To the Editor:
We read with great interest the systematic review and meta-analysis by Bonanni et al.,1 recently published in Anesthesiology, in which the authors investigated how different anesthetic agents (volatile vs. propofol) affected outcomes in patients who had undergone cardiac surgery with cardiopulmonary bypass. The main finding—that volatile anesthetics were superior to propofol with regard to long-term mortality as well as cardioprotective effects—is clinically valuable information. Some methodologic issues should be further discussed and clarified, however, and there is a need for data validation, for three main reasons which we describe below.
First, we are skeptical about the authors’ assertion that the majority of the studies were at low risk of bias (Supplemental Digital Content 2, https://links.lww.com/ALN/C280, in Bonanni et al.1 ). It is almost impossible for anesthesiologists to perform either total intravenous anesthesia or volatile induction and maintenance anesthesia blindly. Performance bias is therefore unavoidable in trials comparing these anesthesia methods,2 and its potential influence will be stronger in some selected studies. A recent trial reported by Landoni et al.3 that was included in the Bonanni et al.1 meta-analysis was terminated early for the reason of futility, reducing the power of the study and leading to an underestimation of the treatment effect.4 This may have induced bias. Because appraisal of risk of bias in included studies is an integral part of systematic review methodology, the authors should clarify why most of the trials they assessed were deemed to have low risk of bias.
Second, the positive results pertaining to improved long-term survival with volatile anesthetics are driven mainly by the study of Likhvantsev et al.,5 but the long-term mortality rate in that study was considerably higher (18.8% in the propofol group) than it was in the other studies analyzed (4.2%), which Likhvantsev et al.5 acknowledged in their study. In addition, we assume that their study entailed high risk of attrition bias, not low risk of bias. In the funnel plot of Figure 2,1 there were 52 deaths in the volatile group (437 patients) and 81 in the propofol group (431 patients) in the study by Likhvantsev et al.5 However, the original study of Likhvantsev et al.5 reported that 1-yr mortality rates were 52 of 292 (17.8%) in the sevoflurane group and 81 of 326 (24.8%) in the total intravenous anesthesia group.5 No explanation of the lost to follow-up rate at 1 yr was provided, but with more than a quarter of patients lost to follow-up—resulting in substantially incomplete outcome data—high risk of attrition bias is arguably inevitable. Accordingly, the authors should clearly state any assumptions or imputation methods to handle missing data, and the effects of imputation should be investigated via sensitivity analyses, which may change interpretations of the results.6
Last, we would like to see the overall quality of the evidence assessed via the grading of recommendations assessment, development, and evaluation (GRADE) framework for relevant outcomes.7 The information thus derived would be valuable to the readers of Anesthesiology.
Competing Interests
The authors declare no competing interests.