We read with great interest the paper by Jiang et al. published in a recent issue of your esteemed journal.1 The authors designed a clinical study comparing general and epidural anaesthesia (GAEA) with only general anaesthesia (GA) for esophagectomy. The authors concluded that GAEA might reduce the inflammatory response and cognitive dysfunction after esophagectomy.
Randomised controlled trials (RCTs) provide the highest level of evidence and, undeniably, are the design of choice to compare the effectiveness of GAEA vs. GA for esophagectomy. However, any study can have systematic errors, and readers should have a critical view and detect the main risk of bias.
Certain methodological aspects of Jiang et al.’s trial favoured bias. First of all, the study lacks a protocol. Protocols are pre-specified documents published in a database reporting the study’s methodology and objectives, so that, the reader can check if the authors complied with the initially proposed plan. Besides, a protocol defines the outcomes to be evaluated. If a protocol is not well established, the reader cannot know if all the evaluated outcomes were reported, predisposing the study to report bias. The study protocol also should predefine if an intention-to-treat (ITT), per-protocol (PP), or as-treated (AS) analysis will be made.2 Theoretically, if there is no pre-specified protocol, the authors could perform any of the three analyses and present only the one that fits better with the result they are looking for. Probably, that is not the case in Jiang et al.’s trial. However, a protocol would give transparency.
In Jiang et al.’s trial, there were eight exclusions related to surgery cancellation. Since there is no specification if the surgery suspension was related to any anaesthetic intercurrence, it is hard to know if this data loss could interfere with the final analysis. If an operation cancellation was related to the studied intervention, this exclusion would not be appropriate, and an ITT would be a better fit than a PP.
Another issue is that the authors did not clarify the randomisation process. Choosing an appropriate randomisation methodology (simple, block, stratified, or other) is essential to avoid selection bias and ensure the balance of groups’ baseline characteristics.3 Despite the authors reporting that the groups' baseline characteristics were similar, no information was provided regarding the surgical techniques. McKeown, Ivor Lewis, and other esophagectomy techniques may differently impact patients' inflammatory responses.
Finally, there is no mention of how the authors estimated the sample size. An improper sample may induce type I and II errors. Readers should have a critical appraisal and know the limitations of any paper, even when the study design suggests the highest level of evidence.
The authors declared no competing interest.
FT: Conceptualisation and supervision.
MV: Writing and editing.