Volatile Agents versus Propofol in Cardiac Surgery: Reply

We thank Yonekura et al.¹  for raising the problem of the risk of bias in our meta-analysis.²  First, we agree that it is almost impossible for anesthesiologists to perform fully blindly either volatile anesthetics or total intravenous anesthesia for anesthesia maintenance. Accordingly, some performance bias in trials comparing two anesthetics is unavoidable. However, such a bias can be markedly reduced if researchers other than anesthesiologists examine data and, more importantly, when outcomes are objective as in our meta-analysis. With regard to the recent article by Landoni et al.,³  the early conclusion of the trial may have in fact underestimated the treatment effect, so the inclusion in our meta-analysis of the published data may in turn have theoretically reduced the treatment effect. Nevertheless, another recent meta-analysis reports low risk of bias with regard to the above-mentioned trial.⁴  Moreover, by considering all recent meta-analyses that report in detail the risk of bias,^4–8  results are similar to ours (61% low risk, 31% unclear risk, 8% moderate/high risk).

Second, in our meta-analysis we considered data on 1-yr mortality reported by Likhvantsev et al.⁹  in the result session (i.e., 52 events in the volatile group and 81 in the total intravenous anesthesia group). However, we related these events to all studied patients (437 in the volatile group and 431 in the total intravenous anesthesia group) rather than to an unexplained number of 292 patients in the volatile group and 326 in the total intravenous anesthesia group. Accordingly, the mortality rate we reported is 11.9% versus 18.8% instead of 17.8% versus 24.8%. When combining data reported in tables 3 and 4, mortality is 17.8% in the volatile group (78 of 437 patients) and 25.3% in the total intravenous anesthesia group (109 of 431). The reason for these differences in results is unknown. According to the above-reported data, the Likhvantsev et al.⁹  study has a high risk of attrition bias, even though this bias is considered low in the meta-analysis by Jiao et al.⁴  or high in that by El Dib et al.⁶ 

Regarding the last point, we agree that the grading of recommendations assessment, development, and evaluation (GRADE) checklist¹⁰  is probably more complete than Cochrane risk of bias, although the latter is predominant in recent meta-analyses including those on our topic. In conclusion, the careful reanalysis of methodologic issues seems to validate the results of our meta-analysis.

The authors declare no competing interests.