Coronary artery disease is a major cause of morbidity and mortality among patients with diabetes.1 The presence of diabetes is associated with at least a doubling of the risk for subsequent cardiovascular disease and cardiovascular death,2–4 and the diagnosis of coronary artery disease is more difficult in patients with diabetes owing to frequent atypical symptoms or silent ischaemia. Furthermore, among patients with diabetes, coronary artery disease presents more often as a diffuse process with involvement of multiple segments and vessels at time of diagnosis, is more likely to progress and more often results in death or myocardial infarction (MI).5,6 These hallmark features of diabetic coronary artery disease have led to unfavourable outcomes, regardless of treatment strategy.
Percutaneous coronary intervention (PCI) has fared especially poorly in this population, with adverse effects attributed to higher rates of restenosis, incomplete revascularisation and disease progression at untreated coronary segments.7–9 While coronary artery bypass graft surgery (CABG), has generally performed better than PCI in most comparison studies, outcomes have remained poorer for patients with diabetes compared with those without diabetes,10,11 and it is uncertain if any revascularisation is preferable to medical therapy in many patients.12 It is likely, however, that patients with large areas of ischaemia and more extensive disease, including left main and three-vessel involvement, are more likely to benefit from revascularisation. For these patients, it is important to determine when PCI is an appropriate strategy.
Historical Clinical Trial Results
Balloon Angioplasty versus Coronary Artery Bypass Grafting
Among patients with left main or multivessel coronary artery disease who require revascularisation, comparative outcomes for PCI versus CABG have been investigated since the early years of percutaneous transluminal coronary angioplasty (PTCA; or balloon angioplasty) and debate continues for patients with or without diabetes. In general, randomised clinical trials performed during PTCA or bare-metal stent (BMS) eras have not shown a significant difference in mortality during short- and long-term follow-up in the overall population, although repeat revascularisation rates have been uniformly higher for PCI.13–17 Within the subgroup of patients with diabetes, however, a mortality benefit for CABG versus PCI is accepted by many, although reported results have not been consistent or always statistically significant.
A mortality benefit for CABG was first reported with five-year follow-up from the Bypass angioplasty revascularization investigation (BARI) randomised clinical trial in 1997. While the results showed no difference in five-year survival overall (89.3 verus 86.3 %) for CABG compared with PTCA,16 survival was significantly higher for those randomised to CABG within the subgroup with treated diabetes (80.6 versus 65.5 %, p=0.003).11 Since the initial publication of the BARI results, other randomised trials of PTCA or BMS versus CABG have also reported on the diabetic subgroup. In the Emory angioplasty versus surgery trial (EAST), survival was similar after three years for patients with or without diabetes regardless of revascularisation strategy. However, survival curves began to diverge after five years and, although the difference did not reach statistical significance, by eight years, survival was measurably higher for patients with diabetes assigned to CABG (75.5 versus 60.1 %, p=0.23).18 In contrast, survival was nearly identical for CABG versus PTCA in patients without diabetes (84.0 versus 82.6 %, p=0.71).
The Coronary angioplasty versus bypass revascularization investigation (CABRI) study reported similar findings at four years with a mortality rate of 22.6 % in the PTCA group compared with 12.5 % in the CABG group, although these results also did not achieve statistical significance.19 Compared with patients without diabetes, those patients with diabetes had higher mortality in the overall cohort (17.8 versus 8.1 %, p=0.001) and in the PTCA group (22.6 versus 9.4 %, p=0.001); but a smaller and non-significant difference in the CABG group (12.5 versus 6.8 %). Despite the absence of statistical significance in CABRI and EAST, the direction and magnitude of benefit for CABG among patients with diabetes support the initial BARI results. The BARI results also proved durable after 10 years of follow-up.20 Only one study in the PTCA era found contrary results. The Randomised intervention treatment of angina (RITA-1) trial reported no difference in mortality for the overall cohort, but actually noted a trend towards increased mortality for CABG versus PTCA within the diabetic subgroup (24.2 versus 6.9 %, p value for interaction = 0.09).21 Overall mortality in RITA-1 was low, possibly related to inclusion of 45 % of patients with single vessel disease, and as in CABRI and EAST there were small numbers of patients in the diabetic subgroup (see Table 1).
Bare-metal stents versus Coronary Artery Bypass Grafting
There have been two randomised studies that compared BMS versus CABG and included results for the diabetic subgroup. Despite an expected benefit related to lower restenosis for BMS compared with PTCA, the results of these studies were similar to prior studies of PTCA versus CABG. In the Arterial revascularization therapies study (ARTS) five-year outcomes for the overall cohort showed no difference in mortality for stenting compared with CABG (8.0 versus 7.6 %, p=0.83), but a significantly higher rate of repeat revascularisation (30.3 % vs. 8.8 %, p<0.001).17 Patients with diabetes, however, had significantly higher mortality after stenting compared with patients without diabetes (13.4 versus 6.8 %, p=0.03), while patients with diabetes assigned to CABG had mortality similar to those without diabetes (8.3 versus 7.5 %, p=0.80). This translated into measurably lower mortality for CABG within the diabetes subgroup, but again the observed difference did not achieve statistical significance. In the Stent or surgery (SoS) study, the benefit for CABG compared with BMS was generalised to the overall cohort without a significant interaction between treatment assignment and diabetes.22 At a median follow-up of six years, mortality was significantly lower for CABG in the overall cohort (6.8 versus 10.9 %, p=0.022). Although the observed difference was greater for the diabetes subgroup (5.4 versus 17.6 %), the p value for interaction was not significant (0.15).
Summary of Percutaneous Transluminal Coronary Angioplasty and Bare-metal Stent Data and Results of Meta-analyses
With the exception of the RITA-1 trial, all of the studies reported above have observed a mortality benefit for CABG versus PCI in the diabetes subgroup. While the treatment interaction with diabetes was not significant in SoS, this is related more to the finding of a mortality effect even among patients without diabetes. This observation from SoS has been attributed in part to an early observed difference in non-cardiac mortality due to cancer, but the difference persisted during six-year follow-up and included a similar difference in cardiac mortality.22,23 Furthermore, although all of the studies except BARI were underpowered to detect clinically meaningful differences in mortality, the consistency of the observed relative risk suggests the observations are unlikely due to chance.
To help address the issue of statistical power, several meta-analyses have assessed mortality differences from the individual studies. Hoffman et al. combined BARI, EAST, RITA-1 and CABRI and concluded that after four years CABG demonstrated a mortality benefit over PTCA, but that this was no longer significant at 6.5 years.24 This meta-analysis reported significant heterogeneity between these studies, which became non-significant with removal of RITA-1 and resulted in a significant benefit for CABG also at 6.5 years. The most convincing evidence comes from a pooled analysis of individual patient data from 10 randomised trials of PTCA or BMS versus CABG, including 615 patients with diabetes.25 This study found no difference in mortality overall or among patients without diabetes, but mortality was lower for CABG among patients with diabetes (hazard ratio 0.70, 95 % CI 0.56–0.87, p value for interaction 0.014).
In addition, although most studies did not report on revascularisation rates separately for patients with and without diabetes or distinguish revascularisation due to restenosis or disease progression, all studies have reported an overall benefit for CABG for reduced repeat revascularisation. The ARTS study provided the most detail, demonstrating a significant difference in revascularisation frequency between BMS and CABG at five years that was greater for patients with diabetes (42.9 versus 10.4 %, p<0.001) than those without diabetes (27.5 versus 8.4 %, p<0.001). Much of this difference occurred in the first year and is likely to be due to restenosis, but the differences continued to increase between one and five years, suggesting a substantial role for disease progression.26,27
Impact of Drug-Eluting Stents
During the early period of exuberant enthusiasm for drug-eluting stents (DES), it was hoped that with avoidance of restenosis and perhaps more complete revascularisation associated with widespread stenting of moderate disease outcomes after PCI would equal or surpass those after CABG among patients with diabetes. It should be noted this enthusiasm was prior to the reported concerns with late catch-up restenosis and stent thrombosis that have been especially problematic in patients treated for more complex coronary disease.28,29 While the specific impact of DES versus CABG is being assessed in the Future revascularization evaluation in patients with diabetes mellitus (FREEDOM) randomised trial,30 some preliminary insights are available from other studies.
The ARTS II study enrolled 607 patients receiving a sirolimus-eluting stent (SES), including 159 patients with diabetes, and compared outcomes to the historical results for BMS and CABG from ARTS I.27,31 After five years, the rate of major adverse cardiac or cerebrovascular events (MACCE), defined as death, MI, repeat revascularisation or stroke, for SES was slightly lower compared with BMS (40.5 versus 53.6 %, p<0.01) but significantly higher than CABG (40.5 versus 23.4 %, p<0.01). This finding was driven by a repeat revascularisation rate for SES that remained lower but approached that of BMS by five years (33.2 versus 43.7 %, p=0.02), mostly due to events occurring in the SES group between years 1–5. Notably, mortality was similar for SES compared with CABG (9.0 versus 8.6 %, p=0.91).
The Coronary artery revascularization in diabetes (CARDIA) trial randomised 510 patients with multi-vessel or complex single-vessel coronary artery disease to stenting versus CABG including 350 patients after the availability of DES.32 This study is limited by only one year follow-up, during which time there was no difference in the composite endpoint of death, MI or stroke within the DES subgroup for PCI versus CABG (11.6 versus 12.4 %). Overall mortality was 3.2 % in both the PCI and CABG groups.
The best available data for DES comes from the diabetes subgroup of the Synergy between percutaneous coronary intervention with taxus and cardiac surgery (SYNTAX) trial. SYNTAX randomised 1,800 patients with three-vessel or left main coronary artery disease to CABG versus paclitaxel-eluting stent (PES) and included 452 patients in the diabetes subgroup.33 After three years of follow-up, the primary endpoint of MACCE was significantly higher in patients with diabetes assigned to PES versus CABG (37.0 veruss 22.9 %, p=0.002).34 This difference was mainly due to a higher rate of repeat revascularisation in the PES group (28.0 versus 12.9 %, p<0.001), although mortality was also numerically higher for PES (13.6 versus 8.7 %, p=0.11). As in the overall SYNTAX population, analysis of subgroups determined by lesion complexity according to the SYNTAX score was also of interest within the diabetes subgroup. For patients with a low SYNTAX score (≤22), MACCE rates were not different for PES versus CABG (29.8 versus 30.5 %, p=0.98), despite repeat revascularisation remaining more frequent (29.8 versus. 16.8 %, p=0.07).
On the other hand, in patients with intermediate (23–32; [36.2 versus 21.0 %, p=0.04]) or high (>32; [45.9 versus 8.5 %, p<0.001]) SYNTAX scores MACCE was significantly higher in the PES group. For the high SYNTAX score subgroup, the composite of death, MI or stroke was also significantly worse for PES (22.9 versus 8.9 %, p=0.03). Given multiple comparisons and subgroup analyses these results should be interpreted with caution and considered hypothesis generating whilst we await the results from FREEDOM.
Rationale and Predictors of Benefit for Coronary Artery Bypass Grafting in Patients with Diabetes
Patients with multivessel or left main coronary artery disease who undergo CABG versus PCI as initial therapy may benefit due to avoidance of restenosis and protection from disease progression at sites not treated by PCI. That restenosis occurs mostly within the first year while the risks of repeat revascularisation continue beyond five years and devices associated with lower restenosis, such as the case for BMS versus PTCA or DES versus BMS, have not significantly improved outcomes, argue that disease progression is a bigger factor. This is an important distinction, since if restenosis were the major issue then improved PCI techniques with progressively lower restenosis risk would offer hope for PCI as a treatment strategy. Likewise, if either restenosis or disease progression were associated only with development of flow-limiting lesions that might require repeat revascularisation, then the controversy over whether repeat revascularisation is an important endpoint if initial surgery can be avoided might also be used as support for PCI as an initial strategy. To the extent, however, that disease progression is manifest as new unstable or vulnerable plaques that lead to MI or death and is not prevented by current medical therapy or amenable to subsequent PCI, but is negated by prior surgical bypass, the debate in favour of CABG gains insurmountable support.9
One question is whether the mere presence of diabetes in association with multivessel or left main coronary disease adequately defines patients at increased risk who will benefit more from CABG than PCI. There are several lines of evidence that suggest that additional selection criteria are needed. In BARI, the mortality benefit for patients with diabetes was limited to those who received at least one mammary artery graft (mortality 2.8 %), whereas those who received only vein grafts had mortality similar to PTCA (18.2 versus 20.6 %).
The benefit of the mammary artery graft is likely due both to avoidance of late adverse effects related to saphenous vein graft failure as well as protection of patients at highest risk, namely those with involvement of the left main or proximal left anterior descending. The latter is supported by observational data from the Duke databank in which the benefit of CABG over PTCA or medical therapy was restricted to patients with three-vessel disease or two-vessel disease with severe proximal left anterior descending artery stenosis.35
The importance of patient selection, based on measurable but also unmeasured risk factors, is also an important issue. For example, in contrast to the BARI randomised trial results, there was no difference in all-cause mortality for CABG versus PTCA (14.9 versus 14.4 %, p=0.86) among patients with diabetes who underwent revascularisation based on physician selection and were followed in a parallel registry.36 Case selection may also explain the discrepant results as discussed above in RITA-1, which enrolled more patients with single-vessel disease and actually demonstrated lower mortality for PTCA compared with CABG.14 The SYNTAX trial offers a quantitative demonstration of the impact of lesion risk that is especially important within the diabetes subgroup.34 Indeed, for low SYNTAX scores, there was no difference in MACE despite a continued higher risk for repeat revascularisation, suggesting other components (death, MI or stroke) might be lower for DES compared with CABG in these patients. Thus, it is unlikely there is a single answer for all patients with diabetes who have multi-vessel or left main coronary disease and physician selection based on overall risk and benefit remains crucial. An algorithm to help select which patients may be appropriate for PCI is included in Table 2. It is hoped that upcoming results from FREEDOM will help add clarity to this persistent and challenging question.