Effect of increasing age on percutaneous coronary intervention vs coronary artery bypass grafting in older adults with unprotected left main coronary artery disease: A meta‐analysis and meta‐regression

Abstract Background Older adults (≥70‐year‐old) are under‐represented in the published data pertaining to unprotected left main coronary artery disease (ULMCAD). Hypothesis: Percutaneous coronary intervention (PCI) might be comparable to coronary artery bypass grafting (CABG) for revascularization of ULMCAD. Methods We compared PCI versus CABG in older adults with ULMCAD with an aggregate data meta‐analyses (4880 patients) of clinical outcomes [all‐cause mortality, myocardial infarction (MI), repeat revascularization, stroke and major adverse cardiac and cerebrovascular events(MACCE)] at 30 days, 12‐24 months & ≥36 months in patients with mean age ≥70 years and ULMCAD. A meta‐regression analysis evaluated the effect of age on mortality after PCI. Odds ratios (OR) and 95% confidence intervals (CI) were estimated using random‐effects model. Results All‐cause mortality between PCI and CABG was comparable at 30‐days (OR0.77, 95% CI 0.42‐ 1.41) and 12‐24‐months (OR 1.22, 95% CI 0.78‐1.93). PCI was associated with a markedly lower rate of stroke at 30‐day follow‐up in octogenarians (OR 0.14, 95% CI 0.02‐0.76) but an overall higher rate of repeat revascularization. At ≥36‐months, MACCE (OR 1.26,95% CI 0.99‐1.60) and all‐cause mortality (OR 1.39, 95% CI 1.00‐1.93) showed a trend favoring CABG but did not reach statistical significance. On meta‐regression, PCI was associated with a higher mortality with advancing age (coefficient=0.1033, p=0.042). Conclusions PCI was associated with a markedly lower rate of early stroke in octogenarians as compared to CABG. All‐cause mortality was comparable between the two arms with a trend favoring CABG at ≥36‐months.PCI was however associated with increasing mortality with advancing age as compared to CABG.

CABG is associated with a higher peri-procedural mortality and stroke while PCI is associated with a higher rate of repeat revascularization. 5,8 Older adults defined as age ≥70 years are generally underrepresented in studies and are a cohort with greater frailty, more comorbidities and higher procedural complication rates. 9,10 Existing data is limited by smaller sample sizes which makes it challenging to draw robust conclusions regarding true efficacy and risks of each therapy.
Our study attempts to explore the optimal revascularization technique for ULMCA disease in the elderly by evaluating all existing data.

| METHODS
A systematic data search was performed using keywords "unprotected left main coronary artery, coronary artery bypass graft" and PCI from January 1, 2003 to April 1, 2019 on MEDLINE. The initial search resulted in 365 citations, a careful review of the abstracts of these citations identified 81 studies that reported comparative outcomes of PCI vs CABG in ULMCA lesions. Sixty six of these identified citations were excluded either because of duplicated data (population reported elsewhere) or if they did not meet our age cutoff (mean age ≥70 years in the PCI arm). The study by Gomez et al 11 compared PCI with CABG in 3 patient cohorts, out of which 2 were in included in our analysis as they met our age cutoff. Finally, abstracts presented in national cardiovascular conferences were searched to identify studies meeting our inclusion criteria ( Figure 1). In the final analysis, 16 studies Each of these studies were reviewed and data were extracted independently by two reviewers (Mahin R. Khan and Waleed T. Kayani). Data pertaining to baseline demographic and clinical variables including clinical presentation, risk stratification scores, coronary anatomy, and procedural variables were obtained. The initial revascularization strategy (CABG or PCI) was the primary independent variable. The primary outcome of our study was all-cause mortality, secondary outcomes included nonfatal myocardial infarction (MI), stroke, repeat revascularization, and major adverse cardiac and cerebrovascular events (MACCE; composite end-point of all-cause mortality, MI, stroke, or repeat revascularization). The outcomes are reported at a 30-day, 12-24 months and ≥36 months follow-up in a cumulative fashion. The studies' defined end points were used to conduct our meta-analyses. Additional comparative analyses performed included analyses to compute inter-group differences between the ≥70-year-old and ≥80-year-old patients. The funnel plot of all-cause mortality shows a symmetrical distribution of studies indicating that a publication bias is less likely (supplementary figure 1). The distributions of continuous and categorical variables were described using mean SD and percentages, respectively. Continuous variables were compared using the 2-tailed student t-test while the categorical variables were compared using chi-square test with Yates' correction, where applicable. Odds ratios (ORs) and their 95% confidence intervals (CIs) were used to summarize the effect of each outcome at the F I G U R E 1 Data search and review method-Flowsheet of identification of the studies that were included in the analyses corresponding follow-up using the random-effects model. Cochran's Q-statistic and I 2 index tests were computed to determine the heterogeneity. An I 2 of 25% was considered to indicate statistically significant heterogeneity. We reported the results using random-effects modeling only, given the inherent heterogeneity of the data. P-values of <.05 were considered statistically significant. Furthermore, a Newcastle-Ottawa scale was employed to assess the quality of all included observational studies. We defined the high-quality studies as those that scored nine stars (maximum) on the scale, studies with moderate quality were defined as those that scored 78 stars. Nine of the included studies scored the maximum of nine stars while all of the remaining observational studies scored at least seven stars on the Newcastle-Ottawa scale (supplementary Table 2).
We also performed meta-regression analyses (on 10 observational studies that reported mortality at 12-24 months) to evaluate whether the effect of PCI on mortality was modulated by age. The metaregression graphs are plotted as log odds ratio of mortality on the yaxis against mean age as a covariate on the x-axis. The metaregression coefficients show the increase in log OR per unit increase in age (covariate). Figure 2 describes the effect of PCI on mortality (plotted as log OR on the y-axis) as a function of age in the PCI arm

| RESULTS
The current meta-analyses included a total of 16 studies (15 observational and 1 randomized) and 4880 patients. Of the included studies, the primary outcome, that is, all-cause mortality was reported in 6 studies at 30 days and ≥36 months while 10 studies reported allcause mortality at 12-24 months. There were several differences in the baseline clinical characteristics of the patients undergoing PCI vs CABG ( Table 1). The patients in the PCI group had a greater prevalence of chronic renal insufficiency and congestive heart failure and were more likely to present with non-ST-elevation myocardial infarction than patients in the CABG group (54.8% vs 50.3%). Patients in the CABG group were more likely to be men and had a higher prevalence of hyperlipidemia and smoking. Patients in the PCI group had a higher prevalence of non-distal (ostial and mid-shaft lesions) left main coronary artery (LMCA) stenosis but a similar frequency of distal LMCA disease as compared to the CABG group. A greater percentage of patients with LMCA with single and double-vessel disease underwent PCI and a greater proportion of patients with LMCA with triplevessel disease underwent CABG. The two groups did not have any statistically significant difference in terms of age, hypertension, diabetes, mean SYNTAX scores, and prevalence of peripheral arterial disease. It is however noteworthy that SYNTAX scores were not reported for the PCI and CABG groups in 11 out of the 17 cohorts included in our analysis. Notably, 220 ( On the meta-regression analyses, at 12-24 months, there was a statistically significant increase in mortality in the PCI group with advancing age (71 ± 7-85 ± 3 years) in the PCI arm (coefficient = 0.1033, 95% CI 0.004-0.203, P = 0.0416) as shown in

| DISCUSSION
In our analysis of 4880 patients with ULMCA disease and a mean age ≥70 years, PCI was associated with lower rates of stroke (in octogenarians at 30 days), higher rates of repeat revascularization (≥12 and ≥36 months) and nonfatal MI as compared to CABG.
Although there was no difference in MACCE and mortality between the two groups at short (30 days) and intermediate-term follow up (12-24 months), long-term follow up (≥36 months) showed a trend favoring CABG that did not reach statistical significance. The initial advantage of PCI in the incidence of stroke became nonsignificant at long-term follow-up, however, the disparity in repeat revascularization persisted. This observation is in conformity with prior studies including randomized controlled trials 1,2,12-15 and registry data [16][17][18] in younger patients.
Results from the SYNTAX, 3 PRECOMBAT, 19 EXCEL 1 trials, and meta-analyses 5,8 comparing PCI and CABG for unprotected LMCA disease also showed no intermediate and long-term mortality difference between the two treatment groups. The NOBLE 2 trial also reported a comparable mortality rate between the PCI and CABG arms but there was a potential concern of the study being lowpowered to detect outcomes due to early termination of the trial secondary to a low event rate. Despite more comorbidities, all-cause . With meta-regression analyses, we observed that advancing age was associated with a higher mortality with PCI ( Figure 2).
Since all of the studies included in the regression analyses were observational, there is a high likelihood of referral bias, where older patients with more comorbidities preferentially underwent PCI instead of CABG.
There was a marked advantage of PCI over CABG in terms of stroke at a 30-day follow-up in the octogenarian population (OR = 0.14), however the advantage was not seen at 12-24 months and ≥36 months follow-up. This trend is in conformity with the NOBLE 2 trial. The elderly sub-group of the EXCEL 20 trial, however, did not detect statistically significant difference in the stroke rates in the both the treatment arms at short and long-term follow-ups, which might be secondary to a low event rate at 30 days in the PCI and CABG groups, that is, 1.3% and 0.6% respectively. The incidence of stroke at ≥36 months in our analysis was 4.8% (PCI) and 6.0% (CABG) which was similar to the stroke rate in the >75-year sub-group (4.9%  Unifying MACCE to include repeat revascularization provides robust scientific validity to our analyses and every effort was made to standardize outcomes wherever possible. Our analyses had limitations. All except one study included in our analyses were non-randomized, which could have confounded the results and introduced a potential selection bias but the quality assessment of the included studies with the Newcastle-Ottawa scale showed that 9 of the included observational studies scored 9 stars (maximum) while 7 studies scored a minimum of 7 stars (moderate quality). We did not have access to patient data. We had limited data on post-interventional medical therapies, degree of symptomatic improvement and the effect of revascularization on patients' quality of life. As we used mean age in the PCI arm to define the age cutoff, there is a chance that 8/17 studies included some patients who might have been <70-year-old, however, on the subgroup analysis of ≥80-year-old patients, we did not see an influence of those studies on outcomes other than repeat revascularization at 12-24 months. Our report was also limited by the absence of a longer-term follow-up where venous graft attrition and divergence in revascularization might become more evident. We did not have access to individual patientdata and end-point definitions could have been different in various studies, however we tried to carefully extract data with standardized outcomes wherever possible. It was also difficult to ascertain the complexity of underlying coronary artery disease, as most studies including the RCT did not report the SYNTAX score in the two treatment arms. Also, three included studies provided propensity-matched outcomes with a potential for a selection bias and consequent effect on outcomes when combined with unadjusted studies. However, a previously performed subgroup analysis after exclusion of these studies did not show an effect on clinical outcomes. 25 Similar to our meta-analyses, the meta regression analysis was performed with the mean age of PCI and CABG arm reported in each study, there is a chance that a proportion of the included patients were <70 years of age, raising concerns about generalizability of the study to all older adults.

| CONCLUSION
Older adults who undergo ULMCA revascularization had no difference in mortality with PCI or CABG at short and intermediate-term follow-ups. At long-term follow-up, CABG showed a favorable mortality trend as compared to PCI. Advancing age was also associated with increasing mortality with PCI. PCI was however associated with a markedly lower rate of early stroke in octogenarians as compared to CABG.