In 1976, the Veteran’s Administration trial of 113 patients comparing coronary bypass surgery (CABG) with medical therapy ushered in the era of revascularization for left main coronary artery (LMCA) disease. Interestingly, the third patient to be treated with balloon angioplasty in the initial series of Andreas Gruentzig was a 43-year-old with LMCA disease who died 4 months post procedure.

Although LMCA disease subsequently remained primarily in the province of the cardiac surgeon, an analysis of 11 randomized trials of 11,518 patients requiring CABG versus PCI identified almost 4500 patients with LMCA disease in which 5 years all-cause mortality in the two groups was almost identical. When stratified by the SYNTAX score there was a non-significant but nonetheless clear trend towards a higher mortality with PCI in those with a SYNTAX score of 33 or more, but in the intermediate and low SYNTAX score groups there was no difference. Similar 5-year results were noted in the SYNTAX trial which demonstrated no significant difference in MACCE between PCI and CABG in LMCA disease patients with low and intermediate SYNTAX scores but a clear benefit statistically for CABG in patients with a score of 33 or more.

What did the guidelines say?

The 2014 European Society of Cardiology Guidelines on revascularization gave PCI for LMCA disease with a low SYNTAX score a Class 1 recommendation; in those with an intermediate SYNTAX score a Class 2A recommendation (as opposed to Class 1 for CABG), and in those with a high SYNTAX score a Class 3 recommendation (versus Class 1 for CABG). In my opinion these guidelines quite accurately reflected prevailing opinions at that time.  The ACCF/AHA guidelines, which were published earlier in 2011, were somewhat softer and really emphasized the role of PCI in patients who were not good surgical candidates.


The EXCEL trial randomzied 1905 patients with low and intermediate SYNTAX scores to PCI or CABG. This demonstrated no significant difference at 3 years in the primary composite endpoint of death, myocardial infarction, and there was also no significant difference in deaths from any cause.  Initially the event rate was slightly higher with CABG due to a higher rate of procedural myocardial infarction but at 3 years the curves had diverged in favor of CABG suggesting that over time there would be a significant difference in favor of surgery.


The NOBLE trial compared PCI with CABG in 1201 patients. This showed a clear benefit with CABG in terms of MACCE but there was no difference in all-cause mortality. There was also a surprising increase in late stroke in the PCI group in contrast to a trend towards a higher rate of early stroke after CABG in other studies.

Possible Consensus? 

Around  that time, I, in conjunction with Deepak Bhatt and Gregg Stone published a commentary in Circulation concluding that the results of these two trials might impact the current ACCF/AHA guidelines by broadening the patient pool who might benefit from PCI. We also emphasized the importance of other factors such as life expectancy, comorbidities, the angiographic extent of disease, and left ventricular function in addition to patient preference in regard to clinical decision making. In this commentary we also stressed the necessity for a longer period of follow-up as the curves were diverging in favor of CABG.

These data set the stage for the 5-year results of the EXCEL trial which demonstrated no statistically significant difference in the primary composite endpoint of all-cause mortality, stroke, or myocardial infarction but a trend which certainly favored CABG. However, the majority of late deaths in the PCI arm were not cardiovascular and notably the majority were due to cancer and infections. There was a higher rate of stroke in the CABG arm, but this was not statistically significant.  A  subsequent meta-analysis of 5 trials of 4612 patients with a follow-up of 1 to 10 years demonstrated a relative risk of all-cause mortality at last follow-up of 1.03 (0.82 – 1.30) for PCI versus CABG, and EXCEL was the only outlier among the 5 trials with a relative risk of 1.35 (1.04-1.78). The relative risk for cardiac death for all 5 trials pooled was 1.03 (0.79-1.34).

Risk of death at latest follow-up. From Ahmad Y et al. EHJ 2020 Mortality after drug-eluting stents vs. coronary artery bypass grafting for left main coronary artery disease: a meta-analysis of randomized controlled trials.

At this stage I felt that the world of LMCA revascularization appeared to be in order. There was a large body of trial data demonstrating a clear role for PCI in selected patients but with a trend towards a benefit from surgery in regard to all-cause mortality and a statistically significant lower risk of repeat revascularization.


In December 2019 the controversy erupted as the Chairman of the Surgical Committee of the trial (David Taggart) took his name off the 5-year results paper published in NEJM, and this was followed by a series of allegations and the withdrawal of the European Association of Cardiothoracic Surgery from the ESC Guidelines. The criticisms of the trial plus the rebuttal from the investigators were aired publically by the BBC, in the print media, and intensively discussed in social media both by physicians and others. These differences of opinion continue to smolder. Is the controversy justified? Taggart has subsequently retracted some of his statements to the effect that the diagnosis of myocardial infarction was changed during the course of the trial and that the data were manipulated or distorted.  Nonetheless, he has reiterated his conviction that there was a failure “to present practical specific data that was potentially important to the interpretation of the trial”.

To my mind, it would appear that the argument rests upon the issue of all-cause versus cardiovascular mortality and this is really a subject open to individual interpretation. As mentioned before, the majority of the excess deaths were due to cancer and infections which is difficult to explain. When one looks at all 5 trials of LMCA disease, there is no signal of any increase in all-cause mortality in the other 4 trials. The other bone of contention arose from the definition of procedural versus spontaneous myocardial infarctions. This has been a subject of discussion, controversy, and uncertainty for the last 30 years dating back to the initial trials of bypass surgery. In the EXCEL trial the pre-specified Protocol Definition used a CPK-M-B level of 10 times the URL (Upper Reference Limit) or 5 times the URL plus other evidence, based on ECG, angiography or imaging. The definition was the same for both PCI and CABG, and in a prior paper it was shown that among patients who were diagnosed with a myocardial infarction using this definition, the diagnosis had a marked impact on mortality after both PCI and CABG. The alternative definition is the Third Universal Definition of Myocardial Infarction using CPK-M-B if troponins were unavailable (as was the case in many centers) in addition to supporting evidence both clinically and with imaging data. Using the Third Universal Definition of Procedural MI, the criteria for diagnosis differed between PCI and CABG and the rate of procedural MI was higher with PCI compared to CABG. The impact of the diagnosis of procedural MI using this definition on mortality has not yet been published, but a detailed study using multiple definitions of myocardial infarction in EXCEL has been completed and is currently under review.

Will this douse the flames and bring everyone to a consensus?  I doubt it despite the recent letter to NEJM on behalf of the investigators stating the results using both definitions of myocardial infarction.

Perhaps the best one can do in the circumstances is to look dispassionately at the data that we have and use some clinical common sense. To quote Voltaire “common sense is not so common”. I also like to use the quotations of 2 friends of mine.  The first from Robert Califf is “for a difference to be a difference it must make a difference” and the second from the biostatistician Stuart Pocock, “a P value is no substitute for a brain”. Given that the absolute magnitude of the differences of the benefits in EXCEL and the other trials is small although favoring CABG, there clearly is a role for PCI in selected patients LMCA disease. Nonetheless, to make a clinical recommendation for an individual patient we need to go beyond the trial data and beyond the P value and take other considerations into account, such as the coronary anatomy and left ventricular function, the extent of coronary disease, age and life expectancy, comorbidities including frailty, chronic kidney disease, diabetes, peripheral vascular and cerebral vascular disease, and also operator proficiency. The latter is often not discussed in randomized trials but is paramount, as is patient preference.

Personally, I believe that recent events have certainly highlighted issues that needed discussion and further analysis. Nonetheless, overall, I believe the controversy has been a distraction. All trials have their limitations but the trials of LMCA revascularization have been well designed, the data are solid, and they have influenced clinical care and should have an impact on the guidelines. I am reminded of the title of the Shakespearean play, “Much Ado About Nothing”. The controversy has not been about “nothing” and has highlighted issues which deserved additional discussion. Nonetheless, I feel that its importance has been overblown.

Editor’s note: This is an ongoing debate. Click here to read responses to this post.

Dr Gersh is an investigator on EXCEL and a co-author of several papers from this trial, including the design paper published in Eurointervention, in addition to being a co-author on the published Rebuttal to the criticisms on behalf of the other trial investigators.

Full Coverage of the PCI versus CABG for LMCA on CVCT Cardiobrief:

+ posts

Leave a Reply