More Risk of Death—All, Cardiac, and Noncardiac—With PCI Over Surgery: Meta-analysis

Focusing solely on mortality, a new meta-analysis in a broad range of patients with coronary artery disease shows that PCI is associated with an increased risk of death when compared with CABG surgery.

Across 23 trials, including those that included patients with left main CAD and multivessel disease, the analysis revealed that PCI was associated not only with a higher incidence of all-cause mortality, but also with a higher risk of cardiac and noncardiac death compared with surgical revascularization. For the researchers, the findings suggest that all-cause mortality is “the most comprehensive and unbiased endpoint for myocardial revascularization trials” and should be the preferred endpoint for future studies.

“Clearly the excess in noncardiac mortality with PCI tells us that the problem with the adjudication of events is real and important,” Mario Gaudino, MD (Weill Cornell Medicine, New York, NY), told TCTMD. Gaudino stressed the group is not implying PCI causes cancer or other noncardiac deaths, but given the heightened risk of noncardiac mortality and the possibility that some cardiac deaths may go unrecognized, he said he believes the primary endpoint of myocardial revascularization trials should be all-cause mortality.

“Based on the data, all-cause mortality is the most-reliable, the most-solid endpoint to be used,” he said. “Yes, you take the risk of counting some events that aren’t necessarily related to your intervention, but it should be the same in the two groups. Most importantly, I think it would be much worse to risk not counting events related to the intervention.”

For David Kandzari, MD (Piedmont Heart Institute, Atlanta, GA), however, relying on all-cause mortality alone as the primary endpoint may be an “oversimplification.”

“Instead, we need to focus on cardiovascular mortality that may be more attributable to the procedure itself,” he told TCTMD. “At the same time, it opens the opportunity for us to find a better consensus and better definition for discriminating between cardiac and noncardiac mortality. While there is a need to better define cardiovascular mortality, it moves past the point of accuracy to instead rely on all-cause mortality.”

Gaudino et al's analysis was published October 12, 2020 in JAMA Internal Medicine.

The EXCEL Controversy

Controversy around the EXCEL trial has rekindled debate surrounding the ideal endpoint for revascularization studies. Aside from the ongoing questions over the periprocedural MI data, including the “missing” Universal Definition MI results, the 5-year EXCEL findings kicked off a heated dispute over the differences in mortality seen with PCI and surgery in that trial. As reported by TCTMD, the 5-year EXCEL results showed that PCI was associated with a significantly increased risk of death (HR 1.38; 1.03-1.85) compared with surgery, but there was no difference in the risk of deaths from cardiovascular causes.

Those findings led David Taggart, MD, PhD (University of Oxford, England), chairman of the EXCEL surgical committee during the design and recruitment phase of the trial, to pull his name from the New England Journal of Medicine publication after he accused the researchers of downplaying the mortality finding. The EXCEL researchers, for their part, countered that the study wasn’t powered for mortality and that the majority of excess deaths in the PCI arm were from noncardiovascular causes. In EXCEL, there was no significant difference in the risk of death from cardiovascular causes.

To TCTMD, Gaudino expressed “immense respect for the EXCEL trial” and its researchers, calling it the best study of left main disease to date, one that will continue to be a source for clinical decision-making and guidelines for many years to come. And while there are ongoing debates surrounding the interpretation of the mortality data, such is par for the course in research. Given the conflicting all-cause and cardiac mortality findings from the study, though, he and his coinvestigators decided to look into the risk of death between PCI and CABG in a broad range of clinical trials. 

The new meta-analysis includes 23 randomized trials, the oldest of which is the SoS trial, dating back to 2002. The study includes several randomized trials comparing PCI and surgery in patients with left main disease, such as EXCEL, NOBLE, SYNTAX, and PRECOMBAT, but also includes numerous studies outside that patient population, such as FREEDOM, BEST, ARTS, and MASS-II, among others.

In total, 13,620 patients were included, with the number of patients in the individual studies ranging from 44 to 1,905. Of the 23 studies, 18 used all-cause mortality as part of their primary composite endpoint.

During a mean follow-up of 5.3 years, PCI was associated with a higher risk of death from any cause (incident rate ratio [IRR] 1.17; 95% CI 1.05-1.29), as well as cardiac-specific causes (IRR 1.24; 95% CI 1.05-1.45). Noncardiac mortality was also significantly increased among patients treated with PCI (IRR 1.19; 95% CI 1.00-1.41).

Interestingly, the pooled IRR for cardiac mortality compared with CABG surgery was higher for patients treated with BMS than those treated with DES, although the test for interaction was nonsignificant. Focusing on patients with left main disease, the benefit of CABG surgery over PCI was reduced, with a lower pooled IRR for all-cause, cardiac, and noncardiac mortality than in patients without left main disease (all tests for interaction were nonsignificant).

Gaudino said their results line up with the oft-cited meta-analysis by Stuart Head, MD (Erasmus University Medical Center, Rotterdam, the Netherlands), published in 2018. In that study, the risk of all-cause mortality was 20% higher in those treated with PCI compared with CABG.

“We know what the data tell us,” said Gaudino. “It’s been consistently telling us the same story. From time to time we get excited and there is a new controversy and then we argue in the journals. PCI is a less-invasive procedure and in the short term it works very well. In the long term, it doesn’t work as well as CABG. Some patients might be interested in short-term outcomes—they have a short life expectancy or are poor surgical candidates—and PCI is a great option. I’m very thankful we have PCI for some patients we see in the office. I’m happy my colleagues can put in a stent. But for other patients, those interested in longer-term outcomes, then you take a higher up-front risk with CABG but then you’re protected for longer.”

Debating Ideal Endpoints

To TCTMD, Gaudino said the debate over using all-cause mortality as the primary endpoint is an old controversy, one not necessarily only related to CABG and PCI trials. “There are advantages and disadvantages of using overall and cause-specific mortality, but overall mortality is easier to capture and is a more-solid outcome, but one can make the case that it includes events not necessarily related to the intervention you want to test,” he said.

While cardiac mortality may be more closely linked to CABG surgery and PCI interventions, “the problem is adjudication of the cause of death, which is always challenging, and there are multiple components that can lead to death,” said Gaudino. Additionally, the reported reason for the death might differ from the larger contributing factor, he said.

“The perfect example is the issue with sepsis,” said Gaudino. “In clinical trials, there is a not-insignificant proportion of patients who die of sepsis. In general life, people don’t die of sepsis very often. Sepsis is something that happens, unfortunately, very often in the ICU. Patients are brought to the ICU for other reasons. In this case, say cardiac reasons, and then after a prolonged stay in the ICU, they get septic and die. Did they die of sepsis? Well, not really. They died, in my opinion, for the reason they were brought to the ICU.”

Kandzari, the lead US interventional investigator for EXCEL, said there are ongoing efforts to increase collaboration between surgeons and interventionalists, but such opinions, as well as the researchers’ conclusions, cloud the issue. In fact, he expressed disappointment in the group’s statement that “even noncardiac deaths after PCI may be in fact related to the procedure and/or subsequent management” of patients. The idea that PCI could possibly be related to noncardiac deaths is a stretch, said Kandzari.  

“To make the suggestion of PCI contributing to noncardiac deaths is an illogical conclusion and a distraction from the primary data they are presenting,” he said. “I don’t see this study impacting practice for either surgeons or interventionalists. At best, it adds only to the inconsistencies and confusion across pooled analyses of comparative studies, many of which have their own inherent limitations. At worst, it endorses erroneous conclusions with gaps in logic like this relationship between PCI and noncardiac death.”

In EXCEL, said Kandzari, the prespecified primary endpoints were agreed upon by surgeons and interventional cardiologists, and “definite” cardiac deaths differed by less than 1% through 5 years of follow-up. When the group conservatively attributed uncertain causes of death to cardiovascular causes, there was a nonsignificant 1.3% difference in cardiovascular deaths between PCI and the bypass surgery arm, he said.

“I still believe cardiac mortality is a very valid endpoint for coronary revascularization procedures, whether it’s surgery or PCI, because it’s equally inappropriate to attribute sepsis or cancer-related deaths to PCI in as much as it would be to surgery.”       

Recently, another meta-analysis that included only trials of patients with left main CAD showed there was no increased risk of death with PCI compared with surgery, noted Kandzari. “So which data do you choose?” he asked.

Gaudino said there are inherent limitations to their analysis, noting that they pooled heterogeneous trials. The patient populations differed in the studies, as did medical management, the types of stents used, and the type of surgical revascularization (off-pump and on-pump CABG surgery, for example), among other variables.