Meet the Trialist: The EMPACT-MI trial findings

Medicom spoke with Prof. Javed Butler (President of the Baylor Scott and White Research Institute in Dallas, and distinguished Professor of Medicine at the University of Mississippi, USA), about the late-breaking EMPACT-MI trial (NCT04509674) results he presented at the recent ACC 2024 Scientific Session, held in Atlanta, USA [1,2].

The double-blind, placebo-controlled EMPACT-MI trial evaluated patients receiving empagliflozin (n=3,260) or placebo (n=3,262) for cardiovascular (CV) outcomes after acute myocardial infarction (MI). Over 17.9 months, the primary endpoint—hospitalisation for HF (HF) or death from any cause—occurred in 8.2% of the empagliflozin group and 9.1% of the placebo group, showing no significant difference (HR 0.90; 95% CI 0.76–1.06; P=0.21). However, empagliflozin was associated with reduced HF hospitalisations compared with placebo (3.6% vs 4.7%, respectively). Overall mortality rates were similar between the 2 groups. These findings highlight that while empagliflozin may reduce the risk of HF hospitalisation, it does not significantly impact overall survival post-MI, suggesting a targeted but limited role in the management of such patients.

What is your interpretation of these rather surprising results of no impact on overall survival versus a clear impact on HF hospitalisation?

“Some risks were taken in the trial design that didn't necessarily pan out. Let me go into more detail about the machinery of how the trial was designed, which can probably explain things. There is a lot of pressure to decrease the cost and the burden of the trials on the patients and on the sites, prioritising pragmatic clinical trials.

As part of the pragmatic clinical trial, our inclusion-exclusion criteria were very general and everything was remote in follow-up. The correlation among the centres was not very good in 2 places: the cause of death, and outpatient HF. There is much pressure to reduce HF admissions and people are increasingly diagnosing and treating patients in the outpatient HF setting. Because we did not have a central adjudication committee as part of a streamlined trial, we went for all-cause mortality, not CV mortality.

However, that choice comes with a cost; if you have 100 people who die post-MI, you can expect about 75–80 of them will die of CV causes. And of those early post-MI mortalities, a lot of things happen—e.g. ventricular septal ruptures, recurrent MIs, or stent thrombosis—that SGLT2 inhibitors can't do anything about. Then there are the other 20–25% who will not die of CV causes but were counted as events here nevertheless; there's no chance that SGLT2 inhibitors will improve their outcome, right? Another problem was that the follow-up was only 18 months, meaning that actual follow-up after HF development was shy of a year, which would limit mortality events to only early events.

The second component beyond all-cause mortality in the endpoint was hospitalisation due to HF. When we designed the trial, we predicted only 10–15% of the HF burden would be in the outpatient setting, and that it would not greatly influence the outcome. We could not have predicted then that the COVID-19 pandemic would intervene and drive the rates of outpatient care well above our anticipated background, making that particular endpoint very hard to realise. On top of that, the 2 regions that were very active in the trial, Israel and Ukraine, entered wars.

This has affected the results in 2 ways; firstly, you don't have HF outpatient diagnosis in the endpoint, but it also changes the ratio in the primary endpoint. At the beginning of the trial, we anticipated that if 100 people were to be in your endpoint of the trial, about 40 of those events would be deaths and 60 would be hospitalisations. Here we had the reverse. We had more deaths than HF hospitalisations. We believe that the specific conditions of this trial led to a systematic under-calculation of HF patients.

We live in a P-value-oriented world, and a significant P-value was not reached. However, if we look at the HF results, we see identical results to what we have seen in diabetes, chronic kidney disease, HFrEF, and HFpEF. In short, no effect on mortality but a 23% reduction in HF time to the first event and a 33% reduction in total HF events. Interestingly enough, although not part of the primary endpoint, and only an exploratory analysis, if you do include outpatient HF in the endpoint, even all-cause mortality and HF hospitalisation total are positive. Then, the HF signal is 23% relative risk reduction time to first hospitalisation, 33% total, and 37% for inpatient and outpatient combined."

It sounds like there's a signal, but you can't draw that hard conclusion based on the trial design?

"Absolutely. To give you one more result; we wanted to pressure test this outpatient test. We, therefore, analysed all patients who were not on HF therapy at the time of discharge and then tracked their medicine use post-discharge (ARNI use, any RAS inhibitor use, MRA use, or diuretics) as a surrogate marker for outpatient HF. We identified that post-discharge outpatients HF were much more common in the placebo arm than the treatment arm. So, if you're a strict statistical frequentist, you would say that the trial is negative. However, if you're a Bayesian statistician, you would say that the totality of evidence suggests that you significantly reduce the risk of HF development post-acute myocardial infarction with empagliflozin."

What about discerning between STEMI and non-STEMI? Can we learn anything about who to treat with SGLT2 inhibitors here?

"This is an unbelievably important question. Currently, STEMI patients have excellent prognosis after revascularisation. Looking at real-world data, the non-STEMI to STEMI ratio is 3:1; 75% of the MI are non-STEMI high-risk patients, and one-fourth are STEMI. We see in both PARADISE-MI and EMPACT-MI the exact opposite [3]. Most of the patients are lower-risk STEMI patients with fewer non-STEMI patients.

In EMPACT-MI, the annualised event risk in STEMI patients was about 8%, and 12% in non-STEMI patients. It's the non-STEMI patient who is at a higher risk, yet the benefit of empagliflozin was higher too; the point estimate was 0. 77 in non-STEMI patients. I think that our data demonstrate that the non-STEMI patient can benefit from early use of non-SGLT2 inhibitors."

What’s your advice to fellow physicians about how and when to use SGLT2 inhibitors in this setting?

"There was an unbelievably good safety signal in EMPACT-MI. Total AEs, AEs leading to discontinuation, contrast-induced nephropathy: none of them were signals or different between the 2 arms. The way I look at this, if somebody comes in with acute MI and they already have an indication for an SGLT2 inhibitor—if your patient has HF, diabetes, or chronic kidney disease—physicians should feel very comfortable starting it during admission because of the safety profile. Now suppose you have an acute MI patient who has none of the other indications; they don't have HF, they don't have diabetes, they don't have CKD. If that patient has non-STEMI, I would be very much inclined to start with SGLT2 inhibition early. However, if they come in with a STEMI and they get revascularised, then I would watch that patient for a few days. If their LV is still depressed, if they're still congested, then I would start an SGLT2 inhibitor. If their stunned myocardium reverses, and if their EF is totally normalised, and they are feeling great, then maybe it's okay just to watch that patient."

1. 
   
   1. Butler J, et al. Empagliflozin After Acute Myocardial Infarction: Results Of The EMPACT-MI Trial. ACC 2024 Scientific Session, 6–8 April, Atlanta, USA.
   2. Butler J et al. N Engl J Med 2024;390:1455-1466.
   3. Mann DL et al. Am Coll Cardiol2024;83:904-14.

Copyright ©2024 Medicom Medical Publishers



Posted on 15 May 202413 June 2024