Article Text

Download PDFPDF

Original research article
Long-term and short-term outcomes of using bilateral internal mammary artery grafting versus left internal mammary artery grafting: a meta-analysis
  1. Sana N Buttar1,
  2. Tristan D Yan1,2,
  3. David P Taggart3,
  4. David H Tian1,4
  1. 1 The Collaborative Research (CORE) Group, Macquarie University, Sydney, Australia
  2. 2 Department of Cardiothoracic Surgery, Royal Prince Alfred Hospital, University of Sydney, Sydney, Australia
  3. 3 Nuffield Department of Surgical Sciences, University of Oxford, John Radcliffe Hospital, Oxford, UK
  4. 4 Royal North Shore Hospital, Sydney, Australia
  1. Correspondence to David H Tian, Level 1, 75 Talavera Road, Macquarie Park, NSW 2113, Australia; drdavidtian{at}


Background A substantial body of evidence demonstrates that myocardial revascularisation using bilateral internal mammary arteries (BIMA) improves long-term survival compared with single/left internal mammary artery (LIMA) grafting. To date, limited analyses have been made regarding other short-term and long-term outcomes in BIMA strategy.

Objectives The primary aim of the present review is to update the difference in long-term survival between BIMA and LIMA grafting and to thoroughly investigate other secondary short-term and long-term clinical outcomes between these two grafting procedures.

Methods Electronic searches were performed using three databases from their inception to November 2015. Relevant studies comparing long-term survival between BIMA and LIMA grafting were identified. Data were extracted by two independent reviewers and analysed according to predefined clinical outcomes.

Results Twenty-nine observational studies were identified, with a total of 89 399 patients. Overall, BIMA cohort had significantly improved long-term survival compared with LIMA cohort (HR 0.78; p<0.00001). BIMA cohort also had significantly reduced hospital mortality rates (1.2% vs 2.1%, p=0.04), cerebrovascular accidents (1.3% vs 2.9%, p=0.0003) and need for revascularisation (4.8% vs 10%, p=0.005), although the incidence of deep sternal wound infection (DSWI) was increased (1.8% vs 1.4%, p=0.0008) in this grafting strategy. Long-term cardiac-free, myocardial infarction-free and angina-free survivals were also superior for the BIMA cohort.

Conclusions BIMA grafting is associated with enhanced overall long-term outcomes compared with LIMA grafting. While the BIMA cohort demonstrates an increased incidence of DSWI, the survival benefits and other morbidity advantages outweigh this short-term risk.

  • Internal mammary artery (IMA)
  • bilateral
  • coronary artery bypass graft (CABG)
  • survival
  • meta-analysis.

Statistics from


The prospect of using bilateral internal mammary artery (BIMA) grafts in coronary artery bypass grafting (CABG) surgery has been a topic of debate since single/left internal mammary artery (LIMA) grafting demonstrated an increase in long-term survival compared with venous grafting.1 Recent studies, including that of the Arterial Revascularization Trial (ART), have demonstrated similar short-term and medium-term outcomes,2–6 in contrast to the increased long-term survival advantage with BIMA grafting shown in other studies and reviews.7–12

Prior reviews have identified the superior late survival advantage in BIMA grafting; however, new studies have emerged since these reviews were conducted.9–12 In addition, existing reviews focused on long-term survival, without thorough analysis of secondary short-term and long-term patient morbidities. Despite the known survival benefits of BIMA, short-term morbidities remain a reason cited by many to avoid BIMA grafting. Therefore, the primary aim of the present meta-analysis was to evaluate long-term survival for BIMA compared with LIMA grafting, as well as clearly define secondary short-term and long-term outcomes. Further subgroup analysis will also aim to delineate outcomes in specific patient populations.


Literature search strategy

Electronic searches were performed in three databases from date of inception to November 2015: MEDLINE, EMBASE and PubMed. To achieve the maximum sensitivity of the search strategy and identify all studies, the following keywords were used: internal, mammary, double, bilateral, multiple, total or complete, single or unilateral, left or right, graft, thoracic, artery, coronary, bypass, revascularization or reconstruction, restoration, implant and unblock, as isolated words and in combination with each another, for example, ‘Mammary’ OR ‘Thoracic’ AND ‘Artery’, ‘Bilateral’ OR ‘Double’ AND ‘Coronary’ And ‘Bypass’. Additionally, database indexing terms, synonyms and abbreviations of these terms were also searched. All retrieved articles were systematically assessed by two independent researchers using the application of the predefined inclusion and exclusion criteria. The reference list of all the included articles and key review articles was further reviewed for other potentially relevant studies.


The primary clinical outcome was long-term survival. Secondary outcomes included short-term (hospital mortality, deep sternal wound infection (DSWI), re-exploration for bleeding, cerebrovascular accident (CVA), myocardial infarction (MI) and revascularisation) and long-term outcomes (cardiac event-free, MI-free and angina-free survival). Analysis was further performed for several specific clinical subgroups including propensity-matched cohorts, modern studies, diabetic cohorts, skeletonised-harvesting and pedicled-harvesting cohorts as these were deemed a priori to be clinically relevant populations. Holm’s sequential Bonferroni correction was applied for these analyses to account for multiple testing.

Long-term survival was defined as time from the surgery to death from any cause. In the majority of the studies, definition of cardiac event-free survival included no occurrence of events such as cardiac death, MI and readmission for re-CABG or percutaneous coronary intervention (PCI). Revascularisation was defined as any repeat CABG or PCI. While most studies did not define the timeline in which revascularisation was performed, others defined it as during follow-up period. Additionally, subgroup analysis was also performed for ‘modern’ studies to determine the influence of general improvements in surgical techniques and technologies. A study was defined as modern if patient enrolment commenced after 1990, so as to ensure up to 25 years of follow-up.

Selection criteria

Eligible studies were those that specifically reported long-term overall survival in patients receiving BIMA versus LIMA grafting for CABG surgery. Furthermore, publications were only included if (1) BIMA and LIMA cohort had at least 100 patients in each arm and (2) they had a median follow-up of at least 4 years. When institutions published duplicate studies with accumulating numbers of patients or increased lengths of follow-up, only the most complete reports were included for quantitative assessment. All publications were limited to human subjects and English language. Abstracts, case reports, conference presentations, editorials and expert opinions were excluded. Review articles were omitted because of potential publication bias and duplication of results.

Quality assessment of selected studies

The assessment scheme was based on the Ottawa-Newcastle system (see online supplementary figure 1). For this meta-analysis, threshold for study inclusion was a rating of six out of seven stars.

Supplementary Material

Supplementary file

Data extraction and critical appraisal

All data were extracted from text, tables and figures. Two investigators (SNB and DHT) independently reviewed each retrieved article. Discrepancies between the two reviewers were resolved by discussion and consensus. The final results were confirmed by the senior investigators (DPT and TDY).

Statistical analysis

Statistics were presented as raw values, percentages, mean or median unless otherwise indicated. As a summary statistic, HR was used for long-term outcomes, while OR was used for short-term outcomes. Random-effect model was tested as it was assumed that there were variations between studies in terms of treatment effect.13 The results using random-effect model were presented to take into account the possible clinical diversity and methodological variation between studies. I2 statistic was used to estimate the percentage of total variation across studies, owing to heterogeneity rather than chance, with values >50% considered as substantial heterogeneity. Possible clinical and methodological reasons for any substantial heterogeneity were explored qualitatively where appropriate. To determine additional factors affecting overall survival, variables with p value <0.20 on univariate regression were entered into a multivariate backwards regression model.

In addition, for long-term outcomes, individual patient survival data were reconstructed using an iterative algorithm that was applied to solve the Kaplan-Meier equations originally used to produce the published graphs. This algorithm uses digitalised Kaplan-Meier curve data to find numerical solutions to the inverted Kaplan-Meier equation.14 The algorithm assumes constant censoring and was calculated in R software (V.3.2.0). The reconstructed patient survival data were then aggregated to form combined survival curves.

Evidence of publication bias was sought using the methods of Egger et al 15 and Begg et al.16 Contour-enhancing funnel plot was performed to aid in interpretation of the funnel plot.17 All p values were two-sided. p Value<0.05 was considered representative of statistically significant, unless corrected by Holm’s sequential Bonferroni method as indicated. All statistical analyses were conducted with Review Manager V.5.2.1 (Cochran Collaboration, Software Update, Oxford, UK), or Comprehensive Meta-analysis V.3 (Biostat) or R (V.3.2.0).


Quantity and quality of evidence

A total of 3678 articles were identified. After exclusion of duplicate and irrelevant references, 120 potentially relevant articles were retrieved (see online supplementary figure 2—PRISMA diagram). According to the inclusion and exclusion criteria, detailed assessment of these texts yielded 29 studies that are included in the present meta-analysis.18–46

From the previous reviews,9–12 five studies were excluded; two studies published in 199547 and 201248 with study period of 1984–1992 and 1985–1995, respectively, failed to report how patients were assigned to the study, information on how the outcome was ascertained as well as lack of reporting on proportion lost to the follow-up. Two studies were updated with newer publications from the same centre49 50 and one study did not have an adequate follow-up period.51 Three known randomised controlled trials (RCTs) were not included in this series as two did not meet the inclusion criteria due to inadequate follow-up period and small sample size,4 5 while one was published outside of the dates stipulated in this meta-analysis protocol.6

 The quality of the studies was moderately high as per requirement of inclusion criteria.


Study characteristics are summarised in online supplementary table 1. Twenty-seven selected studies were retrospective observational studies,18–21 23–36 38–46 while two were prospective series.22 37 In the included series, 89 399 patients underwent CABG surgery, with 20 949 cases undergoing BIMA and 68 450 patients receiving LIMA grafting. Of these patients, 9311 in BIMA and 11 214 in LIMA cohort were selected by propensity matching/stratification in 12 studies.18–20 23 24 26–28 30 33 36 37 Three studies consisted of 10 687 patients with diabetes only,21 33 38 while the majority of the remaining series included mixed general population. Surgical harvesting technique was skeletonised, pedicled or a mixture of the two in seven,18 21 24 26 30 34 40 nine23 25 31 35 36 41 42 44 45 and four20 27 28 37 studies, respectively.

Weighted average follow-up of BIMA studies was 8.6 years compared with 7.0 years in the LIMA arm, with 969 766 patient-years of follow-up in total. Other demographics reported by more than half of the studies included on-pump versus off-pump cardiopulmonary bypass. Exclusion criteria are listed in online supplementary table 1.

Patient baseline, preoperative and intraoperative characteristics

Age was reported in all but two studies,20 21 with a weighted pooled average of 63.4 years (see online supplementary table 2). Female patients accounted for 15.4% and 25.4% in the BIMA and LIMA cohorts, respectively. All but one study43 reported diabetes mellitus, with an overall prevalence of 25.3% in BIMA and 38.8% in LIMA arm. Average of ejection fraction was 52.9% in both groups for all reported studies. More than half of the studies reported patients with prior MI (BIMA: 41.3%; LIMA: 39.6%), peripheral vascular disease (9.2% and 13.7% in the BIMA and LIMA cohorts, respectively) and left main disease accounting for 22.3% of BIMA and 24.5% of LIMA receiving patients. Other relevant preoperative characteristics are listed in online supplementary table 2.

Few studies reported number of coronary grafts, cardiopulmonary bypass time, cross-clamp time and the duration of postoperative hospital stay (see online supplementary table 2).

Long-term survival

Overall long-term survival for all 29 included series was significantly increased in the BIMA cohort (HR 0.78, 95% CI, 0.72 to 0.84, p<0.00001, I2=55%; table 1). Individual HRs for all studies are presented in figure 1. Of these, 12 series had a statistically significant HR that favoured BIMA grafting, while the rest showed no significant difference between the LIMA and BIMA arm. The BIMA cohort also demonstrated superior survival in propensity-matched patients (table 1 and figure 1). Reconstructed individual patient survival data from Kaplan-Meier survival curves of all 29 series demonstrated 1-year, 3-year, 5-year and 10-year survival at 97.6%, 95.1%, 92.3% and 82.1%, respectively, in BIMA and 96.1%, 92.0%, 87.0% and 70.5%, respectively, in LIMA arm (figure 2). Age (HR 1.005, 95% CI 0.986 to 1.024, p=0.609), female gender (HR 2.307, 95% CI 0.922 to 5.776, p=0.074), peripheral vascular disease (HR 1.817, 95% CI 0.389 to 8.474, p=0.448), preoperative CVA (HR 0.54, 95% CI 0.07 to 4.195, p=0.556), pedicled harvest (HR 1.067, 95% CI 0.864 to 1.318, p=0.546), incidence of diabetes (HR 0.973, 95% CI 0.745 to 1.269, p=0.839) and recruitment period (HR 1, 95% CI 0.991 to 1.01, p=0.973) were not associated with survival on meta-regression analysis. There was insufficient information provided by the included studies to analyse other covariates.

Table 1

A summary of primary long-term survival in identified studies comparing bilateral internal mammary arteries (BIMA) versus left internal mammary arteries (LIMA)

Figure 1
Figure 1

The estimate of the HR of each trail in this forest plot of HR of long-term survival in bilateral internal mammary artery (BIMA) versus left internal mammary artery (LIMA) studies corresponds to the middle of the squares, and the horizontal line shows the 95% CI. On each line, the inverse variance and study size is shown. The sum of the statistics, along with the summary HR, is represented by the middle of the solid diamond. A test of heterogeneity between the studies is given below the summary statistics. IV, inverse variance.

Figure 2
Figure 2

Reconstructed overall survival for bilateral internal mammary artery (BIMA) versus single (left) internal mammary artery (LIMA) use in coronary artery grafting. Solid line represents aggregated survival curves of propensity-matched or risk-adjusted patient cohorts, with the grey region indicating 95% CI. Dotted line represents aggregation of all available Kaplan-Meier curves, irrespective of matching or risk adjustment.

Secondary long-term outcomes

Seven studies reported cardiac event-free survival.18 26 34 37 38 41 45 Pooled overall result presents significantly reduced prevalence of cardiac events in the BIMA group (HR 0.55, 95% Cl 0.45 to 0.66, p<0.00001, I2=0%; table 2). Only diabetic subpopulation demonstrated a non-significant trend towards greater cardiac event-free survival in the BIMA cohort (table 2). Kaplan-Meier survival curves of these series demonstrate 1-year, 3-year, 5-year and 10-year cardiac event-free survival at 97.1%, 95.1%, 92.7% and 87.3%, respectively, in BIMA and 95.7%, 91.4%, 88.0% and 76.0%, respectively, in the LIMA group (figure 3A).

Figure 3

Aggregate Kaplan-Meier survival curve for (A) Cardiac-free survival with data of 1946 bilateral internal mammary artery (BIMA) and 1658 left internal mammary artery (LIMA) patients from seven studies, (B) Myocardial infarction-free survival with data of 4359 BIMA and 21 450 LIMA patients from eight studies and (C) Angina-free survival with data of 970 BIMA and 974 LIMA patients from five studies using reconstructed individual patient data derived from published Kaplan-Meier curves.

Table 2

A summary of secondary long-term outcomes in identified studies comparing long-term survival in bilateral internal mammary artery (BIMA) versus left internal mammary artery (LIMA) grafts

MI-free survival was reported in eight studies.22 34 35 37 39 40 42 46 Overall pooled risk of experiencing MI was significantly lower in BIMA than the LIMA group (HR 0.73, Cl 0.64 to 0.83, (CI 95%) p<0.00001, I2=8%; table 2). Subpopulation analysis demonstrated that this difference was not significant in modern studies (table 2). Reconstructed individual data show 1-year, 3-year, 5-year and 10-year MI-free survival at 97.3%, 95.5%, 93.7% and 88.6%, respectively, in BIMA and 96.8%, 94.7%, 92.1% and 84.9%, respectively, in the LIMA cohort (figure 3B).

Five studies reported angina-free survival.34 39 42 45 46 Overall combined result showed significant reduction of angina pectoris incidence in the BIMA group (HR 0.63, Cl 0.54 to 0.75 (CI 95%), p<0.00001; table 2). In subpopulation analysis, pedicled harvested patients maintained this trend with significant lower occurrence in the BIMA arm (table 2). Analysis of reconstructed Kaplan-Meier method presents angina-free survival of 1-year, 3-year, 5-year and 10-year at 98.4%, 95.8%, 90.4% and 75.8%, respectively, in BIMA and 96.4%, 92.5%, 86.8% and 65.6%, respectively, in the LIMA arm (figure 3C).

Secondary short-term outcomes

DSWI was reported in 17 studies (see online supplementary table 3). Overall combined result demonstrated a significant greater risk of DSWI in BIMA than the LIMA cohort (1.8% vs 1.4%, OR 1.37, p=0.0008; (table 3). Overall incidence of hospital mortality, CVA, MI and revascularisation were significantly lower by 21% (1.2% vs 2.1%, p=0.04), 36% (1.3% vs 2.9%, p=0.0003), 23% (2.02% vs 2%, p=0.006) and 24% (4.8% vs 10%, p=0.005) in BIMA than the LIMA arm, respectively (see online supplementary table 3 and table 3). No significant difference was observed in overall re-exploration for bleeding (2.9% vs 3.2%, p=0.51).

Table 3

A summary of short-term outcomes in identified studies comparing long-term survival in bilateral internal mammary artery (BIMA) versus left internal mammary artery (LIMA) grafts

Among propensity-matched studies, only DSWI demonstrated a significant difference between the BIMA and LIMA groups. Results of other subpopulation analyses are shown in table 3.

Publication bias

All 29 studies yielded a Begg’s test score of p=0.680 and Egger’s test score of p=0.884 when assessing long-term survival, while inspection of the contour-enhanced funnel plot revealed absence of publication bias in all 29 studies (see online supplementary figure 3A). However, analysis of short-term mortality using funnel plot demonstrated publication bias with more favourable LIMA results (Begg’s p=0.352; Egger’s p=0.019) (see online supplementary figure 3B). No evidence of publication bias was identified with any other short-term or long-term outcomes.


Existing observational data have demonstrated that BIMA grafting is associated with better late survival compared with LIMA grafting. In the light of the mounting evidence, the recent Society of Thoracic Surgeons Clinical Practice Guidelines published in 2016 on conduit selection recommended that a second skeletonised IMA conduit should be used in patients not at excessive risk of sternal complications (Classification of Recommendation IIa, Level of Evidence B).52 However, despite continuous accumulating evidence in favour of BIMA strategy, the majority of the surgeons are still reluctant to make it a routine procedure. Indeed, continental databases have shown that BIMA grafting is used in only 4% of CABG surgeries in the USA53 and 12% in Europe.54

The present meta-analysis reassessed the late survival benefit between BIMA and LIMA grafting strategies and found that BIMA grafting was associated with superior overall survival. In addition, the incidence of late cardiac events, MI and angina pectoris were also observed to be lower in the BIMA cohort. Such findings are not unexpected—the superior patency rates of arterial grafts are well known, especially that of IMA grafts. Therefore, using two IMA grafts with biological similarities could only be anticipated to enhance the outcomes. However, patient selection and study heterogeneity must also be taken into consideration when evaluating such findings. On average, patients allocated to the BIMA cohort were younger, had slightly better ventricular function and less proportion of diabetes as well as lower prevalence of other relevant preoperative risk factors, thus better outcomes of BIMA grafting may also be attributable to these factors. Nevertheless, in the propensity-matched cohort, where the influence of above factors is greatly mitigated, such findings remain present, supporting the beneficial effects of BIMA grafting in more comparable cohorts.

The question as to why multiarterial grafting with BIMA is not used more frequently still remains puzzling. Numerous authors have cited increased risk of DSWI as a main key concern, particularly in patients with diabetes, obesity and chronic obstructive pulmonary disease.19 55–57 Moreover, the technical complexity of the harvesting technique, the longer operative time associated with it and potential increase in short-term morbidities are also seen as obstacles that prevent widespread uptake of BIMA grafting.58 In the present analysis, BIMA harvest was indeed associated with an increased incidence of DSWI in the overall unmatched cohort, propensity-matched patients and also in patients with diabetes, which is in line with previous reports, in particular, the ongoing ART, which also found the rates of sternal wound complications at 5 years6 and sternal wound reconstruction at 1 and 5 years5 6 to be increased in the BIMA strategy. The well-known reason for this is an acute postoperative de-vascularisation of sternum, which is exacerbated after bilateral IMA harvesting, and thereby increasing the risk of DSWI.59 Deep infection can be thought to increase the risk of sternal necrosis with greater need for sternal resection. Although the occurrence of DSWI is relatively low in the present study (1.8% vs 1.4%, p=0.0008), DSWI serves as a continuing detrimental risk factor for long-term survival.60 61 It should be stressed that even though patients with diabetes are more prone to sternal infection and its poorer healing, they may actually have the most to gain from BIMA grafting as they often present with severe and diffuse coronary artery disease.57 62

Among surgeons, skeletonised harvesting is a well-recognised method to reduce the risk of DSWI associated with BIMA strategy.63 However, concerns of compromised flow capacity and patency of a skeletonised graft, thus fear of poorer end results, have limited its use. Recent studies demonstrate that these concerns are negligible in comparison to pedicle BIMA grafts.64 65 In the present analysis, among patients who underwent skeletonised harvesting, BIMA grafting was shown to increase the odds of causing DSWI compared with LIMA grafting (OR 1.48, 95% Cl 1.12 to 1.94, p=0.005), with minor differences in the actual incidence of DSWI (1.90% vs 1.80%, p=0.005, BIMA vs LIMA). Taken together, these results suggest that for patients who underwent skeletonised IMA harvesting, BIMA grafting has almost negligible clinical difference in terms of DSWI compared with LIMA grafting.

The apparent discrepancy between clinical evidence and clinical practice of BIMA grafting reflects the necessity to promote evidence-based medicine, rather than individual preferences, with further research required to delineate the specific populations, which benefit from BIMA and those for whom BIMA should be avoided. The role of various BIMA configurations, harvesting techniques and the location of the second IMA on clinical outcomes need to be clearly examined, as do short-term outcomes, since most surgeons appear to be more inclined to be influenced by these rather than long-term concerns.66 The importance of allowing younger surgeons to perform BIMA grafting must also be realised as it is necessary for their learning curve and confidence building so that they may consider performing this procedure in future.


First, no existing RCTs qualified to be included in this review. Significant confounding factors and intrinsic patient selection bias associated with included reports weakened the overall pooled effect of this meta-analysis. Second, direct comparison of BIMA and LIMA cohorts should be made carefully as BIMA grafting was typically preferred for those with reduced preoperative risk profile, thus, naturally, more favourable cohort to survive longer regardless of grafting strategy. Further factors, which may have influenced patient selection, including frailty, the quality and size of target vessels, the severity and distribution of stenoses and the quality of conduit, are difficult to match. Finally, the heterogeneity may also reflect the variations in experience and expertise of different centres, given that BIMA grafting is a complex procedure. The ongoing ART will assist to overcome some of the above limitations.5 6


In summary, overall BIMA grafting is associated with superior overall late survival as well as increased overall cardiac event-free, MI-free and angina-free survival. Late survival was not associated with relevant covariates on meta-regression analysis. BIMA grafting does increase the risk of DSWI in general and diabetic population, although this is an infrequent risk. In light of these findings, the long-term benefits of BIMA outweigh its short-term risks, and it should be a more routinely used strategy for CABG surgery.

Key questions

What is already known on this subject?

Long-term survival rate is better in bilateral internal mammary artery (BIMA) than left internal mammary artery (LIMA) grafting. Risk of deep sternal wound infection is increased in BIMA grafting, while reporting of other short-term outcomes is heterogeneous and inconsistent.

What might this study add?

BIMA grafting is not only associated with better long-term survival (HR 0.78, p<0.00001) but also increased secondary long-term outcomes. Rate of hospital mortality (OR 0.79, p=0.04), cerebrovascular accident (OR 0.64, p=0.0003) and need for revascularisation (OR 0.76, p=0.005) was lower, while incidence of deep sternal wound infection was higher in BIMA grafting (OR 1.37, p=0.0008).

How might this impact on clinical practice?

This large meta-analysis encourages greater use of BIMA grafting due to its superior long-term and short-term outcomes in selected patients.



  • Contributors SNB and DHT had full access to all of the data in the study and take responsibility for the integrity of the data and accuracy of the data analysis.

    Study concept and design: SNB, TDY and DHT.

    Analysis and interpretation of data: SNB and DHT.

    Drafting of the manuscript: SNB.

    Critical revision of the manuscript for important intellectual content: TDY, DPT and DHT. Statistical analysis: SNB and DHT.

    Study supervision: TDY.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles