There has been a dramatic change in coronary revascularization trends over the past decade. There was an enthusiastic uptake of drug-eluting stents (DES) after the US Food and Drug Administration approval in 2003 – this peaked at 90% of all stents deployed in the USA in 2005. Concern about early and late stent thrombosis, and the need for long-term dual anti-platelet therapy decreased DES use to 68% at the end of 2008. Overall rates of myocardial revascularization decreased by 15% between 2001 and 2008 – percutaneous coronary intervention (PCI) decreased by 4%, while coronary artery bypass grafting (CABG) decreased by 38%.
The Australian Institute of Health and Welfare (AIHW) annual data suggest a 63% increase in PCI procedures in 10 years from 2000–2001 (21 591 procedures) to 2009–2010 (35 851 procedures). The increase has plateaued somewhat, with only a 4.5% increase in the last 5 years. CABG procedures decreased by 25% in a linear fashion over the same period to 2009–2010 (12 375 procedures; Fig. 1). PCI surpassed CABG as the most common revascularization procedure in Australia in 1997–1998.
There are two Australian registries for PCI and CABG – the Melbourne Intervention Group (MIG, 7 Victorian public hospitals) and the Australian and New Zealand Society of Cardiac and Thoracic Surgeons (ANZSCTS, 21 out of 24 Australian public hospitals, 6 private hospitals offering cardiac surgery), respectively. Progress is underway to form a national registry that would include all hospitals offering CABG and PCI. MIG reported an increase in the overall number of PCI procedures performed between 2004 and 2007, and a greater proportion of all PCIs performed were urgent (overall decrease in elective PCI). The use of DES decreased from 53.9% to 39%. ANZSCTS reported a decrease in the total number of isolated CABG performed in six Victorian hospitals between 2001 and 2006. Patients were older and had a greater number of co-morbidities – nonetheless, 30-day mortality for isolated CABG (including emergency procedures) remained low (1.5–2.2%).
Studies published from these databases represent only a small number of centres and data need to be interpreted with care – not all centres are mandatory and audited; the data presented do not represent the net increase in patients receiving care in the private health sector, which may, in part, explain a decrease in the rate of CABG. Similarly, while AIHW reports data from public and private hospitals, the collection of the data (mainly from public hospitals) may not be complete. These flaws highlight the need for a mandatory national database of myocardial revascularization (and indeed other cardiac interventions) to better reflect the patterns of revascularization and their outcomes in Australia.
There is a consensus for the superiority of PCI over CABG in a number of instances – for culprit lesions in acute coronary syndromes, simple one- and two-vessel disease not involving the proximal left anterior descending artery (LAD), and in patients who may not survive coronary surgery. Similarly, there is a consensus for the benefit of CABG over PCI for three-vessel disease with impaired left ventricular (LV) function, total-vessel occlusion, stent thrombosis or where another surgical procedure is indicated (e.g. in valvular disease).
The results of a number of recently published trials using contemporary revascularization techniques make it timely for a review of the evidence in the ‘grey areas’ of myocardial revascularization for stable coronary artery disease – namely multi-vessel disease with preserved LV function and left main coronary artery disease (Table 1).
|SYNTAX[21, 32]||2009–2011 (3-year results)|| || |
Lower MACCE in CABG at 3 years (CABG 20.2%, PCI 28.0%, P < 0.001)
Three-vessel disease – CABG protective for:
Three-vessel disease – no difference
|New York Registry||2008|| || |
CABG protective against death:
CABG protective against death/MI:
|Weintraub||2012|| || |
No difference in mortality at 1 year
CABG protective against mortality at 4 years
|Yan et al.||2011|| || |
DES protective against:
CABG protective against:
|Capodanno et al.||2011|| || |
No difference between CABG and PCI:
PCI protective against stroke (CABG 1.7%, PCI 0.1%, P = 0.013)
CABG protective against :
|MASS II|| |
2010 (10 years results)
| || |
No difference in overall survival (CABG 74.9%, PCI 75.1%, MT 69%, P = 0.089)
CABG protective against:
No difference in stroke rate (CABG 8.4%, PCI 5.4%, MT 6.9%, P = 0.550)
|BARI 2D||2009|| || |
No difference in 5 years survival between prompt revascularization and medical therapy (revascularization 88.3%, MT 87.8%, P = 0.97)
No difference in MACE at 5 years (revascularization 77.2%, MT 75.9%, P = 0.70)
Significantly lower MACE in CABG stratum (but not PCI) versus MT (CABG 22.4% versus MT 30.5%, P = 0.01)
The historical basis of the benefits of CABG
A meta-analysis of 10-year outcomes in trials comparing CABG with medical therapy, published in 1994, established the survival benefits of CABG. While medical and surgical treatments used in the included trials (conducted in 1972–1984) for coronary artery disease have long since been superseded with current therapy, this meta-analysis established groups who benefit more from surgery.
The meta-analysis showed a disproportionate survival benefit of CABG when compared with medical therapy for patients with three-vessel disease (5.7 months), left main disease (19.7 months) or abnormal LV function (10.6 months). All patients with proximal LAD involvement benefitted from CABG. There was no significant difference in survival between CABG and medical therapy for one- or two-vessel disease (without proximal LAD involvement), normal LV and low-risk patients.
CABG provides additional prophylaxis against ischaemia from de novo lesions proximal to the graft. This differs from PCI, where existing lesions are treated generally at the proximal part of the vessel, and may explain the lower rate of revascularization in surgical patients in most trials.[16, 17]
Evidence to date: difficulties and pitfalls interpreting results
Twenty-six randomized trials, 24 observational and registry studies and several meta-analyses have compared PCI with CABG for multi-vessel disease in the English-speaking literature.[6, 18-20] Only 35% of all patients in these randomized trials had three-vessel disease, and only nine trials included treatment with stents. Only two included DES – the SYNTAX trial and BARI 2D – and these were first- rather than second-generation.[11, 19] There is evidence that second-generation DES have slightly better outcomes than first-generation, and these stents are not included in trials.
No trials comparing PCI with CABG employed surgical techniques, which may improve survival and morbidity – total arterial revascularization[14, 22-24] or off-pump coronary artery bypass (OPCAB) grafting,[25-28] especially when performed without manipulation of the ascending aorta.
Most randomized trials have included only low-risk patients (typically about 5% of the eligible population), excluded patients with poor LV or left main disease and had a short period of follow-up in what is a slowly progressive disease. It is for these reasons that critics of trials comparing PCI and CABG suggest that the benefits of surgery in trials are most likely underestimated.
The pertinent trials comparing CABG and PCI for multi-vessel disease and left main coronary artery disease are summarized in Table 1.
Revascularization (with PCI or CABG) versus medical therapy in stable coronary artery disease
The COURAGE trial showed no benefit to patients with stable angina treated by intensive optimal medical therapy with and without PCI. At 5 years, the composite rate of death, myocardial infarction (MI) or stroke was 20% in the PCI arm versus 19.5% in the medically treated arm (hazard ratio (HR): 1.05, 95% confidence interval (CI): 0.87–1.27, P = 0.62). COURAGE has been criticized by some for including only low-risk patients (with a perhaps overly ambitious goal of 22% reduction in death), a high rate of crossover from medical to PCI arm, low rate of DES and a regime of optimal medical therapy not achievable in the ‘real world’, but is supported by others. Nonetheless, the results of COURAGE have been supported by a recently published meta-analysis, including only randomized trials with contemporary PCI and medical treatments (8 trials, 7229 patients) – this showed no benefit of treatment with PCI for death, MI, repeat revascularization or angina compared with optimal medical therapy.
Mid- to long-term follow-up of two randomized trials comparing revascularization (both PCI and CABG) versus medical therapy have recently been published – MASS II and BARI 2D. MASS II randomized 611 patients with multi-vessel disease (58% three-vessel disease in the CABG group), stable angina and good ventricular function to management with either CABG, PCI (bare-metal stent – BMS) or optimal medical therapy. Primary endpoints evaluated were overall mortality, MI and refractory angina (Canadian class II or III) requiring revascularization. Analysis of results was by intention to treat. At 10 years, the cross-over rate from medical therapy to revascularization was 39%.
CABG was protective against the primary endpoints (PCI versus CABG, HR: 1.85, 95% CI: 1.39–2.47, P < 0.0001). There was no difference in overall mortality between groups at 10 years. Rate of cardiac death and MI was significantly lower after CABG (cardiac death: CABG 10.8%, PCI 14.3%, medical therapy 20.7%, P = 0.019; MI: CABG 10.3%, PCI 13.3%, medical therapy 20.7%, P = 0.010). Rate of additional intervention after CABG was less than one-fifth of that for PCI and medical therapy (CABG 7.4%, PCI 41.9%, medical therapy 39.4%, P = 0.001). There was no significant difference in the rate of stroke (CABG 8.4%, PCI 5.4%, medical therapy 6.9%).
The BARI 2D trial included 2368 patients with type 2 diabetes mellitus and coronary artery disease requiring revascularization. Included patients were nominated an appropriate revascularization strategy (PCI – DES in 34.7% – or CABG) by a physician and then randomized to their nominated revascularization strategy or intensive medical therapy. Primary outcome reported at 5 years was death, and major adverse cardiovascular event (MACE) – composite of death, MI or stroke.
There was no difference in survival or freedom from MACE between revascularization and medical therapy groups. When analysed by revascularization stratum, only in the CABG group was there benefit over medical therapy for MACE.
Revascularization in multi-vessel disease
The SYNTAX trial is the single most informative and relevant trial comparing PCI and CABG in stable multi-vessel disease. In this multi-institutional study, 1800 patients with severe three-vessel and/or left main coronary artery disease were randomized to treatment with either CABG (most often left internal mammary artery (IMA) with additional saphenous vein grafts (SVGs); 15% OPCAB) or PCI (using the TAXUS DES). Patients ineligible for randomization were nested in a separate registry (1077 CABG patients, 198 PCI patients). The primary endpoint was major adverse cardiac or cerebrovascular event (MACCE) comprising all-cause death, MI, need for repeat revascularization and stroke. Patients were further stratified based on their angiographic ‘SYNTAX score’, a novel score of the complexity of coronary disease. While it is subject to some inter-observer difference and takes about 10 min to calculate, it allows quantification of major morphological criteria beyond ‘three-vessel’ and ‘left main’ known to influence outcome.
The SYNTAX trial has overcome many of the shortcomings of other randomized trials comparing PCI and CABG – it included ‘all-comer’ patients with severe stable coronary disease (no MI <48 h), aggressively used drug-eluting rather than BMS (4.6 stents/patient; average total stent length of 86.1 mm) and stratified patients according to the complexity of their disease. Three-year follow-up results have recently been published and follow-up is planned for 5 years.
At 3 years, patients randomized to CABG had a significantly lower rate of the primary endpoint MACCE compared with PCI – 20.2% versus 28% (P < 0.001). Patients in the highest tercile of coronary disease complexity had the greatest benefit when treated with CABG compared with PCI. There was no difference in rates of MACCE for three-vessel disease or left main contrary artery disease at the lowest SYNTAX score tercile. For three-vessel disease, CABG patients had a significantly lower rate of MACCE for intermediate (CABG 16.8% versus PCI 29.4%, P = 0.02) and high (CABG 17.9% versus PCI 31.4%, P = 0.004) SYNTAX scores. These upper two terciles represented 79% of patients with three-vessel disease.
In patients with three-vessel disease, CABG was protective against the composite of death/MI/stroke (CABG 10.6% versus PCI 14.8%, P = 0.04), death (CABG 5.7% versus PCI 9.5%, P = 0.02), MI (CABG 3.3% versus PCI 7.1%), repeat revascularization (CABG 10.0% versus PCI 19.4%, P < 0.001) and MACCE (CABG 18.8% versus PCI 28.8%, P < 0.001). At 3 years, there was no difference in the rate of stroke. For patients with diabetes, CABG had a rate of repeat revascularization less than half that of PCI (CABG 12.9% versus PCI 28.0%, P = 0.001). This contributed to the lower rate of MACCE in diabetic patients (CABG 22.9% versus PCI 37.0%, P = 0.002).
Four-year SYNTAX follow-up results were presented at the European Association for Cardio-thoracic Surgeons in Lisbon, October 2011. These showed continuing divergence of the primary endpoint. CABG was superior to PCI for MACCE (CABG 23.6% versus PCI 33.5%, P < 0.001), all-cause mortality (CABG 8.8% versus PCI 11.7%, P = 0.048), cardiac death (CABG 4.3% versus PCI 7.6%), MI (CABG 3.8% versus PCI 8.3%, P < 0.001) and repeat revascularization (CABG 11.9% versus PCI 23%, P < 0.001). There was no difference in stroke (CABG 3.7% versus PCI 2.3%, P = 0.06) at 4 years of follow-up.
The SYNTAX trial did not include protocol driven assessment of graft/vessel patency. However, graft occlusion was documented in 26 out of 827 (3.14%) CABG patients and stent thrombosis in 36 out of 885 (4.07%) PCI patients. In 10 out of 36 patients, stent thrombosis resulted in fatal MI (no fatal MI for CABG graft occlusion). Nine of the 36 stent thromboses occurred after 1 year – five were on dual anti-platelet therapy and four on aspirin therapy alone. Two of those on aspirin had ceased their clopidogrel in the 2–4 weeks preceding stent thrombosis.
The biggest criticism of surgery in the SYNTAX trial was the increased rate of stroke – 2.2% versus 0.6% in the first year (P = 0.003). The details of timing of these strokes have not been published, but have been presented by the authors (M. Mack, unpublished, 2009). Three of the 19 strokes (0.3%) in the CABG arm occurred preoperatively (no preoperative strokes in PCI) but are included because results were analysed by intention to treat. Nine strokes (1%) occurred in the period between the procedure and 30 days inclusive (2 strokes (0.2%) in the PCI arm), and 7 strokes (0.8%) between 30 days and 1 year (3 strokes (0.3%) in the PCI arm). The higher rate of dual anti-platelet therapy after PCI compared with CABG may have influenced the rate of stroke occurring in the early post-operative period. The rate of stroke for OPCAB was 0.7% (1 stroke in 134 patients). The difference did not reach significance by the third year of follow-up (CABG 3.4% versus PCI 2.0%, P = 0.07).
In the New York Registry, PCI showed a significant survival advantage for two- and three-vessel disease after CABG at 8 years in 7235 matched patients (CABG 78.0% versus PCI 71.2%, HR: 0.68; 95% CI: 0.64–0.74; P < 0.001). Risk of death was lower for subgroups of patients with three-vessel disease involving LAD (HR: 0.53, P < 0.001) and two-vessel disease not involving the LAD (HR: 0.78, P = 0.05). The same group showed a survival advantage in CABG compared with DES at 18 months in two-vessel disease (CABG 96% versus PCI 94.6%, P = 0.003) and three-vessel disease (CABG 94% versus PCI 92.7%, P = 0.03). Substantial regional differences in the PCI to CABG ratio of cardiac hospitals analysed in the New York Registry suggest that patient characteristics, physician biases and hospital culture significantly influence the modality of revascularization.
The American College of Cardiology Foundation and the Society of Thoracic Surgeons Registries combined 86 244 patients undergoing CABG and 103 549 undergoing PCI between 2004 and 2007 for two- or three-vessel disease. Patients were 65 years or older, excluded if they were revascularized emergently or had MI within 7 days and followed-up for a median of 2.7 years. There was no difference in mortality at 1 year (CABG 6.2%, PCI 6.6%). At 4 years, mortality was lower in the CABG group (CABG 16.4% versus PCI 20.8%; relative risk (RR): 0.79, 95% CI: 0.76–0.82).
Our group has recently published a meta-analysis of trials comparing CABG with DES in patients with multi-vessel disease. MACCE at 1 year was significantly higher after PCI than CABG (RR: 1.74, 95% CI: 1.24–2.44, P = 0.001). Repeat revascularization was more frequent after DES (RR: 4.03, 95% CI: 2.7–6.01, P < 0.0001).
CABG remains the treatment of choice for multi-vessel coronary artery disease, with a survival benefit and lower rate of MI evident at approximately 3 years and lower rate of repeat revascularization evident as early as 1 year. The results suggesting superiority of CABG in the large registry trials have been confirmed by the randomized SYNTAX trial, the strength of which lay in the ‘real-world’ population of patients with three-vessel disease it recruited in contrast to the low-risk populations of previous trials. A sub-population of patients with three-vessel disease of low complexity (<22) is treated by PCI with equivalent results. The elevated rate of stroke in the surgical arm of SYNTAX at 12 months is a concern, even though there was no difference evident at 3 years – surgeons need to focus on techniques to reduce neurological injury.
Unprotected left main disease
CABG is recommended as a class I indication for left main coronary artery disease and has been the gold standard treatment for 3 decades.[6, 18] The left main coronary artery is an attractive target for interventional cardiologists because of its large diameter and proximal position in the coronary circulation and is now a class IIa alternative to CABG for low complexity disease (SYNTAX < 22).
Four randomized trials are included in a meta-analysis comparing PCI versus CABG for unprotected left main disease. The range of mean SYNTAX scores in the included trials was 24–30 and proportion with three-vessel disease was 14–68%. At 1 year, there was a difference in MACCE between groups (CABG 11.8% versus PCI 14.5%, P = 0.11). In patients with left main and three-vessel disease, PCI had a significantly higher rate of MACCE than CABG (CABG 10.4% versus PCI 17.9%, P = 0.03). The rate of stroke was lower in PCI (CABG 1.7% versus PCI 0.1%, P = 0.013), but rate of need for repeat revascularization is higher (CABG 5.4% versus PCI 11.4%). There was no difference in the rates of death and MI.
The SYNTAX trial stratified patients with left main disease into terciles based on coronary disease complexity. There was no difference between PCI and CABG in the rate of MACCE for low and intermediate SYNTAX score. However, for high SYNTAX scores (>32), comprising 65% of patients with left main disease, the rate of MACCE was lower in surgery (CABG 21.2% versus PCI 37.3%, P = 0.003). The composite of death/MI/stroke was similar between PCI and CABG.
CABG after previous PCI
A number of patients undergoing PCI as their first revascularization procedure may in the future be treated with CABG due to progression of disease or in-stent stenosis. There is debate as to whether prior treatment with PCI is a risk factor for morbidity or mortality after CABG. A large study (13 184 patients) from the ANZSCTS database reported no significant difference in early or mid-term (6 years) mortality (risk adjusted odds ratio (OR): 1.26, 95% CI: 0.77–2.08; P = 0.35) or MACE (adjusted OR: 1.19, 95% CI: 0.83–1.68, P = 0.34) in those undergoing CABG with prior PCI. Surprisingly, patients without prior PCI had a greater number of co-morbidities and risk factors than those with prior PCI treatment.
Other propensity score-matched studies have suggested higher mortality (OR: 2.0, 95% CI: 1–4–3.0, P = 0.0005) and morbidity in patients undergoing CABG after multiple previous PCI procedure. In a smaller propensity score study of 621 diabetic patients with three-vessel disease, prior PCI was associated with increased in-hospital mortality (OR: 2.97, 95% CI: 1.12–7.86, P = 0.03) and MACE (OR: 2.46, 95% CI: 1.18–5.15, P = 0.02).
Dual anti-platelet therapy is recommended for 6–12 months after insertion of a DES to prevent stent thrombosis. DES may occlude late because of delayed endothelialization and cases of stent thrombosis have been described after cessation of clopidogrel beyond the recommended 12 months.
Patients undergoing non-cardiac surgery are at particular risk of stent thrombosis due to its pro-thrombotic effect and the need to withdraw anti-platelet therapy to avoid bleeding. About 5% of patients undergoing coronary stenting need to undergo surgery in the next year. The rate of perioperative MACE (death/MI/stent thrombosis) is as high as 35% in surgery <1 month after stent deployment but less than 10% 12 months after stent deployment.
The peak age of coronary artery disease is similar for many cancers requiring surgery and for the need for joint replacement. Guidelines recommend non-urgent surgery be delayed 6 weeks after implantation of BMS and 12 months after DES. The balance of the risk of bleeding versus stent thrombosis is difficult if surgery is required within 12 months of DES deployment, and there may be particular difficulty in regional areas without immediate access to cardiac catheterization services.
The use of dual anti-platelet therapy after CABG is controversial. A recently published observational study suggested a lower rate of death after CABG (but, interestingly, not MI) when patients are treated with clopidogrel for 6–12 months. Conversely, the randomized CASCADE trial and a meta-analysis of studies has shown no such benefit.[45, 46] Unlike after DES, however, there is no evidence that cessation of clopidogrel given after CABG results in adverse outcome should this group of patients require non-cardiac surgery.
Improvements in CABG
CABG is a remarkably safe procedure – despite patients presenting for surgery about 10 years older and with more co-morbidities over the last decade, the mortality has remained between 1% and 2%.
Stroke has been cited as the ‘Achilles heal’ of CABG. While there was no significant difference in stroke at 3 years in SYNTAX, increased rate of stroke in the first year was of concern.[21, 32] A number of techniques may improve outcomes in CABG – they are rarely employed in trials comparing PCI and CABG. There is a significantly lower rate of stroke when manipulation of the ascending aorta is avoided (anaortic OPCAB 0.35% versus CABG 1.87%; RR: 0.24, 95% CI: 0.13–0.47, P < 0.001). While there remains some debate as to the benefits of OPCAB surgery versus conventional CABG, an accumulating body of evidence suggests that high-risk patients may be disproportionately benefitted by OPCAB surgery. There may be a mortality after OPCAB than CABG for higher-risk patients – this mortality benefit increases with increasing predicted risk of mortality. This, however, was not reflected in the recently published 30-day results of the randomized CORONARY trial. CORONARY showed similar results in its comparison of CABG and OPCAB in patients of moderate risk profile, highlighting the safety of modern cardiac surgery. There was no difference in neurological outcome between groups, but the anaortic technique was not applied universally in the OPCAB group. The ROOBY randomized trial showed poorer 1-year outcome in OPCAB versus CABG. However, the trial has been criticized – the low-risk population and inexperience of the surgeons in the technically demanding OPCAB technique (resident primary surgeon in 55% of cases) limits the usefulness of the study.[49-51] When compared with PCI in a recently published meta-analysis, OPCAB had lower rates of MACCE and lower need for repeat revascularization, but a similar rate of stroke at 12 months.
Observational registry data suggest better survival and freedom from re-operation with total arterial grafting compared with one IMA and multiple SVGs.[14, 23, 24, 53] There are unfortunately no randomized studies comparing total arterial with IMA/SVG grafting beyond a short period of follow-up (>1 year). In an Australian propensity-matched study, Tatoulis reported survival of 92% versus 74% at 8 years in patients who had undergone total arterial grafting versus those grafted with left IMA and saphenous vein.