BCIS-1: Pre-Procedure Elective IABP Use in High-Risk Patients Reduces Mortality by 34% at 5 Years

The BCIS-1 Study was the first randomized, controlled, multicenter trial to assess the efficacy and safety of elective intra-aortic balloon pump (IABP) use in patients undergoing high-risk percutaneous coronary intervention (PCI). BCIS-1 randomized 301 high-risk patients to either elective IABP use and PCI, or no planned IABP use and PCI, with rescue IABP permitted in the latter group in the event of major hemodynamic compromise. All patients had an ejection fraction of <30% and a large amount of myocardium at risk. The primary endpoint (reported in JAMA in 2010¹) was major adverse cardiac and cerebrovascular events (MACCE) at the time of hospital discharge or 28 days after PCI. The trial was conducted in 17 cardiac centers in the U.K.

The newly reported follow-up study involved collecting mortality data for all study participants for up to six years. The primary endpoint was all-cause mortality. Follow-up study data was reported at the American College of Cardiology Scientific Session on March 24, 2012, in Chicago, Illinois.

Why do a trial looking at elective IABP use?

The results of the initial trial were published in JAMA in 2010. The rationale for BCIS-1 was that there had not been any randomized, controlled data supporting the intra-aortic balloon pumps in high-risk PCI. At the time, the guidelines were very woolly, in the sense that it was left to the discretion of the operator. Given the lack of randomized data, there were no formal, classified guidelines, but a narrative acknowledgement that IABP may be useful in certain high-risk patients or in cardiogenic shock.

BCIS-1 was carried out across 17 centers in the U.K., with 301 patients randomized to either elective intra-aortic balloon pump support during high-risk PCI or to PCI without any planned IABP use. The “no planned IABP” arm could have bailout balloon pumping if a complication occurred during the procedure in the cath lab. Inclusion criteria were that patients needed to have quite severe left ventricular impairment, an ejection fraction of <30%, and a large amount of myocardium at risk from having extensive coronary artery disease. We categorized the amount of coronary artery disease using the BCIS-1 jeopardy score, a modification of the Duke jeopardy score. The maximum score is 12, and to get into the trial you needed a score of 8 or more. In other words, two-thirds or more of the total myocardium had to be subtended by diseased coronary arteries.

The primary endpoint of the original BCIS-1 trial was major adverse cardiac and cerebrovascular events (MACCE) at hospital discharge, a composite of death, myocardial infarction, cerebrovascular event or unplanned revascularization. One of the major secondary endpoints was 6-month all-cause mortality. The other endpoints were all in-hospital endpoints, which included cath lab complications, severe hypotension, ventricular tachycardia or ventricular fibrillation or tachycardia pre-arrest, and bleeding and vascular complications.

The headline result from the original BCIS-1 trial was that there was no difference in in-hospital MACCE. Major adverse events occurred in 16% of the control arm, exactly as expected when we did our power calculations, but the occurrence of MACCE was not reduced by elective IABP use. Essentially, this told us that a strategy of routine IABP use in all these patients could not be recommended.

However, the secondary outcomes were interesting for two reasons. First, there were many more procedural complications in the group that had no planned IABP support, and most of those patients had severe hypotension. As a result, 12% of all patients in the group scheduled to have PCI without IABP needed rescue IABP during the course of their procedure. There was a significant difference in favor of the elective IABP arm.

Can we say there was benefit seen from rescue IABP use in these patients?

The trial was not designed to compare a strategy of rescue IABP versus no IABP use because the only way we could do that would be to not allow rescue IABP insertion in patients who were assigned to have PCI without planned IABP use, even if they did need it during the procedure. But crossover was allowed in BCIS-1, so we cannot say what would have happened to those patients had they not had the rescue use of the balloon pump. The protocol specified that if a patient drew a no-planned balloon-pump treatment assignment, operators should only put in an IABP in extreme instances, where death or a major catastrophe was likely. It turned out that 12% of patients crossed over to rescue IABP use, and in doing so, achieved one of the secondary endpoints of the study, intraprocedural complications.

The other secondary endpoint was 6-month all-cause mortality, assessed by tracking the National Office of Statistics’ database in the U.K. This is a central registry of all individuals in the U.K., which makes it a very tight and robust means of checking mortality. At six months, there were fewer deaths in the elective IABP group although the absolute number of deaths was low at this stage and the difference failed to reach statistical significance, as we reported in JAMA in 2010.

What triggered the five-year follow-up study?

It was the numerical difference in mortality between the two groups at six months, favoring elective IABP use.  Although the confidence intervals were wide, the curves appeared to be diverging at that stage. So, we got IRB approval to extend the follow-up and went back to the central database of the UK Office of National Statistics. Recruitment of this trial finished in January of 2009, with patients randomized between December 2005 and January 2009. We finished mortality data collection in October 2011, giving us follow-up for up to 71 months from randomization. The median follow-up duration was 51 months in the whole cohort.

The first thing to note about the 5-year results is that there were 100 deaths that occurred in this cohort of 301 patients — 33% mortality overall. This confirms the truly high-risk nature of the enrolled population, with a very high long-term mortality. Interestingly, the apparent separation in event curves that was seen early on at 6 months, was borne out in the longer term, achieving statistical significance at long-term follow-up. In the patients who had elective IABP support, 42 deaths occurred, compared to 58 deaths in the no-IABP support, with a hazard ratio of 0.66. These numbers demonstrate a 34% relative risk reduction in mortality in patients who have elective IABP support, which, in effect, is the headline result.

What do you think explains such a dramatic difference between the two groups?

We collected all-cause mortality data via the central U.K. database, so we do not have information on why these patients died, and cannot comment on cardiovascular mortality alone. To that extent, we cannot accurately work out how the balloon pump may have helped, but we looked at a few possible causes and a few possible confounders as well.

The first thing we looked at was whether it was possible this was just a statistical quirk. We used a statistical method, which is to look at time-varying hazard ratios. In other words, we looked at the hazard ratio at 6 months, the original result, at one year, all events occurring after one year, and at the overall cohort. What is quite striking, is that the hazard ratio I gave you of 0.66 is seen all the way through, suggesting a consistent treatment effect at all points. As more deaths occur, the result becomes more statistically significant, and that suggests that it is less likely to be a statistical quirk, but rather a real effect.

One possible mechanism is that the balloon pump may have reduced periprocedural myocardial infarction. In the main trial, we reported no difference in the occurrence of MACCE. Most in-hospital MACCE were myocardial infarctions. However, we used quite a high threshold to define myocardial infarction in the main trial. We used a CKMB threshold of 3x the upper limit of normal. Essentially, that sort of threshold would pick out only large infarctions, not the smaller, troponin-defined infarctions that have, nevertheless, been shown to be important in determining long-term prognosis. It is possible, that there may have been a difference between the IABP and no-IABP groups in smaller, but clinically important, infarctions that we failed to detect in the original trial. It may be that the very low ejection fraction at recruitment (the mean EF in both groups at inclusion was around 23%) means that patients won’t tolerate even a small amount of infarction, but we have no data to confirm or refute that hypothesis.

It is conceivable that there may be a confounder, which is that once the operator knew they had an elective balloon pump with which to do the PCI, they might have done more or different PCI. If this were the case, it might be the more complete revascularization that offers benefit rather than the balloon pump itself. We looked at all of the procedural characteristics to address this: the amount of left mains treated, the amount of proximal left anterior descending coronary arteries treated, the number of vessels, the number of lesions, the number of stents deployed, and so forth. There were no differences between the two groups at all. There was no difference in the amount or type of revascularization that operators performed, regardless of whether they had the comfort of an IABP or not.

It seems to be a true treatment effect. But we have to acknowledge the limitations in that we did not specify all-cause mortality as an endpoint when we originally designed the study. It was a five-year follow-up driven by what we observed in the early data. But, to counter that, there was a very high event rate, many more events than we had anticipated. The mortality rate in this group seems very high, and so that does allow us to do some meaningful statistical comparisons.

There was, in the short term, an increase in bleeding and access site complications in the elective IABP group.

There were actually very few vascular complications, so we could not make any proper statistical comparison, but there was definitely a difference in bleeding complications. We subdivided that into minor and major bleeds. The difference was actually driven by minor bleeds, essentially defined as a decrease in hemoglobin concentration between 2-4 grams per deciliter. Even if there was not any overt bleeding, if blood tests revealed that there was a drop, the patient was classified as having a minor bleed. There was a clear increase in minor bleeds across the balloon pump group. That is recognized not just for IABPs, but all percutaneous left ventricular assist devices. The benefits that you gain need to be balanced against vascular and bleeding complications, which are higher, whatever trial you look at, in the group that have the device, as opposed to those who do not. In general, you should use an IABP or left ventricular assist device if you think that small bleeding or vascular risk is balanced by a bigger MACCE or mortality reduction.

We do not have any information on late bleeding, because the only data we collected after hospital discharge was mortality, rather than bleeding.

You mentioned that only CKMB at 3x the upper limit of normal was considered a periprocedural MI. Was the importance of small periprocedural MIs understood at the time of the original trial design?

Not to the same extent. Back in 2004-05 when the trial was designed, cardiologists were just getting acquainted with cardiac troponin as marker of periprocedural infarction, and we did not really know its significance. Since that time, there has been a lot of data showing that even a small rise in troponin does equate to demonstrable myocardial infarcts on cardiac MRI scans, and also that those patients do have a worse outcome. Perhaps there was less appreciation of the importance at that time.

However, the choice of CKMB was deliberate, because we used a composite endpoint. The components of a composite endpoint needed to be balanced. Death, cerebrovascular accidents or having an emergency extra revascularization are clearly bad outcomes, but would a small troponin rise be equivalent? We wanted to avoid a trial where the composite endpoint would be driven entirely by small troponin rises. For that reason, we chose a high threshold to define periprocedural infarction, which means that you would need to have a big infarct to score a primary endpoint. While this was important from a trial-design perspective, looking at it in retrospect, it slightly reduces our ability to dissect out why these effects might have happened. It may well be that if we had measured troponin, we might be able to see a difference between those patients who died and those who did not.

When the original BCIS-1 trial was presented at the TCT 2009 meeting, some commented that the trial considered myocardial territory at risk, but did not really take into account lesion morphology or any other factors that might increase risk.

That is a really important issue in terms of how cardiologists interpret this and other data, and translate it into their clinical practice. Myocardium at risk is very important, and perhaps the most important thing to consider if you were to have a complication in the cath lab. All the lesions scored in a jeopardy score come into play at that point, and it is that myocardial territory that may determine whether your patient is alive after the event or not. Equally, we know as practicing interventional cardiologists, that if you have a calcified, complex, bifurcating lesion in a given territory as opposed to a simple lesion in that same territory, you treat the case differently in terms of anticipated complications during the procedure. It may be that, in time, those patients in whom we continue to use elective IABP support will be those who have very complex lesion characteristics, a high jeopardy score, as well as very poor left ventricular function at the outset. Unfortunately, the trial does not have the power to come up with a model of the kind of patient who will become compromised and dictate when you should put an IABP in up front.

What can clinicians take away from the BCIS-1 trial and 5-year follow-up?

When the original results were presented, it was seen as a negative trial for IABP support. But what the trial told us was that we cannot use one or two criteria to define high-risk and say, let’s put balloon pumps in all these patients and prevent complications. We need to apply clinical judgment to select out, from within that group, those patients who are going to need IABP support. In the discussion section of the original paper, we strongly recommended that if the patient does meet the high-risk criteria of the trial, a standby IABP approach be used. There should be a balloon pump turned on in the lab and the cath lab staff warned about the possibility of needing it, because if rescue IABP is needed, it is needed quickly.

The follow-up study does have limitations and we do not know whether the results would be reproduced in a larger trial. That said, these new data suggest that there may be differences with elective IABP use that become apparent in the long term. These patients are a group with quite poor outcome in the long term; thirty-three percent mortality at a median of 51 months is almost comparable to some cancers. If we can achieve any sort of improvement in outcome in these patients, it would a good thing. In practice, there will be a certain cohort in whom elective IABP is worth using. If a patient has severe left ventricular impairment and extensive coronary disease, and in addition, has lesion characteristics that make the interventionalist anticipate a long, complex procedure, there may be a case for putting a balloon pump in electively and not waiting for the patient to become compromised.

Dr. Perera can be contacted at divaka.perera@kcl.ac.uk

Reference

1. Perera D, Stables R, Thomas M, Booth J, Pitt M, Blackman D, de Belder A, Redwood S; BCIS-1 Investigators. Elective intra-aortic balloon counterpulsation during high-risk percutaneous coronary intervention: a randomized controlled trial. JAMA 2010 Aug 25;304(8):867-874.