Objective Coronary artery bypass grafting (CABG) using bilateral internal mammary arteries (BIMA) may improve survival over CABG using single internal mammary arteries (SIMA), but may be surgically more complex (and therefore costly) and associated with impaired sternal wound healing. We report, for the first time, a detailed comparison of healthcare resource use and costs over 12 months, as part of the Arterial Revascularisation (ART) Trial.
Methods 3102 patients in 28 hospitals in seven countries were randomised to CABG surgery using BIMA (n=1548) or SIMA (n=1554). Detailed resource use data were collected covering surgery, the initial hospital episode, and for 12 months post randomisation. Using UK unit costs, total costs were calculated and compared between trial arms and for subgroups.
Results Patients randomised to BIMA spent 20 min longer in theatre (95% CI 15 to 25, p<0.001) and also required more treatment for sternal wound problems. Mean (SD) total costs per patient at 12 months were £13 839 (£10 534) for BIMA and £12 717 (£9719) for SIMA (mean cost difference £1122, 95% CI £407 to £1838, p=0.002). No tests for interaction between subgroups and treatment allocation were significant.
Conclusions At 12 months from randomisation, mean costs were approximately 9% higher in BIMA than SIMA patients, primarily due to longer time in theatre and in-hospital stay, and slightly higher costs related to sternal wound problems during follow-up. Follow-up to the primary trial endpoint of 10 years will reveal whether longer-term differences emerge in graft patency or in overall survival.
Trial registration number Controlled-trials.com (ISRCTN46552265).
Statistics from Altmetric.com
During coronary artery bypass grafting (CABG) most patients receive three bypass grafts, one to each of the major coronary arteries. For almost three decades routine practice has been to graft a single internal mammary artery (SIMA) to the left anterior descending coronary artery and to use vein or radial artery grafts to the other coronary arteries.1 Better long-term patency of mammary artery grafts and evidence of improved 10-year survival and reductions in recurrent angina, myocardial infarction and repeat revascularisation suggest that patient outcomes could be further improved by using bilateral internal mammary arteries (BIMA).2 However, concerns that BIMA is technically more challenging and may increase surgery-related mortality and sternal wound complications have restrained use of BIMA in Europe or the USA: it accounted for <5% of US cases in 2009.3–5
A meta-analysis reported a significant reduction in mortality with BIMA over SIMA (HR 0.79, 95% CI 0.75 to 0.84) across nine observational studies including 15 583 patients with mean follow-up exceeding 9 years.6 However, evidence from randomised trials of long-term survival benefits with BIMA is still awaited. The multinational Arterial Revascularisation Trial (ART) (one of the largest trials ever conducted in cardiac surgery) randomised 3102 CABG patients to SIMA (n=1554) or BIMA (n=1548) and will eventually provide valuable information on the impact of BIMA on 10-year survival and the need for repeat revascularisation.7 ART has, however, already reported on the ‘safety’ of the procedure, finding similar clinical outcomes across trial arms at 1 year post randomisation, and a small absolute increase (1.3%) in the need for sternal wound reconstruction with BIMA.8
As CABG is a high volume procedure (in England and Wales, for example, around 16 000 first time surgeries annually) it is important to consider the potential impact on costs as well as effectiveness of a move from SIMA to BIMA grafts. Therefore a health economic evaluation was designed as an integral part of ART and will ultimately report on the cost-effectiveness of BIMA versus SIMA at 10 years. Here we report an analysis of detailed healthcare resource use data collected in each trial arm out to 1 year post randomisation, allowing unbiased comparison of the costs of SIMA and BIMA for the first time, and direct quantification of any short-term cost increases with BIMA on account of additional surgical complexities and impaired wound healing. Secondary aims are to explore resource use and cost differences between SIMA and BIMA for clinical subgroups where surgical outcomes could differ, and to consider variations in resource use across different countries in the trial.
The findings should be of interest to clinicians and healthcare policymakers considering the potential cost implications of moving from SIMA to BIMA, and will also be essential inputs into the final cost-effectiveness analysis of the ART trial, alongside estimates of longer-term costs, repeat revascularisation, survival, and health-related quality of life.
Full details of the ART protocol, baseline and 1 year safety outcomes are published elsewhere.7 8 ART is a multicentre randomised controlled trial involving 28 hospitals across seven countries and randomised CABG patients in a 1:1 ratio to receive SIMA or BIMA grafts. Patients were eligible for ART if they had multivessel coronary artery disease and were to undergo CABG. Patients requiring single grafts or re-do CABG were excluded. The ART study complied with the Declaration of Helsinki. Prior ethical approval was obtained in participating centres and each patient was required to provide written informed consent. The Clinical Trials and Evaluation Unit (CTEU) at the Royal Brompton and Harefield NHS Foundation Trust in London provided central coordination, and the study was sponsored by the University of Oxford.
ART was designed to detect an absolute 5% reduction in 10-year all-cause mortality (from 25% to 20%) with 90% power at a 5% significance level, and randomised 1554 patients to SIMA and 1548 patients to BIMA between 30 June 2004 and 20 December 2007. Data on healthcare resource use and health-related quality of life (EuroQol EQ-5D and shortened WHO Rose Angina Questionnaire)9 10 are also being collected.
Measurement of resource use
During the initial inpatient admission, data were collected for each patient on time in theatre (arrival in the anaesthetic room to final skin closure), total cardiopulmonary bypass time, blood products, platelets and fresh frozen plasma used, use of a cell saver machine for autologous blood transfusion, occurrence and duration of return to theatre, need for additional packed red blood cells (PRBCs), time receiving ventilation, use of an intra-aortic balloon pump, days receiving inotropic support or renal support therapy, and use of haemofiltration drugs.
Patient level data were also collected on treatments for sternal wound infections (antibiotics, debridement, vacuum assisted closure (VAC) dressings, and/or sternal reconstruction), serious adverse events (myocardial infarction, cerebrovascular accident, major bleed/other vascular events, further CABG or percutaneous transluminal coronary angioplasty (PTCA), and other), total hours in intensive therapy unit (ITU), total days in high dependency unit (HDU), and total days on a cardiac surgery ward.
Discharge to home, local hospital, nursing home or elsewhere was recorded, as were medications prescribed. At 6 weeks post randomisation information was collected on medication use, subsequent sternal wound infections and their treatment, serious adverse events, visits to a general practitioner (GP), practice nurse, hospital outpatient clinic, or cardiac rehabilitation clinic, and duration of any hospital readmission. These same data were captured by telephone interview at 12 months post randomisation.
Measurement of costs
Costs were not evaluated separately for each geographical location (by centre or country). Instead, UK unit costs (£ 2013/14) taken from national and local sources were used to value all patient-level resource use data. The perspective used was that of the healthcare system, and out-of-pocket costs such as travel to GP surgeries were not collected. Hospital costs are based on standard tariffs such as NHS reference costs. Where individual drug usage was not available, for example for antibiotics after surgery, assumptions based on clinical opinion and local procedures at one UK hospital were used. Supplementary table 1 and accompanying text provide a detailed description of sources, methods and assumptions.
Approximately 4% of trial resource use data items were missing across the 12 months; these were assumed to be missing at random and multiple imputation (MI) with chained individual linear or logistic regression equations was used to impute missing values for each variable.11 All imputation equations included age, gender, treatment allocation, diabetes and smoking. Five values were imputed for each missing data cell and Rubin’s Rule used to summarise across the imputed datasets.11
Continuous data were summarised using means and SD and categorical data using percentages. When comparing between trial arms, mean differences and 95% confidence intervals for differences were calculated, and two-sample t-tests were applied. All data analyses were performed using STATA 12 (StataCorp, College Station, Texas).
The low frequency of many events (for example, inpatient serious adverse events (SAEs) and sternal wound reconstructions) and the low unit costs of many elements of care (for example, inhalational anaesthetics) mean that even extreme changes to many unit costs and resource use assumptions would not significantly change study results. Sensitivity analyses were used to explore the effect of imputing missing data, and of including only patients who received the surgery they were allocated.
Resource use and costs were compared between BIMA and SIMA arms for the following subgroups: diabetic versus non-diabetic, age ≥70 years versus <70 years, on-pump versus off-pump, prior MI versus no prior MI, New York Heart Association (NYHA) class I and II versus NYHA class III and IV, and Canadian Cardiovascular Society (CCS) class 0, I and II versus CCS class III and IV. Twelve month costs were also compared across the three countries each recruiting more than 100 patients to the trial (UK, Poland, Australia).
Table 1 shows the baseline characteristics for patients in each arm of the trial. The groups were well balanced with respect to demographic and clinical characteristics, and disease/symptom severity.
Sternal wound infections were infrequent but twice as many occurred in patients in the BIMA arm than in the SIMA arm (3.3% v 1.6%, respectively). Sternal wound infection occurring in combination with dehiscence also affected more patients in the BIMA arm (1.6% v 0.6%, respectively). Mortality at 30 days and 12 months was similar across both trial arms.
Table 2 shows mean resource use and costs per patient for the initial inpatient admission by trial arm (intention to treat) for all patients. The cost of initial surgery, cost of initial inpatient admission (not including initial surgery), and the cost between discharge and 12 month follow-up are also shown in figure 1 for each trial arm. Theatre duration in the BIMA arm was on average 20 min longer than in the SIMA arm (95% CI 15 to 25, p<0.002). There were no statistically significant differences between trial arms in time on cardiopulmonary bypass, blood product utilisation or use of cell saver equipment, and similar proportions of patients in both arms were returned to theatre. In the immediate postoperative period, time on ventilation and usage of intra-aortic balloon pumps, inotropes, renal support therapy and haemofiltration were similar between arms.
The proportion of patients receiving treatments for sternal wound problems was small, but significantly more patients in the BIMA arm had treatment involving antibiotics, VAC dressings, and underwent debridement and sternal wound reconstruction (table 2). The proportion of patients experiencing serious events during the initial inpatient episode was similar in both arms, but with a trend towards an increased number of subsequent revascularisations with PTCA in the BIMA arm. On average patients in the BIMA arm spent an additional 2 hours in ITU and stayed a third of a day longer on cardiac surgery wards (table 2).
The total cost of the initial inpatient admission was £10 601 (£6601) in the BIMA arm and £9811 (£6390) in the SIMA arm, giving a mean cost increase of £791 per patient (95% CI £333 to £1248, p<0.001). Results were similar after adjustment for age, sex and diabetes: BIMA was associated with a cost increase of £778 (95% CI £323 to £1233, p=0.001) over the initial inpatient admission. Longer time in theatre and on various hospital wards by patients in the BIMA arm accounted for approximately two thirds of additional costs incurred. Treatments for additional sternal wound problems accounted for a further 7%, and renal support therapy and longer ventilation time for a further 8% each.
Table 3 shows mean resource use and cost per patient in each trial arm from initial hospital discharge to 12 months follow-up. Similar proportions of patients in each arm were transferred to local hospitals, nursing homes and other institutions. At 6 weeks the overall number of patients requiring treatment for sternal wound problems was small but more patients in the BIMA arm received antibiotics and debridement.
The two trial arms were comparable in terms of GP and nurse visits, cardiac rehabilitation visits and hospital readmissions to 12 months, but patients in the BIMA arm had on average 0.5 more (95% CI 0.1 to 0.9, p=0.015) hospital outpatient clinic visits than patients in the SIMA arm. There were no differences between trial arms in the proportion of participants experiencing SAEs, or the cost of these, between discharge and 12 months.
The total cost of the follow-up period from hospital discharge was £3238 (£7118) in the BIMA arm and £2906 (£6203) in the SIMA arm (mean cost difference £332, 95% CI −£141 to £805, p=0.169). The mean (SD) overall total cost per patient out to 12 months from randomisation was £13 839 (£10 534) in the BIMA arm and £12 717 (£9717) in the SIMA arm (mean cost difference £1122, 95% CI £407 to £1838, p=0.002) (figure 1).
Results were similar after adjustment for age, sex and diabetes: BIMA was associated with £1081 (95% CI £376 to £1787, p=0.003) higher costs per patient out to 12 months from randomisation.
One thousand four hundred and ninety-four of the 1554 patients allocated to SIMA and 1294 of the 1548 patients allocated to BIMA received the allocated procedure. Supplementary table 2 reports a ‘per protocol’ analysis comparing costs by trial arm: as in the base case analysis, total inpatient costs were significantly greater for those receiving BIMA than for SIMA, and total costs at 12 months were significantly greater for BIMA than for SIMA (£1243 more costly for BIMA, with 95% CI £501 to £1985, p<0.001).
Supplementary Table 3 reports total costs by trial arm when no multiple imputation was performed and calculations were based on ‘complete cases’ only. As in the base case analysis, the total cost of the initial hospital stay was significantly greater for BIMA than for SIMA, driven by time in theatre and ward stays, and there was a non-significant trend towards BIMA being more costly than SIMA at 12 month follow-up.
Table 4 shows total costs to 12 months for the various subgroup analyses and country comparisons; Supplementary table 4–10 provide full details. Tests showed no evidence of interaction between subgroup and treatment allocation, although additional costs with BIMA for diabetic patients were more than twice those of non-diabetic patients (£2119 v £803 per patient, respectively), for on-pump patients were more than twice those for off-pump patients (£1575 v £623 per patient, respectively), and also were higher for patients with more severe angina and cardiac disease.Subgroup
This analysis was an integral part of the ART Trial, the largest randomised comparison of bypass grafting using single (SIMA) or bilateral (BIMA) mammary arteries. Our comprehensive results show that around 70% of the additional costs observed in the BIMA arm at 12 months (£791, 95% CI £333 to £1248, p=0.001) were incurred during the initial inpatient episode, mainly due to longer time in theatre and on hospital wards. Treatment costs associated with a small increase in sternal wound problems were significant but not sizeable. Other inpatient resource use and clinical events were similar across both trial arms.
Post-hospital discharge, and with the exception of the costs of sternal wound problems at 6 weeks, there were few differences between the two trial arms to 12 months. The absence of manifestations of surgery-related complications requiring treatment in the BIMA arm over this period is reassuring. Any signs of the hypothesised benefits of BIMA—sustained graft patency leading to a reduction in repeat revascularisation, lower use of anti-anginal medication and improved survival —are only likely to become evident in the longer term. Previous trial-based studies of the cost-effectiveness of CABG versus PTCA have consistently found that differences in costs arising from repeat revascularisation and medication use take time to emerge; in the SYNTAX trial these emerged over 5 years and increased when simulated over a lifetime12; in BARI differences did not emerge until after 12 months and persisted for at least 7 years.13 Differences between BIMA and SIMA may take even longer to appear due to the good long-term graft patency already achievable with SIMA, and a full analysis of costs, outcomes and cost-effectiveness is planned at the primary ART endpoint of 10 years.
Exploratory subgroup analyses suggested that the incremental costs of BIMA over SIMA at 12 months could be around £1300 higher for diabetic patients as compared with non-diabetic patients, incurred mainly following hospital discharge. However, there was no evidence of interaction between subgroup and treatment allocation.
Diabetes and BIMA have been reported as independent risk factors for deep sternal wound infection following CABG in a number of observational studies performing multivariate analyses.14 15 However, a recent study including 1 526 360 CABG patients treated in the USA reported that while diabetes mellitus was an independent predictor of deep sternal wound infection, BIMA was not, and was only associated with an increased risk of deep sternal wound infection in patients with chronic complications of diabetes mellitus.16 We found that, during the initial inpatient admission, the higher mean cost per patient of treatment for sternal wound problems in the BIMA arm compared with SIMA was more pronounced among diabetic than non-diabetic patients, although the absolute costs involved were relatively small. This pattern was even stronger for sternal wound problems occurring between discharge and 6 weeks (Supplementary table 4).
We found some evidence that the BIMA versus SIMA difference in cost was greater for on-pump patients: a £1575 excess compared with £623 for off-pump patients. This appears to be driven by the BIMA group: within the SIMA group, off-pump and on-pump patients appeared to have very similar costs, although the non-randomised comparison could be confounded, and there was no evidence of interaction between subgroups and treatment allocation.
The main advantage of our study is the randomised comparison, which greatly reduces the risk of unobserved bias. The retrospective database analysis by Itagaki and colleagues had a very large sample size, but it is impossible to be sure that reported differences between SIMA and BIMA were not confounded by other variables; thus, they report a shorter length of stay for BIMA versus SIMA (9.0 vs 8.0 days) and lower costs ($85 246 vs $92 698), both in the opposite direction to our findings.16
Limitations of our study include the fact that we have applied UK-based unit costs to resource use information from all seven countries in the study, rather than applying local costs by centre and country and then applying an estimation model. Differences in patterns of care between countries may reflect different relative prices: for example, cost differences between BIMA and SIMA related to the initial inpatient admission were highly significant for Poland but not significant for UK and Australia (all with BIMA more costly), while initial surgery costs were lower for BIMA in Australia but significantly higher for BIMA in UK and Poland, and costs of healthcare contacts (GP, etc) were lower for BIMA than SIMA in Australia but higher in UK and Poland. However, these differences were mainly not statistically significant, total cost differences between BIMA and SIMA at 12 months were in the same direction for all countries, and there was no evidence of interaction between subgroups and treatment allocation.
At 12 months from randomisation, mean costs were higher in BIMA than SIMA patients, primarily due to longer time in theatre and in-hospital stay, and slightly higher costs related to sternal wound problems. Follow-up to the primary trial endpoint of 10 years is continuing, and will reveal whether longer-term differences emerge in graft patency or in overall survival.
What is already known about this subject?
Coronary artery bypass grafting using the single internal mammary artery (SIMA) is a safe, effective and high volume procedure, but patient outcomes might be further improved by using bilateral internal mammary arteries (BIMA), which may offer better long-term graft patency and survival. No randomised comparison of the costs of these procedures has been published.
What does this study add?
Using data from the Arterial Revascularisation Trial, we show that at 12 months from randomisation costs were approximately 9% higher in BIMA than SIMA patients, mainly due to longer time in theatre and in-hospital stay, and slightly higher costs related to sternal wound problems during follow-up. These cost differences were larger in some subgroups, such as diabetes versus non-diabetes.
How might this impact on clinical practice?
Our findings will be valuable to clinicians and healthcare policymakers considering the potential cost implications of moving from SIMA to BIMA, for all eligible patients or particular subgroups. Researchers will require these cost estimates to assess the long-term cost-effectiveness of BIMA.
We thank all the patients who are participating in this trial in the seven countries worldwide; the investigators at the participating centres, the members of the steering committee, data and safety monitoring committee, and the clinical-events adjudicators, Dr Jeremy Pearson (British Heart Foundation) and Dr Mark Pitman (UK Medical Research Council), for support throughout; Ms Emma Haines, Ms Sarah Dutton and Mr Edmund Wyatt (Oxford Clinical Trials Research Unit) and Ms Eva Matesanz, Mr Wajid Aslam and Ms Fiona Nugara (Clinical Trials and Evaluation Unit) for data coordination and management.
Contributors AG designed and led the economic analysis. JM and HC conducted the data analysis for this study. DT conceived and leads the ART Trial and is principal investigator. DA and SG provided statistical expertise and analysis. JM and AG drafted the manuscript. DA, UB, HC, MF, SG, BL and DT all contributed to refinement of this study and commented on drafts of the manuscript. All authors approved the final manuscript.
Competing interests None declared.
Patient consent Every patient was required to provide written informed consent. No identifiable information is presented in this analysis.
Ethics approval Local Ethics Committee of each participating hospital.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement All requests concerning data from the ART trial should be directed in the first instanced to the Principal Investigator, David Taggart, at Nuffield Department of Surgical Sciences, University of Oxford, John Radcliffe Hospital, Oxford, OX3 9DU, UK.
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.