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 Inclusion criteria Clinical trials were included in the review if they were randomized, if had compared PCI with stents versus CABG, if included exclusively multi-vessel and/or LMCAD and if had a follow up of at least 1 year. We did not limit our search to DES trials because bare-metal stents (BMS) are still frequently used in many developing countries, had the peculiarity of evaluating patients with less complex coronary artery disease and there is no definitive evidence that BMS are inferior to DES in the outcome mortality. Figure 7 show a flow diagram of the search strategy in the databases. We identified a total of 15 RCT that satisfied the requirements: AWESOM 1 , ERACI II, 2,3 MASS II, 4-6 SOS, 7,8 ARTS, 9,10 LE MANS, 11 SYNTAX, 12-14 CÁRDia, 15 Boldriot et al., 16 PRECOMBAT, 17,18 Va-Cards, 19 FREEDOM, 20 BEST, 21 NOBLE 22 and EXCEL. 23 Three reviewers (PJNA, ATA and JLAF) assessed the quality of the studies using the Cochrane Collaboration’s tool. Data extraction Two reviewers (JLAAF and PJNA) obtained the data from the studies, examining abstracts, results, tables, appendices and figures. A third author (BAAF) checked the results. The main outcomes evaluated were all case-mortality, stroke, AMI and new revascularization. Mortality was divided into early mortality, mortality at one year and late mortality. Early mortality was defined as percentage of deaths that occurred in the first 30 days after the procedure, including deaths after randomization but before the procedure. Late mortality was defined as percentage of deaths reported in the last publication, after at least three years of follow‑up. For the incidence of stroke, we considered the events occurring up to 1 year after the procedure. In twelve studies we obtained the results up to 30 days, in 2 studies 9,12 up to 1 year and in one 21 this observation was unavailable. For the incidence of myocardial infarct, we considered the reported up to one year of the procedure. AMI were reported in 13 trials. 2-5-15 We did not consider in the pooled data the results of NOBLE because it excluded perioperative myocardial infarct in the majority of the patients. New revascularization was divided in any form of new revascularization (PCI or CABG) and new revascularization by alternative procedure (PCI for patients of the CABG group or CABG for the patients of the PCI group). Data synthesis The characteristics of patients from the eligible studies were obtained through a weighted average of published data. For pooling results of mortality and stroke, the numerator was the number of events and the denominator the total of patients. The total of patients was the number of patients effectively followed, including the deaths. Trials were divided into DES era trials and BMS era trials. Trials that used both types of stents 11,15 were classified as DES era trials. We evaluated separately the results of studies in the left main coronary artery and late mortality in the subgroup of patients with diabetes. We also performed analysis of combined major adverse cardiac and cerebrovascular events (MACCE) and assessed the variables age, gender, presence of diabetes, SYNTAX score, and compromised ejection fraction in subgroups based on data published in five trials. Combined MACCE comprised death, AMI, and new revascularization. In order to aggregate the outcomes of mortality and stroke, as well as those of MACCE (in subgroups), we considered whenever possible the absolute number of events and the number of patients followed up. Otherwise, percentages were transformed into absolute numbers. Statistical analysis We measured the relative risk and the risk difference after grouping the results of each outcome. In order to assess the statistical significance of the differences between the DES and the surgery groups, we performed a meta-analysis using the Mantel-Haenszel method, with a random-effect model. We calculated the heterogeneity of the studies using Cochran's Q test and the significance of the measure of the meta-analytic effect using the Z test. Finally, we performed a meta-regression analysis using diabetes, age, gender, ejection fraction, and syntax scores as factors. The differences between the results in the PCI and CABG groups were considered significant if p<0.05. The statistical analyses were performed using the program Review Manager (RevMan), version 5.3 (Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014) and SPSS for Windows v 23, IBM Inc. In order to represent the heterogeneity of the studies, we constructed Forest plots. We used the risk difference to plot these graphs since this is a more stable index. The possibility of publication bias was assessed by visual inspection of funnel plots. Results Studies Characteristics The studies (table 1) included a total of 12,781 patients (6,382 in the CABG group and 6,399 in the PCI group). All studies were, considered of quality A or B in terms of adequate randomization, adequate concealment and inexistence of selection bias, but not in terms of adequate making. In all studies, the PCI and CABG groups were similar, with the exception of VaCards where the PCI group had a higher incidence of the previous revascularization (in most of the cases a previous PCI) and showed a higher percentage of patients with ejection fraction < 55%. The mean age of the patients was 64 years; 74% were male; 42% were diabetics; 28% were smokers; 64% were hypertensive. Unstable angina was the clinical presentation in 34%; mean ejection fraction was 58%. With the exception of AWESOME, all studies tended to exclude patients with previous CABG. The mean SYNTAX score was 26. According to number of arteries affected 20% had two vessel disease, 43% had three-vessel disease and 37% had LMCAD (alone or associated with diseases of other arteries). In the CABG group, at least one arterial graft was used in 90% of the patients. In trials of the BMS era surgery was done almost always using on-pump technique; in trials of the era, the DES off-pump technique was used in 28% of the patients. Some characteristics of the studies deserve special mention: 512