Saturday, 15 August 2015

Fixing a hole

Migraine is common, affecting millions of people worldwide. A patent foramen ovale (PFO – a ‘hole in the heart’ that lets blood cross from the right heart to the left) is common as well, present in about 30% of people. When cardiologists started surgically closing PFOs, they noticed that many patients with migraine got better. As with the discovery of any association in medicine, theories of a causal link soon followed, and doctors started treating migraine by closing the hole in the heart; before properly testing it, of course.

An association between migraine and PFO has been shown in some studies, in that people with a PFO are more likely to have migraines. And when patients underwent PFO closure (for other reasons) doctors noticed that their migraines were often better. When faced with things like that, our brains are quick to kick in – too quick – jumping straight to the conclusion that the association represents causation, and then developing theories as to how (not ‘if’) migraines could be caused by a PFO. Maybe something from the right side of the heart (the venous circulation) is getting through the PFO and going to the brain (arterial circulation)? Migraine is associated with brain changes on MRI that might be caused by emboli –could that be the proof? Or maybe it is some metabolite that is leaking across the PFO?

But migraine is episodic, not constant like blood flow through a PFO. And they are usually well localised to one part of the head, whereas emboli are scattered throughout the brain. And PFO leakage gets worse with age whereas migraine decreases with age. And many people with migraine do not have PFO and vice versa. To me, this means that the people developing these causal theories hadn’t thought it through.

But that didn’t stop people from inserting devices in the heart to close the PFO holes. And in ‘observational’ studies, they saw just what they expected to see: the migraines got better. Was this a case of ‘believing is seeing’?

Later, a blinded placebo (sham) trial (an experimental study) was done comparing PFO closure (using a device inserted through the groin) to sham closure (just an incision over the groin). The results showed that patients who had their PFO closed had fairly good improvements in their migraine measures, just like in the previous observational studies. Unfortunately, the improvement seen was not much better than that seen in the sham group.

The authors tried to make it appear effective, by doing a secondary analysis excluding some patients in the PFO closure group that did very badly (lots of migraines). Then they noticed that the PFO closure group did a little better. Yes, that’s right: if you remove the really bad patients from group A, and then recalculate the scores on the remaining patients, treatment A suddenly looks a lot better. I wish I could do that with my surgical results: “Hey, if you ignore all my patients with very bad outcomes, I look like a pretty good surgeon!”

This editorial (here) in the same issue noted something interesting. When the results of the sham study were initially presented to a major cardiology conference (prior to publication) they were positive, showing a significant reduction in one outcome: the number of days with migraine in the group that had their PFO closed (although the primary outcome and the overall cure rate were still no better). So why did the published study showed no significant difference? Maybe they got excited when they stumbled on some positive results and only later realised that they had forgotten something (like including the patients in the treatment group with bad results)?

The real test of a trial is the primary endpoint – the outcome they put all their money on. They expected 40% of the PFO closure group to have no more migraines, compared to 15% in the sham group – and based these expectations on the results seen in the observational studies. In the end, only 4% had no migraines in the follow up period – exactly the same in both groups.
And as usual, things didn’t go perfectly – just like in life. Some patients never had much of a hole to start with, and some hearts still leaked after the closure was performed, but even allowing for that, it still didn’t make any difference to the migraines.

I think the theory of a causal connection is still alive – those types of theories are hard to kill because they are what we all want to hear. Hopefully not many people are having this procedure done for their migraine. Not just because it probably doesn’t work and is expensive, but because the placebo study also showed a higher than expected rate of complications in the group that had the device inserted (to close the PFO). Things like getting the device stuck in wrong part of the heart, having it float off into the lungs, cardiac tamponade, and pericardial effusion - complications that the authors referred to as “transient”.


Addit: a good summary from 2014 can be found here.

2 comments:

  1. I don't understand how this trial could be double blinded. The surgeons would know which procedure they did.

    ReplyDelete
    Replies
    1. This is why the term "double-blinded" is not recommended - it doesn't tell you which two groups were blinded. Here, it is likely that the patient and the follow up assessor were blinded.

      Delete