ABC | Volume 112, Nº5, May 2019

Original Article Andrade et al Stent versus CABG: a meta-analysis Arq Bras Cardiol. 2019; 112(5):511-523 ERACI II included more them 90% of patients with unstable angina. AWESOME included only patients with high surgical risk; MASS II included predominantly stable angina and had a clinical arm; LE MANS used drug-eluting and BMS, reserving the DESs for left main coronary arteries with a reference diameter < 3.8 mm; CARDia used initially BMS and only assessed patients with diabetes and multivessel disease; SYNTAX evaluated left main coronary artery obstruction and multivessel disease and used first-generation DESs (TAXUS); FREEDOM and VA CARDS exclusively assessed patients with diabetes and multivessel disease; BEST evaluated patients with multivessel disease and used only everolimus-eluting stents; the study by Boudriot et al. evaluated left main coronary artery obstruction and used only sirolimus-eluting stents; EXCEL evaluated left main coronary artery obstruction and used only everolimus-eluting stents; NOBLE evaluated left main coronary artery obstruction and used mostly a biolimus-eluting stent. Outcomes The results are summarized in Figures 1 to 6. Regarding 30‑day mortality, the results favoured PCI (1% versus CABG 1.7%, p = 0.01), but the trials showed moderate overall heterogeneity (I 2  = 49%). The heterogeneity was particularly higher in BMS era trials (I 2  = 83%) and could be attributed to the significant inferior results of surgery in ERACI II and AWESOME. The incidence of stroke was lower with PCI (0.6% versus CABG 1.7%, p < 0.0001), with trials showing low heterogeneity (I 2  = 0). There was no difference in mortality up to one year (PCI 3.3% versus CABG 3.7%, p = 0.25) or up to two year (PCI 6.3% versus CABG 6.0%, p = 0.5). Long-term mortality showed a trend to superiority of CABG (10.6% versus 9.4%, p = 0.04), with trials showing moderate heterogeneity (I 2  = 25%). The differences were significant in trials of DES era (10.1% versus 8.5%, p = 0.01). After excluding FREEDOM (that included only diabetics) the overall difference in long‑term mortality between PCI and CABG became not significant (10.2% versus 9.4%, p = 0.17). The incidence of myocardial infarct was lower with CABG (PCI 6.4 % versus CABG 5.3% at one year and PCI 8,8% versus CABG 6.7 % after 3 or more years), but the trials showed high heterogeneity. In 6 studies of LMCAD (n = 4700), there was no difference in 30 days mortality (0.6% versus 1.1%, p = 0.15) between PCI and CABG, but the incidence of stroke was significantly lower after PCI (0.3% versus 1.1%, p = 0.007). There was no difference in one-year mortality (3% versus 3.7%, p = 0.18) or long-termmortality (8.1% versus 8.1%) between PCI and CABG. Nine trials (n = 4394) reported long-term mortality in diabetics (AWESOME, ARTS, ERACII, MASS II, SOS, SYNTAX, CARDia, FREEDOM and BEST). After pooling of results, CABG was associated with significantly lower long-term mortality (13.7% versus 10.3% CABG, p < 0.0001); After excluding the diabetic patients of these nine trials the overall difference in long-term mortality between PCI and CABG was no longer significant (9.2% versus 9.2%). The data regarding new revascularization are shown in figure 5. The superiority of surgery over PCI was consistent in all 15 trials. However, if we consider the risk of new revascularization by alternative procedure there was a trend to superiority of PCI in ARTS and in all studies of DES era. Subgroups results Five trials reported long-term results of major Adverse Composite Events (death, myocardial infarct and stroke) in subgroups. In three of them (SYNTAX, PRECOMBAT and BEST) the results were obtained through the collaborativemeta‑analysis of Lee et al. 24 (Figure 4). The pooled data showed that CABG, compared to PCI, was associated with a lower incidence of MACCE (18.4% vs 14.4%, p < 0.0001). The subgroups in which PCI had worse results, when compared with CABG, by meta-regression analysis were presence of diabetes (23% versus 17.5, p < 0.0001) and a high SYNTAX score (22.7 vs. 16.3%, p = 0.001). There was no difference between PCI and CABG in non-diabetics (14.1% versus 12.3%, p = 0.11), low SYNTAX score patients (14.1% vs. 13.3% scores, p = 0.4) and LMCAD patients (14.7% vs 14.1%, p = 0.5). Female sex and old age less significantly influenced the results. Left ventricular dysfunction did not influence the results. Figure 5 shows that the meta-Adjusted value of p for diabetes was 0.03 (adjusted for age or sex) and 0.09 (adjusted for SYNTAX score). The same figure shows that the meta-adjusted value of p for SYNTAX score was 0.03 (adjusted for diabetes). Discussion To our knowledge this meta-analysis is the most comprehensive and up to date overview of randomized trials that compared coronary stents (DES and BMS) versus CABG. It is also the only major meta-analysis of the stent era that evaluated mortality at different times (up to 30 days, up to one year and after three or more years of follow-up). Another peculiarity of the present meta-analysis was the statistical meta-regression analysis of sub-groups. The superiority of PCI on 30 days mortality is in accord with the New York state Registry 33 and with the meta-analysis of Palmerini et al. 32 This superiority should be seen with caution considering the heterogeneity of the trials and cannot be extended to patients with high SYNTAX score, considering the mortality curve of the study of Cavalcante et al. 30 The significant difference, favoring PCI found in the incidence of stroke is a relevant finding. A recent study showed that, after death (relative weight 0.23), stroke is the most feared event for patients (relative weight 0.18), being considered more important them longevity (relative weight 0.17), myocardial infarct (relative weight 0.14) and risk of repeat revascularization (relative weight 0.11). 34 The lack of difference in intermediate mortality was an expected finding, having been reported in almost all trials. The trend superiority of surgery in long-term mortality was shown in other meta-analysis 26,29,31 and is probably related to the higher percentage of diabetics in recent trials. Our results of long-term mortality (HR 1.13) were similar to the results of Smit el al. 26 (HR 1.11) and Lee et at. 29 (HR 1.18). They were much less unfavourable to PCI them the reported by Benedetto et al (HR 1.5). 31 The reason for this is that 513