Percutaneous Coronary Intervention Compared to Coronary Artery Bypass Grafting in Patients with Non-ST-Elevation Acute Coronary Syndrome – A Systematic Review and Meta-Analysis


 Objectives: Non-ST-elevation acute coronary syndrome (NSTE-ACS) affects millions of patients. Although an invasive strategy can improve survival, the optimal treatment (i.e., percutaneous coronary intervention [PCI] or coronary artery bypass grafting [CABG]) is not clear. We performed a meta-analysis of studies reporting outcomes between PCI and CABG in patients with NSTE-ACS.Methods: MEDLINE, EMBASE and Cochrane Library were assessed. The primary outcome was long-term mortality. Inverse variance method and random model were performed.Results: We identified 13 observational studies (48,891 patients). No significant difference was found in the primary endpoint (CABG vs. PCI, incidence rate ratio [IRR]= 0.93, 95% confidence interval [CI] 0.70; 1.23). CABG was associated with lower long-term major adverse cardiovascular events (MACE) (IRR= 0.64, 95% CI 0.54; 0.76) and lower long-term re-revascularization (IRR= 0.37, 95% CI 0.30; 0.47). There was no significant difference in long-term myocardial infarction (CABG vs. PCI, IRR= 0.96, 95% CI 0.50; 1.84) and peri-operative mortality (CABG vs. PCI, odds ratio [OR]= 1.36, 95% CI 0.94; 1.95).Conclusion: For the treatment of NSTE-ACS, CABG and PCI are associated with similar rates of long-term mortality and myocardial infarction. CABG is associated with lower rates of long-term MACE and re-revascularization. Randomized comparisons in this setting are necessary.


Introduction
Non-ST-elevation acute coronary syndrome (NSTE-ACS) represents a large proportion of patients with acute coronary syndromes (ACS) impacting millions of patients worldwide [1,2].
Recently, the percentage of patients presenting with non-ST-elevation myocardial infarction (NSTEMI) has increased in both Europe and United States [1,2]. Although a routine invasive strategy in NSTE-ACS may improve long-term survival and reduce late myocardial infarction [3], the optimal treatment method remains controversial. Speci cally, it remains unclear as to whether percutaneous coronary intervention (PCI) or coronary artery bypass grafting (CABG) provide better long-term outcomes [4].
To date, there is no randomized comparison of PCI vs. CABG surgery in the speci c setting of NSTE-ACS. While there have been a number of nonrandomized studies published, their results have been inconclusive. Additionally, no systematic review and meta-analysis addressing this topic exists in the literature.
In this analysis we set out to systematically review the literature on the impact of the treatment modality on clinical outcome in NSTE-ACS.

Methods
This analysis was prospectively registered on the International Prospective Register of Systematic Reviews in Health and Social Care (PROSPERO, ID number CRD42020214423). Ethical and internal review board approval was not required for this analysis as no human or animal subjects were involved.

Study selection and eligibility criteria
The study selection followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) strategy [5]. After de-duplication, records were screened by two independent reviewers (TC and HK). Any discrepancies and disagreements were resolved by a third author (TD). All titles and abstracts were reviewed against pre-de ned inclusion and exclusion criteria. Studies were considered for inclusion if they were written in English, reported direct comparison between PCI patients and CABG patients and had at least 1 outcome of interest reported. Studies evaluating refractory angina, chronic coronary disease, conference abstracts, proceedings and case reports were excluded.
Following the rst round of screening, full text was pulled for selected studies for a second round of eligibility screening. Reference lists for articles in these selected studies were also searched for any relevant articles not captured by the original search strategy.

Data abstraction and quality assessment
The data extraction and the quality assessment were performed independently by two different investigators (TC and HK) and veri ed by a third investigator (TD) for accuracy. The following variables were extracted: age, sex, left ventricular ejection fraction, hypertension, diabetes, smoking status, prior cerebrovascular accident, prior myocardial infarction, prior percutaneous coronary intervention and number of vessels addressed (1, 2 or 3). Table 2

Outcomes and effect summary
The primary outcome was long-term mortality (> 6 months). Mean follow up by study is summarized in Table 1.
Secondary outcomes were long-term major adverse cardiovascular events (de ned in Supplementary Table 3), long-term re-revascularization, longterm myocardial infarction and peri-operative mortality (in-hospital and 30-day events). For the primary outcome, a sensitivity analysis comparing adjusted and low risk of bias studies vs. unadjusted studies was also performed.

Data analysis
Peri-operative binary outcomes were reported as odds ratios (OR) while long-term outcomes, were reported as incidence rate ratio (IRR); for both estimates the generic inverse variance method was used and 95% con dence intervals (CIs) were also presented.
Random effect meta-analysis was performed using "metafor" and "meta" package [7,8]. CABG was the reference for all pairwise comparisons.
Heterogeneity was reported as low (I 2 = 0-25%), moderate (I 2 = 26-50%), or high (I 2 > 50%) [9]. Leave-one-out analysis for the primary outcome was performed to assess the robustness of the obtained estimate. Meta-regression was used to explore the effects of: year of publication, mean followup time, age, sex, left ventricular ejection fraction, hypertension, diabetes, smoking status, prior cerebrovascular accident, prior myocardial infarction, prior percutaneous coronary intervention and number of vessels addressed (1, 2 or 3) on the IRR of the primary outcome.
Statistical signi cance was set at the 2-tailed 0.05 level, without multiplicity adjustments. All statistical analyses were performed using R (version 3.3.3, R Project for Statistical Computing) within RStudio.

Description of included studies
A total of 7,520 records were identi ed through database searching. After duplicate records were removed, 5,927 citations were retrieved and their titles and abstracts were screened; 13 studies were included in the nal analysis, with a total of 48,891 patients. The full PRISMA ow diagram outlining the study selection process is available in Fig. 1 [5,10,11]. A complete list of studies included in the nal analysis is presented in Table 1.
All studies were observational: 5 of them were unadjusted and 8 adjusted (adjustment details are shown in Table 1). Ten studies were multicenter; 3 originated from the United States, 2 from Israel, 1 each from Netherlands, Australia, Germany, Saudi Arabia, Poland, South Korea, Portugal and China.
The number of patients in each study ranged from 450 to 15,281. The mean age ranged from 56.1 to 75 years. The percentage of female sex in each study ranged from 18.0 to 43.5%. The prevalence of hypertension ranged from 43.0 to 87.3%, diabetes from 22.0 to 100.0%, smoking from 5.0 to 74.0%, prior cerebrovascular accident from 2.1 to 11.9% and prior myocardial infarction from 21.3 to 100.0%. The percentage of 1-vessel disease ranged from 0.0 to 47.2 %, 2-vessels from 9.3 % to 67.0 % and 3-vesels from 16.3 to 100.0% (Supplementary Table 4).

Primary outcome
Detailed results of the meta-analysis are outlined in Graphical Abstract and summarized in Table 2.
No signi cant difference between the groups was found in long-term mortality (IRR = 0.93, 95% CI 0.70; 1.23, p = 0.83, Fig. 2). This nding was con rmed in the sensitivity analysis comparing adjusted and low risk of bias studies vs. unadjusted studies (p-interaction = 0.58, Supplementary   Fig. 1).
Leave-one-out analysis con rmed the solidity of the pooled estimate ( Supplementary Fig. 2). Funnel plot did not demonstrate evidence of publication bias (Supplementary Fig. 3).
At meta-regression, the year of publication and the proportion of patients with prior PCI were inversely associated with the IRR for the primary outcome (beta=-0.07, p = 0.01 and beta=-0.05, p = 0.004, respectively). The proportion of 1-vessel disease was associated with the IRR for the primary outcome (beta = 0.01, p = 0.04, Supplementary Table 5). That means that recent studies and studies with higher percentage of patients with prior PCI reported lower long-term mortality in the CABG group; and studies with higher percentage of patients with 1-vessel disease reported lower long-term mortality in the PCI group.

Discussion
Our analysis suggests that for the treatment of NSTE-ACS, CABG and PCI are similar with respect to long-term mortality and myocardial infarction rates. The analysis further suggests that CABG is associated with decreased rates of long-term MACE and re-revascularization. This is the rst meta-analysis to address this important topic; our results are relevant as a substantial number of patients worldwide present with NSTE-ACS every year; in Germany over a period of 10 years 2.77 million cases of NSTE-ACS were recorded [1]. In the United States there are more than 1 million hospitalizations per year due to acute coronary syndrome [12], with the proportion of patients with non-ST-elevation myocardial infarction being over 50% of all infarctions and increasing over time [2]. In other studies the annual incidence of NSTE-ACS has been reported with 88 per 100 000 inhabitants [13].
Previous studies have suggested that an invasive strategy might be superior to a conservative one [3,14], but have not summarized a recommendation for a speci c invasive therapy (PCI or CABG). Currently, only 4-10% of the patients with NSTE-ACS receive CABG and 30-40% of them PCI [1,15].
Our ndings support the current guidelines, where no clear recommendation for PCI or CABG is given and a suggestion that the criteria applied for patients with stable coronary artery disease should be applied to stabilized patients with NSTE-ACS is made [4]. Both American and European guidelines recommend a heart-team approach to revascularization decisions in NSTE-ACS [4,16] and that, factors such as extent and complexity of the coronary artery disease, as well as other factors should be considered.
It has been proposed that the survival bene t of PCI or CABG may be primarily due to an infarct-preventing mechanism rather than to revascularization per se [17,18]. This theory might provide a mechanistical explanation of our results, showing no mortality difference in 2 invasive treatments treating the same acute events. This line of argumentation nds supports in a meta-analysis we performed demonstrating an association of a survival advantage of CABG over PCI only in cases when there was also a difference in the rate of myocardial infarction [19].

STUDY STRENGTH AND LIMITATIONS
This is the rst meta-analysis to address this important topic. Moreover, we analyzed 5 different outcomes and performed different subgroup analyses and a meta-regression of 14 different pre-operative factors. However, this work has the intrinsic limitations of observational series, including the risk of methodological heterogeneity of the included studies, residual confounders and ecological fallacy of meta-regression. In addition, treatment allocation bias is likely present in all observational series comparing two interventions with different operative risk and invasiveness.

Conclusion
Our analysis suggests that for the treatment of NSTE-ACS, CABG and PCI are similar with respect to long-term mortality and myocardial infarction rates. The analysis further suggests that CABG is associated with decreased rates of long-term MACE and re-revascularization. Randomized comparisons in this speci c setting are necessary.  Forest plot for long-term mortality. Forest plot showing pooled rates of long-term mortality in patients with non-ST-elevation acute coronary syndrome (NSTE-ACS) treated with coronary artery bypass grafting (CABG) or percutaneous coronary intervention (PCI). Abbreviations: CABG= coronary artery bypass grafting, CI= con dence interval, IRR= incidence rate ratio, NSTEMI= non-ST-elevation myocardial infarction, PCI= percutaneous coronary intervention, UAP= unstable angina pec-toris, WHO= world health organization. Figure 3