CTA indicates computed tomography angiogram; IMA, internal mammary artery; LIMA, left internal mammary artery; RIMA, right internal mammary artery.
A, Major adverse cardiovascular events (MACEs) represent a composite outcome of cardiovascular death, myocardial infarction, stroke, or revascularization. B, Revascularization (all revascularization procedures were done by percutaneous coronary intervention).
eTable 1. Baseline Characteristics by Patient Group
eTable 2. Clinical Outcomes at the End of the Study by IMA Harvesting Technique for 1256 Patients With CTA Included or Without CTA Who Received an IMA
eTable 3. Clinical Outcomes at the End of the Study by IMA Harvesting Technique for 254 Patients Without CTA Who Received an IMA
eTable 4. Clinical Outcomes at the End of the Study by IMA Graft Status
Data Sharing Statement
Customize your JAMA Network experience by selecting one or more topics from the list below.
Lamy A, Browne A, Sheth T, et al. Skeletonized vs Pedicled Internal Mammary Artery Graft Harvesting in Coronary Artery Bypass Surgery: A Post Hoc Analysis From the COMPASS Trial. JAMA Cardiol. 2021;6(9):1042–1049. doi:10.1001/jamacardio.2021.1686
Is skeletonized vs pedicled harvesting of the internal mammary artery associated with higher rates of graft occlusion and clinical outcomes after coronary artery bypass graft surgery?
In this post hoc analysis that included 1002 patients undergoing coronary artery bypass graft surgery in the Cardiovascular Outcomes for People Using Anticoagulation Strategies (COMPASS) randomized clinical trial, the rate of internal mammary artery graft occlusion at 1 year was 9.6% with skeletonized harvesting vs 3.9% with pedicled harvesting, a significant difference.
In this study, skeletonized harvesting of the internal mammary artery was associated with a higher rate of occlusion and worse clinical outcomes; further investigation of the skeletonized technique in a randomized clinical trial is needed.
The relative safety and patency of skeletonized vs pedicled internal mammary artery grafts in patients undergoing coronary artery bypass graft (CABG) surgery are unknown.
To investigate the association of skeletonized vs pedicled harvesting with internal mammary artery graft patency and clinical outcomes 1 year after CABG surgery.
Design, Setting, and Participants
This study was a post hoc analysis of the multicenter, randomized, double-blind, placebo-controlled Cardiovascular Outcomes for People Using Anticoagulation Strategies (COMPASS) clinical trial, which enrolled 27 395 patients from 602 centers in 33 countries from March 2013 through May 2016. Eligibility criteria for the trial included CABG surgery for coronary artery disease with at least 2 grafts implanted and an estimated glomerular filtration rate of at least 30 mL/min. A total of 1002 of 1448 patients were randomized to the CABG arm of the COMPASS trial and underwent skeletonized (282 [28.1%]) or pedicled (720 [71.9%]) internal mammary artery harvesting. The patients had evaluable angiography results 1 year after surgery. Data were analyzed from October 11, 2019, to May 14, 2020.
Patients underwent graft harvesting with either the pedicled technique or skeletonized technique.
Main Outcomes and Measures
The primary outcome was graft occlusion 1 year after CABG surgery, as assessed by computed tomography angiography.
A total of 1002 patients underwent skeletonized (282 [28.1%]; mean [SD] age, 65.9 [8.1] years; 229 men [81.2%]; 194 White patients [68.8%]) or pedicled (720 [71.9%]; mean [SD] age, 64.8 [7.6] years; 603 men [83.8%]; 455 White patients [63.2%]) internal mammary artery harvesting. Rates of internal mammary artery graft occlusion 1 year after CABG surgery were higher in the skeletonized group than in the pedicled group (33 of 344 [9.6%] vs 30 of 764 [3.9%]; graft-level adjusted odds ratio, 2.41; 95% CI, 1.39-4.20; P = .002), including the left internal mammary artery to left anterior descending artery (21 of 289 [7.3%] vs 25 of 725 [3.4%]; graft-level adjusted odds ratio, 2.10; 95% CI, 1.14-3.88, P = .02). After a mean follow-up of 23 months, skeletonized graft harvesting was also associated with a higher rate of major adverse cardiovascular events (20 [7.1%] vs 15 [2.1%]; adjusted hazard ratio, 3.19; 95% CI, 1.53-6.67; P = .002) and repeated revascularization (14 [5.0%] vs 10 [1.4%]; adjusted hazard ratio, 2.75; 95% CI, 1.10-6.88; P = .03).
Conclusions and Relevance
This post hoc analysis of the COMPASS randomized clinical trial found that harvesting of the internal mammary artery during CABG surgery using a skeletonized technique was associated with a higher rate of graft occlusion and worse clinical outcomes than the traditional pedicled technique. Future randomized clinical trials are needed to establish the safety and patency of the skeletonized technique.
ClinicalTrials.gov Identifier: NCT01776424
Coronary artery bypass grafting (CABG) is one of the most commonly performed surgical procedures and improves clinical outcomes in appropriately selected patients.1 The greater saphenous vein is often used in CABG, but arterial conduits, specifically the left internal mammary artery (LIMA), are used routinely owing to higher long-term patency rates.2,3 The use of the right internal mammary artery (RIMA) is gaining in popularity and has been included routinely in large multiarterial grafting trials, such as the Arterial Revascularization Trial (ART)4,5 and the currently ongoing Randomized Comparison of the Clinical Outcome of Single vs Multiple Arterial Grafts (ROMA) trial.6
Traditionally, the internal mammary artery (IMA) (LIMA, RIMA, or both) is harvested under direct vision as a pedicle (en bloc) with the help of linear incisions along the course of the IMA. The pedicle is 1 to 2 cm wide and contains the IMA, veins, fascia, and nerve. Harvesting of the IMA with the more difficult skeletonized technique (direct dissection) rather than the usual pedicled technique is frequently performed to improve IMA flow and facilitate the use of bilateral IMAs during CABG surgery. However, little is known about the effect of skeletonization of the IMA on long-term graft patency or on clinical outcomes. As the skeletonized technique is more prone to injuries to the IMA, a higher occlusion rate could have important consequences for patients. The objectives of this observational analysis using the COMPASS trial data set were to assess the effect of using a skeletonized or pedicled IMA on graft patency, as evaluated by computed tomography angiography (CTA) 12 months after CABG, and on clinical outcomes.
The COMPASS study was a double-blind randomized clinical trial with a 3 × 2 partial factorial design, conducted from March 2013 through May 2016, and involving 27 395 patients with stable coronary artery disease or peripheral artery disease from 602 centers in 33 countries.7 The trial evaluated rivaroxaban 2.5 mg administered twice daily in combination with aspirin, 100 mg once daily; rivaroxaban, 5 mg twice daily with matched placebo once daily; and aspirin, 100 mg administered once daily with matched placebo twice daily for the prevention of the composite of cardiovascular (CV) death, stroke, or myocardial infarction (MI) among those with a history of stable coronary artery disease or peripheral artery disease. The antithrombotic comparisons including the CABG substudy were stopped early (mean follow-up of 23 months) for efficacy after a recommendation by the independent data and safety monitoring board. The combination of rivaroxaban and aspirin compared with aspirin alone reduced the primary outcome by 24% (hazard ratio [HR], 0.76; 95% CI, 0.66-0.86; P < .001).8
A predefined group of patients undergoing CABG surgery were eligible for the COMPASS study if they had at least 2 grafts implanted and an estimated glomerular filtration rate of at least 30 mL per minute. Eligible participants were randomized between 4 and 14 days after CABG surgery in a 1:1:1 ratio.9 The primary outcome of COMPASS CABG was the proportion of coronary bypass grafts that had failed with complete occlusion of the graft (graft-level analysis). The secondary outcome was the proportion of patients with a failed graft (patient-level analysis). Graft failure was assessed with CTA at 1 year after surgery, or with conventional coronary angiography if this had been performed for a clinical indication. Two experienced readers (including T.S.) blinded to all patient data evaluated the CTA results in a core laboratory. Grafts were evaluated for image quality and graft patency10 and categorized as patent or occluded. The main trial protocol was approved by the relevant health authorities and institutional review boards. All participants provided written informed consent, and self-identified race/ethnicity was collected by research staff to evaluate potential differences in drug treatment effects by racial/ethnic group. This study followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline.
The combination of rivaroxaban and aspirin or rivaroxaban alone did not reduce the graft occlusion rates compared with aspirin alone (combination vs aspirin, 113 [9.1%] vs 91 [8.0%] occluded grafts; odds ratio [OR], 1.13; 95% CI, 0.82-1.57; P = .45; rivaroxaban alone vs aspirin, 92 [7.8%] vs 92 [8.0%] occluded grafts; OR, 0.95; 95% CI, 0.67-1.33; P = .75). Graft characteristics, such as the quality of the conduit itself and the quality of the target vessel (native artery), were reported by the surgeons at the time of the CABG surgery as being excellent or acceptable. From the COMPASS CABG data set, we extracted the exact graft location, conduit type, and method of harvesting for veins (open or minimally invasive) and the IMA (pedicled or skeletonized). The proximal anastomosis site and the distal anastomosis sites were also collected, as was the highest stenosis of the target vessel. The overall occlusion rate of all grafts combined was 8.3%, with an occlusion rate of 4.5% for LIMA, 8.6% for radial arteries, 9.6% for veins, and 21.4% for RIMA.
The analyses were conducted for all patients in the COMPASS study who had CABG and underwent CTA and received skeletonized or pedicled IMA harvesting. Baseline characteristics were compared among patients who received a skeletonized vs a pedicled IMA, using a t test for continuous variables and a χ2 test for categorical variables.
Both graft-level and patient-level results were adjusted for drug treatment allocation and propensity score (age, cholesterol, hypertension, angiotensin-converting enzyme inhibitor or angiotensin II receptor blocker, calcium channel blocker, diuretic, β-blocker, lipid-lowering agent, nonsteroid anti-inflammatory drug, and nontrial proton pump inhibitor). For the graft-level analysis, the primary outcome of graft occlusion was analyzed using linear mixed effects logistic regression with logit link and random effects to account for repeated measures. For the patient-level analysis (including clinical events), the number of patients with an occluded graft or who experienced a clinical event were evaluated using negative binomial models with a logarithmic link and random effects to account for the center. In addition, a similar analysis was conducted separately for the patients in whom the LIMA was used, those in whom the LIMA to the left anterior descending artery (LAD) was used, and those in whom the RIMA was used. The ORs are reported for graft-level analysis, and the risk ratios are reported for patient-level analysis along with the corresponding 95% CIs. For the clinical outcomes, the composite of CV death, MI, stroke, or revascularization was assessed using Cox regression models, with HRs and the corresponding 95% CIs reported. Finally, we present the clinical outcomes at the end of the study by IMA graft status (at least 1 occluded vs none occluded). All P values were 2-sided, and P < .05 was considered significant. Statistical analyses were conducted from October 11, 2019, to May 14, 2020, using SAS software, version 9.4 (SAS Institute).
Of the 27 395 individuals enrolled in COMPASS, 1448 patients were randomized between 4 and 14 days after CABG surgery. Of the 1448 patients, 306 (21.2%) did not complete a CTA. Of the remaining 1142 patients, 140 (12.3%) were not included in our grafts analysis because they did not receive an IMA (89 [63.6%]), the CTA was not evaluable (14 [10.0%]), or the harvesting technique could not be confirmed (37 [26.4%]) (Figure 1). Baseline characteristics of the 140 patients (12.3%) who underwent a CTA and were excluded were similar to those who were included (1002 [87.7%]), although excluded patients were older, smoked tobacco less frequently, and were less frequently of White race/ethnicity (Table 1; eTable 1 in Supplement 1). Of the 1002 included patients, the baseline characteristics of those who received a skeletonized IMA (282 [28.1%]; mean [SD] age, 65.9 [8.1] years; 229 men [81.2%]; 194 White patients [68.8%]) were generally similar to patients who received a pedicled IMA (720 [71.9%]; mean [SD] age, 64.8 [7.6] years; 603 men [83.8%]; 455 White patients [63.2%]), but those who received a skeletonized IMA were older, had higher cholesterol levels (mean cholesterol [SD], 4.1 [1.2] mmol/L vs 3.9 [1.2] mmol/L) and more hypertension (229 [81.2%] vs 524 [72.8%]), and had different medication profiles. Of the 306 patients who did not complete a CTA scan, 256 (83.7%) received an IMA graft, and the harvesting technique was confirmed. Baseline characteristics of these patients are presented in eTable 1 in Supplement 1.
Among 1002 patients, 1108 IMA grafts were evaluated, of which 63 (5.7%) were occluded (Table 2). The primary outcome of graft occlusion occurred in 33 of 344 IMA grafts (9.6%) in the skeletonized group compared with 30 of 764 IMA grafts (3.9%) in the pedicled group (adjusted OR, 2.41; 95% CI, 1.39-4.20; P = .002). Analysis of the secondary outcome of patient-level graft occlusion showed that 57 of 1002 patients (5.7%) had at least 1 IMA graft occluded. The secondary outcome occurred in 28 of 282 patients (9.9%) in the skeletonized group compared with 29 of 720 patients (4.0%) in the pedicled group (adjusted risk ratio, 2.90; 95% CI, 1.67-5.05; P < .001).
At the end of the trial, patients with skeletonized grafts had a higher risk of major adverse cardiovascular events (MACEs)—CV death, MI, stroke, or revascularization—than patients who received pedicled grafts (20 events [7.1%] vs 15 events [2.1%]; adjusted HR, 3.19; 95% CI, 1.53-6.67; P = .002) (Figure 2A and Table 3). All revascularization procedures were done by percutaneous coronary intervention. Repeated revascularization was required in 14 of 282 patients (5.0%) with skeletonized grafts compared with 10 of 720 patients (1.4%) with pedicled grafts (adjusted HR, 2.75; 95% CI, 1.10-6.88; P = .03) (Figure 2B and Table 3). The primary clinical outcomes of the main COMPASS trial (CV death, MI, or stroke) were similar for patients who received a skeletonized IMA graft compared with patients who received a pedicled IMA graft (adjusted HR, 2.50; 95% CI, 0.89-6.98; P = .08) (Table 3).
When all patients with (1002 patients) and without (254 patients) CTA who received an IMA with a confirmed harvesting technique were included in the analysis (1256 patients total), similar clinical results were obtained (eTable 2 in Supplement 1). The results of the 254 patients without CTA are also provided separately (eTable 3 in Supplement 1).
For LIMA use, graft occlusion occurred in 22 of 300 grafts (7.3%) in the skeletonized group compared with 25 of 732 grafts (3.4%) in the pedicled group (adjusted OR, 2.13; 95% CI, 1.16-3.91; P = .02) (Table 2). Specifically, the LIMA to LAD graft had a higher rate of occlusion (7.3% vs 3.4%; adjusted OR, 2.10; 95% CI, 1.14-3.88, P = .02). For RIMA use, graft occlusion occurred in 11 of 44 grafts (25.0%) in the skeletonized group compared with 5 of 32 grafts (15.6%) in the pedicled group (adjusted OR, 2.88; 95% CI, 0.62-13.49; P = .18). A higher rate of occlusion with skeletonized vs pedicled IMA was seen with regard to the quality of the conduit, quality of the target, proximal and distal anastomosis site, and the proximal target vessel stenosis (Table 2). As the graft characteristics are not independent of each other, these results should be interpreted globally rather than individually.
Compared with patients without IMA graft occlusion, patients with at least 1 occluded IMA graft were more likely to experience a MACE (adjusted HR, 4.62; 95% CI, 2.17-9.83; P < .001), revascularization (adjusted HR, 5.52; 95% CI, 2.29-13.31; P < .001), or hospitalization for CV causes (adjusted HR, 2.72; 95% CI, 1.69-4.38; P < .001) at the end of the study (eTable 4 in Supplement 1).
Our results, based on data from 68 academic centers in 25 countries around the world, indicate that the rate of graft occlusion of skeletonized IMAs was higher than that of pedicled IMAs. This lower patency was seen in LIMA grafts overall but also with the important LIMA graft to the LAD artery. Skeletonized IMAs were also associated with a higher rate of clinical events and revascularization procedures.
Skeletonized IMA grafts are reported to provide a lengthier conduit (length increased by 3.7 cm)11 to reach more targets with higher blood flow11,12 and possibly fewer sternal infections.13,14 The skeletonized IMA technique requires dissection and ligation of branches directly on the IMA rather than dissecting the IMA en bloc as done in the pedicled IMA technique, which avoids direct contact with the IMA itself. This close dissection is possibly associated with trauma to the IMA, thereby increasing the risk of thrombus, laceration, or lumen narrowing that may lead to graft occlusion.
Few studies12,15-19 have directly compared the results of pedicled and skeletonized harvesting techniques with regard to early graft occlusion after cardiac surgery. Except for a small trial of 200 patients, all other studies were nonrandomized and are from centers favoring the skeletonized over pedicled technique in their practice (two-thirds of patients received a skeletonized IMA over a pedicled IMA). In a meta-analyses of 5 studies with a total of 1988 patients (1764 grafts), the authors concluded that, when compared with pedicled grafts, the odds of occlusion of skeletonized grafts were higher, but this difference was not significant (OR, 1.35; 95% CI, 0.41-4.47; P = .80).20
Subsequently, a large nonrandomized study11 comparing skeletonized and pedicled harvesting techniques reported similar rates of graft occlusion 1 year after off-pump coronary artery bypass surgery among skeletonized (9 of 778 [1.2%]) and pedicled (12 of 795 [1.5%]) LIMA grafts (P = .47). In this single-center study, patients were not randomized to the harvesting technique but were recruited consecutively, because patients in the first recruitment period received only a pedicled IMA, and patients in the second recruitment period received only a skeletonized IMA. Angiographic follow-up was significantly shorter in the skeletonized IMA group (mean [SD] time, 19.4 [7.0] months vs 40.0 [15.4] months; P < .001). Among 1859 patients (903 skeletonized IMAs, 956 pedicled IMAs) included in the follow-up, there was a trend toward excess mortality (log-rank test, P = .12) and a higher rate of cardiac death, MI, and surgical reoperation (log-rank test, P = .13) in the skeletonized IMA group. In our analysis, the time-to-event curves for MACE were still diverging beyond 2 years. Therefore, it is possible that a longer follow-up in their study11 might have detected significant inferior results in the skeletonized IMA group.
Skeletonized IMA grafts help surgeons to perform more bilateral IMA grafting with a RIMA graft on a lateral or inferior wall target vessel. Bilateral internal mammary artery grafting was not superior to a single IMA graft after 5 or 10 years of follow-up in a large randomized trial (Arterial Revascularization Trial [ART]),4,5 and bilateral IMA was associated with more sternal infections than a single LIMA, but no angiographic follow-up was done in ART.4,5,21 The use of skeletonized IMAs by some investigators may have influenced the results. No difference in patency of the RIMA was seen in our study, but few RIMAs were used, which limited the statistical power, and the absolute rate of occlusion was 9.4% higher with skeletonized RIMA than pedicled RIMA grafts (25.0% compared with 15.6%). The RIMA grafts performed poorly in comparison with vein grafts (9.6%) and radial artery grafts (8.6%). Furthermore, the poor patency of RIMA grafts (skeletonized or pedicled) seen in our analysis was worrisome and could have been a major confounding variable in ART as well. The ROMA trial is similar in design to ART but allows for a radial artery or a RIMA as the second arterial graft. The ROMA trial6 has no angiographic component and may have the same confounding variables. The poor performance of RIMA grafts (both techniques) should be further evaluated in large studies.
An important objective of CABG surgery is to improve event-free survival. Although it has been assumed that graft occlusion is a major determinant of clinical events, this viewpoint remains controversial. We found that patients undergoing skeletonized harvesting of the IMA had a higher risk of a MACE (composite of CV death, MI, stroke, or revascularization) compared with patients undergoing pedicled IMA harvesting, which was mainly driven by stroke and revascularization. We also found that patients with at least 1 occluded IMA graft had a higher risk of a MACE, which was driven mainly by revascularization. The latter results were consistent with those of the large (n = 1539 patients) Project of Ex-Vivo Vein Graft Engineering via Transfection IV (PREVENT IV) trial, which reported a higher rate of acute MACEs (composite of death, MI, or repeated revascularization) in patients with IMA graft failure than in patients without IMA graft failure (14.4% vs 4.9%; adjusted HR, 3.92; 95% CI, 2.30-6.68; P < .001), mostly owing to a higher rate of repeat revascularization.22 Together, these results suggest that the skeletonized technique increases the risk of early IMA occlusion and consequently the risk of revascularization by percutaneous coronary intervention. Future studies should clarify whether occlusion of grafts terminating on the lateral wall are associated with MACEs and whether conduit choice (saphenous vein, RIMA, or radial artery) influences this relationship.
Our analysis had several limitations. Our investigation was not prespecified but was prompted by a report from a secondary analysis of the PREVENT IV trial23 comparing different methods of vein harvesting, which showed worrisome results.24 This was not a randomized clinical trial but a post hoc observational study nested within a large and rigorous randomized clinical trial, and there were risks of selection biases. The imaging follow-up was not evaluated in 446 of 1448 patients (30.8%), but their clinical results were similar to the cohort of patients who received CTA. The baseline characteristics of the patients were also similar, and the results were adjusted with a propensity score. In addition, the majority of patients were recruited and randomized in COMPASS after their CABG surgery, which minimized treatment allocation bias. In observational data, causal estimation relies on the (untestable) assumption of “no unobserved confounding”; therefore, such bias was a risk in this post hoc analysis. We did not collect information about the surgeons or their level of experience. The surgical technique was not randomized, and surgeons used the technique of their choosing. Furthermore, the patency results of veins and LIMA grafts were similar to or better than those of other randomized clinical trials with patency, such as Randomized On/Off Bypass (ROOBY),25 PREVENT IV,23 Danish On-Pump Vs Off-Pump Randomization Study (DOORS),26 and Different Antiplatelet Therapy Strategy After Coronary Artery Bypass Grafting (DACAB).27 Finally, there was a substantial increase in the stroke rate in the skeletonized group, but these results should be interpreted with caution.
Many surgeons skeletonize IMA grafts in their surgical routine. However, the results of this post hoc analysis of a randomized clinical trial suggest that the potential clinical benefits of skeletonized IMA grafts are potentially thwarted by higher graft occlusion (including the important graft of LIMA to LAD artery), worse clinical outcomes, and more revascularization procedures. A randomized clinical trial is needed to address the safety concerns regarding the skeletonized technique.
Accepted for Publication: April 14, 2021.
Published Online: June 16, 2021. doi:10.1001/jamacardio.2021.1686
Correction: This article was corrected on August 18, 2021, to fix the spelling of author Faisal G. Bakaeen’s surname.
Corresponding Author: André Lamy, MD, Population Health Research Institute, 20 Copeland Ave, Room C1-112, Hamilton, ON L8L 2X2, Canada (firstname.lastname@example.org).
Author Contributions: Drs Lamy and Lee had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Lamy, Dagenais, Bakaeen, Alboom, Salim.
Acquisition, analysis, or interpretation of data: Lamy, Browne, Sheth, Zheng, Noiseux, Chen, Bakaeen, Brtko, Stevens, Alboom, Lee, Copland, Salim, Eikelboom.
Drafting of the manuscript: Lamy, Alboom.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Alboom, Lee, Eikelboom.
Obtained funding: Lamy, Salim.
Administrative, technical, or material support: Lamy, Browne, Sheth, Zheng, Chen, Bakaeen, Brtko, Copland.
Supervision: Lamy, Noiseux, Brtko, Stevens, Salim.
Other – Screening and investigation of the patients: Brtko.
Conflict of Interest Disclosures: Dr Eikelboom reported receiving grants and personal fees from Bayer, Boehringer Ingelheim, Bristol Myers Squibb, Pfizer, Daiichi Sankyo, Janssen, AstraZeneca, Eli Lilly, GlaxoSmithKline, and Sanofi-Aventis and personal fees from Servier outside the submitted work. Dr Salim reported receiving grants and personal fees from Bayer, Boehringer Ingelheim, AstraZeneca, Bristol Myers Squibb, and Cadila Pharmaceuticals and reported receiving consultant and speaker fees and travel expenses from Bayer. No other disclosures were reported.
Group Information: The COMPASS Investigators appear in Supplement 3.
Data Sharing Statement: See Supplement 2.