Evaluation of the predictive performance of a 'Diprifusor'TCI system

Anaesthesia, 1998, 53, Supplement 1, pages 61–67 ................................................................................................................................................................................................................................................

Evaluation of the predictive performance of a ‘Diprifusor’ TCI system C. F. Swinhoe1, J. E. Peacock2, J. B. Glen3 and C. S. Reilly2 1 Anaesthetic Department, Barnsley District General Hospital, Gawber Road, Barnsley, South Yorkshire, UK 2 University Department of Anaesthesia, The Royal Hallamshire Hospital, Sheffield, South Yorkshire, UK 3 Zeneca Pharmaceuticals, Alderley Park, Macclesfield, Cheshire, UK Summary

The predictive performance of a ‘Diprifusor’ target controlled infusion system for propofol was examined in 46 patients undergoing major surgery, divided into three age groups (18–40, 41–55 and 56–80 years). Measured arterial propofol concentrations were compared with values calculated (predicted) by the target controlled infusion system. Performance indices (median performance error and median absolute performance error) were similar in the three age groups, with study medians of 16.2% and 24.1%, respectively. Mean values for ‘divergence’ and ‘wobble’ were ¹7.6%.h¹1 and 21.9%, respectively. Measured concentrations tended to be higher than calculated concentrations, particularly following induction or an increase in target concentration. The mean (SD) propofol target concentration of 3.5 (0.7) mg.ml¹1 during maintenance was lower in older patients, compared with higher target concentrations of 4.2 (0.6) and 4.3 (0.7) mg.ml¹1 in the two younger age groups, respectively. The control of depth of anaesthesia was good in all patients and the predictive performance of the ‘Diprifusor’ target controlled infusion system was considered acceptable for clinical purposes. Keywords Equipment ; ‘Diprifusor’, target controlled infusion. Anaesthetics, intravenous ; propofol. ...................................................................................... Correspondence to: Dr C. F. Swinhoe ‘Diprifusor’ is a trademark, the property of Zeneca Ltd

Target controlled infusion (TCI) devices allow the anaesthetist to provide anaesthesia by controlling the theoretical (calculated) concentration of the drug in the central compartment. Rapid changes in the depth of anaesthesia are therefore possible with similar ease to that achieved with inhalational anaesthesia. It must be accepted, however, that pharmacokinetic variation between patients will result in variation in the actual concentration of the drug. Unlike inhalational anaesthesia, where the end-tidal concentration of vapours can be monitored, on-line blood concentration monitoring is not practical at the present time. Before the accuracy of computer-assisted infusion systems is accepted, their performance must be verified in a population of patients by confirmation of blood concentrations. The objectives of the present study were to assess the predictive performance of the ‘Diprifusor’ TCI system incorporating the pharmacokinetic parameters proposed by Gepts et al. [1], as modified by Marsh et al. [2], for the Q 1998 Blackwell Science Ltd

administration of propofol. Performance may vary with time and chosen target concentrations. Thus, it was assessed by comparing measured arterial blood concentrations of propofol with concentrations predicted by the ‘Diprifusor’ system in groups of subjects of different ages, at various time-points during anaesthesia and for a range of calculated blood concentrations. In addition, the target concentrations required to achieve clinically acceptable induction and maintenance of anaesthesia were recorded. Methods

A total of 46 patients (ASA I, II and III) undergoing elective surgery expected to last for 2–6 h and requiring cannulation of an artery for direct arterial pressure monitoring, were recruited to this open, non-comparative trial. They were grouped into three age groups: 18–40, 41–55 and 56–80 years. Patients were not studied if they had serious impairment of respiratory, cardiovascular, hepatic, 61

C. F. Swinhoe et al. • Predictive performance Anaesthesia, 1998, 53, Supplement 1, pages 61–67 ................................................................................................................................................................................................................................................

renal, haemopoetic or endocrine function; if their body weight was more than 20% above ideal; or if they were receiving medication likely to influence the course of anaesthesia. The study was approved by the South Sheffield Research Ethics Committee and all patients were required to provide written informed consent. Patients for major head and neck or abdominal surgery were premedicated 1 h before surgery with temazepam 20 mg. Neurosurgical cases were premedicated with propranolol 1 mg.kg¹1 orally the night before (22.00 h) and on the day of surgery (06.00 h), unless contraindicated, and with glycopyrrolate 0.2 mg on the morning of surgery. A radial artery cannula was inserted before induction of anaesthesia using local infiltration anaesthesia for the monitoring of arterial pressure and collection of blood samples for the analysis of propofol concentrations by high-performance liquid chromatography [3]. Immediately before induction of anaesthesia with propofol, all subjects received alfentanil 30 mg.kg¹1. Propofol was administered with a prototype ‘Diprifusor’ system in which ‘Diprifusor’ TCI software was incorporated in an external purpose-built computer (backbar computer) linked via a serial port to a Graseby 3400 syringe pump. The propofol blood target concentration (CT) for induction of anaesthesia was set at 4 mg.ml¹1 in the two younger age groups and 3 mg.ml¹1 in the oldest age group. If anaesthesia was not induced within 5 min, the CT was increased sufficiently to complete the induction of anaesthesia, as directed in the protocol. Following loss of consciousness, vecuronium 0.1 mg.kg¹1 was given to facilitate endotracheal intubation. Supplementary doses (0.025 mg.kg¹1) were given as required, or an infusion of vecuronium was used with monitoring using train-of-four nerve stimulation. The patients’ lungs were mechanically ventilated with a nitrous oxide/oxygen (2:1) mixture to maintain an arterial oxygen saturation above 95% and an end-tidal CO2 concentration between 3.5 and 5.5 kPa. During maintenance of anaesthesia, the propofol CT could be increased or decreased on the basis of clinical judgement of the depth of anaesthesia in relation to the degree of surgical stimulation. In the absence of signs of light anaesthesia, it was titrated downwards to avoid maintaining propofol concentrations higher than clinically necessary. Analgesia was provided with a continuous infusion of alfentanil at a rate of 0.5–2 mg.kg¹1.min¹1. The propofol and alfentanil infusions were given via a dedicated intravenous cannula; other cannulae and monitoring were used as required. At the end of the procedure, nitrous oxide and the infusions of propofol and alfentanil were discontinued. Residual neuromuscular blockade was reversed with neostigmine 2.5 mg and glycopyrrolate 0.5 mg. 62

Arterial blood samples for estimation of propofol concentrations were taken at the following intervals: before and at 2 and 5 min after the start of the infusion; immediately before surgical incision; immediately before and 2 and 5 min after a significant increase (> 25%) in CT; immediately before a significant decrease (> 1.5 mg.ml¹1) in CT, at the time of and 3 min after achieving this lower target; before and 2 and 5 min after stopping the infusion; and on eye opening. As far as possible, any change in CT was made at least 30 min after the previous change. The propofol concentration calculated by the TCI system was recorded every time a sample was taken. In addition, the quality of induction was assessed as: good (smooth induction, no problems); adequate (minor problems, but easily managed) or poor (significant problems). Maintenance of anaesthesia, the overall ease of control of anaesthesia and cardiovascular stability were also assessed, and the total doses of alfentanil and propofol were recorded. At each time-point where calculated and measured concentrations were available, the percentage performance error was calculated as: Performance error (%) ¼

CM ¹ CCALC × 100 CCALC

where CM and CCALC are the measured and calculated blood propofol concentrations, respectively. Performance error data were evaluated as recommended by Varvel et al. [4]. The median performance error (MDPE) is a signed value and represents the direction (over- or underprediction) of the performance error (bias) rather than the size of the error (precision), which is represented by the median absolute performance error (MDAPE). Other measures of performance are ‘divergence’, which reflects any timerelated changes in performance, and ‘wobble’, a measure of intrasubject variability in performance error. For each group and the overall population, median values for MDPE and MDAPE were calculated from individual patient medians. For all four performance measures, MDPE, MDAPE, divergence and wobble, mean values with 95% confidence limits were calculated for the study population. The distribution of performance errors within the study population was estimated by calculating within-subject and between-subject variances of log (measured/calculated) blood concentrations. Performance errors were evaluated further using box and whisker plots and scatter diagrams. The relationship between age and indices of performance was examined using regression analysis. Propofol CT values were summarised as time-weighted averages. Descriptive statistics were calculated as appropriate for each variable, dependent on the distribution of the variable. No formal statistical tests were performed. Q 1998 Blackwell Science Ltd

Anaesthesia, 1998, 53, Supplement 1, pages 61–67 C. F. Swinhoe et al. • Predictive performance ................................................................................................................................................................................................................................................

Table 1 Patient characteristics. Values are mean (SD). Age group; years

Age; years Sex; M:F Weight; kg Height; cm ASA; I:II:III Duration of anaesthesia; min

18–40

41–55

56–80

29 (6.6) 6:4 68 (17.4) 171 (13.1) 2:8:0 199 (98)

48 (4.6) 6:9 69 (14.1) 170 (14.1) 1:13:1 190 (99)

68 (6.7) 18:3 72 (10.3) 173 (6.3) 0:19:2 256 (127)

Results

Table 1 presents the demographic data, by age group, of the 46 patients enrolled in the study. There were no significant differences between the three groups other than age and a preponderance of male patients in the oldest age group. The duration of maintenance infusion was similar across the groups (Table 1). One male patient in the older age group was withdrawn as a result of severe intra-operative haemorrhage and massive blood transfusion. Performance indices (MDPE and MPAPE) for the three age groups and for all patients are shown in Table 2. The values were similar in the age groups, with overall median values of 16.2% for MDPE and 24.1% for MDAPE. For the study population, the mean values (95% confidence intervals) were 20.2% (12.1–28.4%) for MDPE; 31.1% (25.3– 36.8%) for MDAPE; ¹7.6%.h¹1 (¹12.8%.h¹1 to ¹2.5%.h¹1) for divergence; and 21.9% (19.1–24.7%) for wobble. Regression analysis failed to reveal a linear relationship between age and any of the four indices of performance. Figure 1 shows a scatter diagram with measured propofol concentrations plotted against the calculated (predicted) values displayed when samples were collected. The 68% confidence interval for the ratio of measured to calculated concentrations extended from 0.79 to 1.63 (mean 1.14). Thus, if one sample per patient was taken, 68% of blood propofol concentrations would be expected to lie within ¹21–63% of the model prediction. Consistent with the positive value for MDPE, measured concentrations tended to be greater than calculated

Table 2 MDPE and MDAPE by age group and for all patients. Values are group medians (range).

Age group; years

MDPE;% MDAPE;%

Q 1998 Blackwell Science Ltd

concentrations. However, of the 586 determinations of arterial blood propofol concentration in this study, only 23 (3.9%) indicated performance errors greater than 100%. Figure 2 shows the variation in performance error at different time periods. The errors were greater in the early part of the study and after an increase in CT (period 5). Following a decrease in CT, there was a tendency for overprediction of the true value (period 8). When MDPE (one value per patient) was plotted against the calculated concentration for the middle maintenance period, no obvious relationship was seen (Fig. 3). However, when samples from all patients were examined at specific calculated concentrations, performance error tended to increase at the higher concentrations (Fig. 4). Examples of results obtained from individual patients showing the best, median and worst performance (based on ranked MDAPE values) are shown in Fig. 5. The propofol CT was adequate for induction in all but one patient in group 3, who required an increase in concentration to 4 mg.ml¹1. The mean (SD) induction dose of propofol was 1.10 (0.05), 1.15 (0.23) and 0.84 (0.18) mg.kg¹1 in the three groups, respectively, and the mean (range) induction time was 58 (40–86), 67 (36–240) and 58 (37–210) s, respectively. The quality of control of induction was assessed as good in all but one patient, in whom it was considered adequate. The overall mean (SD) maintenance CT was higher in the younger age groups [4.2 (0.6) and 4.3 (0.7) mg.ml¹1, respectively] than in the older age group [aged 56–80 years; 3.5 (0.7) mg.ml¹1]. This resulted in overall lower mean (SD) rates of administration of propofol in the oldest age group [8.9 (1.4), 9.1 (1.6) and 7.3 (1.5) mg.kg¹1.h¹1 in the three groups, respectively]. The mean (SD) alfentanil infusion rates in the three groups were 0.58 (0.16), 0.53 (0.17) and 0.49 (0.18) mg.kg¹1.min¹1, respectively. The quality of maintenance and ease of control of anaesthesia were assessed as good in all patients. Haemodynamic effects and recovery times were similar to those expected with conventional administration techniques using the same drugs. An episode of sinus bradycardia was noted in one patient and hypotension occurred in four patients, all in the oldest age group.

18–40

41–55

56–80

All patients

13.8 (¹20.7–63.6) 22.6 (6.3–63.6)

17.7 (¹15.8–46.4) 25.0 (13.5–48.4)

16.2 (¹11.1–84.1) 24.2 (10.8–84.1)

16.2 (¹20.7–84.1) 24.1 (6.3–84.1)

63

C. F. Swinhoe et al. • Predictive performance Anaesthesia, 1998, 53, Supplement 1, pages 61–67 ................................................................................................................................................................................................................................................

Measured concentration (µg.ml-1)

16 14 12 10 8 6 4

Figure 1 All measured propofol

2 0 0

1

2

3 4 5 6 Calculated concentration (µg.ml -1)

There was a technical problem with the infusion system in one patient. An apparent communication failure, resulting in an inability to cancel the Graseby pump alarm, selfcorrected and the pump restarted, giving the appropriate bolus to regain the target concentration. The problem was attributable to a mains failure and an uncharged battery in the infusion pump. Discussion

Performance errors were similar across the three age groups, as shown by the values for MDPE and MDAPE.

Performance error (%)

200

100

0

-100

0

1

2

3

4

5 6 7 8 Time period

9

10 11 12 13

Figure 2 Box and whisker plots of performance error (%) related to different time periods. The horizontal line within each box is the median; lower and upper boundaries are the 25th and 75th percentiles; vertical whiskers extend over the range; dots show outliers. Time periods: 1 ¼ 2 min infusion; 2 ¼ 5 min infusion; 3 ¼ 1 min before incision; 4 ¼ 1 min before CT increase; 5 ¼ 2 min after increase; 6 ¼ 5 min after increase; 7 ¼ 1 min before CT decrease; 8 ¼ time CT reached; 9 ¼ 3 min after CT reached; 10 ¼ 1 min before infusion end; 11 ¼ 2 min after stopping infusion; 12 ¼ 5 min after stopping infusion; 13 ¼ eyes open.

64

7

8

concentrations against calculated concentrations, with the line of identity (solid line) and limits of the 68% population distribution (dotted line).

The positive value for MDPE indicates a tendency for measured blood concentrations of propofol to be higher than calculated concentrations. The performance error also tended to be greater and positive after induction or an increase in concentration and to become negative after a reduction in CT (Fig. 2). The difference between measured and calculated blood concentrations tended to decrease during maintenance of anaesthesia, i.e. MDPE moved towards zero, whether from a positive value after an increase in CT or a negative value after a decrease. This is demonstrated overall by the negative divergence (¹7.6%.h¹1) seen for all patients. The tendency for performance error values to increase at higher calculated concentrations (Fig. 4) may be related to incomplete mixing at greater rates of drug delivery. Many of the values obtained at lower calculated concentrations were observed as the calculated concentration fell towards a lower target at the end of the procedure. An improved predictive performance during zero pump flow has been observed in another study [5]. It is not possible to measure effector-site concentrations of propofol. These are likely to be lower than blood concentrations immediately after increases in CT and higher immediately after decreases in it. In both cases, this would tend to reduce the bias if measured concentrations could be compared with effector-site concentrations. Six patients in the study received propranolol, a drug that may affect hepatic clearance [6]. However, when data from these patients were excluded from the analysis, the MDAPE and MDPE values were only slightly lower, at 23.8% and 14.9%, respectively, indicating that any conclusions based on the results from the overall study population are valid. Other studies have examined the predictive performance of TCI systems using the same pharmacokinetic parameters as those in ‘Diprifusor’. Davidson et al. [7] Q 1998 Blackwell Science Ltd

Anaesthesia, 1998, 53, Supplement 1, pages 61–67 C. F. Swinhoe et al. • Predictive performance ................................................................................................................................................................................................................................................

Figure 3 MDPE (%) against the time-averaged middle maintenance period-calculated propofol concentration in each patient.

found mean values of bias and precision of 21% and 30%, respectively, which are similar to those obtained in the present study. On the other hand, Coetzee et al. [8] found a minimal degree of bias (median MDPE ¹7%) when propofol administered by TCI was supplemented with sufentanil rather than alfentanil, as in the present study. Clearly further studies are required to clarify the influence of ancillary agents on the pharmacokinetics of propofol and the predictive performance of the ‘Diprifusor’ system. The bias with this system (MDPE 16%) is smaller than the difference between end-tidal and arterial partial pressures of inhalational anaesthetics: after 15 min of isoflurane administration, a mean ratio of arterial to end-tidal partial pressure of 0.78 has been observed [9]. After 1 h of administration, the arterial concentration remained about 20% lower than the end-tidal concentration. There is obviously a greater discrepancy between inspired and arterial partial pressure [10]. Some performance error with TCI systems may be due to errors in the blood sampling technique or variability in the measurement of blood concentrations. The outlying results shown in Figs 2 and 4 have been

included in the overall calculations. This also goes some way towards explaining the wide range of MDPE and MDAPE in Table 2, which may be partly explained by incomplete mixing at greater rates of drug delivery and the inability to measure effector site concentration of propofol as discussed earlier. The principal source of error, however, is pharmacokinetic variability between patients, which may be genetic or a result of haemodynamic variations during anaesthesia or the administration of other medications. Information on the predictive performance of any TCI system is important to allow comparison with other systems and to provide a baseline for possible future improvements. From a clinical point of view, information on bias and precision is of little practical value. Interpatient pharmacokinetic variability must always be expected, and titration of the propofol CT to achieve a desired effect will be essential to account for this. Such titration is facilitated by the ability to make proportional changes in blood concentration quickly to achieve the desired effect with the ‘Diprifusor’ TCI system. The small amount of divergence noted in this study indicates that the amount of titration required to

Figure 4 Box and whisker plots of the performance error (%) related to particular calculated concentrations.

Q 1998 Blackwell Science Ltd

65

C. F. Swinhoe et al. • Predictive performance Anaesthesia, 1998, 53, Supplement 1, pages 61–67 ................................................................................................................................................................................................................................................

Figure 5 Plots from three individual

patients of measured concentrations (dots) superimposed on calculated concentration profile (solid line) indicated by the ‘Diprifusor’ system. (a) Best; (b) median; (c) worst performance.

accommodate any bias resulting from a mismatch between the patient and the model is likely to be reasonably constant over time. More important than accuracy is the desirability of standardising the pharmacokinetic model, and hence the performance of ‘Diprifusor’ [11], so that there is a transferable learning curve for all such delivery systems for propofol. 66

It has been suggested that the performance of a TCI system is clinically acceptable if the bias (MDPE) is no greater than 10–20% [12] and the MDAPE falls between 20% and 40% [12, 13]. Assessed on these criteria and the good control of depth of anaesthesia achieved in this study, the accuracy of the ‘Diprifusor’ TCI system can be considered clinically acceptable. Q 1998 Blackwell Science Ltd

Anaesthesia, 1998, 53, Supplement 1, pages 61–67 C. F. Swinhoe et al. • Predictive performance ................................................................................................................................................................................................................................................

References 1 Gepts E, Camu F, Cockshott ID, Douglas EJ. Disposition of propofol administered as constant intravenous infusions in humans. Anesthesia and Analgesia 1987; 66: 1256–63. 2 Marsh B, White M, Morton N, Kenny GNC. Pharmacokinetic model driven infusion of propofol in children. British Journal of Anaesthesia 1991; 67: 41–8. 3 Plummer GF. Improved method for the determination of propofol in blood by high-performance liquid chromatography with fluorescence detection. Journal of Chromatography and Biomedical Applications 1987; 421: 171–6. 4 Varvel JR, Donoho DL, Shafer SL. Measuring the predictive performance of computer-controlled infusion pumps. Journal of Pharmacokinetics and Biopharmaceutics 1992; 20: 63–94. 5 White M, Kenny GNC. Evaluation of a computerised propofol infusion system. Anesthesiology 1989; 71: A278. 6 Schneck DW, Luderer JR, Davis D, Vary J. Effects of nadolol and propranolol on plasma lidocaine clearance. Clinical Pharmacology and Therapeutics 1984; 36: 584–7. 7 Davidson JAH, MacLeod AD, Howie JC, White M,

Q 1998 Blackwell Science Ltd

8

9

10

11 12

13

Kenny GNC. Effective concentration 50 for propofol with and without 67% nitrous oxide. Acta Anaesthesiologica Scandinavica 1993; 37: 458–64. 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. Frei FJ, Zbinden AM, Thomson DA, Reider HU. Is the end-tidal partial pressure of isoflurane a good predictor of its arterial partial pressure? British Journal of Anaesthesia 1991; 66: 331–9. Dwyer RC, Fee JPH, Howard PJ, Clarke RSJ. Arterial washin of halothane and isoflurane in young and elderly adult patients. British Journal of Anaesthesia 1991; 66: 572–9. Glen JB. Development of ‘Diprifusor’: a TCI system for propofol. Anaesthesia 1998; 53 (Supplement 1): 13–21. Schu¨ttler J, Kloos S, Schwilden H, Stoeckel H. Total intravenous anaesthesia with propofol and alfentanil by computer assisted infusion. Anaesthesia 1988; 43 (Supplement): 2–7. Glass PJA, Jacobs JR, Reeves JG. Intravenous drug delivery. In: Milder RD, ed. Anaesthesia, 3rd edn. New York: Churchill Livingstone, 1990: 367–88.

67