Letters to the Editor: Letters & Announcements
To the Editor:
We believe that the article by Monk et al. (1) should not have been published. The article presented the conclusion that cumulative time with Bispectral Index (BIS) scores <45 was an independent predictor of 1-yr mortality in patients undergoing major, noncardiac surgery. We doubt the validity of that result. Our principal concern is that the statistical analysis methods were determined post hoc. Although the study is described as prospective, a footnote indicates that the data analysis was performed by a statistician who was not involved in the design of the study. Furthermore, a preliminary description of this investigation (2) employed different statistical methods and different BIS thresholds for the analysis of mortality subgroups. It is an established principle of complex observational studies of this nature that the statistical methods be established before data acquisition. In addition, the statistician was an employee of Aspect Medical, a company whose business prospects might well be influenced by the results of the investigation. That statistician was also involved in the development of the manuscript. We submit that even if the science inherent to this investigation were impeccable, this inclusion of an Aspect employee inevitably and irrevocably taints the credibility of the results.
What should the Board have done? Flatly reject the paper? We think not. The issue is too important. The Journal should have asked—and should now ask—the authors for the original protocol, including the originally proposed statistical methods. If the methods have been changed, the authors should be called on to explain why. Whatever the response, an independent statistician should, on the basis of the original hypotheses and protocol, recommend appropriate statistical methods. That statistician should receive the entire dataset and perform the analysis. The Editorial Board or its designees should undertake the reporting of those results. This is an extraordinary process for dealing with any manuscript. However, we believe that publishing it was an extraordinary error that warrants an unusual course of redress.
We call to the Editorial Board’s attention one specific reason why clarification of this dataset is important. Myles et al. (3) reported that the use of BIS monitoring in patients at high risk for awareness reduced the incidence of that phenomenon. However, the average BIS value in monitored patients in that study was 44.5. Ekman et al. (4) reported that BIS monitoring reduced the incidence of awareness as compared with a historical control group. In that investigation, the average BIS value during maintenance of anesthesia was 38.5. Clinicians will see the implication quickly. The use of a monitor that results in sufficiently deep hypnosis to prevent awareness in Australia and Sweden results in average BIS values that are associated with increased mortality in Florida. This paradox must be resolved. The validity of the mortality-hypnosis relationship reported by Monk et al. deserves examination and verification by the independent statistical examination of the data set that we propose.
John C. Drummond, MD, FRCPC
Piyush M. Patel, MD, FRCPC
Department of Anesthesiology
The University of California, San Diego
VA Medical Center, San Diego
1. Monk TG, Saini V, Weldon BC, Sigl JC. Anesthetic management and one-year mortality after noncardiac surgery. Anesth Analg 2005;100:4–10.
2. Weldon B, Mahla M, van der Aa M, Monk T. Advancing age and deeper intraoperative anesthetic levels are associated with higher first year death rates [abstract]. Anesthesiology 2002;96:A1097.
3. Myles PS, Leslie K, McNeil J, et al. Bispectral index monitoring to prevent awareness during anaesthesia: the B-Aware randomised controlled trial. Lancet 2004;363:1757–63.
4. Ekman A, Lindholm ML, Lennmarken C, Sandin R. Reduction in the incidence of awareness using BIS monitoring. Acta Anaesthesiol Scand 2004;48:20–6.