It is controversial whether drug-eluting stents (DESs) are safe and effective when generalized to “real-world” patients with unprotected left main coronary artery disease. This meta-analysis compared the safety and efficacy of DESs to coronary artery bypass grafting (CABG) in real-world patients with unprotected left main coronary artery. We identified comparative, observational, DES versus CABG studies published from January 2000 through May 2012. All studies included ≥100 patients and reported end points with follow-ups ≥6 months. We included adjusted risk estimates and, when no adjusted estimate was available, crude estimates. Data were grouped according to follow-up times of ≤2, ≤3, and >3 years. We included data from 25 observational studies representing 7,230 patients. No differences were detected between CABG and DES in overall mortality (≤2 years, adjusted risk ratio [RR], 0.83, 95% confidence interval [CI] 0.53 to 1.28; ≤3 years, adjusted RR 0.60, 95% CI 0.20 to 1.66; >3 years, adjusted RR 0.58, 95% CI 0.29 to 1.17) or in major adverse cardiac and cerebrovascular events (≤2 years, adjusted RR 1.22, 95% CI 0.86 to 1.73; ≤3 years, adjusted RR 1.70, 95% CI 1.35 to 2.15; >3 years, adjusted RR 1.23, 95% CI 0.87 to 1.73). Compared to DESs, CABG showed a significant decrease in target vessel revascularization (≤2 years, adjusted RR 3.72, 95% CI 2.50 to 5.52; ≤3 years, adjusted RR 3.92, 95% CI 2.54 to 6.04; >3 years, adjusted RR 3.45, 95% CI 2.14 to 5.57). In conclusion, DESs and CABG were not significantly different in short- and long-term rates of death or major cardiovascular/cerebrovascular events, but DESs showed a higher risk of target vessel revascularization compared to CABG.
Coronary artery bypass grafting (CABG) is the standard of care for managing unprotected left main coronary artery disease. However, recent guidelines from the European Society of Cardiology and European Association for Cardiothoracic Surgery have advocated the broader use of percutaneous coronary intervention for the management of unprotected left main coronary artery disease. To date, 2 meta-analyses of 4 randomized controlled trials (RCTs) have compared CABG to drug-eluting stents (DESs) and percutaneous coronary intervention for the management of unprotected left main coronary artery disease. Most of these RCTs did not investigate the long-term safety and effectiveness of DESs and CABG in patients with unprotected left main coronary artery disease. However, population-based observational studies tend to resemble clinical practice; thus, those studies may include patients with high risk of cardiovascular disease or co-morbidities. The objective of the present study was to systematically compare the effects of DESs and CABG on cardiovascular outcomes including major adverse cardiac and cerebrovascular events (MACCEs), target vessel revascularization (TVR), and all-cause mortality based on observational studies in patients with unprotected left main coronary artery disease.
Methods
We searched electronic databases for articles published from January 2000 through March 2012. The databases included MEDLINE, EMBASE, Cochrane Library, Google Scholar, epidemiologic research Web sites, abstracts of scientific meetings, and bibliographies of relevant studies without language restrictions. The key search words included “coronary artery bypass graft surgery,” “drug-eluting stent,” “sirolimus-eluting stent,” “paclitaxel-eluting stent,” and “unprotected left main disease.” Citations were screened at the title/abstract level and full texts were retrieved. Studies included in this meta-analysis were based on predetermined inclusion criteria: (1) the study should compare percutaneous coronary intervention with DES implantation to CABG; (2) the indication should be unprotected left main coronary artery disease; (3) primary end points should be mortality, MACCEs (defined as occurrence of death, myocardial infarction, stroke, or TVR), or TVR; and (4) the study design should be observational. Exclusion criteria were (1) duplicate publication, (2) ongoing/unpublished study, (3) studies published only as an abstract or in conference proceedings, (4) study with ≤6-mont follow-up, and (5) RCTs.
All data were extracted independently by 2 evaluators; discrepancies were resolved by consensus. The following outcomes were extracted: (1) MACCEs, defined as the composite of death, myocardial infarction, stroke (hemorrhagic or ischemic), or TVR; (2) all-cause death; and (3) TVR. Clinical end point definitions were similar among the studies examined.
We used the appropriate Newcastle-Ottawa Scales to assess the quality of observational studies based on the presence of potential confounders. We used quality at the analysis stage as a means of interpreting the results. We assigned risk-of-bias categories based on the number of Newcastle-Ottawa Scale items judged inadequate in each study ( Table 1 ). In the meta-analysis, we adhered to the recently updated guidelines for Preferred Reporting of Items for Systematic Reviews and Meta-Analyses.
| Study | Country of Patient Recruitment (study period, month/year) | Patient Number | Follow-Up (years) | Age (years) | Men (%) | Main Outcomes | Quality ⁎ | |||
|---|---|---|---|---|---|---|---|---|---|---|
| DES | CABG | DES | CABG | DES | CABG | |||||
| Palmerini et al, 2006 | Italy (1/2002–6/2005) | 157 | 154 | 1.2 | 73.0 | 69.3 | 70.1 | 76 | death/MI/TVR | low bias |
| Chieffo et al, 2006 | Italy (3/2002–7/2004) | 107 | 142 | 1 | 63.6 | 67.5 | — | — | death/TVR/MACCE | medium bias |
| Lee et al, 2006 | USA (since 2003) | 50 | 123 | 0.6 | 72 | 70 | 50 | 76 | death/TVR/MACCE | medium bias |
| Palmerini et al, 2007 | Italy (1/2003–4/2006) | 98 | 161 | 2 | 81 † | 78 † | 54 | 66 | death/TVR/MI | low bias |
| Sanmartín et al, 2007 | Spain (1/2000–10/2005) | 96 | 245 | 3.2 | 66.0 | 66.0 | 81 | 87 | death/TVR/MACCE | low bias |
| Mäkikallio et al, 2008 | Finland (1/2005–1/2007) | 49 | 238 | 1 | 72 | 70 | 59 | 80 | death/TVR/MACCE | medium bias |
| White et al, 2008 | USA (4/2003–1/2007) | 120 | 223 | 1.6 | 70.9 | 69.4 | 58 | 77 | death/MACCE | medium bias |
| Rodés-Cabau et al, 2008 | Canada (1/2002–1/2008) | 104 | 145 | 1.9 | 85 | 82 | 56 | 92 | death/TVR/MACCE | medium bias |
| Wu et al, 2008 | USA (10/2003–12/2004) | 56 | 56 | 0.7 | 68.6 | 68.9 | 69.6 | 69.6 | death/TVR | medium bias |
| Brener et al, 2008 | USA (1997–2006) | 97 | 190 | 3 | 68 | 68 | 72 | 74 | death | low bias |
| Ghenim et al, 2009 | France (1/2004–12/2007) | 105 | 106 | 1 | 80.7 | 79.6 | 62 | 70 | MACCE/TVR | low bias |
| Liu et al, 2009 | China (1/2004–6/2006) | 89 | 206 | 2 | 81.6 | 76.5 | 71 | 44 | death/TVR | medium bias |
| Buszman et al, 2009 | Poland (1/2005–4/2007) | 63 | 75 | 1 | 64 | 65 | 44 | 61 | death/TVR/MACCE | high bias |
| Cheng et al, 2009 (DES) | China (6/2000–3/2007) | 94 | 216 | 2.2 | 67.6 | 66.6 | 74.5 | 75.9 | death/TVR/MACCE | medium bias |
| Chieffo et al, 2010 | Italy (3/2002–7/2004) | 107 | 142 | 5 | 63.6 | 67.5 | — | — | death/TVR/MACCE | low bias |
| Kang et al, 2010 | Korea (1/2003–12/2006) | 205 | 257 | 2.8 | 64.2 | 65.7 | 70.2 | 73.9 | death/TVR/MACCE | low bias |
| Park et al, 2010 (DES) | Korea (1/1995–4/1999) | 176 | 219 | 5 | 61.1 | 62.4 | 71.0 | 74.0 | death/TVR/MACCE | low bias |
| Wu et al, 2010 | China (2/2003–12/2006) | 131 | 245 | 4 | 61.9 | 63.6 | 76 | 83 | death/TVR/MACCE | low bias |
| Huang et al, 2010 | China (1/1996–6/2006) | 148 | 336 | 5.4 | 66 | 66 | 88 | 79 | death | medium bias |
| Shimizu et al, 2010 | Japan (4/2004–12/2007) | 64 | 89 | 2.2 | 71 | 70 | 81 | 85 | death/TVR/MACCE | low bias |
| Rittger et al, 2011 | Germany (4/2004–12/2007) | 95 | 205 | 1.0 | 71.6 | 68.7 | 71 | 76 | death/TVR/MACCE | medium bias |
| Capodanno et al, 2011 | Italy (since 2002) | 173 | 173 | 3 | 64.9 | 65.3 | 79.8 | 76.3 | death/TVR/MACCE | low bias |
| Caggegi et al, 2011 | Italy (since 2002) | 222 | 361 | 1 | 67.2 | 65.8 | 76.1 | 78.7 | death/TVR/MACCE | medium bias |
| Zhao et al, 2011 | China (1/2004–12/2006) | 56 | 116 | 2.35 | 51.4–61.5 | 54.8–72.0 | 73.2 | 71.6 | death/TVR/MACCE | low bias |
| Kawecki et al, 2012 | Poland (2005–2008) | 34 | 111 | 0.1 | 63.6 | 65.6 | 68 | 73 | death/MACCE | medium bias |
Individual cohort studies reported results as risk ratio (RR), hazard ratio, odds ratio, or dichotomous frequency data. For this analysis, we recalculated the confidence interval (CI) by converting the frequency data into RRs because log values for the lower and upper limits were not symmetrical. Where possible, we pooled primary data that included adjusted odds ratios, RRs, or hazard ratios based on the propensity score and/or a multivariate logistic regression analysis and/or a multivariate Cox proportional hazards regression analysis; otherwise, we used raw outcome data to yield unadjusted odds ratios. We combined RRs with a random-effects model based on the inverse variance method. Given the heterogenous nature of observational analyses, we divided the studies into subgroups of unadjusted estimates and adjusted estimates for separate analyses. Heterogeneity was assessed with the I 2 statistic, where I 2 values of 30% to 60% represented a moderate level of heterogeneity. We assessed publication bias with the Begg adjusted rank correlation test and the Egger regression asymmetry test.
For sensitivity analyses, estimates derived from studies based on propensity score matching were considered in a separate substratum. Any individual study could be used in several categories (e.g., when unadjusted, adjusted, and propensity-matched data were reported); however, the highest-quality estimate was selected for overall meta-analysis based on the following ranking: adjusted (propensity matched, Cox/logistic multivariate analysis) greater than unadjusted. We repeated the primary analysis using a randomized-effects model. The purpose of this analysis was to assess the influence of statistical heterogeneity on overall findings by comparing the result to the more conservative random-effects model. We also examined the degree to which excluding single studies 1 by 1 influenced summary results. We adapted to ensure analyses were based on robust data, included only studies that scored ≥6 stars on the scale’s assessment of selection of patients, comparability, and outcome. All meta-analyses were conducted with STATA 10.0 (STATA Corp. LP, College Station, Texas).
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
