Today’s article marks the ending of one of the better parables in medicine. It’s the story of a drug that seemed like a really good idea, moved the needle on a lot of surrogate endpoints, and ultimately proved to be ineffective for improving any outcome that might actually be important to patients. It’s the story of nesiritide, recombinant human B-type natriuretic peptide (BNP) . We’ll be examining it via the analogous TRUE-HF study by Milton Packer and colleagues, a trial of ularitide (an analog of another natural natriuretic peptide and vasodilator, similar to BNP) for cardiovascular mortality in acute heart failure.
With the incoming FDA commissioner talking about the need to speed approval processes, let’s slow down and take a good look at what happens when we use surrogate outcomes and safety data in lieu of large RCT data with clinically important endpoints. It’s a pretty good real-life example that runs counter to the argument that physicians can be left to make their own judgments about a drug’s efficacy independent of the FDA, and a reminder that as we move to speed the approval process for novel therapeutics we must do so without diminishing our ability to confidently tell our patients that these drugs really work.
If You’re Only Going to Read One Paragraph
Ularitide minimally improved symptomatic dyspnea when administered early in acute decompensated heart failure, but failed to provide a survival benefit or to prevent rehospitalization. These results are similar to those seen in trials of other synthetic analogs of naturally occurring vasodilatory hormones. The story of nesiritide, an analogous medication with analogous study outcomes, is illustrative of why strict FDA standards are necessary. Even if they do slow down the approval of novel therapeutics, they pay dividends in ensuring that physicians have the data they need to choose the best therapy for their patients.
Patients, Intervention, Comparator, and Outcomes
2157 patients were randomized to 48 hours of weight-based ularitide administration versus placebo. To be enrolled, patients had to have an unplanned ED visit or direct admission for CHF, worsening dyspnea over the past week, CXR consistent with heart failure, NT-proBNP > 2000 pg per mL, and the ability to start ularitide within 12 hours of clinical evaluation. Additionally, patients had to have received a dose of 40mg IV furosemide or equivalent and have persistent dyspnea 2 hours thereafter.
There was significant concern for hypotension with ularitide administration, so patients were excluded if they had a BP less than 116mmHg or required any hemodynamic support. There is a long list of other exclusion criteria in the supplementary appendix, notable mostly for ACS, severe chronic pulmonary disease, creatinine clearance < 30, hemoglobin < 9, or suspected infection. It’s a long list, but not an unreasonable one; I think the majority of CHF patients I treat would still have been included in this trial if they were still dyspneic after the first furosemide dose.
The co-primary outcomes were cardiovascular death over the median 15 month follow up and a hierarchical outcome describing the clinical course of patients over the 48 hours of administration. This hierarchical endpoint is a bit weird — effectively, it tries to classify patients as “improved,” “worse,” or “unchanged.” If the patient died or required initiation of IV or mechanical hemodynamic support, they were considered to be worse. Which, I mean, reasonable, death is kind of the opposite of what we go for. If they did not die or need pressors, they were asked to rate their self-perceived clinical status on a 7-point Likert scale from markedly improved to markedly worse; if they rated in either of the top two boxes at 6 hours, 24 hours, AND 48 hours after initiation of therapy, they qualified as “improved” overall. Those who did not meet the above criteria for worsening but used one of the middle three boxes were rated as “unchanged.” Those who used one of the bottom two boxes were rated as “worse” regardless of clinical course. It’s a pretty crude method of symptom management, and as far as I can find is not well-validated as a measure of symptoms in acute decompensated heart failure, but it’s not objectionable on its face.
The list of secondary outcomes include both length of stay and overall length of stay, need for initiation of hemodynamic support, 30-day readmissions, as well as changes in serum BNP and serum creatinine levels.
A Bit of Context
The hypothesis in both nesiritide and ularitide was essentially that having an elevated BNP is bad — after all, it’s only produced when your heart is under strain. The hypothesis assumed that heart strain can be tied to “myocardial microinjury.”
Packer et al., DOI: 10.1056/NEJMoa1601895
Certainly we know that troponins can be elevated by acute decompensation of heart failure, so it’s not a ridiculous premise that BNP might be a marker for the same injury on a smaller scale. So, by acutely offloading the ventricle with vasodilators, we could spare the heart the stress of operating at a time of high oxygen demand (i.e. operating while volume overloaded), which would result in decreased remodeling and better long-term outcomes.
I’m going to sum up nesiritide’s sordid story, but I’m only going to do almost as good a job as one of my favorite editorials of all time, The Lost Decade of Nesiritide. Go give Eric Topol’s piece a read if you have some time this weekend, it’s a really nice piece of writing and a good cautionary tale.
Nesiritide was approved on the basis of a single RCT in which it was shown to generate a decrease in self-reported dyspnea at 3 hours compared to placebo. Granted, only 30% of patients had received any dose of diuretic prior to enrollment in the study, which is entirely inconsistent with standard of care, and sure there was a trend towards increased 30 day mortality with nesiritide (8.6% vs 5.5%, P = 0.20) and patients on nesiritide had a statistically longer length of stay (10 vs 8.1 days, P = 0.008), but hey — it’s a positive RCT. Besides, patients receiving nesiritide had lower pulmonary capillary wedge pressures than those on placebo, and lower wedge = more better. So nesiritide was approved. Of course, the manufacturers were responsible stewards of this approval, and physicians likewise ensured that the drug was reserved for its proper indication and not used out of proportion to its proven efficacy.
Nah, they marketed the hell out of it. More concerning than its widespread inpatient use despite its concerning safety profile and no obvious benefit apart from earlier improvement in dyspnea, a piece in the New York Times in 2005 detailed the widespread off-label use of nesiritide infusion in outpatients as a “tune up” therapy. Doctors used the same codes they used to bill for the infusion of chemotherapy in a practice without anything even resembling an evidence base. And if you wonder why, consider that Scios, the company that brought nesiritide to market, set up a reimbursement hotline dedicated exclusively to nesiritide and published numerous guides on how to get paid for the outpatient infusion of the drug — which practice, again, was not based on any meaningful evidence.
And then, the hammer fell. An RCT of nesiritide vs placebo again demonstrated self-reported improvement in dyspnea (on the same 7-point Likert measure our current study employs), but not at a threshold prespecified for clinical significance. The difference here might well have been that 100% of patients received diuretics. The investigators point out cheekily that we don’t know the comparative effectiveness of nesiritide with standard therapy like diuretics as monotherapy, but we do know this; diuretics cost about 0.1% the price of a vial of nesiritide, and don’t have nearly the rates of hypotension forcing drug discontinuation.
Oh, and also, again no mortality benefit, and no decrease in rates of rehospitalization.
We spent north of a billion dollars on nesiritide. And yes, we spend $3.2 trillion on healthcare, but when we’re doing things like proposing taking away the $140 million dollar National Endowment for the Arts because we just can’t afford it, surely we have proven that savings on the scale of a billion dollars are important to us.
Back to ularitide. I’m going to be brief here, because this is really just a different verse in the same song. Patients were similar between the placebo and intervention groups. The median age was ~68 and 9% of patients were black. 65% of patients had an EF < 40%, and 39% had a history of diabetes. Similar rates dropped out before study drug administration, but patients receiving ularitide were more likely to require dose reductions or discontinuation of drug than those in the placebo group (P<0.001 for both), mainly due to hypotension.
The drug did not improve cardiovascular mortality at one year; 21.7% of patients receiving ularitide and 21.0% receiving placebo died of cardiovascular causes, P=0.75. There were no differences by prespecified risk strata (broken down by EF, baseline NT-proBNP, and troponins). The hierarchical outcome, the symptoms-based one, also did not differ significantly betwen the two groups . None of the secondary outcomes demonstrated any benefit.
What did show significant difference were biomarkers indicating that the drug was indeed working. BPs dropped faster on ularitide than placebo, and NT-proBNP dropped 47% more in this group than in the comparator. Troponin T levels, however, had no significant difference.
Vasodilators in heart failure sound like a great idea, and they certainly make numbers look better. It was totally reasonable to wonder whether they would help people with acute decompensations of heart failure, and I’m glad this trial was funded. However, much like albumin in cirrhotics, just because it seems like it should work does not mean that it does, in actuality, work (don’t tell the liver doctors I said that).
So the next time you hear somebody talking about how there are too many barriers to getting a new drug approved, ask them what they think about nesiritide. Remind them that in a state of nature, once it is legal to market a drug, pharmaceutical companies will do everything in their power to sell it to as many people as they can. And mostly well-meaning physicians, who are as prone to marketing as any other human, are less likely to be careful stewards of the evidence than to be unintentional shills for these same companies who, out of desire to help their patients, steer people towards unproven drugs and very proven adverse effects. The FDA has been our safeguard against this, and its ability to insist on meaningful clincal data underpinning any drug approval should be strengthened, not weakened. Make things faster, sure, though we are already faster than most. But don’t do it by lowering the standard of evidence.
Read the original study here.
Coming Up On IM HEAT
On Monday, I get to talk about the follow up to TOPCAT, one of my favorite studies and a reason you might finally have something to offer your HFpEF patients other than Lasix and prayer. Tuesday has another guest piece from Risa, our ID specialist, and it looks like we’re going to have another guest piece lined up for Thursday — no spoilers about that just yet. Plus I promise to write about at least one paper from a different journal next week to avoid this blog becoming “Bill rehashes NEJM three days a week and adds jokes.” I mean, you’d still read that blog, yes, but we can do better.