Mesmerized by My Own Beat

On the history of anticoagulation and the role of randomized evidence in cardiology.

You might have noticed a pattern in last week’s blog history of the megatrials in cardiology. All the breakthrough trials I covered were essentially drug trials of anticoagulant interventions like aspirin, streptokinase, and heparin. This, of course, wasn’t just a coincidence.

Though randomized trials are often pitched as the ultimate arbiters of causality, you need to have a good reason to do one. You can’t (and don’t) just propose a random trial with a control group. Not only must there be considerable uncertainty about the outcome to get a trial approved, but there must also be some reason to do the trial in the first place. There has to be some intuition that leads some investigator to propose some intervention, and they have to have enough faith in that intervention to devote massive resources to proving it’s efficacious. Though randomized trials sit far above expert opinion in the hierarchy of clinical evidence, expert opinion necessarily must still drive the operation.

So what was the core hypothesis driving the series of anti-coagulant megatrials? Whether clots were a primary cause of heart attacks was a hot debate in cardiology well into the 1980s. The prominent theory for much of the 20th century posited that it was arterial hardening and valve narrowing that caused heart attacks. In this model, blood clots were caused by the heart attack and not the other way around. This felt consistent with the evidence, as most of those who died from heart attacks didn’t have blood clots in their autopsies.

It wasn’t until 1980 that a paradigm-shifting study showed that blocked arteries were overwhelmingly common in the early stages of heart attacks. The investigators corroborated earlier findings that the clots were less prevalent the following day. This suggested that some biological mechanism was naturally breaking down the clots during heart attacks, and suggested a mode of attack for treatment. Cardiologists spent a decade finding new and creative ways to break down blood clots in the heart and intervene as early as possible, leading to marked improvements in survival.

The large pragmatic drug trials were important for building practice. They did find a fairly robust suite of therapies for treating heart attacks. Could an individual cardiologist have determined a better regimen? Possibly. But the megatrial infrastructure of cardiology trials proved that, for the right clinical hypothesis, professional societies of trialists could maximize population outcomes to achieve a reasonable standard of care with 3-6 times lower mortality than the prior one.

Now, I also mentioned a trial that wasn’t about coagulation: CAST. As a reminder, the CAST trial studied the value of drugs that quelled irregular heartbeats in heart attack patients. The trial found that these therapies led to a major increase in mortality, and it ended the practice. But why were doctors convinced that this was a good idea in the first place?

I suppose it might sound reasonable to make a sick heart look “healthier” by messing with its rhythm. But what’s the model behind why that would be a good idea? I’m sure there is some grand rounds somewhere that lays out the full hypothesis, but I couldn’t find one, and the literature associated with the CAST trial doesn’t provide it. Instead, the trial report cites a bunch of correlational studies showing that arrhythmia is correlated with mortality in heart attack patients. The main driving hypothesis seems to be this statistical correlation. Expert opinion on anti-arrhythmia drugs was decidedly mixed before CAST, and that’s part of the reason the trial went forward.

It’s worth dwelling on one remarkable feature of the CAST trial. The investigators were allowed to compare against a placebo. Without the placebo arm, the bad practice of treating arrhythmias would not have been ended. It remains a heated debate in medical ethics as to when it’s acceptable to compare an intervention to a placebo rather than to the standard of care. There is a strong case for placebos. A cascade of RCTs has a great deal of path dependence. Every trial slightly adjusts the standard of care along a single axis. This is what optimizers call random search, and there is nothing to prevent it from ending up in a disadvantageous local minimum. Thus, trials against the standard of care might keep you in an iatrogenic region of medical policy. It is quite possible that the standard of care needs to be completely reimagined to achieve the next major therapeutic improvement. As evidence, this was exactly what was needed to force cardiology to shift to anticoagulant therapy. CAST remains a classic reminder that it is not just ethical but often imperative to test against placebos.

Though often held up as a poster child for the megatrial, the full history of cardiology looks an awful lot like the rest of science and engineering. There isn’t a single magic bullet that automatically led to improved therapies. There was a mixture of expert opinion and institutional debate. Small studies inspired huge randomized trials. The megatrials were guided by biological plausibility and mechanistic intuition. The RCT was clearly an important part of improving practice, but it was part of a complex curation of expertise, rather than a cut-and-dried gold standard.

However, some people stubbornly maintain that RCTs are the only pure way to get to the truth of what works. This belief led to one of the weirder moves in medical history that’s still with us today: The rise of the postmodern thinking of evidence-based medicine. This is the topic of my next post.