Study Shows Benefits of Impella, Despite Early Termination

Reed Miller

April 03, 2011

April 3, 2011 (New Orleans, Louisiana) Even though the PROTECT II trial of the Impella 2.5 circulatory assist device (Abiomed, Danvers, MA) was stopped early for futility, the trial's lead investigator told a standing-room-only audience here at the American College of Cardiology (ACC) 2011 Scientific Sessions that the per-protocol analysis of the trial shows that the device is superior to intra-aortic balloon pump (IABP) support for a subset of high-risk coronary intervention patients [1].

Principal investigator Dr William O'Neill (University of Miami, FL) announced the final results of both the intention-to-treat and per-protocol analyses of the study, which was stopped December 6, 2010 when a prespecified interim review found that the study as designed would not reach its end point. Despite that apparent failure, O'Neill said that "we've learned that the device provides superb hemodynamic support during high-risk interventions. It really allows the operator to do a more complete procedure. That translates into [fewer] late events."

"When we operate on these patients, we're often 'skating rapidly over thin ice.' We do brief balloon inflations, brief stent deployments, and minimize contrast load to get in and get out as soon as we can. This device gives us the luxury of being able to take our time and do a more complete and safe procedure," he said. "It will translate into a large increase in use for these high-risk patients."

The Impella 2.5 is designed to be inserted across the aortic valve through a single femoral site to pump blood from the left ventricle into the ascending aorta. PROTECT II was a randomized, multicenter trial comparing the Impella system with an IABP in patients requiring hemodynamic support during nonemergent high-risk PCI in an unprotected left main coronary or the last patent conduit and an LVEF under 35% or three-vessel disease and an LVEF over 30%. O'Neill said PROTECT II enrolled a higher-risk population than any previous randomized PCI trial. He noted that about two-thirds of the patients in the trial were formally deemed inoperable by surgeons, their SYNTAX score was around 30, and the average LVEF was about 24%.

The primary end point for PROTECT II was the composite rate of 10 major adverse events, including death, MI, stroke, and repeat revascularization 20 to 40 days after the procedure.

The trial began in late 2007 with a plan to enroll 654 patients at 50 centers. The study designers predicted that patients randomized to hemodynamic support by Impella would have a 20% adverse-event rate and that those randomized to IABP would have a 30% rate. However, the overall adverse-event rates were nonsignificantly worse in the IABP group than in the Impella group in the interim analysis.

A total of 447 patients were enrolled in the intention-to-treat study, but only 426 met all the criteria to be included in the per-protocol analysis. The per-protocol analysis showed that the Impella patients had 21% fewer major adverse events at 90 days than the IABP patients (40.8% vs 51.4%; p=0.029).

In the prespecified subgroup of patients treated without atherectomy (88% of patients in the study), Impella patients had significantly fewer adverse events than IABP patients at 30 days postprocedure (29.5% vs 42.4%; p=0.009) and 90 days postprocedure (35.9% vs 51.1%; p=0.003), for an overall difference in adverse events of 29%.

However, in the 12% of per-protocol study patients treated with atherectomy during their PCI, there was no overall statistical difference (p=0.316) in major adverse events at 90 days. In this subgroup, the Impella arm showed a significant increase in periprocedural creatine kinase (CK-MB), a myocardial enzyme linked to MI and higher mortality risk (p=0.030), but the Impella patients had fewer repeat revascularizations at 90 days (p=0.006).

O'Neill also pointed out that Impella's advantage over IABP improved during the course of the trial, suggesting a learning curve. The prespecified "roll-in subject" analysis showed that, when the first patients in both arms were excluded, Impella was significantly better at 90 days for the primary end point (n=307; p=0.027).

"Knowing what I know now, I would have insisted on a one-year follow-up for the patients, because these are very high-risk patients with poor ventricular function and multivessel coronary disease," O'Neill told heartwire . "One part of the story is getting them out of the hospital safely, but the next part is how they do in the short and intermediate term."

Operators Too Aggressive With Atherectomy With Impella

O'Neill explained that the study design assumed a lower rate of adverse events in both the IABP and Impella group than was actually seen in the trial and that the adverse events turned out to be especially high because the operators were too aggressive with rotational atherectomy when the patient was supported by Impella.

Overall, about 70% of patients treated with Impella and atherectomy had an adverse event, but only about 35% of patients treated with an IABP and atherectomy did. Unfortunately, there was no way to blind a study comparison of Impella and IABP, and "it would have been very difficult to protocolize atherectomy use--that was the major confounder. It's just a problem you get into with an unblinded trial," O'Neill told heartwire . "You really can't control a lot of the nuances of behavior in the cath lab. The physicians just do the best job they can with the high-risk patients."

O'Neill said that it is not entirely clear why the operators used atherectomy more frequently and more aggressively (ie, more and longer passes) with the Impella than with the IABP, but he suspects it was because the operators knew the Impella would provide better hemodynamic support. Atherectomy, which is used in 3% to 5% of PCI cases in the US, is known to trigger the release of CK-MB, which is linked to MI, he explained.

"But when it's apples to apples--Impella vs balloon pump without the confounder of atherectomy--there was a very significant decrease in the adverse event rates at 90 days. That's a very consistent and scientifically valid observation," O'Neill said.

However, O'Neill cautioned that "the message is not 'don't use atherectomy [with Impella].' The message is 'use it gingerly.' Use it like you think you have a balloon pump in there with very few passes and short-duration runs, and that will not drive up the event rates."

Doubts Remain About a "Negative Study"

Following O'Neill's presentation, Dr Ron Waksman (Washington Hospital Center, Washington, DC) stepped to the microphone to reiterate his long-held skepticism about the PROTECT II design. He cited his group's previous futility analysis of the study that predicted it would be underpowered to detect a difference in adverse event rates. "But maybe we were right for a different reason, which is actually the bottom line. With all of the conclusions and speculations on the atherectomy, with all due respect, this is, in my view, a negative study. I don't understand how we come to a conclusion based on this result." Waksman also suggested that the composite end point of the trial combined too many different adverse events to provide meaningful information. "We don't know what is driving the differences shown. . . . We cannot correlate, physiologically, any of those to the outcome presented."

O'Neill responded: "When we started this analysis, we basically just conjectured what the incidence rate of the end points would be based on at least 30 single-center, multicenter observational studies," he said. "At a minimum, this will be a landmark for these kinds of studies in the future," because, unlike those previous small studies, the adverse events in PROTECT II were independently adjudicated. He compared it to the BARI trial, which showed no difference overall between angioplasty and surgery but showed bypass surgery was better for diabetics in a prespecified subpopulation analysis. "In a prespecified analysis, with the confounding of a termination of the study prior to the complete enrollment, we do think we have a very strong message."

In an interview with heartwire , O'Neill dismissed Waksman's criticisms. "His major complaint is that we were never going to reach the [necessary] power because we projected that event rates were going to be too high, and that's based on who-knows-what-the-hell data he has, but in fact he has been proven wrong, because the event rates were higher than we projected. We projected that the balloon-pump group was going to have a 30% rate, and it ended up having a 50% rate," O'Neill explained. "He's absolutely, unequivocally wrong."

Dr Roxana Mehran (Mount Sinai Medical Center, New York, NY), the chair of the session in which O'Neill presented the PROTECT II data, told heartwire that "the [PROTECT II] data were presented in a very scientific manner, because they presented the intention-to-treat as well as the per-protocol treatment. There was a very important significant interaction in the way patients were treated when they were randomized to the Impella arm, and it was important to dissect that out. What this study shows you is that, when you do an open-label study like this, that the investigators will use their own bias to actually drive certain important differences in the procedure.

"Dr Waksman's comments are well taken and he should be congratulated for coming forward and speaking to the scientific way of looking at this--that this is a negative study because it was designed for superiority and it was stopped early," she added. "But I think that these kinds of studies inform us tremendously on how to treat very, very high-risk patients. Having access to these kinds of adjunct devices will be very important in the high-risk population."


O'Neill said he is working with other investigators to create a PROTECT II scoring system, comparable to the SYNTAX score, which will help interventionalists define which patients should have hemodynamic support from a device like Impella while undergoing PCI, based on their risks factors. "There are probably subgroups that definitely do need it, and we expect that better hemodynamic support is going to translate into better benefit in higher-risk patients."


O'Neill disclosed he has a consulting agreement with Medtronic. Mehran disclosed consulting fees/honoraria from Cardiva, Ortho-McNeil, Regado, the Medicines Company, Abbott Vascular, AstraZeneca, and Cordis, and research grants from Bristol-Myers Squibb/Sanofi-Aventis. Waksman disclosed consulting fees/honoraria from Medtronic, Abbott, Biotronik, the Medicines Company, Merck, and Boston Scientific.