乌鸦传媒

This article is a collaboration between 乌鸦传媒 and:

The negative results of three studies should have closed the book on patent foramen ovale (PFO) closure in cryptogenic stroke, but somehow the book is still being picked up.

In the RESPECT trial, out of the 980 patients randomized to closure of the PFO or medical therapy and followed for an average of approximately 2.5 years, those in the treatment arm had a nonsignificant 51% reduced risk of recurrent stroke in the intention-to-treat analysis (HR 0.49, 95% CI 0.22 to 1.11, P=0.08), reported John D. Carroll, MD, from the University of Colorado Denver, and colleagues.

In the PC Trial, of the 414 patients randomized to treatment or medical therapy and followed for an average of approximately 4 years, closing the PFO resulted in a nonsignificant 37% reduced risk of recurrent stroke (HR 0.63, 95% CI 0.24 to 1.62, P=0.34), according to Bernhard Meier, MD, from Bern University Hospital in Switzerland, and colleagues.

Both studies, which used the Amplatzer PFO Occluder, were published March 21 in the New England Journal of Medicine and were initially reported at the 2012 Transcatheter Cardiovascular Therapeutics meeting.

An earlier trial called CLOSURE I had similar rates of the primary endpoint between treatment with the STARFlex Septal Closure System and medical therapy, also leading to a nonsignificant finding (5.5% and 6.8%, respectively, P=0.37).

"Unlike the results of CLOSURE I, however, the results of the PC Trial and RESPECT have encouraged those who believe that closure of a patent foramen ovale might be an effective therapeutic option for this common clinical problem," Steven R. Messé, MD, from the Hospital of the University of Pennsylvania in Philadelphia, and David M. Kent, MD, from Tufts University in Boston, wrote in an accompanying editorial.

"Advocates of closure will surely focus on the substantial relative effect size of the point estimates in both trials, the significance of the per-protocol and as-treated analyses in RESPECT, the arbitrariness of the conventional P-value threshold of 0.05, and various other intriguing signals," Messé and Kent pointed out.

Skeptics will "focus on the fact that three consecutive trials have failed to reject the null hypothesis in their primary intention-to-treat analyses, on the imprecision in the effect estimates, and on potential sources of bias, which included uneven dropout rates and unblinded (and imbalanced) referral for adjudication of endpoints," they added.

"Thus, we are left for the moment to make decisions under conditions of uncertainty," Messé and Kent concluded.

Low Statistical Power

The overall rate of events were low in both RESPECT and the PC Trial.

In RESPECT -- a 25-event-driven trial -- nine patients in the treatment arm met the primary endpoint of a composite of recurrent nonfatal ischemic stroke, fatal ischemic stroke, or early death after randomization, compared with 16 controls.

In the PC Trial, which began 14 years ago and had no formal rules for stopping, the primary endpoint -- a composite of death, nonfatal stroke, transient ischemic attack, or peripheral embolism -- occurred in seven and 11 patients in the treatment and control arms, respectively.

"[L]ow statistical power can lead to spurious, misleading results and can make trials more sensitive to potential bias," the editorialists Messé and Kent pointed out. "Misclassification of even one or two events can have dramatic effects on the P values in trials with low outcome rates."

In fact, during a discussion at the TCT meeting, Sanjay Kaul, MD, of Cedars-Sinai Medical Center in Los Angeles, said of RESPECT that with one fewer event in the closure arm, there would be a P value of 0.05. And with one fewer event in the closure arm and one more event in the control arm, there'd be a P value of 0.03. "We have to respect the fragility of the data," he said.

In the RESPECT trial, Carroll and colleagues outlined results from secondary analyses.

For example, in the pre-specified per-protocol analysis, there was a significant between-group difference in the rate of recurrent stroke that favored treatment, with six versus 14 events (HR 0.37, 95% CI 0.14 to 0.96, P=0.03).

They also found a significant difference in recurrent stroke in the as-treated cohort, which again favored treatment with five versus 16 events (HR 0.27, 95% CI 0.10 to 0.75, P=0.007).

However, Carroll and colleagues said that these data should be interpreted "with caution, owing to potential bias arising from nonrandom factors that may have accounted for nonadherence to the protocol."

But they added that there is some merit to these other analyses, because three of the nine primary events of recurrent ischemic stroke in the closure group of the intention-to-treat cohort "occurred in patients who did not have a device in place at the time of the recurrent stroke."

Furthermore in RESPECT, event rate point estimates favored treatment over medical therapy as time progressed, going from 1.3% versus 1.7% at 1 year to 2.2% versus 6.4% at 5 years.

One analysis suggested that PFO closure was more beneficial in patients with a grade 3 right-to-left shunt (P=0.01) and in those with an atrial septal aneurysm (P=0.02).

The closure group also had fewer moderate, large, or massive strokes (14% versus 69%, P=0.06).

Various analyses of the PC Trial data, however, did not reveal any significance, including when they evaluated the per-protocol results, the individual components of the primary endpoint, or the rate of stroke employing the definition of stroke used in RESPECT. Meier and colleagues found a trend that favored treatment in patients with an atrial septal aneurysm.

Both studies had no differences in adverse event rates between the PFO group and controls. The mean age of patients in the PC Trial was 44 and in RESPECT, it was 46.

Other Limitations

The PC Trial was limited by the inclusion of death in the primary composite endpoint, which would count deaths that were not specific to the studied condition.

The inclusion of transient ischemic attack could have diluted effects because it is a less clear-cut endpoint, the authors said, noting the difference in the estimated hazard ratios for stroke (0.20) and TIA (0.71).

Also, the long recruitment time was a limitation, as it could limit the generalizability of the findings.

Also a limitation was the lower-than-expected patient retention rate and the fact that more events were discounted in the medical therapy arm than in treated patients.

The RESPECT trial was limited by the difference in the dropout rate (17.2% in controls versus 9.2% in the closure group). Also, the results could be biased because of the preference for high-risk patients to be treated outside the trial, the researchers said.

From the American Heart Association: