Performance of alfentanil target-controlled infusion in normal and morbidly obese female patients

British Journal of Anaesthesia 19 (4): 551 6 (212) Advance Access publication 24 June 212. doi:1.193/bja/aes211 Performance of alfentanil target-controlled infusion in normal and morbidly obese female patients O. Pérus 1 *, A. Marsot 2, E. Ramain 1, M. Dahman 3,A.Paci 4, M. Raucoules-Aimé 1 and N. Simon 2 1 Pôle Anesthésie-Réanimations, CHU de Nice, Hôpital de l Archet 2, 151, route de Saint-Antoine de Ginestière, BP 379, 622 Nice Cedex 3, France 2 Service de Pharmacologie Médicale et Clinique AP-HM, Aix-Marseille Université, Marseille, France 3 Service de Chirurgie Viscérale, CHG La Fontonne, Antibes, France 4 Service interdépartemental de Pharmacologie et d Analyse du Médicament (SiPAM), Institut de cancérologie Gustave Roussy, 114 rue Edouard Vaillant, 9485 Villejuif cedex, France * Corresponding author. E-mail: ol_perus@yahoo.fr Editor s key points All alfentanil pharmacokinetic (PK) models are derived from populations with normal BMI. Maitre and colleagues alfentanil target-controlled infusion is acceptable for clinical application to morbidly obese patients. Maitre and colleagues model should be applied with caution to that population of patients, especially when high-concentration targets are applied. Data from this study were also accurately described by a new threecompartment model with BMI as the covariate. Background. Available alfentanil pharmacokinetic (PK) sets for target-controlled infusion (TCI) were derived from populations with normal BMI. The performance and accuracy of the models devised by Maitre and colleagues and Scott and colleagues were evaluated in a population including morbidly obese patients. Methods. Alfentanil TCI using Maitre and colleagues model was administered to 1 obese and six non-obese women (BMI 19.5 57.4 kg m 22 ) undergoing laparoscopic surgery. The initial effect-site target concentration was 1 ng ml 21. Alfentanil arterial plasma concentrations were sampled from TCI onset to 22 min after its termination. Stanpump w software calculated predicted alfentanil concentrations. Data were analysed with a non-linear mixed-effect model (NONMEM, version 7.2), including calculations of the median performance error (MDPE) and the median absolute performance error (MDAPE). Scott and colleagues model was evaluated retrospectively. Results. Using Maitre and colleagues model, MDPE and MDAPE (range) for the whole population were 13.3% and 23.9%, respectively. With Scott and colleagues model, MDPE and MDAPE were 23.7% and 5.1%, respectively. We created a three-compartment model with BMI as the covariate (CL), yielding MDPE 1.1% and MDAPE 3.6%. Conclusions. Maitre and colleagues PK set underestimated the predicted concentrations in our mixed-weighted population, but its bias and accuracy were acceptable for clinical application. Scott and colleagues model was inaccurate. The NONMEM model seemed to be more accurate during the infusion and for high concentrations, but it needs to be validated in a larger population. Keywords: alfentanil; opioids; overweight; pharmacokinetics; target-controlled infusion Accepted for publication: 9 April 212 Obesity induces physiological modifications that are likely to affect drug tissue distribution and elimination. 1 In clinical practice, to avoid the risk of tissue drug accumulation, 2 the anaesthetist has to rely on short-action substances. Among opioids, alfentanil is useful, because of its low liposolubility and its rapid efficacy. The benefits of alfentanil target-controlled infusion (TCI) may be extended to the obese population, but the available alfentanil pharmacokinetic (PK) models were derived with data obtained from non-obese populations. Among these PK sets, Maitre and colleagues 3 and Scott and colleagues 4 models are the most used in the literature and seem to provide the best accuracy. This study was undertaken to evaluate the accuracy of Maitre and colleagues and Scott and colleagues PK sets using the alfentanil PK parameters described by Maitre and colleagues 3 for morbidly obese and non-obese patients undergoing laparoscopic surgery, and to determine a PK model describing alfentanil PK parameters in this mixed population. Meetings at which the work has been presented: PAGE 29: Population Approach Group in Europe, Saint Petersburg, Russia, June 23 26, 29, abstract no. 145; Congrès de la SFAR: Paris, France, September 23 26, 29, abstract no. R427; P2T 21: Congrès de Physiologie, Pharmacologie et Thérapeutique, Bordeaux, France, March 23 25, 21, abstract no. 244. & The Author [212]. Published by Oxford University Press on behalf of the British Journal of Anaesthesia. All rights reserved. For Permissions, please email: journals.permissions@oup.com

BJA Methods This prospective study was approved by the Nice University Ethics Committee (Comité de Protection des Personnes se livrant à la Recherche Biomédicale, CIR 23 no. 2-35), and registered with EudraCT (ref: 211-1439-21). Written informed consent was obtained from morbidly obese patients undergoing laparoscopic bariatric surgery, and normal-weight patients (BMI,3 kg m 22 ) undergoing laparoscopic surgery. Morbidly obese and non-obese female patients 18 68 yr old with ASA I or II were eligible for enrolment. Morbidly obese patients undergoing bariatric surgery had a BMI of.35 kg m 22 in combination with a comorbidity, or a BMI of.4 kg m 22 without any comorbidity. BMI was defined as: body weight (kg)/height 2 (m 2 ). All patients underwent laboratory tests to exclude significant illness or pregnancy. Exclusion criteria were: history of alcohol abuse or illegal drug use, renal or hepatic disease, gastro-oesophageal reflux, concurrent medication included drugs known to interact significantly with opioids, or cytochrome P45. Anaesthesia procedure One hour before surgery, patients were premedicated with oral hydroxyzine (1 mg). After local anaesthesia (Emla & cream, Astra, Rueil-Malmaison, France), the following blood vessels were cannulated: one forearm vein for administration of anaesthetics, a second forearm vein for fluid replacement, and a radial artery for continuous arterial pressure monitoring and blood samplings. After adequate preoxygenation, anaesthesia was induced with a propofol bolus (2.5 mg kg 21 ) followed after 9 s by alfentanil in the TCI mode. Atracurium (.6 mg kg 21 ) was injected i.v. to facilitate tracheal intubation. Anaesthesia was maintained with desflurane, whose administration was adjusted to obtain the bispectral index values between 4 and 6. Muscle relaxation was monitored with a train-of-four nerve simulator. Alfentanil was administered with a BD Pilot Anesthesia & pump (Fresenius-Viale, Brezins, France) connected via a serial RS-232 interface to a personal computer running Stanpump & software (SL Shafer, Department of Anesthesia, Columbia University, New York, NY, USA). The PK set for alfentanil described by Maitre and colleagues 3 used the total body weight (TBW) to calculate the physiological parameters. The initial alfentanil effect-site concentration was set at 1 ng ml 21, and then 6 ng ml 21 5 min later. The predicted target concentrations (effect-site and plasma) and the corresponding alfentanil infusion rate were recorded every 1 s by Stanpump & software. During surgery, the target concentration was adjusted to mean arterial pressure (MAP) and heart rate (HR): MAP or HR variation exceeding 15% above or below baseline values induced a 1 ng ml 21 increment in the target alfentanil concentration in the same direction; the predicted plasma concentration had to reach the predicted effectsite concentration before a new target could be requested. Alfentanil target concentration was set at ng ml 21 at pneumoperitoneum exsufflation, and Stanpump & software was stopped after the last blood sample. Blood sample processing and alfentanil assay Arterial blood samples (4 ml) for alfentanil concentration measurements were obtained before a target change 1, 5, 3, 6, and 12 min after starting alfentanil infusion, and, 5, 1, 2, 3, 4, 5, 6, 9, 12, 18, and 22 min after its end. The corresponding predicted plasma and effect-site alfentanil concentrations of Stanpump software were recorded at the exact moment of the sampling. All blood samples were immediately heparinized and centrifuged at 48C, and plasma was separated (35 rpm for 1 min) and frozen ( 758C) until analysis. Plasma alfentanil concentrations were determined by high-performance liquid chromatography coupled with mass spectrometry in the single ion-monitoring mode (Département de Pharmacologie Clinique, Institut Gustave-Roussy, Villejuif, France). In accordance with the ICH guidelines, precision (repeatability and intermediate precision) of the method was evaluated. Repeatability (intra-day precision), expressed as the coefficient of variation of repeatability (CVr), was performed for each level of quality control (QC) six times and intermediate precision (interday precision), expressed as the coefficient of variation of intermediate precision (CVi), was evaluated for each level of QC over 5 days. The accuracy was measured by the deviation or bias (%) of the mean found concentration (n¼6) from the actual concentration. The CV values for repeatability (CVr) and for intermediate precision (CVi) were found between.8% and 4.5% and between 1.1% and 6.%, respectively. The bias values representing the accuracy of the method were found below 4.3%. External validation Data were analysed using Excel 23 (Microsoft Corporation, Seattle, WA, USA). The predictive performance of Maitre and colleagues 3 PK model of alfentanil was assessed by using the prediction error (PE). For each blood sample, the per cent PE of the predicted plasma alfentanil concentrations was calculated as follows: PE = Cm Cp 1 Cp where Cm and Cp are the measured and predicted plasma alfentanil concentrations, respectively. As recommended by Varvel and colleagues, 5 intrasubject data analysis consisted of an evaluation of four indicators of predictive performance for the n subjects: (i) The median PE (MDPE): the per cent MDPE reflects the bias of TCI in the ith subject; MDPE i (%) =median{pe ij }, j = 1,... Pérus et al. 552

Alfentanil target-controlled infusion and obesity BJA (ii) The median absolute PE (MDAPE): the per cent MDAPE indicates the TCI precision in the ith subject. MDAPE i (%) =median{ PE ij }, j = 1,..., N i (iii) The divergence: it characterizes performance stability over time. We used the modified test purposed by Glen and Servin: 6 divergence is defined as the slope of the signed PE plotted against time, and is expressed in per cent per hour. (iv) The wobble: it measures the intrasubject variability of PE, and is defined as: Wobble (% min 1 )=median{pe ij MDPE i }, j = 1,..., N i where N i is the number of PE values obtained for the ith subject. A second model was evaluated, the Scott and colleagues 4 model. Based on this PK data set and the infusion rates stored during Maitre and colleagues model evaluation, we calculated the new predicted plasma alfentanil concentrations for each patient. Then, predictive performance was assessed by using the same method. Population PK modelling Population PK analysis was performed using a non-linear mixed effects model as implemented using NONMEM computer program (version 7.2, ICON Solutions, Dublin, Ireland). 7 The first-order conditional estimation method was used throughout. Different structural models for alfentanil PKs were investigated: one, two, and three compartments with linear elimination. Intersubject variability of the different PK parameters was estimated with an exponential error model. Several error models (additive, proportional or both) were investigated to describe residual variability. The performance of the model was judged by both statistical and graphic methods. The minimal value of the objective function as calculated by NONMEM was also used to assess the goodness-of-fit. An increase in the goodness-of-fit is accompanied by a decrease in objective function, and this decrease is asymptotically distributed as a x 2 distribution. Furthermore, standard errors were calculated by the use of the COVARIANCE option of NONMEM. For graphic model diagnostics, the following graphs were compared: observed concentrations (depending variable, DV) vs predictions (PRED), corrected weighted residuals (CWRES) vs time, corrected weighted residuals vs PRED, individual predictions (IPRED) vs DV, and normalized predictive distribution error vs PRED or TIME. Graphics were obtained with R for Windows Software (R-2.14.1, The R Foundation for Statistical Computing). A first analysis was performed to find the base model that best described the data. Once it was defined, the influence of each covariate on the PK parameters was tested via an upward model building. These covariates were height, body weight, lean body mass (LBM) calculated with Janmahasatian and colleagues 8 formula, BMI, and age. The diagnostic plots described above, the change in objective function, and the change in parameter variability were noted to select those which improved the model prediction. A decrease in the objective function value (OFV) of at least 6.61 (x 2 distribution with one degree of freedom for P,.1) relative to the base PK model was required for the addition of a single parameter in the model. Covariates defined as relevant during the screening step were included in a so-called full model. Then a backward elimination procedure was performed in which each covariate was removed in turn from the full model and the difference in OFV between the full and each reduced model was examined. An increase in OFV.7.88 (P,.5) was required to retain the covariate in the final model. Bootstrap procedures were performed using Wings for NONMEM (www.wfn.sourceforge.net) to evaluate the 95% confidence intervals non-parametrically. One thousand bootstrapped data sets were generated by re-sampling subjects from the original data set with replacement. These data sets were analysed using the final model described previously. Finally, the 2.5th and 97.5th percentiles of the parameter estimates were taken to build the 95% bootstrap percentile confidence intervals. Results Ten morbidly obese and six non-obese or overweight patients were included in this study from 24 to 27. One nonobese patient was excluded because of a disconnection between the computer and the electric pump. No adverse effect was encountered during alfentanil administration. All patients were successfully extubated in the recovery room. Patient characteristic data are reported in Table 1. The mean BMI for morbidly obese patients was 43 (5.6) kg m 22. Alfentanil infusion duration was in mean 117 (61) min; total alfentanil dose was 5.3 (2.3) mg. Target-concentration was modified 3 (.5) times during alfentanil infusion. The mean plasma alfentanil concentration measured at the end of the infusion was 7.9 (5.8) ng ml 21. Figure 1 shows the observed alfentanil plasma concentration vs time in obese and non-obese patients. The results of the external validation analysis are reported in Table 2. MDPE and MDAPE were calculated for obese and non-obese patients, for three periods: during and after alfentanil infusion, and for the whole study period. Maitre and colleagues model Maitre and colleagues model underpredicted plasma alfentanil concentrations during and after infusion in both obese and non-obese patients, and the bias differed significantly from zero. Values of MDPE for the whole population were 13.3%, and of MDAPE were 23.9%. Maitre and colleagues PK set performed within acceptable limits during infusion in obese patients, but only after infusion in non-obese patients. Figure 2A shows the performance error vs time in obese and 553

BJA Pérus et al. Table 1 Patient characteristic data Patient Surgery Age (yr) BMI (kg m 22 ) Alfentanil Infusion duration (min) Total dose (mg) 1 Gastric bypass 49 42.3 19 7.8 2 Gastric bypass 31 45.7 8 5 3 Gastric bypass 33 42.2 51 4 4 Sleeve gastrectomy 48 57.4 166 7.9 5 Gastric bypass 56 41.2 131 5.6 6 Gastric bypass 48 39 212 7.9 7 Gastric bypass 18 4.7 154 7.1 8 Gastric bypass 37 44.5 136 7.1 9 Gastric bypass 38 38.3 145 7.5 1 Gastric bypass 39 38.9 19 7.6 11 Cholecystectomy 3 21.1 67 3.2 12 Cholecystectomy 54 19.5 49 2.4 13 Cholecystectomy 68 23.9 1 3.5 14 Diagnostic laparoscopy 48 28.9 3 2.3 15 Cholecystectomy 44 26.7 75 3.9 4 35 Obese Non obese Measured concentration (mg litre 1 ) 3 25 2 15 1 5 1 5.11 6 12 19 2 21 22 23 24 25 28 31 37 41 Time (min) Fig 1 Measured alfentanil plasma concentration vs time in obese and non-obese patients, respectively. non-obese patients for Maitre and colleagues model. Figures 3A and 4A show the predicted vs observed plasma alfentanil concentrations and the normalized predicted distribution error vs predicted concentration for the whole period, respectively. Scott and colleagues model Scott and colleagues model underpredicted plasma concentrations during infusion in obese patients, and overestimated these afterwards. The model showed an overall poor performance in both obese and non-obese patients (MDPE were 23.7%, and MDAPE 5.1% for the whole population, respectively). An acceptable performance of Scott and colleagues model was only observed during the infusion period in non-obese patients. NONMEM model A three-compartment model with a first-order elimination provided a better description of the data than did a twocompartment model and was then used as the base 554

555 Table 2 Accuracy analysis of Maitre and colleagues model, Scott and colleagues model, and the NONMEM PK model. Values are percentages. MDPE, median of performance error (bias); MDAPE, absolute median of performance error (accuracy). Divergence is expressed as % min 21 Obese patients Infusion Maitre and colleagues model 3 NONMEM model Scott and colleagues model 4 MDPE MDAPE Wobble Divergence MDPE MDAPE Wobble Divergence MDPE MDAPE Wobble Divergence Mean 25.5 36.5 12.3 2.7 22.6 12.4 7.2. 38.3 4.6 2.2 2.4 25th; 75th percentiles 11.9; 4.9 24.5; 43.1 8.5; 21.9 2.9; 2.4 237.8; 16.7 4.6; 37.8 2.5; 18.6.;. 18.3; 49.9 38.3; 5.3 14.8; 27.1 2.5; 2.2 After infusion Mean 4.1 26.8 7.6.2 8.4 37.8 6.. 245.1 48.6 12.6.1 25th; 75th percentiles 211.4; 42.2 13.3; 42.2 4.5; 11.7.;.3 258.5; 77.8 6.5; 8.4 1.1; 26.8.;. 258.7; 55.8 43.3; 1.5 7.2; 32..1; 1.1 Whole period Mean 31.1 23.8 11.2 2.1 2.2 25.9 9.4. 225.9 52.4 23..1 25th; 75th percentiles 29.7; 37.5 13.3; 4.5 7.9; 14.3 2.2;.1 257.8; 65.3 6.4; 77.1 4.; 32.1.;. 252.; 218.7 43.3; 61.7 14.5; 31.9.;.1 Non-obese patients Infusion Mean 43.6 43.6 1.8 2.6 37.4 4.5 2.4. 215.1 16.5 9.1.1 25th; 75th percentiles 7.; 86.4 25.6; 86.4 4.3; 27.1 25.4;. 214.6; 83. 2.7; 83. 6.; 29.8 2.1;.1 216.5; 26. 15.5; 33.6 7.4; 15.5 2.7;.7 After infusion Mean 216.1 34.2 6.2. 212.9 36.7 8.5. 246.2 46.2 13.4.1 25th; 75th percentiles 231.; 51.8 31.; 51.8 5.9; 9.5.;.4 234.1; 69.2 21.4; 69.2 5.6; 21.2.;. 272.2; 21.9 45.9; 74.2 11.8; 36.6.1;.4 Whole period Mean 23.9 23.9 8.6.1 29.5 36.7 12.1. 242.6 47.7 2.9. 25th; 75th percentiles 3.; 89.6 7.7; 89.6 6.1; 9.5.;.2 232.1; 68.6 21.4; 68.6 6.3; 21.7.;. 268.6; 216.1 43.2; 68.6 14.7; 21.1.;.2 Alfentanil target-controlled infusion and obesity BJA

BJA Pérus et al. A Performance error (%) B Performance error (%) 4 35 3 25 2 15 1 5 5 1 15 4 35 3 25 2 15 1 5 5 1 15 1 5 6 12 145 155 165 175 185 195 25 235 265 325 Time (min) 1 5 6 12 19 2 21 22 23 24 251 31 37 Time (min) Fig 2 Performance error vs time for Maitre and colleagues model (A) and NONMEM model (B) for obese (solid lines) and non-obese patients (dotted lines). model. The interindividual variability (IIV) was retained for clearance (CL), volume of compartment 1 (V 1 ), clearance of compartment 2 (Q 2 ), clearance of compartment 3 (Q 3 ), and volume of compartment 3 (V 3 ) and were modelled as exponential. An exponential model for the residual variability ensured a good adequacy between the observed and predicted values. During the selection process, we found that BMI was a significant factor influencing the IIV in clearance (CL). Between the base and the final model, the OFV decreased by 15 units and the intersubject variability on clearance from 61.1% to 51.1% (1.%). None of the other covariates tested was retained. The relationship observed between BMI and CL is described in Figure 5. The final PK estimate parameters and 95% confidence intervals obtained from the bootstrap procedures are summarized in Table 3. h-shrinkage was evaluated for the parameters for which IIV was identified, since poor individual estimates of the parameter could result in distorted parameter covariate relationship: 9 h-shrinkage values were 1.6% for CL, 7.9% for Q 2, 4.8% for Q 3 and 28.7% for V 3, respectively. Figure 2B shows the performance error vs time in obese and non-obese patients for the NONMEM model. Figures 3B and 4B show the predicted vs observed plasma alfentanil concentrations and the normalized predicted distribution error vs predicted concentration for the whole period, respectively. Discussion Clinical outcomes of patients anaesthetized with TCI and its accuracy should be used to select a PK set. It was shown that a mean 2 3% variation of measured concentrations above or below targeted concentrations with a maximum of 5 6% can be considered clinically acceptable. 1 Different possible sources of variability are encountered when using a PK model for TCI, especially if patients receiving TCI do not belong to the same population as that used to develop the original PK model. In morbidly obese patients, PK changes 11 12 have been reported for highly lipophilic drugs. 556

Alfentanil target-controlled infusion and obesity BJA A Predicted concentrations (mg litre 1 ) 12 1 8 6 4 2 B Predicted concentrations (mg litre 1 ) 3 2 1 5 1 15 2 25 3 35 Observed concentrations (mg litre 1 ) 5 1 15 2 25 3 35 Observed concentrations (mg litre 1 ) Fig 3 Correlation between predicted and measured alfentanil concentrations for the whole period using Maitre and colleagues model (A) or NONMEM model (B). The solid line represents the ideal correlation as opposed to that obtained. A 3 B 2 2 1 1 npde npde 1 1 2 2 3 2 4 6 8 1 12 Predicted concentrations (mg litre 1 ) 3 5 1 15 2 25 3 35 Predicted concentrations (mg litre 1 ) Fig 4 Plot of the normalized predictive distribution errors (NPDE) vs predicted alfentanil concentration with Maitre and colleagues model (A) or NONMEM model (B). The solid line represents the ideal correlation as opposed to that obtained. Only a few studies are available concerning the administration of alfentanil in the TCI mode, 13 18 and ours is the first one evaluating an alfentanil TCI in a population including obese patients. However, the small number of our patients induced limitations for covariate analysis, 19 and probably contributed to alter the results of Maitre and colleagues 3 and Scott and colleagues 4 PK sets performance, especially in non-obese patients (Table 2). We selected Maitre and colleagues PK model because it was used in all previous studies on TCI administration of alfentanil, during the intraoperative period 13 16 18 and/or 13 14 17 for postoperative analgesia. According to Coetzee and colleagues, 1 our MDPE and MDAPE overall values of 13.3% and 23.9%, respectively, for the whole period with Maitre and colleagues PK set would be sufficient to provide acceptable accuracy for clinical use. Our alfentanil plasma concentrations provided good residual analgesia, and according to Stanski and Hug, 2 an alfentanil concentration of 1 ng ml 21 is sufficient to provide spontaneous ventilation. 557

BJA Pérus et al. Van den Nieuwenhuyzen and colleagues obtained similar results. They successfully used Maitre and colleagues PK set for postoperative analgesia in three different studies, and found excellent bias and accuracy values. 13 14 17 In their first studies, 13 14 alfentanil was administered during surgery targeting plasma concentrations (4 and 15 ng ml 21, respectively), and blood samples were collected exclusively after operation. The median values of minimal effective analgesic concentrations were in a range 43 65 ng ml 21. Even though no adverse effect was described in the other studies, the predictive performances of Maitre and colleagues PK model differed markedly. Clearance (litre min 1 ).7.6.5.4.3.2.1 2 3 4 5 Body Mass Index (kg m 2 ) Fig 5 Correlation with NONMEM model between BMI and CL. Barvais and colleagues 16 compared nine different PK set in eight volunteers during a 4 h alfentanil infusion at 5 ng ml 21. All PK sets, except Scott and colleagues model which slightly overpredicted concentrations, underestimated plasma alfentanil concentrations. Maitre and colleagues MDAPE was inaccurate, exceeding 5%. The authors concluded that PK models derived from patients receiving a large and rapid bolus injection were not accurate for alfentanil TCI administration, and recommended to use models derived from a slow injection. These findings were confirmed in the elderly in another study. 18 Others 15 found Maitre and colleagues PK set to have poor predictive performance with MDPE and MDAPE values at 53% in two groups of patients, resulting in underestimations of alfentanil concentrations. The initial plasma concentration target was 1 ng ml 21 for the first group, and the second group had sequential concentrations from 4 to 7 ng ml 21. Arterial blood samples were collected during the intraoperative period. Those authors stopped their investigation of Maitre and colleagues model after 11 patients from the second group because of the poor bias and accuracy values, and pursued the study with Scott and colleagues PK set, which yielded better results. The authors did not find a complete explanation for the differences observed in the two PK parameter sets, but found that the inaccuracy of Maitre and colleagues model was high after changing the target concentration. We observed similar results in our study, when we compared the Maitre and colleagues model bias and accuracy during the different periods of alfentanil infusion: performance was poorer after changing the target concentration, and at the onset of alfentanil infusion: the plot of predicted vs observed concentrations (Fig. 3A) showed a good correlation for concentrations below 11 ng ml 21, but a lack of fit for higher concentrations for which the model Table 3 Population PK parameters of alfentanil NONMEM model [values (SE)]. CL, clearance; V, volume of distribution; Q, clearance of the compartment Parameter Basic model Final model 1 bootstrap replicates Mean estimate %SE Mean estimate %SE Mean estimate 2.5 97.5% quantiles CL (litre min 21 ).226 15.3 CL (litre min 21 )¼n 1 BMI.245.243.178.33 n 1.647 12.4.641.469.791 V 1 (litre) 3.67 8.6 2.7 7.2 3.33 2.7 4.2 Q 2 (litre min 21 ) 1.35 13.7 1.71 21.1 1.44.94 1.95 V 2 (litre) 1.3 7.2 9.28 9.2 1.71 9.15 12.6 Q 3 (litre min 21 ).331 22.2.389 15.7.36.266.464 V 3 (litre) 36.3 15.9 38.6 13.6 44.9 27.5 65.3 v 2 CL.611 48.4.511 32.9.453.268.644 v 2 V 1.285 37.9 v 2 Q 2.573 28.2.619 36.6.525.174.757 v 2 Q 3.65 41.3.56 42.4.53.3.773 v 2 V 3.349 48.5.31 46.3.542.24.812 Residual variability, s 2.116 22.7.123 22.2.137.98.184 558

Alfentanil target-controlled infusion and obesity BJA underestimated the observed concentrations. We chose to target the alfentanil effect-site concentration, because of the difficulty of rapidly modifying the drug effect while targeting the plasma, as each increase in the effect-site target induced a consequent peak plasma alfentanil concentration. The performance of Maitre and colleagues model was particularly bad at the first sample, 1 min after the start of alfentanil infusion (the only one sample realized before effect-site and plasma target concentration equilibration), and some hypothesis may explain this situation. Mertens and colleagues 21 compared alfentanil PKs during a continuous infusion, alone, or with a continuous infusion of propofol started 1 min before the start of alfentanil infusion: propofol altered alfentanil PKs, increasing alfentanil plasma concentrations, including the initial alfentanil plasma peak. In our study, a single propofol bolus was injected during induction 9 s before the start of alfentanil infusion, and it probably contributed to increase the initial alfentanil plasma peak. On the other hand, the technique of blood sampling may play an important role on possible artifacts: some authors recommend to evaluate PK studies of drugs using arterial samples, observing that venous concentrations are lower than arterial concentrations; 1 22 conventional compartment models assume that drug added in the central compartment is instantaneously completely mixed and appears in the arterial concentration, but in practice, this is false, and this is a reason why PK models do not perform well in the first minutes. 23 For drugs undergoing tissue elimination such as remifentanil, it was observed that venous sampling reflected more exactly the effect compartment concentrations. 24 We also observed that Scott and colleagues model performed within acceptable performance in non-obese patients during alfentanil infusion. This PK set seemed to be more accurate than Maitre and colleagues one for high concentrations, but it did not perform with sufficient precision in obese patients, and during the post-infusion period in both populations. These observations could explain the poor results obtained by Raemer and colleagues, 15 because they used much higher alfentanil target concentrations than van den Nieuwenhuyzen and colleagues, 13 14 who began blood sampling at lower alfentanil concentration. The NONMEM population PK model seemed to be more accurate during the infusion and for high concentrations. But because of the small size of our population, it needs to be retested and validated in a larger population of patients, including men. Despite the relationship between BMI and CL, IIV observed with our model is relatively high. Maitre and colleagues devised their model based on data from 45 non-obese patients, 3 and validated it for 19 non-obese patients: 25 their PK pharmacodynamic analysis identified three variability factors: age, sex, and weight. Entering a modified weight on the TCI device system could improve the reliability of the PK set: Egan and colleagues 11 showed that administering a lean-person dose of remifentanil to an obese patient induced an excessively high plasma concentration, and recommended to adjust remifentanil dose to LBM. But the problem is that existing LBM formulas or nomograms are not applicable to obese patients, because of the lack of data. 26 Anaesthesiologists must find a compromise between IBW, and TBW for drug administration to obese patients. La Colla and colleagues 27 described in a case report a super-obese patient in whom remifentanil was successfully administered in the TCI mode using the IBW. Servin and colleagues 28 suggested an adjusted IBW formula for propofol administration, but finally, it seemed that the performance of the TCI device system was better with propofol administered using TBW. 27 For sufentanil, we previously demonstrated that TCI administration using TBW was accurate for obese patients with the PK set described previously. 29 Because of its pharmacological properties, alfentanil has lower clearance and far less redistribution than sufentanil; it could be administered as a function of LBM. But repeating our analyses with LBM showed no improvement for Maitre and colleagues model. Our results suggest that the concentration predicted by Maitre and colleagues model was underestimated, which may lead to overdosing alfentanil for obese patients. In conclusion, the accuracy of the PK model described by Maitre and colleagues 3 for alfentanil TCI is acceptable for clinical application to morbidly obese patients, but it should be applied with caution to that population of patients, especially when high-concentration targets are applied. Declaration of interest None declared. Funding This study was fully supported by the University Hospital of Nice, as part of an incentive contract for Clinical research (CIR 23). References 1 Casati A, Putzu M. Anesthesia in the obese patient: pharmacokinetic considerations. J Clin Anesth 25; 17: 134 45 2 Ogunnaike BO, Jones SB, Jones DB, Provost D, Whitten CW. Anesthetic considerations for bariatric surgery. Anesth Analg 22; 95: 1793 85 3 Maitre PO, Vozeh S, Heykants J, Thomson DA, Stanski DR. Population pharmacokinetics of alfentanil: the average dose plasma concentration relationship and interindividual variability in patients. Anesthesiology 1987; 66: 3 12 4 Scott JC, Ponganis KV, Stanski DR. EEG quantitation of narcotic effect: the comparative pharmacodynamics of fentanyl and alfentanil. Anesthesiology 1985; 62: 234 41 5 Varvel JR, Donoho DL, Shafer SL. Measuring the predictive performance of computer-controlled infusion pumps. J Pharmacokinet Biopharm 1992; 2: 63 94 6 Glen JB, Servin F. Evaluation of the predictive performance of four pharmacokinetic models for propofol. Br J Anaesth 29; 12: 626 32 559

BJA Pérus et al. 7 Beal SL, Sheiner LB, Boeckmann AJ, eds. NONMEM User s Guide. Ellicott City, MD: Icon Development Solutions, 1989 26 8 Janmahasatian S, Duffull SB, Ash S, Ward LC, Byrne NM, Green B. Quantification of lean bodyweight. Clin Pharmacokinet 25; 44: 151 65 9 Karlsson MO, Savic RM. Diagnosing model diagnostics. Clin Pharmacol Ther 27; 82: 17 2 1 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 11 Egan TD, Huizinga B, Gupta SK, et al. Remifentanil pharmacokinetics in obese versus lean patients. Anesthesiology 1998; 89: 562 73 12 Schwartz AE, Matteo RS, Ornstein E, Young WL, Myers KJ. Pharmacokinetics of sufentanil in obese patients. Anesth Analg 1991; 73: 79 3 13 van den Nieuwenhuyzen MC, Engbers FH, Burm AG, et al. Computer-controlled infusion of alfentanil for postoperative analgesia. A pharmacokinetic and pharmacodynamic evaluation. Anesthesiology 1993; 79: 481 92; discussion 27A 14 van den Nieuwenhuyzen MC, Engbers FH, Burm AG, Vletter AA, van Kleef JW, Bovill JG. Computer-controlled infusion of alfentanil versus patient-controlled administration of morphine for postoperative analgesia: a double-blind randomized trial. Anesth Analg 1995; 81: 671 9 15 Raemer DB, Buschman A, Varvel JR, et al. The prospective use of population pharmacokinetics in a computer-driven infusion system for alfentanil. Anesthesiology 199; 73: 66 72 16 Barvais L, Cantraine F, D Hollander A, Coussaert E. Predictive accuracy of continuous alfentanil infusion in volunteers: variability of different pharmacokinetic sets. Anesth Analg 1993; 77: 81 1 17 Van den Nieuwenhuyzen MC, Engbers FH, Burm AG, Vletter AA, Van Kleef JW, Bovill JG. Target-controlled infusion of alfentanil for postoperative analgesia: a feasibility study and pharmacodynamic evaluation in the early postoperative period. Br J Anaesth 1997; 78: 17 23 18 Barvais L, D Hollander A, Schmartz D, Hendrice C, Cantraine F, Coussaert E. Predictive accuracy of alfentanil infusion in coronary artery surgery: a prebypass study in middle-aged and elderly patients. J Cardiothorac Vasc Anesth 1994; 8: 278 83 19 Ribbing J, Jonsson EN. Power, selection bias and predictive performance of the population pharmacokinetic covariate model. J Pharmacokinet Pharmacodyn 24; 31: 19 34 2 Stanski DR, Hug CC Jr. Alfentanil a kinetically predictable narcotic analgesic. Anesthesiology 1982; 57: 435 8 21 Mertens MJ, Vuyk J, Olofsen E, Bovill JG, Burm AG. Propofol alters the pharmacokinetics of alfentanil in healthy male volunteers. Anesthesiology 21; 94: 949 57 22 Viby-Mogensen J, Ostergaard D, Donati F, et al. Pharmacokinetic studies of neuromuscular blocking agents: good clinical research practice (GCRP). Acta Anaesthesiol Scand 2; 44: 1169 9 23 Struys MM, Coppens MJ, De Neve N, et al. Influence of administration rate on propofol plasma-effect site equilibration. Anesthesiology 27; 17: 386 96 24 Kabbaj M, Vachon P, Varin F. Impact of peripheral elimination on the concentration effect relationship of remifentanil in anaesthetized dogs. Br J Anaesth 25; 94: 357 65 25 Maitre PO, Ausems ME, Vozeh S, Stanski DR. Evaluating the accuracy of using population pharmacokinetic data to predict plasma concentrations of alfentanil. Anesthesiology 1988; 68: 59 67 26 Coetzee JF. Total intravenous anaesthesia to obese patients: largely guesswork? Eur J Anaesthesiol 29; 26: 359 61 27 La Colla L, La Colla G, Albertin A, Poli D, Baruffaldi Preis FW, Mangano A. The use of propofol and remifentanil for the anaesthetic management of a super-obese patient. Anaesthesia 27; 62: 842 5 28 Servin F, Farinotti R, Haberer JP, Desmonts JM. Propofol infusion for maintenance of anesthesia in morbidly obese patients receiving nitrous oxide. A clinical and pharmacokinetic study. Anesthesiology 1993; 78: 657 65 29 Slepchenko G, Simon N, Goubaux B, Levron JC, Le Moing JP, Raucoules-Aime M. Performance of target-controlled sufentanil infusion in obese patients. Anesthesiology 23; 98: 65 73 56