Free
Education  |   August 2001
Propofol Dosing Regimens for ICU Sedation Based upon an Integrated Pharmacokinetic– Pharmacodynamic Model
Author Affiliations & Notes
  • Juliana Barr, M.D.
    *
  • Talmage D. Egan, M.D.
  • Nancy F. Sandoval, M.D.
  • Katayoun Zomorodi, Ph.D.
    §
  • Carol Cohane, R.N.
    §
  • Pedro L. Gambus, M.D.
    ‖‖
  • Steven L. Shafer, M.D.
    #
  • * Assistant Professor of Anesthesia, Department of Anesthesia, Stanford University, and Staff Anesthesiologist, VA Palo Alto Health Care System. † Associate Professor of Anesthesia, Department of Anesthesia, University of Utah, Salt Lake City, Utah. ‡ Staff Anesthesiologist, Department of Anesthesia, Oakland Children’s Hospital, Oakland, California. § Research Associate, Department of Anesthesia, Stanford University. ‖‖ Staff Anesthesiologist, Dpto. De Anestesiologia, Hospital Universitari, Barcelona, Spain. # Professor of Anesthesia, Department of Anesthesia, Stanford University, and Staff Anesthesiologist, VA Palo Alto Health Care System.
  • Received from the Department of Anesthesia, Stanford University School of Medicine, Stanford, California, and the VA Palo Alto Health Care System, Palo Alto, California.
Article Information
Education
Education   |   August 2001
Propofol Dosing Regimens for ICU Sedation Based upon an Integrated Pharmacokinetic– Pharmacodynamic Model
Anesthesiology 8 2001, Vol.95, 324-333. doi:
Anesthesiology 8 2001, Vol.95, 324-333. doi:
ArticlePlus
Click on the links below to access all the ArticlePlus for this article.
Please note that ArticlePlus files may launch a viewer application outside of your web browser.
PROPOFOL is an intravenous sedative–hypnotic agent that is commonly administered for sedation of patients in the intensive care unit (ICU) who are being treated with tracheal intubation and mechanical ventilation. The pharmacokinetics of propofol in critically ill patients being given continuous intravenous infusions for sedation 1–4 differ significantly from the pharmacokinetics of propofol in healthy surgical patients being given intravenous boluses or short-term infusions of propofol for anesthesia. 5–7 Studies assessing the dose–response relationship of propofol in ICU patients have also shown significant variations in the mean time to emergence and extubation following discontinuation of propofol infusions. 8–12 To date, the exact relationship between the pharmacokinetics, depth of sedation, and propofol dosing in ICU patients has not been fully characterized. The purposes of this study were (1) to develop an integrated pharmacokinetic–pharmacodynamic model of propofol infusions in ICU patients; (2) to test this model prospectively in a second group of ICU patients; and (3) to develop rational dosing guidelines for propofol sedation in ICU patients based on the integrated pharmacokinetic–pharmacodynamic model for propofol.
Materials and Methods
Study Design
After institutional review board approval (Stanford University School of Medicine, Stanford, CA), written informed consent was obtained from 30 adult male medical or surgical ICU patients (age range, 21–80 yr) requiring endotracheal intubation and mechanical ventilation with sedation for more than 24 h. Individuals were excluded from the study if they had severe hepatic or renal insufficiency, significant hemodynamic instability, primary hyperlipidemia, pancreatitis, neurologic impairment, or a known allergy to eggs or propofol at the time of enrollment. Eligible individuals received no propofol within 30 days of enrollment into the study.
All subjects were given 10 mg/ml undiluted open-label intravenous propofol (Diprivan; AstraZeneca International, Wilmington, DE) for sedation. Propofol was administered to all subjects by means of a target-controlled infusion (TCI) to achieve predefined propofol plasma concentrations. Appropriate precautions were taken to prevent microbial contamination of the propofol infusion. The TCI system consisted of an 80306-20 laptop computer (Everex, Inc., Fremont, CA) with an MS-DOS operating system (Microsoft, Inc., Redmond, WA) running STANPUMP software. 1This computer was connected to an infusion pump (Harvard Pump 22; Harvard Apparatus, Inc., South Natick, MA) through a serial interface. The STANPUMP software controlled the infusion pump rate to target a specific plasma propofol concentration using a previously derived pharmacokinetic model for propofol. The first 20 subjects (i.e.  , learning group  ) were given propofol through the TCI system using a pharmacokinetic model for propofol derived from healthy surgical patients being given short-term propofol infusions for anesthesia (J. B. Dyck, M.D., Department of Anesthesia, Stanford University, Stanford, CA, written communication, February 1991). The last 10 subjects (i.e.  , test group  ) were given propofol through the TCI system using a pharmacokinetic model for propofol derived from the learning group. Subjects were given intravenous or epidural infusions of fentanyl (up to 200 μg/h) as needed for analgesia.
Subjects were allowed to emerge from the residual effects of general anesthesia (surgical patients) or previously administered sedative agents (medical patients) before the administration of propofol. An intravenous infusion of propofol was initiated at a target plasma propofol concentration of 0.5 μg/ml. This target concentration was increased every 5 min by 0.25–0.5 μg/ml until a sedation score (SS) of 5 or 6 was achieved as defined by the modified Ramsay Sedation Scale (table 1). 13 This level of sedation was maintained as long as subjects remained endotracheally intubated and mechanically ventilated except for daily neurologic assessments, when the target concentration was decreased by 0.25–0.5 μg/ml every 10 min until an SS of 2 was achieved. After the neurologic assessment, the target concentration was increased until the subject’s SS was again 5 or 6 (fig. 1). If a subject became agitated during the study, the propofol target concentration was incrementally increased to maintain an SS of 5 or 6. If a subject became hypotensive during the study, the propofol infusion was temporarily suspended until the subject was hemodynamically stable. Once subjects were ready to be weaned from mechanical ventilation, the propofol target concentration was decreased by 0.25–0.5 μg/ml every 10 min until an SS of 2 was achieved. This level of sedation was maintained until subjects were ready for tracheal extubation, at which point the propofol infusion was discontinued. Subjects were withdrawn from the study if they developed significant hemodynamic instability or hypertriglyceridemia (serum triglyceride concentration > 500 mg/dl) or if they required neuromuscular blockade or a surgical procedure requiring general anesthesia.
Table 1. Modified Ramsay Sedation Scale 13 
Image not available
Table 1. Modified Ramsay Sedation Scale 13 
×
Fig. 1. Propofol titration scheme using a target-controlled, intravenous infusion system. X = stepwise changes in the target plasma propofol concentration over time; SS = sedation score.
Fig. 1. Propofol titration scheme using a target-controlled, intravenous infusion system. X = stepwise changes in the target plasma propofol concentration over time; SS = sedation score.
Fig. 1. Propofol titration scheme using a target-controlled, intravenous infusion system. X = stepwise changes in the target plasma propofol concentration over time; SS = sedation score.
×
Plasma propofol assays were performed on 3-ml arterial blood samples obtained from subjects at the following times: at baseline; immediately before any changes made in the propofol target concentration; every 4 h during maintenance propofol infusion periods; and every 4 h for up to 5 days after discontinuation of the propofol infusion. Venous samples were obtained in lieu of arterial samples during the postinfusion period if a subject no longer had an indwelling arterial catheter. Blood samples were collected in 7-ml heparinized glass tubes and immediately placed on ice. Samples were centrifuged at 2,000 rpm for 15 min, and the plasma fraction was transferred to a 3-ml polypropylene tube and frozen at 4°C until assayed. Propofol assays were performed by ICI Pharmaceuticals Group (currently AstraZeneca International) using a modification of the whole-blood method of Plummer 14 (ICI Pharmaceuticals Group method 8P-03 and 8P-05). Plasma propofol concentrations were determined using liquid–liquid extraction followed by reverse-phase high-pressure liquid chromatography with fluorescence detection. Two different assay sensitivity ranges were used. The high-sensitivity range covered the standard from 2 to 200 ng/ml, and the low-sensitivity range covered the standard from 0.05 to 10 μg/ml. The limit of quantitation was 2 ng/ml. The use of target-controlled infusions, together with high-resolution sampling of plasma propofol concentrations and SS measurements, resulted in a highly accurate data set from which to derive pharmacologic models. 15 
Statistical Analyses
All statistical analyses were performed by members of the Department of Anesthesia, Stanford University School of Medicine (Stanford, CA). The demographics of the learning group (n = 20) and the test group (n = 10) were compared using the two-tailed t  test with the summary results expressed as mean (SD). The propofol infusion data for the learning group were compared to those of the test group using the Wilcoxon rank test with the summary results expressed as median (range). The value P  < 0.05 was considered to be statistically significant in both cases.
Multiple nonlinear logistic regression analyses, using both naïve pooled data and mixed-effects modeling approaches, were performed to characterize the pharmacokinetic and pharmacodynamic models for propofol in the learning group. Then, these models were prospectively tested in the test group. Model performance was assessed both numerically and graphically in both groups. A detailed summary of the pharmacokinetic and pharmacodynamic analyses can be found in Appendix 1 (Web Enhancement). The newly derived pharmacokinetic and pharmacodynamic models for propofol were integrated to construct dosing regimens for light and deep sedation with propofol in ICU patients.
Results
The demographic profiles of subjects in the learning group (n = 20) and the test group (n = 10) are summarized in table 2. There were no significant differences between the two groups in age, body habitus, or severity of illness. The propofol administration profiles for each group are summarized in table 3. There were no significant differences between the two groups in median duration of infusion, total propofol dose, or steady-state infusion rates. There was significant variation within both groups in infusion duration and total propofol dose.
Table 2. Subject Demographics (n = 30)
Image not available
Table 2. Subject Demographics (n = 30)
×
Table 3. Propofol Infusion Data (n = 30)
Image not available
Table 3. Propofol Infusion Data (n = 30)
×
Reasons for discontinuing the propofol infusions are summarized for all subjects in table 4. Nineteen subjects (63%) had their propofol infusions discontinued in preparation for extubation. In 18 of these subjects, extubation was successful, including 15 subjects who had extubation within 45 min of stopping the propofol infusion. Extubation was significantly delayed (> 24 h) in two subjects after discontinuation of the propofol infusion. One subject with a cervical myelopathy developed acute respiratory failure during weaning, and another subject had prolonged respiratory depression attributable to epidural morphine (respiratory depression attributable to propofol in the latter subject was ruled out on the basis of subclinical plasma propofol concentrations). Five subjects (17%) had persistent hemodynamic instability requiring discontinuation of the propofol infusion. One subject developed significant hypertriglyceridemia (peak serum triglyceride concentration = 1,148 mg/dl) while receiving propofol in conjunction with parenteral lipids for nutrition. His serum triglyceride concentrations returned to normal within 48 h of discontinuing both the propofol and lipid infusions.
Table 4. Reasons for Discontinuing Propofol Infusion (n = 30)
Image not available
Table 4. Reasons for Discontinuing Propofol Infusion (n = 30)
×
Three subjects were noted to have mild disorientation to place and time lasting 3–5 days after discontinuing their propofol infusions. All of these subjects had been given propofol for greater than 7 days and had been given large cumulative doses of propofol ranging from 308 to 1,421 mg/kg. A fourth subject became agitated and paranoid within 24 h of discontinuing his propofol infusion. This subject had a history of extensive ethanol use preoperatively and was given propofol for less than 48 h after major vascular surgery. He was treated with benzodiazepines and haloperidol for presumed ethanol withdrawal, and his mental status changes subsequently resolved.
Pharmacokinetics and Pharmacodynamics
The revised pharmacokinetic model for propofol was derived from 1,006 observations (93%) obtained from 19 of the 20 subjects in the learning group. One subject in the learning group was excluded from the pharmacokinetic analysis altogether, because of insufficient data. The data were best described both numerically and graphically by a three-compartment mammillary model using a naïve pooled data approach with lean body mass and fat body mass as covariates (table 5). This revised model differs significantly from the original model both numerically and in terms of its superior ability to predict plasma propofol concentrations. A detailed comparison of the original and revised pharmacokinetic models for propofol is summarized in Appendix 2 (Web Enhancement).
Table 5. Original versus Revised Pharmacokinetic Parameters for Propofol
Image not available
Table 5. Original versus Revised Pharmacokinetic Parameters for Propofol
×
The revised pharmacokinetic model for propofol was prospectively applied to the test group (n = 10). The performance of the revised model was similar in both the learning group and the test group (table 6). The performance variability of the revised model in the test group is reflected in the differences between the residual error plots for all subjects in the learning and test groups (figs. 2A and B). Figures 3A and Bdemonstrate comparable individual performances of the revised model in both groups.
Table 6. Performance Measures of Revised Pharmacokinetic Model for Propofol*
Image not available
Table 6. Performance Measures of Revised Pharmacokinetic Model for Propofol*
×
Fig. 2. Residual error plots as a measure of performance for the revised pharmacokinetic model in (A  ) the learning group (n = 19) and (B  ) the test group (n = 10), expressed as measured/predicted plasma propofol concentrations over time during the infusion and postinfusion periods. For each subject, solid lines depict the infusion period and dashed lines, the postinfusion period.
Fig. 2. Residual error plots as a measure of performance for the revised pharmacokinetic model in (A 
	) the learning group (n = 19) and (B 
	) the test group (n = 10), expressed as measured/predicted plasma propofol concentrations over time during the infusion and postinfusion periods. For each subject, solid lines depict the infusion period and dashed lines, the postinfusion period.
Fig. 2. Residual error plots as a measure of performance for the revised pharmacokinetic model in (A  ) the learning group (n = 19) and (B  ) the test group (n = 10), expressed as measured/predicted plasma propofol concentrations over time during the infusion and postinfusion periods. For each subject, solid lines depict the infusion period and dashed lines, the postinfusion period.
×
Fig. 3. Median and worst individual performances for the revised pharmacokinetic model in (A  ) the learning group and (B  ) the test group. The accuracy of the revised model is preserved when tested prospectively in individuals in the test group.
Fig. 3. Median and worst individual performances for the revised pharmacokinetic model in (A 
	) the learning group and (B 
	) the test group. The accuracy of the revised model is preserved when tested prospectively in individuals in the test group.
Fig. 3. Median and worst individual performances for the revised pharmacokinetic model in (A  ) the learning group and (B  ) the test group. The accuracy of the revised model is preserved when tested prospectively in individuals in the test group.
×
The pharmacodynamic model for propofol sedation was derived from 643 observations (100%) obtained from all 20 subjects in the learning group. The number of pharmacodynamic observations is less than the number of pharmacokinetic observations because SS measurements were suspended during the postinfusion period once the SS was 1 or 2, whereas plasma propofol concentrations were measured for up to 5 days after infusion. The results of the pharmacodynamic analyses are summarized in table 7. A mixed-effects model, relating the probability of sedation to plasma propofol concentration, best described the pharmacodynamic data as follows: where P(Sedation ≥ SS) is the probability that the sedation score is ≥ N (where N = 2, 3, . . .6), C is the plasma propofol concentration, C50,SSis the plasma propofol concentration at which P(Sedation ≥ SS) = 50%, and γ is the slope of the probability curve.
Table 7. Summary of Pharmacodynamic Models for Propofol*
Image not available
Table 7. Summary of Pharmacodynamic Models for Propofol*
×
Although the percentage of correct (observed SS = predicted SS) and close (observed SS = predicted SS ± 1) predictions of sedation in both the learning and test groups are similar for the naïve pooled data and mixed-effects modeling approaches, the mixed-effects analysis yields a more accurate pharmacodynamic model in terms of minimizing the objective function. Figure 4demonstrates that the mixed-effects approach also provides a better graphic description of the probability of sedation. On this basis, the mixed-effects model was chosen as the final pharmacodynamic model.
Fig. 4. Predicted probability of sedation for a sedation score (SS) ≥ 4. Comparing the performance of the derived pharmacodynamic models in the test group (n = 10) using naïve pooled data (NPD; dashed line) and mixed-effects modeling (MEM; solid line) approaches. Measured probabilities are derived from observed sedation scores at various plasma propofol concentrations using the technique of Somma et al.  The mixed-effects pharmacodynamic model provides a superior graphic fit of the data.
Fig. 4. Predicted probability of sedation for a sedation score (SS) ≥ 4. Comparing the performance of the derived pharmacodynamic models in the test group (n = 10) using naïve pooled data (NPD; dashed line) and mixed-effects modeling (MEM; solid line) approaches. Measured probabilities are derived from observed sedation scores at various plasma propofol concentrations using the technique of Somma et al. 
	The mixed-effects pharmacodynamic model provides a superior graphic fit of the data.
Fig. 4. Predicted probability of sedation for a sedation score (SS) ≥ 4. Comparing the performance of the derived pharmacodynamic models in the test group (n = 10) using naïve pooled data (NPD; dashed line) and mixed-effects modeling (MEM; solid line) approaches. Measured probabilities are derived from observed sedation scores at various plasma propofol concentrations using the technique of Somma et al.  The mixed-effects pharmacodynamic model provides a superior graphic fit of the data.
×
Equation 1describes the probability of being at or deeper than a given level of sedation. For clinical purposes, it is more useful to know the probability of being at a specific or discrete level of sedation for a given propofol concentration. Therefore, the probability of being at a discrete sedation score is defined in as follows:MATH
The probability curves for discrete levels of sedation are depicted in figure 5. With the exception of the curves for SS = 1 and SS = 6, the remaining probability curves are asymmetrically distributed. The maximal probability of SS = 1 (i.e.,  no sedative effect) approaches 1 as the plasma propofol concentration approaches zero. The maximal probability of SS = 6 (deeply sedated, unresponsive to any stimuli) approaches 1 as the plasma propofol concentration approaches infinity. Because of their asymmetric distribution, the plasma propofol concentrations corresponding to the peak probabilities (i.e.,  probability modes) for SS = 2, 3, 4, and 5 differ from the C50,SSfor each probability curve (table 8). The pharmacodynamic model derived for propofol using the mixed-effects modeling approach predicts light (SS = 3 ± 1) and deep (SS = 5 ± 1) levels of sedation in the test group with 73% accuracy (table 7).
Fig. 5. Probability curves for discrete sedation scores (i.e.,  SS = 1, 2, . . . 6). The distribution of probability curves is asymmetric. The peak probability (modes) of each curve defines the plasma propofol concentrations at which a discrete level of sedation is most likely to occur.
Fig. 5. Probability curves for discrete sedation scores (i.e., 
	SS = 1, 2, . . . 6). The distribution of probability curves is asymmetric. The peak probability (modes) of each curve defines the plasma propofol concentrations at which a discrete level of sedation is most likely to occur.
Fig. 5. Probability curves for discrete sedation scores (i.e.,  SS = 1, 2, . . . 6). The distribution of probability curves is asymmetric. The peak probability (modes) of each curve defines the plasma propofol concentrations at which a discrete level of sedation is most likely to occur.
×
Table 8. Comparison of Estimated C50,SS*to Sedation Score Modesfor Propofol (n = 20)
Image not available
Table 8. Comparison of Estimated C50,SS*to Sedation Score Modesfor Propofol (n = 20)
×
Dosing Regimens for Propofol Sedation
By integrating the results of the pharmacokinetic and pharmacodynamic analyses, we were able to develop dosing regimens for propofol in ICU patients targeting specific levels of sedation over time. Table 9summarizes the propofol infusion regimens necessary to maintain either light (SS = 3) or deep (SS = 5) levels of seda-tion in a typical ICU patient (i.e.,  61-yr-old man; weight, 81 kg; height, 176 cm) for up to 14 days. The plasma propofol concentrations corresponding to the probability modes for SS = 3 and SS = 5 probability curves were used as target plasma propofol concentrations for light and deep sedation, respectively. The emergence times from light sedation (i.e.,  the time it takes for the SS to decrease from 3 to 2) remained fairly rapid (i.e.,  < 35 min) for propofol infusions lasting 3 days or less. For infusions of longer duration, the emergence time from light sedation continued to increase, but a plateau was reached at approximately 3.5 h for 14-day infusions. By contrast, the emergence times from deep sedation with propofol (i.e.,  the time required for SS to decrease from 5 to 2) rapidly increased with longer propofol infusions. After a 24-h infusion, the time to emergence from deep sedation with propofol was 25 h. The emergence time from continuous deep sedation with propofol for 7–14 days was approximately 3 days. The differences in emergence times from light versus  deep sedation as a function of infusion duration is depicted graphically in figure 6. A 58% decrease in the plasma propofol concentration (from 0.6 to 0.25 μg/ml) was required for emergence from light sedation (SS = 3 → 2) with a steady-state emergence time of approximately 3.5 h. An 88% decrease in the plasma propofol concentration (from 2 to 0.25 μg/ml) was required for emergence from deep sedation (SS = 5 → 2) with a steady-state emergence time of approximately 75 h.
Table 9. Representative Dosing Guidelines for Propofol Infusions to Maintain Light and Deep Sedation in ICU Patients*
Image not available
Table 9. Representative Dosing Guidelines for Propofol Infusions to Maintain Light and Deep Sedation in ICU Patients*
×
Fig. 6. The predicted time required for plasma propofol concentration (PPC) to decrease by 88% and 58% as a function of the duration of the propofol infusion. An 88% decrease is required for emergence from deep sedation (i.e.,  sedation score = 5 → 2). A 58% decrease is required for emergence from light sedation (i.e.,  sedation score = 3 → 2).
Fig. 6. The predicted time required for plasma propofol concentration (PPC) to decrease by 88% and 58% as a function of the duration of the propofol infusion. An 88% decrease is required for emergence from deep sedation (i.e., 
	sedation score = 5 → 2). A 58% decrease is required for emergence from light sedation (i.e., 
	sedation score = 3 → 2).
Fig. 6. The predicted time required for plasma propofol concentration (PPC) to decrease by 88% and 58% as a function of the duration of the propofol infusion. An 88% decrease is required for emergence from deep sedation (i.e.,  sedation score = 5 → 2). A 58% decrease is required for emergence from light sedation (i.e.,  sedation score = 3 → 2).
×
Discussion
Propofol is routinely administered as a continuous intravenous infusion for sedation of patients in the ICU. The unique lipophilic properties of propofol enable it to cross the blood–brain barrier quickly, resulting in a rapid onset of sedation. The high metabolic clearance rate of propofol and its rapid redistribution into peripheral tissues account for the rapid emergence from sedation with short-term infusions with propofol, even though the elimination half-life of propofol is quite long (table 10). 1–4 The reported observed emergence times from propofol sedation in ICU patients is highly variable (table 11). 8–12 Previous studies of propofol sedation in ICU patients have focused on either the pharmacokinetics or dose–response relationships of propofol infusions. The principal purpose of this study was to develop a prospectively tested, integrated pharmacokinetic–pharmacodynamic model of propofol infusions in ICU patients. This integrated model was then used to simulate propofol infusion regimens in a typical ICU patient with predicted patterns of sedation and emergence times. The results of this study demonstrate that the emergence time from propofol sedation varies considerably and is a function of the depth of sedation, the duration of the infusion, and patient size.
Table 10. Comparison of Pharmacokinetic Models for Propofol in ICU Patients
Image not available
Table 10. Comparison of Pharmacokinetic Models for Propofol in ICU Patients
×
Table 11. Propofol Dose–Response Studies for Sedation of ICU Patients
Image not available
Table 11. Propofol Dose–Response Studies for Sedation of ICU Patients
×
Observed Side Effects of Propofol Infusions
Five subjects (17%) enrolled in this study developed significant hypotension requiring discontinuation of their propofol infusions. Although hypotension attributable to systemic vasodilation is a known side effect of propofol, 16 the incidence of significant hypotension in this study was higher than reported in previous propofol studies. 12,17 This is probably the result of using target-controlled infusions in the present study, which typically deliver an initial bolus dose of propofol with each increase in the target plasma propofol concentration. Bolus dosing of propofol significantly increases the risk of hypotension in critically ill patients and is not recommended for the routine management of ICU patients being given propofol for sedation.
One subject developed significant hypertriglyceridemia while being given propofol. Mild hypertriglyceridemia occurs commonly with propofol infusions, because of its lipid carrier, and is of little clinical consequence. Severe hypertriglyceridemia (i.e.,  serum triglyceride concentrations > 1,000 mg/dl) in ICU patients receiving propofol infusions occurs rarely and is typically associated with high propofol infusion rates, concurrent administration of parenteral lipids for nutrition, or baseline hypertriglyceridemia. 18,19 The one subject in the present study who developed significant hypertriglyceridemia had normal baseline serum triglyceride concentrations. His hypertriglyceridemia quickly resolved once his parenteral lipid and propofol infusions were discontinued.
Three subjects had mild confusion lasting several days after discontinuation of their propofol infusions. Although each subject was cooperative and able to follow commands, the subjects were intermittently disoriented to place and time. These symptoms resolved completely after 3–5 days in all cases. This confusion was attributed to persistent effects of propofol because all three subjects had been given propofol for greater than 7 days and had detectable concentrations of propofol in their plasma for up to 5 days after infusion. A fourth subject became acutely agitated after discontinuation of his propofol infusion. Acute withdrawal syndrome has been reported in ICU patients after prolonged sedation with propofol. 20,21 In this case, the subject’s agitation was attributed to acute ethanol withdrawal rather than withdrawal from propofol because the subject had a history of ethanol abuse and had been given propofol for only 48 h. There was no evidence of withdrawal from propofol in any of the subjects who were given prolonged infusions of propofol in the current study.
Pharmacokinetics and Pharmacodynamics
The pharmacokinetics of intravenous propofol infusions in the current study were best described by a three-compartment model with lean body mass and fat body mass as covariates. The addition of lean and fat body masses as model covariates significantly improved the accuracy of the model, which was preserved when prospectively tested in a similar group of ICU patients being given propofol infusions for sedation. This list of covariates is by no means complete, and a larger study including ICU patients with a wider degree of illness would potentially yield additional pharmacokinetic model covariates. The pharmacokinetic model derived in this study differs significantly from the simple three-compartment models for propofol infusions previously described for ICU patients (table 10). 1–4 Although the metabolic clearance rate of propofol estimated in the present study is comparable to clearance rate estimates for propofol in previous ICU studies, the estimated volume of distribution (Vdss) is three to nine times larger than previous estimates of Vdssin ICU patients. This larger Vdssresults from the longer duration of propofol administration, together with a longer plasma propofol sampling period (5 days vs.  ≤ 72 h) compared with previous studies of propofol infusions in ICU patients. The variability of the Vdssestimated in the current study is explained by the lipophilic properties of propofol and the significant influence of fat body mass on its Vdss. The clinical significance of this large Vdssis that with long-term propofol administration, significant drug accumulation and saturation of peripheral tissues occur, especially in obese patients. With increasing peripheral tissue saturation, the rate at which plasma propofol concentrations decrease after discontinuation of the propofol infusion becomes less dependent on redistribution and more dependent on metabolic clearance. This results in a much slower rate of decrease in plasma propofol concentration over time, which potentially increases the emergence time from sedation in obese patients.
The pharmacodynamic model derived in the current study relating the probability of sedation to plasma propofol concentration was based on a sigmoid model and was best described using a mixed-effects modeling approach. The accuracy of this pharmacodynamic model, like the pharmacokinetic model, was preserved when prospectively tested in a similar group of ICU patients. The derived pharmacodynamic model does not include the potential effects of epidural or intravenous fentanyl on propofol sedation because the effect of fentanyl cannot be incorporated into this model without an accurate fentanyl infusion history or known fentanyl plasma or cerebrospinal fluid fentanyl concentrations. Eleven subjects in the learning group and nine subjects in the test group were given either intravenous or epidural fentanyl infusions together with propofol. There were no differences in the observed relationship between depth of sedation and plasma propofol concentrations in these subjects when compared with subjects who were given propofol alone, suggesting that the effect of fentanyl on sedation was minimal in these patients. In a more diverse group of ICU patients (e.g.,  those with hepatic failure, renal failure, higher Apache II scores) or in patients being given higher doses of opioids for analgesia, the synergistic sedative effect of opioids on propofol sedation might become significant. Although tolerance to propofol has been previously reported, 22 there was no evidence of tolerance in subjects being given prolonged infusions of propofol in the current study.
Developing Dosing Regimens for Propofol Sedation
Integrating the derived pharmacokinetic and pharmacodynamic models for propofol infusions in ICU patients enables us to simulate various propofol infusion regimens, the corresponding patterns of sedation, and the predicted times to emergence from sedation in these patients. As shown in table 9, propofol infusion rates and the resulting times to emergence from propofol sedation differ considerably for light and deep levels of sedation. In addition, the observed differences between emergence times for light and deep sedation increase significantly as the duration of propofol sedation increases. Propofol retains its short-acting properties in patients who are lightly sedated. The estimated emergence time in a typical patient ranges from 13 min to 3.5 h for infusions lasting longer than 24 h. Patients who are deeply sedated are given larger total doses of propofol, resulting in significant drug accumulation and slower decreases in plasma propofol concentrations after discontinuation of the infusion. Combined with the greater required decrease in plasma propofol concentration to achieve emergence from sedation (figure 6), emergence times from deep sedation with propofol are much longer, ranging from 24 to 72 h for propofol infusions lasting more than 24 h in a typical patient. In morbidly obese patients, the predicted emergence times from both light and deep sedation with propofol would be expected to be even longer, because of the larger Vdssfor propofol in these patients. Regardless of a patient’s body habitus, to maintain a constant level of sedation, propofol infusion rates must be decreased over time to account for this cumulative drug effect.
None of the subjects in the current study had significant delays in emergence from sedation with propofol. This is attributable to the fact that all subjects had their propofol infusions frequently titrated to a specific level of sedation and were allowed to emerge to a light level of sedation before  being weaned from mechanical ventilation. The variability in observed emergence times from propofol sedation reported in previous studies of ICU patients can be explained by the differences in the duration of infusion and the depth of sedation maintained (table 11). Subjects who were lightly sedated with propofol for 12 h–4 days had an observed sedation emergence time of 5–90 min, which is consistent with the emergence times predicted by the integrated model in the current study. 8,12 In contrast, subjects in the study of Barrientos-Vega et al.  9 were more deeply sedated with propofol for an average of 6 days and had an average propofol emergence time of 35 h, which is also consistent with the results of the current study.
Propofol is a unique sedative–hypnotic agent with a rapid onset and offset of sedation with short-term administration. In critically ill patients being given continuous intravenous infusions of propofol for sedation, the offset of sedation varies considerably and is a function of the depth of sedation maintained, the duration of the infusion, and the patient’s size and body composition. Emergence times from sedation with propofol in ICU patients can be minimized by employing the following propofol sedation strategies: (1) Titrate propofol infusion rates to maintain a light level of sedation in ICU patients at all times (i.e.,  patient is asleep but responds to simple commands); (2) frequently reassess the patient’s depth of sedation and adjust the propofol infusion rate every 3–6 h for the first 24 h, then daily thereafter, to minimize propofol accumulation and decrease the risk of oversedation; (3) patients requiring deep sedation with propofol should have their propofol infusions suspended on a daily basis to allow them to emerge to a lighter level of sedation, and the propofol infusion should then be resumed at the minimal infusion rate required to maintain deep sedation; and (4) propofol dosing in morbidly obese patients should be based on their ideal body weight rather than their actual body weight and titrated to the desired level of sedation to prevent significant drug accumulation and oversedation.
The authors thank Eran Geller, M.D., M.S., Medical Director, Intensive Care Unit Service, VA Palo Alto Health Care System, Palo Alto, California, and Professor of Anesthesia, Stanford University, Stanford, California, and the Medical and Surgical Intensive Care Unit nursing staff at the VA Palo Alto Health Care System for their assistance and support during this study.
References
Albanese J, Martin C, Lacarelle B, Saux P, Durand A, Gouin F: Pharmacokinetics of long-term propofol infusions used for sedation in ICU patients. A nesthesiology 1990; 73: 214–7Albanese, J Martin, C Lacarelle, B Saux, P Durand, A Gouin, F
McMurray TJ, Collier PS, Carson IW, Lyons SM, Elliott P: Propofol sedation after open hear surgery: A clinical and pharmacokinetic study. Anaesthesia 1990; 45: 322–6McMurray, TJ Collier, PS Carson, IW Lyons, SM Elliott, P
Bailie GR, Cockshott ID, Douglas EJ, Bowles BJ: Pharmacokinetics of propofol during and after long-term continuous infusion for maintenance of sedation in ICU patients. Br J Anaesth 1992; 68: 486–91Bailie, GR Cockshott, ID Douglas, EJ Bowles, BJ
Frenkel C, Schuttler J, Ihmsen H, Heye H, Rommelsheim K: Pharmacokinetics and pharmacodynamics of propofol/alfentanil infusion for sedation in ICU patients. Intensive Care Med 1995; 21: 981–8Frenkel, C Schuttler, J Ihmsen, H Heye, H Rommelsheim, K
Shafer A, Doze VA, Shafer SL, White PF: Pharmacokinetics and pharmacodynamics of propofol infusions during general anesthesia. A nesthesiology 1988; 69: 348–56Shafer, A Doze, VA Shafer, SL White, PF
Schuttler J, Stoeckel H, Schwilden H: Pharmacokinetic and pharmacodynamic modeling of propofol (‘Diprivan’) in volunteers and surgical patients. Postgrad Med J 1985; 61: 53–4Schuttler, J Stoeckel, H Schwilden, H
Gepts E, Camu F, Cockshott ID: Disposition of propofol administered as a constant-rate infusion in humans. Anesth Analg 1987; 66: 1256–63Gepts, E Camu, F Cockshott, ID
Beller JP, Pottecher T, Lugnier A, Mangin P, Otteni JC: Prolonged sedation with propofol in ICU patients: Recovery and blood concentration changes during periodic interruptions in infusion. Br J Anaesth 1988; 61: 583–8Beller, JP Pottecher, T Lugnier, A Mangin, P Otteni, JC
Barrientos-Vega R, Sanchez-Soria MM, Morales-Garcia C, Robas-Gomez A, Cuena-Boy R, Ayensa-Rincon A: Prolonged sedation of critically ill patients with midazolam or propofol: Impact on weaning and costs. Crit Care Med 1997; 25: 33–40Barrientos-Vega, R Sanchez-Soria, MM Morales-Garcia, C Robas-Gomez, A Cuena-Boy, R Ayensa-Rincon, A
Chamorro C, deLatorre FJ, Montero A, Sanchez-Izquierdo JA, Jareno A, Moreno JA, Gonzalez E, Barrio M, Carpintero JL, Martin-Santos F, Otero B, Ginestal R: Comparative study of propofol versus midazolam in the sedation of critically ill patients: Results of a prospective, randomized, multicenter trial. Crit Care Med 1996; 24: 932–9Chamorro, C deLatorre, FJ Montero, A Sanchez-Izquierdo, JA Jareno, A Moreno, JA Gonzalez, E Barrio, M Carpintero, JL Martin-Santos, F Otero, B Ginestal, R
Weinbroum AA, Halpern P, Rudick V, Sorkine P, Freedman M, Geller E: Midazolam versus propofol for long-term sedation in the ICU: A randomized prospective comparison. Intensive Care Med 1997; 23: 1258–63Weinbroum, AA Halpern, P Rudick, V Sorkine, P Freedman, M Geller, E
Ronan KP, Gallagher JT, George B, Hamby B: Comparison of propofol and midazolam for sedation in intensive care unit patients. Crit Care Med 1995; 23: 286–93Ronan, KP Gallagher, JT George, B Hamby, B
Ramsay MAE, Savege TM, Simpson BRJ, Goodwin R: Controlled sedation with alphaxalone-alphadolone. Br Med J 1974; 2: 656–9Ramsay, MAE Savege, TM Simpson, BRJ Goodwin, R
Plummer GF: Improved method for the determination of propofol in blood by high performance liquid chromatography with fluorescein detection. J Chromatogr 1987; 421: 171–6Plummer, GF
Gustafsson LL, Ebling WF, Osaki E, Harapat S, Stanski DR, Shafer SL: Plasma concentration clamping in the rat using a computer-controlled infusion pump. Pharm Res 1992; 9: 800–7Gustafsson, LL Ebling, WF Osaki, E Harapat, S Stanski, DR Shafer, SL
Robinson BJ, Ebert TJ, O’Brien TJ, Conlinco MD, Muzi M: Mechanisms whereby propofol mediates peripheral vasodilation in humans. A nesthesiology 1997; 86: 64–72Robinson, BJ Ebert, TJ O’Brien, TJ Conlinco, MD Muzi, M
Higgins TL, Vared J, Estafanous FG, Coyle JP, Haumei KK, Goodale DB: Propofol versus midazolam for intensive care unit sedation after coronary artery bypass grafting. Crit Care Med 1994; 22: 1415–23Higgins, TL Vared, J Estafanous, FG Coyle, JP Haumei, KK Goodale, DB
Lowrey TS, Dunlap AW, Brown RO, Dickerson RN: Pharmacologic influence on nutrition support therapy: Use of propofol in a patient receiving combined enteral and parenteral nutrition support. Nutr Clin Pract 1996; 11: 147–9Lowrey, TS Dunlap, AW Brown, RO Dickerson, RN
Eddleston JM, Shelly MP: The effect on serum lipid concentrations of a prolonged infusion of propofol–hypertriglyceridaemia associated with propofol administration. Intensive Care Med 1991; 17: 424–6Eddleston, JM Shelly, MP
Valente JF, Anderson GL, Branson RD, Johnson DJ, Davis K, Porembka DT: Disadvantages of prolonged propofol sedation in the critical care unit. Crit Care Med 1994; 22: 710–2Valente, JF Anderson, GL Branson, RD Johnson, DJ Davis, K Porembka, DT
Cammarano WB, Pittet J, Weitz S, Schlobohm RM, Marks JD: Acute withdrawal syndrome related to the administration of analgesic and sedative medications in adult intensive care unit patients. Crit Care Med 1998; 26: 676–84Cammarano, WB Pittet, J Weitz, S Schlobohm, RM Marks, JD
Buckley PM: Propofol in patients needing long-term sedation in intensive care: An assessment of the development of tolerance. Intensive Care Med 1997; 23: 969–74Buckley, PM
Fig. 1. Propofol titration scheme using a target-controlled, intravenous infusion system. X = stepwise changes in the target plasma propofol concentration over time; SS = sedation score.
Fig. 1. Propofol titration scheme using a target-controlled, intravenous infusion system. X = stepwise changes in the target plasma propofol concentration over time; SS = sedation score.
Fig. 1. Propofol titration scheme using a target-controlled, intravenous infusion system. X = stepwise changes in the target plasma propofol concentration over time; SS = sedation score.
×
Fig. 2. Residual error plots as a measure of performance for the revised pharmacokinetic model in (A  ) the learning group (n = 19) and (B  ) the test group (n = 10), expressed as measured/predicted plasma propofol concentrations over time during the infusion and postinfusion periods. For each subject, solid lines depict the infusion period and dashed lines, the postinfusion period.
Fig. 2. Residual error plots as a measure of performance for the revised pharmacokinetic model in (A 
	) the learning group (n = 19) and (B 
	) the test group (n = 10), expressed as measured/predicted plasma propofol concentrations over time during the infusion and postinfusion periods. For each subject, solid lines depict the infusion period and dashed lines, the postinfusion period.
Fig. 2. Residual error plots as a measure of performance for the revised pharmacokinetic model in (A  ) the learning group (n = 19) and (B  ) the test group (n = 10), expressed as measured/predicted plasma propofol concentrations over time during the infusion and postinfusion periods. For each subject, solid lines depict the infusion period and dashed lines, the postinfusion period.
×
Fig. 3. Median and worst individual performances for the revised pharmacokinetic model in (A  ) the learning group and (B  ) the test group. The accuracy of the revised model is preserved when tested prospectively in individuals in the test group.
Fig. 3. Median and worst individual performances for the revised pharmacokinetic model in (A 
	) the learning group and (B 
	) the test group. The accuracy of the revised model is preserved when tested prospectively in individuals in the test group.
Fig. 3. Median and worst individual performances for the revised pharmacokinetic model in (A  ) the learning group and (B  ) the test group. The accuracy of the revised model is preserved when tested prospectively in individuals in the test group.
×
Fig. 4. Predicted probability of sedation for a sedation score (SS) ≥ 4. Comparing the performance of the derived pharmacodynamic models in the test group (n = 10) using naïve pooled data (NPD; dashed line) and mixed-effects modeling (MEM; solid line) approaches. Measured probabilities are derived from observed sedation scores at various plasma propofol concentrations using the technique of Somma et al.  The mixed-effects pharmacodynamic model provides a superior graphic fit of the data.
Fig. 4. Predicted probability of sedation for a sedation score (SS) ≥ 4. Comparing the performance of the derived pharmacodynamic models in the test group (n = 10) using naïve pooled data (NPD; dashed line) and mixed-effects modeling (MEM; solid line) approaches. Measured probabilities are derived from observed sedation scores at various plasma propofol concentrations using the technique of Somma et al. 
	The mixed-effects pharmacodynamic model provides a superior graphic fit of the data.
Fig. 4. Predicted probability of sedation for a sedation score (SS) ≥ 4. Comparing the performance of the derived pharmacodynamic models in the test group (n = 10) using naïve pooled data (NPD; dashed line) and mixed-effects modeling (MEM; solid line) approaches. Measured probabilities are derived from observed sedation scores at various plasma propofol concentrations using the technique of Somma et al.  The mixed-effects pharmacodynamic model provides a superior graphic fit of the data.
×
Fig. 5. Probability curves for discrete sedation scores (i.e.,  SS = 1, 2, . . . 6). The distribution of probability curves is asymmetric. The peak probability (modes) of each curve defines the plasma propofol concentrations at which a discrete level of sedation is most likely to occur.
Fig. 5. Probability curves for discrete sedation scores (i.e., 
	SS = 1, 2, . . . 6). The distribution of probability curves is asymmetric. The peak probability (modes) of each curve defines the plasma propofol concentrations at which a discrete level of sedation is most likely to occur.
Fig. 5. Probability curves for discrete sedation scores (i.e.,  SS = 1, 2, . . . 6). The distribution of probability curves is asymmetric. The peak probability (modes) of each curve defines the plasma propofol concentrations at which a discrete level of sedation is most likely to occur.
×
Fig. 6. The predicted time required for plasma propofol concentration (PPC) to decrease by 88% and 58% as a function of the duration of the propofol infusion. An 88% decrease is required for emergence from deep sedation (i.e.,  sedation score = 5 → 2). A 58% decrease is required for emergence from light sedation (i.e.,  sedation score = 3 → 2).
Fig. 6. The predicted time required for plasma propofol concentration (PPC) to decrease by 88% and 58% as a function of the duration of the propofol infusion. An 88% decrease is required for emergence from deep sedation (i.e., 
	sedation score = 5 → 2). A 58% decrease is required for emergence from light sedation (i.e., 
	sedation score = 3 → 2).
Fig. 6. The predicted time required for plasma propofol concentration (PPC) to decrease by 88% and 58% as a function of the duration of the propofol infusion. An 88% decrease is required for emergence from deep sedation (i.e.,  sedation score = 5 → 2). A 58% decrease is required for emergence from light sedation (i.e.,  sedation score = 3 → 2).
×
Table 1. Modified Ramsay Sedation Scale 13 
Image not available
Table 1. Modified Ramsay Sedation Scale 13 
×
Table 2. Subject Demographics (n = 30)
Image not available
Table 2. Subject Demographics (n = 30)
×
Table 3. Propofol Infusion Data (n = 30)
Image not available
Table 3. Propofol Infusion Data (n = 30)
×
Table 4. Reasons for Discontinuing Propofol Infusion (n = 30)
Image not available
Table 4. Reasons for Discontinuing Propofol Infusion (n = 30)
×
Table 5. Original versus Revised Pharmacokinetic Parameters for Propofol
Image not available
Table 5. Original versus Revised Pharmacokinetic Parameters for Propofol
×
Table 6. Performance Measures of Revised Pharmacokinetic Model for Propofol*
Image not available
Table 6. Performance Measures of Revised Pharmacokinetic Model for Propofol*
×
Table 7. Summary of Pharmacodynamic Models for Propofol*
Image not available
Table 7. Summary of Pharmacodynamic Models for Propofol*
×
Table 8. Comparison of Estimated C50,SS*to Sedation Score Modesfor Propofol (n = 20)
Image not available
Table 8. Comparison of Estimated C50,SS*to Sedation Score Modesfor Propofol (n = 20)
×
Table 9. Representative Dosing Guidelines for Propofol Infusions to Maintain Light and Deep Sedation in ICU Patients*
Image not available
Table 9. Representative Dosing Guidelines for Propofol Infusions to Maintain Light and Deep Sedation in ICU Patients*
×
Table 10. Comparison of Pharmacokinetic Models for Propofol in ICU Patients
Image not available
Table 10. Comparison of Pharmacokinetic Models for Propofol in ICU Patients
×
Table 11. Propofol Dose–Response Studies for Sedation of ICU Patients
Image not available
Table 11. Propofol Dose–Response Studies for Sedation of ICU Patients
×