Hydroxycarbamide in children with sickle cell anaemia after first‐dose vs. chronic therapy: pharmacokinetics and predictive models for drug exposure

Abstract

Aims

The purposes of this work were to: (1) compare pharmacokinetic (PK) parameters for hydroxycarbamide in children receiving their first dose (HC_new) vs. those receiving chronic therapy (HC_chronic), (2) assess the external validity of a published PK dosing strategy, and (3) explore the accuracy of dosing strategies based on a limited number of HC measurements.

Methods

Utilizing data from two prospective, multicenter trials of hydroxycarbamide (Pharmacokinetics of Liquid Hydroxyurea in Pediatric Patients with Sickle Cell Anemia; NCT01506544 and Single‐Dose (SD) and Steady‐State (SS) Pharmacokinetics of Hydroxyurea in Children and Adolescents with Sickle Cell Disease), plasma drug concentration vs. time profiles were evaluated with a model independent approach in the HC_new and HC_chronic groups. Various predictive scenarios were analysed to evaluate whether systemic exposure with hydroxycarbamide could be accurately predicted.

Results

Absorption of hydroxycarbamide was rapid, variable and dose independent. Dose‐normalized peak plasma concentrations and drug exposure (AUC) were higher, and weight‐normalized apparent oral clearance was lower in the HC_new group. We assessed a PK‐guided dosing strategy along with other predictive scenarios and found that inclusion of plasma samples only slightly improved the accuracy of AUC predictions when compared to a population‐based method.

Conclusions

Children naïve to hydroxycarbamide exhibit a different PK profile compared to children receiving chronic therapy. Accuracy of population‐based dosing is sufficient to target AUCs in individual patients. Further clearance/bioavailability studies are needed to address the factors responsible for variability in the disposition of hydroxycarbamide.

What is Already Known about this Subject

• Plasma concentrations of hydroxycarbamide, a critical therapy for children with sickle cell anaemia (SCA), are linear to the orally‐administered dose, and the clinical benefits appear to be dose‐dependent.

• Hydroxyurea exhibits a substantial interpatient variability both in pharmacokinetics (PK) and clinical response.

• An individualized PK‐guided dosing strategy using plasma samples performed at initiation of therapy has been developed to rapidly predict an individual's maximum tolerated dose (MTD) of hydroxycarbamide.

What this Study Adds

• This study, for the first time, suggests that apparent clearance of hydroxycarbamide is lower in children while being treated chronically compared to those that are naïve to therapy, and further work to investigate the mechanistic causes for changes in bioavailability over time should be pursued.

• We externally validated the PK‐guided dose individualization strategy for hydroxycarbamide treatment for children with SCA proposed by Dong et al.; however, enrichment of PK sampling in this cohort did not measurably add to the prediction accuracy compared to various population‐based dosing algorithms.

Introduction

Individuals with sickle cell anaemia (SCA; HbSS and HbSβ⁰thalassemia) suffer a wide spectrum of recurrent and debilitating clinical manifestations 1. Hydroxycarbamide (hydroxyurea) is an approved therapy in the United States by the Food and Drug Administration (FDA) for adults and by the European Medicines Agency (EMA) for children and adults.

Hydroxycarbamide is a small, water‐soluble molecule that remains non‐ionized in the gastrointestinal tract and is approximately 100% bioavailable when administered orally 2. Plasma concentrations are linear to the administered dose 3, and at doses used to treat SCA, elimination is linear 4 with 50% excreted as the parent compound in urine 3.

Despite FDA and EMA approval and several decades of experience with hydroxycarbamide in children, there remain several gaps in our knowledge regarding the pharmacokinetics (PK) of hydroxycarbamide. To date, published paediatric PK data have been obtained following a first‐dose of hydroxycarbamide and limited data exist during chronic therapy 5, 6. In these first‐dose studies, a large degree of variation was observed in the absorption rate of hydroxycarbamide, and some variation in maximal drug concentration (C _max) and total drug exposure (AUC_Inf) was also noted 5, 6.

In children treated with chronic hydroxycarbamide therapy escalated to maximum tolerated dose [MTD, defined by an absolute neutrophil count (ANC) goal of 2000–4000 × 10⁶ l⁻¹], a wide range (14.2–35.5 mg kg⁻¹ day⁻¹) of individual MTD doses are identified after an average of 8 months of dose escalation 6, 7, 8, and exposure is not significantly altered with weight‐based dosing schemes in children older than 2 years of age 9.

A recently proposed PK‐guided dosing strategy based on plasma drug samples obtained 15–20 min, 50–60 min, and 3 h after the initial dose of hydroxycarbamide aimed to reduce the time required to reach MTD 10; however, the authors were unable to perform an internal or external validation of their findings.

The purposes of this work are as follows: (1) to compare PK parameters for hydroxycarbamide in children receiving their first‐dose (HC_new) vs. those on chronic therapy (HC_chronic), (2) to assess the external validity of the published PK dosing strategy by Dong et al., and (3) to explore the accuracy of AUC predictions based on a different number of plasma drug concentration measurements.

Methods

Study design

Previously generated data from two prospective, open‐label, single‐dose, multicentre pharmacokinetic trials, ‘Pharmacokinetics of Liquid Hydroxyurea in Pediatric Patients with Sickle Cell Anemia’ (NCT01506544) and ‘Single‐Dose (SD) and Steady‐State (SS) Pharmacokinetics of Hydroxyurea in Children and Adolescents with Sickle Cell Disease’, conducted at eight medical centres in the United States were utilized for this analysis. Both trials received approval from local institutional ethics committees and participants provided informed consent.

Study population

Children (≤18 years of age) with either HbSS or HbSβ⁰thalassemia were eligible for participation. Inclusion criteria included being in a ‘well’ state (i.e., no acute SCA manifestations or other acute illness), weight of ≥10 kg, body mass index (BMI) ≥5th and ≤95th percentiles, and normal hepatic, renal and gastrointestinal function. Children were ineligible if they had a malignancy, received a recent blood transfusion, or had cytopenia noted on screening laboratory assessment. Additionally, children were excluded for concomitant medication usage that could potentially affect the analysis.

This analysis compares children who had never been administered hydroxycarbamide (HC_new) with those that had been treated for ≥3 months at a stable dose (HC_chronic). All children received a physical examination and a medical history was taken, including a detailed medication history (prescription, over‐the‐counter medications and nutritional supplements). Individuals in the HC_chronic cohort withheld their routine dose of hydroxycarbamide for 24 h prior to study drug administration. All children fasted for 8 h prior to receiving the study drug, and they abstained from food or drink 2 h post ingestion, when an age‐appropriate diet was initiated. In order to minimize repeated venipuncture, all blood samples were obtained via a peripheral indwelling venous cannula (PIV). Patency of the PIV was maintained with infusion of sterile 0.9% sodium chloride.

Both cohorts (HC_new and HC_chronic) received a single dose of hydroxycarbamide rounded up to the nearest 100 mg. The HC_new cohort was administered approximately 20 mg kg⁻¹ , and the HC_chronic cohort received their normal daily dose. Blood PK samples were collected pre‐dose, then at 15, 30, 45 and 60 min, and 1.5, 2, 4, 6 and 8 h after the study drug was administered under direct supervision.

Bioanalytical methods

For PK evaluations, whole blood (2 ml) was collected into heparinized tubes, inverted a minimum of 8–10 times, placed immediately on ice, and centrifuged at 750g for 10 min at room temperature. Plasma was separated and then stored at ≤−80°C until shipped to Children's Mercy Hospital on dry ice, where it was stored at ≤−80°C until analysis. Samples were analysed using a validated gas chromatography–mass spectrometry (GC–MS) technique that was linear from 0.1–100 μg ml⁻¹ of hydroxycarbamide 11, 12. The intra‐ and inter‐assay coefficients of variation (CV) were consistently <10% across concentrations spanning the range of linearity 11.

Noncompartmental PK methodology

Plasma concentrations vs. time profiles were evaluated with a model‐independent approach. The maximum plasma concentration (C _max) was obtained directly from the plasma concentration vs. time profile. The area under the plasma concentration vs. time curve (AUC_0‐n) was determined using the mixed log‐linear rule. Extrapolation of the AUC to infinity (AUC_inf) was calculated by summation of AUC_0‐n + C_n/λ_z, where C_n represents the observed plasma concentration at the last quantifiable post‐dose time point and λ_z is the apparent terminal elimination rate constant (calculated from a curve fit of the apparent terminal elimination phase with 1/y weighting). The total plasma clearance (CL/F) and volume of distribution during the terminal phase (V_z/F) were calculated from the AUC_inf. The CL/F and V_z/F were standardized for participant weight (CL/F/BW and V_z/F/BW). To standardize for daily administered dose of hydroxycarbamide and difference in body mass, AUC_inf and C _max were divided by the daily prescribed dose (mg kg⁻¹) (AUC_inf/(Dose/BW), and C _max/(Dose/BW)). This allowed comparison of those parameters between both cohorts (HC_new and HC_chronic).

Statistical analysis

Noncompartmental PK modelling

Descriptive statistics are provided for continuous variables. Ninety‐five per cent confidence intervals are provided for demographic and baseline laboratory parameters. Comparison of two groups was performed with either a two‐sample t‐test or exact Wilcoxon Rank Sum test, depending on normality of the data. Pearson and Spearman correlations were used to assess the relationship between two continuous variables, depending on the normality. Normality was tested using the Shapiro–Wilk test, and all hypothesis tests are considered significant at the α = 0.05 level. Analyses were conducted using SAS software version 9.3 (Cary, NC). Figure 1, Figure 2B and Figure 3 were produced with GraphPad Prism v.6.07 (La Jolla, CA), and all others with SAS software.

Figure 1.

(A) Mean hydroxycarbamide (HC) (μg ml⁻¹) and (B) dose‐normalized HC concentration (μg ml⁻¹ per 1 mg kg⁻¹ HU) per time (h) profiles in children on chronic HC therapy (HC_chronic) and children who are receiving the first dose of drug (HC_new). Error bars represent standard deviation of the mean

Figure 2.

(A) Coefficient of determination (R ²) of the linear correlation between HC concentration and observed AUC_inf at different single sampling time points in 59 children. Plasma hydroxycarbamide (HC) concentrations have a positive relationship with AUC_inf beginning 1.5 h after the initial dosage (P < 0.0001; Pearson) which is maintained until the eighth hour. (B) Scatterplot of a selected relationship between a single measured plasma HC level in 59 children at 4.0 h post HC ingestion and the observed AUC_inf (Supplemental Figure S1)

Figure 3.

Boxplot summarizing the distribution of AUC_inf prediction errors for five tested scenarios. Closed symbols represent naïve participants whereas open symbols represent chronically treated participants. Scenario 1 is based on the equation: 3.48*C_1.5hr + 24.3 using a single plasma hydroxycarbamide (HC) measurement obtained 1.5 h after ingestion. Scenario 2 is a previously published population prediction model 9 with no sampling. Scenario 3 is a MAP‐Bayesian prediction based on a single plasma HC measurement at 1.5 h after ingestion. Scenario 4 is a previously reported MAP‐Bayesian prediction based on three plasma HC measurements (15–20 min, 50–60 min, and 3 h) 10. Scenario 5 is a MAP‐Bayesian predictions based on 12 plasma measurements

Single time point sampling for prediction of drug exposure (internal model)

Single post‐dose plasma hydroxycarbamide concentrations were examined as a function of AUC_inf to determine whether systemic exposure could be reliably predicted from a single hydroxycarbamide plasma concentration in both HC_new and HC_chronic cohorts. A linear regression model was used to assess the association between plasma concentrations and AUC_inf. The goodness of fit was calculated using the coefficient of determination, the adjusted R ².

External validation of models for prediction of hydroxycarbamide exposure

Comparison of different strategies of AUC predictions based on a varying number and time(s) of hydroxycarbamide measurements was performed utilizing an independent data set of plasma concentrations obtained pre‐dose, then at 15, 30, 45 and 60 min, and 1.5, 2, 4, 6, 8, 12, 18 and 24 h in 21 children 3. The methods used to develop and validate a maximum a posteriori Bayesian (MAP) estimator based on limited plasma samples are described in supplemental material. The predictive performance of MAP‐Bayesian estimation was evaluated by the comparison of AUC_inf estimates with reference AUC_inf values obtained with the linear‐trapezoidal rule in five separate scenarios:

1. Scenario 1. AUC predictions based on the previously proposed equation: 3.48*T1.5 + 24.3 3. These predictions are based on a single plasma hydroxycarbamide measurement obtained 1.5 h after ingestion.

2. Scenario 2. Population predictions with no sampling based on a previously published model that assumed a one compartment disposition with an absorption described by a series of transit compartments, allometric (with theoretical exponent) relationship between body weight and apparent clearance, an isometric relationship between body weight and apparent volume of distribution and different mean absorption time for liquid and capsule formulation. Since AUC is related to the apparent clearance and dose, the individual AUC was calculated according to the following equation: Dose/(11.9 [l h⁻¹] × (BW [kg]/70 [kg])^0.75), where 11.9 l h⁻¹ corresponds to the typical apparent clearance of a 70 kg adult.

3. Scenario 3. MAP‐Bayesian predictions based on a single plasma hydroxycarbamide measurement 1.5 h after ingestion and a previously published model 9.

4. Scenario 4. MAP‐Bayesian predictions based on three plasma hydroxycarbamide measurements at time windows suggested by 10 (15–20 min, 50–60 min, and 3 h) and a previously published model 9.

5. Scenario 5. MAP‐Bayesian predictions based on all available plasma measurements for a particular subject and a previously published model 9.

Results

In total, 59 (HC_new, n = 7; HU_chronic, n = 52) children with HbSS at a mean (range) of 9.5 (2.4–17.9) years of age had a complete plasma hydroxycarbamide profile. The average (mean [95% CI]) age (11.4 [8.8–14.1] vs. 9.2 [7.9–10.6] years; P = 0.19) and body weight (40.5 [23.6–57.4] vs. 31.3 [26.7–36.0] kg; P = 0.07 were similar between the HC_new and HU_chronic groups, respectively. The HC_new group received a lower mean dose per kilogram compared to children in the HC_chronic group (17.9 vs. 23.8 mg kg⁻¹, P < 0.01). No participant withdrew from the study prior to completion, and all participants tolerated the hydroxycarbamide dose and all study related procedures without apparent adverse events attributed to them.

Noncompartmental PK results

Mean plasma concentrations (μg ml⁻¹) and dose normalized mean plasma concentrations (μg ml⁻¹ per 1 mg kg⁻¹) vs. time (h) profiles of study participants in both the HC_new and HC_chronic cohorts are illustrated in Figure 1 and non‐compartmental PK parameters are summarized in Table 1.

Table 1.

Noncompartmental pharmacokinetic parameters for hydroxycarbamide in new and chronic dosing

Parameters, mean (SD)	HCNew (n = 7)		HCChronic (n = 52)
	Mean (SD)	95% CI	Mean (SD)	95% CI	P‐value
Cmax (μg ml−1)	35.5 (14.6)	22.0–49.0	33.8 (9.4)	31.1–36.4	0.85*
Cmax/BW/Dose (μg ml−1 per 1 mg kg−1)	1.96 (0.62)	1.4–2.5	1.40 (0.39)	1.3–1.6	0.025*
Tmax (h)	0.9 (0.58)	0.4–1.4	0.8 (0.47)	0.6–0.90	0.68*
AUCINF (μg h−1 ml−1)	105.1 (26.2)	80.9–129.3	107.3 (27.5)	99.6–114.9	0.84**
AUCINF/BW/Dose (μg h−1 ml−1/(mg kg−1))	5.9 (1.0)	5.0–6.7	4.6 (1.0)	4.3–4.9	0.002**
λz (h−1)	0.34 (0.03)	0.31–0.36	0.32 (0.05)	0.31–0.34	0.50**
Vz/F/BW	0.52 (0.04)	0.48–0.55	0.61 (0.22)	0.54–0.67	0.43*
Cl/F/BW (l h−1 kg−1)	0.17 (0.03)	0.15–0.20	0.23 (0.06)	0.21–0.25	<0.001*

C _max, maximum observed plasma concentration; C _max/(Dose/BW), dose/weight‐normalized maximum concentration; T _max, time to maximum plasma concentration; AUC_INF, area under the concentration vs. time curve calculated by summation of AUC_0‐n + C_n/λ_z, where C_n represents the observed plasma concentration at the last quantifiable post‐dose time point and λ_z is the apparent terminal elimination rate constant; λ_z, terminal elimination slope constant; Vz/F/BW, apparent volume of distribution normalized for body weight; Cl/F/BW, clearance normalized for body weight. Parameters for first and chronic dosing were compared with the

^* Exact Wilcoxon rank sum test or

^** two sample t‐test.

The mean time from administration of hydroxycarbamide to maximal plasma concentration (T _max) was 0.8 h but a wide range (0.2 to 2.0 h) was observed, corresponding to a large CV (61.5%). Mean maximal peak plasma concentrations (C _max) were similar (34.0 μg ml⁻¹) in the HC_new and HC_chronic cohorts, but exhibited a broad range (15.2–65.6 μg ml⁻¹) and a moderate degree of inter‐patient variability with a CV of 30%. After dose/weight normalization, the C _max was 1.4‐fold higher in the HC_new group compared to the HC_chronic group (2.0 vs. 1.4 mg l⁻¹ per 1 mg kg⁻¹ dose, P = 0.025). There was less variability observed in post‐peak plasma concentrations (C _max) and in dose‐normalized plasma concentration (C _max/dose) and those in hydroxycarbamide's elimination phase (i.e., concentrations at 4–12 h post‐dose, Figure 1A and 1B). As expected from the visual depiction of the mean plasma concentration vs. time data (Figure 1A), the mean elimination half‐life for hydroxycarbamide did not differ between the HC_new (2.1 h) and HC_chronic (2.3 h) cohorts.

The mean area under the plasma drug concentration–time curve extrapolated to infinity (AUC_inf) or body exposure to hydroxycarbamide was similar between groups (107 μg h⁻¹ ml⁻¹) but had a wide range (51.7–182.2 μg h⁻¹ ml⁻¹) and a moderate degree of inter‐patient variability with a CV of 25.3%. After dose normalization, the HC_new group had a 1.3‐fold higher exposure to hydroxycarbamide (5.9 vs. 4.6 mg l hr⁻¹ per 1 mg kg⁻¹, P = 0.002). The mean body weight‐normalized volume of distribution (Vz/F/BW) was 0.59 l kg⁻¹ with a moderate degree of inter‐patient variation with high CV (35.6%) and a range of 0.3–1.5 l kg⁻¹, but no difference was observed between HC_new and HC_chronic. The average body weight‐normalized apparent clearance (Cl/F/BW) of hydroxycarbamide for all children was 0.22 l hr⁻¹ kg⁻¹, which had a moderate degree of inter‐patient variability with CV (27.2%) and a range of 0.1–0.5 l hr⁻¹ kg⁻¹. Apparent oral clearance was 26% lower in the HC_new group compared to the HC_chronic group (0.17 vs. 0.23 l hr⁻¹ kg⁻¹, P < 0.001).

Limited sampling strategy for prediction of drug exposure (internal model)

In the entire cohort, AUC_inf has a positive relationship with timed plasma concentrations, which becomes and maintains strong significance starting 90 min after the initial dose (Figure 2A). For example, 4 h after hydroxycarbamide ingestion, the fitted model is AUC_inf = 35.24 + 6.81 × plasma concentration, which gives a coefficient of determination (R ²) for the linear regression model of 0.77 (i.e. 4 h after hydroxycarbamide ingestion, the plasma concentration explains about 78% of the variance of AUC_inf) (Figure 2B). The HC_chronic group displays similar relationships where timed plasma concentrations are strongly association with AUC_inf beginning at 90 min after the initial dose (and this association is maintained all the way until the eighth hour) (Figure 2A). The HC_new group displays a similar pattern except it is not as definitive. The first significant association between plasma concentrations and AUC_inf starts at 4 h after the initial dose (R ² = 0.7), and this relationship is maintained (Figure 2A).

Validation of Bayesian estimators

Five different prediction scenarios (described in methods section above) were tested utilizing the maximum Bayesian estimation methodology described in the supplemental materials. The median prediction error (MDPE) and median absolute prediction error (MDAPE) of AUC predictions were low (<10% and <20%) and comparable across all scenarios (Figure 3 and Table S1). In summary, the accuracy of prediction for drug exposure (AUC_inf) to hydroxycarbamide was not improved with the addition of single, multiple or numerous plasma concentrations when compared to population‐based model predictions.

Discussion

In these prospectively collected data, we report the novel finding that dose‐dependent PK parameters vary between children who were naïve to hydroxycarbamide and those treated for more than 3 months at a stable dosage. Specifically, the dose and weight‐normalized peak plasma concentrations (C _max/BW/dose) and drug exposure (AUC_inf/DW/dose) were lower in children treated with hydroxycarbamide for more than 3 months. We externally validated the published PK‐guided dosing strategy by Dong et al. 10; however, we report that population predictions (without any plasma measurements) are accurate enough to make a basis for dose adjustments. In the five predictive scenarios we tested, there was little benefit to be gained from use of MAP Bayesian estimation with additional plasma measurements in terms of accuracy and precision of AUC predictions.

Overall, the non‐dose‐normalized PK parameters for hydroxycarbamide from our study were comparable to other published populations 5, 6, 9 (Table S2). The apparent rate of absorption of hydroxycarbamide was rapid and showed a large degree of inter‐individual variation, but was not affected by the dose administered or by the chronicity of therapy. The large degree of variation in absorption time has partially been explained by potential variations in transmembrane transporters that have been implicated in the absorption of hydroxycarbamide 6, 13.

Wide ranges in individual peak plasma concentrations were observed, which is consistent with other reports in both younger and older children 5, 6, 9. Children naïve to hydroxycarbamide achieved similar maximum plasma concentrations (C _max) and drug exposures (AUC_inf) compared to those treated with chronic therapy, albeit with a lower administered dose. With dose and body weight normalized, naïve children have a higher plasma concentration per dose administered (C _max/(Dose/BW), drug exposure per dose administered (AUC_inf/(Dose/BW)), and a lower clearance and volume of distribution normalized for body weight (CL/F/BW). The mechanistic basis for these differences is not yet clear.

These novel observations have clinically relevant implications. First, the bioavailability (CL/F) of hydroxycarbamide changes with chronic therapy; specifically, given a fixed dose of hydroxycarbamide, drug exposure declines during chronic therapy. Understanding this relationship is vitally important and may provide insight into why paediatric patients appear to respond better to therapy when compared to adult populations 14, 15. Second, understanding why bioavailability declines over time with exposure to hydroxycarbamide may provide some insight into why some individuals respond well to hydroxycarbamide and others respond poorly.

We were able to externally validate the PK‐guided dose individualization strategy for hydroxycarbamide treatment for children with SCA proposed by Dong et al. 10. We also evaluated the predictive performance of other scenarios utilizing MAP‐based AUC predictions with none, one and 12 plasma samples and a direct relationship between AUC and HC concentration measurement 1.5 h after drug administration. The prediction error, MDPE and MDAPE were low (<10% and <20%) and comparable across all the tested scenarios. Our data suggest that the enrichment of PK sampling as performed in our study did not measurably add to the prediction accuracy attained with paediatric population‐based dosing algorithms.

In summary, weight‐normalized PK parameters suggest that apparent clearance of hydroxycarbamide is lower in children with chronic dosing compared to those that are naïve to therapy. Further work to investigate the mechanistic causes for changes in bioavailability over time in children treated with hydroxycarbamide should be pursued.

Competing Interests

J.H.E. receives funding support from Daiichi Sankyo, Global Blood Therapeutics, and Eli Lilly and Co. and serves as an advisor for Daiichi Sankyo. The other authors have no potential conflicts to report.

This study was supported by the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD), under the authority of the Best Pharmaceuticals for Children Act (BPCA) contract no. HHSN201000003I/task order no. HHSN27500004 (K.A.N.), by a grant [5K23HL077684‐02 (K.A.N.)] from the National Heart Lung and Blood Institute (NHLBI) of the National Institutes of Health (Bethesda, MD), and by the American Lebanese Syrian Associated Charities (ALSAC).

The authors would like to thank the Pediatric Trials Network for their support and the children and families who participated in the two pharmacokinetic studies analysed for this manuscript. The authors also appreciate the efforts of the laboratory personnel, study coordinators, and nursing staff at each of the participating institutions.

Contributors

J.H.E. contributed to the overall research design, analysed the results, and wrote the manuscript; P.W., J.M. and G.K analysed results and provided critical review and editing of the manuscript; R.I.P. enrolled patients and provided critical review and editing of the manuscript; J.P. enrolled patients and provided critical review and editing of the manuscript; U.G. designed the protocol, analysed samples and provided critical review and editing of manuscript; G.K designed the protocol and provided critical review and editing of manuscript; K.A.N. contributed to the overall research and protocol design, served as study chair, executed the protocol, analysed results and wrote the manuscript.

Supporting information

Table S1 Median prediction error (MDPE) and median absolute prediction error (MDAPE) across tested scenarios

Table S2 Summary of published paediatric pharmacokinetic parameters of hydroxycarbamide

Figure S1 Scatterplots of all relationships between a single measured plasma hydroxycarbamide (HC) level in 59 children at 0.25 (Time 1), 0.5 (Time 2), 0.75 (Time 3), 1.0 (Time 4), and 1.5 (Time 5), 2.0 (Time 6), 4 (Time 7), 6 (Time 8), and 8 (Time 9) hours post HC ingestion and the observed AUC_inf. DV, dependent variable. Green line is the mean; red line median; dotted lines represent the 95% confidence interval. The goodness of fit was calculated using the coefficient of determination, the adjusted R ²

Estepp, J. H. , Wiczling, P. , Moen, J. , Kang, G. , Mack, J. M. , Liem, R. , Panepinto, J. A. , Garg, U. , Kearns, G. , and Neville, K. A. (2018) Hydroxycarbamide in children with sickle cell anaemia after first‐dose vs. chronic therapy: pharmacokinetics and predictive models for drug exposure. Br J Clin Pharmacol, 84: 1478–1485. doi: 10.1111/bcp.13426.