Article Text

Original article
Bilateral internal mammary artery grafts, mortality and morbidity: an analysis of 1 526 360 coronary bypass operations
  1. Shinobu Itagaki1,
  2. Paul Cavallaro2,
  3. David H Adams1,
  4. Joanna Chikwe1
  1. 1Department of Cardiothoracic Surgery, The Mount Sinai Medical Center, New York, New York, USA
  2. 2Department of Cardiothoracic Surgery, The Mount Sinai School of Medicine, New York, New York, USA
  1. Correspondence to Dr Joanna Chikwe, Department of Cardiothoracic Surgery, Mount Sinai Medical Center, 1190 Fifth Avenue, New York, NY 10029, USA; Joanna.chikwe{at}


Objectives The objective of this study was to investigate the impact of bilateral internal mammary artery (BIMA) on early outcomes after coronary artery bypass grafting.

Design Retrospective database analysis.

Setting US hospitals.

Patients 1 526 360 patients (mean age 65 years, 73% male) from the Nationwide Inpatient Sample from 2002–2008 who underwent isolated coronary artery bypass grafting with at least one internal mammary artery.

Interventions Single versus BIMA bypass grafting.

Main outcome measures Inhospital mortality, deep sternal wound infection (DSWI).

Results The rate of BIMA use was 3.9%. Use of BIMA was independently associated with slightly lower inhospital mortality (unadjusted rate 1.1% vs 1.7%, adjusted OR 0.86, 95% CI 0.79 to 0.93). The DSWI rate was 1.4%. The independent predictors of DSWI were female gender (OR 1.06), congestive heart failure (OR 6.22), chronic pulmonary disease (OR 1.57), obesity (OR 1.17), diabetes mellitus (OR 1.04; OR 1.51 with chronic complication) and chronic renal failure (OR 2.13; OR 2.63 with dialysis). The use of BIMA was not an independent predictor of DSWI (OR 1.03, 95% CI 0.96 to 1.10). BIMA was associated with higher incidence of DSWI in patients with chronic complications of diabetes mellitus (OR 1.90, 95% CI 1.51 to 2.41).

Conclusions BIMA grafting is associated with increased risk of DSWI only in patients with severe, chronic diabetes. The incremental morbidity and mortality of DSWI does not justify denial of BIMA in the majority of patients.


Statistics from


The left internal mammary artery is the conduit of choice to bypass the left anterior descending coronary artery in coronary artery bypass graft surgery (CABG).1 The best conduit to bypass the circumflex or right coronary artery remains controversial. Since its first clinical application in the 1970s,2 observational data suggest that the right internal mammary artery may have superior patency and associated survival benefits compared with vein and radial artery grafts.3–14 In the 2011 American College of Cardiology/American Heart Association guidelines, bilateral internal mammary arteries (BIMA) use was a Class IIa recommendation.1 BIMA is used relatively infrequently in the USA (<5% of isolated coronary bypass operations),15–17 and one of the main reasons for this is evidence suggesting increased risk of deep sternal wound infection (DSWI) with BIMA use, particularly in high-risk patients with diabetes mellitus18–20 but data is conflicting.6 ,7 ,18–22 The present study was therefore designed to investigate the effect of BIMA grafting on the rate of DSWI after CABG using national outcome data from the Nationwide Inpatient Sample (NIS).


Data source and population

The NIS was used to identify patients who underwent isolated CABG between 2002 and 2008. The NIS is the largest all-payer inpatient care database and contains discharge data from over eight million hospitalisations per year, which is approximately 20% of all hospitalisations in the USA. International Classification of Disease, the ninth edition (ICD-9) diagnosis codes were used to identify all patients in NIS who underwent CABG (36.10, 36.11, 36.12, 36.13, 36.14, 36.15, 36.16). Patients who underwent concomitant valve procedures (35.20, 35.21, 35.22, 35.23, 35.24, 35.25, 35.26, 35.27, 35.28, 35.11, 35.12, 35.13, 35.14) were excluded. Patients were stratified into two groups based on whether only single internal mammary artery (SIMA) was used or BIMA was used, using ICD-9 procedure codes of SIMA use (36.15) and BIMA use (36.16). Patients who did not receive SIMA or BIMA were excluded. A total of 1 526 360 patients were included in the analysis. Patient demographics, comorbidities and provider characteristics are listed in table 1. High-risk subgroups of diabetes mellitus (diabetes mellitus with chronic complications) and renal failure (end-stage renal failure requiring dialysis) were also identified. The ICD-9 diagnosis codes used to code for these comorbidities, based on Agency for Healthcare Research and Quality comorbid disease categories23 are provided in online supplementary table S1.

Table 1

Patient and provider characteristics

To include the effect of surgical volume on outcomes, annual surgeon CABG volume was also obtained. Eleven states specifically reported unique physician identifiers for the study time period: Florida, Iowa, Kentucky, Maryland, New Hampshire, Nevada, New York, Pennsylvania, Texas, Virginia and West Virginia. Annual surgeon CABG volume was determined by calculating the total number of CABGs performed by an individual surgeon divided by the number of years that the surgeon was surveyed. Annual surgeon CABG volume was categorised into quartiles, such that a similar number of patients would fall into each quartile (very low <50 cases: 29.6%, low 50–75 cases: 20.8%, high 75–100 cases: 18.1% and very high >100 cases: 31.5%).

The study protocol was reviewed by the Mount Sinai School of Medicine Institutional Review Board and was compliant with a waiver of informed consent, and the Health insurance Portability and Accountability Act regulations.

Clinical outcomes of interest

The primary endpoint was deep sternal infection, which was identified using ICD-9 diagnosis codes (998.59, 998.32, 998.31, 519.2).24 The other endpoints were inhospital mortality, length of hospital stay and total costs of admission.

Statistical analysis

Continuous variables are expressed as means with SDs. Categorical variables are expressed as proportions. Differences between groups were detected using the χ2 test for categorical variables, Student's t test for normally distributed continuous variables and the Mann-Whitney U test for non-normally distributed continuous variables. Binary logistic regression analysis was performed to determine independent predictors of BIMA use and DSWI. Logistic regression for BIMA use was performed using variables related to patient demographics and hospital characteristics. The effect of BIMA use for the rate of DSWI was investigated using logistic regression accounting for age, gender, elective status and clinically relevant patient comorbidities. The same analysis was repeated in the subgroup analysis of each high-risk group which was found to be significant in the first logistic regression model with the whole study population. Annual surgeon CABG volume was also separately included into the model only with patients with this information available. Results are demonstrated as OR and 95% CI. All tests were two tailed. A p value of <0.05 was considered statistically significant. The statistical analysis was performed using IBM SPSS Statistics for Windows, V.20.0 (SPSS, IBM Corporation, Armonk, New York, USA).


National trends in BIMA use

The overall BIMA use rate was 3.9% over the study period of 2002–2008, with a gradual increase from 3.9% in 2002 to 4.5% in 2008. The patient demographics, hospital characteristics and clinical comorbidities for the overall population and according to SIMA or BIMA use are listed in table 1. Generally, BIMA patients were younger (60.1 years vs 64.8 years, p<0.001), more likely to be male (81.8 vs 72.9%, p<0.001) and more predominantly Caucasian (83.6 vs 80.8%, p<0.001) with fewer comorbidities, and higher socioeconomic status. The strongest independent predictors of BIMA use were male gender, Caucasian race and higher household incomes (table 2). Hospitals in Northeast urban area with small bed size also tended to use BIMA more frequently. Higher annual surgeon CABG volume had no significant impact on the rate of the use of BIMA.

Table 2

Predictors of bilateral internal mammary artery use based on patient demographics, hospital characteristics and annual surgeon CABG volume

Operative outcomes

Operative outcomes are shown in table 3. In logistic regression analysis after adjusting for age, gender, elective status and comorbidities, the use of BIMA was independently associated with slightly lower inhospital mortality (unadjusted rate 1.1% vs 1.7%, adjusted OR 0.86, 95% CI 0.79 to 0.93, p<0.001). The overall rate of DSWI was 1.4%: 1.4% in SIMA patients and 1.3% in BIMA patients (p<0.001).

Table 3

Operative outcomes

Predictors and impact of DSWI

Overall the strongest independent predictors of DSWI were congestive heart failure (OR 6.22, 95% CI 5.80 to 6.68), chronic pulmonary disease (OR 1.57, 95% CI 1.53 to 1.62), diabetes mellitus with chronic complications (OR 1.51, 95% CI 1.44 to 1.59) and renal failure (OR 2.13, 95% CI 2.05 to 2.22), all p<0.001. BIMA use was not found to be an independent predictor of DSWI (adjusted OR 1.03, 95% CI 0.96 to 1.10, p=0.416). Higher annual surgeon CABG volume was associated with lower DSWI rate (adjusted OR 0.87, 95% CI 0.81 to 0.93, p<0.001). BIMA use remained a non-significant risk factor for DSWI after further adjustment for surgical volume (p=0.935). In a subgroup analysis of patients with obesity, diabetes and renal failure, BIMA use was an independent predictor of DSWI only in patients with chronic complications of diabetes mellitus (5.4% of the whole population) (unadjusted rate 4.2% vs 2.5%, adjusted OR 1.90, 95% CI 1.51 to 2.41, p<0.001) (table 4). DSWI on average increased length of hospital stay by almost a week and the cost of admission by more than $1 00 000, with a fivefold increase in inhospital mortality (unadjusted rate 8.3% vs 1.6%, adjusted OR 3.77, 95% CI 3.57 to 3.98, p<0.001).

Table 4

Effect of BIMA use on the incidence of DSWI in high-risk subgroups


The main finding of this study is that despite established evidence of the safety and survival benefits of bilateral internal mammary grafting, in contemporary US practice fewer than 5% of patients undergoing primary isolated CABG receive BIMA grafts.

Rationale for BIMA use

The primary rationale for more liberal use of BIMA is the long-term survival benefit demonstrated in large-scale observational studies such as the propensity-matched analysis of 8123 SIMA patients and 2001 BIMA patients that underwent primary isolated CABG at the Cleveland Clinic between 1971 and 1989.8 Survival of BIMA patient groups versus SIMA patient groups at 10, 15 and 20 years was 81% versus 78%, 67% versus 58% and 50% versus 37%, respectively (p<0.001), with continued widening of the survival curves at 20 years. In a more recent meta-analysis of comparative studies including 11 269 SIMA patients and 4693 BIMA patients, Taggart et al5 reported a HR of 0.8 (95% CI 0.70 to 0.94, p<0.01) for long-term mortality in patients who underwent BIMA compared with patients who underwent SIMA.

Why is BIMA use so low?

Despite the potential long-term survival benefits, BIMA use in most countries is limited to a minority of patients. In Europe, the proportion of isolated coronary bypass operations where two or more arterial conduits are used (including all BIMA, but also including SIMA and radial artery) averages 20%, but ranges from 76.6% down to 1.2%.25 In the USA BIMA is currently used in fewer than 5% of cases.17 There are several reasons why BIMA is so infrequently used in contemporary practice. In two separate surveys the reasons cited most commonly by cardiac surgeons for not using BIMA more often were concerns about increased early morbidity and mortality,26 ,27 with the potential risk of sternal wound infection mentioned by 35% of surgeons.27 BIMA harvest results in greater devascularisation of the sternum, reducing wound healing and potentially increasing risk of sternal wound complications, which carry a significant incremental mortality. The putative long-term benefits of using BIMA for coronary bypass surgery are consequently outweighed in the minds of many surgeons by the short-term considerations dominant in contemporary practice where key quality benchmarks are entirely focused on 30-day and inhospital outcomes, including surgical wound infections.


Several large observational studies and a recent meta-analysis report a slightly higher incidence of DSWI in patients that undergo CABG with BIMA.4 ,18–20 ,28 More recently, in the analysis of the 1-year results of a multicentre randomised controlled trial designed to identify 10-year survival difference in 3102 patients undergoing primary CABG using BIMA or SIMA, the incidence of sternal wound reconstruction was 0.6% in SIMA patients compared with 1.9% in BIMA patients,29 (relative risk 3.24, 95% CI 1.54-6.83). While the absolute increase in risk of DSWI is small, at around 1%; the relative risk of inhospital mortality in patients with sternal wound infection compared with those without this complication has been reported to range from twofold to more than 10-fold, depending on patient comorbidity.30

While our data confirms the profound impact of DSWI on patient outcomes (we observed a fivefold increase in operative mortality in patients with DSWI), we found that BIMA use does not appear to be an independent predictor of DSWI in contemporary US practice. This may reflect differences in perioperative management since some of the earlier comparative studies were conducted, particularly changes in antibiotic prophylaxis and glycaemic control, which have subsequently been included among publically reported outcome benchmarks in the USA, and used to determine monetary reimbursement for hospitals. Patient variables and differences in operative techniques not included in our analysis may also play a role. We show that surgeons avoid using BIMA in patients traditionally viewed as particularly high risk for DSWI—with particularly low use in patients with diabetes, obesity and in female patients. BIMA use has, however, been shown to provide a long-term survival benefit in patients with diabetes.6 ,10 ,31 While our data suggests that this approach can be employed in most patients with diabetes without incremental short-term morbidity, the potential long-term survival benefit in the 5% of patients with chronic complications of diabetes such as retinopathy, nephropathy or peripheral neuropathy must be carefully weighed against the absolute 1–2% increase in risk of sternal wound infection with this technique in these patients.

Technical considerations

In comparison with saphenous vein conduit, BIMA poses specific technical constraints limiting location of the coronary bypass graft, including reduced conduit length and poor response to competitive flow. These anatomical considerations, however, fall a long way short of explaining the reasons for the infrequent use of BIMA in the wider community: fewer than a third of surgeons surveyed cited this as a common concern,27 and within the Arterial Revascularisation Trial only 9% of patients (142/1548) randomised to BIMA received one internal mammary artery graft instead of two because target or conduit was considered unsuitable.29 Finally, while the additional time required for BIMA has been described as a deterrent to its use,26 ,27 the total incremental time added to bypass surgery in the Arterial Revascularisation Trial (which excluded surgeons who had performed fewer than 50 such procedures) was only 23 min, with no increase in bleeding or other technical complications.29

Study limitations

Although the NIS is a national, mandatory database containing detailed demographic, clinical, socioeconomic and provider information, there are several inherent limitations with using this dataset for clinical outcomes research. The first is that the NIS is primarily an administrative dataset, relying on non-clinical personnel to code patient, provider and procedure factors using the ICD-9CM classification which, in comparison with clinical registries such as the Society of Cardiothoracic Surgeons of Great Britain and Ireland, and the Society of Thoracic Surgeons adult cardiac surgery databases, offers much less precise, and clinically relevant diagnostic and procedural definitions.32 For example, four codes were used in this study to identify a broad definition of DSWI, 998.59 (postoperative wound infection), 998.32 (dehiscence of external operation wound), 998.31 (dehiscence of internal operation wound), 519.2 (mediastinitis). These codes are not mutually exclusive and the definition and severity of DSWI differs depending on which variables are considered to be DSWI. The total rate of 1.4% correlates with the literature but still DSWI in this study may include superficial sternal wound infection as well. As shown in online supplementary table S2, there is a spectrum of ‘DSWI’ but each level still significantly affects the postoperative course and most importantly BIMA use was not associated with increased rate of each level of DSWI. The second limitation is that due to a lack of any outcome data after the initial hospitalisation we are unable to fully quantify the incidence of late DSWI, which may be more influenced by BIMA use. The third limitation is the inability to better control for potential confounding variables because of the lack of information available regarding the decision to use specific conduit, the harvesting technique and other factors impacting DSWI (eg, the postoperative glucose control, and timing and nature of prophylactic antibiotics) or mortality (eg, under-revascularisation).


In this analysis of more than a million patients who underwent isolated CABG in the USA, the 4% of patients who received BIMA experienced slightly reduced operative mortality with no incremental risk of sternal wound infection. With the exception of patients that have complications of chronic diabetes, concerns about increased risk of DSWI do not justify use of a SIMA in the majority of patients undergoing isolated CABG.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Contributors All authors included meet the criteria of authorship as outlined below, including final approval of the published manuscript. No authors who met these criteria were excluded. JC: Conception and design, data acquisition, data analysis and interpretation, drafting and critical revision of the manuscript. SI: Data acquisition, data analysis and interpretation, drafting and critical revision of the manuscript. PC: Data acquisition, data analysis and interpretation, critical revision of the manuscript. DHA: Critical revision of the manuscript. The authors gratefully acknowledge the help of Antonio Polanco and Matthew Seigerman in preparation of this manuscript.

  • Competing interests None.

  • Ethics approval Reviewed by Mount Sinai School of Medicine Institutional Review Board.

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

  • Data sharing statement The Nationwide Inpatient Sample dataset on which this study was based is available to healthcare researchers. NIS releases for data years 1988–2010 are available from the HCUP Central Distributor. The 2010 NIS may be purchased for $50 for students and $350 for all others on a single DVD-ROM with accompanying documentation.

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