Radial artery versus right mammary artery: the fight is always open, and statistics rule

Excerpt

Magouliotis et al. elaborated a comprehensive systematic review and meta-analysis with the aim of comparing the long-term survival in coronary patients receiving either single arterial (SAG) or multiple arterial grafts (MAG) during coronary artery bypass grafting (CABG) surgery [1]. The authors were able to identify 39 articles including a total sample size of more than 180,000 patients (SAG:124,284; MAG:56,175). The median follow-up was 17.5 years in MAG and 11.6 years in SAG group. A series of subgroups analysis was performed comparing SAG vs MAG in patients over 70 years, in diabetics, in females and in radial artery (RA) vs right internal mammary artery (RIMA). Interestingly, patients who received MAG had a significantly better outcome in the overall population and with the subgroups. In RA vs RIMA subgroup, patients with RIMA as second conduit had a longer survival compared to RA patients. This meta-analysis keeps open some topics of interest: first, the excellent long-term results of the MAG approach [2‐4]; second, the underused application of MAG surgery. Another interesting debated topic concerns the use of either RA or RIMA as second conduit. Based on the reconstructed Kaplan–Meier curves, the authors have observed an advantage of RITA over RA in the long-term overall survival. They tried to explain this difference with several shareable reasons. The Kaplan–Meier curves show a clear difference in term of longest follow-up in each group, with the RA group being the lowest. Although the authors have demonstrated rigor in performing their meta-analysis, it suffers from a limitation in statistical analysis. We believe that testing the proportional hazard assumption is mandatory, as the two curves seem to slightly diverge after 10 years and then probably cross at about 18 years. In addition, analysis should be truncated at the shortest follow-up (RA group) in order to balance the survival of both groups as the hazard-ratio probably is affected by the longest follow-up of RIMA group. If one visually inspects the Kaplan–Meier curves of the overall population and the other subgroups, it is clear visible the curves diverge progressively. In these situations, it is easily to assume the hazard-ratio proportionality. In contrast, the reconstructed Kaplan–Meier curves for RA vs RIMA seem to show equipollent hazards of both curves. That said, we surely appreciated the results obtained from this elegant study, which encourage wider use of MAG in daily activities. However, considering the recent evidence on the comparable results of RIMA vs RA as second conduit [5], we should always ask ourselves whether statistics are always right and whether we have always applied the most appropriate statistical approach. …