Population Pharmacokinetics of Intravenous Salbutamol in Children with Refractory Status Asthmaticus

For this PK study, patients were recruited from Erasmus MC-Sophia Children’s Hospital’s pediatric intensive care unit (PICU), a tertiary PICU in The Netherlands. Approval was obtained from the Institutional Review Board, along with written informed consent from parents or legal representatives and patients (12–18 years).

Eligible patients were children with status asthmaticus aged between 0 and 18 years of age who were admitted to the PICU because of respiratory distress and who required intravenous salbutamol. Those patients requiring extracorporeal membrane oxygenation were excluded from the study.

Patients were treated according to our institutional clinical protocol for status asthmaticus (electronic supplementary Appendix I). A continuous salbutamol infusion was started at a rate of 0.1 μg/kg/min. Subsequently, the infusion rate was increased every 10 min depending on clinical symptoms (wheezing, dyspnoe, saturation) and asthma score, with a maximum infusion rate of 10 μg/kg/min (electronic supplementary Appendix I).

Blood for salbutamol concentrations was sampled using an optimized sampling strategy for PK analysis. Ideally, the first blood sample was taken before the start of intravenous salbutamol (plasma drug level after inhaled salbutamol only), but because of the acute setting at admission to the Emergency Department or PICU, and no informed consent having been received at that point in time, the first blood sample was only available if leftover material taken from clinical care was available. During the infusion of salbutamol, serial blood samples were taken at time points of 10, 15–30, 40–60 min, 2–4, 4–6 and 8–16 h after an infusion rate change. Further samples were taken prior to discontinuation, as well as 10, 30–60 min, 2–6 and 12–24 h after discontinuation of the infusion. Blood (0.5 mL) was drawn from an indwelling arterial line or central venous catheter, or, in case this was not available, together with peripheral blood sampling for clinical purposes. Plasma samples were analysed for R- and S-salbutamol concentrations separately using a validated liquid chromatography–tandem mass spectrometry (LC-MS/MS) method; the limit of quantification of the method was 1.0 µg/L, with a linearity of between 1.0 and 500 µg/L for both R- and S-salbutamol (electronic supplementary Appendix II).

As a PD parameter, the Qureshi asthma score was collected at the same time points as blood sampling. This validated score is routinely used in the intensive care unit to monitor symptom progression [12, 13]. It is a 5-item scale (breathing frequency [categories for age], oxygen saturation, auscultation, retractions and dyspnea), with a minimum of 1 and maximum of 3 for each item, resulting in a range of 5–15. A higher score indicates a more severe exacerbation. Possible adverse reactions from salbutamol, such as tachycardia, arrhythmias, hypotension, hypokalemia, hyperglycemia and lactic acidosis were recorded.

Furthermore, age, bodyweight, sex, diagnosis, drug doses, and available kidney and liver function indices were collected for all patients from our computer-based patient data management system.

The population PK model was developed using nonlinear mixed-effects modelling (NONMEM, version 7.2; ICON Development Solutions, Ellicott City, MD, USA). The analysis was performed using logarithmically transformed concentrations and the first-order conditional estimation method with interaction (FOCE-I) [ADVAN5 TRANS1]. Tools used to evaluate and visualize the model were RStudio (version 0.98.1049), R (version 3.1.2), XPose (version 4.5.3) and PsN (version 4.6.0), all with the graphical interface Pirana (version 2.9.0) [14].

Model development was a three-step process: (1) selection of a structural and error model; (2) covariate analysis; and (3) internal validation of the model.

In the first step, a structural model was developed to describe the PK of R- and S-salbutamol. The salbutamol concentration before the start of the intravenous concentration related to the inhalation therapy was estimated based on the available data. The concentration at this time was estimated assuming an equal distribution between R- and S-salbutamol. For each patient, the baseline concentration was estimated using a parameter for the baseline dose, with interpatient variability (IPV). To determine the structural PK model, one- and two-compartment models were tested for both enantiomers. PK parameters were estimated in terms of central volume of distribution (V_c), peripheral volume of distribution (V_p), clearance (CL), and intercompartmental clearance (Q). For each parameter, it was determined whether the parameter could be estimated for both enantiomers together, or whether it had to be estimated separately. To account for the variability in PK parameters due to the varying size of the individual children, the parameter values were standardized to a body weight of 70 kg using allometric scaling, in which the exponential parameter was initially fixed at a value of 0.75 for CL and Q, and a value of 1 for the volumes of distribution [15]. During the covariate analysis, it was evaluated whether estimation of the exponent improved the goodness of fit. The addition of IPV, described using an exponential model, was evaluated for each PK parameter. Residual variability between observed and predicted plasma concentrations was described using an additional error model for logarithmically transformed data.

Once the base model was selected, covariates were tested for their influence on PK parameters. Covariates tested were bodyweight (estimation of exponent for allometric scaling), age, sex, creatinine, alanine transaminase, and urea concentrations. Continuous covariates were normalized to the population median and modeled using an exponential model, while categorical variables were modelled proportionally. Covariates were included using forward inclusion (p < 0.05) and backward elimination (p > 0.001). Additional criteria for the inclusion of covariates in the model were graphical evaluation of the parameter–covariate relationship and decrease in the IPV of the parameter involved.

The final model was validated using normalized prediction distribution errors (NPDEs), using 2000 simulated datasets. This is a simulation-based diagnostic, which can be used to evaluate models developed on datasets with variable dosing regimens.

After validation, the model was used for simulations to evaluate the effect of different dosing regimens with or without a loading dose. The R- and S-salbutamol concentrations were simulated using the final model for a typical child weighing 15 kg. As a result of salbutamol inhalation therapy, a concentration of approximately 20 μg/L was observed just before the start of the intravenous administration. The intravenous salbutamol administration is started as well as this salbutamol concentration, which is described by the model using the baseline dose. The intravenous infusion rate in the first 10 min varied between 0.5 (no loading dose) and 1.5 μg/kg/min (with loading dose), followed by a continuous infusion of 0.5 μg/kg/min for the following 6 h.

Salbutamol concentrations were obtained from 19 children treated with intravenous salbutamol at the PICU between September 2011 and February 2013. Patients were between 9 months and 15 years of age (median 4.9 years) and body weight ranged from 7.8 to 70 kg (median 18 kg) (see Table 1).

A total of 117 plasma samples were analyzed for R- and S-salbutamol. In 8 of the 117 samples obtained from seven different patients, extremely high concentrations of both enantiomers were measured. Six of these concentrations could be explained due to skin contamination of fingerprick blood after nebulizing salbutamol [16]. These samples were subsequently removed from the database, resulting in a final dataset containing 111 samples (median of three samples per patient, range 1–22).

Data could best be described using a two-compartment model, with separated clearances for R- and S-salbutamol (Fig. 1). Immediately after infusion, salbutamol is divided towards two central compartments (k = 10 h⁻¹ fixed), with equal volumes of distribution of 12.9 L for both enantiomers. The covariate analysis showed that body weight was a significant covariate using a power function. The exponent of the power function was fixed at 0.75 for clearance and Q, and at 1 for V_c and V_p. Estimation of the exponent resulted in similar values and did not improve the model. The parameter estimations are presented in Table 2. No other covariates resulted in a significant improvement of the model. Goodness-of-fit plots showed a good fit of the model (Fig. 2). This final model resulted in a CL_R-salbutamol of 5.13 L/h, CL_S-salbutamol of 2.78 L/h, and V_c of 2.76 L for a typical child weighing 15 kg.

The model was evaluated using a bootstrap analysis. Median parameter values, as well as the 5th and 95th percentiles, were in agreement with the model estimations and errors. The distribution of errors was close to normal in the NPDEs (electronic supplementary material 2), with no significant trends in NPDE versus time and NPDE versus predictions.

The simulation in Fig. 3 shows that a loading dose leads to higher R- and S-salbutamol concentrations in the first hour after the start of intravenous salbutamol, compared with the infusion without a loading dose. The infusion rate of 0.5 μg/kg/min without a loading dose resulted in R-salbutamol concentrations of 22.2 μg/L and S-salbutamol concentrations of 29.2 μg/L at 15 min after the start of infusion therapy. With a loading dose, the R-salbutamol concentration was 36.7 μg/L and the S-salbutamol concentration was 45.7 μg/L at 15 min. Longer treatment with intravenous salbutamol results in increasing concentrations of mainly S-salbutamol.