Population pharmacokinetics and dosing recommendations for the use of deferiprone in children younger than 6 years

Abstract

Aims

Despite long clinical experience with deferiprone, there is limited information on its pharmacokinetics in children aged <6 years. Here we assess the impact of developmental growth on the pharmacokinetics of deferiprone in this population using a population approach. Based on pharmacokinetic bridging concepts, we also evaluate whether the recommended doses yield appropriate systemic exposure in this group of patients.

Methods

Data from a study in which 18 paediatric patients were enrolled were available for the purposes of this analysis. Patients were randomised to three deferiprone dose levels (8.3, 16.7 and 33.3 mg kg⁻¹). Blood samples were collected according to an optimised sampling scheme in which each patient contributed to a maximum of five samples. A population pharmacokinetic model was developed using NONMEM v.7.2. Model selection criteria were based on graphical and statistical summaries.

Results

A one‐compartment model with first‐order absorption and first‐order elimination best described the pharmacokinetics of deferiprone. Drug disposition parameters were affected by body weight, with both clearance and volume increasing allometrically with size. Simulation scenarios show that comparable systemic exposure (AUC) is achieved in children and adults after similar dose levels in mg kg⁻¹, with median (5–95^th quantiles) AUC values, respectively, of 340.6 (223.2–520.0) μmol l⁻¹ h and 318.5 (200.4–499.0) μmol l⁻¹ h at 75 mg kg⁻¹ day^–1, and 453.7 (297.3–693.0) μmol l⁻¹ h and 424.2 (266.9–664.0) μmol l⁻¹ h at 100 mg kg⁻¹ day^–1 given as three times daily (t.i.d.) doses.

Conclusions

Based on the current findings, a dosing regimen of 25 mg kg⁻¹ t.i.d. is recommended in children aged <6 years, with the possibility of titration up to 33.3 mg kg⁻¹ t.i.d.

What is Already Known about this Subject

• Deferiprone pharmacokinetics has been characterised in adults and adolescents

• After oral administration, deferiprone is rapidly and well absorbed, and plasma levels show peak concentrations within 1 h of administration

• Essentially no pharmacokinetic information is available in children aged <6 years despite long clinical experience with this iron chelator

What this Study Adds

• The pharmacokinetics of deferiprone has been characterised in children younger than 6 years after administration of single oral doses.

• Body weight is a covariate on clearance and volume of distribution across the paediatric population.

• The approved dosing regimen for deferiprone yields exposure in children comparable to that observed in adults and adolescents.

• A dosing regimen of 25 mg kg⁻¹ three times daily is recommended for children below 6 years, with the possibility of titration up to 33.3 mg kg⁻¹ three times daily.

Table of Links

LIGANDS
Deferiprone

This Table lists key ligands in this article that are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY 1.

Introduction

Patients with haemoglobinopathies, such as β‐thalassaemia or sickle cell disease, which affect the ability to synthesise haemoglobin may require life‐long blood transfusion therapy to survive. This chronic intervention results in a series of potential complications, with iron overload being an inevitable consequence within a few years. Chelation therapy is therefore required to prevent potentially fatal iron‐related complications. In most cases, the disease is diagnosed within the first year of life and blood transfusion regimen is started immediately after diagnosis. Chelation therapy is subsequently initiated when serum ferritin levels reach a threshold of about 1000 μg l⁻¹, which occurs on average about 1 year after the start of blood transfusions 2, 3, 4, 5, 6. Deferiprone is a hydroxypyridinone, which was authorised in Europe in 1999 for the treatment of iron overload in patients with β‐thalassaemia major when deferoxamine is contraindicated or inadequate. The recommended dose of deferiprone is 75 mg kg⁻¹ day^–1 given as a three times daily (t.i.d.) regimen; the dose can be increased up to 100 mg kg⁻¹ day^–1, if necessary. When administered orally, deferiprone is rapidly and well absorbed. Plasma levels show peak concentrations (C_max) within 1 hour of administration. Food reduces its absorption rate without much of an effect on the overall exposure to the drug. In patients with β‐thalassaemia, the administration of deferiprone at doses of 37.5 mg kg⁻¹ twice‐daily yields C_max of 34.6 mg l⁻¹ and area under the plasma concentration–time curve (AUC) of 137.5 mg l⁻¹ h 7, 8. By contrast, peak serum concentrations were 17.53 mg l⁻¹ and 11.82 mg l⁻¹ in fasting and fed states, respectively after a dose of 25 mg kg⁻¹ 9. Deferiprone is for the most part inactivated by glucuronidation (>85%) and >90% of the drug is cleared from plasma within 6 hours of ingestion, with an elimination half‐life of 1–2.5 h in patients affected by β‐thalassaemia 6, 7, 10, 11, 12, 13, 14, 15, 16, 17. The chelating effect of deferiprone results from the formation of a 3:1 complex with iron, which is eliminated mainly through the kidneys, as is the free parent drug.

Despite the extensive clinical experience with deferiprone, pharmacokinetic (PK) data in children are sparse, and there are effectively no data in children younger than 6 years. To cover this gap, deferiprone was included in the list of priorities prepared by the Paediatric Committee – European Medicines Agency. The main objective of this analysis is to characterise the systemic exposure to deferiprone in paediatric patients aged <6 years using a model‐based approach and to assess the effect of demographic and physiological factors on the drug's disposition properties. Furthermore, it is our endeavour to identify the dose levels yielding drug exposure comparable to adults.

Methods

Clinical study

This experimental and modelling study was a multicentre, randomised, single‐blind, single‐dose PK study aimed to evaluate the PK of deferiprone in children aged <6 years affected by transfusion‐dependent haemoglobinopathies.

The PK of deferiprone was analysed using data collected from the DEEP‐1 Pharmacokinetic Study (EudraCT, 2012–000 658‐67, clinicaltrial.gov reference number: NCT01740713), in which enrolled patients were randomised to one of three dose levels (8.3, 16.7 and 33.3 mg kg⁻¹). Deferiprone was administered as a single oral dose (80 mg ml⁻¹ solution). This study was sponsored and performed by the DEEP Consortium (www.deep.cvbf.net) according to an approved paediatric investigation plan (EMEA‐001 126‐PIP01–10). Patients undergoing a chronic transfusion program (receiving at least 150 ml kg⁻¹/year of packed red blood cells) and, if naïve to any chelation therapy, having ferritin levels >800 ng ml⁻¹ were considered eligible for the study. In addition, amongst other criteria, patients with haemoglobin levels <8 g dl⁻¹, abnormal liver function and severe heart dysfunction secondary to iron overload, or serum creatinine levels above the upper normal level were excluded from the study. The study protocol was approved by national Ethics Committees and parental consent was obtained for patients' enrolment. All experimental procedures were performed in accordance to good clinical practice guidelines and to the 1964 Helsinki declaration and its later amendments. In brief, 18 children aged <6 years (9 male and 9 female) who received the active medication were included in the analysis. Recruitment of up to 30 patients was envisaged in the protocol to ensure an appropriate sample size, if data from 18 evaluable subjects did not suffice to provide parameter estimates with adequate precision (< 40%). In practice, the use of nonlinear mixed‐effects modelling allowed completion of the study with the data of the first 18 evaluable subjects by providing accurate and precise estimates of the main parameters of interest. Blood samples (2 ml) for the evaluation of deferiprone concentrations were collected according to WHO guidelines, with total blood volume collected per patient not exceeding the maximum recommended values. A matrix was used for sampling purposes, including one pre‐dose sample and the following sampling times after dosing: 0.167, 0.25, 0.333, 0.67, 0.83, 0.916, 1.083, 1.167, 1.25, 1.416, 4.5, 5.5, 6, 7 and 8 h. A maximum of five postdose samples were collected per subject according to three different sampling schemes derived from an optimal design analysis previously performed by our group. Further details on the sampling scheme can be found elsewhere 18. Blood samples were drawn by peripheral venous catheter following discard of 2 ml of blood; catheters were filled with saline (i.e., saline lock) between sampling times. Samples were collected in citrate tubes and maintained at 4°C in water and ice until centrifugation; a maximum interval of 1 hour was allowed between sample collection and centrifugation. Samples were then centrifuged at 2000 g for 10 minutes at 4°C and plasma was thereafter transferred into a cryovial and stored at –80°C until analysis.

Bioanalysis

Deferiprone plasma concentrations were analysed by the laboratory of the Division of Pharmacology (Leiden, the Netherlands) using a validated method previously developed by ApoPharma (Toronto, Canada) consisting of high‐performance liquid chromatography with UV detection without internal standard 19. Extraction of deferiprone from supernatant was performed after precipitation of plasma proteins by trichloroacetic acid (15%) and centrifugation at 10 000 g for 20 min at 4°C. The analytical column used for the analysis was a Hamilton PRP‐1 and separation of the chromatogram of interest was achieved using an isocratic mobile phase (pH 7.0). Mean retention time for deferiprone was 5.945 min (standard deviation: 0.087 min). The analytical range was between 3.13 and 800 μmol l^–1 (equivalent to 0.43 to 111 μg ml⁻¹) and an R² value >0.98 was used as acceptance criterion for the calibration curve. The lower limit of quantification was 0.238 μmol l^–1 (equivalent to 0.033 μg ml⁻¹). Inter‐ and intraday accuracy and precision were always <6%, except for the interday precision at 3.13 μmol l^–1, which was found to be 10.7%.

Pharmacokinetic modelling

Nonlinear mixed effects modelling was performed in NONMEM version 7.2 (Icon Development Solutions, Ellicott City, MD, USA). Model building criteria included: (1) successful minimisation; (2) standard error of estimates; (3) number of significant digits; (4) termination of the covariance step; (5) correlation between model parameters; and (6) acceptable gradients at the last iteration.

Fixed and random effects were introduced into the model in a stepwise manner. Interindividual variability in PK parameters was assumed to be log‐normally distributed. A parameter value of an individual i (posthoc value) is therefore given by the following equation:

$${\mathrm{\theta }}_{\mathrm{i}}={\mathrm{\theta }}_{\mathrm{TV}}\times {\mathrm{e}}^{\mathrm{\eta i}}$$

where θ_TV is the typical value of the parameter in the population and η_i is assumed to be a random variable with zero mean and variance ω². Residual variability, which comprises measurement and model error, was described with a proportional error model. This means for the j ^th observed concentration of the i ^th individual, the relation Y_ij:

$${\mathrm{Y}}_{\mathrm{ij}}={\mathrm{F}}_{\mathrm{ij}}+{\mathrm{\epsilon }}_{\mathrm{ij}}\times \mathrm{W}$$

where F_ij is the predicted concentration and ε_ij the random variable with mean zero and variance σ². W is a proportional weighing factor for ε.

Goodness of fit was assessed by graphical methods, including population and individual predicted vs. observed concentrations, conditional weighted residual vs. observed concentrations and time, correlation matrix for fixed vs. random effects, correlation matrix between parameters and covariates and normalised predictive distribution error 20, 21. Comparison of hierarchical models was based on the likelihood ratio test. A superior model was also expected to reduce intersubject variability terms and/or residual error terms.

With the objective of increasing the stability of the model and reducing the uncertainty around the parameters of interest, the Normal‐Inverse Wishart distribution was used as prior during the estimation steps in NONMEM 22. Primary PK parameters estimated with a previously developed model in adults 23 were used as prior information for the PK analysis of deferiprone in the target paediatric population.

Covariate analysis

Continuous and categorical covariates were tested during the analysis. The relationship between individual PK parameters (posthoc or conditional estimates) and covariates was explored by graphical methods (plot of each covariate vs. each individual parameter). Relevant demographic covariates (body weight, height, age and sex) were entered one by one into the population model (univariate analysis). After all significant covariates had been entered into the model (forward selection), each covariate was removed (backward elimination), one at a time. The model was run again and the objective function recorded. The likelihood ratio test was used to assess whether the difference in the objective function between the base model and the full (more complex) model was significant. The difference in – 2Log likelihood between the base and the full model is approximately χ² distributed, with degrees of freedom equal to the difference in number of parameters between the two hierarchical models. Because of the exploratory nature of this investigation, for univariate analyses, additional parameters leading to a decrease in the objective function of 3.84 were considered significant (P < 0.05). During the final steps of the model building process, only the covariates that resulted in a difference of objective function of at least 7.88 (P < 0.005) were kept in the final model.

Model validation

The validation of the final PK model was based on graphical and statistical methods, including visual predictive checks 20. Given the importance of the validation procedures for the subsequent use of a model for simulation purposes, in this study we have included a wide range of diagnostic methods to assess the accuracy of the parameter estimates and the predictive performance of the model 21. Despite the relatively small sample size, bootstrap was used to identify bias, stability and accuracy of the parameter estimates (standard errors and confidence intervals). The bootstrap procedures were performed in PsN v3.5.3 (University of Uppsala, Sweden) 24, which automatically generates a series of new data sets by sampling individuals with replacement from the original data pool, fitting the model to each new data set. Subsequently, parameter estimates were used to simulate plasma concentrations in subjects with similar demographic characteristics, dosing regimens and sampling scheme as in the original clinical studies. Mirror plots were also generated to evaluate the variance–covariance structure of the parameters in the model, which is reflected by the degree of similarity between the original fit and the pattern obtained from the fitting of the simulated data sets using the final PK model.

PK bridging and dosing recommendations

To optimise the dosing regimen of deferiprone in the target population, simulations were performed to identify the doses which resulted in systemic exposure values similar to the adult reference population 23. Simulations were carried out to explore how differences in demographic covariates might affect steady‐state exposure to deferiprone during the course of treatment. Sampling frequency and times were based on a serial sampling scheme for the purposes of estimating AUC, C_max and steady‐state concentration (C_ss) over the dosing interval. Integration of the concentration vs. time data was based on the trapezoidal rule to ensure realistic estimates of variability. The adequacy of the simulated dosing regimens was assessed graphically by determining the fraction of the paediatric population reaching systemic exposure comparable to the target value observed in adults.

A hypothetical clinical trial including a 1‐week treatment according to a t.i.d. regimen was chosen for the purpose of this simulation exercise. Each scenario consisted of 1000 simulations. Two dosing regimens were simulated in both populations: 75 and 100 mg kg⁻¹/day as three daily doses of 25 and 33.3 mg kg⁻¹ respectively. The PK parameters of interest (AUC, C_max and C_ss) were estimated after administration of the first dose on Day 7.

A PK model developed in adult healthy subjects 23 was used to simulate deferiprone exposure in the reference patient population. A cohort of 100 subjects (50 male and 50 female) with a body weight distribution of mean 55 and standard deviation 7.5 kg was used to describe the demographic characteristics of a typical adult thalassaemic population.

The final PK model developed for children younger than 6 years was used to simulate deferiprone exposure in the population of interest. A cohort of 100 subjects (50 males and 50 females) with a body weight distribution of mean 16 and standard deviation 2.0 kg was used to describe the demographic characteristics in this group of patients.

Results

Population PK modelling

Data from 18 evaluable children (9 male and 9 female) were used for the PK analysis. Patients were randomised to three dose levels (8.3, 16.7 and 33.3 mg kg⁻¹) with six patients assigned to each group. Sixteen patients were diagnosed with β‐thalassaemia major and two with thalassodrepanocytosis. Median (range) body weight, height and age of the children were respectively 15.8 (11–22.5) kg, 99.2 (83–117) cm and 3.4 (1.2–5.9) years. An overview of the baseline demographic characteristics is presented in Table 1.

Table 1.

Baseline and demographic characteristics of the pharmacokinetic analysis population

n = 18	Mean	SD
Weight (kg)	16.08	3.18
Height (cm)	98.95	9.16
Age (years)	3.62	1.33

The PK of deferiprone after oral administration to paediatric patients was described by a one‐compartment open model with first‐order absorption and elimination processes. The absorption rate constant (K_a) represents a first order process. The disposition parameters included (apparent) clearance (CL/F) and (apparent) volume of distribution (V/F).

Between subject variability (BSV) was tested on each parameter, and was identified on CL/F and V/F. In addition, it should be noted that an omega block was used during the estimation of BSV for CL/F and V/F, accounting for the expected correlation between these two parameters. The inclusion of the omega block significantly decreased the – 2Log likelihood.

Different error models were tested to characterise residual variability; e.g., additive, proportional, exponential, combined etc. The proportional error model provided the best results and was kept to describe the residual variability.

Prior parameter distributions were used during the estimation procedures in NONMEM to facilitate parameter search for the absorption rate constant (K_a) and the BSV for CL/F and V/F. The use of priors allowed a better description of the data, reducing significantly the uncertainty around the parameters. The prior information was derived from a population PK analysis performed in healthy adults receiving deferiprone as a 100 mg ml⁻¹ solution 23. The following values were used for the different parameters: 8.2 h⁻¹ for K_a with an uncertainty of 4.02; 0.057 (23.8%) interindividual variation on CL/F and 0.0278 (16.6%) interindividual variation on V/F with an omega block of 0.0345. There were 54 degrees of freedom for the prior on the BSV parameters given that 55 individuals were used for the final population PK model in healthy adults.

During covariate model selection, the effect of weight, height, gender, and age was tested on the different parameters. This was performed after visual inspection of the correlations between covariates and model parameters. The inclusion of body weight on CL/F and V/F according to fixed allometric scaling 25 led to the highest improvement in the model fitting and allowed a better description of the data, increasing the model performance. The exponent was fixed to 0.75 and 1 for CL/F and V/F, respectively. The final parameter estimates are summarised in Table 2.

Table 2.

Final population pharmacokinetic (PK) parameters of deferiprone in children aged <6 years and bootstrap results

Model predicted primary PK parameters
	Estimate	SE	Bootstrap a (mean)	CV (%)
CL/F (l h –1 )	8.3*	0.569	8.30*	8.07
V/F (l)	18.7*	1.16	18.74*	7.95
K a (h –1 )	9.13	1.41	8.91	10.54
WT on V/F (fix allom.)	1 FIX	/	1 FIX	/
WT on CL/F (fix allom.)	0.75 FIX	/	0.75 FIX	/
Error (prop)	0.0953	0.0182	0.0916	39.3
IIV CL/F b	0.0644	0.0115	0.0642	11.37
IIV V/F b	0.0392	0.0077	0.0393	13.23
Block CL‐V	0.031	0.0058	0.0313	12.14

Model predicted secondary PK parameters stratified per dose level
	Median (5 th and 95 th quantiles)
	8.3 mg kg −1	16.7 mg kg −1	33.3 mg kg −1
AUC 0–8 (μmol l −1 h)	116.7 (90.6–129.0)	210.0 (173.1–266.6)	428.8 (291.4–547.8)
C max (μmol l −1 )	61.7 (45.1–80.7)	119.8 (106.0–154.0)	229.5 (179.7–278.1)
T max (h)	0.33 (0.19–0.92)	0.33 (0.21–0.63)	0.37 (0.27–0.42)
C ss (μmol l −1 )	2.1 (1.6–2.3)	3.7 (3.1–4.9)	7.7 (5.1–10.0)
C min (μmol l −1 )	1.5 (0.92–2.6)	1.9 (0.79–5.5)	6.8 (3.1–13.9)

^a0 minimisation terminated out of 500

^bEta shrinkage was −11% and 0% for CL/F and V/F respectively

^* Values refer to parameter value for median body weight (15.8 kg), AUC = area under the concentration‐time curve, C_max = peak concentration, C_min = trough concentration, CV = coefficient of variation, CL/F = apparent clearance, C_ss = steady state concentration, fix allom. = fixed allometric exponent, IIV = inter individual variability, K_a = absorption rate constant, prop = proportional, SE = standard error, T_max = time at which peak concentration is reached, V/F = apparent volume of distribution, WT = body weight

A bootstrap analysis was performed to assess model stability. The mean parameter estimates from the bootstrap analysis were found to be in close agreement with the final model estimates, and the CV values were found to be all below 15%, indicating that the final estimates have sufficient precision. The results of the bootstrap analysis can also be found in Table 2.

Internal model validation diagnostics were satisfactory. Individual predicted profiles and goodness‐of‐fit plots revealed that the model provides an adequate and nonbiased description of the data, as shown in Figures 1 and S1. In addition, normalised predictive distribution error summaries (Figure S2) show that the discrepancy between predicted and observed values can be assumed to be normally distributed. The predictive performance of the model in subsequent simulations was deemed critical to achieve the objective of our analysis. To this purpose, visual predictive checks were used to assess whether the variance and covariance structures have been well characterised (Figure 2). Overall these diagnostic techniques confirm that the final model is suitable for the purposes of data simulation.

Figure 1.

Pharmacokinetic model diagnostics. Goodness‐of‐fit plots. Upper panels show the observed data (Obs) vs. individual predictions (IPred) (left) and the observed data vs. population predictions (Pred) (right). Lower panels show the conditional weighted residuals (CWRES) vs. population predictions (left) and the CWRES vs. time (left)

Figure 2.

Deferiprone concentration vs. time profiles. Visual predictive check: observed data are plotted using blue circles; the black solid line represents the median of the simulated data; the red dashed lines represent the 5th and 95th percentiles of the simulated data. The left, mid and right panels show, respectively, dose groups 1 (8.3 mg kg⁻¹), 2 (16.7 mg kg⁻¹) and 3 (33.3 mg kg⁻¹)

PK bridging and dosing recommendations

The results of the simulations are shown in Figures 3 and 4 and Table 3. Similar exposure is achieved in adults and children in terms of AUC and C_ss after administration of the currently recommended dosing regimen, with median (5–95^th quantiles) AUC values respectively of 340.6 (223.2–520) and 318.5 (200.4–499) μmol l⁻¹*h at 75 mg kg⁻¹ day⁻¹ and 453.7 (297.3–693) and 424.2 (266.9–664) at 100 mg kg⁻¹ day⁻¹, given as t.i.d. doses. The simulations also showed a 29% increase in C_max in children when compared to the adult population.

Figure 3.

Predicted deferiprone exposure expressed as AUC_0–8 (upper panel), C_max (mid panel) and C_ss (lower panel) for children aged <6 years receiving 75 mg kg⁻¹ day^–1. The black line represents the median of the reference population (adult thalassaemic population), whereas the orange lines represent first and third quartiles and the red lines represent 5^th and 95^th percentiles of the same reference population. Percent of total indicates the percentage of cases for each beam of 1000 simulations with 100 patients in each simulated trial

Figure 4.

Predicted deferiprone exposure expressed as AUC_0–8 (upper panel), C_max (mid panel) and C_ss (lower panel) for children younger than 6 years receiving 100 mg kg⁻¹ day^–1. The black line represents the median of the reference population (adult thalassaemic population), whereas the orange lines represent first and third quartiles and the red lines represent 5^th and 95^th percentiles of the same reference population. Percent of total indicates the percentage of cases for each beam of 1000 simulations with 100 patients in each simulated trial

Table 3.

Summary statistics of the secondary PK parameters obtained in the simulated PK bridging study. Parameter distributions are shown in the histograms in Figures 3 and 4

	75 mg kg −1 day −1						100 mg kg −1 day −1
	Adults			Children			Adults			Children
	AUC	C max	C ss	AUC	C max	C ss	AUC	C max	C ss	AUC	C max	C ss
Median	318.5	132.2	5.5	340.6	170.7	5.9	424.2	176.0	7.4	453.7	227.4	7.9
1 st quartile	263.9	109.2	4.6	286.6	145.0	5.0	351.5	145.4	6.1	381.8	193.2	6.6
3 rd quartile	383.0	159.0	6.7	404.7	200.5	7.0	510.0	211.9	8.8	539.0	267.1	9.4
5 th quantile	200.4	81.6	3.5	223.2	114.9	3.9	266.9	108.7	4.6	297.3	153.1	5.2
95 th quantile	499.0	205.6	8.7	520.0	253.0	9.0	664.0	273.9	11.5	693.0	337.0	12.0

AUC, area under concentration–time curve (μmol l⁻¹ h); C_max, maximum concentration (μmol l⁻¹); C_ss, steady‐state concentration (μmol l⁻¹)

The performance of an individualised dosing regimen (i.e., based on weight‐banded doses) was also tested on the target population, but the results show that it does not significantly affect drug exposure in children when compared to a single fixed dose level in mg kg⁻¹ (at 75 mg kg⁻¹ day^–1; data not shown).

Our results suggest that the currently approved dosing regimen for the adult population is suitable for children aged <6 years. This regimen yields similar and effective exposure, accounting for the effect of body weight on the disposition of deferiprone.

Discussion

In spite of the changes in legislation for the approval of new medicines for children, the dose rationale for many of the drugs currently approved for paediatric diseases remains unsupported or relies upon weak empirical evidence. Accurate dosing recommendations are critical for the implementation of concepts such as personalised medicines and essential for the advancement of therapeutics in children. In this context, model‐based approaches can be crucial for therapeutic decisions when limited evidence is available. This is certainly the case for rare diseases such as haemoglobinopathies, especially when considering young paediatric patients, for which practical and ethical constraints make paediatric clinical investigation a true challenge 26, 27.

The need for better understanding of the PK, efficacy and safety in the paediatric population led to the establishment of the DEEP consortium (www.deep.cvbf.net). Within this project, a model‐based approach has been used to explore the implications of potential PK differences and ensure adequate dosage in the age <6 years group. Supporting evidence for the dose rationale was deemed critical before progressing with the evaluation of efficacy and safety in this group of patients. More specifically, the lack of experimental data available on the paediatric use of deferiprone, and in particular deferiprone PK in this group of patients, limited our ability to assess whether doses used in adults, adjusted linearly for differences in body weight (i.e., doses in mg kg⁻¹) produce comparable exposure across the two populations.

There should be little doubt about the therapeutic relevance of defining the appropriate starting dose and dosing regimen for chronic interventions, as in the case of iron chelating agents for the management of iron overload in transfusion‐dependent haemoglobinopathies. Modelling and simulation (M&S) techniques have become an invaluable tool for the evaluation of the dose rationale and personalisation of dosing regimens for subgroups of patients and special populations, allowing the characterisation and quantification of the contribution of different sources of variability to an agent's overall PK properties, reducing at the same time the experimental burden on such a vulnerable population 28, 29, 30. In addition, M&S techniques can be used in conjunction with other advanced statistical concepts to optimise protocol design, increasing the quality of the information gathered. Two concrete advantages of the approach include the reduction in the number of patients required and the use of sparse blood sampling. Here we have shown the implementation of these concepts in the design, conduct and analysis of a clinical study. Our results clearly show the importance of establishing the dose rationale before evaluating the efficacy and safety of deferiprone in paediatric patients affected by transfusion‐dependent haemoglobinopathies.

PK modelling

The PK of deferiprone after oral administration to paediatric patients was successfully characterised by a model‐based approach. As shown in the results section, a one‐compartment open model with first‐order absorption and elimination processes described the concentration vs. time profile of deferiprone, allowing precise and accurate characterisation of the main PK parameters of interest (Table 2). Body weight was found to be a significant predictor of changes in the distribution and elimination processes of the drug; the relationship with CL/F and V/F was described by fixed allometric scaling. Furthermore, the use of prior information from the adult population allowed for the characterisation of the absorption profile. This is another example of how M&S techniques can overcome the limited evidence generated in the clinical study. Of note is the fact that the use of a dosing regimen based on mg kg⁻¹ deferiprone did produce comparable systemic levels, despite the nonlinear (allometric) relationship between body weight and clearance. There are a number of possible explanations for our findings. In fact, previous publications have shown linear correlation between dose, body weight and clearance for biologicals and some small molecules. In general, such a linear correlation is likely to occur for drugs with small volumes of distribution, which correspond to the drug distribution in plasma and lymph (e.g., warfarin) or total body water (e.g. theophylline). By contrast, deferiprone shows a relatively large apparent volume of distribution (i.e., approximately 1.12 l kg⁻¹), but these values could well be the consequence of the 3:1 ratio for deferiprone–iron complex formation, rather than due to distribution beyond total body water 31, 32.

Dosing recommendations

Given that paediatric patients are likely to initiate chelation therapy approximately after 1 year from the start of blood transfusions, the use of chelating agents is not clinically justified before age 1.5–2 years. The therapeutic context in which deferiprone should be used in this patient population has therefore been accurately captured by the PK modelling approach presented here. It can be assumed that differences in drug disposition are determined by the effect of size (i.e. body weight). The impact of metabolic maturation at the start of chelation therapy can be considered minor.

The availability of a population PK model in children allows bridging concepts to applied, enabling the assessment of the dosing requirements to achieve drug levels that correspond to the efficacious exposure in adults. Using the PK parameter estimates from the current study and from a model developed previously in adults 23, simulations were performed to demonstrate how exposure to deferiprone in children younger than 6 years compares to drug levels in the adult patient population after administration of the currently recommended dosing regimen.

As shown in Figures 3 and 4, AUC and C_ss distributions are comparable at 75 mg kg⁻¹ day⁻¹ and 100 mg kg⁻¹ day⁻¹ respectively, whereas an increase in peak concentrations (C_max) is predicted in children. This increase is most probably due to differences in the volume of distribution between the two groups, and is expected to have limited clinical implications. Overall exposure (AUC and C_ss) is the determinant of the response, and changes in C_max are not expected to modify the safety profile of the drug. This is confirmed by studies previously published in the literature, in which children exposed to a dosing regimen of 100 mg kg⁻¹ day⁻¹ show a safety profile similar to that reported for deferiprone in adults 33, 34, 35.

In conclusion, based on these findings, a dosing regimen of 25 mg kg⁻¹ t.i.d. (75 mg kg⁻¹ day^–1) is recommended for children younger than 6 years, with the possibility of titration up to 33.3 mg kg⁻¹ t.i.d. (100 mg kg⁻¹ day^–1), if necessary. It is worth mentioning that this dose will be used to conduct an efficacy–safety comparative phase III study and will be adopted in future modifications of the summary of product characteristics.

Competing Interests

All authors have completed the Unified Competing Interest form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare: F.B. had financial support from the DEEP consortium (sponsored by the European Union); no financial relationships with any organisations that might have an interest in the submitted work in the previous 3 years; no other relationships or activities that could appear to have influenced the submitted work.

This contribution is part of the DEferiprone Evaluation in Paediatrics (DEEP) project, supported by the FP7 Framework Research Program “HEALTH‐2010.4.2–1: Off‐patent medicines for children”. We would also like to thank ApoPharma for their effort in the development of the new oral formulation, which is suitable for the young paediatric population.

Supporting information

Figure S1 Individual plots: observed data are plotted using blue circles; the black solid line represents the population prediction (Pred) and the red solid line represents the individual predictions (IPred). Panel A shows patients in dose group 1 (8.3 mg kg⁻¹); panel B shows patients in dose group 2 (16.7 mg kg⁻¹); and panel C shows patients in dose group 3 (33.3 mg kg⁻¹)

Figure S2 Normalised prediction distribution errors: upper panels show the QQ‐plot of the distribution of the normalised predictive distribution errors (NPDEs) for a theoretical N (0, 1) distribution (left) and the histogram of the distribution of the NPDE together with the density of the standard normal distribution (right). Lower panels show the NPDEs vs. time (left) and NPDEs vs. individual predictions (right)

Supporting info item

Bellanti, F. , Del Vecchio, G. C. , Putti, M. C. , Maggio, A. , Filosa, A. , Cosmi, C. , Mangiarini, L. , Spino, M. , Connelly, J. , Ceci, A. , Della Pasqua, O. , and on behalf of the Consortium DEferiprone Evaluation in Paediatrics (DEEP) (2017) Population pharmacokinetics and dosing recommendations for the use of deferiprone in children younger than 6 years. Br J Clin Pharmacol, 83: 593–602. doi: 10.1111/bcp.13134.