A Respectable Result? Or Do We Need More Closure?

I believe it was my brother who began a sermon with an image of hell. In this image, hell was described as existing eternally in a room with tables filled with sumptuous food and drink, but with each individual shackled to the tables and unable to reach the food to enjoy it. Heaven was described as the same room, but with each individual able to help one another to enjoy the food (or something along those lines). I was reminded of this image of hell when I listened to the results of the RESPECT randomized trial. It is full of some compelling data regarding benefit of PFO closure to prevent recurrent stroke, but because the primary endpoint had a P-value of 0.08, I am anxious that we will not be able to make use of that delicious data. As an active participant in RESPECT and a proctor for closures of atrial septal defects for St. Jude, the sponsor of the RESPECT study, I am of course inclined to want there to be data supporting what I have been doing for the last decade – closing PFOs to prevent recurrent strokes. But I have consistently maintained a sense of equipoise regarding the role of PFO closure.

In brief, the RESPECT study enrolled patients with cryptogenic stroke and a PFO to optimal medical therapy or to PFO closure with the Amplatzer PFO occluder device. The primary outcome was freedom from stroke over the duration of the study. All inclusion and endpoint events were imaging documented cerebrovascular accidents – no TIAs. Due to an anticipated low event rate, the study continued to enroll until there were a total of 25 neurologic events in the study, which took 8 years to achieve. 

By intention-to-treat analysis there were 9 strokes in the PFO arm, versus 16 in the medical management arm, which was associated with a P-value of approximately 0.08. However, of the 9 strokes in the PFO arm, 3 occurred in patients who did not receive a device, either because of patient refusal after randomization, or the stroke occurred between randomization and the procedure. One of the strokes occurred in a patient with a large residual shunt who was subsequently found to have a sinus venosus atrial septal defect which was closed after the endpoint stroke. Three of the other strokes in the device arm were lacunar strokes, suggesting there was a non-PFO mechanism for the endpoint stroke. In fact the imaging size of the strokes in the medical management arm were significantly larger than those in the PFO closure arm. As was presented, if one analyzes the study either by what treatment they actually received, or by per-protocol rather than intention to treat, P-values less than 0.05 were seen. Furthermore, the curves appear to diverge, suggesting that there is greater benefit with time to having the PFO closed, which is extremely relevant to the young patients so common in the PFO/cryptogenic stroke population. Thus, although a primary endpoint of 0.08 does not allow for what we would conventionally consider definitive acceptance of the finding of this study, my sense of equipoise is certainly shaken, and I suspect that even the harshest critic might reconsider their rejection of the findings if they themselves found themselves having suffered a cryptogenic stroke in the face of a PFO. After all, even using the intent-to-treat population, the incidence of stroke was reduced by about 50%, with only 24 patients needed to treat to prevent one stroke over 5 years. As an added benefit, PFO closure in RESPECT was extremely safe, so it seems the harm of having a PFO closed after a cryptogenic stroke seems primarily to be economic. 

And now for the hell part I alluded to above. Although I would like to see a P-value of 0.05 to have “ultimately” convincing evidence, I am skeptical that we will be able to see such a future study. There is an ongoing study, REDUCE, which might provide us additional information (although it is using a different device), but I wonder how difficult it will be to convince patients to undergo randomization, once they are told the results of RESPECT, even when the finding is described as indeterminate. The major question is how will the FDA respond? If the FDA holds to the P-value of the primary endpoint and denies approval of any device with a label for PFO closure, then PFO closure may become less and less possible to perform, as ever more insurance companies refuse to pay for the use of off-label devices (ASD occluder devices). This lack of labeling could therefore effectively destroy a treatment option which at the worst looks extremely promising. However, the dilemma is understandable, because approving device for PFO closure could open up floodgates, with significant economic implications. It additionally has the risk to the FDA of establishing some precedence by accepting trial data that did not meet the traditional 0.05 P-value. 

At the end of all this one is left with the “sniff” test: who among us thinks that the failure to achieve the P-value of 0.05 is sufficient to abandon this therapy in patients with stroke versus who among us thinks that the proof-of-concept was supported by these data? For those us who can smell the delicious food on the table in front of us, we need to acknowledge that we will not be able to partake unless we are helped by others because of our commitment to evidence-based medicine, and the shackles which we willingly accept to avoid over-reaching in our desire to understand what is the best way to care for our patients. We will need others (the neurology community, the FDA and their advisors in particular, and even the insurance industry) to conscientiously apply their own individual sniff test to these data. As interventional cardiologists we are at their mercy, perhaps that is the way it should be. 

Dr. Goldberg is the Director of Cardiac Catheterization Laboratory at the University of Washington Medical Center in Seattle, Washington where he is a Clinical Associate Professor of Medicine. He is also the Chief Clinical Officer of Cardiac Dimensions, Inc. He is a fellow in the Society of Cardiovascular Angiography and Intervention, as well as in the American College of Cardiology. Email: stevgold@u.washington.edu