Radial artery versus saphenous vein versus right internal thoracic artery for coronary artery bypass grafting: different conduits or different trials?

Information from numerous observational studies and secondary analyses from randomized controlled trials suggest that increased use of arterial grafts during coronary artery bypass graft (CABG) surgery may provide better myocardial revascularization and improved outcomes [1, 2]. However, the Arterial Revascularization Trial (ART) is the largest randomized controlled trial (3102 patients) to compare the use of bilateral internal thoracic artery (BITA) versus single internal thoracic artery (SITA) grafting and failed to show improved all-cause mortality at 10 years with the BITA approach in an intention-to-treat analysis [3]. Importantly, a post hoc as-treated analysis showed a significant survival advantage in patients who received >1 arterial graft in comparison to those who received a single arterial graft and supplemental veins [3]. Moreover, various limitations to the ART trial may explain the primary finding, including ∼14% of patients assigned BITA who actually received SITA, and about a fifth of patients in both groups who received a radial artery (RA) graft. These aspects may have preferentially benefited the SITA group in the intention-to-treat analysis, making detection of a clinically relevant difference more difficult. Interest in the RA as an additional conduit in CABG has been growing over the last 2 decades and improvements in harvesting and handling have increased confidence in its use. An individual patient data meta-analysis of randomized trials with a total of 1036 patients showed improved patency and composite cardiovascular outcomes for patients receiving LITA + RA compared to those who received LITA + SVG at 5 years [4]. A network meta-analysis by Gaudino et al. [5] with direct and indirect comparisons of observational studies and randomized trials including about 150 000 patients concluded that there could be similar proportional reductions in the mortality of about a quarter with the use of BITA or LITA + RA compared to the use of LITA + SVG and concluding ‘the use of the RA or the RITA is associated with a similar and statistically significant long-term clinical benefit compared with the SV’. The report by Gaudino et al. [6] in this journal includes individual patient data from 4 randomized controlled trials: ART, CORONARY, PREVENT IV and RAPCO. The main question addressed by this analysis was whether use of the RA or RITA could provide better mortality outcomes compared to SVG in patients undergoing CABG and receiving LITA as the standard approach at about 8 years of follow-up. It is important to note that almost 60% of the BITA group in this analysis came from ART and that data from RAPCO were excluded. The mean age of the LITA + RA group (61 years) was on average 3 years younger than the other groups (P = 0.004), and 41% of the patients in the LITA + RA group received only 2 grafts in marked contrast to the other groups (P < 0.001), suggesting a lesser severity of coronary artery disease. It is unlikely that these important differences in baseline characteristics can be fully adjusted in statistical models. The main findings of the report were that all-cause mortality in the propensity-matched model was ∼40% lower in the LITA + RA group compared to the LITA + SVG group [hazard ratio (HR) 0.62, 95% confidence intervals (CIs): 0.51—0.76; P = 0.003] and also about 40% lower compared to the BITA group (HR 0.59, 95% CI: 0.48–0.71; P = 0.001). Interestingly, the results for composite cardiovascular outcomes were similar for the LITA + RA and BITA groups compared to LITA + SVG with HR around 0.75 in both groups. There was little difference between mortality outcomes of BITA versus LITA + SVG in the unadjusted analysis (HR = 0.92 for mortality and 0.96 for MACCE; Supplementary Material, Table S5). In the propensity-matched analysis, the HR for mortality in the BITA versus LITA + SVG comparison was 1.07 (Table 3). For the LITA + RA versus LITA + SVG comparison, the unadjusted HR was 0.66 for mortality while the adjusted HR was 0.62. There were many layers to the statistical methods that led to a complex analysis including meta-analysis of 3 trials, propensity score matching, matched triplicates and matching carried out with replacement. The final analysis is therefore limited by several issues. First, in a meta-analysis, the statistical model should incorporate a random effect, stratification, or a similar approach to account for the heterogeneity of treatment effects across trials. Second, combining propensity score matching and meta-analysis is unusual, with some experts suggesting that matching should be carried out preferentially from within the trials [7], but it is not apparent that this was done. Third, matching with replacement usually requires adjustment of standard errors, as does the triplicate membership. Although these aspects were apparently incorporated into the model, it is not clear what was done about missing covariate data. Finally, the impact of differential follow-up of patients in the 3 trials is not clear. ART, which contributed most of the BITA patients, had 10-year follow-up compared to 5 years in the other 2 trials. These complexities and limitations suggest that the conclusions of the analysis may not be reliable. The direct intention-to-treat comparison of BITA versus SITA in ART failed to demonstrate a significant mortality reduction of a second ITA graft, but there was no signal of increased mortality with BITA in comparison to LITA + SVG [3]. Indeed, a post hoc as-treated analysis showed improved survival with multiple arterial grafting [3]. This is consistent with a large body of contemporary and historic observational data reporting a survival benefit of BITA grafting compared to LITA + SVG that increases with longer follow-up. Conversely, the meta-analysis by Gaudino et al. [4] of individual patient data of randomized trials of RA versus SVG failed to show any signal of mortality reduction in the RA group although event rates were low. The conclusion of Gaudino et al. [6] that LITA + RA could provide a 40% reduction in mortality compared to BITA is biologically implausible and is inconsistent with a large and well-established existing literature [5]. Indeed, we know of no surgical intervention that has ever been associated with a 40% relative reduction in mortality in CABG patients, especially over just 8 years of follow-up (i.e. 5% difference per year!). Thus, we are concerned that the conclusions reached in this analysis may have adverse consequences for patients undergoing CABG by reducing the use of BITA grafting, which may not be biologically justified. Consequently, an independent review of this complex statistical analysis may be warranted.