The superiority of bilateral internal thoracic artery (BITA) grafting has been regarded for decades as a well-founded belief. The biological rationale is compelling: internal thoracic artery grafts maintain angiographic patency of approximately 90% at 10 years, compared with around half of saphenous vein grafts. And the observational evidence has been remarkably consistent. Our group contributed to this body of evidence: in a meta-analysis of 45 propensity score-matched populations, we documented a significant long-term survival benefit with BITA (HR = 0.78; 95% CI = 0.71-0.86), which persisted in the subgroup of patients with diabetes (HR = 0.65; 95% CI = 0.43-0.99). In a subsequent study comparing BITA with the combination of the left internal thoracic artery and radial artery, we found no survival differences within the first decade, but a significant benefit beyond 10 years. In other words, we started from a position favorable to multiple arterial grafting, and supported it with our own data.

The randomized evidence, however, has always been less enthusiastic. The Arterial Revascularization Trial (ART), designed to test this hypothesis, showed no mortality difference at either 5 or 10 years on an intention-to-treat basis. The usual response from the field—and a reasonable one—was that follow-up was still too short: if the mechanism is the durability of arterial grafts, the benefit should emerge when saphenous vein grafts begin to fail. At 10 years, there was also a suggestion of divergence in the curves for the composite outcome in favor of BITA, which provided the rationale for extending follow-up for another 5 years.

The newly published study is a secondary analysis of the trial, based on the intention-to-treat principle and censored at 15 years after randomization. The primary outcome was all-cause mortality; the secondary outcome was the composite of death, myocardial infarction, or stroke. The study also included analyses adjusted for age, sex, diabetes, and ejection fraction, a per-protocol analysis, a shared frailty model accounting for surgeon effects, and a prespecified as-treated analysis.

The results leave little room for interpretation. At 15 years, 585 of 1548 patients in the BITA group (37.8%) and 584 of 1554 in the single internal thoracic artery (SITA) group (37.6%) had died: HR 1.00 (95% CI = 0.89-1.12; p = .97). The composite of death, myocardial infarction, or stroke occurred in 43.9% versus 45.8% (HR 0.94; 95% CI = 0.85-1.05), with no differences in any of its individual components. Multivariable adjustment did not change the result (HR = 0.99), nor did the per-protocol analysis, and the model accounting for surgeon effects again yielded an HR of 1.00. The incipient separation observed at 10 years did not progress.

The authors acknowledge the main objection to their trial: contamination between groups. A total of 215 patients (14%) assigned to BITA ultimately received only a single arterial graft, whereas 359 patients (23%) in the control group also received a radial artery graft. The as-treated analysis did show a difference, 34.9% mortality among the 1690 patients receiving 2 or more arterial grafts versus 41.1% among the 1330 receiving only 1; adjusted HR 0.84 (95% CI = 0.75-0.95), although the authors themselves caution that this is a nonrandomized comparison. Their conclusion is that the multiple arterial graft hypothesis still needs to be tested in randomized clinical trials.

COMMENTARY:

The first point is to acknowledge that this study closes one argument. For years, ART’s neutral result was explained by insufficient follow-up. That argument established a testable condition; the condition has now been met, and the prediction has failed. An HR of 1.00 with more than 1000 accumulated deaths is not a problem of statistical power: it is an answer. Indeed, the authors themselves emphasize that the extended follow-up provides greater statistical power and confirms the absence of further divergence between the randomized strategies after 10 years.

The as-treated analysis deserves more skepticism than it usually receives. It is not randomized: it compares patients in whom the surgeon was able to complete multiple arterial grafting with those in whom the surgeon decided not to do so, and that decision is made intraoperatively according to coronary target and conduit quality, patient frailty, conduit availability, technical complexity, and other factors. An HR of 0.84 obtained in this way measures, indistinguishably, the effect of the conduit and the effect of being the type of patient in whom that conduit can be used. This is not merely a theoretical objection, and it has now been quantified. Put differently, it confirms that, for whatever reason, patients who undergo or can be offered multiple arterial grafting have a better prognosis. Among 1,291,314 Medicare beneficiaries undergoing surgery between 2001 and 2019, conventional risk adjustment attributed a median survival of 10.74 years to multiple arterial grafting versus 10.33 years with single arterial grafting: almost a 5-month advantage. When the analysis was repeated using the individual surgeon’s multiple arterial grafting rate during the preceding 12 months as an instrumental variable, a way of emulating randomization, the difference disappeared: 10.38 versus 10.38 years. The advantage was not attributable to the conduit, but to the patients in whom it was used.

That said, the study has a limitation that affects its interpretation: cause of death was unavailable between years 10 and 15, and events during this period were not independently adjudicated. Fifteen years after coronary surgery, in a cohort with a mean age of 64 years at randomization, a substantial proportion of mortality will be due to cancer, renal disease, or infection, and the choice of conduit has no plausible pathway to prevent, for example, the development of malignancy. An HR of 1.00 for all-cause mortality is compatible both with the absence of an effect and with a modest cardiovascular benefit diluted by increasingly noncardiovascular mortality. Without cause-specific mortality data, these two hypotheses cannot be distinguished.

There is also another side of the equation that this analysis does not address and that deserves more attention than it usually receives in discussions about conduit selection. In ART itself, at 5 years, the rate of sternal wound complications was 3.5% with BITA versus 1.9% with SITA (p = .005), and, the relevant finding, the need for sternal reconstruction was 3 times higher: 1.9% versus 0.6% (p = .002). For a long time, this has been regarded as an acceptable price to pay, an early morbidity that resolves and is subsequently forgotten. It is not. In a meta-analysis of 24 studies including more than 400,000 patients, deep sternal wound infection was associated with almost twice the overall mortality (IRR = 1.99; 95% CI = 1.66-2.38), a similar excess in mortality during follow-up (IRR = 2.02), and twice the rate of major cardiovascular events (IRR = 2.04). In other words, we are not comparing an uncertain benefit against trivial harm, but rather an unproven benefit against demonstrated harm with prognostic consequences. That asymmetry changes the conversation.

It is also worth considering what ART itself tells us about the long-term implications of these complications. In a subsequent analysis of the complete cohort (3102 patients; median follow-up, 10 years), 115 patients (3.7%) developed a sternal wound complication, 85 deep infections and 30 sterile dehiscences requiring reconstruction, and BITA was an independent predictor of this event (OR = 1.75; 95% CI = 1.12-2.72), together with diabetes, female sex, previous stroke, and chronic obstructive pulmonary disease. Ten-year survival was 79.4% among patients without this complication versus 64.3% among those who developed it, a 15-percentage-point difference that persisted after multivariable adjustment (HR = 1.84; 95% CI = 1.30-2.62) and in the analysis excluding the first 6 months of follow-up (HR = 1.82), arguing against this being explained solely by the acute mortality associated with the event and its treatment. It is the same trial, the same cohort, and the same follow-up underlying the HR of 1.00 for mortality: within ART, BITA did not improve survival, but it did increase the probability of a complication associated with worse survival. Gaudino himself, when discussing these data, put it plainly: the excess mortality associated with deep sternal wound infection probably offsets, or at least neutralizes, the benefit of the second arterial conduit.

Something should also be said about the ROMA trial, which is routinely invoked as the definitive answer yet to come. It has generated publications on its protocol, design, funding, and organization for years, but no clinical results. We will have to read them when they arrive; until then, decisions must be made on the evidence currently available.

What should we take into the operating room? BITA is safe in experienced hands and its use remains defensible; however, the statement that it “improves survival” can no longer be regarded as an established fact, and those of us who, like ourselves, have published data supporting that position have an obligation to say so as clearly as we did before. The decision should be individualized rather than turned into doctrine. In a young, nondiabetic patient with a good life expectancy and favorable anatomy, BITA remains a reasonable option. In a patient with 2 or more risk factors for mediastinitis, the balance has shifted, and we now have randomized data showing that avoiding BITA is not costing that patient years of life.

Perhaps the most useful interpretation is this: BITA has ceased to be a moral imperative and has become a well-tolerated technical option. Choosing whether to use it in each patient can once again be a surgical decision rather than an act of faith.

REFERENCE:

Taggart DP, Gaudino M, Gerry S, Jasinski M, Lees B, Gray AM, et al. Long-term outcomes following bilateral vs single internal thoracic artery grafts: a secondary analysis of a randomized clinical trial. JAMA Cardiol. 2026 Sep 2. doi: 10.1001/jamacardio.2026.3232.