ADJUNCTIVE NON-PHARMACOLOGICAL ANALGESIA FOR INVASIVE MEDICAL PROCEDURES: A RANDOMIZED TRIAL:

ELVIRA V. LANG

ADJUNCTIVE NON-PHARMACOLOGICAL ANALGESIA FOR INVASIVE MEDICAL PROCEDURES: A RANDOMIZED TRIAL: ELVIRA V. LANG

[Elvira V Lang (gave permission to use this article) {Dr Elvira V Lang, Department of Radiology, West Campus CC308, Beth Israel Deaconess Medical Center, 330 Brookline Avenue, Boston, MA 02215, USA(e-mail: elang@caregroup.harvard.edu)}, Eric G Benotsch, Lauri J Fick, Susan Lutgendorf, Michael L Berbaum, Kevin S Berbaum, Henrietta Logan, David Spiegel:

Summary: Background Non-pharmacological behavioral adjuncts have been suggested as efficient safe means in reducing discomfort and adverse effects during medical procedures. We tested this assumption for patients undergoing percutaneous vascular and renal procedures in a prospective, randomized, single- centre study.

Methods: 241 patients were randomized to receive intraoperatively standard care (n=79), structured attention (n=80), or self-hypnotic relaxation (n=82). All had access to patient-controlled intravenous analgesia with fentanyl and midazolam. Patients rated their pain and anxiety on 0B 10

scales before, every 15 min during and after the procedures.

Findings: Pain increased linearly with procedure time in the standard group (slope 0× 09 in pain score/15 min, p<0× 0001), and the attention group (slope 0× 04/15 min; p=0× 0425), but remained flat in the hypnosis group. Anxiety decreased over time in all three groups with slopes of 0× 04 (standard), 0× 07 (attention), and 0× 11 (hypnosis). Drug use in the

standard group (1× 9 units) was significantly higher than in the attention and hypnosis groups (0× 8 and 0× 9 units, respectively). One hypnosis patient became haemodynamically unstable compared with ten attention patients (p=0× 0041), and 12 standard patients (p=0× 0009). Procedure times were significantly shorter in the hypnosis group (61 min) than in the standard group (78 min, p=0× 0016) with procedure duration of the attention group in between (67 min).

Interpretation: Structured attention and self-hypnotic relaxation proved beneficial during invasive medical procedures. Hypnosis had more pronounced effects on pain and anxiety reduction, and is superior, in that it also improves haemodynamic stability. Lancet 2000; 355: 1486B 90

Beth Israel Deaconess Medical Center/Harvard Medical School (E V Lang MD); Department of Radiology, Boston, MA (E V Lang), University of Iowa, Iowa City, IA; Departments of Radiology (E V Lang, E G Benotsch PhD, L J Fick BS, K S Berbaum PhD), Psychology (S Lutendorf PhD), Division of Public Health Services and Research, University of Florida, Gainsville, FL (H Logan PhD); Institute for Social Science Research, University of Alabama, Tuscaloosa, AL (M L Berbaum PhD); and Stanford University, Department of Psychiatry and Behavioral Sciences, Stanford, CA (D Spiegel PhD), USA

Introduction: Minimally invasive, image-guided, percutaneous medical procedures increasingly replace open surgery. Technical refinement minimizes tissue injury and largely obviates the need for general anaesthesia, but patients may still experience distress, which can tax the coping mechanisms of even well-functioning individuals.1 Most physicians rely on intravenous conscious sedation with narcotics and sedatives to manage pain and anxiety.2 These drugs, however, can induce cardiovascular depression, hypoxia, apnoea, unconsciousness, and, rarely, death, even in dosages usually well tolerated.3,4 The operator typically has to weigh the risks of medically induced over sedation against the risks of uncontrolled discomfort and restlessness. An approach that provides comfort while reducing or eliminating the need for intravenous drugs is, therefore, highly desirable.

Biobehavioral A non-pharmacological@ analgesia in the form of imagery, relaxation training, and hypnosis has been used successfully to treat procedure pain.5B 10 Clinical practice guidelines for acute pain management, published by the US Public Health Service, mention relaxation exercises and cognitive approaches, but do not elaborate.2 Behavioral methods still need testing in larger clinical studies. To address this need, we designed a prospective randomized trial comparing the standard approach of intravenous conscious sedation alone with the adjunctive use of two behavioral non-pharmacological interventions: structured attention and self-hypnotic relaxation. We tested the hypothesis that adjunctive non-pharmacological analgesia would reduce patients= perceived pain and anxiety during interventional radiological procedures, reduce the amount of intravenous conscious sedation needed and make the procedure safer. Since operating teams (and hospital administrators) are very sensitive to factors that could prolong the patient= s stay in the procedure room, we also assessed how non-pharmacological analgesia adjuncts affect procedure time.

Methods: Selection and randomization of patients: The study was approved by the Institutional Board for Human Subjects Review. Eligible individuals were adults referred for percutaneous transcatheter diagnostic and therapeutic peripheral vascular and renal interventions, who were able and willing to give written informed consent. Exclusion criteria were severe chronic obstructive pulmonary disease, psychosis, intolerance of midazolam or fentanyl, pregnancy, inability to hear or understand English. After one of the operators had obtained informed consent for the planned invasive medical procedure, a research assistant asked the patient to participate in a research study to assess whether a relaxation exercise would enhance comfort during invasive procedures. Patients were told that the chance of having this relaxation exercise would be one in three, and that, irrespective of whether they would have this relaxation exercise or not, they would have access to as much medication for comfort as they wanted within safe limits. Consenting patients were then screened with the Mini Mental-State Exam.11 If they passed, patients were randomly assigned to received standard treatment (A standard group@ ), structured attention (A attention group@ ), or the self- hypnotic relaxation (A hypnosis group@ ).

Characteristic Standard Attention Hypnosis: group (n=79) group (n=80) group (n=82),  Age* 57 (18B 92) 57 (18B 84) 54 (19B 82),  Weight (kg)* 74 (44B 118) 80 (40B 143) 80 (45B 147) Male/female 36/43 40/40 38/44, Ethnicity, White 74 75 78, Black 5 5 3, Native American 0 0 1,

Disease categories: 1=benign, no threat to limb or life; 2=benign, threat to limb or organ, no threat to life; 3=malignant, 4=acutely life-threatening. Anaesthesia class=American Society of Anesthesiologists Physical Status Classification: 1=healthy patient, 2=mild systemic disease, 3=severe systemic disease, 4=threat to life.

Random numbers generated a sequence of group assignments for consecutive patients. Group assignments were contained on a card in a sealed envelope. Envelopes were not opened until the patients were about to undergo surgery. Group attribution was based on the intent to treat so that a patient assigned to a group remained in that group irrespective of their reaction to the test treatments.

Test treatments: All test treatments were done during surgery. Standard-group patients received care typical for the institution; they were attended to by the department= s special-procedure nurses, who were instructed to behave naturally, do their best to comfort the patient, but to abstain from induction of imagery and hypnosis. In the attention group, an additional provider displayed standardized structured attentive behavior. In the hypnosis group, the additional provider displayed the same structured attention as in the attention group and in addition gave standardized guidance to self-hypnotic relaxation. The additional provider wore scrubs like the regular team members and, throughout the procedure, sat close to the patient= s head separated from the imaging tower and operators by a mobile lead-glass shield. The surgical team worked on the opposite side of the imaging tower and behind another ceiling-mounted lead-glass shield, and could not hear what the provider said to the patient. There were no visual clues that would have permitted the operating team to distinguish between attention and hypnosis conditions.

Structured attentive behavior included eight key components and guidance to self-hypnotic relaxation an additional three. These 11 key components were standardized in a treatment manual.12 The components were: matching the patients= verbal and nonverbal communication patterns; attentive listening; provision of the perception of control (A Let us know at any time what we can do for you@ .); swift response to patient= s requests; encouragement; use of emotionally neutral descriptors (such as A What are you experiencing?@ , focus on a sensation of fullness, numbness, coolness, or warmth when painful stimuli were imminent); avoidance of negatively loaded suggestions (A How bad is your pain?@ , A You will feel a sting and burn now@ ); reading a standardized hypnotic induction script; addressing anxiety and worries according to the script, if needed. For the first 29 hypnosis patients, a script was used that included progressive muscle relaxation and having patients count backwards from 100. Because procedure interruptions, particularly at the time of counting, made it awkward for the providers to restore the patient= s state of concentration, we developed a script12 that allowed a rapid restoration of a focused state with ease. It was used for the final 53 hypnosis patients. Patients were instructed to roll their eyes upwards, close their eyes, breathe deeply, and concentrate on a sensation of floating. Self-generated imagery was used to help patients focus on a safe and pleasant experience in this relaxed state.

Training of the providers was done anew to reflect what an average competent healthcare worker could do with specific training, not what experts in the field could achieve. Four providers (a nurse, two medical students, and a psychology graduate student) were trained in the structured attention and hypnosis interventions by 24 h classroom instruction and roleplay, study of the treatment manual and video, supervised clinical practice, and a second workshop lasting 8 h. Interactions with patients were videotaped and used for feedback. The trainees did not participate in the study until they were able to execute reliably all of the key components. Psychologists and physicians supervised throughout the training and study.

Fidelity of treatment administration: Our procedure followed recommendations by Moncher and Prinz13 for achieving and monitoring fidelity of treatment administration. A standardized manual defined prescribed and proscribed techniques for each of the three treatments. Training in the intervention techniques was given to all providers. Adherence to the treatment protocol was ensured by analysis of videotaped sessions, consistent with previous research in health-care settings6 and with recommendations by reviewers in the field.14B 16 All interventional procedures were videotaped and 55 (23%) tapes were randomly selected and rated for adherence. Two research assistants who were not otherwise involved in the study independently rated the videotapes with an inter-rater reliability of the manipulation check17 of 0× 81. Analysis of videotapes has been shown to be successful for assessing treatment fidelity.13

Pain, anxiety, and drug use: Pain and anxiety were assessed by self-certification on a scale of 0B 10 before surgery and every 15 min during it. Since dimmed lights and immobilization of patients in the radiography equipment made use of visual scales cumbersome, we used verbal scales with 0=no pain and 10=worst pain imaginable, and 0=no anxiety and 10=terrified. Such verbal pain scales have been validated for clinical research.18,19 Reliability and validity of the verbally- administered anxiety rating has been shown previously.20

To reduce the possibility of unblinded experimenter bias and to ensure that patients in the three treatment groups had the same access to drugs, patient-controlled analgesia/sedation (PCA) was used. PCA is well suited for acute pain management during and after medical procedures and is felt to enhance comfort while providing patients with a means of control.21,22 In a pilot trial before this study, use of a PCA pump was tested as a means of further blinding, but was found to be potentially hazardous. Drug-induced cardiorespiratory emergencies are treated differently from those of other origins, and rapid knowledge of the drug history becomes important. Entering the recording mechanism of a PCA pump, however, can cause undue delay. Therefore, patients were given a button to press to alert the attending nurse, rather than a machine, to deliver drugs through an indwelling intravenous access. The dose regimen was chosen within the standard of care for

interventional procedures2,23 as 0× 5 mg of midazolam plus 25 g of fentanyl per request up to four times with a lockout time of 5 min, then with a lockout time of 15 min. Medication was withheld during the lockout times and if systolic blood pressure was below 89 mm Hg, oxygen saturation fell below 89%, or the patient developed slurred speech or became difficult to arouse. In rare cases, fentanyl and midazolam were administered without the patient= s request for reasons of safety, such as when systolic blood pressure exceeded 180 mm Hg and did not normalize after 20 mg: 366 patients eligible, 96 did not consent, 13 did not pass mental screening, 16 had procedures cancelled, 241 randomized, 79 received 80 had test 82 received full control intervention test intervention intervention and A attention control@ and were were analyzed and were analyzed for for primary for primary endpoint endpoint end point.

Procedural safety: Patients were monitored by ECG and pulse oximetry continuously during and after the procedure, and by automated blood-pressure measurements at least every 10 min during the procedure, and at least every 15 min during recovery. Haemodynamic instability was defined as any cardiovascular reaction that needed treatment or interruption of the procedure. Adverse effects in each of the three treatment groups were counted and included: oxygen desaturation less than 89%, need for placing oxygen tubing, prolonged hypoxaemia; haemodynamic instabilities such as prolonged new- onset bradycardia, new-onset hypertension, hypotension, or cardiac arrhythmia; rebleeding from puncture sites; oversedation with somnolence, unresponsiveness, disorientation; vomiting; patient behaviour prone to distract the operators such as crying, sobbing, talking to the operator, complaining loudly, and grabbing personnel.

Patients: The study was done over 11 months in the interventional radiology division of a single university medical centre with a mix of about 55% inpatients and 45% day patients. 336 consecutive patients, who presented during the regular working hours of the research assistant, were invited to enrol in this study. 66 declined to participate. Among the 270 consenting patients, 13 did not pass the Mini Mental-State Exam and 16 had their procedures cancelled. The study thus comprised 241 patients. Ages ranged from 18 to 92 years (median 56 years); 114/241 (47%) were men. Randomisation resulted in homogeneous groups of patients; there were no significant differences among the three groups in key characteristics including baseline pain and anxiety levels, type and technical complexity of procedures, disease status, and anaesthesia class (table 1). No patient withdrew from the study during the time of observation.

Statistical analysis: Effects of treatment on total units of drugs requested and administered, and on total procedure duration were assessed by univariate ANOVA with a between-patient factor for treatment group (standard, attention, hypnosis).24 Before analysis, logarithmic transformations were applied to remove skewness from the data (ln[x+1], or ln[x] if x could not be 0); however, all results were presented in terms of the original scales.24 Tukey tests were done to determine the pairs of treatment means that were significantly different from one another.24

The repeated-measures analysis of pain responses was designed to characterize and compare trends in pain ratings for the three treatment conditions over time.25,26 The analysis used reports from as many as 13 successive 15 min intervals from 241 patients; the average number of reports from patients was 4× 7 (median 3). The dependent variable for this analysis was ln(pain score+1) to correct skewness; residuals appeared normally distributed and no outliers were identified. For descriptive flexibility, the statistical model included separate parameters for intercepts, linear-order and higher-order trends. These latter quadratic and cubic trend components were not significant and were excluded from the model reported here. Correlations among residuals differed according to the time between observations, declining with increasing separation and reaching negligible levels after six intervals; there was a slight decrease in error variability in later intervals, but not enough to model. These linear mixed models were estimated by restricted maximum likelihood in BMDP,

version 5, which provides unbiased estimates of the intercepts and slopes;27 comparisons among slopes were done by two-tailed Wald statistics. A similar analysis was done for anxiety.

The frequencies of eight types of adverse effects in the three treatment groups were compared pair wise. One-tailed Fisher exact tests28 tested the predictions that adverse effects would be more frequent in the standard group than either the attention or hypnosis group and more frequent in the attention than hypnosis group. Since there were no differences in the outcomes for patients who had hypnosis script I and and those who had script II, the results are given for the entire hypnosis group.

Results: The trial profile is shown in figure 1. Figure 2 presents the number of patients remaining in their procedure as a function of duration (15-min intervals). The figure suggests procedures in the hypnosis group needed less time than procedures in the attention group, which in turn needed less time than those in the standard group. ANOVA confirmed the difference in groups= means (F2238=6× 60, p=0× 0016). Average procedure duration was significantly shorter in the hypnosis group than in the standard group (61 vs 78 min). Procedure duration for the attention group was between, but not significantly different from, that of the other two groups (mean=67 min).

Groups differed from respect to medication received (F2238=12× 49, p<0× 0001). Patients in the standard group had significantly higher drug use (1× 8 drug units requested, 1× 9 drug unitsreceived) than those in the attention group (0× 8 units requested and received) and in the hypnosis group (0× 9 units requested and received). Drug deliveries exceeded requests in ten patients (total 14 drug units) in the standard group, eight patients (total 9× 5 drug units) in the attention group, and three patients (total 2 drug units) in the hypnosis groups. Drug deliveries were fewer than

requested in five patients (6× 5 drug units) in the standard group, in none in the attention group, and in one patient (3 drug units) in the hypnosis group.

Figures 3 and 4 present the average pain and anxiety scores as a function of procedure duration. Initially and for several 15-min periods, treatment groups did not differ in their levels of pain and anxiety. Subsequently, differences emerged with succeeding intervals. After an hour, absolute pain and anxiety were highest in the standard group, intermediate in the attention group, and lowest in the hypnosis group, and these relative positions stayed the same. All procedures for patients in the hypnosis group were completed by the tenth interval where that curve ends; the curves for the other conditions become somewhat erratic in the final intervals owing to the small number of cases (figure 2).

The repeated measures analysis of pain (raw data shown in figure 3) showed that pain increased linearly with procedure time in the standard group (slope=0× 09 increase in pain score per 15 min, p<0× 0001) and in the attention group (slope=0× 04, p=0× 0425, but not in the hypnosis

group (slope= 0× 03, p=0× 234, a decrease in pain score over time, but not significantly different from 0, and hence interpreted as no change in pain over time, or flat). The flat trend of predicted pain as a function of procedure duration in the hypnosis group was significantly less than the positive trends in the standard (p<0× 0001) and attention groups (p=0× 0259). The difference in the trends of predicted pain medication. The trend towards less pain and anxiety over time, found with structured attention, reached significance with the addition of hypnosis. Whether less anxiety lessened the pain or vice versa, or whether anxiety and pain experience responded individually or interactively to the non-pharmacological intervention, remains unclear.

Pain perception increased with procedure time under standard care conditions. Similarly, Schutz and colleagues identified length of procedure as a negative predictor of satisfaction with pain management in a group of patients undergoing colonoscopy with intravenous conscious sedation.29 These findings are consistent with laboratory research showing that exposure to acute pain makes

individuals more attentive to external cues, such that they report increasing pain over time even in the absence of a painful stimulus.30 Pain scores in the standard group increased while anxiety scores decreased from baseline over time. This further complicates speculation about the impact of anxiety reduction on pain perception.31

Lower use of analgesics and anti-anxiety medication with adjunctive hypnosis compared with a control condition was also reported in clinical studies in which physicians, not the patients, controlled intravenous conscious sedation during invasive medical procedures.6,8,9,32 Reduced drug use in the attention and hypnosis groups resulted in a strong trend towards fewer overall and severe episodes of oxygen desaturation (significant for the

comparison of standard and attention groups); only in the hypnosis group were there fewer episodes of haemodynamic instability. Thus, hypnosis has beneficial effects that cannot be explained solely by reduced drug use. These findings confirm the results of smaller scale studies showing areduction of drug use and improved haemodynamic stability when adjunct hypnosis was used during invasive medical procedures.9,32

Self-hypnotic relaxation saved 17 min of theatre time despite the time invested in the hypnotic induction compared with standard care. Structured attention alone showed a strong trend towards savings (11 min) that might have reached significance in a study with more patients. It is likely that these savings resulted from lower frequencies of events diverting the surgical team= s attention. When we took into account prolonged hypoxaemia, haemodynamic instability,

rebleeding, oversedation, and vomiting (table 2) as such distractions, 34 events were counted in the standard group, 22 in the attention group, and eight in the hypnosis group.

The additional provider in the attention and hypnosis groups might possibly have changed patients= demand characteristics, thus contributing to the lower self-reports of pain and anxiety and fewer drug requests. Demand characteristics alone, however, would not explain why self- hypnotic relaxation had more profound effects on pain and anxiety reduction than structured attention. In particular, the greater haemodynamic stability in the hypnosis group would be difficult to explain based on a biased response. Since drug deliveries were nearly identical in the attention and hypnosis groups, medication can be excluded as a confounding factor.

Contributors: The original hypothesis and concept of the study was generated by E Lang. The experimental plan was elaborated by E Lang, E Benotsch, Susan Lutgendorf, Kevin Berbaum, Henrietta Logan, and David Spiegel. The hypnotic intervention use in the study was formulated for the study needs by D Spiegel, E Lang, Lauri Fick, and Susan Lutgendorf. The components of the study pertaining to psychological parameters and adherence checks were designed and supervised by E Benotsch, S Lutgendorf, and H Logan. The intervention was applied by Lauri Fick and E Benotsch. E Lang and L Fick were responsible for patient recruitment. K Berbaum and M Berbaum did the statistical analyses and wrote the statistical components of the paper. All authors were involved in writing and reviewing the paper.

Acknowledgments: Supported by grant RO1-MH56274 cofunded by the National Institute of Mental Health and The Office for Alternative Medicine. We greatly appreciate the help received throughout this study by Eleanor Laser, Susan Small, Keela Herr, and Paula Mobily.

References:

1 Horne DJ, Vatmanidis P, Careri A. Preparing patients for invasive medical procedures and surgical procedures. 1: Adding behavioral and cognitive interventions. Behav Med 1994; 20: 5B 13.

2 Acute pain management guideline panel. Acute pain management: operative or medical procedures and trauma (Clinical practice guideline. AHCPR Pub. No.92-0032) Rockville, MD: US Department of Health and Human Services, 1992.

3 Arrowsmith JB, Gerstman BB, Fleischner DE, Benjamin SB. Results from the American Society for Gastrointestinal Endoscopy/US Food and Drug Administration collaborative study on complication rates and drug use during gastrointestinal endoscopy. Gastrointest Endosc 1991; 37: 421B 27.

4 Bailey PL, Pace NL, Ashburn MA, Moll JWB, Easat KA, Stanley TA. Frequent hypoxia and apnea after sedation with midazolam and fentanyl. Anesthesiology 1990; 73: 826B 30.

5 Smith JT, Barabasz A, Barabasz M. Comparison of hypnosis and distraction in severely ill children undergoing painful medical procedures. J Counseling Psychol 1996; 43: 187B 95.

6 Weinstein EJ, Au PK. Use of hypnosis before and during angioplasty. Am J Clin Hypn 1991; 34: 29B 37.

7 Patterson DR, Everett JJ, Burns GL, Marvin JA. Hypnosis for the treatment of burn pain. J Cons Clin Psychol 1992; 5: 713B 17.

8 Faymonville ME, Mambourg PH, Joris J, et al. Psychological approaches during conscious sedationC hypnosis versus stress reducing strategies: a prospective randomized study. Pain 1997; 73: 361B 67.

9 Lang EV, Joyce JS, Spiegel D, Hamilton D, Lee K. Self-hypnotic relaxation during interventional radiological procedures: effects on pain perception and intravenous drug use. Int J Exp Clin Hyp 1996; 44: 106B 19.

10 John ME, Parrino JP. Practical hypnotic suggestion in ophthalmic surgery. Am J Ophthalmol 1983; 96: 540B 42.

11 Folstein MF, Folstein SE, McHugh PR. A Mini Mental State@ : a practical method for grading the cognitive state of patients for the clinician. J Psychiatric Res 1975; 12: 189B 98.

12 Lang EV, Spiegel D, Lutgendorf S, et al. Nonpharmacological analgesia and anxiolysis for interventional radiological procedures. Sem Intervent Radiol 1999; 16: 113B 23.

13 Moncher FJ, Prinz RJ. Treatment fidelity in outcome studies. Clin Psychol Rev 1991; 11: 247B 66.

14 Kazdin AE. Methods of psychotherapy research: research and practice. New York: Oxford University Press, 1995.

15 Ludwick-Rosenthal R, Neufeld RWJ. Stress management during noxious medical procedures: an evaluative review of outcome studies. Psychol Bull 1988; 104: 326B 42.

16 Waltz J, Addis ME, Koerner K, Jacobsen NS. Testing the integrity of a psychotherapy protocol: assessment of adherence and competence. J Consult Clin Psychol 1993; 61: 620B 30.

17 Shrout PE, Fleiss JL. Intraclass correlations: use in assessing rater reliability. Psychol Bull 1979; 86: 420B 28.

18 Paice JA, Cohen FL. Validity of a verbally administered numeric rating scale to measure cancer pain intensity. Cancer Nurs 1997;

20: 88B 93.

19 Murphy D, McDonald A, Power A, Unwin A, McSullivan R.

Measurement of pain: a comparison of the visual analogue with a nonvisual analogue scale. J Clin Pain 1988; 3: 197B 99.

20 Benotsch EG, Lang EV, Lutgendorf SK, Fick L. Rapid anxiety assessment in medical patients. Ann Behav Med 1998; 20: 138. 21 Duggleby W, Lander J. Patient-controlled analgesia for older adults. Clin Nurs Res 1992; 1: 107B 13.

22 Schelling G, Weber W, Mendl G, Braun H, Cullman H. Patient controlled analgesia for shock wave lithotripsy: the effect of self- administered alfentanil on pain intensity and drug requirement. J Urol 1996; 155: 43B 47.

23 Lang EV, Chen F, Fick LJ, Berbaum KS. Determinants of intravenous conscious sedation for arteriography. J Vasc Interv Radiol 1998; 9: 407B 12.

24 Kirk RE. Experimental design: procedures for the behavioral sciences (3rd edn). Pacific Grove, Calif: Brooks/Cole Publishing Company 1995: 72B 112, 144B 54, 164B 208.

25 Jennrich RI, Schluchter MD. Unbalanced repeated-measures models with structured covariance matrices. Biometrics 1986; 42: 805B 20. 26 Laird NM, Ware JH. Random effects models for longitudinal data. Biometrics 1982; 38: 963B 74.

27 Dixon WJ. BMDP statistical software manual, volume 2 (to accompany release 7). Berkeley: University of California Press, 1992: 1311B 52.

28 Siegal S, Castellan NJ Jr. Nonparametric statistics for the behavioral sciences. 2nd edn. New York: McGraw-Hill, 1988: 103B 11.

29 Schutz SM, Lee JG, Schmitt CM, Almon M, Baille J. Clues to patient dissatisfaction with conscious sedation for colonoscopy. Am J Gastroenterol 1994; 89: 1476B 79.

30 Bayer TL, Coverdale JH, Chiang E, Bangs M. The role of prior pain experience and expectancy in psychologically and physically induced pain. Pain 1998; 74: 327B 31.

31 Barber TX. The effects of A hypnosis@ on pain. A critical review of experimental and clinical findings. Psychosom Med 1963; 25: 303B 33.

32 Faymonville ME, Fissette J, Mambourg PH, et al. Hypnosis as adjunct in conscious sedation for plastic surgery. Reg Anesth 1995; 20: 145B 51.