The incidence of awareness with recall during anesthesia has been estimated to be 0.2%-0.3% during general surgery [1-3]. There are voluminous data of the effects of various anesthetic and sedative drugs on memory . However, few studies correlate the amount of different anesthetics to the incidence of awareness during clinical anesthesia . There are case reports citing psychiatric syndromes as a result of intraoperative awareness with recall. These psychological problems, which may persist for long periods of time, have usually involved sleep disturbances, depression, anxiety, and a preoccupation with death [6-8]. We designed a study about the incidence of awareness with recall during modern general anesthesia, the doses of anesthetics related to this phenomenon, and the nature and prognosis of psychiatric disorders after intraoperative awareness.
We conducted a prospective, open, cross-sectional study of the incidence of awareness, and a structured interview method was used. The psychiatric part of the study was of a single-blind, case-controlled, and follow-up design. The study was completed in a secondary care center serving a population of approximately 200,000.
We were interested in patients with explicit recall of a period of general anesthesia. For study purposes, patients reporting recall of awareness were divided into three groups: Group 1 = patients with unclear memories or dreams, which could be of intraoperative origin; Group 2 = patients with short periods of awareness occurring either intraoperatively or during the period of awakening from anesthesia; and Group 3 = patients with long-lasting, clear, and undisputed recall of the intraoperative period. For statistical calculations, either the patients in Groups 2 and 3 or the patients in Groups 1-3 were combined. The patients with awareness were compared with control patients without awareness.
The patients were also asked about dreaming. The dreams reported were classified as pleasant, neutral, or unpleasant, according to the emotional content of the dreams.
After obtaining permission from the institutional ethics committee, all patients undergoing elective surgery under general anesthesia in the Central Hospital of Paijat-Hame, Lahti, Finland between August 1994 and August 1995 were considered for the study. Patients excluded from the study were those less than 12 yr old, those not speaking Finnish, those unable to respond because of their disease, and those who refused the interview. For the screening of possible awareness, all the patients included in the study were structurally interviewed immediately before leaving the postanesthesia care unit. Questions asked during the interviews were those originally described by Brice et al. . The interviews were performed by the attending nurse anesthetists of the recovery room. The patients who, in the interview, reported recall of any incident from the period of general anesthesia were re-interviewed later during the day of operation or the day after, and were then referred for a psychiatric interview.
Matched control patients for the psychiatric testing were recruited as follows: the control patients were matched with their respective patient with awareness for the following variables: age (+/- 5 yr), sex, weight (+/- 5 kg), body mass index (+/- 3%), type of surgery, date of surgery (+/- 3 wk), and ASA physical status. One control patient was found for each patient with awareness.
The anesthesiologists were told that a study of awareness during general anesthesia was being conducted in the hospital. No additional information was given, and no attempt was made to standardize the anesthesia methods used in the hospital.
The doses of anesthetics used for general anesthesia of patients was estimated from random samples. First, a sample of 103 anesthetic records was collected from the 3-mo period immediately preceding the interview period. We supposed that this period represents the normal practice of the anesthesiologists. During the interview period, we collected samples of 108 anesthetic records during each of the four 3-mo periods of the study. Thus, we had a database of 535 random anesthetic records. We studied possible change in the doses of anesthetics over time and compared the doses received by the patients with awareness with those of the patients without awareness. This database was also used for Poisson regression analysis (see below). For comparison of the doses of anesthetics received by the patients with or without awareness, the records of patients not interviewed for some reason were excluded from the sample. After this exclusion, 247 anesthetic records were left for the analysis.
For study purposes, a continuous administration of an anesthetic was defined as follows: administration of an inhaled or IV anesthetic as started within 15 min after induction and continued throughout the anesthesia until the end of surgery. To be defined as continuous administration, a break of no longer than 15 min in the administration was allowed .
The dose of the inhaled anesthetics was calculated from the anesthetic records as follows. The inspired concentration of the anesthetic was multiplied by the time (in minutes) that this concentration was used. These totals for different concentrations and times were then added to a grand total. This was divided by the anesthesia time. This Figure wasthen used as an estimate of how much anesthetic the anesthesiologist intended to give to the patient .
If the patients who experienced awareness and the matched controls were able and willing to give informed consent, they were interviewed by one of the authors (RL) immediately after the operation. If the patient had already left the hospital, the first interview was conducted on the telephone, and the first face-to-face interview was conducted the first 2 wk after the operation. The patients were followed-up and reinterviewed 2 and 6 mo after the operation.
The interviewer was blinded to the possible experience of awareness during the operation. All diagnoses were made according to strict Diagnostic and Statistical Manual (3rd revised edition; DSM-III-R) criteria by using the Structured Clinical Interview for DSM-III-R Nonpatient edition (SCID-NP)  and Structured Clinical Interview for DSM-III-R Personality Disorders (SCID-II) .
Patients completed the Symptom Checklist 90-Revised, a measure of psychological distress . The Symptom Checklist 90-Revised was used to make the diagnostic structured interview complete. Patients were also asked to complete the Impact of Event Scale, a measure that indexes symptoms that characterize posttraumatic stress disorder (PTSD) .
The chi squared test was used to compare frequencies, Student's t-test was used for independent samples, and analysis of variance (with Tukey's honestly significant difference post hoc test) was used as appropriate. A P value of 0.05 was considered statistically significant.
Poisson regression was used to create models of factors influencing the occurrence of awareness. For this analysis, the anesthetic data of all patients with possible awareness (Groups 1-3) were used. The outcome (awareness or no awareness) formed the dependent factor. The independent factors entered into the Poisson regression analysis included body mass index, ASA physical status, laparoscopic or open surgery, type of surgery (orthopedic surgery, obstetrical and gynecological surgery, otorhinolaryngological or eye surgery, and abdominal or thoracic surgery each formed a type), type of premedication (amnestic/nonamnestic), continuous or noncontinuous mode of administration of the inhaled anesthetic, administration of nitrous oxide (60%-75% inspiratory fraction), the doses atracurium, fentanyl, and propofol (per weight and anesthesia time), the dose of thiopental per weight, and the dose of isoflurane (mean inspiratory fraction percentage) administered.
During the study period, 4818 patients (1215 male and 3603 female) underwent elective surgery under general anesthesia. Of these patients, 2612 (54.2%) were interviewed (608 male and 2004 female). Of the elective male and female patients, 50.0% and 55.6% were interviewed, respectively (P < 0.001). Most patients not interviewed were not entered into the study because the nurse anesthetists in the postanesthesia care unit were too busy to enter all eligible patients. One hundred twenty-one patients were interviewed more than once (i.e., they underwent surgery more than once, and hence were interviewed on separate occasions). Therefore, 2484 separate patients were interviewed during the study.
A variety of drugs were used as the general anesthetic. Of the patients, 60% received premedications, usually diazepam (50%) or oxycodone (9%). Either thiopental (in 59% of patients) or propofol (38%) was used for the induction of anesthesia, and anesthesia was maintained with nitrous oxide and a volatile anesthetic (isoflurane 49%, enflurane 11%) supplemented with an opioid (fentanyl 70%, alfentanil 28%). Muscular relaxation was usually achieved with the use of atracurium. For short surgical procedures, only an induction bolus dose and, possibly, some additional boluses, of an IV anesthetic were used. During the study, there was a significant change over time in the amount of isoflurane and atracurium used (Table 1).
There was no significant change in the incidence of awareness with recall over time during the study (Table 1). One patient (Patient 1, Table 2) felt pain during the period of awareness (Table 2). Another (Patient 10, Table 2) possibly felt pain during the awareness period, but the pain may also have been of postoperative origin. There were no differences in sex or age between patients without awareness and those with awareness either between Groups 2 and 3 or among Groups 1-3.
The doses of anesthetics received by the patients with awareness and recall (Groups 1-3) were compared with the random patient database collected for the analysis of the doses of anesthetics used in the hospital (Table 3). There were significant differences in the doses of propofol and isoflurane between patients with and without awareness in the Poisson regression model. Continuous dosing of the principal anesthetic and use of amnestic premedication were used more often in patients without awareness, but the differences between the groups were not significant (P = 0.06 and 0.07, respectively). There were no significant differences in the Poisson regression models between the groups with and without awareness for the following factors: ASA physical status, body mass index, type of surgery, and laparoscopic or nonlaparoscopic surgery.
Women reported dreams significantly (P < 0.01) more often than men (14.2% vs 9.4% for women versus men). The patients reporting dreams were significantly (P < 0.001) younger than those who did not report dreams (41 +/- 15 vs 45 +/- 17 yr, respectively). Eight women (0.4% of women interviewed) and no men remembered unpleasant dreams, but the difference between the sexes was not significant. Two of the eight patients with unpleasant intraoperative dreams also had possible awareness (Group 1).
Of the 19 patients with awareness, 5 (26%) underwent psychiatric interviews. Another 2 patients were originally evaluated for the psychiatric interviews. One (Patient 2, Table 2) did not remember the experience of awareness the day after the operation and was in too poor a physical condition to be enrolled in the psychiatric study. The other patient (Patient 5, Table 2) was too incoherent to follow the structured psychiatric interview. The remaining 12 patients reported memories or replied in the interview in a way that was not considered as suggesting intraoperative awareness by the recovery room nurses. These patients were only found during the final scrutiny of the interview charts in the end of the study, and therefore could not have been enrolled in the psychiatric evaluation.
In the psychiatric interviews, there was one 31-yr-old woman (Patient 4, Table 2), who felt intense fear during the extubation and had experienced sleep disturbances afterward, but she did not meet the criteria for PTSD. None of the other patients interviewed and followed-up by the psychiatrist experienced any after effects.
Three of the five patients with awareness had a history of major depression and also had preoperative symptoms of depression and anxiety. None of the control group patients had any preoperative psychiatric disorders. The mental state of two depressive patients did not significantly change after the operation, but one patient was treated with antidepressant medication and recovered.
In the present study, the incidence of awareness with recall (0.4%-0.7%) and the number of patients who remembered pain (2 of 19 patients with awareness) were of a magnitude similar to that reported in other studies [1-3]. The patients with awareness received smaller doses of principal anesthetics than the patients without awareness. Previous clinical studies have not compared the doses or the modes of administration of anesthetics used in patients with and without awareness. Perhaps amnestic premedication and continuous administration of anesthetics are ways to reduce awareness during general anesthesia [4,14]; this suggestion seems to be supported by our findings.
An unresolved question in studies of this type is the timing of the interview [2,15]. Perhaps the interview should take place relatively soon after the patient awakens from the anesthesia . Brice et al.  have described a patient who did not recall awareness recalled a week earlier at the end of an anesthetic. In the present study, there was also one patient (Patient 2, Table 2) who had clear recall of awareness during the first interview in the recovery room but denied any memories of anesthesia when interviewed in the ward the next day. Thus, it seems that the interview should take place as soon as possible. Whether this should be followed-up by a later interview (one to two weeks after anesthesia) cannot be concluded from the present study.
Another problem is with the definition of this kind of complication. In clear cases of intraoperative awareness with recall, there is no problem in deciding whether the patient was aware during anesthesia. The decision is more difficult with a patient with confusing memories. Therefore, we have developed a classification of awareness into three grades ranging from clear, undisputed awareness to spurious, possible awareness. Grade 1 awareness in the present study includes patients remembering dreams, illusions, or hallucinations . The difficulty in defining awareness makes it difficult to compare studies of the incidence of awareness. Most previous studies have not clearly stated their definition of awareness [2,3].
The psychiatric sequelae after awareness in the present study were not common. It seems that in the present study, the experiences of awareness were not severe enough to cause PTSD. Preoperative mental disorders may bear some relation to awareness during general anesthesia, as was also supposed in some previous studies . In the present study, depressive preoperative mental state seemed to be associated with intraoperative awareness. It may be that patients with certain psychological constitutions are more prone to at least expressing the experience of awareness during general anesthesia. Further studies are needed to evaluate these issues.
The incidence of dreaming in the present study was greater than that in the study of Liu et al.  but lower than that in earlier studies in which potent halogenated anesthetics were not used [9,17]. In the current study, most of the dreams were emotionally pleasant or neutral. Some of the patients who experienced unpleasant dreams were classified as having experienced possible (Group 1) intraoperative awareness. Women reported having experienced significantly more dreams than men. It seems important to follow the content of dreams in detail to uncover all cases of awareness and to report the sex distribution of the study population when reporting the incidence of perioperative dreaming.
It is possible that the performance of studies exploring the quality of anesthesia may cause change in the way anesthesiologists anesthetize their patients [5,16]. We made no effort to blind the anesthesiologists from the study. There was a significant increase over time in the use of isoflurane, which may have been caused by knowledge of the study. However, normal baseline incidence of awareness during anesthesia may only be detected by blinding the anesthesia providers from the study.
The rate of interviews was relatively low (54% of electively operated patients) in the present study. Low and variable interview rates seem to be a problem in this type of study, in which interviews of large patient populations are required . More disappointing was the low proportion of patients with awareness (26%) who were enrolled to the psychiatric interview. This small sample makes it impossible to draw any conclusions about the incidence and nature of possible psychological problems after awareness during anesthesia.
It was not possible to measure the end-tidal concentration of the inhaled anesthetics in every operating room. Therefore, we had to estimate the dose of the anesthetics from the inspiratory dial settings recorded on the anesthetic records. This carries the possibility of gross error, as it may be that the dial is set to a certain value, but the patient received no anesthetic because of a technical error (vaporizer empty, not properly fit on the machine, main switch in off position, etc.). Similar error is possible with the IV anesthetics, as we did not measure patients' blood concentrations.
In conclusion, a small incidence of awareness still complicates general anesthesia today. The present study shows that patients are predisposed to awareness during anesthesia by receiving doses of general anesthetics. In addition, the lack of amnestic premedication and noncontinuous dosing of anesthetics may increase the risk of this complication. Continuous attention should be paid to these factors because intraoperative awareness is still a problem, although its prevention should be feasible in most cases.
We are indebted to all the nurse anesthetists of the Paijat-Hame Central Hospital for performing the interviews, and to Mr. Jarmo Jarvinen, MSc, for help with the patient data management. We also express our gratitude to professor Dr. Seppo Sarna, PhD, for giving invaluable advice regarding the statistical methodology, and to biostatisticians Mr. Arttu Laine and Ms. Hanna Tolonen for performing the Poisson analysis.