Pharmacokinetics of Propofol Administered by Target-controlled Infusion to Alcoholic Patients

After institutional approval and written informed consent, 15 alcoholic and 15 control patients, aged 18–65 yr, scheduled to undergo elective otolaryngologic surgery were planned to be included in this prospective, parallel group study to assess the influence of chronic alcohol intake on the pharmacokinetics of propofol. Three patients (two controls and one alcoholic patient) were excluded because of technical problems either with the infusion device (two patients) or with blood sampling (one patient). Therefore, to reach the required number of 15 in each group, 17 alcoholic and 16 control patients were ultimately included in the study. For a patient to be included in the alcoholic group, the following criteria had to be met: chronic daily alcohol intake greater than 75 g pure alcohol, signs of addiction to the drug such as inability to reduce alcoholic consumption despite obvious deleterious effects and morning tremor alleviated by an alcoholic beverage, biologic sequel of alcohol abuse such as erythrocyte volume greater than 95 μm³and γ-glutamyl transferase activity greater than 35 U/l, but no clinical or biologic signs of cirrhosis. Erythrocyte volume and γ-glutamyl transferase activity were chosen as markers of chronic alcoholism because they are easily measured, and if none of them are reliable enough on their own to support the diagnosis of alcoholism, when they are combined, their sensitivity in detecting alcoholism is superior to 80%. 6,7Patients were eligible for inclusion in the control group if their daily alcohol intake was less than 25 g pure alcohol, with a normal erythrocyte volume and γ-glutamyl transferase activity. Table 1is used to estimate the amount of daily alcohol intake.

Patients’ baseline characteristics in the two groups were compared by Wilcoxon rank-sum test for continuous variables and Fisher exact test for qualitative variables. A value of P < 0.05 was considered as statistically significant.

One hour before induction of anesthesia, patients received 0.15 mg/kg diazepam orally. After a bolus dose of 10 μg/kg alfentanil followed by an infusion at a constant rate of 10 μg · kg⁻¹· h⁻¹, anesthesia was induced with propofol using a commercial TCI device (Diprifusor®, AstraZeneca, London, United Kingdom, implemented in a Master-TCI infusion device, Fresenius Vial SA, Brezins, France) at an initial target blood concentration (TC) of 6 μg/ml. The TC was adjusted in 0.5-μg/ml steps to allow laryngoscopy, topical anesthesia of the vocal cords, and tracheal intubation and then at least every 15 min during maintenance according to clinical signs (TC was increased by 0.5 μg/ml if the patient moved and/or mean arterial blood pressure increased above 130% of baseline values and/or heart rate was greater than 90 beats/min; TC was decreased if anesthesia seemed adequate or if systolic arterial pressure was less than 80 mmHg). No muscle relaxants were used. The level of hypnosis in our patients was not monitored by electroencephalographic analysis, and the adequacy of anesthesia was assessed only on hemodynamic parameters and movements in response to surgical stimuli. The lungs of the patients were ventilated to normocapnia with 50% N²O in oxygen. The alfentanil infusion was stopped at the beginning of skin closure, and TC was set to zero at the end of the dressing, when nitrous oxide administration was discontinued.

The propofol dose at loss of consciousness was recorded as well as movements in response to skin incision. Alcoholic patients frequently underwent longer surgical procedures than control patients. Consequently, the mean propofol infusion rate during the first 90 min of maintenance was calculated. The time to opening eyes after the end of propofol infusion was recorded, and blood samples were also taken at that time for measurement of blood propofol concentration.

The range of predicted alfentanil plasma concentrations from intubation to skin closure was estimated in all the patients by simulation of the infusion with Stanpump #implemented with the pharmacokinetic model of Maitre et al.  8 

Characteristics of propofol anesthesia between the two groups were compared by Wilcoxon rank-sum test. A value of P < 0.05 was considered as statistically significant.

Venous blood samples were drawn before propofol administration; after tracheal intubation; before each TC modification; when opening eyes on verbal command; and 2, 4, 6, 8, 10, 15, 30, 45, 60, 75, 90, 120, 180, 240, 300, 360, and 480 min after the end of propofol infusion. Propofol was measured by high-performance liquid chromatography as previously described. 9 

The history of propofol infusion, including all the changes in infusion rates and their times, was printed on-line for each patient. These data were implemented in the NONMEM data sheet. To reduce the amount of data, successive infusion rates were averaged when their difference was less than 10%.

The pharmacokinetic analysis was performed using the nonlinear mixed-effects modeling approach implemented in the NONMEM computer program (version V). 10Propofol pharmacokinetics was best described by a three-compartment mammillary model. The population characteristics of the pharmacokinetic parameters (fixed and random effects) were estimated using the subroutines ADVAN11 and TRANS4 from the library of programs provided with the NONMEM-PREDPP package. The model was parameterized in terms of volume of the central compartment (V¹), elimination clearance (Cl¹), intercompartmental clearances (Cl², Cl³), and volumes of the shallow peripheral (V²) and deep peripheral (V³) compartments.

Interindividual variability (random effects) was assessed according to a proportional error model associated with each fixed effect parameter: The kth component of the individual parameter p^jof the jth subject was modeled by p^jk=θ^kmean× exp(η^kj), where θ^kmeanis the population mean and η^kjis assumed to be a Gaussian random variable with zero mean and variance ω. 2 

The concentration measurements profile in the jth individual was assumed to be affected by a proportional error described by the relation: C^ij(t) = f(p^j, D^j, t^ij) × (1 +εe^ij), where C^ijis the ith concentration of the jth subject; t^ijis the time of the ith measurement; D^jis the dosing history of the jth subject; f is the pharmacokinetic model; and εe^ijdescribes the departure of the model from the observations and contains contributions from intraindividual variability, assay error, and model misspecification. The distribution of the residuals, εe, is assumed to be a random Gaussian variable with zero mean and variance ς2.

Empirical Bayes estimates of pharmacokinetic parameters for each subject were obtained using the POSTHOC option in NONMEM.

The decrease in blood propofol concentration over time after an infusion often shows a rebound increase during recovery. 11Consequently, the effect of awakening (opening eyes) on pharmacokinetics was modeled by allowing each parameter (V¹, V², V³, Cl¹, Cl², and Cl³) to change after awakening. 12The change in mean parameters was modeled as a percentage decrease from the initial parameter value. In the first stage of the analysis of the effect of awakening, each model parameter was sequentially allowed to change while the other parameters remained constant. The importance of these variations was evaluated based on changes in the minimum objective function (OF). A modification of a parameter after opening eyes was added to the model based on a decrease in OF, which is estimated by NONMEM as −2 log likelihood of the data. The changes in OF between two nested models has a distribution that is approximately chi-square with degrees of freedom equal to the difference between the number of parameters in the full and reduced model. Thus, when the change of a parameter resulted in a decrease of the OF of at least 4 units, it was considered as significant (P < 0.05) and included in the model. In a second stage of the analysis, the retained parameters were allowed to change simultaneously, and only the combination of parameters resulting in a further decrease of the OF of at least 8 units was retained. In addition, for each parameter whose mean was affected, it was tested whether interpatient variability was the same before and after opening eyes.

Alcohol status, age, body weight, lean body mass, 13and fat body mass (weight minus lean body mass) were considered as relevant covariates to be investigated in this analysis. The influence of each covariate on the individual Bayesian estimates of the parameters obtained from the basic model was examined. This first screening allowed the recognition of covariate/parameter associations to be included in the multivariate NONMEM model. This model was thus built forward by successively including the most significant covariates. Then, the adequacy of the model was assessed backward by removing step by step all the covariates included if their removal did not modify the OF. Covariates were added to the model based on an improvement in residual plots and a decrease in OF (duplication). A more conservative test based on a reduction of 8 units in the (P < 0.005) was used because of the multiple comparisons that are made for this final step.

The structural model was built up using the first-order estimation method. To estimate the final parameters, the conditional estimates method of the individual etas during the computation of the OF was used, with preservation of the dependence of the error on the individual parameters (FOCE with interaction).

The model performance was assessed both numerically and graphically, using individual and population predictions. The population weighted residual and the absolute population weighted residual were calculated for each sample, using the final model and the individual covariates. The median weighted residual and the median absolute weighted residual were used as overall measures for goodness of fit of the final model.

Alcoholic and control patients had similar ages, weights, and sex ratio (table 2).

This study dealt primarily with propofol, and thus, the amount of opioid supplementation was kept low, with an infusion rate of alfentanil that maintained the predicted concentration at a value less than 30 ng/ml throughout the procedure. The total amount of propofol and the predicted and measured concentrations during all phases of anesthesia were not different in the two groups of patients, and this similar potency was confirmed by the fact that the measured concentration when opening eyes was not modified by alcoholic status of the patients (table 3). No relation was found between the measured blood propofol concentration at the time of surgical incision and the occurrence of movements on skin incision.

Eight hundred thirty-five blood samples were analyzed in the 30 patients. The changes in blood propofol concentrations over time in the two groups are shown in figure 1. The pharmacokinetics of propofol in the whole population could be described by a three-compartment model, with a minimum OF of −180.23. After recovery (opening eyes), there were significant independent changes in mean Cl¹, V², and V³. When these parameters were allowed to change simultaneously with eyes opening, the minimum OF was further reduced to −267.28. The percentage change (+22% for Cl¹, −19% for V², and −17% for V³) was of the same order of magnitude for all parameters, and as a result, the OF was not significantly different (OF =−267.18) if the same magnitude of change was assumed for all the parameters. This solution being simpler, it was kept in the final model.

The introduction of alcoholism as a covariate resulted in a significant decrease in the OF when it was applied both to Cl²and to the interindividual variability for V¹(OF =−287.41), but the other parameters were not influenced by alcoholism. Chronic alcohol intake did not influence the change in pharmacokinetic parameters estimated after opening eyes. The introduction of weight, lean body mass, or age as covariates did not improve the OF. Conversely, the introduction of fat body mass as a covariate for both V³and Cl³significantly improved the OF (−295.37).

Thus, this study failed to demonstrate any major influence of alcoholism on propofol pharmacokinetics. Considering the interindividual variability, we assessed retrospectively the power of a study including 15 patients in each group to detect a difference in clearance with a test significance level of 0.05, based on the mean and SD of the Bayesian estimates of clearance values during the infusion and after opening eyes. Intraoperatively, the power to detect a 50% change in clearance was 80%, whereas after opening eyes, because of a lesser variability, the power to detect a 20% change was 99%.

Figure 2shows the NONMEM CONTROL file of the final model. The pharmacokinetic parameters for the final model are shown in table 4. The medians of the Bayesian estimates of the individual pharmacokinetic parameters before and after awakening, as a function of alcohol status, are shown in table 5.

The precision (median absolute weighted residual) and bias (median weighted residual) of the model were, respectively, 60.1% and −6% (fig. 3), and that of the individual Bayesian estimates were 9.55% and 0.77%. The weighted residuals are plotted against time for each individual in figure 4, and the ratio of the measured over predicted concentrations calculated for each sample are given in figure 5. The best, median, and worst fits, based both on individual median absolute weighted residuals calculated from the final population model estimations and on the individual Bayesian estimates, are shown in figure 6.

Despite that chronic alcoholism is a common disease worldwide, its influence on the pharmacokinetics of anesthetic drugs has not been extensively studied. Indeed, most studies have specifically excluded patients with cirrhosis of the liver. Swerdlow et al.  2investigated the pharmacokinetics and pharmacodynamics of thiopental in 11 patients with a history of excessive alcohol intake compared to 9 control patients or volunteers who were social drinkers. The alcoholic population had consumed ethanol 9–17 days before the study, with no evidence of acute intoxication or acute withdrawal at the time of the study, and was therefore similar to our own alcoholic population. The authors found no differences in thiopental pharmacokinetic and pharmacodynamic parameters between alcoholics and nonalcoholics. Sporadic reports of increased propofol requirements at induction of anesthesia have been published. 4There is a single prospective clinical study 5that found differences in propofol requirements as a function of drinking habits but could not differentiate pharmacokinetic or pharmacodynamic causes for this observation. In our study, the amount of propofol required (in conjunction with a low concentration of alfentanil) to maintain anesthesia in both alcoholic and nonalcoholic patients was not different, but the absence of monitoring of depth of anesthesia together with venous rather than arterial blood sampling precluded any detailed pharmacodynamic analysis.

Chronic alcoholism was associated with few changes in the pharmacokinetics of propofol because only the intercompartmental clearance to the shallow peripheral compartment was significantly altered, and interindividual variability of the volume of the central compartment was significantly greater in the alcoholics. These findings may be related to the heterogeneity of the alcoholic population, 14to the vasomotor changes induced by chronic alcohol intake, 15,16and to the fact that chronic alcoholism is frequently associated with heavy smoking. Our results are in good accordance with those of Liu et al.  17in rats selected for differential ethanol sensitivity.

It has been demonstrated that weight and age were significant covariates of propofol pharmacokinetics. 18,19However, in our population, weight and age were purposely kept within a narrow range, which explains why the introduction of those covariates did not significantly improve our model. Similarly, our alcoholic noncirrhotic patients did not have abnormal levels of plasma protein concentrations, specifically albumin.

Coadministered drugs might influence pharmacokinetics and modify comparisons if their behavior differed as a function of the alcoholic status of the patients. Propofol has a flow rate dependent clearance, and thus, pharmacokinetic interaction occurs mainly through hemo-dynamic changes. The heart rate and blood pressure profiles were not different in our two groups of patients. Pharmacokinetic interactions between propofol and alfentanil have been described. Propofol is a potent enzyme inhibitor, and because alfentanil has an intermediate clearance dependent on the CYP450 activity, when propofol and alfentanil are infused simultaneously, alfentanil concentrations are increased by approximately 20%. 20,21Alfentanil (which in this study was used at very low dosages) has no influence on the pharmacokinetics of propofol.

The pharmacokinetics of propofol described in this study before awakening are in good agreement with previously published results. 18,19Nevertheless, no study so far has attempted to model what happens on recovery, despite the fact that the increase in propofol concentration at that time is common knowledge and was evident in early publications on propofol pharmacokinetics. 11This phenomenon has been described with other lipophilic drugs 22but without any clear explanation. When no more drug is administered to a patient, a rise in plasma drug concentration may be related to a release of the drug from some sort of depot, to a decrease in peripheral protein binding or to changes in peripheral distribution. Clearance changes alone would induce mainly a change in the decrease slope but cannot account for a rise. For fentanyl, it has been hypothesized that the drug was released from a lipid depot at recovery, 23but the pharmacokinetics of propofol are characterized by a very slow return from the deep compartment (k31 is small when compared to k10), which should preclude any significant increase in blood concentration at recovery. Modifications in protein binding during recovery are difficult to determine, specifically because nowadays, intraoperative control over hypocapnia reduces the relative importance of postoperative respiratory acidosis. Another hypothesis is that the cardiac output and regional blood flows are modified during recovery, when the cardiovascular sympathetic drive is increased. 19Our findings of an increase in clearance with recovery (and propofol clearance depends on hepatic blood flow, which in turn varies with cardiac output) associated with reduced distribution volumes of the peripheral compartments support this hypothesis. Clearance is increased but cannot completely blunt this phenomenon. The interindividual variability of the pharmacokinetic parameters is larger during anesthesia than after opening eyes, and this in turn might be due to variations in sympathetic response both to anesthetic drugs and to surgical stimuli. These findings may explain the high performance error of pharmacokinetic models established during general anesthesia when they are applied to TCI devices for use during conscious sedation. 24,25 

Chronic alcoholism induces only mild changes in the pharmacokinetics of propofol. Conversely, propofol pharmacokinetics are markedly different during anesthesia and surgery or after opening eyes in the recovery period.