Precision and bias of target controlled propofol infusion for sedation{dagger}

M. A. Frölich*, D. M. Dennis, J. A. Shuster and R. J. Melker

Departments of Anesthesiology, Statistics, Pediatrics, and Biomedical Engineering, University of Florida Colleges of Medicine and Engineering, Gainesville, Florida, USA

* Corresponding author. University of Alabama, Department of Anesthesiology, University of Alabama at Birmingham, Jefferson Tower, Room 920, 619 South 19th Street, Birmingham, AL 35249-0001, USA. E-mail: froelich{at}uab.edu

Accepted for publication November 18, 2004.


    Abstract
 Top
 Footnotes
 Abstract
 Introduction
 Subjects and methods
 Results
 Discussion
 References
 
Background. The purpose of this study is to test precision and systematic bias of a target controlled infusion (TCI) of propofol in human volunteers at two sedative concentrations.

Methods. We studied the ‘Diprifusor’ model (Marsh Pharmacokinetics and a Graseby® 3400 infusion pump) in 18 human volunteers at two sedative target plasma concentrations (0.5 and 1.0 µg ml–1). Twenty minutes after infusion start or change and 20 min after discontinuation of the infusion plasma propofol concentrations were measured using liquid chromatography–mass spectroscopy (LC-MS). Plasma propofol concentrations were compared with concentrations predicted by the TCI system. Agreement of those two measures (precision and bias) was determined using regression analysis.

Results. We found little systematic bias but poor precision. When setting the TCI system to deliver a plasma concentration of 1.0 µg ml–1 one can predict the actual plasma concentration with 95% confidence only within a range of 0.44–1.38 µg ml–1.

Conclusions. This finding helps to explain differences in responses to propofol sedation; pharmacokinetic variability appears to be an important factor.

Keywords: anaesthetics, i.v., propofol ; anaesthetic techniques, computer-assisted continuous infusion ; equipment, infusion pump ; model, computer ; pharmacokinetics, model ; statistics, regression analysis


    Introduction
 Top
 Footnotes
 Abstract
 Introduction
 Subjects and methods
 Results
 Discussion
 References
 
Target controlled infusion (TCI) systems have been developed to provide improved convenience and ease of control during i.v. anaesthesia. The basic principle is that the anaesthetist sets and adjusts the target blood concentration and infusion rates are altered automatically according to a pharmacokinetic model.

TCI systems are based on a three compartment pharmacokinetic model; these compartments are the central compartment (blood or plasma), highly perfused tissue (heart, brain, muscle) and poorly perfused tissue (adipose tissue, bone). At equilibrium, the drug moves from one compartment to another at a constant rate based on inter-compartmental distribution rate constants. These rate constants are used to mathematically predict plasma and brain (effect site) concentrations.

The pharmacokinetics of propofol have been comprehensively studied in the past.13 The pharmacokinetic variability of propofol has been the concern of many investigator and efforts have been made to adjust for some of the sources of this biologic variability (age, body weight, pre-existing medical conditions, genetic, and environmental factors). Therefore, various investigators have refined the infusion model to maximize predictive accuracy of propofol plasma or brain concentration.4 5 The most popular of these models developed by Marsh and colleagues6 was chosen for the commercially available Diprifusor® (Astra Zeneca). This model was found to have good delivery performance in a recent laboratory report when hypnotic propofol concentrations are chosen.7 However, when applied to clinical care TCI systems it may not be as accurate as suggested previously.8 9

TCI systems are used for clinical practice as well as research; in many studies investigators use predicted propofol values as a surrogate for actual plasma concentrations.1012 The accuracy of the Diprifusor® model has been investigated at anaesthestic (hypnotic) doses of propofol but little information is available on the precision of the Diprifusor® over the sedative dosage range.

The aim of this study is to determine precision and bias of the Diprifusor® model at sedative concentrations.


    Subjects and methods
 Top
 Footnotes
 Abstract
 Introduction
 Subjects and methods
 Results
 Discussion
 References
 
After institutional review board approval, 18 human volunteers were recruited by public advertisement. During a screening visit, subjects underwent physical examination, gave a medical history and underwent laboratory screening with a basic metabolic panel, a complete blood count and, if female, a urine pregnancy test. Only subjects without medical problems and normal laboratory test results were enrolled. Other exclusion criteria were obesity (BMI >30 kg [m2]–1), age less than 18 yr, a remarkable social history, and the use of any medication on a regular basis. Volunteers participated on a single occasion and were paid for their involvement.

An i.v. line for propofol infusion was placed in one arm and a sampling i.v. line was inserted in the other. Subjects 3–18 received propofol in two sedative effect site concentrations, 1.0 (moderate sedation) and 0.5 (mild sedation) µg ml–1. The first two subjects were started on a slightly higher dose for both levels (1.5, 1.2, and 0.75 µg ml–1). In order to avoid potential problems with over-sedation, doses were reduced and kept consistent at 1.0 and 0.5 µg ml–1 starting with subject 3. The study was designed to alternate mild and moderate sedation with a no drug infusion control condition necessary to obtain psychophysical measurements that are reported separately. There were three of four time points per subject during which blood samples were obtained (illustrated and in Fig. 1). A total of 66 blood samples were obtained and compared with the calculated concentration provided by the TCI software.



View larger version (17K):
[in this window]
[in a new window]
 
Fig 1 Illustration of study sequence. This figure shows the schematic study sequence for three subjects and shows the number of blood samples obtained per subject (arrows). The two bars represent the mild and moderate sedation state and the thick line the ‘no infusion’ condition. The dotted line represents the calculated target plasma concentration of propofol.

 
We used a Graseby infusion pump (model 3400), equipped with a RS-232 port for interfacing a computer and Stanpump software. We used the pharmacokinetic model reported by Marsh.6 Subjects were monitored with electrocardiogram, arterial pressure, and pulse oximetry and received dextrose 5% in saline infusion at 100 ml h–1. Following blood with draw, samples were immediately spun using a centrifuge (25 000 r.p.m. for 10 min). Plasma was separated and stored at –75°C until analysed. Propofol was extracted from plasma using solid-phase extraction. Analyses were performed in triplicate using liquid chromatography–mass spectroscopy (LC-MS). Quality control samples (known samples for determining the accuracy of the method) were 94–98% accurate. The precision of the assay (%CV) varied between 2.6 and 4.1%.

Data were analysed using regression analysis. Measured propofol plasma concentration (LC-MS, gold standard) was the response variable (Y) and the propofol plasma concentration predicted by Stanpump was used as explanatory variable (X).

This allows a more flexible calibration than the Bland–Altman method13 used in other medical publications because it allows a true calibration equation. Bland and Altman consider as errors any literal Y–X difference. Thus, the Bland–Altman method forces the line to have a slope equal to 1.0 and allows an intercept term (which they call bias). If the regression line produces a slope near 1.0 and an intercept near 0.0, then regression will be very similar to Bland–Altman with only a very minor loss of two degrees of freedom for error. However, if the regression line has slope or intercept moderately different from 1.0 and 0.0, respectively, then the regression line will produce a more reliable predictor of the gold standard than the method described by Bland and Altman. Both a linear and a quadratic regression model were tested.


    Results
 Top
 Footnotes
 Abstract
 Introduction
 Subjects and methods
 Results
 Discussion
 References
 
Seven female and 11 male subjects were enrolled. Their age ranged from 19 to 28 yr, height from 156 to 193 cm, and weight from 43 to 90 kg. None of the subjects recruited withdrew from the study. All subjects maintained spontaneous ventilation during and after propofol infusion.

Results of the regression analysis are displayed in Table 1. The regression coefficients for the linear regression were an intercept of 0.09, a slope b of 0.82 and a root mean square (RMS) error of 0.24. The coefficient of determination, R2, was 0.59. The SD of all blood propofol levels was 0.37 µg ml–1. The SD of the Diprifusor® model was 0.24 µg ml–1. The precision of the Diprifusor® can also be described in terms of confidence. If we set our Diprifusor infusion pump to deliver 1.0 µg ml–1, we can be 95% confident that our measured propofol concentration will be between 0.44 and 1.38 µg ml–1; this wide range illustrates the lack of precision (Fig. 2).


View this table:
[in this window]
[in a new window]
 
Table 1 Results of general linear model. ;

 


View larger version (18K):
[in this window]
[in a new window]
 
Fig 2 Scatterplot and fit line. Relationship between propofol concentrations as predicted by a TCI pump (Diprifusor®) (X-variable) and blood propofol concentrations (Y-variable). The prediction equation and R2 is provided. There are also two sets of lines for the 95% confidence (prediction) intervals for both the estimate of the mean of blood propofol data (dashed) and for the actual spread of blood propofol measures around a predicted concentration (solid). The thick solid line is the regression fit. There are two clusters of samples, at the 0.5 µg ml–1 and the 1.0 µg ml–1 mark with several sample points on the lower end of concentrations representing samples obtained after the pump had been turned off.

 
The difference in the mean propofol concentration provided by Stanpump and those measured by LC-MS was negligible (–0.01) indicating that there is no overall bias (over- or underestimation) in the prediction. The quadratic model produced a RMS error of 0.24 µg ml–1, which is no significant improvement over the linear regression model.


    Discussion
 Top
 Footnotes
 Abstract
 Introduction
 Subjects and methods
 Results
 Discussion
 References
 
Target controlled propofol infusion is used extensively both in research and clinical practice. This method is currently the most advanced pharmacologic model for predicting propofol blood concentrations. The Diprifusor® model is commercially available in Europe. Many studies have described the use of propofol TCI in anaesthetic doses14 but little data have been published about its use in the low concentration range (sedation). We noted a negligible bias but poor precision. The SD of the estimated propofol concentration (0.47 µg ml–1) is very close to that noted in a preliminary study by Veselis and colleagues (0.41 µg ml–1)15 despite their use of a different study design and analytical method. This finding reflects the considerable pharmacokinetic variability when TCI is used for sedation.

Information about the precision of propofol TCI is important because this model is widely applied as a surrogate for plasma propofol concentration both for clinical16 and research applications.11 17 A study by Kaike and colleagues18 exemplifies this potential problem; investigators used propofol TCI to predict levels of propofol anaesthesia in 16 subjects to study regional cerebral blood flow as measured by positron emission tomography. Taking predicted propofol levels as surrogate for actual (measured) propofol levels introduces error, which is compounded by the subsequent analysis of regional cerebral blood flow. Disregard for the propofol measurement error certainly presents a challenge for this and many similar studies.

The focus of our study on the sedative dose range and the chosen precision analysis requires comment. Several infusion pumps have been used for the study of TCI performance characteristics. Traditionally the Harvard 22 pump is used. Unfortunately, the latter pump is not approved for clinical use. We therefore selected an infusion pump, the Graseby 3400 pump, which is approved for clinical use. Many pumps used clinically tend to be less accurate at low infusion rates and it is recommended to check infused volumes predicted by software with actually delivered volumes (read at the syringe). We made this comparison before each change in the infusion rate and found the Graseby 3400 pump to perform accurately to within 1 ml of the infused volume. Another point of discussion relates to the evaluation of pump performance. Rather than focusing on a single time point, many investigators are also interested in pump performance over time. A comprehensive method to evaluate computer assisted infusions using temporal characteristics like divergence and wobble has been described by Varvel and co-workers.12 In the present study, we observed data from several subjects at three time points. Thus, only part of the analysis described by Varvel and colleagues—precision and bias—is suitable for our experiment. Precision and bias are derived from a regression analysis and are described using traditional statistical terms.

Another point of discussion relates to blood sampling. Many investigators use arterial blood sampling for pharmacological studies as opposed to venous sampling. One might postulate that arterial propofol sampling might have resulted in a more precise agreement of predicted and measured propofol levels because tissue uptake of propofol might result in an arterio-venous concentration difference and potentially also a difference in the variability of measured propofol concentrations. This issue has been studied by Johnston and colleagues19 who did not find a significant difference in either variance or mean propofol concentrations when comparing arterial and arterialized venous sampling. We used simple venous sampling and sampled from a large antecubital vein without the use of a tourniquet at a comfortable room temperature to avoid subjects shivering. This is a potential limitation of our study results. However, one would expect a systematic bias towards higher (or lower) measured propofol concentrations if tissue extraction indeed were to be a major confounder.

In summary, this study addresses a simple question; what is the precision of the Diprifusor® model 20 min after a sedative infusion is started or changed? We detected that the actual dose delivered has a wide estimated range (0.44–1.38 µg ml–1 if 1 µg ml–1 is targeted); thus, the predicted concentration is not very precise. This finding aids us in understanding why patients appear to react differently to comparable sedative doses of propofol and alerts researchers to use actual rather than predicted propofol levels.


    Footnotes
 Top
 Footnotes
 Abstract
 Introduction
 Subjects and methods
 Results
 Discussion
 References
 
{dagger} Presented in abstract format at the World Congress of Anaesthesiology, Paris, 2004. Back


    References
 Top
 Footnotes
 Abstract
 Introduction
 Subjects and methods
 Results
 Discussion
 References
 
1 White PF. Propofol pharmacokinetics and pharmacodynamics. Semin Anesth 1988; 7: 4–20

2 Kanto J, Gepts E. Pharmacokinetic implications for the clinical use of propofol. Clin Pharmacokinet 1989; 17: 308–26[ISI][Medline]

3 Gepts E, Camu F, Cockshott ID, Douglas EJ. Disposition of propofol administered as constant rate intravenous infusions in humans. Anesth Analg 1987; 66: 1256–63[Abstract]

4 Tackley RM, Lewis GT, Prys-Roberts C, Boaden RW, Dixon J, Harvey JT. Computer controlled infusion of propofol. Br J Anaesth 1989; 62: 46–53[Abstract]

5 Struys MM, De Smet T, Depoorter B, et al. Comparison of plasma compartment versus two methods for effect compartment—controlled target-controlled infusion for propofol. Anesthesiology 2000; 92: 399–406[ISI][Medline]

6 Marsh B, White M, Morton N, Kenny GN. Pharmacokinetic model driven infusion of propofol in children. Br J Anaesth 1991; 67: 41–8[Abstract]

7 Schraag S, Flaschar J. Delivery performance of commercial target-controlled infusion devices with Diprifusor module. Eur J Anaesthesiol 2002; 19: 357–60[ISI][Medline]

8 Hoymork SC, Raeder J, Grimsmo B, Steen PA. Bispectral index, serum drug concentrations and emergence associated with individually adjusted target-controlled infusions of remifentanil and propofol for laparoscopic surgery. Br J Anaesth 2003; 91: 773–80[Abstract/Free Full Text]

9 Fechner J, Albrecht S, Ihmsen H, Knoll R, Schwilden H, Schuttler J. Predictability and precision of ‘target-controlled infusion’ (TCI) of propofol with the ‘Disoprifusor TCI’ system. Anaesthesist 1988; 47: 663–8[CrossRef]

10 Gale T, Leslie K, Kluger M. Propofol anaesthesia via target controlled infusion or manually controlled infusion: effects on the bispectral index as a measure of anaesthetic depth. Anaesth Intensive Care 2001; 29: 579–84[ISI][Medline]

11 Lequeux PY, Cantraine F, Levarlet M, Varvais L. Absence of explicit and implicit memory in unconscious patients using a TCI of propofol. Acta Anaesthesiol Scand 2003; 47: 833–7[CrossRef][ISI][Medline]

12 Varvel JR, Donoho DL, Shafer SL. Measuring the predictive performance of computer-controlled infusion pumps. J Pharmacokinet Biopharm 1992; 20: 63–94[ISI][Medline]

13 Bland JM, Altman DG. Measuring agreement in method comparison studies. Stat Methods Med Res 1999; 8: 135–60[ISI][Medline]

14 Coetzee JF, Glen JB, Wium CA, Boshoff L. Pharmacokinetic model selection for target controlled infusions of propofol. Assessment of three parameter sets. Anesthesiology 1995; 82: 1328–45[ISI][Medline]

15 Veselis RA, Glass P, Dnistrian A, Reinsel R. Performance of computer-assisted continuous infusion at low concentrations of intravenous sedatives. Anesth Analg 1997; 84: 1049–57[Abstract]

16 Passot S, Servin F, Allary R, et al. Target-controlled versus manually-controlled infusion of propofol for direct laryngoscopy and bronchoscopy. Anesth Analg 2002; 94: 1212–6[Abstract/Free Full Text]

17 Absalom AR, Kenny GN. Closed loop control of propofol anaesthesia using bispectral index: performance assessment in patients receiving computer-controlled propofol and manually controlled remifentanil infusions for minor surgery. Br J Anaesth 2003; 90: 737–41[Abstract/Free Full Text]

18 Kaike KK, Metsähonkala L, Teräs M, et al. Effects of surgical levels of propofol and sevoflurane anesthesia on cerebral blood flow in healthy subjects studied with positron emission tomography. Anesthesiology 2002; 96: 1358–70[CrossRef][ISI][Medline]

19 Johnston KR, Vickers MD, Mapleson WW. Comparison of arterialized venous with arterial blood propofol concentrations during sub-anaesthetic infusions in volunteers. Br J Anaesth 1996; 76: 401–4[Abstract/Free Full Text]