47

Pain. 43 (1990) 47-55 Elsevier

PAIN 01693

Profiles of opioid analgesia in humans after intravenous bolus administration: alfentanil, fentanyl and morphine compared on experimental pain C.R.

Chapman

a,b, H.F.

Hill

a, L. Saeger

b and J. Gavrin

a.b

’ Clinical Research Division, Fred Hutchinson Cancer Research Center, Seattle, WA 98104 (U.S.A.), and ’ Department of Anesthesiology, (Received

University of Washington School of Medrcine, Seattle, WA 98195 (U.S.A.)

18 October

1989, revision received 16 May 1990, accepted

18 June 1990)

This report examines the relationship of plasma drug concentration to analgesic effect following bolus doses of Summary alfentanil, fentanyl and morphine and assesses individual differences in analgesic response among volunteers. We predicted that the 3 opioids would yield disparate analgesic profiles because their physicochemical and pharmacokinetic characteristics differ. Ten healthy volunteers received intravenous bolus doses of either alfentanil, fentanyl, morphine or normal saline on different days. We stimulated their teeth electrically and measured brain evoked potential (EP) and pain report (PR) repeatedly over 2 h to assess analgesic effect. Concurrently, we drew 18 blood samples to assess opioid plasma concentrations during the test period. The relationship between opioid plasma concentration and analgesic effect was well defined for alfentanil but ambiguous for morphine. Fentanyl exhibited a marked hysteresis. We observed noteworthy individual differences in analgesic response with all 3 drugs but these differences were greatest for morphine and least for alfentanil. Inter- and intrasubject variability in analgesic response across drugs is related to the physicochemical properties of the drugs tested. Key words:

Alfentanil;

Fentanyl;

Morphine;

Concentration-effect

Introduction Systemically administered opioids differ markedly from one another in rate of onset, analgesic potency and duration of action. Because the drugs are infrequently studied in humans apart from clinical settings, their differences in analgesic effect remain imprecisely documented and clinicians rarely use them to maximal advantage. In this report and its 2 companions, we quantify and contrast the analgesic effects of opioids in human.

Correspondence to: Harlan F. Hill, Ph.D., Pain and Toxicity Research Program AB-122, Fred Hutchinson Cancer Research Center, 1124 Columbia Street, Seattle, WA 98104, U.S.A.

0304-3959/90/$03.50

0 1990 Elsevier Science Publishers

relationship;

Opioid

analgesia;

Experimental

pain

We compare 3 mu receptor-selective opioid agonists differing in physicochemical characteristics: fentanyl, alfentanil and morphine. This, the first of 3 reports, investigates how the physicochemical properties of an opioid determine its analgesic profile after intravenous bolus administration. In addition, it examines the extent of variability in analgesic response across subjects. The second report illustrates how pharmacokinetic tailoring procedures can fit drugs individually to experimental subjects and achieve accurate steady-state plasma opioid concentrations. The final paper shows that manipulating steady-state plasma concentrations achieves strong experimental control of analgesic effect. Taken together, these reports characterize the therapeutic windows

B.V. (Biomedical

Division)

48

of the 3 drugs and demonstrate the feasibility and benefit of stabilizing plasma concentration in human laboratory research. This report targets 2 problem areas: the relationship of changing opioid plasma concentration to analgesic effect and the influence of individual differences in analgesic response among subjects. We address the first problem by comparing drugs with high (fentanyl), medium (alfentanil), and low (morphine) lipid solubility. The opioid plasma concentration-effect relationship is elusive because analgesia lags behind plasma concentration to varying degrees with different drugs. The extent of such lag, termed hysteresis, depends upon the physicochemical characteristics of the drug, which, along with the fraction of unionized free drug in plasma, determine the rate of diffusion from plasma to sites of action at opioid receptors in the central nervous system (CNS). We predicted that alfentanil, fentanyl and morphine would yield markedly different profiles of analgesic effect over a 2 h period following intravenous boluses at approximately equianalgesic doses. The physicochemical differences among the drugs necessarily restricted our ability to detect and quantify analgesic effects following an intravenous boius. We further predicted that the differing hysteresis of fentanyl, alfentanil and morphine would result in disparate qualities of analgesic assessment. Drugs with less hysteresis should yield clearer profiles of analgesic effect. Similarly, drugs with comparatively slow equilibration between blood and brain such as morphine [21], should yield an ambiguous analgesic time course. We therefore predicted that alfentanil would yield a more distinct pattern of analgesic response than either fentanyl or morphine. Our second purpose in this study was to assess the extent of individual differences in analgesic response to opioids under conditions of changing plasma concentration. Pharmacokinetic and pharmacodynamic variation among subjects can obscure analgesic effect in group data characterized by means. Moreover, the influence of these variations may be greater with some drugs than with others because of differing physicochemical charWe predicted that individual difacteristics.

ferences would be least for alfentanil because of its minimal hysteresis and rapid CNS penetration and most marked for morphine because of its slower CNS equilibration.

Methods Overview Volunteers experienced repetitive electrical tooth pulp stimulation at baseline and after drug administration. We delivered opioids by intravenous bolus and assessed the changing plasma concentration of drug over 2 h. Drug-induced decreases in pain report (PR) and dental evoked potential (EP; peak-to-peak amplitude at 150-250 msec) quantified analgesia. We used a repeated measures design: each subject was tested repeatedly on each of 4 different days. Subjects We enrolled 10 male volunteers, 21-40 years of age, within +lO% of normal body weight for height. Body weight ranged from 61.4 to 90.0 kg with a mean of 74.6 kg (morphine, 74.5; fentanyl, 74.8; alfentanil, 74.6; saline, 74.4 kg). All subjects signed informed consent as approved by the Fred Hutchinson Cancer Research Center Institutional Review Board. Drug administration Each subject participated in 4 experimental sessions during which he received bolus intravenous injections of either morphine (142 pg/kg), fentanyl (2 pg/kg), alfentanil (15 pg/kg) or normal saline (6 ml). We derived the doses of opioids from animal and clinical studies that indicate a potency ratio of about 100: 10 : 1.5 for morphine, alfentanil and fentanyl[2,13]. The order of drug administration was counterbalanced. At least 7 days separated any 2 sessions for each subject. Dental stimulation We employed our standard method for electrical dental stimulation to create safe and reliable experimental pain [15]. Several previous studies support the applicability of this testing model for

49

the study of opioid and other analgesic states [4-6,8,11]. The subject held a probe with a conductive rubber tip (cathode) at the edge of an upper central incisor. We taped the anode, a Beckman hollow disk electrode filled with conductive gel, ipsilaterally beside the nasal ala. A Grass Model 44 stimulator with stimulus isolation and constant current units delivered electrical pulses. Each stimulus consisted of a 5 msec square wave pulse of up to 80 PA. Before baseline (preinjection) measurement, we repeated stimuli of increasing intensity at 5 see intervals and identified a stimulus that the subject consistently rated as strong pain. That stimulus intensity remained invariant for the remainder of the experimental session. The computer varied the interstimulus interval randomly from 4 to 6 set during baseline and postinjection measurement periods.

peak-picking routine to identify major waveform peaks. To procure EPs over blocks of 50 trials, the summation averaged 50 samples of electrical brain activity extending from 100 msec prestimulus to 400 msec poststimulus. We obtained EPs for three blocks before drug administration and for 18 blocks over 120 min following drug injection.

Evoked potential data collection At the beginning of each session, we attached recording electrodes to the subject’s scalp with collodion and to the face with adhesive tape. We placed a ground electrode at the center of the forehead just below FPz. To detect eye blinks and ocular rotation, we placed cup electrodes above and below the external canthus of the left eye and filled each with conductive gel to provide low resistance (< 5 kSZ). An LSI-11/23 microcomputer system controlled stimulation and event recording. The computer sampled stimulus related brain electrical activity at vertex with reference to linked ears. Grass Model 12 amplifiers amplified and filtered signals to a band~d~ of 0.05-300 Hz before sending them to the computer. Channel amplifications of 50,000 insured that eye blinks and large horizontal or vertical eye movements would saturate the amplifier. When the computer detected activity in the eye channel on a given trial, it aborted data collection and repeated the trial. The computer digitized and stored amplified signals on a disk with periodic dumping to tape. We processed signals off line, including signal averaging and digital filtering to an upper cut-off of 50 Hz, using a graphics display with a software

Procedure After attachment of recording electrodes and measurement of body weight, the subject sat in a hospital bed inside a sound attenuate isolation room. We placed a 20-gauge intravenous catheter in a branch of the basilic vein of his right arm for blood sampling. A similarly placed contralateral catheter permitted drug or saline injection, Each day, when the ‘strong pain level had been established, we began baseline testing. The computer delivered 3 blocks of 50 stimuli each to the subject for measurement of baseline EPs and PRs. We separated stimulation blocks by 30-60 sec. During these interludes the subject repositioned the probe electrode, when necessary, and provided the PR for the previous block of trials. At the end of the 20 min baseline period, we administered either saline or one of the 3 opioids intravenously over 60 sec. Immediately following the injection, we repeated the stimulation and recording sequence. The sequence recurred every 20 min for 2 h after injection.

Pain report We collected PR data by giving the subject a 6-point pain intensity rating scale at the end of each subset of 50 trials. The categories were: (0) very faint sensation, (1) very faint pain, (2) faint pain, (3) mild pain, (4) moderate pain, and (5) strong pain. We have used this scale for over a decade in both dental EP and sensory decision theory studies. It has proven sensitive to both gradations in dental stimulation intensity and various analgesic interventions [7].

Plasma drug concentrations We drew 5 ml samples of blood into heparinized syringes just before the drug or saline injection and at the following times after injection: 1, 2, 3,

4, 5. 7, 10. 15, 20, 30, 45, 60, 90. 120. 180. 240 and 300 min. We immediately placed each sample on ice until the end of the test session. After completing testing we separated plasma from each sample and froze it at -60°C for later assay. We measured concentrations of morphine in plasma with a modification of the method by Todd et al. [24]. This method uses high performance liquid chromatography with electrochemical detection and nalorphine as an internal standard. The detection limit of the method was 1 ng/ml with coefficients of variation of 6.9%, 5.3% and 3.9% at 5, 25 and 100 ng/ml. respectively. Alfentanil and fentanyl plasma concentrations were measured by separate radioimmunoassays for each drug [ 17,181 with sensitivities of about 1 ng/ml for alfentanii and 0.1 ng/ml for fentanyl. C’onc~ordunce of effect curiuhles und plusmu concentrutiofl We could not synchronize time of data collection. opioid plasma concentration. and effect perfectly across subjects and sessions. Therefore, we recorded the time of each testing block at the midpoint of the EP data collection. Block data collection times varied across individuals and within individuals across test sessions. The standard deviation across all sessions was 1.6 min for the first block, less than 1 min for 2 of the following blocks, and maximal at the last block where it reached 4.6 min. We adjusted the data to maximize correspondence in time among measures. Using the exponential concentration versus time curves for each subject, we matched plasma opioid concentration to effect measurement at all time points. This allowed us to estimate the plasma concentration of drug corresponding to midpoints of the 1X testing blocks for each subject in each session. Correction of evoked potential for huhituution Habituation occurs regularly in EP data regardless of stimulus modality [1,3,19,20]. In boius studies where we must stimulate the subject at a rapid rate, the peak-to-peak amplitude of the dental EP typically habituates [ll]. That is. EP amplitude gradually diminishes during the experimental session unless there are large intervals (several

minutes) of rest between trial blocks. Dental pain ratings do not habituate to the stimulation pattern used in our studies, a phenomenon consistent with the observation that nociceptors do not habituate to repeated stimulation in contrast with other sensory end organs [22]. This suggests that habituation in EP data is a central phenomenon limited to the EP and strictly independent of analgesia. To account for the possible confounding of analgesia and habituation in the EP data, WC corrected each subject’s EP data for the drug sessions using the habituation pattern he demonstrated when given saline. We fitted an EP time effect curve for saline and subtracted this from the corresponding time-effect data for each of the subject’s opioid sessions. This procedure. which preserved the other individual differences among subjects. removed the effect of habituation from the EP data collected during each of the drug xessions. After this adjustment the EP data were comparable to the PR scores, which did not show habituation in the saline condition. Baseline EP amplitude varied across days for some subjects. Therefore, we calculated drug effect magnitude for each subject on each day as percent change from the baseline EP amplitude. To facilitate visual comparison, we also transformed the PR data to percent change from baseline. This transformation preserved individual differences in analgesic response but removed within subject baseline variation over days. We examined correlations of the transformed PR and EP amplitude scores with plasma concentrations of the .i opioids from 0.~ 120 min after drug administration. Results

Three-dimensional effect profiles The 6 panels in Fig. 1 display the mean EP and PR scores for the 10 subjects for each of the drugs. These plots demonstrate the recovery of each of the 2 effect variables following bolus injection. Comparison of the 3 drugs across the floors of the boxes reveals the expected exponential pattern of change in plasma concentration of drug over time: a gradual elimination phase follows a rapid distributional phase. The initial decline in plasma

ALFENTANIL. EP

ALFENTANIL

FENTANYL

EP

MORPHINE EP

PR

Fig. 1. Three-dimensional profiles of analgesic effect following bolus injections. The 6 plots represent the effects of 3 drugs on 2 effect variables: evoked potential (EP) amplitude and pain report (PR). Each datum is the mean for 10 subjects. The 18 data points represent the data averaged over subjects for the 18 blocks of trials that followed injection. Baseline scores are not displayed. Mean times in min following intravenous bohts administration appear on the abscissas of the floors; the mean plasma concentrations are shown on the ordinates of the floors. The floor of each box therefore reveals the pattern of decreasing plasma concentration over 2 h following intravenous bolus of the drug. The z-axis, and consequently the height of the stem descending from each datum, indicates the magnitude of each effect variable as time and plasma concentration changed. On the back wall of each box, which corresponds to zero plasma concentration, the data for the saline session appear as triangles. The saline data permit visual assessment of the changes resulting from drug administration. Prebolus baseline means are represented as zero on the z-axis; consequently. the saline means vary around zero.

concentration (distribution phase) was less pronounced for alfentanil than for either fentanyl or morphine. In general, mean PR data were less variable than EP scores. Both variables responded to the changing (declining) plasma concentrations of fentanyl and alfentanil after the bolus doses. Al-

fentanil rapidly decreased EP amplitude and PR with a peak effect at about 2 min after injection. The duration of EP and PR effects was about 20 min. Fentanyl also decreased EP amplitude and PR rapidly, manifesting a longer duration of effects than alfentanil. Compared to the saline data, morphine clearly affected both the EP and PR,

52

but the effect was poorly defined. We perceive no trend in relationship between plasma morphine concentration and analgesic effect across time after the bolus dose.

testing the 3-way interaction of DRUG x LCON X MIN. The third model provided the key test since the gain in explained variance associated with the 3-way interaction determined whether or not the time-concentration-effect profile differed significantly across the 3 drugs. Table I summarizes the outcomes for the three models, applied separately to the 2 effect variables. It presents the multiple regression coefficient (R),the standard error of estimate, the significance level of the regression, and the ‘explained’ variance (R').R2 indicates the percent of variance in the effect variable accounted for by the predictors. The effect of increasing model complexity was greater for PR than for EP data. The explained variance increased 0,017 from the second to the third model for the PR (P -c0.01) but only 0.006 for the EP (not significant). Thus, the differences across the drugs in time-plasma con~entration-effect profile were sig~ficant for the PR but not for the EP. A power analysis of the increase in explained variance from the second to the third regression model revealed that power was 0.91 for PR but only 0.47 for EP. Doubling the sample size to n = 20 would have increased power to 0.76 for the EP. Thus, small sample size may account for our failure to find an EP 3-way interaction effect in the EP results.

Statistical analysis of profiles We performed separate multiple regression analyses to compare response profiles across drugs and to assess the importance of individual differences. The effect variables, EP and PR, were the criteria. The predictors were drugs (DRUG), log plasma concentration (LCON), time (MIN), and the log concentration x time interaction (LCON X MIN). Drugs were coded as dummy variables [9,10]. To meet the statistical assumptions for regression analysis, we transformed plasma concentration scores for each subject and each drug into percent of highest value. This made the plasma concentration changes comparable across drugs. We further transformed the individual plasma concentration scores loga~th~cally. This transformation allowed the complex and inherently nonlinear plasma concentration by time relationship to approximate linear trend. We undertook the first analysis to determine whether the different drugs produced significantly different profiles (patterns} of analgesic effect over time. To achieve this, we sequentially tested three multivariate regression models of increasing complexity using the entire data set. The first model tested the effects of DRUG, MIN and LCON, omitting all interaction effects. The second model extended the first by including the interaction of LCON x MIN. The third extended the second by TABLE

Individual differences By using 2 regression models (with and without representation of subjects), we were able to assess the influence of individual differences in analgesic effects. Fig. 2 compares the magnitude of R2 for 2

I

MULTIPLE

REGRESSION

MODELS

OF INCREASING

COMPLEXITY ~..

R

S.E.

P

R"

MIN X LCON

0.187 0.214

+ 0.214 + 0.213

< 0.001 < O.oool

0.035 0.046

DRUG, MIN, LCON, MIN x LCON, DRUG x MIN x LCON

0.229

F0.213

< 0.0001

0.052

DRUG, DRUG,

MIN X LCON

0.328 0.330

f 0.943 + 0.943

< O.oool i 0.0001

0.108 0.109

DRUG, MIN, LCON, MIN x LCON. DRUG x MIN x LCON

0.354

f 0.936

i 0.0001

0.126

Indicators EP Model 1 Model 2 Model 3 PR Model 1 Model 2 Model 3

DRUG, DRUG,

MIN. LCON MIN, LCON,

MIN, LCON MIN, LCON,

53

R2 100% EVOKED

PAIN REPORT

POTENTIAL 80% 60%

Alfentonil

Fentonyl I

Model A

659

Morphine

Alfentanil

Fentanyl I

Model B

Model A

m

Morphine Model B

Fig. 2. Comparison of explained variance for 2 multiple regression models. We compared the results of 2 regression models in order to assess the importance of individual differences. EP and PR were criterion variables. The predictors were log plasma concentration (LCON), time after bolus injection (MIN) and their interaction, LCON x MIN. The plots contrast the outcomes for each drug and each effect variable by model. The left panel depicts R2 for EP and the right for PR. Model A ignored individual differences among subjects: criterion = LCON + MIN + LCON x MIN. Model B took the individual differences into account through the use of n - 1 dummy variables (Sl . S9): criterion = LCON + MIN + Sl S9 + LCON X MIN.

types of multivariate regression analyses performed upon data grouped by drug. Because we were repeatedly analyzing data from the same subjects and therefore increasing the likelihood of a type 1 error, we restricted statistical significance to probabilities smaller than P = 0.01. For Model A alfentanil and fentanyl yielded significant multiple R values for both EP and PR but morphine failed to achieve statistical significance with either measure. For Model B all multiple regressions were statistically significant. The small explained variances for Model A suggest that there were noteworthy individual differences across subjects. Alfentanil yielded the greatest explained variance in the absence of control for individual differences. The striking gain in explained variance from Model A to Model B confirms the influence of individual differences. R2 is markedly larger for all of the drugs. This increment is particularly salient for morphine for which R2 increased more than lo-fold. The regression analyses performed for Fig. 2 included the LCON X MIN interaction factor, which represents the relationship between changing plasma concentration over time and analgesic

effect. This factor was a significant predictor at P < 0.0001 for alfentanil for both effect variables with both models. It failed to achieve statistical significance for either fentanyl or morphine with either effect variable in either model. This observation suggests that fentanyl’s hysteresis and morphine’s comparatively slow CNS equilibration markedly complicate the relationship between brain effect and plasma drug concentration when the latter is unstable.

Discussion As predicted, our ability to characterize opioid analgesia during changing plasma concentrations differed with drug. We ascribe this primarily to variations across drugs in physicochemical characteristics. Alfentanil yielded good measurement of analgesic effect, fentanyl acceptable measurement, and morphine poor measurement. High lipid solubility is clearly not the most desirable factor for a predictable opioid analgesic since fentanyl did not yield the best defined analgesic profile. The moderate rather than high lipid solubility of

54

alfentanil yielded the most predictable drug effects. Very high lipid solubility favors drug uptake at extensive nonspecific binding sites [23]. These differential effects reflect the different physicochemical properties of these three opioids. They differ in percent of unionized drug at physiological pH and in lipid solubility. Alfentanil is over 100 times lipid soluble as morphine and fentanyl is more lipophilic than alfentanil [16]. At pH 7.4 alfentanil is about 90% unionized whereas morphine is about 20% and fentanyl less than 10% unionized [14,16]. Thus, alfentanil may have the highest capacity to diffuse from plasma to brain with fentanyl diffusing somewhat more slowly and morphine diffusing very slowly [12]. After diffusion into brain the high lipophilicity of fentanyl may cause substantial amounts of the drug to bind nonselectively (e.g.. to lipids) so that the concentration of fentanyl at opioid receptor sites does not increase appreciably until nonselective binding approaches saturation [23]. The lower lipid solubility of alfentanil could account for both its rapid onset and the high correspondence between alfentanil plasma concentration and brain effect. Alfentanil may be lipophilic enough to afford rapid penetration to brain receptors involved in opioid analgesia but not so lipid soluble that it binds extensively to nonselective brain sites. As a result, alfentanil concentration at brain opioid receptor sites covaries with plasma drug concentrations. For these drugs, the timing of CNS effects is as interesting as the potency differences. Our findings with alfentanil and fentanyl concur with those of Scott et al. [23] who used spectral edge to measure opioid effect. We observed that alfentanil produced a nearly immediate effect on both pain measures. There was a lag in the peak effect of fentanyl, and morphine was considerably slower to produce peak effect and to recover. Our results with morphine may be due to (1) time required for morphine to reach opioid receptor sites responsible for analgesia once it has entered the lipid-rich environment of brain, and (2) the slow efflux of this drug from brain tissue. In studies that measured brain flux of these 3 opioids, Owen et al. [21] showed that morphine, alfentanil and fentanyl enter brain from blood equally rapidly but differ

markedly in the rate at which they diffuse back from brain to blood. Clearance of opioid from brain may be very high for alfentanil. moderate for fentanyl and very low for morphine. Measurement error. inevitable in any study. was greater here for EP than PR. This difference may reflect the unique sensitivity of the EP to habituation. The problem of EP habituation is minor when data collection can be interspersed with adequate periods of rest (see following papers). In bolus administration studies, data collection must proceed at a rapid pace and habituation is unavoidable. The method we used to correct the EP necessitated the untestable assumption that habituation is essentially the same for a given subject during drug testing as during testing with saline. If our data have met this assumption imperfectly, the greater variability in the EP data may be due to either inconsistent patterns of habituation for individuals across sessions or an interaction of drug effect with habituation. Unfortunately. we are unable to further evaluate the importance of this assumption. In conclusion, this report demonstrates three useful observations. First, even when drugs share a common mechanism for producing analgesia. they manifest markedly different analgesic profiles when their physicochemical characteristics differ. The 3 opioids studied here are all mu-selective. but their onset, duration and potency depend upon their pharmacokinetics, and these are determined by the lipophilicity and other characteristics of the individual drugs. These differences have received little attention in clinical pain management. Second, comparison of opioid analgesic states is suboptimal when plasma concentrations are unstable. Major pharmacokinetic and other individual differences among subjects tend to obscure analgesic effect because they govern plasma opioid concentration and effect tends to lag behind plasma concentration. This report underscores the importance of controlling or accounting for 18-22 changing plasma concentrations in clinical pain control trials. The complications ensuing from neglect of this factor differ with the drug in question and depend on both the physicochemical nature of the drug and the extent of individual differences among subjects.

55

Finally, individual differences in pharmacokinetics and pharmacodynamics are great and can obscure assessment of analgesic states in group data. These differences are more marked with some opioids than with others. Our results suggest that study of opioid analgesia under conditions of unstable plasma drug concentration (common practice in clinical studies) requires close attention to individual differences. Use of standardized doses and evaluation of group means maximize the deleterious effects of between subjects variability.

10 11

12

13 14

Acknowledgements This work was supported by grants from the National Cancer Institute (CA 385529, the National Institute on Drug Abuse (DA 05513), and by Janssen Research Foundation.

References Bourbon, W.T., Will, K.W., Gary, H.E. and Papnicolaou, A.C., Habituations of auditory event-related potentials: a comparison of self-initiated and automated stimulus trains, Electroenceph. Clin. Neurophysiol., 66 (1987) 160-166. Bower, S. and Hull, CT., Comparative pharmacokinetics of fentanyl and alfentanil, Br. J. Anaesth., 51 (1982) 871-877. Callaway, E., Habituation of averaged evoked potentials in man. In: H.V.S. Peeke and M.J. Herz @is.), Habituation, Vol. 2, Physiological Substrates, Academic Press, New York, 1973, pp. 153-174. Chapman, CR. and Benedetti, C., Nitrous oxide effects on cerebral evoked potential: partial reversal with a narcotic antago~st, Anesthesiology, 51 (1979) 135-138. Chapman, CR., Colpitts, Y.M., Benedetti, C. and Butler, S., Event-related potential correlate of analgesia: comparison of fentanyl, acupuncture and nitrous oxide, Pain, 14 (1982) 327-337. Chapman, CR., Colpitts, Y.M., Benedetti, C., Kitaeff, R. and Gehrig, J.D., Evoked potential assessment of acupunctural analgesia: attempted reversal with naloxone, Pain, 9 (1980) 183-197. Chapman, CR. and Jacobson, R.C., Assessment of analgesic states: can evoked potentials play a role? In: B. Bromm (Ed.), Pain Measurement in Man, Elsevier, Amsterdam, 1984, pp. 223-255, Chen, A.C.N. and Chapman, CR., Aspirin analgesia evaluated by event-related potentials in man: possible central action in brain, Exp. Brain Res., 39 (1980) 359-364. 9 Cohen, J. and Cohen, C., Applied Multiple Regression/

15

16

17

18

19

Correlation Analysis for the Behavioral Sciences, Erlbaum, Hillsdale, NJ, 1975. Draper, N. and Smith, H.S., Applied Regression Analysis, 2nd edn., Wiley, New York, 1981. Hill, H., Walter, M., Saeger, L., Sargur, M., Sizemore, W. and Chapman, R., Dose effects of alfentanil in human analgesia, Clin. Pharmacol. Ther., 40 (1986) 178-186. Hull, C.J., The pharmacokinetics of opioid analgesics, with special reference to patient-controlled administration. In: M. Harmer, M. Rosen and M.D. Vickers (Eds.), PatientControlled Analgesia, Blackwell Scientific Publications, Oxford, 1985, pp. 7-17. Kay. B., Postoperative pain relief, Anesthesia. 36 (1981) 949-951. Kaufman, J.J., Semo, N.M. and Koski, W.S., Micromanometric titration measurement of the pK,s and partition and drug distribution coefficients of narcotics and narcotic antagonists and their pH and temperature dependence, J. Med. Chem., 18 (1975) 647-655. Martin, R.W. and Chapman, CR., Dental dolorimetry for human pain research: methods and apparatus, Pain, 6 (1979) 349-364. Meuldermans, W.E.G., Hurkmans, R.M.A. and Heykants, J.J.P., Plasma protein binding and distribution of fentanyl, sufentanil, alfentanil and lofentanil in blood, Arch. Int. Pharmacodyn. Ther., 257 (1982) 4-19. Michiels, M.. Hendriks, R. and Heykants, J.. A sensitive radioimmunoassay for fentanyl. Plasma level in dogs and man, Eur. J. Clin. Pharmacol., 12 (1977) 153-158. Michiels, M., Hendriks, R. and Heykants, J.. Radioimmunoassay of the new opiate analgesics alfentanil and sufentanil. Pharmacokinetic profile in man, J. Pharm. Pharmacol., 35 (1983) 86-93. Miltner, W., Larbig, E. and Braun, C., Habituation of subjective ratings and event-related potentials to painful intracutaneous electrical stimulation of the skin. J. Psychophysiol., 1 (1987) 221-228.

20 Molenaar. P.C.M., The analysis of multiple habituation profiles of single trial evoked potentials. Biol. Psychol., 24 (1987) I-21. 21 Owen, H., Mather, L.E., Runciman, W.B., Carapetis. R.J. and Upton, R.N., The lockout interval in patient-controlled analgesia: a rational basis for choice? Br. J. Anaesth., 59 (1988) 1328-1329. 22 Perl. E.R., Afferent bases of nociception and pain: evidence from the characteristics of sensory receptors and their projection to the spinal dorsal horn. In: J.J. Bonica (Ed.), Pam, Raven Press, New York, 1980, pp. 19-46. 23 Scott, J.C., Ponganis, K.V. and Stanski, D., EEG quantification of narcotic effect: the comparative pharmacodynamics of fentanyl and alfentanil. Anesthesiology, 62 (1985) 234-241. 24 Todd, R.D., Muldoon, S.M. and Watson, R.L., Determination of morphine in cerebrospinal Buid and plasma by high performance liquid chromatography with electrochemical detection, J. Chromatogr. Biomed. Appl., 232 (1982) lOl110.