Key words: Anesthetics, intravenous: eltanolone (pregnanolone). Computer-controlled infusion. Pharmacokinetic-pharmacodynamic modeling.

Like several other structurally related steroids, pregnanolone (3alpha-hydroxy-5beta-pregnane-20-one), a metabolite of progesterone, has been known for nearly 40 y to be anesthetically active without causing notable endocrinologic side effects. [1] The problem of finding a suitable solvent restricted its clinical use until 1990, when the first volunteer studies began. [2] Because of its low water solubility, pregnanolone was formulated as a stable oil-in-water emulsion with soybean oil and called eltanolone. After various authors [3-6] reported largely satisfactory results using a bolus intravenous dose of eltanolone to induce anesthesia, we studied the concentration-effect relation of this new steroid hypnotic agent using pharmacokinetic-pharmacodynamic modeling.

Twelve healthy men, ages 24-31 y (27 +/- 2 y; mean +/- SD), were enrolled in the study. All volunteers were within 30% of normal weight (according to the Metropolitan Life Insurance Company tables for 1980: 76 +/- 5 kg; mean +/- SD) and showed normal borderline electroencephalographic (EEG) results. None had documented drug allergies; known alcohol, drug, or medication abuse; or had participated in clinical studies of nonapproved drugs in the 2-week period before admission. Results of prestudy physical examinations, including blood pressure, electrocardiograms, and routine laboratory tests for all participants, were normal. The men fasted (no food, drink, or smoking) from at least 6 h before administration of the study drug to 6 h after cessation of the infusion.

The study protocol was approved by the Medical Ethics Review Committee of Freiburg, Germany, as well as by the local ethics committee and was conducted in accordance with the revised Declaration of Helsinki. All volunteers gave written informed consent.

A cannula was inserted into the radial artery of the nondominant arm and into a forearm vein of the contralateral side of each man. Hemodynamic and EEG monitoring devices were applied and after a rest of about 10 min stable baseline values of blood pressure, heart rate, oxygen saturation, and EEG were recorded. Eltanolone emulsion (4 mg/ml) was infused (Braun Perfusor Secura FT; Braun, Melsungen, Germany) via the indwelling venous cannula as two repetitive infusion cycles. The infusion rate was controlled by a computer (Toshiba 1850 C, Neuss, Germany) that was programmed to achieve venous serum concentrations of 2.0-2.5 micro gram/ml of unchanged eltanolone after about 20 min of infusion. The anticipated increase of the serum concentration was 0.1 micro gram [centered dot] ml sup -1 [centered dot] min sup -1 in the first two volunteers and was then reduced to 0.075 micro gram [centered dot] ml sup -1 [centered dot] min sup -1. The resulting infusion rate varied between 6 and 10 mg/min of eltanolone, which could only be achieved by using two parallel installed infusion pumps. Pharmacokinetic parameters obtained from bolus-dose studies [2,3] were used to program the software (IVA-PUMP, version 3.2; Bonn, Germany) controlling the infusion pumps. They were the only pharmacokinetic data available at the beginning of the study. Results of recently performed bolus-dose studies also suggested that under the described conditions, a thorough observation of the induction phase might be possible. [4,6,7] Although the EEG median frequency was used as a continuous parameter for the hypnotic effect of eltanolone, clinical signs such as cessation of counting, loss of response to verbal command, and loss of the corneal reflex also were used as indicators for the course of anesthesia. The infusion was stopped when burst suppression of 1-2 s occurred in the EEG recordings. It was restarted when the volunteers were awake and fully orientated to person, place, and time. At the end of the second awakening period, monitors were removed and the volunteers were observed clinically until the next morning. During the two infusion cycles, oxygen was administered through a probe to the nose with a flow of 3 l/min. A maximum amount of 500 ml Ringer's solution was infused to each man, and arterial P^CO2was analyzed from the collected blood every 6 min. To avoid interference with natural sleep, loud verbal commands were set every 2 min ("open your eyes") when the men stopped counting until the appearance of burst suppression and from the disappearance of burst suppression until the men were fully orientated.

Blood samples to determine eltanolone concentrations were collected from the indwelling arterial cannula contralateral to the infusion. The dead space of the cannula (0.1 ml) was flushed using a small amount of isotonic saline after each blood sample. Before the blood samples were collected for analysis, 1 ml blood was withdrawn and discarded. A control sample (10 ml) was drawn 3 min before the start of infusion and further samples of 5 ml each were drawn after the start of the first infusion cycle. The two infusion cycles were divided into two periods each, periods A and B and periods C and D, respectively. Period A was the first infusion, terminated by the appearance of burst suppression in the raw EEG, followed directly by period B, the first recovery period, terminated by the complete orientation of the volunteer. Period C was the second infusion, terminated again by the appearance of burst suppression patterns, followed by period D, which lasted until 12 h after the cessation of the second infusion. During periods A and C, blood samples were drawn every 3 min. During period B, blood was collected every 3 min until 30 min and then every 5 min until 60 min. The same procedure was carried out during period D; in addition, blood samples were obtained at 90, 150, 210, 270, 390, 510, 630, and 720 min after cessation of the second infusion. The blood was collected in labeled tubes without anticoagulant and was left to clot for at least 45 min at room temperature. The serum was separated by centrifugation within 1.5 h after collection and transferred to labeled polypropylene tubes. The tubes were stored at -20 degrees Celsius pending analysis.

The serum samples were analyzed for unchanged eltanolone using a specific gas chromatographic mass spectrometric assay with selected ion monitoring (Hewlett Packard gas-chromatograph 5890, and mass selective detector 5971A). Separation was performed on a methyl silicone, 25 m x 0.2 mm capillary column with 0.11-mm film thickness. Two microliters were injected in the splitless mode for 2 min. The column oven was heated to 120 degrees Celsius, followed by a temperature gradient of 45 degrees Celsius min sup -1 to 280 degrees Celsius and the injection part to 300 degrees Celsius. Helium was used as the carrier gas. The mass spectrometer was operated at 70 eV ionization potential. The drug was extracted from serum with diethyl ether/pentane 1:3 (VIV). The extract was evaporated and subsequently a derivatization with N,O-bis (trimethyl-silyl) trifluoracetamide was performed. 3Beta-hydroxy-5alpha-pregnan-20-one was used as an internal standard. The daily reproducibility of the method was 5 +/- 1.6% (mean +/- SD) with accuracy of 90-110%. The limit of sensitivity was 0.003 micro gram/ml. The analysis was performed in the Analytical Department, Pharmacia, Stockholm, Sweden.

The pharmacokinetic parameters were calculated (see appendix 1) and fitted using the NONMEM program, version IV, a nonlinear mixed-effects modeling algorithm developed by Beal and Sheiner*.

Electroencephalographic recording began 10 min before eltanolone was administered and was continued until the early second recovery period, when the men had regained full orientation as defined previously. Electroencephalographic data were recorded using the CATEEM electrodiagnostic monitoring system (Computer-aided topographical electroencephalometry; Medisyst, Linden, Germany). Bipolar cerebral activity was monitored using an electrocap (Electro-Cap International, Eaton, OH) with the electrodes positioned according to the international 10-20 system and using Cz placement as a physical reference. The gap between the cavities of the electrodes and the tissue was bridged by an electrolytic jelly. The signals were preamplified with a battery-powered amplifier close to the head, and the resulting digital code was transmitted by optical fiber to the CATEEM system. Artifact-free serial epochs of 4 s were digitized at a rate of 512 Hz (12 bit) using bandpass filters between 0.45 and 35 Hz. After smoothing the signals, the frequency analysis was performed by fast Fourier transformation. High-input impedance ensured a sufficient signal-to-noise ratio with the electrode impedances ranging from 1 to 100 k Ohm. An additional display of the unprocessed raw EEG was used to check the parameters derived and to recognize the burst-suppression patterns.

For each lead, EEG analysis included the changes within the individual band power and median frequency, defined as the frequency that divides the area below the power spectrum curve into two equal parts. The individual frequency bands were set as follows: delta, 1.25-4.5 Hz; theta, 4.75-6.75 Hz; alpha¹, 7.0-9.5 Hz; alpha², 9.75-12.5 HZ; beta¹, 12.75-18.5 Hz; and beta², 18.75-35.0 Hz. To correlate the EEG changes with the serum concentrations, the values of the central lead of the dominant hemisphere (C³) were used from the 17 channels derived. Because the CATEEM system excludes the subdelta band (0.5-1.25 Hz) from the power spectral analysis, the raw EEG signal of this lead was transferred to a separate personal computer-based analysis tool. There the signal of each epoch (8 s) was converted (128 Hz, 16 bit), the power spectrum from 0.5 to 32 Hz was derived by fast Fourier transformation, and the median frequency was subsequently calculated (EEG-BASIC, version 2.4; developed by Schwilden, Bonn, Germany), thus producing more information during periods of pronounced hypnosis. The appearance of burst-suppression patterns was analyzed visually in the raw EEG, and these periods were excluded from the power spectral analysis.

In addition, the raw EEG signal was examined for amplitude artifacts and steep slopes, and a test on normal distribution was performed. The derived median frequencies were averaged to get one value per minute. The best-fit serum concentrations of the pharmacokinetic modeling were correlated to the corresponding values of the median frequency using pharmacodynamic modeling. As a basis for the modeling process, the sigmoid E^maxmodel was chosen to account for the nonlinearity of the concentration-effect relation (see appendix 1). Then a nonlinear regression analysis was done to identify the pharmacodynamic model parameters. In case of a time-lag (hysteresis) between the onset of the effect and the course of the serum concentration, an additional effect compartment was defined. Thus k^e0was calculated within the fitting process by minimizing the hysteresis to describe the equilibration time between the central compartment (arterial blood) and the biophase or effect compartment. The course of the eltanolone concentration in the effect compartment was estimated from this procedure.

The mean, SD, and SEM were used to express the central tendency of the data. The Wilcoxon matched-pairs test was performed to calculate significant changes. Probability values less than 0.05 were considered significant. The statistical evaluation was done using the program STATISCA for Windows, release 4.5 (StatSoft, Tulsa, OK).

Twelve young and healthy men were enrolled in the study. One man (volunteer 4) was excluded from the data analysis because of a seizure during the first awakening period, which led to his withdrawal from the study. Minor undesirable effects such as involuntary movements, excess salivation, coughing, or inspiratory stridor occurred in 100%, 25%, 25%, and 42% of the volunteers, respectively. In one man, oxygen saturation decreased to less than 90% but returned to a normal level immediately after oxygen was administered via a face mask. All participants breathed spontaneously during the entire observation period. Although the systolic blood pressure decreased significantly (P < 0.05) after the infusion of eltanolone, the diastolic and mean pressure remained largely stable. Heart rate increased significantly (P < 0.05) during both infusion cycles (Table 1).

The total doses administered to the 11 volunteers during the two infusion cycles varied between 149 and 236 mg (186 +/- 25 mg; mean +/- SD). The serum concentration of eltanolone versus time of a representative volunteer (number 8) is shown in Figure 1(a) together with the best-fit line and the predicted curve. Figure 2(a) depicts the time course of the measured concentrations for all volunteers expressed as means and SDs at the defined times of blood sampling. Table 2summarizes the pharmacokinetic parameters using the best individual fits of the 11 volunteers evaluated. Altogether, eltanolone showed a high total clearance, a small central volume of distribution, and a relatively short half-life. With the bolus pharmacokinetic data used to manage the computer-controlled infusion device, serum concentrations during induction were systematically underestimated, whereas they were systematically overestimated after the infusion was stopped.

Changes in the EEG median frequency after the infusion of eltanolone emulsion of a representative volunteer (number 8) are shown in Figure 1(b). The time course for all volunteers is plotted in Figure 2(b) as means and SDs at the times of blood sampling. Median frequency decreased significantly from the ninth minute after the start of the first infusion (P < 0.01) until the 24th minute of the first recovery period (P < 0.05). The maximal decrease during the first cycle was achieved 3 min after cessation of the infusion, with 0.84 +/- 0.09 Hz (mean +/- SD; P <0.001). In the second cycle, median frequency was significantly less than the baseline from the sixth minute after the start of the infusion (P <0.01) until the 30th min of the second recovery period (P < 0.05). Using the appearance of burst-suppression patterns as an end point, the infusion time lasted 19.3 +/- 2.9 min (mean +/- SD) in the first loop and 14.8 +/- 2.5 min in the second loop. Full orientation was regained 35 +/- 6 min after the cessation of the infusion in the first cycle and 38 +/- 7 min in the second cycle, respectively.

(Table 3) summarizes the pharmacodynamic modeling parameters of the individual volunteers. The C^p50 was 0.46 +/- 0.09 micro gram/ml. Furthermore, we found a hysteresis between arterial serum concentration and response, with an equilibration half-life of about 8 min (k^e0= 0.087 +/- 0.013 min sup -1) and a relatively steep concentration-response relation (gamma = 6.2 +/- 1.7). Using the EEG median frequency and the arterial serum concentrations, the concentration-versus-effect plot of Figure 3(a) of a representative volunteer (number 8) describes a wide hysteresis loop. As shown in Figure 3(b) using the estimated effect compartment concentrations of eltanolone, a consistent sigmoid relation between concentration and effect was derived. The relation between the concentration in the effect compartment and the EEG median frequency is plotted for each volunteer in Figure 4, illustrating the interindividual variability in pharmacodynamics. With regard to clinical signs such as falling asleep (cessation of counting) or loss of response to loud verbal commands, sufficient conformity was found (Figure 5). Only small differences remain in the concentrations for the loss and the reappearance of clinical signs, respectively.

Eltanolone emulsion is a new steroid intravenous anesthetic that has already been found to be a potent hypnotic in previously published dose-finding studies by judging clinical criteria for successful induction of anesthesia. [4,5] van Hemelrijck and colleagues [6] examined the relative potency of eltanolone using logistic regression analysis. They found it to be 3.2 times more potent than propofol and six times more potent than thiopental. In the present study, in which we applied a pharmacokinetic-pharmacodynamic modeling procedure with a sigmoid E^maxfunction after computer-controlled infusion, eltanolone also proved to be a potent hypnotic. Its C^p50 of 0.46 micro gram/ml indicates a hypnotic potency comparable to that of etomidate (0.32 micro gram/ml) [8] and about four times greater than that of propofol (2.3 micro gram/ml). [9] 

The pharmacokinetic data obtained in this study showed the rapid elimination characteristics of eltanolone. The total clearance of 25 ml [centered dot] min sup -1 [centered dot] kg sup -1 is nearly in the same range as that of propofol at 28 ml [centered dot] min sup -1 [centered dot] kg sup -1. [10] As was expected based on comparable studies with other hypnotics, [9,11] the central volume of distribution after infusion and arterial blood sampling was found to be significantly (P < 0.05) smaller (about four times) than that reported after bolus injection and venous blood collection (0.11 +/- 0.01 l/kg instead of 0.42 +/- 0.23 l/kg). This explains the systematic underestimation of serum concentrations during induction by the computer simulation, which applied the bolus data set. With regard to the best-fit data of the NONMEM model-dependent pharmacokinetic analysis, it is evident that the predicted peak values regularly remain less than the measured serum concentrations. This might be due to nonlinear kinetic behavior at higher concentrations with a decreasing total clearance, which is, however, not supported by the results of previous bolus-dose studies. Furthermore, it might show a fundamental problem of traditional pharmacokinetic models, making the assumption that each compartment is continuous and that the drug administered mixes instantaneously. [12] In reality, especially at high infusion rates, compartments might be discontinuous and separated by circulation delays. [13] Conventional compartment models do not incorporate circulation delays, and the applied three-compartment model failed to describe exactly the measured peak concentrations of eltanolone in this study.

With respect to the pharmacodynamic parameters of eltanolone, this study revealed a high hypnotic efficacy. It was, however, accompanied by a hysteresis (8 min equilibration half-life) that clearly exceeds the values of other hypnotics, such as thiopental (1.5 min) and propofol (3 min). [9,14] Etomidate and ketamine do not show any measurable hysteresis. [8,15] Because the physicochemical properties of eltanolone do not provide obvious reasons for this hysteresis, we speculated that a delayed diffusion into the brain or protracted receptor binding causes the time lag between concentration and the corresponding effect.

Another possibility was discovered by Wang and colleagues, [16] who found that an albumin solution of 5alpha-pregnanolone was more potent in inducing anesthesia than an intralipid emulsion of the same compound. Thus we could conclude that the release of steroids out of lipid emulsions might be delayed. Nevertheless, it was possible to achieve induction times between 40 and 50 s after administering a bolus dose of 0.75 mg/kg (about double the median effective dose) of eltanolone emulsion in recently performed studies in patients, with only minor hemodynamic side effects. [4-7] This is probably due to the rapid initial decline of the serum concentrations after bolus injection, caused by redistribution, that leads to a quick adaptation of the peak serum concentration in the effect compartment, despite the long equilibration half-life. Thus, if the dose is large enough, the delay of induction will be of minor clinical relevance after bolus injections. However, during total intravenous anesthesia, if the effect of eltanolone must be increased or decreased rapidly, hysteresis will at first lead to a delayed response, and the steep concentration-effect relation (gamma = 6.2) will cause a sudden reaction. Therefore, control of the anesthetic effect is limited.

Another aspect will be the danger of overdose unless the meaning of the small k^e0is well understood or computer-controlled infusion devices programmed to achieve specific drug concentrations at the effector site in brain are used. The bolus-dose studies in the patients noted previously [4-7] confirm the relatively small hemodynamic side effects up to doses of 0.8 mg/kg body weight of eltanolone. Myint and colleagues [17] suggested that the induction dose in elderly patients (60 y or older) should not exceed 0.6 mg/kg, and moreover they found indications for a much slower pharmacodynamic response compared with propofol and thiopental. Kallela and coworkers [18] registered significantly longer recovery times after 0.8 mg/kg eltanolone was administered compared with 2 mg/kg propofol.

As a representative monoparametric value that summarizes the complex changes within the entire electrical cortex activity, EEG median frequency again proved to be a practical and reliable continuous measure for the central effect of a single hypnotic agent. [8,19,20] Changes in the EEG power spectrum and median frequency, found after infusion of eltanolone, correlated well with the clinical course of anesthesia and were comparable to those described after the administration of other hypnotics such as thiopental, etomidate, and propofol. [8,21,22] The pharmacodynamic end point used in this study was the appearance of burst suppression of the raw EEG wave form. Calculation of the median frequency, however, becomes unstable as the EEG nears the discontinuous phenomenon of burst suppression. To incorporate EEG activity at high serum concentrations in the pharmacodynamic modeling procedure, frequencies as low as 0.5 Hz had to be considered. Therefore, it was necessary to extend the CATEEM system by the capability offered through an additional analysis tool (EEG-BASIC). The comparison between the course of median frequency and clinical signs revealed good conformity. The small discrepancies remaining are probably caused by time-recording inaccuracies for the clinical signs, particularly with respect to the steepness of the concentration-effect relation.

Although eltanolone is thought to have anticonvulsive properties by acting at the gamma-aminobutyric acid receptor, [23,24] one volunteer experienced a seizure after the first infusion during awakening. This coincided with seizure activity in his raw EEG recordings showing polyspike waves. Medical history and results of all prestudy examinations, including his baseline EEG, were normal, as were those of the poststudy screening on the day after infusion. Seizure-like movements have been reported after all known hypnotics, but in nonepileptic patients mostly without EEG seizure activity. [25,26] This incidence would make it necessary to examine systematically under which circumstances eltanolone could develop proconvulsant effects, because this might be a severe contraindication of this drug.

At this stage of evaluation, eltanolone was shown to be a potent hypnotic agent with rapid elimination characteristics and relatively minor cardiovascular and respiratory side effects in healthy volunteers and patients. Induction characteristics were satisfactory after bolus injection, and the most frequent undesirable effects reported were involuntary movements and myoclonis. The equilibration half-life of 8 min between central compartment and effect compartment is a clear disadvantage compared with other hypnotics in that control of the compound is reduced. Furthermore, it involves the potential disadvantage of drug accumulation and prolongation of recovery if larger-than-necessary doses of eltanolone are used to induce anesthesia rapidly or to control rapidly the sudden responses to noxious stimulation during an operation. These findings, together with possible proconvulsant properties and increased incidence of urticaria, have led to the termination of the eltanolone project.

The serum concentration c(t) was calculated by convolution using the following formulas: Equation 1where g(t) describes the concentration after a bolus dose and I(t) is the infusion rate applied. The macro constants A¹, A², A³, lambda¹, lambda², and lambda³were computed from the pharmacokinetic parameters k¹², k²¹, k¹³, k¹³, Cl, and Vc. [27] 

The pharmacodynamic effect was described by the modified Hill equation: Equation 2where E(t) is the EEG median frequency measured at time t, E⁰is the baseline median frequency, E^maxis the maximum effect, C^p50is the half-maximum effect concentration, and gamma defines the steepness of the concentration effect curve. The concentration in the effect compartment C^e(t) is given by Equation 3where c(t) is the serum concentration and k^e0is the elimination rate out of the effect compartment.

*Beal SL, Sheiner LB: NONMEM User's Guide. San Francisco, University of California, 1994.