Beta-Blockers Post-MI? No for Me, Even After ABYSS

The ABYSS trial found that interruption of beta-blocker therapy in patients after myocardial infarction (MI) was not noninferior to continuing the drugs. 
I will argue why I think it is okay to stop beta-blockers after MI — despite this conclusion. The results of ABYSS are, in fact, similar to REDUCE-AMI, which compared beta-blocker use or nonuse immediately after MI, and found no difference in a composite endpoint of death or MI. 
Translation of the ABYSS trial results to patient care is a case where we must look past the paper’s abstract and conclusions. The key problem is the authors’ choice of primary endpoint, which obscures the correct clinical answer. 
The ABYSS Trial
ABYSS investigators randomly assigned nearly 3700 patients who had MI and were prescribed a beta-blocker to either continue (control arm) or stop (active arm) the drug at one year. 
Patients had to have a left ventricular ejection fraction (LVEF) at least 40%; the median was 60%. 
The composite primary endpoint included death, MI, stroke, or hospitalization for any cardiovascular reason. ABYSS authors chose a noninferiority design. The assumption must have been that the interruption arm offered an easier option for patients — eg, fewer pills. 
Over 3 years, a primary endpoint occurred in 23.8% of the interruption group vs 21.1% in the continuation group. 
In ABYSS, the noninferiority margin was set at a 3% absolute risk increase. The 2.7% absolute risk increase had an upper bound of the 95% CI (worst case) of 5.5% leading to the not noninferior conclusion (5.5% exceeds the noninferiority margins). 
More simply stated, the primary outcome event rate was higher in the interruption arm. 
Does This Mean We Should Continue Beta-Blockers in Post-MI Patients?
This led some to conclude that we should continue beta-blockers. I disagree. To properly interpret the ABYSS trial, you must consider trial procedures, components of the primary endpoint, and then compare ABYSS to REDUCE-AMI. 
It’s also reasonable to have extremely pessimistic prior beliefs about post-MI beta-blockade because the evidence establishing benefit comes from trials conducted before urgent revascularization became the standard therapy. 
ABYSS was a pragmatic open-label trial. The core problem with this design is that one of the components of the primary outcome (hospitalization for cardiovascular reasons) requires clinical judgment — and is therefore susceptible to bias, particularly in an open-label trial. 
This becomes apparent when we look at the components of the primary outcome in the two arms of the trial (interrupt vs continue): 
For death, the rates were 4.1 and 4.0%
For MI, the rates were 2.5 and 2.4%
For stroke, the rates were 1.0% in both arms
For CV hospitalization, the rates were 18.9% vs 16.6%
The higher rate CV hospitalization alone drove the results of ABYSS. Death, MI, and stroke rates were nearly identical. 
The most common reason for admission to the hospital in this category was for angiography. In fact, the rate of angiography was 2.3% higher in the interruption arm — identical to the rate increase in the CV hospitalization component of the primary endpoint. 
The results of ABYSS, therefore, were driven by higher rates of angiography in the interrupt arm. 
You need not imply malfeasance to speculate that patients who had their beta-blocker stopped might be treated differently regarding hospital admissions or angiography than those who stayed on beta-blockers. Researchers from Imperial College London called such a bias in unblinded trials “subtraction anxiety and faith healing.”
Had the ABYSS investigators chosen the simpler, less bias-prone endpoints of death, MI, or stroke, their results would have been the same as REDUCE-AMI. 
My Final Two Conclusions
I would conclude that interruption of beta-blockers at 1 year vs continuation in post-MI patients did not lead to an increase in death, MI, or stroke. 
ABYSS, therefore, is consistent with REDUCE-AMI. Taken together, along with the pessimistic priors, these are important findings because they allow us to stop a medicine and reduce the work of being a patient. 
My second conclusion concerns ways of knowing in medicine. I’ve long felt that randomized controlled trials (RCTs) are the best way to sort out causation. This idea led me to the believe that medicine should have more RCTs rather than follow expert opinion or therapeutic fashion. 
I’ve now modified my love of RCTs — a little. The ABYSS trial is yet another example of the need to be super careful with their design.
Something as seemingly simple as choosing what to measure can alter the way clinicians interpret and use the data. 
So, let’s have (slightly) more trials, but we should be really careful in their design. Slow and careful is the best way to practice medicine. And it’s surely the best way to do research as well. 
John Mandrola practices cardiac electrophysiology in Louisville, Kentucky, and is a writer and podcaster for Medscape. He espouses a conservative approach to medical practice. He participates in clinical research and writes often about the state of medical evidence.