HFpEF — the thing you write on your admission H&P when your patient’s ejection fraction (EF) looks better than the patient himself. It’s well-established that congestive heart failure (CHF) can occur in patients whose EF is normal, and that these patients have symptoms and prognoses not dissimilar to patients with a decreased EF. Yet the efficacy of most therapies with a mortality benefit in CHF, from beta blockers to ACE inhibitors to implantable cardiac defibrillators, appears to be limited only to patients with a reduced EF. So the treatment of heart failure with a preserved ejection fraction (HFpEF) has historically been diuretics, blood pressure control, and a can-do attitude. It has been frustrating.
But when some surrogate endpoint data demonstrated echocardiographic and exercise tolerance improvement with aldosterone antagonist therapy, cardiologists began to hope that maybe at least one therapy used in patients with a decreased EF might benefit HFpEF patients as well. The TOPCAT trial by Bertram Pitt and colleagues (published wayyyy back in April of 2014 — more of a lukewarm trial than a Hot one) was a multinational RCT of spironolactone versus placebo for HFpEF patients. But how can I justify talking about a three year-old trial on a blog about the new hot evidence? Simple: there’s new correspondence from the study authors, and it is super cool, albeit in a moderately awful kind of way. You see, TOPCAT was a negative trial — the primary outcome was no different between spironolactone and placebo. But when you exclude the 1700 or so patients recruited from Russia and Georgia? Positive trial. Granted, that’s not how science works; you don’t just get to exclude half of your patients post-hoc because it makes your data look better. But what if you had, like, a really really good reason? Let’s take a look.
And let’s just get this out of the way right now. Apparently not everyone knows about this ancient and moderately terrible cartoon, but the theme song has been stuck in my head literally the entire time I’ve been writing this and I wanted you to suffer with me.
Good? Good. Back to it.
If You’re Only Going to Read One Paragraph
The TOPCAT trial of spironolactone versus placebo in patients with heart failure with a preserved ejection fraction demonstrated decreased rates of hospitalization but no difference in a composite primary outcome of hospitalization, cardiovascular mortality, or cardiac arrest. However, the trial would have been positive with a hazard ration of 0.82 and a NNT for one year of just 46 if the patients from Russian and Georgian centers had been excluded. Given that patients in these regions had much lower rates of the primary outcome than those in other regions, as well as new revelations that Russian/Georgian patients were significantly less likely to have serum levels of spironolactone metabolites consistent with their reported dose of the drug, it is reasonable to wonder if these patients were recruited or treated in a manner inconsistent with the study protocol. As such, we can wonder if spironolactone does actually have a benefit for HFpEF, though this should be reproduced in future studies. This is illustrative of the problems of recruiting for multinational trials in places whose medical culture and practice may be different than ours.
Patients, Intervention, Comparator, and Outcomes
In TOPCAT, investigators recruited patients 50 years of age or older with at least one sign (jugular venous distension, chest X-ray with pulmonary edema, or lower extremity edema) and one symptom (orthopnea, paroxysmal nocturnal dyspnea, or dyspnea on moderate exertion) of heart failure, with EF 45% or better based on echocardiography or nuclear study, and with either hospitalization in the last 12 months for which heart failure was a major component or elevated serum BNP level in the last 30 days. Read those last two requirements again and note how while both are objective criteria, one is more objective than the other. That’s important later.
Patients also had to have a systolic blood pressure < 140 or be on at least three agents for BP control (i.e. have treatment to the existing standard of care), as well as have a serum potassium < 5.0 mmol/L (i.e. have a reasonable chance at tolerating spironolactone). They were excluded for acute coronary syndrome or revascularization in the past 90 days, severe COPD, known infiltrative cardiomyopathy (i.e. things that modify prognosis), as well as history of hyperkalemia or glomerular filtration rate < 30mL/min (i.e. things that make spironolactone a not-great idea). Overall, a pretty reasonable set of exclusions.
Patients were stratified based on whether they were included by BNP versus by hospitalization, then randomized 1:1 to placebo versus spironolactone. Spironolactone was started at 15mg daily, and then increased every four weeks to a maximum dose of 45mg daily if the patient was able to tolerate it with respect to creatinine, potassium, and blood pressure.
The primary outcome was a composite of cardiovascular mortality, cardiac arrest, and hospitalization for heart failure. Secondary outcomes included each of these individual components, as well as MI and stroke. Important safety outcomes included potassium and creatinine levels, as well as rates of drug discontinuation.
Important to our current conversation is that investigators kept patient blood samples collected during the trial in a “biorepository” (just call it a freezer, guys), available for reference post-hoc. Not every letter to the editor comes with its own supplementary appendix, but this one is necessary for us to establish that participants in this biorepository (OK it’s growing on me, kind of fun to say) were more likely to be recruited from the BNP arm than the hospitalization arm, but were otherwise stastically similar to the study population as a whole. With this blood on hand, the researchers realized they could look for circumstantial evidence of medication compliance. They tested the samples for canrenone, a known metabolite of spironolactone. And the results are… well, wait for it, we’ll get there.
A Bit of Context
I’m not going to spend a ton of time going over prior investigations into aldosterone antagonists for heart failure, but, y’know, it works. Briefly, the Randomized Aldactone Evaluation Study (RALES) found a mortality benefit in patients with an EF of 35% or less and NYHA class 3-4 symptoms already on an ACE-I and diuretics (and digoxin, because it was 1999 and it seemed like a good idea at the time). The Eplerenone Post–Acute Myocardial Infarction Heart Failure Efficacy and Survival Study (EPHESUS) trial demonstrated decreased all-cause mortality and hospitalization for cardiovascular cause in patients with LV dysfunction post-MI. Then the EMPHASIS-HF replicated the results of RALES in NYHA class 2 patients. All of these studies have decreased EF as an inclusion criteria. There is no prior large study of clinical outcomes for aldosterone antagonists in HFpEF patients.
TOPCAT randomized 3445 patients in 6 countries (the USA, Canada, Brazil, Argentina, Russia, and Georgia) as above. 1678 of those were from one of the two Eastern European centers. 71.5% of patients were randomized via the hospitalization inclusion criteria rather than via elevated BNP level. The primary outcome was not significantly different between the two groups (18.6% in the spironolactone group vs 20.4% in placebo, P=0.14). Of the events observed satisfying the primary outcome, about two thirds of these were hospitalization rather than death or cardiac arrest, and when you break out hospitalization itself as an outcome there actually was a benefit to spironolactone, with a hazard ration of 0.83 (P=0.04). So a negative trial, but with some evidence of a signal of benefit in the secondary outcomes. But then things get weird quickly.
Patients in Russia and Georgia had events satisfying the primary outcome at a rate of only 8.4% in the placebo group, compared to 31.8% at the other centers. We can speculate as to why this could be, and this is one supplementary appendix you will actually have fun reading — start at page 20 and watch as the authors work to establish the idea that Russian and Georgian patients may have had very different levels of illness at time of inclusion, and to build a case that the patients recruited at these centers were not representative of the disease state they were trying to investigate. The study authors actually published a separate paper just to explore this possibility, and their argument goes roughly like this: Russian/Georgian patients were more likely to be included on the basis of hospitalization for heart failure, whereas the split between hospitalization and BNP was more even in the Americas. Meanwhile, and patients included via BNP at all centers were more likely to be sicker in a number of important ways, from being older (73.4 vs 67.0) to being more likely to have CKD, A fib, or prior PCI or CABG. Combining this with the fact that the event rate in the placebo group for Russian/Georgian patients was far lower than in previous trials of HFpEF and the fact that the stratum included by BNP levels from all centers actually did demonstrate decreased rates of the primary outcome (HR 0.65, P=0.003) and you start to wonder if the Russian patients recruited via hospitalization had HFpEF in the sense in which we usually use that term. For their part, the TOPCAT investigators advance the theory that hospitalization is used differently in these countries than it is in the Americas, and thus that this criteria opened the study to a different population than was intended.
So, the story goes, patients from Russia and Georgia, by and large were recruited on the basis of recent hospitalization, and because hospitalization is used differently in these centers than in the Americas, these patients were not as sick and thus did not receive a benefit from a therapy that was effective in patients recruited via another objective measure. A reasonably good argument. But if we’re building a case against the validity of the Eastern European patient data, the new metabolite data is the smoking gun.
You see, beyond the different event rates and comorbidities, the investigators also note one more unexplained thing — patients from the Eastern European regions had less increase in serum potassium and serum creatinine to a given dose of spironolactone. If you can read that sentence without wondering “Were they actually taking it?” then… well, I just don’t believe that you can. And neither did the TOPCAT investigators.
So those who are taking it at all have similar levels regardless of region, and about 30% of people in Russia had no detectable concentration. Ok, I’m sold — more people weren’t taking it in the Eastern Europe than in the Americas. But does the presence of this metabolite actually track with effects of the drug?
Pack it in folks —we’re done here. Unless Eastern European patients are immune to spironolactone, they weren’t taking their meds with enough frequency to experience any physiologic effect related to the drug.
So, to sum up: a significant proportion of patients in TOPCAT had event rates that were totally at odds with what we would expect from the disease process they supposedly had, were more likely to be recruited through criteria that were more subjective (hospitalization for heart failure vs BNP concentration), and had much lower detectable levels of drug metabolites than their American counterparts controlling for patient-reported dose, with corresponding physiologic effects (less hyperkalemia, less increase in creatinine). Oh, and if you exclude these patients from the initial analysis, TOPCAT is a positive trial.
So how does this change my practice? Well, I decided to buy into TOPCAT fairly early on, so I’ve been doing spironolactone for these patients for years without great evidence to back me up. When PATHWAY-2 demonstrated that spironolactone was a more effective fourth line agent for hypertension than alpha or beta blockers with less hyperkalemia than people feared, that really sealed the deal for me. I can’t say we have level one evidence for it — unfortunately for all the reasons in the world to throw these patients out, we just can’t pretend that they weren’t part of the study, and can’t exclude them post-hoc without tainting our results But I think it is reasonable to believe on the basis of what we know about TOPCAT that spironolactone may have a mortality benefit in HFpEF. I look forward to another trial to prove this, but I’m not holding my breath. There are no aldosterone antagonists still on patent, after all, so there’s no drug company lining up to finance the trial. In the meantime, I use spironolactone in HFpEF, and you should consider doing the same.
But what does this imply for future international trials? Well, I suspect we will see closer oversight and less room for subjective error. More direct and ongoing measures of medication adherence, stricter adjudication of inclusion criteria and events, that sort of thing. International trials are the way of the future, to be sure — they allow us to get more patients enrolled, and lend credence to the generalizability of findings. But doing research in countries with their own medical and scientific cultures has its dangers, and TOPCAT is a warning to those who would follow in its footsteps.
Read the original article here.
Tomorrow on IM HEAT
Resident ID expert Risa Fuller talks about new monoclonal antibodies for prevention of C diff recurrence! Because you weren’t yet being forced to remember enough fake words ending in “-mab.” Get excited!