Bayesian Backing for Surgery in EXCEL Clash Over Left-Main Disease

June 08, 2020

More ammunition for the debate over interpretation of the EXCEL trial has hit print. A new analysis of the trial yielded probabilities for 5-year outcomes after revascularization for left-main (LM) coronary disease that favored the surgical approach. The findings were meant to serve as an alternative to the usual hazard ratios in an effort to clarify the trial's clinical message, the study's author says.

The results add to the public and often loudly argued clash over contrasting interpretations of EXCEL, one with profound clinical implications, as well as charges of impropriety in the trial's execution dating to at least early October 2019, as chronicled by | Medscape Cardiology.

The new bayesian analysis of EXCEL — which compared coronary bypass (CABG) surgery and percutaneous coronary intervention (PCI) for LM disease — saw substantial probabilities of excess mortality and other clinical events with PCI, whether the trial was considered alone or combined with three comparable earlier studies.

That challenges the EXCEL trialists' interpretation that there were few, if any, important differences in all-cause mortality or rate of myocardial infarction (MI) and some other outcomes in patients assigned to PCI or CABG.

The new study "suggested with a reasonably high probability that PCI was associated not only with statistical inferiority but also with clinical inferiority to CABG" for the treatment of LM disease, says the report published online June 1 in JAMA Internal Medicine.

The bayesian approach provides a new analytical twist in applying EXCEL to practice by accounting for the effect of the field's invasive-therapy learning curve in patients with LM disease based on the three prior trials, explained the report's sole author, James M. Brophy, MD, PhD, McGill University, Montreal, Quebec, Canada, to | Medscape Cardiology.

For example, after including the results of those earlier trials — PRECOMBAT, SYNTAX, and NOBLE — the probability of excess mortality for PCI, compared to CABG, was 85%. And there was a 47% probability that the excess deaths would exceed 1 per 100 patients treated. The figures based only on EXCEL were higher.

"I'm not trying to be proscriptive and say every patient who has left-main disease must go to CABG," Brophy said; his report cautions that the choice of revascularization strategy still must be tailored. "All I'm saying is that this is information, hopefully in a way that's transparent and understandable, that can lead to the sort of discussion that we need between the healthcare provider and the patient."

An accompanying editorial supports the analysis in agreeing it shows a "high probability of a clinically important benefit with CABG over PCI."

The overall mortality results "clearly favor CABG with additional reductions, compared with the use of PCI, in nonprocedural MI and repeat revascularization procedures, and no increase in stroke," writes Sanjay Kaul, MD, Cedars-Sinai Medical Center, Los Angeles, California.

"The suggestion by EXCEL investigators that the treatment strategies are comparable is questionable at best, and at worst risks unnecessary and avoidable deaths of patients with LM CAD."

But a leading EXCEL trialist who examined the new report says it contains "several nonstatistical errors" and "glaring omissions," and knocked its conclusion of an 85% excess-mortality probability with PCI.

The finding of a 0.9% difference in 5-year all-cause mortality "may be statistically correct, although the posterior probability of more deaths with PCI was only 85%, ie, not meeting the level (95%) that is usually required to confidently be certain there is a true difference," Gregg W. Stone, MD, Icahn School of Medicine at Mount Sinai, New York City, told | Medscape Cardiology by email.

"However, even if real, this is a difference in mortality of < 0.2%/year (!), a clinically meaningless rate given the extra morbidity of CABG," Stone said in the emailed statement.

Stone also criticized the analysis for pooling events across the four trials without respecting distinctions between procedural MI, which is identified by biomarker assay, and spontaneous clinical MI.

"Thus they combined apples with oranges. They should instead have analyzed each component event separately, for example procedural and nonprocedural MI, which would have shown that the former favored PCI and the latter favored CABG," Stone said.

"Considering both the periprocedural and long-term outcomes, there are no major differences in cardiovascular or all-cause death and or total MI; and late quality of life is similar."

"There were differences in how each study defined nonfatal MI, that's certainly true," Brophy said. The bayesian analysis incorporated MI data according to each trial's own definition; and notably, NOBLE did not count periprocedural MIs.

"There's going to be some variability for sure" he acknowledged. "The models account for this to a certain extent. I don't think it invalidates the idea of pooling data together. Clearly in each study, the trialists thought that what they were measuring in terms of MI was the important concept."

Regarding the probability of death with PCI in the analysis, Brophy said "the conventional thinking is that these results are just not statistically significant and we can ignore them. But the bayesian would challenge that interpretation."

Many people would at least consider some caution if they believed there was an 85% chance of increased mortality "and a 47% chance that it’s going to be a greatly increased risk," he proposed.

"Even though you don't have the 95% confidence that you would want from the frequentist standpoint, I think the bayesian does bring some extra information to the table. A lot of people might still want to take their umbrella even if there's going to be an 85% chance that it will rain."

In the bayesian analysis of only EXCEL, there was a 95% probability of excess of the trial's primary endpoint — a composite of death, nonfatal MI, and stroke — for patients in the PCI group and 87% probability that the difference exceeded 1 excess event per 100 patients treated, Brophy reported.

For mortality in EXCEL, the probability of an excess with PCI compared to CABG was 99%, with a 94% probability that the difference exceeded 1 death per 100 patients.

Composite endpoints in PRECOMBAT, SYNTAX, and NOBLE were not consistently identical to those of EXCEL, so Brophy went to each trial's data to derive the composite of death, nonfatal MI, and stroke for each one, as necessary.

Then, after pooling outcomes for all four trials, there was a 96% probability of an excess primary endpoint with PCI and 86% probability that the difference exceeded 1 additional event per 100 patients. The corresponding probabilities for mortality alone were 85% and 47%, respectively.

Brophy urged caution, however, in applying the results to individual patients. The evidence as a whole "in most cases favors CABG," he said. "But one could easily imagine a scenario where somebody is very high surgical risk, in which case these averages may not apply. Or somebody has a very limited life expectancy." Under such circumstances, "maybe you should go with the less invasive, quicker fix. So I think you have to personalize this data."

Neither Brophy nor Kaul had disclosures. Stone has reported receiving honoraria or consulting fees from Miracor, TherOx, Neovasc, Valfix, Ancora, Robocath, HeartFlow, Gore, Ablative Solutions, Abiomed, and MAIA Pharmaceuticals; serving as an officer or director for Orchestra Biomed; and having ownership or partnership interest in Biostar family funds, Qool Therapeutics, Aria, Cagent, Applied Therapeutics, Spectrawave, and Ancora.

JAMA Intern Med. Published online June 1, 2020. Article, Editorial

Follow Steve Stiles on Twitter: @SteveStiles2. For more from | Medscape Cardiology, follow us on Twitter and Facebook.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.