Use Rate and Outcome in Bilateral Internal Thoracic Artery Grafting: Insights From a Systematic Review and Meta‐Analysis
Background This meta‐analysis was designed to assess whether center experience affects the short‐ and long‐term results and the relative benefits of bilateral internal thoracic artery grafting (BITA) for coronary artery bypass grafting.
Methods and Results MEDLINE and EMBASE were searched to identify all articles reporting the outcome of BITA in patients undergoing coronary artery bypass grafting. The BITA center experience was gauged according to the percentage use of BITA in the institutional overall coronary artery bypass grafting population (%BITA). The primary outcome was long‐term all‐cause mortality. Secondary outcomes were operative mortality, perioperative myocardial infarction, perioperative stroke, deep sternal wound infections (DSWIs), and major postoperative adverse event. The rates of the primary and secondary outcomes were calculated after adjusting for %BITA. Primary and secondary outcomes were also compared between the BITA and the single internal thoracic artery arms in the adjusted studies. Meta‐regression was used to evaluate the effect of %BITA on the primary and secondary outcomes. Thirty‐four studies (27 894 patients undergoing BITA) were included. In the pooled analysis, the incidence rate for long‐term mortality was 2.83% (95% confidence interval, 2.21%–3.61%). %BITA was significantly and inversely associated with long‐term mortality and the rate of DSWI. In the pairwise comparison, %BITA was significantly and inversely associated with the risk of long‐term mortality and DSWI in the group undergoing BITA.
Conclusions BITA series with higher %BITA report significantly lower long‐term mortality and DSWI rate as well as higher long‐term survival advantage and lower relative risk of DSWI in their BITA cohort. These findings suggest that a specific volume‐outcome relationship exists for BITA grafting.
- bilateral internal thoracic artery
- coronary artery bypass graft
- coronary artery bypass graft surgery
What Is New?
Our analysis suggests the existence of a use rate to outcome effect for bilateral internal thoracic artery grafting.
What Are the Clinical Implications?
Our findings suggest the possibility that the creation of specialized tertiary centers for coronary surgery, similar to those that exist for aortic surgery and transplantation, may improve the outcomes of bilateral internal thoracic artery grafting.
The relationship between center or operator experience and outcome has extensively been described in medicine and in surgery.1 The volume/outcome (V/O) effect is particularly evident for technically complex procedures, such as off‐pump surgery or valve repair procedures.2 This has resulted in recommendations for minimum center‐ and surgeon‐specific procedural volumes, as well as for specialized referral centers for highly complex cardiovascular and cancer operations.1
The use of bilateral internal thoracic artery (BITA) increases the technical complexity of the CABG operation.4 Previously published studies on the V/O effect in CABG did not stratify the results according to the type of technique used, although in the great majority of the published series, BITA was used only in a small minority of patients.
We hypothesized that, because of the more complex nature of the procedure, a specific center experience to outcome relationship exists for BITA grafting; therefore, we aimed at investigating this by using a meta‐analytic approach.
We conducted this systematic review and meta‐analysis following the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses statement.5 Table S1 illustrates the Meta‐Analysis of Observational Studies in Epidemiology guidelines checklist. The data, analytic methods, and study materials will not be made available to other researchers for purposes of reproducing the results or replicating the procedure.
Search Strategy and Study Selection Criteria
OVID versions of MEDLINE and EMBASE were searched from January 1972 to June 2017 to identify all articles reporting the outcome of BITA in patients undergoing CABG.
The following keywords were used: “bilateral,” “double,” “mammary,” “thoracic,” “artery,” “multiple,” “total,” “arterial,” “revascularization,” and “coronary.” Their combinations were searched using the term “AND.” All citations were screened for study inclusion independently by 2 investigators (A.D.F and M.G.). In case of disagreement, a consensus was reached. In addition, the bibliography of all studies and meta‐analyses was searched to identify further publications (backward snowballing).
Inclusion criteria for analysis were single‐institution study, sample size of at least 100 patients, and English language. We excluded studies in which the percentage of BITA use of the individual center (number of patients undergoing BITA/total number of patients undergoing CABG in the center in the study period=%BITA) could not be extracted. In case of overlapping between studies or multiple publications from the same center, only the publication with the largest sample size was included.
The critical appraisal of the quality of included studies was assessed using the Newcastle‐Ottawa Scale for observational studies.6 The highest possible score is 9 stars; <6 stars was considered low quality, whereas ≥6 stars was considered high quality (Table S2).
The following data were abstracted: study period, country, institution, total sample size, number of patients undergoing BITA, %BITA, annual CABG volume of the individual center (total number of CABGs in the study/the study period), study design, and follow‐up duration. The following patient characteristics were abstracted: age, female sex, diabetes mellitus, left ventricular ejection fraction, number of grafts per patient, number of internal thoracic artery grafts per patient, use of internal thoracic artery sequentials, use of skeletonization technique for BITA harvesting, and chronic obstructive pulmonary disease.
For descriptive purposes, the studies were divided according to quartiles of %BITA (boundaries for the quartiles were 17.1%, 29.2%, and 50.3%; the range of %BITA was 3.7%–64%). In all the other analysis, %BITA was analyzed as a continuous variable.
For the BITA versus single internal thoracic artery (SITA) comparison, data were abstracted from the adjusted series only (covariate adjusted or propensity matched). Crude event rates, unadjusted and adjusted hazard ratios, 95% confidence intervals (CIs) for BITA and SITA, and log p‐rank values were abstracted. For the secondary outcomes, number of events was extracted for each outcome.
Continuous variables were expressed as median (25th–75th percentile) or as mean±SD. Categorical variables were reported as frequency (percentage).
The primary outcome was long‐term all‐cause mortality.
The secondary outcomes were operative mortality, perioperative myocardial infarction, perioperative stroke, deep sternal wound infections (DSWIs), and major postoperative adverse events, defined as the composite of operative mortality, perioperative myocardial infarction, perioperative stroke, and DSWIs. Major postoperative adverse event was derived only from studies that reported all 4 individual outcome components.
Analytic Plan and Statistical Analysis
In the pooled analyses, the incident rate or the pooled event rates (PERs) of the primary and secondary outcomes in the BITA series were calculated according to the %BITA.
In the pairwise comparisons including only the adjusted studies, the relative risks of the primary and secondary outcomes in the BITA series were calculated according to the %BITA.
To account for the differential follow‐up times of the primary outcome in the various studies, an underlying Poisson process with a constant event rate was assumed with a total number of events observed within a treatment group of the total person‐time of follow‐up for that treatment group calculated from study follow‐up. A log‐link function was used to model the incidence rate (IR), and a random effect was used. When the number of events was not available from text or tables, the number of events was derived from the unadjusted Kaplan‐Meier curves using GetData Graph Digitizer software 2.26 (http://getdata-graph-digitizer.com) using a previously described method.7
For secondary outcomes, the PERs with 95% CIs were calculated.
BITA versus SITA comparison.
For the primary outcome, the generic inverse variance (DerSimonian‐Laird) method was used to pool the natural logarithm of the IR ratio across studies to account for potentially different follow‐up durations between the studies. We estimated the IR ratio through several means, depending on the available study data. When hazard ratios were provided, we took the natural logarithm of the hazard ratio; the standard error was derived from the 95% CI or log‐rank P value.8 When event rates were not readily available, they were extracted from Kaplan‐Meier curves.7, 9 The standard error was estimated from the number of events in each arm.8 For secondary outcomes, individual and pooled odds ratio (OR) with 95% CIs were used.
In the pooled and pairwise analysis, univariable meta‐regression was used to explore the association between %BITA and the primary and secondary outcomes.
A mixed‐effects meta‐regression model that contained both study‐specific covariates and random‐effect components was used to allow for the division of heterogeneity into an explained (by the covariates) and an unexplained (the random‐effects) component.10 Each study was weighted by the inverse of the variance of the estimate for that study, and between‐study variance was estimated with DerSimonian‐Laird estimator.
In both sets of analyses, a multivariable meta‐regression model was used to assess the association between %BITA with the primary outcome while also adjusting for age, sex, diabetes mellitus, and annual CABG hospital volume. A separate multivariable meta‐regression model, including %BITA, sex, diabetes mellitus, annual CABG volume, and skeletonization, was used to assess for the analysis of DSWI.
The Cochran Q statistic and the I2 test were used to assess studies’ heterogeneity. For the primary outcome, if significant heterogeneity was detected (I2>75%), a leave‐one‐out sensitivity analysis was performed to assess for single comparison driven inference. Funnel plots and Egger regression test were used to assess for potential publication bias. If publication bias was suspected, visual assessment of the cumulative forest plot and Duval and Tweedie's trim and fill methods were used for further assessment.
A random‐effect model (inverse variance method)11 was used for all the analysis. Hypothesis testing for equivalence was set at the 2‐tailed 0.05 level.
All analyses were performed using R, version 3.3.3 (R Project for Statistical Computing) using the following statistical packages: “meta” and “metafor”12, 13 within the RStudio, 0.99.489 (http://www.rstudio.com) and Comprehensive Meta‐Analysis V 3.0 (2006; Biostat, Inc, Englewood, NJ).
The literature search identified 2899 potentially eligible studies. Twenty‐two additional articles were identified through backward snowballing. The Preferred Reporting Items for Systematic Reviews and Meta‐Analyses flow diagram is reported in Figure 1.
Studies’ and Participants’ Characteristics
A total of 34 studies including 27 894 patients who had CABG using BITA were included.14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47 Details of the individual studies are shown in Tables 1 and 2 and Table S3. The weighted mean follow‐up time was 7.7±1.2 years. For the pairwise comparison, 27 adjusted studies (12 propensity matched) were included (75 334 patients; 19 290 BITAs and 56 044 SITAs). Eight studies (13 292 patients) were included in the analysis of the composite major postoperative adverse event.
The included studies were published from 1989 to 2016, and the sample size ranged from 147 to 17 609.
In the pooled analysis, the IR for long‐term mortality in the overall population was 2.83%/year (95% CI, 2.21%/year–3.61%/year; Table 3). The leave‐one‐out analysis is shown in Figure S1, and the funnel plot and the cumulative analysis are shown in Figure S2. %BITA was significantly and inversely associated with long‐term mortality in the univariable meta‐regression (β=−0.02, P=0.02; Figure 2A) and the multivariable meta‐regression (β=−0.03, P=0.04; Figure 2B).
In the pairwise comparison with SITA, the use of BITA was associated with a significantly lower long‐term mortality (IR ratio, 0.78; 95% CI, 0.72–0.84; Figure S3). %BITA was significantly and inversely associated with the IR ratio for long‐term mortality in both the univariable meta‐regression (β=−0.006, P=0.01; Figure 3A) and the multivariable meta‐regression (β=−0.008, P=0.03; Figure 3B).
In the pooled analysis, the PER for operative mortality was 1.6% (95% CI, 1.2%–2.2%), the PER for myocardial infarction was 1.6% (95% CI, 1.1%–2.4%), the PER for perioperative stroke was 1.1% (95% CI, 0.9%–1.4%), the PER for DSWI was 2.2% (95% CI, 1.7%–2.7%), and the PER for major postoperative adverse event was 5.7% (95% CI, 4.7%–6.8%) (Table 3). %BITA was significantly and inversely associated with DSWI, according to the univariable and multivariable meta‐regressions (β=−0.001 [P=0.006] and β=−0.02 [P<0.001], respectively; Table 3 and Figure 4). %BITA did not influence the other secondary outcomes (Table 3 and Figure S4).
In the pairwise comparison with SITA, BITA use was associated with a significantly higher incidence of DSWI (OR, 1.58; 95% CI, 1.15–2.19) and a significantly lower rate of perioperative stroke (OR, 0.76; 95% CI, 0.61–0.94). %BITA was significantly and inversely associated with the OR for DSWI by univariable and multivariable meta‐regressions (β=−0.020 [P=0.02] and β=−0.03 [P=0.005], respectively; Figure 5).
No significant differences were found for the other secondary outcomes (Figure S5).
An inverse relationship between hospital volume and clinical outcome has been described extensively in surgery.1 Some data suggest that the V/O relationship can be more evident for more complex procedures, such as off‐pump CABG, or higher‐risk patients.2
The V/O effect in CABG has been the focus of a large amount of research. Despite controversy related to the methodological quality of the sources used in the published studies and the lack of a clear‐cut explanation, it is usually accepted that hospitals that perform a high annual volume of CABG and have more experience with the procedure have better outcomes than hospitals that perform a smaller number of procedures.1, 2, 3 The use of BITA during CABG adds technical complexity to the operation. In a survey of all UK consultant cardiac surgeons, the perceived increased technical difficulty and need of a learning curve were the most frequent reason to explain the low adoption rate of BITA.4
In the recently published ART (Arterial Revascularization Trial), only 83.6% of the patients randomized to BITA received the assigned treatment (versus 96.1% in the conventional CABG group).48 This high crossover rate in the BITA series is a testament to higher technical complexity of the operation, and it is even more meaningful if one considers that only expert BITA surgeons were allowed to participate in ART. However, it also raises the possibility that the BITA surgeons were not all equally experienced in BITA grafting because the crossover rate varied from 0% to 42.9% on a center level and from 0% to 100% for the 168 participating surgeons, suggesting the need for appropriate and documented experience for participation in trials involving complex technical procedures. Thus, as complexity of the coronary surgery increases with the addition of a BITA grafting strategy, institution experience with BITA may play an ever‐increasing role on outcomes. However, to date, this subject has not been investigated in detail.
Our data suggest that a relationship between the rate of BITA use at the center level and the clinical results exists at least for the 2 most important outcomes associated with BITA grafting: long‐term survival and incidence of DSWI. In our analysis, long‐term mortality was significantly and inversely associated with %BITA, with better survival reported by centers with high %BITA. In the pairwise comparison with SITA, the long‐term survival benefit associated with the use of BITA was significantly associated with %BITA, with centers with high %BITA reporting a significantly larger survival advantage for patients undergoing BITA. The effect of %BITA on long‐term mortality remained significant even when entering the annual hospital volume as a covariate in the meta‐regression model, suggesting the existence of an “experience effect” specific for BITA grafting and independent from the V/O relationship for standard CABG.
The rate of DSWI and the increase in the risk of DSWI in the BITA group were also significantly and inversely associated with %BITA. Centers with high %BITA reported a lower incidence of DSWI in the BITA series and a lower relative risk of DSWI in the BITA group compared with the SITA series. Furthermore, the incidence and risk of the short‐term outcomes, such as operative mortality, perioperative myocardial infarction, and perioperative stroke, were not influenced by the %BITA.
Taken together, our findings seem to suggest that the reasons for the reported difference in outcomes between centers at high and low %BITA are not strictly technical, because outcomes that are heavily influenced by technical factors, such as perioperative myocardial infarction, stroke, and operative mortality, were not significantly associated with %BITA. One explanation for our results may be better patient selection and grafting strategy in centers at high %BITA. It is possible that more experienced centers were more proficient in selecting appropriate patients who would benefit from BITA grafting and the use of the arterial grafts.
It is notable that 67% of the studies in the highest quartile of %BITA versus 38% in the lowest quartile used BITA sequentials (P=0.03). It has been shown that an increase in the number of BITA anastomoses is associated with better clinical outcome.49
For DSWI, the adoption of the skeletonized technique for harvesting was similar between high and low BITA users (42.9% in the first quartile and 57.1% in the fourth quartile; P=0.56), and the association between the OR for DSWI and %BITA was confirmed, even in the multivariable meta‐regression model after adjusting for skeletonization. These results suggest that BITA skeletonization alone is not the explanation for the reported difference in DWSI.
This analysis must be interpreted in the context of some limitations. We used %BITA as opposed to BITA volume as a marker of experience with BITA because we believe that the rate of use is a stronger surrogate measure of familiarity, comfort, and skill in the operation than the absolute volume of procedures performed. However, this assumption is based on the authors’ opinion, and has never been objectively validated. We did not capture individual surgeon's experience, which may be more important than center's experience. Also, the included studies used different surgical protocols and definition of outcomes and were in different stages of their BITA learning curve, leading to heterogeneity in the analyzed data. Most important, an unavoidable publication bias exists, because all centers were in some way experienced in the use of BITA (although at different levels). Our analysis probably does not capture the results of inexperienced centers or beginners in BITA grafting who are unlikely to publish their results. In addition, meta‐regressions can only be used to assess association and do not infer causality. Nonetheless, despite these limitations, the reproducibility of our results, on the basis of multiple different statistical approaches, supports the robustness of our reported findings.
In conclusion, our analysis suggests the existence of a use rate to outcome effect for BITA grafting. In our study, centers that used BITA more frequently reported a reduced risk of sternal complications and achieved better long‐term survival compared with SITA. Our findings suggest the possibility that the creation of specialized tertiary centers for coronary surgery, similar to those that exist for aortic surgery and transplantation, may improve the outcomes of BITA grafting.
Table S1. MOOSE Checklist for Meta‐Analyses of Observational Studies
Table S2. Summary of Critical Appraisal of Included Studies Using the Newcastle‐Ottawa Scale for Cohort Studies
Table S3. Risk Factor Distribution in the Populations of the Studies Included in the Primary Analysis
Figure S1. The “Leave‐one‐out” analysis for the primary outcome.
Figure S2. The pooled analysis for long term mortality: (A) Funnel plot with trim and fill method and (B) Cumulative meta‐analysis.
Figure S3. The pairwise comparison for long term mortality among the adjusted studies using the incident rate ratio.
Figure S4. The effect of the percentage of BITA use on the pooled event rate of (A) peri‐operative myocardial infarction, (B) peri‐operative stroke, (C) major postoperative adverse events (MAE), (D) operative mortality.
Figure S5. The effect of the percentage of BITA use on the odds ratio of (A) peri‐operative myocardial infarction, (B) peri‐operative stroke, (C) major postoperative adverse events (MAE), (D) operative mortality.
- ↵Kim LK, Looser P, Swaminathan RV, Minutello RM, Wong SC, Girardi L, Feldman DN. Outcomes in patients undergoing coronary artery bypass graft surgery in the United States based on hospital volume, 2007 to 2011. J Thorac Cardiovasc Surg. 2016;151:1686–1692.
- ↵Catarino PA, Black E, Taggart DP. Why do UK cardiac surgeons not perform their first choice operation for coronary artery bypass graft? Heart. 2002;88:643–644.
- ↵Liberati A, Altman DG, Tetzlaff J, Mulrow C, Gøtzsche PC, Ioannidis JPA, Clarke M, Devereaux PJ, Kleijnen J, Moher D. The PRISMA statement for reporting systematic reviews and meta‐analyses of studies that evaluate healthcare interventions: explanation and elaboration. BMJ. 2009;339:b2700.
- ↵Wells G, Shea B, O'Connell D, Peterson J, Welch V, Losos M. Newcastle‐Ottawa scale (NOS) for assessing the quality of nonrandomised studies in metaanalysis. 2000. Available at: http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp. Accessed May 7, 2017.
- ↵Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta‐analyses. BMJ. 2003;327:557.
- ↵Viechtbauer W, Viechtbauer MW. Package “metafor.” https://cran.r-project.org/web/packages/metafor/metafor.pdf. Accessed May 7, 2017.
- ↵Schwarzer G, Schwarzer MG. Package “meta.” Meta‐Anal R. https://cran.r-project.org/web/packages/meta/meta.pdf. Accessed May 7, 2017.
- ↵Benedetto U, Amrani M, Gaer J, Bahrami T, de Robertis F, Simon AR, Raja SG; Harefield Cardiac Outcomes Research Group . The influence of bilateral internal mammary arteries on short‐ and long‐term outcomes: a propensity score matching in accordance with current recommendations. J Thorac Cardiovasc Surg. 2014;148:2699–2705.
- ↵Calafiore AM, Di Giammarco G, Teodori G, Di Mauro M, IacoIacò AL, Bivona A, Contini M, Vitolla G. Late results of first myocardial revascularization in multiple vessel disease: single versus bilateral internal mammary artery with or without saphenous vein grafts. Eur J Cardiothorac Surg. 2004;26:542–548.
- ↵Dewar LR, Jamieson WE, Janusz MT, Adeli‐Sardo M, Germann E, MacNab JS, Tyers GFO. Unilateral versus bilateral internal mammary revascularization. Circulation. 1995;92:8–13.
- ↵Endo M, Nishida H, Tomizawa Y, Kasanuki H. Benefit of bilateral over single internal mammary artery grafts for multiple coronary artery bypass grafting. Circulation. 2001;104:2164–2170.
- ↵Gansera B, Günzinger R, Eichinger W, Neumaier P, Breuer M, Kemkes BM. End of the millenium—end of the single thoracic artery graft? Two thoracic arteries—standard for the next millenium? Early clinical results and analysis of risk factors in 1,487 patients with bilateral internal thoracic artery grafts. Thorac Cardiovasc Surg. 2001;49:10–15.
- ↵Ioannidis JP, Galanos O, Katritsis D, Connery CP, Drossos GE, Swistel DG, Anagnostopoulos CE. Early mortality and morbidity of bilateral versus single internal thoracic artery revascularization: propensity and risk modeling. J Am Coll Cardiol. 2001;37:521–528.
- ↵Itoh S, Kimura N, Adachi H, Yamaguchi A. Is bilateral internal mammary arterial grafting beneficial for patients aged 75 years or older? Circ J. 2016;80:1756–1763.
- ↵Locker C, Schaff HV, Dearani JA, Joyce LD, Park SJ, Burkhart HM, Suri RM, Greason KL, Stulak JM, Li Z, Daly RC. Multiple arterial grafts improve late survival of patients undergoing coronary artery bypass graft surgery: analysis of 8622 patients with multivessel disease. Circulation. 2012;126:1023–1030.
- ↵Medalion B, Mohr R, Ben‐Gal Y, Nesher N, Kramer A, Eliyahu S, Pevni D. Arterial coronary artery bypass grafting is safe and effective in elderly patients. J Thorac Cardiovasc Surg. 2015;150:607–612.
- ↵Navia DO, Vrancic M, Piccinini F, Camporrotondo M, Dorsa A, Espinoza J, Benzadon M, Camou J. Myocardial revascularization exclusively with bilateral internal thoracic arteries in T‐graft configuration: effects on late survival. Ann Thorac Surg. 2016;101:1775–1781.
- ↵Pettinari M, Sergeant P, Meuris B. Bilateral internal thoracic artery grafting increases long‐term survival in elderly patients. Eur J Cardiothorac Surg. 2014;47:703–709.
- ↵Pusca SV, Kilgo PD, Vega JD, Cooper WA, Vassiliades TA, Chen EP, Lattouf OM, Guyton RA, Puskas JD. Propensity‐score analysis of early outcomes after bilateral versus single internal thoracic artery grafting. Innovations (Phila). 2008;3:19–24.
- ↵Rosenblum JM, Harskamp RE, Hoedemaker N, Walker P, Liberman HA, de Winter RJ, Vassiliades TA, Puskas JD, Halkos ME. Hybrid coronary revascularization versus coronary artery bypass surgery with bilateral or single internal mammary artery grafts. J Thorac Cardiovasc Surg. 2016;151:1081–1089.
- ↵Stevens LM, Carrier M, Perrault LP, Hebert Y, Cartier R, Bouchard D, Fortier A, El‐Hamamsy I, Pellerin M. Single versus bilateral internal thoracic artery grafts with concomitant saphenous vein grafts for multivessel coronary artery bypass grafting: effects on mortality and event‐free survival. J Thorac Cardiovasc Surg. 2004;127:1408–1415.
- ↵Taggart DP, Altman DG, Gray AM, Lees B, Gerry S, Benedetto U, Flather M. Randomized trial of bilateral versus single internal‐thoracic‐artery grafts. N Engl J Med. 2016;375:2540–2549.
- ↵Glineur D, Boodhwani M, Hanet C, de Kerchove L, Navarra E, Astarci P, Noirhomme P, El Khoury G. Bilateral internal thoracic artery configuration for coronary artery bypass surgery. Circ Cardiovasc Interv. 2016;9:e003518.