乌鸦传媒

After two days of deliberation on what could become the first approved drugs for heart failure with preserved ejection fraction (HFpEF), FDA advisors voted in favor of both -- but not for HFpEF exactly.

Rather, members of the FDA's Cardiovascular and Renal Drugs Advisory Committee argued that they wanted to see indications for sacubitril/valsartan (Entresto) and the mineralocorticoid receptor antagonist spironolactone (Aldactone) that specified a narrower mid-range of ejection fraction (EF) -- perhaps 45% to 55% or 57% or just dubbed "mildly-reduced."

Moreover, the reluctance to endorse a HFpEF indication per se wasn't based solely, or even primarily, on the fact that trials with both agents missed their primary endpoints.

Sacubitril/valsartan

The advisory panel voted almost unanimously (12 "yes" to one "no") for expansion of the indication for sacubitril/valsartan beyond the currently-approved heart failure with reduced ejection fraction (HFrEF).

The applicant, Novartis, proposed that expanded indication to read: To reduce worsening heart failure (total heart failure hospitalizations and urgent heart failure visits) in patients with chronic heart failure and preserved ejection fraction with left ventricular ejection fraction below normal.

However, the panelists were fairly unanimous in rejecting the use of the "preserved" terminology to describe the group with potential to benefit from a broader indication.

The application centered on the PARAGON-HF trial, which missed its primary aim of reducing HF hospitalization and death from cardiovascular causes with sacubitril/valsartan versus valsartan alone in those with a left ventricular EF of 45% or higher (RR 0.87, P=0.06).

"The idea of dichotomizing the success of studies by a P value being less than or greater than 0.05 has no basis in law, either natural or federal or in regulations, and it is barely mentioned in guidance," said the FDA's Norman Stockbridge, MD, PhD, in opening up the day's deliberations.

Stockbridge pointed out examples where the agency has approved drugs despite a neutral primary endpoint finding, such as enalapril on the basis of 's secondary endpoint benefits and bivalirudin (Angiomax) on the basis of .

Many panelists said they found the consistency of the relative efficacy across different ways of looking at PARAGON-HF reassuring.

• A pre-specified exploratory analysis adding urgent hospital visits that didn't result in admission to the primary composite endpoint tipped into nominal statistical significance favoring sacubitril/valsartan (RR 0.86, two-sided P=0.04)
• A sensitivity analysis of investigator-reported, unadjudicated primary endpoint events did the same thing (RR 0.84, two-sided P=0.01)
• And the FDA's independent readjudication to give "partial credit" for events excluded due largely to missing documentation likewise tipped the results toward statistical significance (RR 0.86, two-sided P=0.04)

Ravi Thadhani, MD, MPH, of Massachusetts General Hospital and Harvard Medical School in Boston, said he voted in favor of expanding the indication based on the totality of data added to the safety profile of the agent. He cited the fragility of the P value in that just seven additional primary endpoint events would have changed the trial to a positive one.

And there's biological plausibility, a number of panelists noted.

The two subgroups who appeared to benefit significantly in the primary analysis were women (for whom the threshold for low EF is higher) and those with EF on the lower end of what's considered HFpEF (below the median 57% in a trial that enrolled patients with an EF ≥45%).

Other drugs working on neurohumoral pathways, like (Atacand) and , also seem to be beneficial in patients with a LVEF in lower part of the HFpEF range, noted John McMurray, MD, of the University of Glasgow in Scotland, who presented as a PARADIGM-HF primary investigator.

However, that group from 45% to 57% could just represent "the best of the HFrEF patients," argued panelist Scott Emerson, MD, PhD, a biostatistician at the University of Washington in Seattle. He declared himself not at all convinced that the trial showed benefit for HFpEF. "We're really just talking about moving the threshold for what is HFrEF."

The semantics were key for panelist Paul Ridker, MD, MPH, of Brigham and Women's Hospital in Boston, although he agreed that the combination of PARAGON-HF with the PARADIGM-HF trial in HFrEF supported benefit in some in-between group.

"Words matter more than we are giving them credence here," he said. "Preserved ejection fraction sounds normal, so maybe that's not the word we want. I don't like lower limits of normal, because I don't really know what that is. I must say I don't like mid-range either because I think the clinical community is not as up to date with this as is the heart failure community."

"Mildly reduced, I as an echocardiographer know exactly what that is," he added. "That would be the middle ground that would find us an expansion of this label to a place I'm comfortable with, without going someplace I think the data become quite murky."

Given that the trials didn't have EF measured by a core lab, some substantial proportion of those in the 45% to 50% range could have been in the 30% to 40% range by another reader, meeting the PARADIGM-HF trial criteria, noted panelist Steven Nissen, MD, of the Cleveland Clinic.

Ejection fraction measurements likely have a standard deviation of plus or minus 5 percentage points, or even more at some labs, acknowledged Scott Solomon, MD, of Brigham and Women's and PARADIGM-HF trial co-chair.

In one experiment by his group, about one-third of 100 patients initially scored in the 40% to 50% EF range by the core lab were deemed outside of that range when redone by another reader, he said. "That's even ignoring the potential biological variability," he said. "There's some degree of imprecision there. We live with it."

Even so, Ridker argued, "it's also the best that we have, and it's clearly the way this drug will get used because out in the real world this is the fundamental way ejection fraction is measured."

With a significant interaction term for treatment effect by ejection fraction (P=0.035) and no benefit seen in those with an above-median EF, Nissen agreed that EF has to play a role in prescribing sacubitril/valsartan.

"If you look at all the analyses, no matter how you cut it, regardless of gender or anything else, the lower the ejection fraction the more the appearance of benefit you see in PARAGON. There's not much evidence that truly HFpEF benefits," he said. "Where you draw that line is really hard, but I was convinced by the post hoc analyses that something was going on there."

Stockbridge said the agency "will work out a reasonable way of describing the heart failure spectrum, but preserved and reduced is probably not a very useful description."

The only "no" vote, by committee chair Julia Lewis, MD, a nephrologist at Vanderbilt Medical Center in Nashville, centered on multiple concerns: lack of any prior drugs that work for HFpEF, that the trial didn't directly address mid-range HF, and hypotension as a side effect that could cause harm.

Spironolactone

In the second day of deliberations, the same panel looked at expanding the indication for spironolactone beyond HFrEF into HFpEF.

Again, the panel voted in favor of some indication expansion but with less certainty: eight voted yes, four no (Lewis again, along with Nissen and both of the statisticians on the panel), and one abstained (Ridker).

appeared to be in part due to poor results with spironolactone among the 49% of participants from sites in Eastern Europe (Russia and Georgia), whereas it bested placebo by a relative 18% for the primary endpoint in the remaining participants from North and South America.

The primary event rate in Eastern Europe overall was about one-fifth of that in the Americas and went the wrong way for the drug (HR 1.10 vs 0.82); cardiovascular death in that region was also more common with spironolactone (HR 1.31 vs 0.74).

Excluding such a large region from the data considered for a drug indication would be unprecedented, but Stockbridge pointed out that approval for ticagrelor (Brilinta) in prevention after acute coronary syndrome excluded North America and relied on data from the rest of the world.

The Eastern European participants -- being younger, more often with a history of coronary disease or aspirin treatment, and less often entering the trial on the basis of elevated natriuretic peptide levels -- they had substantially lower than expected event rates in the placebo group (2.3 vs 12.6 per 100 patient-years), and there were strong signals of poor drug adherence versus participants in the Americas.

In fact, the TOPCAT trial presenters argued that many of the Eastern European participants didn't actually have heart failure and were enrolled fraudulently for financial or quality of care incentives. They wanted that region's data adjusted off or thrown out, leaving the positive data from the Americas.

However, "There is insufficient evidence to conclude the two regions are different enough that the overall results should not be considered for the primary endpoint," cautioned FDA biostatistician Ququan Liu, MD, who presented the agency's analysis.

The interaction between region and treatment effect wasn't even close to significant in a statistical sense (P=0.12). Also, patients who qualified for the trial by prior HF hospitalization showed no treatment effect in either region, which undermines the explanation that different populations between two regions led to different outcomes, noted Lu.

"If spironolactone is approved for this indication, this may set a precedent on what is considered as substantial evidence for approval," she added.

That was a big concern cited by panelists in how they voted, along with the poor overall conduct of the trial.

Those who voted in favor of some indication tried to thread the chasm by suggesting an indication for the secondary endpoint of heart failure hospitalization alone, which in the overall trial had a significant benefit without any fancy statistical wrangling (HR 0.83, P=0.042).

Again, there was a concern about not expanding to all of HFpEF, but with a poorer database to go on. While those below the median of 56% EF had a spironolactone benefit with a hazard ratio of 0.78 (P=0.02), there was a nonsignificant interaction term for treatment effect by EF (P=0.22).

"Like many of you, I am very concerned about the precedent of dropping half the data. I think that would be wrong, even though in this case I believe half the data is wrong," Ridker said. Even though he abstained from voting, he said, "I think I would be very comfortable with a secondary endpoint of heart failure hospitalization of people with mildly-reduced [EF]. But the problem is, we never saw that data today."

If the FDA can combine that data from TOPCAT with that of the in patients with an EF ≤35%, maybe that's the way forward, Ridker suggested.

"I'm equally concerned about establishing a precedent that we cannot do generic trials on reasonable budgets that our federal government would do," said C. Noel Bairey Merz, MD, of Cedars-Sinai Medical Center in Los Angeles, in explaining her vote for an indication based on the secondary endpoint of heart failure hospitalization.

Notably, the FDA initiated the application for HFpEF itself as it couldn't find any company willing to apply. TOPCAT was sponsored by the National Institutes of Health, and spironolactone is available from multiple manufacturers in generic form.

The FDA does not have to act in accord with the approval recommendation of its advisory committees but often does.

Comment