Bilateral versus single internal mammary artery bypass grafts for coronary arterial atherosclerosis
Abstract
This is a protocol for a Cochrane Review (Intervention). The objectives are as follows:
To review the effects of bilateral versus single internal mammary artery grafts for coronary bypass grafting.
Background
The clinical and prognostic benefits of coronary revascularisation for certain subgroups of patients with ischaemic heart disease are well accepted (Yusuf 1994) and each year approximately one million patients world‐wide undergo coronary artery bypass grafting (CABG). While CABG provides excellent short and intermediate term outcome the long‐term results are limited by vein graft failure.
Most patients undergoing CABG require three bypass grafts and the 'standard' operation uses a single internal mammary artery (SIMA) and two vein grafts. Even in the absence of data from randomised trials the clinical and survival advantages of an IMA graft in CABG are widely accepted (Cameron 1996; Loop 1986; Yusuf 1994). In comparison to the use of vein conduits an IMA graft to the left anterior descending (LAD) coronary artery improves survival and reduces the incidence of late myocardial infarction, recurrent angina and the need for further cardiac interventions (Cameron 1996; Loop 1986; Yusuf 1994). Ten years after CABG 90%‐95% of left IMA are patent and disease free whereas three quarters of vein conduits are blocked or severely diseased (Fitzgibbon 1996). Vein graft failure leads to a high risk of recurrent angina, late myocardial infarction and the need for further intervention (Eagle 1991) so that by 10 to 15 years after the initial operation as many as 30% of patients may require redo CABG (Weintraub 1994; Kaul 1995). There is a significant increase in both the risk and costs of redo surgery.
As vein grafts are more frequently used to graft non‐LAD coronary vessels it has been suggested that their inferior patency rates might reflect their use in anatomical territories with less favourable run‐off. Current evidence strongly refutes this explanation with several reports of long‐term patency of IMA anastomoses to non‐LAD coronary vessels far superior to that obtainable with vein grafts (Calafiore 2000; Dion 2000). Calafiore and colleagues reported angiographic patency rates of 99% in bilateral IMA (BIMA) grafts eighteen months after CABG with an estimated actuarial survival of 95% in 1800 patients at eight years (Calafiore 2000). Dion and colleagues reported that 96% of multiple arterial anastomoses to various coronary vessels were patent seven years after CABG in 161 patients who consented to repeat angiography (Dion 2000).
The possibility that the use of bilateral IMA conduits might result in clinical and survival benefits exceeding those of a single IMA graft has also been examined in previous studies (Berreklouw 1995; Buxton 1998; Dewar 1995; Farinas 1999; Johnson 1989; Lytle 1999; Morris 1990; Naunheim 1992; Pick 1997; Sergeant 1997). Interpretation of these studies is, however, complicated by lack of randomisation, insufficient patient numbers and inadequate length and/or completeness of follow‐up. Furthermore, as the initial use of bilateral IMA was largely confined to younger and lower risk patients, any long term benefits were attributed to the inherently more favourable survival characteristics of these patients obscuring any potential benefit of bilateral IMA conduits per se. We compared the effects of BIMA versus SIMA in systematic review in which survival data from seven studies (11 269 single and 4693 bilateral IMA grafts) were analysed (Taggart 2001) The bilateral group had significantly better survival than the single group (hazard ratio for death 0·81; 95% CI 0·70‐0·94). Exclusion of methodologically weak studies improved survival rates with bilateral IMA grafts. Because no study was a randomised trial, the results were more uncertain than is indicated by the 95% CI. Nevertheless, bilateral IMA grafts seem to give better survival rates than single grafts. A further systematic review was published in 2002 (Rizzoli 2002), which include the same seven studies and show similar favourable hazard ratios of BIMA over SIMA for survival as seen in the review by Taggart et al (Taggart 2001).
Since the publication of the Taggart systematic review (Taggart 2001) we know of two more published studies (Endo 2001; Ioannidis 2001) and two more RCTs that are ongoing but have not yet published long term results (ART 2007; CARACCASS 2007; Pawlaczyk 2006a; Pawlaczyk 2006b; Taggart 2006). A systematic search of the literature may reveal more studies therefore the aim of this systematic review is to identify more studies reporting clinical outcome with single or bilateral IMA conduits to determine if there are any harms or benefit from the use of bilateral IMA. Conduct of a formal meta‐analysis is deemed to depend on the existence of studies of adequate methodological quality with sufficient outcome data. In particular, preoperative data regarding age, sex, ventricular function and the presence of diabetes are sought. The primary outcome measure is survival.
Objectives
To review the effects of bilateral versus single internal mammary artery grafts for coronary bypass grafting.
Methods
Criteria for considering studies for this review
Types of studies
Non‐randomised controlled studies comparing single with bilateral internal mammary artery bypass grafting of coronary arteries.
Randomised controlled trials comparing single with bilateral internal mammary artery bypass grafting of coronary arteries.
Studies must include at least 100 patients in each group followed for at least four years.
Types of participants
Adult patients with coronary artery disease requiring coronary artery bypass graft surgery
Types of interventions
Unilateral or bilateral bypass graft onto coronary arteries of patients undergoing bypass surgery.
Types of outcome measures
PRIMARY OUTCOME MEASURE
Survival
SECONDARY OUTCOME MEASURES
(1) Morbidity e.g. length of stay in hospital, re‐hospitalisation;
(2) Redo CABG;
(3) Myocardial infarction.
Search methods for identification of studies
The Cochrane Controlled Trials Register, MEDLINE, EMBASE will be searched from inception to date. No language restrictions will be applied. The reference lists of any identified papers will be scanned and expert advice sought.
Two authors will independently inspect the citation lists obtained from the electronic searches and from searching bibliographies and select citations for consideration. Either full length articles or abstracts will be obtained for each citation of interest. Differences of opinion will be resolved by discussion and, when necessary, in consultation with a third reviewer.
The following strategy will be used to search CENTRAL on the Cochrane Library and will be adapted appropriately for other databases.
#1 CORONARY ARTERY BYPASS (*ME)
#2 MYOCARDIAL REVASCULARIZATION (*ME)
#3 cabg
#4 (coronary near bypass)
#5 (myocard* near revasculari*)
#6 (heart near revasculari*)
#7 MAMMARY ARTERIES (*ME)
#8 (mammary near arter*)
#9 (((((((#1 or #2) or #3) or #4) or #5) or #6) or #7) or #8)
#10 (single near artery)
#11 (single near bypass)
#12 (single near graft*)
#13 (single near mammary)
#14 (single near ima)
#15 (single near imas)
#16 (double near artery)
#17 (double near bypass)
#18 (double near graft*)
#19 (double near mammary)
#20 (double near ima)
#21 (double near imas)
#22 (multiple near arter*)
#23 (multiple near bypass)
#24 (multiple near graft*)
#25 (multiple near mammary)
#26 (multiple near ima)
#27 (mutliple near imas)
#28 (bilateral near arter*)
#29 (bilateral near bypass)
#30 (bilateral near graft*)
#31 (bilateral near mammary)
#32 (bilateral near ima)
#33 (bilateral near imas)
#34 (unilateral near arter*)
#35 (unilateral near bypass)
#36 (unilateral near graft*)
#37 (unilateral near mammary)
#38 (unilateral near ima)
#39 (unilateral near imas)
#40 (unilateral near thoracic)
#41 (bilateral near thoracic)
#42 (mutliple near thoracic)
#43 (double near thoracic)
#44 (single near thoracic)
#45 bimas
#46 bima
#47 sima
#48 (((((((((#10 or #11) or #12) or #13) or #14) or #15) or #16) or #17) or #18) or #19)
#49 (((((((((#20 or #21) or #22) or #23) or #24) or #25) or #26) or #27) or #28) or #29)
#50 (((((((((#30 or #31) or #32) or #33) or #34) or #35) or #36) or #37) or #38) or #39)
#51 (((((((#40 or #41) or #42) or #43) or #44) or #45) or #46) or #47)
#52 (((#48 or #49) or #50) or #51)
#53 (#9 and #52)
Data collection and analysis
INCLUSION CRITERIA
An 'in/out' form will be prepared and used to assess studies for inclusion or exclusion using predetermined inclusion criteria. The in/out form will be tried on three reports, as a pilot, and if necessary refinements made. The inclusion of studies will be assessed using the in/out form independently by two reviewers and differences of opinion will be resolved by discussion and, when necessary, in consultation with a third reviewer. In all cases disagreements about any study inclusions will be resolved by consensus and a third reviewer will be consulted if disagreement persists.
QUALITY ASSESSMENT OF NON‐RANDOMISED STUDIES
Quality of studies will be assessed according to six criteria in three domains: cohort selection, cohort comparability, and outcome (see below). The assessment scheme is partly based on the Ottawa‐Newcastle system (Wells 2000), in particular the use of stars for each item.
Domain 1: Cohort selection will be assessed on the answers to three questions:
(1) Were details of criteria for assignment of patients to treatments provided? (One star if relevant details provided);
(2) How representative was the exposed cohort? (One star if representative of typical patient undergoing CABG, no star if groups of patients were selected or selection of group was not described);
(3) How was the non‐exposed cohort selected? (One star if drawn from the same community as the exposed cohort; no star if drawn from a different source, or selection of groups not described).
Domain 2: Cohort comparability will be assessed on the basis of study design or analysis of cohort differences:
(4) No differences between the two groups, or differences controlled for, in particular with reference to age, sex, ventricular function, or diabetic status (two stars). (One star will be assigned if one of these four characteristics was not reported, even if there were no other differences between the groups, and the other characteristics had been controlled for. No star will be assigned if the two groups differ).
Domain 3: Outcome will be assessed by two criteria:
(5) Assessment of outcome (one star for information ascertained by record linkage or interview; no star if this information was not reported or ascertained in some other way);
(6) Adequacy of cohort follow up (one star if no patient or fewer than 20% of patients lost to follow up; no star if more than 20% of patients were lost to follow up, or if the researchers did not provide relevant information).
DATA EXTRACTION
A data extraction form will be used to record relevant data from reports selected for inclusion. The data extraction form will be tried on three reports, as a pilot, and if necessary refinements made. The data extracted will be assessed independently by two reviewers and differences of opinion will be resolved by discussion and, when necessary, in consultation with a third reviewer. If necessary study authors will be contacted to obtain clarification of data or to obtain missing data.
QUALITY ASSESSMENT OF RANDOMIZED CONTROLLED TRIALS
The quality of trials will be assessed in terms of concealment of allocation, losses to follow up, and blind assessment of outcomes as described in the Cochrane Handbook (Clarke 2001).
STATISTICAL ANALYSES
The hazard ratio (HR) with 95% confidence interval will be estimated for each study with adequate data. If necessary, the log hazard ratio and its variance will be estimated from survival curves and tabulated results using the method described by Parmar and colleagues (Parmar 1998).
Meta‐analysis of non‐randomised studies will be deemed possible for those studies where either the patient characteristics (age, sex, ventricular function, diabetes) were similar at baseline or where multiple regression analysis was used to adjust for any differences.
The overall log hazard ratio can be obtained by computing the weighted average of the individual log hazard ratios, with the weights inversely proportional to the variance of the log hazard ratio of each study. Briefly, this means that large studies with a smaller standard error will be given more weight than smaller studies, which have a larger standard error. We intend to use a random effect meta‐analysis.
If any randomised studies are found their results will not be pooled with the results of non randomised studies.
If substantial heterogeneity (P < 0.1) is detected, the reviewers will look for possible explanations (e.g. participants and intervention) for this.
Regression modelling will be considered to examine the effect of age, diabetes, and disease severity on treatment effect, if there are enough studies to sustain such an analysis.
Dichotomous outcomes will be expressed as relative risks, and 95% confidence intervals (CI) will be calculated for each study.
Sensitivity analysis will be carried out, excluding studies of low methodological quality.
