Previously, I posted the results of the RESPECT study, which tested PFO closure vs. medical therapy. Briefly, the primary analysis failed to show PFO closure to be beneficial, but the primary analysis was deemed “invalid”. So they did a secondary analysis, which also failed to show benefit. Tertiary and quaternary analyses did suggest benefit, however, so now we’re supposed to get all excited.
Yesterday we received the results of the PC trial (data here, but you need to fill out a free registration), and the story is largely the same. 414 subjects were randomized 1:1. The primary outcome was a composite of death, non-fatal stroke, TIA, or peripheral embolism (composite outcomes making it easier to imply benefit). After a mean follow-up of 4 years, there were 7 events in the closure group (3.4%) and 11 in the medical group (5.2%). This difference didn’t reach statistical significance. A bunch of subgroup analyses failed to show anything either, except to contradict the RESPECT finding that having an atrial septal aneurysm predicted greater response to PFO closure.
Of note, the annualized event rate in the medical arm was only 1.3%/year–a much lower rate than was relied upon in calculating the sample size. Thus, the study was underpowered to detect the
hoped-for hypothesized 66% relative risk reduction. The low event rate supports my own speculation about the causal role for PFO in cryptogenic stroke:
I think that having a PFO probably does slightly increase one’s stroke risk. The magnitude of this increase is very small, but with ~ 75 million Americans harboring one, a few will manifest with stroke. Then you follow a few hundred such patients for a few years and what happens to them? Not much.
Here’s another example: The ISUIA study suggested that the rupture risk for small anterior circulation aneurysms is 0%. However, all of us in the stroke field have seen plenty of subarachnoid hemorrhages from such aneurysms. Why? 2.5% of people have an occult, unruptured, cerebral aneurysm. Even if the rupture risk is tiny, some of those several million people harboring one will hemorrhage. But that doesn’t mean that every patient who’s found to have such a lesion should have it treated.
A final analogy. The probability of hitting the lottery is very low, but with millions of people playing, there are some winners. Now suppose we enroll that handful of lottery winners in a registry and tally how many of them win a second time. Probably none will. Indeed, when I counsel cryptogenic stroke patients about their recurrence risk, I often invoke the analogy of the “anti-lottery”.
In the prior PFO post, I suggested that we might get some input from a better clinical trialist than I. So here’s Dr. Patrick Lyden of Ceders-Sinai, one of the top strokeologists in the world* (emphasis added):
A “non-statistically significant reduction” is also known as “failure to achieve the primary endpoint” , or better known as “no evidence of benefit”. Those confidence intervals strongly suggest no benefit with that device.
I concur with . . . the need to continue and complete [the REDUCE trial, for which UW will be a site]. Very importantly, the failure of RESPECT and PC argue powerfully against any use of devices outside of a controlled clinical trial.
* Disclosure: He’s also one of my mentors.