To the Editor:—
We read with interest the study by Pollard et al. 1suggesting that the incidence of awareness during anesthesia is less than one tenth of that which has been previously described. We are concerned about the implications of these results and wish to comment on the methodology used in this study.
There are several prospective studies in the literature2,3that suggest the incidence of awareness during general anesthesia is approximately 1 in 1,000. The common features of all of these studies are that they used the same brief questionnaire and that they interview patients on a second occasion approximately 1 week after surgery.
The data gathered by Pollard et al. 1was not designed to elucidate the incidence of awareness. Rather, it is a quality improvement database. We have some concerns about the robustness of the data. It is our general impression that quality assurance databases are not collected with the same rigor as research databases. From the article, is seems that there were more than 175,000 completed interviews (2 interviews per patient), which means more than 350,000 data points. Were the answers to all questions recorded in every case? Or are there missing data? If so, how much is missing and how were missing data treated? Another cause for concern is the failure to capture approximately 17% of the total database for unknown reasons. This may also have led to an underestimation of the incidence of awareness.
The authors used for detection of awareness a modification of the interview described by Liu et al. 4(who modified the questionnaire of Brice et al. 5), which has been the standard for years. Instead of asking the patients the crucial question “Do you remember anything between going to sleep and waking up?” they asked, “Did you have any dreams while you were asleep for surgery?” thus confusing dreaming and awareness. A further possible confusion may have been created by adding what seems like irrelevant questions, e.g. , “Were you put to sleep gently?” The argument that no one has proved that one set of questions is better than the other is factual but superfluous. The Liu et al. 4modification of the interview described by Brice et al. 5has been used by several independent investigators over the years who have demonstrated its reliability, robustness, and absence of suggestibility. It is difficult to think of a further modification for the better.
Other studies report an incidence of dreaming around 6%. The authors specifically asked about dreaming but presented no data related to this question. How many patients dreamed? Were these dreams ever disturbing?
The other, prospective, studies on the incidence of awareness conducted the second interview later than in the study of Pollard et al. ,1where the second interview was conducted at 48 h. Sandin et al. 2found that only 65% of cases of awareness are elucidated on the first interview. The other 35% of cases are reported on the second interview a week (or more) later. This may have contributed to a substantial underestimation of the incidence of awareness.
In our opinion, the elimination of the “crucial” question from the modified Brice interview4,5means that the data from this study cannot be compared directly with the prospective data reported elsewhere. It is well known6that patients do not spontaneously report episodes of awareness to their anesthesiologist. If you don't ask about awareness, patients are not going to tell you about it.
*Emory University School of Medicine, Atlanta, Georgia. firstname.lastname@example.org