors (natriuretic peptides, bradykinin, adrenomedullin), which is characteristic of patients with heart failure.
McMurray: The incremental benefit was presumably due to the effect of neprilysin inhibition added to renin-aldosterone system blockade. Neprilysin inhibition augments natriuretic peptides and probably other vasoactive substances (including bradykinin). We believe that these other substances are beneficial in heart failure patients.
JOIN THE DISCUSSION
Join Dr. McMurray and Dr. Packer in reflecting on the findings from PARADIGM-HF.
To view all of our coverage from the ESC meeting, go to our ESC.14 Headquarters page.
24 Responses to “Will There Be a PARADIGM Shift in Treatment of Heart Failure?”
NEJM Journal Watch — Recent Cardiology Articles- Sex-Based Disparity in CABG Outcomes: Are We Moving Toward Parity?
- Can Genetic and Social Risk Scores Improve Cardiac Risk Prediction?
- Extensive Ablation vs. Pulmonary Vein Isolation in Persistent Atrial Fibrillation
- Can a Novel Stent Design Improve Long-Term Outcomes of Percutaneous Coronary Intervention?
- Are Angiographically-Derived Coronary Physiology Measurements Ready for Use in the Cath Lab?
a low-cost question:
why LCZ was not compared to Valsartan alone instead of Enalapril?
Because enalapril is still considered the gold standard, at least for clinical trials. It’s never been beaten.
Thank you
next step will be
LCZ 697 (or LCZ 696 season II) a combo Sacabitril and Enalapril…
Is this enthousiasm for One sponsorized study not too much excessive?
Time will tell us.
However many thanks to CardioEx team for their helpful analysis and discussions.
It is always a very good question to ask whether one trial — no matter now compelling — should be enough to change clinical practice. There is a mthematical way to answer this. Remember that we consider a clinical trial to be positive on the primary endpoint (in general) when the P value is less than 0.05, meaning that there is a 1 in 20 chance that the reported result is not real. If two trials are positive in a concordant direction, the likelihood of a false positive result is 0.05 x 0.05 divided by 2, or 0.00125, meaning a 1 in 1000 chance that the reported result is not real. The general formula is (0.05) to the nth power divided by n. If you apply this formula to the findings of the PARADIGM-HF trial, then the observed P value for the primary endpoint — which was 0.0000002 — indicates that the strength of evidence is actually greater than FOUR independent positive trials with concordant and statistically significant results. The likelihood that the observed effect is not real is 1 in 5 million! So one cannot reasonably refer to the findings of PARADIGM-HF as having the strength of evidence of only one study.
Regarding Dr. Usdin reference to a “sponsored” study, I am sure that he understands that all studies are “sponsored”, since someone has to pay for them. He seems to imply that an “industry” sponsored trial is somehow more suspect than one sponsored by the NIH. But if he has ever actually participated in an NIH-sponsored trial, I know he would feel very differently.
One of the good things about having been around for a while is remembering prior studies, and this one sounds exactly like CHARM-ADDED. To recall, between March, 1999, and November, 1999, CHARM-ADDED enrolled 2548 patients with New York Heart Association functional class II–IV CHF and left-ventricular ejection fraction 40% or lower, and who were being treated with ACE inhibitors. CHARM randomly assigned patients candesartan (n=1276, target dose 32 mg once daily) or placebo (n=1272). At baseline, 55% of patients were also treated with blockers and 17% with spironolactone. The primary outcome of the study was the composite of cardiovascular death or hospital admission for CHF. The median follow-up was 41 months. 483 (38%)
patients in the candesartan group and 538 (42%) in the placebo group experienced the primary outcome (unadjusted hazard ratio 0·85 [95% CI 0·75–0·96], p=0·011; covariate adjusted p=0·010). Candesartan reduced each of the components of the primary outcome significantly, as well as the total number of hospital admissions for CHF.
So, let’s consider PARADIGM-HF. Curiously, John McMurray is the lead author of both CHARM and PARADIGM. But, as a believer in CHARM results, I would venture that enalapril is a very convenient straw man, and that the former CHARM investigators (and Novartis) should have used candesartan plus enalapril as the control group to compare to the new combination drug (valsartan recently having lost patent protection). The results of PARADIGM-HF, hazard ratio in the LCZ696 group, 0.80; 95% confidence interval [CI], 0.73 to 0.87; P<0.001 are within the 95% CI of the CHARM-ADDED results, and clinically equivalent.
So I don't quite get the enthusiasm. I plan to continue using candesartan (newly generic) + ACEI in my HF patients. It is cheap and has withstood the test of time.
SORRY FORGOT REFERENCE: McMurray JJV et al. Lancet 2003; 362: 767–71. Published online Sept 1, 2003
Tony, the FDA reviewed the CHARM-Added data and rejected the claim that candesartan had value when added to an ACE inhibitor.
The effect of the combination on cardiovascular death and on all-cause mortality was not robust or persuasive.
Furthermore, Tony misses one of the important parts of our data, i.e., LCZ696 had fewer adverse effects than enalapril.
In contrast, the combination of candesartan and an ACE inhibitor in the CHARM-Added trial (and other trials) produced more side effects than an ACE inhibitor alone.
Agree with colleague Usdin,
the comparison should be like valsartan vs valsartan+LCZ OR enalapril vs enalapril+LCZ. Even better would be enalapril vs LCZ – in this case we would know what the effect of LCZ is (yet approximately).
Also, the enalapril was used in the common dose, but valsartan in increased dose.
As a result one may not be sure that the difference (effect size) is because of LCZ. The effect of LCZ is not known yet. All tricks with inadequate comparator are here together.
I am confused by Dr. Vlassov’s comments. We cannot give enalapril together with LCZ696 because of the risk of serious angioedema. We made that point very clearly in our paper. Similarly, it would not have made sense to have combined valsartan and LCZ696, since valsartan is part of the LCZ696 complex. Dr. Vlassov seems to think that valsartan 160 mg daily is somehow superior to enalapril 10 mg BID. But he should remember that valsartan 160 mg daily failed to show any effect on mortality, whereas enalapril 10 mg BID was shown to be effective in reducing the risk of cardiovascular death. Those who consider valsartan 160 mg BID to be a first choice in patients with heart failure (over enalapril 10 mg BID) are not practicing evidence-based medicine.
I agree with Dr Vlasov’s general comment, but still contend that the only ARB with clear activity in HF is candesartan. If you think back to Val-HeFT, valsartan had NO effect on mortality, and in fact increased it in patients on ACEI + BBLK (kind of a problem). The study was foisted on us as positive due to a funky complex primary endpoint, vigorous marketing, prestige of investigators, and insufficiently critical review, I think.
With regards to losartan, not clear that this weak short-acting drug is active in HF.
Tony
You are right. Valsartan did not reduce the risk of death when compared with placebo. The data are better with candesartan, but in fairness, it is not very impressive. The CHARM-Alternative trial, which compared candesartan to placebo did not show a significant reduction in mortality, and the effect size was smaller than reported in the SOLVD-Treatment trial.
Tony, the FDA reviewed the CHARM-Added data and rejected the claim that candesartan had value when added to an ACE inhibitor.
The effect of the combination on cardiovascular death and on all-cause mortality was not robust or persuasive.
Furthermore, Tony misses one of the important parts of our data, i.e., LCZ696 had fewer adverse effects than enalapril.
In contrast, the combination of candesartan and an ACE inhibitor in the CHARM-Added trial (and other trials) produced more side effects than an ACE inhibitor alone.
The key question in PARADIGM HF is blood pressure control and why treatment clearly beneficied more to those patients less sick. Is specifically interesting the significant interaction between treatment effect and NYHA class in the primary outcome but not in death, so the main interaction was in hospitalization an end point very sensitive to appropiate BP control.
Also an additional arm in the trial of VALSARTAN 160 alone would help to clarify the especific effect of ARNI treatment in the populations and outcomes uder investigation.
We hope that we have a new drug that may help our HF pts to live better and longer. But prior to accept that I think that the role of BP difference in outcomes should be completely clarified.
If you read our paper, you will see that we specifically addressed your question. Using conventional statistical modeling procedures, we asked whether the difference in blood pressure between the LCZ696 and enalapril groups could account for the difference between the two groups in the risk of cardiovascular death or hospitalization for heart failure. We found that the blood pressure had NO effect on the outcome. So this should provide consider reassurance that we are not looking at a BP difference as the explanation.
Dr. Valencia suggests a significant interaction between the treatment effect and NYHA class. To be fair, the P value we reported for this interaction was “nominal”, meaning it was not adjusted for multiple comparisons. If we had done so, the interaction would not be significant. More importantly, there was no interaction of NYHA class and the effect on cardiovascular mortality, arguably a more important endpoint. So Dr. Valencia should be greatly reassured by our findings.
In the paper, we specifically said that we had analyzed the relationship between the effect on blood pressure and the effect on cardiovascular death and hospitalization for heart failure.
There was no relationship whatsoever.
This means that the advantage of LCZ696 was not related to its additional effects on blood pressure.
What it is stated in the paper is far from a demonstration that BP control do not played a key role in trial results:
-“As compared with the value at randomization, the mean systolic blood pressure at 8 months was 3.2±0.4 mm Hg lower in the LCZ696 group than in the enalapril group (P<0.001) (Section 8 in the Supplementary Appendix). However, when the between-group difference in blood pressure was modeled as a time-dependent covariate, it was not a determinant of the incremental benefit of LCZ696."
With this data the reader know what you say but not what you do not say:
-The effect of treatment in the different components of BP are unknown (SBP, DBP and PP are all important aspects in the relation between blood pressure and outcomes in pts with LV dysfunction) in any of the study group.
-We know that when you modelled "between-group difference in blood pressure as a time-dependent covariate" there was no association with outcomes, but what components of the blood pressure were modeled, what additional models were tested, and the results of this analysis are not provided nor stated.
-Finally there are several data as pointed previously that suggest that those pts less sicker were those with greater benefit in the trial as suggested in subgroup analysis by age, NYHA class, prior hospitaliation for heart failure and prior treatment with aldosterone antagonists. These HF patients would have better BP at randomization and greater potential benefit of lowering blood pressure by the treatment under investigation.
I fear that Dr. Valencia is having difficulty understanding how to interpret subgroup effects. One does not interpret subgroup effects by looking at the confidence intervals of each subgroup and determining if each crosses the line of neutrality (risk ratio = 1.0). Doing so is tantamount to performing within-group comparisons in a parallel-group trial, which one should never do. The right way to look at subgroup effects is to look for the consistency of the effect across subgroups, which was particularly striking across all prespecified subgroup in this trial, particularly with respect to cardiovascular death, which was the driving endpoint for the study. There is simply no support for Dr. Valencia’s feeling that sicker patients responded less well to LCZ696. However, it is important to note that most of the patients in the trial were clinically stable patients with mild symptoms, precisely the type of patients that physicians think are doing well. This means that physicians need to understand that patients who appear to be “doing well” are actually NOT “doing well”. They still die; they are still hospitalized for heart failure; and they still have worsening of symptoms and quality of life. But they really do better with LCZ696 than enalapril.
The subgroups analysis are useful to look for the consistency of the effect across subgroups, and I am sure you know, it also provide valuable information during exploration of the data, help to construct models, may suggest the presence of confusion or interaction in the results… Are exploratory, and may provide new hypothesis to test. And may help to know or clarify what´s going on with the data.
I realize that you know what I am pointing out, it seem that the effect of ARNI was different in magnitude in some subgroups (see HR). That is different to say thart there were no effect. I am also sure that Dr Packer is able to differentiate between confusion and interaction and the clues needed to do so, although with the data provided in the PARADIGM HF paper we are not able to know if there is confusion or interaction in the fact suggested by subgroup analysis that less sicker pts did better with ARNI (please see the magnitude of the HRs for the end-point according to subgroups previously pointed: age, NYHA class, prior HF hospitalization or prior treatment with aldosterone antagonists). Wether this relations relations suggests counterintuitive evidence that those pts being less sicker, with lower neurohormonal (NH) activation and lower risk of events during follow up do better with aggresive NH active treatments (ARNI) than those sicker with greater NH activation and greater risk for events analyzed in the trial, or it´s the result of trial design or data is confused by other factor such are BP control or others.
I am sure that you are able to distinguish between confusion, statistical interaction and biological interaction. I am sure that you are able to go beyond confidence intervals and p value for interaction terms.
We only could say, because we don´t have access the data, suggestions and doubts related with the alternatives to compare and results in the PARADIGM HF. I am really interested in apply to my pts the better treatment options and I really hope the product tested in PARADIGM HF do so, but I think that more information and analysis are needed.
Thank you so much for your question. We have examined the data very carefully in many ways to address your question. You are quite right. One needs o go beyond confidence intervals and interaction P values. Therefore, we looked at ALL variables that might reflect the severity of heart failure and looked as to whether they influenced the size of the treatment effect. When we did so, the sicker patients had a slightly larger treatment effect based on some variables and a slightly smaller treatment effect based on other variable. Taken together, the magnitude of the benefit in “sicker” patients was entirely consistent with that seen in the entire trial.
Milton, thanks for your reply. We may have to agree to disagree. FDA follows many rules when deciding whether to approve a drug, many are opaque to the practicing cardiologist. I suspect that not controlling the specific ACEI in the control arm of CHARM was confusing, and they did not want to issue a blanket approval.
Notwithstanding, there was added benefit to candesartan as add-on to ACEI (HR 0.85. FDA-requested analyses demonstrated that the benefit was not dependent on specific agent nor on dose of agent. I don’t remember Dr. McMurray ever retracting results, so the conclusions of the 2003 paper stand.
“In conclusion, the addition of candesartan to
an ACE inhibitor and other treatments, including a
blocker, is generally well tolerated in patients with CHF
and a low left-ventricular ejection fraction and leads to a
clinically important reduction in cardiovascular mortality
and morbidity.”
I also disagree that the results of CHARM were not robust. The HR and 95%CI were 0.85, 0.75-0.96; the HR was well within the 95%CI of PARADIGM. The p-value reflects the precision of the estimate, which in such a massive trial as PARADIGM, is quite good. We must not lose sight of the similarity in treatment effect, however.
Regarding side effects, the run-in period really makes an interpretation of side effect profile really difficult – even above my pay grade. This is the kind of design pharma likes but we should resist.
Finally, valsartan does not have a well-demonstrated beneficial effect in HF: 1] note absolute lack of mortality effect in Val-HeFT; or an add-on effect in LV dysfunction – note absolute lack of benefit over ACEI in VALIANT.
So I suspect that the benefit in PARADIGM is due to the sacubutil component. But we may never know. I also think FDA may have a problem with this odd drug combination.
Great to debate with you again.
Tony
I do not think that the FDA decision-making process is so difficult to understand. Having served as a consultant to FDA and a member of its Advisory Committees for the last 25 years, I can attest to the extraordinary logic of their thinking. Many of the decisions made by the FDA that seems inexplicable to cardiologists are generally related to two factors: (1) the FDA has access to far more data; and (2) most cardiologists are not very familiar with the principles of analysis of clinical trials.
Tony, in all fairness, the results you refer to from the CHARM trials is derived from a meta-analysis of two of the component trials. When candesartan was compared with placebo in the CHARM-Alternative Trial, the drug produced no significant effect on mortality. A similarly disappointing result was seen with valsartan in the Val-HeFT trial. That is the reason why no ARB has a mortality reduction claim for use in patients with heart failure.
I agree that the benefit seen in PARADIGM-HF is due to the sacubatril component. But we do not need to splice this out. Neprilysin inhibition (used without an inhibitor of the renin-angiotensin system) is ineffective in heart failure, and combining a neprilysin inhibition with an ACE inhibitor is dangerous. The only effective and safe approach is to combine an ARB with a neprilysin inhibitor. So there is nothing particularly odd in combining the two. It is the only logical thing to do.
I guess I am nitpicking on candesartan, but the report of the LV dysfunction patients of CHARM showed an important reduction in the primary outcome (time to first event by intention to treat) was cardiovascular
death or CHF hospitalization for each trial, with all-cause mortality a secondary end point in the pooled analysis of the low LVEF trials. Of the patients in the candesartan group, 817 (35.7%) experienced cardiovascular death or a CHF hospitalization as compared with 944 (41.3%) in the placebo group (HR 0.82; 95% CI 0.74 to 0.90; P<0.001) with reduced risk for both cardiovascular deaths (521 [22.8%] versus 599 [26.2%]; HR 0.84 [95% CI 0.75 to 0.95]; P=0.005) and CHF hospitalizations (516 [22.5%] versus 642 [28.1%]; HR 0.76 [95% CI 0.68 to 0.85]; P<0.001.
This was a pre-specified pooled analysis, with a robust enough p to withstand any correction for multiplicity. So to bring this back to PARADIGM, I am still left wondering how an additional 18% reduction in events in the enalapril group by the addition of candesartan would have fared when compared to the new combination drug. Obviously we will never know.
Tony
You are right. We will never know.
But don’t you think the difference in side effects matters?
Milton
Of course it does. Our will be interesting to follow the drug’s regulatory and then clinical path. Best regards as always. T