Pharmacokinetics and Pharmacodynamics of Midazolam and Metabolites in Nonventilated Infants after Craniofacial Surgery

The study was performed in the Pediatric Surgery Intensive Care Unit (PSICU) of the Erasmus Medical Center–Sophia Children’s Hospital, Rotterdam, The Netherlands. The study protocol was approved by the ethics committee of the Erasmus Medical Center–Sophia Children’s Hospital. Written informed consent was obtained from the parents. The studied patients and the design of the randomized study are presented in detail in the article of Prins et al. ,18in which the safety assessments in the patients receiving midazolam and propofol are described and shortly repeated in the article of Peeters et al. ,19in which the population pharmacokinetic and pharmacodynamic model for the children allocated to receive propofol is described. For brevity, parts of the methods are mentioned in this article when relevant.

Data of children who required sedative medication (agitated group) according to the COMFORT-B score (score ≥ 17) and who were randomly allocated to receive midazolam were included in the analysis. Criteria for eligibility in the study included age between 1 month and 2 yr; admission to the PSICU after major craniofacial surgery; and no respiratory infections, epilepsy, hypertriglyceridemia or family histories of hypercholesterolemia, or allergic history to midazolam, propofol, eggs, or soybean oil. Characteristics of patients in the midazolam group are shown in table 1. The median age was 11.1 (3.2–24.7) months, and the median weight was 9.4 (5.1–12) kg. For the description of the postoperative sleep pattern of the agitated group, we also included in the analysis data from 20 infants (14 male, 6 female; age, 9.4 [3.8–17.3] months; weight, 8.8 [4.8–12.5] kg) who were randomly assigned to receive propofol for whom more than two COMFORT-B observations before propofol administration were available. In 9 infants (5 male, 4 female; age, 8.8 [4.0–12.4] months; weight, 8.3 [5.5–9.6] kg), no sedation was necessary (nonagitated group). One infant received a bolus of midazolam at the end of the operation, before entering the PSICU. During the stay in the PSICU, no sedation was needed. In contrast to Prins et al. ,18this particular infant was included in the midazolam group instead of the group in which no sedation was needed. All were full-term babies without overt growth retardation. All patients had normal hepatic and renal functions. Genotype analysis identified 22 carriers of the CYP3A5 allele among the 24 infants who received midazolam (2 heterozygous CYP3A5*1/*3, classified as extensive metabolizers, and 20 homozygous CYP3A5*3/*3, classified as poor metabolizers), and in 2 infants, no result was obtained. Three carriers of the CYP3A7*1C were identified (3 heterozygous). The allele CYP3A4*1B was not detected.

From arrival at the PSICU, depth of sedation was evaluated using the COMFORT-B score, which rates six behavioral items.14,20Alertness, calmness, muscle tone, body movement, facial tension, crying (nonventilated children), or respiratory response (ventilated children) are scored on a five-point scale, resulting in a total score varying from 6 (no distress) to 30 (severe distress). The interobserver reliability proved to be good for all nurses and the principal investigator (κ > 0.65). In addition, the BIS was recorded continuously and noted at 15-min intervals (Bispectral® A 2000 version 3.12; Aspect Medical Systems, Natick MA, with pediatric BIS® sensors). The BIS ranges from 100 (awake) to 0 (isoelectric electroencephalogram). Midazolam was initially given as 0.1-mg/kg bolus followed by a continuous infusion of 0.05 mg · kg⁻¹· h⁻¹, titrated up after an additional bolus or down by 0.025 mg · kg⁻¹· h⁻¹. In 21% of the infants, the starting dose was insufficient. To determine whether restlessness was induced by pain, the trained nurses also obtained the visual analog scale. Patients received standard four daily doses of 120–240 mg acetaminophen rectally in the PSICU after a loading dose of 40 mg/kg rectally 2 h before extubation during the operation.21 

Arterial blood samples (500–1,000 μl) were collected in each infant at the following times: at baseline before the start of the midazolam bolus, approximately 45 or 30 min, 90 or 60 min, 120 min, 4 h, 6 or 8 h, and 10 h after the start of the midazolam infusion, just before and 1 h after dose adjustment, just before discontinuation of the midazolam infusion, and 30 or 45, 60 or 90, 120, and 180–240 min after the end of the infusion (median of 11 samples per child). If the arterial line was no longer available (dislocation, obstruction), venous samples were collected from a central line, routinely present in the superior caval vein. In five infants, venous blood samples were obtained with three, three, five, eight, and nine venous samples taken per infant, respectively. After collection, the samples were centrifuged and stored at −80°C until analysis.

Midazolam, 1-OH-midazolam, and 1-OH-midazolamglucuronide concentrations were measured in serum using high-performance liquid chromatography with ultraviolet detection at 230 nm. The mobile phase was prepared as follows: 400 μl phosphoric acid, 85%, and 146 μl triethylamine were added to 530 ml water. The pH was adjusted to 3.2 with 10% potassium hydroxide, and 470 ml acetonitrile was added. Temazepam was used as an internal standard. Borate buffer, 500 μl, 0.05 m (pH 9.2), was added to 200 μl serum. After liquid–liquid extraction with 6 ml dichloromethane, the organic layer was evaporated to dryness at 37°C. The residue was reconstituted in 200 μl of mobile phase, and 75 μl was injected onto the analytical column (Lichrosphere 100RP-18 encapped 5 μm; Merck, Darmstadt, Germany). Total (conjugated and unconjugated) drug concentrations of 1-OH-midazolam were measured after enzymatic hydrolysis of 200 μl serum with 100 UI β-glucuronidase (Roche Diagnostics, Almere, The Netherlands) for 24 h at 37°C. The differences between total and unconjugated 1-OH-midazolam concentration was taken as the 1-OH-midazolamglucuronide concentration. The limits of quantification were 11 μg/l for midazolam and 6 μg/l for 1-OH-midazolam using 200 μl serum. Interassay and intraassay coefficients of variation were less than 8% and 13%, respectively. Total recovery was larger than 90% for both compounds.

Providing data for a large genomic study, DNA was isolated from EDTA blood (MasterAmp; Epicenter Technologies, Madison, WI). CYP3A4*1B, CYP3A5*3, and CYP3A7*1C analyses were performed, using polymerase chain reaction restriction fragment length polymorphism assays, as described previously.22–24 

The analysis was performed in NONMEM (Non-Linear Mixed Effect Modeling; version V, release 1.1; GloboMax LLC, Hanover, MD)25by use of the first-order conditional estimation (method 1) with η–ϵ interaction. S-plus (version 6.2; Insightful software, Seattle, WA) was used to visualize the data. Population pharmacokinetic and pharmacodynamic data were sequentially analyzed. Discrimination between different models was made by comparison of the objective function. A value of P < 0.005, representing a decrease of 7.8 in the objective function, was considered statistically significant. In addition, goodness-of-fit plots (observed vs.  individually predicted, observed vs.  population predicted, time vs.  weighted residuals, and population predictions vs.  weighted residuals) were used for diagnostic purposes. Furthermore, the confidence interval of the parameter estimates, the correlation matrix, and visual improvement of the individual plots were used to evaluate the model.

Covariates were plotted independently against the individual post hoc  parameter estimates and the weighted residuals to visualize potential relations. The following covariates were tested: body weight, age, body surface area, body mass index, sex, and sampling (venous or arterial). The pharmacokinetic parameters were also tested for correlation with heart frequency, blood pressure, and the genotypes (CYP3A4*1B, 3A5*3, 3A7*1C). Potential covariates were separately entered into the model and statistically tested by use of the objective function. A significant covariate that most reduces the objective function was left in the model. Additional covariates had to reduce this objective function further to be retained in the model. The choice of the model was further evaluated as discussed previously.

The internal validity of the population pharmacokinetic and pharmacodynamic models was assessed by the bootstrap resampling method (repeated random sampling to produce another data set of the same size but with a different combination of individuals). Parameters obtained with the bootstrap replicates were compared with the estimates obtained from the original data set.

Midazolam and metabolite data were fitted simultaneously, and concentrations were expressed as μm. The molecular weights of midazolam, 1-OH-midazolam, and 1-OH-midazolamglucuronide are 325.77, 341.77, and 517.9, respectively. The pharmacokinetic model used is schematically depicted in figure 1. The midazolam data were described with a two-compartment model, parameterized in terms of volume of the central compartment (V¹), volume of the peripheral volume (V²), intercompartmental clearance (Q), and clearances to 1-OH-midazolam (Cl¹) and other metabolites (Cl⁰). In the absence of information on the ratio of metabolite formation in children, Cl¹was assumed to be 60% of the elimination clearance of midazolam Cl^e(the sum of Cl⁰and Cl¹) as reported in the literature for adults.1,10,26The formation of 1-OH-midazolam and 1-OH-midazolamglucuronide was best described with a one-compartment model. Cl³is the clearance of 1-OH-midazolam, and Cl⁴is the clearance of 1-OH-midazolamglucuronide. The volume of distribution of 1-OH-midazolam (V³) was modeled as a fraction of the sum of V¹and V²of midazolam, because estimation of this parameter was found to be unstable by the bootstrap resampling. The individual value (post hoc  value) of the parameters of the i  th subject was modeled by

where θ^meanis the population mean and ηⁱis assumed to be a gaussian random variable with zero mean and variance ω2. The intraindividual variability was best described with a combined additive and proportional error model for midazolam assuming a constant coefficient of variation over the complete concentration range superimposed on a constant absolute error (equation 2) and a proportional error model for the metabolites (equation 3), respectively. This means for the j  th observed concentration of the i  th individual, the relation (Y^ij):

where c^predis predicted midazolam or metabolite concentration and ϵ^1,2,3,ijare random variables with mean zero and variance σ².

Depth of sedation (S) was characterized as a function of postoperative natural sleep pattern (PNSP) and midazolam effect (MEF):

The PNSP was described as a function of three equations:

in which BSL is the level of sedation at arrival at the PSICU, PAEFF is the postanesthesia effect, and CNR is the circadian night rhythm.

The postanesthesia effect was assumed to wash out in time postoperatively by an E^maxmodel, resulting in a more awake sedation level to a maximum estimated score (S^max) for the COMFORT-B and 100 (fully awake) for the BIS.

where PAE^maxis the maximal effect from BSL to the maximal score S^max. T^PSis the time (minutes) postsurgery, and T^50,PSis the time (minutes) postsurgery at half maximum postanesthesia effect. Incorporation of the postanesthesia effect of the COMFORT-B score resulted in a reduction of the objective function by 80.1 points, which was highly significant.

Circadian night rhythm was modeled by

in which O denotes the onset of the natural night dip in minutes from 12:00 h. The end of the circadian night dip (wake-up time) was assumed at 7:00 h, because at this time point, the light is turned on, nursing care is optimized, and the parents arrive at the PSICU. A is amplitude of the night dip (units COMFORT-B or BIS), and 2 π/Fr is frequency of the oscillations (minutes). Introduction of the CNR improved the goodness of fit as reflected in a decrease in objective function of 18.9 points for the COMFORT-B and 119.3 points for the BIS.

Midazolam effect (MEF) was related to the pharmacokinetic model–predicted individual midazolam concentration (C^1,ij) by a simple E^maxmodel:

where E^max,iis the maximum possible midazolam effect (equal to S^max−6 on the COMFORT-B scale and 100 on the BIS scale) in the i  th subject. EC⁵⁰is the concentration (μm) at half maximum effect, in which the interindividual variability was assumed to be log-normally distributed. The significant increase in objective function when the midazolam effect is eliminated from the COMFORT-B and BIS model (50.3 and 119 points, respectively) demonstrated the effect of midazolam.

Using the BIS, EC⁵⁰was modeled with the MIXTURE subroutine in NONMEM (P < 0.005). A mixture model assumes that the population consists of two or more subpopulations, each approximating a normal distribution, where each subpopulation may have its own model. The ratio of the fraction and the corresponding typical EC⁵⁰are estimated, and NONMEM assigned patients to one of the subpopulations.

For the influence of the active metabolite 1-OH-midazolam (C^2,ij) in the presence of the midazolam concentrations (C^1,ij), an additive interaction model was tested, in which the maximal effect (E^max) of midazolam and 1-OH-midazolam was assumed to be equal and the Hill factor was 1 for the two compounds:

Because all infants had a normal renal function, the metabolite 1-OH-midazolamglucuronide, which is only of clinical relevance in renal failure when accumulation occurs,27was assumed to be without effect. The interindividual variabilities (ηⁱs) were symmetrically distributed zero-mean random variables with a variance ω2. The intraindividual variabilities in the COMFORT-B (equation 10) and BIS (equation 11) were best characterized by a proportional and an additive error model, respectively.

where Y^ijrepresents the observed effect in the i  th subject at the j  th time point.

The pharmacokinetic model was derived from a median of 9 midazolam, 8 1-OH-midazolam, and 8 1-OH-midazolamglucuronide observations obtained per infant. Median 1-OH-midazolam/midazolam and (1-OH-midazolam + 1-OH-midazolamglucuronide)/midazolam ratios were 0.37 in 158 samples and 2.3 in 144 samples, respectively. The pharmacokinetic parameter values and their confidence interval and the values obtained from the bootstrapping are shown in table 2. The fits of 250 bootstrap replicates of the data set demonstrated the stability of the model. These mean parameter estimates were within 17% of those obtained with the original data set. However, it should be noted that the estimated volumes of distribution of the metabolites must be taken with caution, because accurate estimates can only be obtained by separate administration and are affected by the assumed fraction of midazolam metabolized to 1-OH-midazolam. One individual who needed up to 0.2 mg · kg⁻¹· h⁻¹midazolam showed very low midazolam and 1-OH-midazolam concentrations (two and five times lower, respectively, as compared with the population mean), indicated by a high individual Cl¹(0.18 l/min) and Cl³(0.59 l/min) and a low 1-OH-midazolam/midazolam ratio of 0.18. Considering the large effect of this individual on the variability, an extra factor (fa) was estimated for this infant, which resulted in a significant decrease in objective function (P < 0.001). This infant was heterozygous for the allele CYP3A7*1C. Figure 2shows the diagnostic plots for parent midazolam pharmacokinetic data. A representative example of measured and predicted serum concentrations of midazolam and its two metabolites for a median fit are shown in figure 3. None of the explored covariates (body weight, age, body surface area, body mass index, sex, heart frequency, blood pressure, sampling [venous or arterial], and the genotypes [CYP3A4*1B, 3A5*3, 3A7*1C]) were identified as significant, although there was a trend toward a positive linear correlation between age and elimination clearance (fig. 4). In this figure, the appearance of the allele expression is also given.

The data set included 632 COMFORT-B observations from 53 infants, yielding a median of 13 (3–25) observations per infant and a total of 3,570 BIS observations, 75 (4–496) per infant. The population parameters of the pharmacodynamic model are reported in table 3. The bootstrap validation (100 times) confirmed the precision of the parameters. Age was found to be a significant covariate for the baseline BSL (state of comfort at arrival) in the PSICU, according to a slope-intercept model centered to the median value. Nonagitated infants, in whom sedative administration was not necessary (COMFORT-B < 17), displayed a delayed postanesthesia washout (T^50,PS1,794 vs.  537 min). In addition, they showed a night dip (CNR), which was implemented in the model using the dip of a circadian rhythm. The nighttime observations decreased a maximum of 3.5 units on the COMFORT-B (amplitude) from 20:00 h onward (equal to 478 min from 12:00 h) and 14.7 values on the BIS from 17:30 h onward. In the agitated infants, no night dip (CNR) could be identified. Using the BIS as a pharmacodynamic endpoint, the postanesthesia effect could not be described because of the large observed interindividual and intraindividual variability in response (table 3). The effect of midazolam on the COMFORT-B was highly variable, with an interindividual coefficient of variation in EC⁵⁰of 89%. Using the BIS, an estimated 57% of the infants did not display a significant response on midazolam (“nonresponders”). The EC⁵⁰for the subpopulation “responders” was 0.63 μm, with an interindividual variability of 66%. No covariates, age included, could be detected. Splitting the patients in two age groups, ≥ 1 yr and < 1 yr, according to the age for which the BIS was validated, the EC⁵⁰was 0.34 μm for two responders in the age group ≥ 1 yr. The other eight patients did not display a response on the BIS. For the age group < 1 yr, 61% displayed a response on the BIS. The EC⁵⁰was 0.69 μm, with an interindividual variability of 70%.

For the influence of 1-OH-midazolam on the pharmacodynamics, an additive interaction model was tested according to the Materials and Methods section. However, this model was unable to estimate the values of the EC⁵⁰of midazolam and 1-OH-midazolam separately. Further simplification of this model, assuming equal values for EC⁵⁰for both components, did not result in a significant decrease in objective function or interindividual variability but only a shift of EC⁵⁰from 0.58 to 0.81 μm.

The observed and predicted depth of sedation characterized by COMFORT-B and BIS for a responder (A) and a nonresponder (B) and their corresponding midazolam concentrations are shown in figure 5. In figures 6A and B, the simulated relation between time, two different dose regimens of midazolam, midazolam concentration, and predicted population response is demonstrated in terms of depth of sedation using COMFORT-B, based on the derived pharmacodynamic model. The influence of the covariate age on the baseline using the COMFORT-B is shown in figure 6B. Figure 6Cshows the postoperative natural sleep pattern of the nonagitated infants who did not need sedative medication.

A population pharmacokinetic and pharmacodynamic model of midazolam and its metabolites 1-OH-midazolam and 1-OH-midazolamglucuronide based on the validated COMFORT-B scale is described to refine postoperative sedative treatment in nonventilated infants aged 3 months to 2 yr after surgery in the PICU.

In defining the optimal dose for children, population pharmacokinetic and pharmacodynamic modeling is useful. Key factors in this respect are that the pharmacokinetic–pharmacodynamic correlation can be established in the clinical situation at the basis of sparse sampling. Furthermore, application of the population approach enables the characterization of interindividual variability as well as the source of this variability on the basis of covariate analysis.

The pharmacokinetic model derived in this study estimated a total clearance of midazolam of 157 ml/min (16.7 ml · kg⁻¹· min⁻¹) in nonventilated children, which is two to five times higher than clearance described in ventilated critically ill children or in ventilated children after cardiac surgery. Hughes et al.  28estimated a median clearance of 3.1 ml · kg⁻¹· min⁻¹from steady state concentrations in critically ill infants aged 1 month to 1 yr. De Wildt3found a mean clearance of 5.0 ml · kg⁻¹· min⁻¹in intensive care patients aged 2 days to 17 yr. In addition, they found a 2.5 times lower ratio for 1-OH-midazolam/midazolam concentrations. Mathews et al.  29reported a clearance of 9.6 ml · kg⁻¹· min⁻¹in children aged 2–8 yr as continuous infusion after cardiac surgery. Compared with nonventilated children aged 6 months to 2 yr after a single dose before minor in-hospital or day-stay procedures, the clearance found in our population was slightly higher (16.7 vs.  11.3 ml · kg⁻¹· min⁻¹)5but was comparable to clearance described in nonventilated healthy adults after a bolus injection (16.1 ml · kg⁻¹· min⁻¹).30Our pharmacokinetic analysis demonstrates that nonventilated infants after major surgery may require relatively high doses of midazolam. Because midazolam is an intermediate extraction ratio drug, this may be attributable to the relatively healthy state of the children and to the absence of mechanical ventilation, which affects the hepatic blood flow.

A high degree of interindividual variability in clearance (coefficient of variation, 54%) was seen, which could not be attributed to body weight, venous or arterial sampling, frequencies of CYP3A4, 5 or 7 variant alleles, or hemodynamic parameters, probably because of the narrow variability in patient characteristics and condition. The clearance tended to be related to age, but the relatively small number of infants older than 15 months may be the cause that the trend did not reach statistical significance. As a consequence, dose recommendations may be less appropriate for infants aged 15–24 months. Regarding arterial and venous sampling, there is evidence that differences are only relevant during the rapid distribution phase,31and this does not seems to be of clinical relevance when midazolam is used as continuous infusion in sedated children. In one heterozygous CYP3A7*1C infant who needed high doses of midazolam, the incorporation in the model of a multiplication factor of Cl¹and Cl³resulted in a significantly better fit, which means that the oxidation and glucuronidation was between two and three times faster than in the other infants. The allele CYP3A7*1C is associated with continued high hepatic and intestinal CYP3A7 expression.32The clinical relevance of the finding is yet unclear because it is limited by the frequency of the alleles as relates to the number of patients because it has been analyzed as part of a large DNA data study. Large studies may answer the question of whether this investigated allele plays a significant role.

Depth of sedation may be difficult to assess in children. The COMFORT scale is validated in the PICU and measures six behavioral items as well as two physiologic items (mean arterial pressure and heart rate).33Because the physiologic items are controlled in the intensive care unit, the COMFORT-B score was developed in Canada by Carnevale and Razack15and is routinely used in most PICUs in The Netherlands.14The BIS is a processed electroencephalographic parameter developed using adult data and is objective and easy to use, but is not yet validated for children below the age of 1 yr. The impact of age on the BIS is still debated, with divergent findings.17,34 

Using the COMFORT-B as pharmacodynamic endpoint, depth of sedation was described as a function of a baseline, a postanesthesia effect, a CNR, and the midazolam effect. Age was found to be a significant covariate for the baseline (the state of comfort at arrival) in the PSICU. This indicates that young children may be more sensitive to the environment and emotional distress than older infants. Nonagitated children displayed on the COMFORT-B a night dip starting at 20:00 h and a slower washout period of the postanesthesia effect (1,794 vs.  537 min, respectively) compared with agitated infants. In agitated children, no night dip was observed. In these infants, the midazolam effect was implemented using an E^maxmodel. Using the BIS, a large residual error and a large interindividual variability were found, resulting in the inability to detect the postanesthesia effect. This confirms the clinical observation that the BIS highly fluctuates in particular in lightly sedated children. This may be explained by the fact that light sedation may be influenced more by the environment.35Fifty-six percent of the infants did not show a response of midazolam on the BIS, whereas midazolam influenced the COMFORT-B in all infants, although the interindividual variability in EC⁵⁰was large (89%) and no patient characteristics (covariates) could increase the predictability. Taking into account the age for which the BIS is validated, 8 of the 10 patients aged ≥ 1 yr did not show a response of midazolam on the BIS. Therefore, the BIS seems less sensitive and less specific for the effect of midazolam than the COMFORT-B score.

In this study, no separate EC⁵⁰could be identified for midazolam parent and metabolite, because the concentration profiles ran parallel in time while the concentration of the metabolite 1-OH-midazolam was low throughout the entire treatment period (ratio 1-OH-midazolam/midazolam is 0.37). Therefore, the observed effect was only ascribed to midazolam, using a simple E^maxmodel. Sampling immediately after the bolus may have provided a different ratio of metabolite and parent drug, which would enable identification of contribution of 1-OH-midazolam to the effect. It has been shown before in a study after oral or separate intravenous administration that 1-OH-midazolam has pharmacologic activity.4,12However, after intravenous administration, the concentration of the metabolite is relatively low compared with oral use. Also in adults after coronary artery bypass grafting, no effect of 1-OH-midazolam could be detected,36whereas 1-OH-midazolam levels were above 10 μg/l in only 11% of the patients and the ratio was at most 0.20.37 

Currently, no population pharmacodynamic studies in adults are available for comparison of the sensitivity of infants to adults using these sedation scales. In adults, the Ramsay score is often used to assess the level of sedation. Using the Ramsay scale, the midazolam concentrations in adults associated with 50% probability of a level of sedation 2 (cooperative), 3 (drowsy or asleep, easily responded to commands), and 4 (asleep, brisk response to a glabellar tap) were 0.017, 0.22, and 0.52 μm, respectively.36In the current study, after a bolus of 1 mg and a continuous infusion of 0.5 mg/h, the predicted concentration in the infants is 0.16 μm, corresponding to values between 12 and 14 on the COMFORT-B (lightly sedated). Although comparison is difficult, it seems that the midazolam concentration to achieve light sedation in infants is comparable to that in adults.

In a previous article, we described a pharmacokinetic and pharmacodynamic model for propofol in this population group19and discussed the safety of propofol compared with midazolam.18As found for midazolam, clearance of propofol was also higher than the values reported in the literature. The results of the current analysis demonstrate that midazolam shows a less predictable effect than propofol, because the interindividual variability in EC⁵⁰(89% vs.  47%) on the COMFORT-B is higher, whereas the residual (intraindividual) variability and elimination half-life are comparable (30% vs.  32% and 16.8 vs.  18.6 min, respectively). The results indicate that propofol may be preferred over midazolam as a sedative in intensive care, which should be further studied taking into account the safety recommendations.

The pharmacokinetic and pharmacodynamic population model shows that a loading dose of 1 mg followed by a continuous infusion of 0.5 mg/h midazolam is the optimal initial dose for a desired COMFORT-B score of 12–14 during the first night after major surgery in nonventilated infants aged 1 yr. Because of large interindividual variability, further individual titration is important for midazolam. Although no significant effect of age on the clearance could be detected, the initial dose recommendation may be less suitable for application in infants older than 15 months.

The authors wish to thank Ilse P. van der Heiden and Marloes van der Werf (Research Analysts, Department of Clinical Chemistry, Erasmus Medical Center, Rotterdam, The Netherlands) for genotyping, and the medical and nursing staff of the Pediatric Surgical Intensive Care Unit (Erasmus Medical Center–Sophia Children’s Hospital, Rotterdam, The Netherlands) for their help and cooperation.