CABG Over PCI for Severe HFrEF Patients With Complex CAD: Observational Data

PCI in patients with heart failure with severely reduced ejection fraction (HFrEF) was associated with significantly worse survival and more major adverse cardiovascular events compared with CABG, a large, retrospective, propensity-matched study from Ontario, Canada, shows.

Over a median 5.2 years of follow-up, PCI was associated with a relative 60% higher mortality rate versus CABG, equaling nearly a 5% higher absolute rate of death.

While subject to all the shortcomings of nonrandomized data, the study provides important new insights in a field sorely lacking in information to help physicians and patients decide on the best treatment strategy, according to Louise Sun, MD (University of Ottawa Heart Institute/Institute for Clinical Evaluative Sciences, Canada), and colleagues writing in JAMA Cardiology this week.

To TCTMD, Sun said she believes the stark difference in outcomes between CABG and PCI in this retrospective analysis should be enough to nudge guideline writers to more strongly favor CABG for patients with complex CAD and HFrEF. “This is lower down on the evidence pyramid, but in a setting where we lack large and definitive randomized trials, a large observational study based on a real-life cohort is quite generalizable and is really the next best thing.” 

European guidelines recommend CABG over PCI for patients with reduced LVEF and multivessel CAD, whereas US guidelines support CABG but make no specific recommendation for PCI, saying that the choice of revascularization can be made based on consultation with a surgeon and interventionalist. That equivocation may stem from the fact that STICHES, the 10-year extension study from the landmark STICH trial, carved out a role for revascularization over medical therapy in patients HFrEF, but only included CABG, not PCI. At the time, some interventional cardiologists were hopeful that less invasive PCI might, in time, also prove beneficial in these complex patients. With the randomized REVIVED-BCIS2 trial, now underway, comparing PCI to best medical therapy in patients with ischemic heart failure, investigators hope to deliver a more concrete answer.

Eric Velazquez, MD (Yale School of Medicine, New Haven, CT), who led the STICHES trial, told TCTMD that he and an “international group” of investigators are actively planning a randomized controlled trial to address this question. “For now, this analysis reinforces what, to me, makes sense but practically may not be done very often—HFrEF patients being considered for PCI should have a heart team review to see if CABG could be offered as well so the patients are fully informed and we can make the best decisions with the patients involved and their utilities addressed.”

Ontario Data

For their analysis, Sun and colleagues reviewed data on all Ontario residents ages 40 to 84 with LVEF < 35% who underwent CABG or PCI between October 2008 and December 2016 for left main, left anterior descending, or multivessel CAD.

Of the total cohort of 12,113 patients 2,397 PCI patients were matched with 2,397 CABG patients using 30 baseline characteristics.

Over the first 30 days, there were no differences in all-cause or cardiovascular death, but other endpoints—with the exception of stroke—favored CABG.

30-Day Outcomes

Over a median of 5.2 years, patients treated with PCI were significantly more likely to die from any cause, to die of cardiovascular disease, to undergo subsequent revascularization, and to be rehospitalized for MI or heart failure. Again, only stroke was higher among the CABG-treated patients.

Long-term Outcomes

In a landmark analysis, patients who were event-free at 30 days were still likely to experience higher rates of all-cause death, cardiovascular death, and MACE over the long term if they were treated with PCI instead of CABG.

Additional analyses looking at disease type showed that all-cause mortality was significantly reduced at both 30 days and longer term for all three CAD groups (LAD-only, left main or two-vessel disease, and three-vessel disease). For cardiovascular mortality, however, only patients with three-vessel disease saw significant reductions with CABG versus PCI over long-term follow-up. MACE was significantly lower for all three disease categories at a median of 5.2 years, while diabetes was predictive of reduced cardiovascular deaths with CABG only over the longer term.

Polarizing Data

Speaking with TCTMD, Sun stressed that this particular patient group represents a “small chunk of the caseload.” At Ottawa University Heart Institute, she said, they do approximately 1,800 cases per year, of which 800 would be isolated CABG. And of these, she estimates, about 100 would be patients who also have heart failure with severely reduced ejection fraction, and that’s at an institution that specializes in complex cases. 

As for how to explain the striking superiority of CABG in this observational study, Sun said this is partly a reflection of the complex comorbidities these patients have, particularly the likelihood of very diffuse CAD, which other studies have shown is better treated with CABG. She also acknowledged that despite the best efforts of investigators to propensity match the two groups, there is likely residual confounding that they could not account for in the analysis.

“Certainly I'm not expecting a lot of clapping from the interventionalists when this study is published, and I think we will get criticism in terms of people picking this apart and saying there's a lot of selection bias,” Sun said. “We've done everything that we can think of to take care of imbalances between the groups we're comparing, but the problem is there's always going to be something that we either cannot think of, or something that's kind of inherent in the selection of patients for each procedure.”

Velazquez, in an editorial accompanying the study, details a number of potential factors that might have skewed the results, but concludes, “While unmeasured confounders might partially explain the countervailing 30-day outcomes favoring CABG over PCI, they are unlikely to fully explain the longer-term results.” Left unexplored, he added, is the extent to which patient and physician preferences led to the choice of one therapy over another in this series and, indeed, would continue to drive decision-making in the real world.

To TCTMD, he added: “If CABG can be done, I think it appears from [this paper] that it should be strongly considered as the standard approach to revascularizing these patients for now, pending further data. Many patients may neither want to consider CABG or may not be good candidates, and we cannot tell how many of those type of patients were in this analysis, and thereby there will always be a role for PCI—the question of how big will need further study.”

A separate but related issue, Velazquez added, is the number of patients with heart failure who never undergo tests looking for CAD. He pointed to a recent paper in Circulation: Heart Failure drawing on numbers from the Get With The Guidelines Heart Failure registry, by Kyle D. O’Connor, BA (Duke Clinical Research Institute, Durham, NC), and colleagues. That study showed that among more than 17,000 patients hospitalized with new-onset heart failure, only a minority (39%) ever undergo CAD screening and less than one-quarter do so during their index hospitalization.

“I think the Sun and O’Connor papers reinforce the importance of paying attention to CAD in heart failure and getting better data to treat it to enhance outcomes for the population at large,” Velazquez said.

Also commenting on the study for TCTMD, Dharam Kumbhani, MD (UT Southwestern Medical Center, Dallas, TX), pointed to the continuing debate over left main revascularization, noting that the optimal revascularization method in patients with low ejection fraction is shaping up to be equally polarizing for interventionalists and surgeons. “In the absence of randomized controlled trials,” he said, “it’s going to be hard for one side to win.”

The methods used by Sun et al are “really solid,” he stressed. “They did all the things that I would look for as an editor,” in terms of how well they matched patients for their analysis, the proportion of unmatched patients, the long-term follow-up, and the use of falsification endpoints to rule out the possibility that other factors not captured in the propensity matching were driving the outcomes. At a minimum, he continued, the study is hypothesis-generating and may help to establish power calculations for a subsequent randomized trial.

“Can this ever be a substitute for a randomized controlled trial? No,” Kumbhani said. “But I think these are helpful data and I would not be dismissive of them. I don’t know if this necessarily moves the field forward, but it will certainly add to the debate and discussion.”