Two More Cardiology Highlights from 2023

Randomized controlled trials had a great showing last year.

Regular readers know how much I love the randomized controlled trial. Let’s do two topics on the RCT that I covered in my Top Ten Stories of 2023 piece.

The Case of Extracorporeal Life Support or ECLS

Extracorporeal membrane oxygenation (ECMO) may be the highest-level supportive care we can offer. It is, basically, a heart-lung machine. It’s used to support the sickest patients.

Two trials published this year provided clarity on whether it is useful.

One scenario that doctors considered dire enough to use ECLS occurs when CPR does not work to restore circulation.

In January, Dutch investigators published results of the INCEPTION trial, in which extracorporeal cardiopulmonary resuscitation (CPR) or conventional CPR (standard advanced cardiac life support) were compared.

The trial enrolled 160 patients who had out-of-hospital cardiac arrest, due to a ventricular arrhythmia, and initial bystander CPR without return of spontaneous circulation within 15 minutes. They found no significant improvement in the primary endpoint of survival with a favorable neurologic outcome. (odds ratio, 1.4; 95% CI, 0.5-3.5; P = .52).

Another dire scenario that doctors thought ECLS might work for was in patients who have acute MI that is large enough to cause cardiogenic shock. Cardiogenic shock simply means that the heart is not pumping enough blood to support organ function, such as brain and kidneys. A perfect place for heart-lung support, right?

In the ECLS-SHOCK trial, a German team studied the use of extracorporeal life support (ECLS) in the treatment of acute infarct–related cardiogenic shock.

Slightly more than 400 patients were randomly assigned to ECLS plus medical therapy vs medical therapy alone. The primary outcome of all-cause death occurred in 48% and 49% of the ECLS and control group, respectively (RR, 0.98; 95% CI, 0.80-1.19; P = .81). Bleeding complications occurred 2.4-fold more often in the ECLS group, and peripheral vascular complications were nearly threefold more common in the ECLS group.

I was drawn to the ECLS story because it demonstrates the power of RCTs over plausible rational thinking. That is…it’s totally rational to think putting people on what is essentially a heart-lung machine will save at least some lives.

These two RCTs show that it does not. Proponents of ECLS might say, and in fact, even did during the medical meeting that ECLS-SHOCK was presented, that ECLS can work in the “right” patients. My answer to this is always the same: when it comes to invasive, risky, and costly interventions, the onus is on proponents to show us which patients these are—with an RCT.

---

Another RCT-related story from 2023 involved attempts to simulate randomized trials with observational data. If doable, this would be a huge advance, because a) RCTs are complicated and costly, and b) there are oodles of available data to analyze. I consider the ability to reliably simulate RCT data from real-world data a Nobel Prize-level achievement.

I subtitled one of my choices for stories of the year:

The Sobering Results of the RCT Duplicate Project

The power of an RCT is that randomization, not a clinician, chooses treatment. This balances both known and unknown baseline characteristics, reduces bias, and allows clinicians to make causal inferences.

Real-world evidence, such as that in healthcare databases, is plentiful. But none of it is randomized, which makes causal inference much more difficult.

The goal of the RCT Duplicate initiative is to learn to what extent real-world evidence could be used to provide RCT-level causal inference.

JAMA published the results of an attempt to emulate 32 RCTs from insurance claims data. The group of experts from Harvard Medical School first chose trials on their feasibility to be emulated. This required the database and the trials to include similar metrics, such as treatments, comparators, and outcomes. They then quantified the degree of agreement in the trial emulation vs the actual RCT.

The results were mixed. The upside was that some of the trials could be emulated closely with nonrandomized data. Three of the DOAC vs vitamin K antagonist (VKA) trials (RE-LY, ROCKET AF, ARISTOTLE) were closely emulated using real-world evidence.

But many trials did not emulate. For instance, the EINSTEIN-PE trial of rivaroxaban vs VKA yielded a nonsignificant hazard ratio of 1.12 while the real-world-evidence trial emulation found a highly significant HR of 0.67 (95% CI, 0.55-0.80).

Results went in the opposite direction with PARADIGM-HF, an RCT of sacubitril/valsartan vs enalapril in patients with heart failure. The RCT reported a hazard ratio of 0.80 (95% CI, 0.73-0.87) but the trial emulation attempt found no significant difference, with an HR of 1.02 (95% CI, 0.97-1.14).

The authors were honest in their conclusions. They wrote that real-world evidence can reach similar conclusions as RCTs, but this may be difficult to achieve.

I highlight this paper and initiative because of its sobering conclusions regarding causal inference from nonrandomized data.

That is, in this best-case scenario (experts in trial emulation using selected trials), real-world evidence did not consistently and reliably emulate an RCT. My understanding, though, is that these experts will keep at it.

---

Thanks for reading. And thanks to the shocking number of people who support this Substack. Thank you x 1000.

In other news, we are off to a great start over at Cardiology Trials. We’ve posted four of the most important trials in acute coronary syndrome. And we’ve added voice over for those who like to listen rather than read.

On Sensible Medicine, I am back to the Study of the Week. The last one critiques a dubious study on hearing aids.

I’ve also discovered a great Substack from Computer Science professor Ben Recht called Arg Min.

JMM