Population Pharmacokinetics of Intranasal Dexmedetomidine in Infants and Young Children

This prospective pharmacokinetic study was approved by the Guangzhou Women and Children’s Medical Center (Guangzhou, China) Review Board (Institutional Review Board 201507) and registered before the first patient enrollment at the Chinese Clinical Trial Registry (ChiCTR-OPC-16008589, Principal investigator: B. L. Li, Date of registration: June 3, 2016).

Written informed consent was obtained from all the guardians of the subjects recruited in this study before surgery. This study adhered to the revised Declaration of Helsinki of the World Medical Association (Ferney-Voltaire, France) and International Conference on Harmonization and Good Clinical Practice Guidelines.

Subjects were enrolled from June 2016 to November 2017 at the Guangzhou Women and Children’s Medical Center. We enrolled children between 3 and 36 months of age, American Society of Anesthesiologists (Schaumburg, Illinois) Physical Status I and II, with simple vascular malformations as classified by the International Society for the Study of Vascular Anomalies (Milwaukee, Wisconsin) Criteria, and requiring intervention radiological procedures. The exclusion criteria included a history of allergy or hypersensitivity to dexmedetomidine; severe hepatic impairment; hematological, cardiovascular, endocrine, metabolic, and gastrointestinal diseases; exposure to dexmedetomidine or any other sedative within a week; and the presence of active respiratory symptoms, rhinorrhea, and vascular malformations in or near the nasal cavity that might influence nasal drug absorption.

All subjects received 2 mg · kg⁻¹ propofol, 0.3 µg · kg⁻¹ sufentanil, and 0.2 mg · kg⁻¹ cisatracurium besylate at anesthesia induction and had laryngeal mask airway placement or tracheal intubation. After intravenous induction of anesthesia, intranasal dexmedetomidine at 2 µg · kg⁻¹ was administered. Undiluted preservative-free dexmedetomidine (100 µg · ml⁻¹; Ai Bei Ning, JiangSu Singchn Pharmaceutical Co. Ltd., China) was used. The solution was drawn into a 1-ml tuberculin syringe and attached to a mucosal atomization device (MAD Nasal, Teleflex Incorporated, USA). The dead space of the atomization device was approximately 0.15 ml, and it was primed with dexmedetomidine so that exactly 2 µg · kg⁻¹ dexmedetomidine was drawn to the tuberculin. An equal volume of the drug was administered to each nostril of the participants. A single pediatric anesthesiologist (B.L. Li) with extensive experience using the atomizer device performed all nasal dexmedetomidine administrations. General anesthesia was maintained with sevoflurane. Intranasal administration at 2 to 3 µg · kg⁻¹ dexmedetomidine is commonly used for procedural sedation.^6,10  To prevent interventional puncture site rebleeding caused by postoperative emergence agitation,¹¹  we used intranasal dexmedetomidine at 2 µg · kg⁻¹ as an adjuvant for anesthesia. Vital signs, including oxygen saturation measured by pulse oximetry (Spo₂), pulse rate, noninvasive systolic blood pressure (SBP), and sedation score (University of Michigan Sedation Score), were measured at baseline and every 5 min until the discharge criteria were reached. Pulse rate and noninvasive SBP were recorded when their values were lower or higher than 20% of the age-defined normal range limits. Hypoxia was defined as Spo₂ equal to or less than 93%.

The following clinical data were collected and evaluated as covariates due to their potential influence on dexmedetomidine pharmacokinetics: postnatal age,¹²  postmenstrual age (defined as the sum of gestational and postnatal age), weight, sex, albumin, bilirubin, hemoglobin, glucose, liver function (alanine aminotransferase and aspartate aminotransferase), creatinine, creatinine clearance,¹³  and coadministered drugs.

The dexmedetomidine sampling strategy adopted herein was designed according to the D-optimal criterion using PopED (Harmonic Software Inc.‚ USA) in the R language. Pharmacokinetic parameters of intranasal dexmedetomidine from healthy adult subjects were extrapolated to children based on allometric scaling.⁴  Sampling windows were estimated around each sampling time to ensure that the design was clinically feasible. The D-optimality product criterion was evaluated using the following parameters: Fisher information matrix, normalized efficiency, and coefficients of variation. The final blood sampling schedule by age was as follows: 3- to 12-month-old subjects (Group 1), 13- to 23-month-old subjects (Group 2), and 24- to 36-month-old subjects (Group 3) at 6, 18, 120, 180, and 360 min; 6, 18, 60, 240, and 360 min; and 6, 18, 120, 240, and 360 min, respectively, with a minimal subject number of 50 in total. Subsequently, a 1-ml blood sample was collected from an indwelling intravenous cannula into heparin sodium tubes. After collection, the samples were centrifuged at 4°C for 10 min at 3,000 rpm · min⁻¹ and then stored at −80°C until analysis.

Plasma dexmedetomidine concentrations were quantified by validated ultrahigh-performance liquid chromatography–tandem mass spectrometry using a stable isotope-labeled internal standard.¹⁴ 

The dexmedetomidine concentration in plasma was analyzed using a ultrahigh-performance liquid chromatography system (Thermo Fisher Scientific Inc., USA) consisting of an Ultimate 3000 RSLC system with binary pumps and an S surveyor autosampler (Thermo Fisher Scientific Inc.) coupled with a TSQ Ultra triple-quadrupole mass spectrometer (Thermo Fisher Scientific Inc.). Samples were separated on an Acquity BEH C₁₈ column (2.1 mm × 50 mm, 1.7 µm particle size; Waters, USA) set at 40°C. The mobile phase consisted of acetonitrile (A) and 1‰ formic acid water solution (B) at a flow rate of 0.3 ml · min⁻¹. The total run time of each sample was 3.1 min. The conditions of the gradient elution were set as follows: 0 to 0.5 min, 28% A; 0.5 to 1.5 min, 28 to 90% A; 1.5 to 2.0 min, 90% A; 2.0 to 2.1 min, 90 to 28% A; and 2.1 to 3.1 min, 28% A.

Mass spectrometric detection was performed on a TSQ Quantum Ultra triple-quadrupole mass spectrometer (Thermo Fisher Scientific Inc.) equipped with an electrospray ionization interface. Dexmedetomidine and deuterated medetomidine were monitored under positive ion-switching electrospray ionization conditions and quantified in the selected reaction monitoring mode with transitions of mass-to-charge ratio 201.3 to 95.1 and 204.2 to 98.0, respectively.

The ultrahigh-performance liquid chromatography–tandem mass spectrometry method was validated with a lower limit of quantification of 0.05 ng · ml⁻¹. The linear range was 0.05 to 10 ng · ml⁻¹ (r² > 0.99) for dexmedetomidine. The within-batch and between-batch precision levels were less than 7.67%, whereas the accuracy ranged from –3.06 to 11.2%. The bioassay was fully validated according to the Food and Drug Administration (Silver Spring, Maryland) Guidelines.¹⁵  The bioassay showed good linearity, acceptable precision and accuracy, negligible matrix effects, and excellent extraction efficiency.

The concentration-time data for dexmedetomidine were modeled by first-order conditional estimation with interaction using the nonlinear mixed-effect modeling program Phoenix NLME (Version 7.0, Certara L.P. Pharsight, USA).

The clinical team recorded the precise drug administration and sampling time using dedicated bedside reporting documentation and then transcribed them into a case report form. Missing observations or concentration data were excluded from the analysis. Missing covariate values were replaced by previous values recorded before surgery from the same individual or interpolated for time-dependent covariates. The rest, if not resolved, was replaced with the median value from the study population.¹⁶  Concerning the management of plasma concentrations below the quantification limit the M3 Method was used to fit the pharmacokinetics model.¹⁷  The M3 method allowed the below the quantification limit observations to be retained but handled them as censored observations under the assumption that all the concentrations were normal. The likelihood for all the data to be maximized with respect to the model parameters, and the likelihood for a below the quantification limit observation in particular, were taken to be the likelihood that the observation was indeed below the quantification limit.¹⁷ 

By visually inspecting the raw data and reviewing the literature, it was deemed likely that a one- or two-compartment disposition model would suffice, with a possible need for an absorption lag. Therefore, one- or two-compartment open models were compared, and each model had first-order absorption with or without a lag time to describe the absorption phase. The interindividual variabilities were assumed to follow log-normal distributions (η on CL/F, η on Q/F, η on V₁/F and V₂/F with covariance) and were implemented in the base model as (equation 1)

where P_i is the estimated pharmacokinetic parameter value for the i^th subject. P_pop is the mean pharmacokinetic parameter, and η_i is the interindividual variability between the log-transformed individual-specific parameter and a typical parameter. Independent and identically distributed random variables were normally distributed around 0 with variance ω² and the variable i for the i^th individual.

The models were parameterized using the first-order absorption rate (Ka), apparent central volume of distribution (V₁/F), apparent peripheral volume of distribution (V₂/F), apparent clearance of central compartment (CL/F), apparent clearance of peripheral compartment (Q/F), and lag time.

A combined proportional and additive model was evaluated to estimate the residual variability based on a visual inspection of the routine diagnostic plots and improvement of the objective function value. The residual variability is expressed as Cm_ij = C_pij + C_pij*ε_prop.ij +ε_add,ij, where Cm_ij is the j^th observed concentration of the i^th subject, C_pij is the j^th model-predicted concentration of the i^th subject, and ε_prop.ij and ε_add,ij are the random variables with a mean of zero and variance σ²_prop,ij and σ²_add,ij. The residual variability in sample collection, analytical determination, and model-misspecification risk might result in estimated variance (σ²).¹² 

Demographic characteristics, such as age, weight, and liver function (alanine aminotransferase, aspartate aminotransferase), could serve as covariates and could be used to investigate their influences on pharmacokinetics.¹⁸ 

The allometric power models related to pharmacokinetic parameters are represented in equation 2 and equation 3,

where F_size is the fractional difference in allometrically scaled size compared to an individual with a median weight, weight_median is the standard median weight of the study population, and the coefficient is an empirically derived constant; and

where F_size is the fractional difference in allometrically scaled size compared to an individual with a weight of 70 kg. Weight represents an individual’s body weight (i), and 70 kg is the standard adult body weight. For both allometric models described in equation 2 and equation 3, apparent peripheral clearance and volume of distribution were standardized with a body weight of 70 kg or median weight using the allometric coefficients of 0.75 for clearance and 1 for distribution.¹⁸ 

Collinearity of age and size is a fundamental feature, but they are not mutually exclusive.¹⁹  Age-dependent changes in body weight affecting drug disposition should also be considered.²⁰  After examining the influence of body weight on the basic model, a sigmoid maximum effect model was tested to account for differences in body size and maturation function of dexmedetomidine clearance on postmenstrual age.¹²  The model allows gradual maturation of clearance in the early life stage (equation 4),¹⁸ 

where F_mat is the fraction of adult dexmedetomidine clearance value, TM₅₀ is the postmenstrual age at which the clearance is 50% of the mature value, and γ is the Hill coefficient for clearance.

Clearance (CL/F) could then be described as follows (equation 5):

where CL_pop is the population estimated value of clearance.

After delineating size and age, the forward and inclusion-backward elimination method was performed to analyze other potential covariates in the nested model. In the forward step of the covariate screening period, a covariate was included if the best improvement in the goodness-of-fit was found, and a statistically significant decrease of at least 10.60 in objective function value (P < 0.005, with 1 degree of freedom) was obtained. Subsequently, all significant covariates were added. The backward deletion was performed using a stringent statistical significance of P < 0.001 to preserve the influenced covariates in the final model for an increase in objective function value of greater than 13.82 (with 1 degree of freedom). Model convergence, reasonable estimates of parameter values, and precision were also considered for covariate selection.

A nonparametric bootstrap resampling method was applied to evaluate the stability and performance of the final model. The original dataset was resampled at the subject level to generate 1,000 new datasets. The 1,000 resampled datasets were used to obtain the 2.5th and 97.5th percentiles of the simulated model parameters. If the model was valid, the parameter estimates derived from the original dataset were similar to the median and were included in the 2.5th and 97.5th percentiles. The final model was evaluated using a prediction-corrected visual predictive check.²¹  This method generated virtual predictions and observations based on some values obtained by Monte Carlo simulations. The percentiles of the simulated data were compared with the corresponding percentiles of the observed data.

A wide range of intranasal dosages was reported to be used in children;²²  the most commonly reported dose was 1 to 4 µg · kg⁻¹.^23,24  Therefore, the pharmacokinetic profiles at dosage regimens of 1 to 4 µg · kg⁻¹ for children (postmenstrual age, 101 weeks; weight, 10 kg) were simulated via Monte Carlo simulations. All the children in the simulation were assumed to be term births with a gestational age of 40 weeks. For each scenario, 1,000 replications were performed. The time-concentration profiles for the male term children whose ages ranged from 3 months to 3 yr (3, 6, and 9 months, and 1, 2, and 3 yr) were also simulated. Clinical covariates were based on the 50th percentile weight estimates per age for boys as provided by the Clinical Growth Charts from Chinese children references.²⁵ 

Model building was conducted using Phoenix NLME (Version 7.0, Certara L.P. Pharsight). Graphs were prepared using GraphPad Prism 8 (GraphPad Software Inc., USA). The patients’ characteristics were summarized as the median (interquartile range) and range of observations, whichever was appropriate. All model parameters are reported as estimated values with relative standard errors. A two-tailed unpaired t test was performed to compare the time to reach the maximum plasma concentration (T_max) and the maximum concentration (C_max) between different age groups by simulated intranasal dexmedetomidine values. The results were analyzed using P < 0.05 for statistical significance.

A total of 140 Chinese children were recruited from June 9, 2016, to November 9, 2017. Three patients dropped out of the study because of the cancellation of scheduled surgery; thus, 137 children completed the study and were included in the analysis. The demographic information is summarized in table 1.

The median (interquartile range) duration of anesthesia was 30.0 min (24.0 to 44.5 min). The median (interquartile range) wake-up time (from anesthesia start to the time to reach University of Michigan (Ann Arbor, Michigan) Sedation Score [0 to 1]) was 43.5 min (32.5 to 55.8 min), and the median (interquartile range) time for the patients to be discharged to the ward after anesthesia was 50 min (35 to 65 min). Eight patients experienced hypotension during the operation, and one 33-month-old patient required epinephrine to treat hypotension after his SBP fell to 58 mmHg. No episodes of oxygen desaturation, bradycardia, or hypertension occurred.

There were 685 scheduled blood sample collections; however, the investigator failed to collect 37 blood samples because of other clinical commitments. An additional 45 samples were missed due to blocked cannula, and 17 samples were not collected because of parental refusal. In total, 586 samples were included in the final analysis, and all the included samples were taken within the effective window of each optimal sampling schedule. For the concentration data below the lower limit of quantification, a likelihood-based approach was applied where lower limit of quantification data were flagged and treated as categorical data.^17,26  In this study, 32 (5.46%) plasma concentrations were below the quantification limit, and the values for below the quantification limit observations that were considered normally distributed were replaced by random values somewhere between negative infinity and the lower limit of quantification.²⁶  The plasma concentrations of dexmedetomidine over time for the three groups divided by age (3- to 12-month-old subjects in Group 1, 13- to 23-month-old subjects in Group 2, and 24- to 36-month-old subjects in Group 3) are presented in Supplemental Digital Content 1 (http://links.lww.com/ALN/C857). The measured concentrations and individual/population predictions versus time are shown in Supplemental Digital Content 2 and 3 (http://links.lww.com/ALN/C858, http://links.lww.com/ALN/C859).

A two-compartment model described intranasal dexmedetomidine pharmacokinetics with first-order elimination from the central compartment. Adding the second compartment yielded a smaller improvement in fit than that yielded from the one-compartment model (Δ objective function value, –4.6; diagnostic plots shown in Supplemental Digital Content 4 (http://links.lww.com/ALN/C860). Although the objective function value improvement was minor, the two-compartment model was still selected. The reasons were that the one-compartment model exhibited larger errors and interindividual variances concerning the associated pharmacokinetic parameters (17 to 122%) and showed poorer diagnostic plots when compared with the two-compartment model.

The basic models were estimated with or without lag time (−2LL = −681.3, if the lag time was absent). Different lag times were also compared (data not shown); the lag times were fixed at 0.5 min (Δ objective function value, −80.3) and 5 min (Δ objective function value, −150.3) but with V₁/F lower than 10 l, and 10 min (Δ objective function value, 424.3). Finally, the basic model with an estimated lag time of 1.62 min (Δ objective function value, −12.8) fit the data. Given that the estimate of the proportional residual error (σ²_{prop, ij}) asymptotically approached zero, this type of error was removed from the model, and only the additive model (σ²_{add, ij}) was retained (model 3 and model 4 in table 2).

The weight-based allometric method (standardization for a 70-kg adult, equation 3) was applied to the structural model (allometric coefficients of 0.75 for CL/F and Q/F and 1 for V₁/F and V₂/F), which caused a slight decrease in objective function value (Δ objective function value, –11.2). By comparison, a linear weight-normalized model with standardization for median weight yielded no improvement compared with that in the model standardized for a weight of 70 kg (Δ objective function value, 24.3). The addition of maturation of dexmedetomidine clearance based on postmenstrual age improved the model to an even greater extent than a 70-kg standardization allometric model (Δ objective function value, –96.9). The maturation parameters for clearance were estimated as follows: TM₅₀ = 44.7 (coefficient of variation percentage, 17.3%) and γ = 2.96 (coefficient of variation percentage, 10.7%). Moreover, gestational age and postnatal age alone or together were not superior to postmenstrual age alone after adding to clearances (CL/F) and volumes (V₁/F, V₂/F), and other covariates (sex, creatinine, aspartate aminotransferase, and alanine aminotransferase) were not statistically justified when they were incorporated into the model. The description of the covariate analyses is provided in table 2.

Model diagnostics indicated acceptable goodness-of-fit for the final model (fig. 1). Finally, the model was best described by a two-compartment model with first-order elimination, an allometric scaling with estimates standardized to 70-kg weight, and maturation of clearance. The parameter estimates for the intranasal dexmedetomidine population pharmacokinetic model and bootstrap results are presented in table 3.

The reliability and stability of the final model were verified by bootstrapping (table 3) and prediction-corrected visual predictive check (fig. 2). The median of the bootstrap fixed-effects parameter estimates was within 5% of the population estimates from the original dataset for all parameters. The final model revealed a good fit between the predicted and observed dexmedetomidine concentrations, and the 5th, 50th, and 95th prediction intervals simulated from the posterior distribution of the final model parameter estimates were overlaid with the 5th, 50th, and 95th percentiles from the observed data.

Intranasal dexmedetomidine pharmacokinetics were best described by the two-compartment model incorporating weight and postmenstrual age. The simulated concentration-time curves of the dose regimens across 1 to 4 µg · kg⁻¹ are plotted in fig. 3, A and B. The plasma concentration curves for an intranasal dose at 2 µg · kg⁻¹ were simulated in different typical boys with the 50th percentile estimates of weight per age reported in the Clinical Growth Charts for Chinese children.²⁵  The age (including postmenstrual age) and weight of each group were set as follows: (A) a 3-month-old child (postmenstrual age, 53 weeks; body weight, 6.8 kg), (B) a 6-month-old child (postmenstrual age, 66 weeks; body weight, 8.4 kg), (C) a 9-month-old child (postmenstrual age, 79 weeks; body weight, 9.3 kg), (D) a 1-year-old child (postmenstrual age, 92 weeks; body weight, 10.1 kg), (E) a 2-year-old child (postmenstrual age, 144 weeks; body weight, 12.5 kg), (F) a 3-year-old child (postmenstrual age, 196 weeks; body weight, 14.7 kg) (fig. 3C). The terminal half-life parameters of each group were 1.8 ± 0.8 h, 1.6 ± 0.7 h, 1.5 ± 0.6 h, 1.5 ± 0.7 h, 1.5 ± 0.8 h, and 1.5 ± 0.8 h, respectively. Ninety-five percent of the simulated individuals who received 2 µg · kg⁻¹ intranasal dexmedetomidine would achieve the target concentration of 0.3 ng · ml⁻¹ within 20 min. There was no significant difference in the T_max (P = 0.056) or the C_max among the different age groups (P = 0.721). A comparison between the pharmacokinetic parameters of intranasal dexmedetomidine is summarized in table 4.

The current study described the pharmacokinetics of intranasal dexmedetomidine in a large cohort of Chinese children aged 3 months to 3 yr. Using an optimal sampling method combined with nonlinear mixed-effects analysis and simulation, we demonstrated that a target plasma concentration of 0.3 ng · ml⁻¹ would be reached within 20 min in 95% of the simulated individuals treated with intranasal dexmedetomidine at 2 µg · kg⁻¹, and the C_max at 0.563 ng · ml^-1 would be attained at 61 min.

Given the paucity of relevant literature, comprehensive information on the pharmacokinetic profiles of intranasal dexmedetomidine administration in pediatric populations, especially among infants and young children, is still lacking. To date, only two studies based on small sample sizes have reported the pharmacokinetics of intranasal dexmedetomidine administration in this population. One focused on the peak plasma concentration in 18 African-American and White children aged 6 to 48 months,⁸  and the other conducted pharmacokinetic modeling of 13 Chinese children aged 4 to 10 yr.⁷  Miller et al. found that the average C_max values were 0.199 ng · ml⁻¹ and 0.355 ng · ml⁻¹ after intranasal administration of 1 and 2 µg · kg⁻¹ dexmedetomidine, respectively.⁸  On the contrary, it was 0.748 ng · ml⁻¹ in the study by Wang et al.⁷  after intranasal administration of 1 µg · kg⁻¹ dexmedetomidine. In the current study, the average C_max values were 0.499, 0.525, and 0.506 ng · ml⁻¹ for the infant, 1-year-old, and 2-year-old groups, respectively. The simulated intranasal dexmedetomidine C_max values of 0.563 ng · ml⁻¹ at 2 µg · kg⁻¹ and 0.260 ng · ml⁻¹ at 1 µg · kg⁻¹ were consistent with those reported by Miller et al.⁸  with a similar age group. However, they were lower than that in the study by Wang et al.⁷  The discrepancy in the results might be attributed to the patients’ age range and clinical status as well as the bioavailability of dexmedetomidine by different administration methods.²⁷  Wang et al.⁷  recruited older children who underwent different surgeries. In contrast, we enrolled only relatively healthy children with simple vascular malformations in the current study. Dexmedetomidine was administered into the nose by simple drops in the study by Wang et al., whereas an atomization device was used in the study by Miller et al. and our study. Although no notable difference was found between administration by an atomizer or by drops in adult volunteers and children,^4,5  these studies were performed on awake patients in different positions. The bioavailability resulting from different administration methods in anesthetized children would warrant further clarification.

The pharmacokinetic model of intranasal dexmedetomidine was established using an allometric two-compartment disposition model. Clearance changes relatively with weight and organ maturation. Considering the collinearity of weight and age, the use of the allometric scaled model combined with a sigmoidal maturation function facilitated the prediction of mature adult value (70 kg) for comparison across studies. Allometric scaling with an exponent of 0.75 describes drug clearance in children over 2 yr of age, but the allometric exponent then tends to rise with decreasing age.²⁸  Several different methods have been suggested to account for this trend, but they have produced a similar model fit.¹²  We found that the scaled allometric model fit our data better than the linear model. Therefore, we chose the sigmoidal maturation function plus allometric scaling as was used in the study by Potts et al. so that parameters could easily be compared.²⁹ 

The bioavailability of intranasal dexmedetomidine by atomizer was 83.8% with a systemic clearance of 62.4 l · h⁻¹ per 70 kg in 6- to 44-month-old children,⁸  whereas the bioavailability was 40.6 to 82% with a systemic clearance of 33.9 to 44 l · h⁻¹ per 70 kg in healthy adults.^4,30  However, the drug dexmedetomidine in the study by Yoo et al.³⁰  was a more concentrated veterinary formulation, and they used nasal spray instead of the conventional atomization or simple drops for drug administration. Therefore, the bioavailability in the study by Yoo et al. is not comparable. As age-related dexmedetomidine clearances have been demonstrated by Potts et al. using both allometric and linear models,⁹  our apparent clearances estimates were lower than those estimated in older children²⁹  and healthy adults,^4,30  but similar to those reported in the studies by Miller et al. and Wang et al.^7,8  The terminal elimination parameters in our study are also consistent with the studies by Miller et al. and Wang et al. with a 1.8-h terminal half-life. Ebert et al. reported significant changes in cardiac output and heart rate at plasma concentrations of dexmedetomidine that exceeded 1.2 ng · ml⁻¹.³¹  Although body size and age are the two main factors contributing to clearance parameter variability, the hemodynamic effects on clearance under a high dosage of intravenous dexmedetomidine (1 to 6 µg · kg⁻¹) in the study by Potts et al.²⁹  are still worthy of attention. In the study by Potts et al., data were pooled from four separate studies, including those involving patients who underwent different types of surgery, which could contribute to the discrepancy.

Compared with empirical sampling, optimal design in conjunction with simulation scenarios has been proven to improve the precision and accuracy of the pharmacokinetic parameters of dexmedetomidine.^9,32  In this study, we used PopED in R language to determine the optimal design and produce maximal information in pharmacokinetic analysis by providing a strategic sampling schedule with a reduced number of subjects and sampling points in a population of infants and young children. Five blood samples from the three groups were obtained within each optimal sampling window using the D-optimal method. The final pharmacokinetic model provided good accuracy and robustness, with variability ranging from 5 to 21%.

Adult data suggested that dexmedetomidine-mediated sedation and analgesia were dose-dependent.³³  Kim et al. found that the effect-site concentration of dexmedetomidine is strongly correlated with the depth of sedation.³⁴  They reported that concentrations of 0.57, 0.89, and 1.19 ng · ml⁻¹ were associated with mild, moderate, and deep levels of sedation, respectively. The target effect-site concentration between 0.2 and 0.4 ng · ml⁻¹ resulted in a significant level of sedation in healthy volunteers.^35,36  Potts et al. reported that children were aroused from dexmedetomidine infusion sedation at a plasma concentration of 0.304 ng · ml⁻¹, and adequate sedation for children in intensive care units was associated with plasma concentrations of 0.4 and 0.8 ng · ml⁻¹ for moderate and deep sedation, respectively.^9,29  Based on these target concentration values, a dexmedetomidine plasma concentration between 0.3 and 1.0 ng · ml⁻¹ was set as the estimated therapeutic window to produce adequate sedation in our pediatric cohort after general anesthesia.

The simulation results revealed that intranasal administration at 2 µg · kg⁻¹ dexmedetomidine would reach a plasma level of 0.45 ng · ml⁻¹ at 20 min after administration, whereas 3 µg · kg⁻¹ would achieve a plasma level of 0.66 ng · ml⁻¹. Hence, the usual dose of 2 µg · kg⁻¹ or above would be associated with moderate and deep sedation at 20 min after drug administration. These doses would be adequate for nonpainful procedural sedation that lasts for up to 2 h. The simulated C_max obtained by intranasal dexmedetomidine at 3 to 4 µg · kg⁻¹ (0.780 to 1.03 ng · ml⁻¹) would be similar to intravenous dexmedetomidine between 1 and 2 µg · kg⁻¹ (0.783 to 1.24 ng · ml⁻¹).^8,37  Future pharmacokinetic studies are warranted to validate whether a higher dose of intranasal dexmedetomidine would show dose proportionality, as we have assumed in our simulations. This protocol could be transferred into dose-effect–supportive software and guide individualized treatment in clinical practice.

Nevertheless, this study has several limitations. First, although conventional covariates were evaluated, they did not influence the pharmacokinetic parameters. The probable reason was that we enrolled relatively healthy patients whose baseline laboratory values were almost normal. Given that dexmedetomidine is metabolized primarily by UGTs (UGT1A4 and UGT2B10) and CYPs (CYP2A6), information on its pharmacogenetics was not available in this study. Second, despite having used an optimized method for sampling design, the current data did not produce stable Ka and lag time estimates. Third, as intravenous dexmedetomidine administration was not included in this study, the bioavailability of intranasally administered dexmedetomidine was not estimated. Last, the simulations in our study were based on the assumption that the pharmacokinetic profile was linear concerning the dose. However, the absorption of intranasal dexmedetomidine in children is related to the nasal mucosal surface area and anatomy. These factors might be associated with lower bioavailability and behave differently at high dosages, and point to a need for further intranasal dose proportionality studies.

Using an optimal sampling schedule in conjunction with allometrically scaled and maturation models, pharmacokinetic parameters of intranasal dexmedetomidine for children aged 3 months to 3 yr were comprehensively evaluated. Model simulations indicated that intranasal administration at 2 µg · kg⁻¹ dexmedetomidine would be associated with mild to moderate sedation within 20 min, and C_max would be achieved at 61 min. This dose would be feasible for nonpainful procedural sedation that lasts for up to 2 h. An increase in dosage might increase maximum concentrations and prolong the duration of sedation. Model simulations with different dosages should be applied to predict individualized dosing regimens and help to reduce the potential adverse effects associated with overdose.

The authors gratefully acknowledge the assistance of Ao Zheng‚ M.D., Department of Anesthesiology, Guangzhou Women and Children’s Medical Center, Guangzhou, China, for his assistance as a research coordinator in this study. The authors also gratefully acknowledge the assistance of Yao Liu‚ M.D., Si Y. Wang‚ M.D., and Fu L. Jiang‚ Ph.D., Institute of Clinical Pharmacology, School of Pharmaceutical Sciences, Sun Yat-Sen University, Guangzhou, China, for plasma concentration determination and data analysis of the study.

This work was supported by the National Natural Science Foundation of China (Beijing‚ China; grant No. 81901385), National Key Research and Development Program of China during the 13th 5-yr plan (China; grant No. 2018ZX09734-003), and Guangzhou Women and Children’s Medical Center/Guangzhou Institute of Pediatrics (Guangzhou‚ China; grant No. Pre-PI-2019-011).

The authors declare no competing interests.

Full protocol available at: sxjess@126.com. Raw data available at: sxjess@126.com.

Supplementary Figure 1: Dexmedetomidine Concentrations versus Time, http://links.lww.com/ALN/C857

Supplementary Figure 2: Modeled and Measured Concentration over Time, http://links.lww.com/ALN/C858

Supplementary Figure 3: Plasma Concentrations over Time for Individuals, http://links.lww.com/ALN/C859

Supplementary Figure 4: Diagnostic Scatter Plots of the One-Compartment Model, http://links.lww.com/ALN/C860