Section Editor(s): Miller, Ronald D. Section EditorSebel, Peter S. MB, BS, PhD, MBA^*; Bowdle, T. Andrew MD, PhD^†; Ghoneim, Mohamed M. MD^‡; Rampil, Ira J. MD^§; Padilla, Roger E. MD^||; Gan, Tong Joo MB, BS, FRCA, FFARCSI^¶; Domino, Karen B. MD, MPH^#

Awareness with recall after surgery under general anesthesia is an infrequent but well described adverse outcome (¹). Despite its relatively infrequent occurrence, awareness is of significant concern to patients (²) and is often associated with significant adverse psychological sequelae, including symptoms associated with posttraumatic stress disorder (PTSD) (^3–5). The occurrence of awareness is often the consequence of light-anesthetic techniques or smaller anesthetic doses (^6,7).

All data on the incidence of awareness during anesthesia come from outside the United States (US). Sandin et al. (³) reported an overall incidence of 0.16% in 11,785 patients treated at 2 hospitals in Sweden; the rate was 0.18% when neuromuscular blocking drugs were used and was 0.11% in their absence. Long-term follow-up of the patients who reported awareness showed a frequent incidence (approaching 50%) of PTSD 2 yr after the incident, even though patients initially did not report much distress (⁸). In a study from Australia, Myles et al. (²) reported an incidence of awareness of 0.10%; it was the highest risk factor for patient dissatisfaction after anesthesia.

Traditional clinical monitoring modalities during anesthesia are ineffective in preventing awareness. For instance, hypertension and tachycardia are generally not associated with reports of awareness (^5,7), and end-tidal anesthetic concentration monitoring is also ineffective (³). Recently, a monitor that uses a processed electroencephalogram (EEG) derivative, the Bispectral Index^® (BIS^®; Aspect Medical Systems, Newton, MA) has been introduced into clinical practice for monitoring anesthetic effects on the brain (^9,10). It may have the ability to measure the hypnotic component of the anesthetic state (¹⁰). The effectiveness of BIS monitoring to prevent awareness is unknown (¹¹), and it has even been suggested that guiding anesthetic administration by using BIS monitoring may be associated with an increased incidence because of deliberate reductions in anesthetic dose on the basis of BIS data (¹²).

Although the incidence of awareness is thought to be infrequent, many patients remain concerned about this potential adverse experience (^1,2), and this issue has attracted considerable public media attention. The incidence of awareness may vary between countries or institutions depending on their respective anesthetic practices and patient populations. This multicenter prospective cohort study was therefore undertaken to establish the incidence of awareness with recall during routine general anesthetic practice in the US and to determine (where possible) the BIS values associated with intraoperative awareness events.

Methods

The IRBs at seven geographically dispersed academic medical centers approved this prospective, nonrandomized, descriptive cohort study. Patients with informed consent (written or verbal, depending on site) were enrolled between April 2001 and December 2002. Inclusion criteria were patients scheduled to receive general anesthesia, aged ≥18 yr, apparently normal mental status (excluding obviously impaired patients), and ability to provide informed consent. Patients were excluded if they were not expected to survive, were transferred directly to the intensive care unit (ICU) and were not tracheally extubated within 1 wk, could not speak English, or had abnormal mental status that precluded answering the required questions. A sample size of 20,000 patients was estimated on the basis of previous incidence studies outside the US (³). Anesthetic care, including anesthetic drugs and use (or otherwise) of the BIS monitor (A2000 or A1050; Aspect Medical Systems) during the time of this study was entirely at the discretion of the attending anesthesiologist and was not influenced by participation in this study. In general, the attending anesthesiologist was not aware of patient participation in the study. Each patient was interviewed by research staff with the same structured interview, modified from Brice et al. (¹³), that was used in prior incidence studies (^3,13,14) (Table 1). Patients were interviewed first in the postanesthesia care unit (PACU) (if they were sent to the PACU and not directly to the ICU). At one site, the IRB required the interview to be conducted after the patient had left the PACU. A follow-up interview was attempted at least 1 wk after anesthesia.

Table 1

The principal investigators classified each patient report as awareness, possible awareness, dreaming, or no awareness, according to the definitions described in Table 2. Basic patient demographic and treatment data (e.g., age, sex, ASA physical status, type of surgery, and disposition after anesthesia) were also recorded on a standardized case report form used by all participating sites. In cases in which awareness was detected and BIS monitoring had been used, available BIS trends were retrospectively downloaded from the A2000 monitors’ electronic memory. Each patient with awareness was offered follow-up care according to each institution’s standard practice.

Table 2

Descriptive statistics were used to describe the incidence of awareness in the study population. Comparisons between groups were conducted with Fisher’s exact or χ² tests with Yates’ correction. Logistic regression models (SPSS; SPSS Inc., Chicago, IL) were used to determine associations of patient demographics with awareness and dreaming. Variables found to be significant on univariate analysis were entered into the forward-selection multivariate model. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated. P < 0.05 was accepted as significant.

Results

A total of 20,402 patients were initially enrolled, but 827 patients (4%) were excluded because they could not be interviewed after surgery (n = 793) or did not meet the inclusion criteria (n = 34), resulting in a final evaluable population of 19,575 patients. Eighty-five percent (n = 16,544) of the evaluable patients were interviewed in the recovery room, and 67% (n = 13,123) completed the postoperative follow-up interview (Table 3). The postoperative follow-up interview occurred between 1 and 2 wk in most patients (n = 9535; 73%), with follow-up at more than 2 wk in 24% (n = 3216) and less than 1 wk in 3% (n = 372). In the recovery room, 49 patients (0.30% of interviewed patients) reported remembering something between going to sleep and waking (yes to Question 3 in Table 1), whereas 80 (0.61%) reported intraoperative memories on the postoperative follow-up interview. Six percent (994 of 16,544) of patients reported dreaming (yes to Question 4 in Table 1) during the recovery room interview, and 3.43% (439 of 13,123) reported intraoperative dreams on the postoperative follow-up.

Table 3

From these interviews, 25 awareness cases (0.13%) were identified (see Table 4 for detailed patient descriptions of recollections). Awareness during anesthesia occurred at a fairly consistent rate of 1–2 cases per 1000 patients interviewed at each institution (range by site, 0.09%–0.21%;Fig. 1). In addition, 46 additional cases (0.23%) of possible awareness and 1183 reports of intraoperative dreaming (6.04%) were identified. Fourteen of the cases of awareness were identified only at the follow-up interview. Auditory perceptions and being unable to move or breathe were described by nearly half of the patients with awareness (Table 5). Anxiety/stress, pain, and the sensation of the endotracheal tube were also reported (Table 5).

Table 4

Table 5

Figure 1

The demographic characteristics of patients with awareness and possible awareness compared with no awareness are listed in Table 6. Logistic regression analysis demonstrated an association of awareness with increased ASA physical status, final disposition to the ICU, and procedure (abdominal, thoracic, cardiac, and ophthalmology versus other). ICU disposition was eliminated from the multivariate model (Table 7). Sex and age did not influence the incidence of awareness. There were no significant predictors of possible awareness.

Table 6

Table 7

The incidence of dreaming varied markedly by site (1.1%–10.7%; P < 0.001). Dreaming was associated with the following patient characteristics: younger age (OR, 2.43; 95% CI, 2.03–2.91 for <40 yr compared with ≥60 yr), lower ASA physical status (OR, 1.48; 95% CI, 1.29–1.70 for ASA status I–II compared with ASA status III–V), undergoing elective surgery (OR, 2.53; 95% CI, 1.04–6.15 compared with emergency surgery), and undergoing surgery on an ambulatory basis (OR, 1.40; 95% CI, 1.02–1.91 compared with disposition to ICU).

Thirty-eight percent of all cases in the study were monitored for portions of each case by using BIS. However, the use of BIS monitoring varied widely between study sites from 0% to 74% (P < 0.001), and not all cases that were BIS-monitored were monitored from induction to emergence. Figure 2 illustrates a case of awareness with BIS monitoring. BIS was consistently >60. There was no significant association between the use (or otherwise) of BIS and the incidence of awareness.

Figure 2

Discussion

This study demonstrates that the incidence of awareness with recall after general anesthesia (0.13%) in the US is comparable to that described in other countries (^2,3,6,14). Awareness during anesthesia therefore appears to be a ubiquitous phenomenon that occurs at an incidence of 1 to 2 cases per 1000, irrespective of geographic location and potential differences in anesthetics and techniques. Assuming that approximately 20 million general anesthetics are administered in the US annually, we can expect, on the basis of our data, approximately 26,000 cases of awareness to occur each year, or approximately 100 per workday.

Our estimate of the incidence of awareness is relatively conservative. If the cases of possible awareness are also considered, then the overall incidence of awareness increases to 0.36%. It is interesting to note that the incidence data from Sweden included several cases that were described as “possible” cases (³) on the basis of inability to confirm the reports. If we compare the incidence of awareness between this study and the data from Sweden including “possible” cases of awareness, our incidence of awareness is approximately twice that reported previously. It is also possible that knowing that they were participating in a study of awareness may have increased the incidence of patients’ self-reports. However, this would be true of all awareness incidence studies.

The detection of awareness depends on the interview technique, timing of the interview, and structure of the interview. A single short postoperative visit by an anesthesiologist without use of a structured interview is unlikely to elicit many cases of awareness. We used the same structured interview that has been used in prior investigations (^6,13,14). We interviewed patients in the PACU and again after seven days because it has previously been demonstrated that approximately 35% of cases are detected only at a delayed postoperative interview (³). Approximately one half of the cases in our study were detected only at the second interview. The loss of follow-up at the postoperative interview would therefore bias our results in the direction of underestimating the incidence of awareness during anesthesia.

The descriptions of the awareness cases identified in this study closely resemble those reported previously (^3–6). As might be expected, a significant proportion of the awareness episodes occurred either during endotracheal intubation or at surgical incision, i.e., times when the level of patients’ stimulation is highest. Patients reported auditory recollections, sensations of not being able to breathe, paralysis, panic, and pain (Tables 4 and 5), consistent with previous reports (^3–6). Our study did not assess the long-term psychological sequelae of intraoperative awareness.

Awareness is caused by the administration of general anesthesia that is inadequate to maintain unconsciousness and to prevent recall during surgical stimulation. Common causes include large anesthetic requirements, equipment misuse or failure, and smaller doses of anesthetic drugs (¹). Our finding of an increased risk of awareness with sicker patients (ASA physical status III–V) undergoing major surgery (Table 7) may reflect the use of smaller anesthetic doses and light anesthetic techniques in sicker patients. However, specific details of anesthetic doses and intraoperative hemodynamics in patients with awareness compared with those without awareness were not obtained in this incidence study. Although female sex and younger age have been suggested as risk factors for intraoperative awareness on the basis of analysis of closed malpractice claims (⁷), our study found no association between sex and age and awareness during anesthesia.

Dreaming during anesthesia was described by 6% of patients in our study, and this is consistent with the common occurrence of perioperative dreaming reported in several small European studies (^13,15,16). Dreaming was more frequently reported in the recovery room than later after surgery; this is also consistent with earlier studies (¹⁵). Dreaming was associated with younger, healthier patients undergoing ambulatory surgery. The widely varying incidence in dreaming by study site (1.1%–10.7%) may reflect differences in patients or anesthetic drugs, or, alternatively, it may reflect bias in patient selection or responses between the geographically dispersed centers (¹⁶). The significance of dreaming and its relationship to awareness during anesthesia is unclear.

In many cases, awareness during anesthesia is a potentially avoidable adverse anesthetic outcome. In light of follow-up studies suggesting that such “victims of awareness” (⁸) may exhibit significant psychological aftereffects, such as PTSD, attempts to further reduce its incidence are warranted. Because awareness occurred despite the usual clinical monitoring of anesthetic depth (e.g., blood pressure, heart rate, and end-tidal anesthetic monitoring) in this study and others (^3,5,7), a monitor of cerebral function and depth of anesthesia may be of theoretical benefit.

One such monitor, the BIS monitor, is a complex processed EEG derivative that assigns a numerical value to the probability of consciousness. Recovery of consciousness during general anesthesia without any recall (in the absence of surgical stimulus) has generally been associated with BIS values >60 (^17,18). Cases of awareness during surgical stimulation with high BIS values (>60) have also been reported (^19,20). Although there is at least one case report of awareness with a BIS of apparently <50 (²¹), BIS was subsequently found to be >60 at the probable time of awareness (²²). In the present study, a number of the cases of awareness in which BIS was used also had high BIS values (see, for example, Fig. 2). We were unable to positively identify any cases of awareness with BIS values <60, but no firm conclusions can be drawn from this observation. This study was not designed to test the efficacy of BIS monitoring because the population that received additional monitoring was not randomly selected or matched to those who did not, and no specific guidelines for BIS were used. Other emerging data suggest that BIS monitoring is effective in reducing the incidence of awareness. Ekman et al. (²³) investigated the incidence of awareness when the anesthetic was guided with BIS and found a 77% reduction in the incidence of awareness (²³) compared with historical data (³). Myles et al. (²⁴) also found that, in a double-blind study of patients at high risk for awareness, BIS-guided anesthesia resulted in an 82% reduction in the incidence of awareness.

In summary, the incidence of awareness during general anesthesia in the US was 0.13%. It occurred at a rate of 1–2 per 1000 patients interviewed at each site.

Statistical processing support was provided by Aspect Medical Systems, Inc. We thank Jeff Sigl, PhD, and Paul Manberg, PhD, from Aspect Medical Systems for providing statistical analysis and helpful suggestions during the preparation of this manuscript. We also appreciate the assistance of Scott Acker, RN, Antonio Adam, MD, Kerith Brandt, Catherine Dobres, CRNA, Samantha Goldstein, BA, Meghan Holmes, MA, Kui-Ran Jiao, MD, David Kramer, Yumi Lee, MD, Jason Leggio, CRNA, Tanya Lipto, RN, Lee McClurkin, RN, Rachel Pessah, BS, Jacqueline Sumanis, CRNA, Meghan Swardstrom, and Thu Tran, BS.

References