Paediatric PK studies: what are they?

PK studies absorption, distribution, metabolism and excretion of drugs. PK studies guide the determination of a drug’s optimal route of administration and formulation, and appropriate dose and dose schedule(s). PK studies include (1) early-phase studies including PK components, comprising (a) first-dose-in-child studies and (b) phase II and III studies; and (2) phase IV postmarketing PK studies (including, eg, dose optimisation or therapeutic drug monitoring (TDM) studies).

Paediatric research is challenging because children are considered vulnerable and sample sizes are small. Nevertheless, accurate dosing is paramount to safe and effective treatment, which makes paediatric PK studies essential. The USA and European Union (EU) have adopted paediatric clinical trial regulations to address historic deficiencies in this research.4 These reflect the recent paradigm shift in attitudes recognising that ‘the time has come to protect children and young people through research not from research’.5

Knowledge gaps in developmental physiology and pharmacology

Physiological development, including physical growth and the maturation of organs, transporters and enzymes, creates size and age-dependent variability in PK parameters. PK measures should therefore be related to measures of growth, including age, weight or body surface area (BSA).6 The heterogeneity in paediatric patient populations contributes to large PK variability: neonates, for example, can have a tenfold weight range (eg, 400 g–4 kg) in addition to differing gestational ages and postmenstrual ages. Critical illness, renal replacement therapy and cardiopulmonary bypass can also influence PK, as well as drug–drug interactions.7 Furthermore, genetic differences (polymorphisms) contribute to variability in drug disposition and response. A deeper understanding of the interplay between genetic polymorphisms, growth and development is urgently needed to be able to provide personalised dosing to children of different ages, underlying disease states and treatments.8

Challenges in paediatric PK

Traditional PK study designs involve a single dose with multiple blood samples at fixed intervals, covering a time period of up to approximately 10 half-lives. Multiple samples per patient, with varying routes of administration, enable calculation of individual drug exposure (area under the plasma drug concentration–time curve: AUC), and the rate and extent of absorption. Such studies generally require the same number of samples from all subjects taken at the same time in the dosing interval. These designs present ethical and practical barriers for children. Similar challenges arise later in the drug life-cycle when needing to undertake verification of bioequivalence or cross-over relative bioavailability studies in children in order to exclude age-by-formulation interactions.

First, the number of samples and the blood volume per sample that can safely be taken in children—especially neonates—are limited.9 Second, it is impractical to collect many samples over a long period, without interfering with patient care and length of stay. Phlebotomy-related barriers partly concern the acceptability of additional sampling in children, particularly when venepuncture is not needed for clinical care. Similar challenges arise later in the drug life-cycle when needing to undertake verification of bioequivalence or cross-over relative bioavailability studies in children in order to exclude age-by-formulation interactions.

In neonatal studies, differences in postmenstrual age, body composition and end-organ perfusion create extensive PK variability, reflecting rapid phases of growth and maturation. Usually, prior to neonatal PK studies, evidence should be established in less vulnerable populations.3 However, particularly in conditions unique to newborns, neonatal PK data should be obtained a priori.3 Current regulations stipulate that the volume of blood samples for research should not exceed 1% at any one time (0.8 mL/kg) or 3% within 1 month, that is, 1.2 mL for a neonate weighing 500 g, which is problematic for many assays.3 10

Sampling techniques for blood and other biological matrices

Advances in analytical technology have dramatically reduced the blood volume required, yet novel biological sampling methods have been relatively underused. Microsampling techniques (eg, dried blood spots) are now possible with ultrasensitive analytical methods (eg, liquid-chromatography with tandem mass-spectrometry), potentially reducing sample volumes to 5–10 μL, sometimes also reducing invasiveness and improving analyte stability.10

A thorough understanding of different paediatric blood sampling techniques, summarised in table 1, is paramount for PK study design. Research methods which reduce the burden of blood sampling, including so-called ‘scavenged’ sampling11 (using left-over blood from clinical samples) or opportunistic sampling (sampling at the same time as clinical blood tests), can be used but also require careful validation, to ensure they provide equivalent estimation of PK parameters when compared to preplanned sampling time points.12

The collection of alternative matrices to blood requires special consideration in children. Whenever possible, PK/PD information should be obtained via less invasive matrices such as urine and saliva. If an indwelling bladder catheter is present and a reliable quantitative assay available, urine samples can also be obtained to minimise necessary blood sampling.3 Limitations of these approaches include analytical complexity, potential impact of collection methods and environmental factors (eg, temperature and pH) on analyte recovery, and practical challenges. Urinary sampling methods are summarised in table 2. The urine output must be collected over a prolonged (prespecified) period to measure cumulative drug (or metabolite) excretion per unit time, but collection periods are often logistically challenging in practice, deviating from clinical routines.

Saliva has shown feasibility for caffeine TDM in preterm infants13 but requires characterisation of the relationship between saliva and plasma concentrations. Several factors can drive variability in salivary measurements, including salivary pH and drug physicochemical properties (molecular weight, pKa). Importantly, salivary sampling in neonates can also be time-consuming, for example requiring prolonged oral cavity suctioning (up to 30 min), thereby potentially posing additional ethical concerns. Often samples from tissues of interest are more informative in understanding PK/PD, so suitable opportunities to use left-over clinical samples (eg, cerebrospinal fluid) should be taken. However, the cost of developing and validating bioanalytical assays in different sampling matrices must be incorporated into study feasibility assessments. All assays should be developed in accordance with relevant guidelines to meet regulatory requirements.14

Ethical issues in paediatric PK studies

The ethical principles of paediatric research are well described elsewhere.15 PK research presents specific challenges surrounding the informed consent process and the risk–benefit of study participation. Potential research participants (and/or their parents) need sufficient information and time to make an informed decision. PK samples, however, may be required soon after drug administration. If this follows an emergency procedure, parents may be absent or under too much stress to consent. In these circumstances, deferred consent may be appropriate subject to ethics committee approval. For newborns, prenatal consent can be suitable.

To be ethically acceptable, participation of children should be limited to research of minimal risk and burden, or with the potential for direct benefit.16 Risk assessments can be difficult in paediatric studies with minimal adult safety data, for example if drug toxicity prevents healthy adult studies. Furthermore, dose-finding PK studies may not provide direct participant benefit, posing the dilemma of balancing individual research-associated burdens with intended long-term benefits to future children.

PK modelling as a solution

To circumvent some of the challenges described above, a model-based approach to PK study design and analysis can be employed. In population PK (popPK) modelling, data from multiple subjects are analysed simultaneously, thus allowing patients to contribute data with varying sample numbers and timing. Modelling is used to estimate typical values of PK parameters for the population. Capturing information from a larger number of patients enables improved understanding of PK variability and thus improved generalisability of study results.17 However, PK parameters for individuals can also be estimated within the same process, making this method highly efficient. PopPK modelling using non-linear mixed effects (NLME) is now the standard method of choice. PK data (drug concentration–time profiles) are generally non-linear; hence, PK model fitting requires non-linear regression. The mixed-effects approach gives less biased and more precise estimates of variability than other methods.18 NLME is implemented in many general and specialist statistical tools and enables quantification of interindividual variability, intraindividual variability and residual unexplained variability. During PK model building, significant factors determining PK variability are identified (covariates, eg, age, weight, renal function). Within a mixed-effects popPK model, there are ‘fixed effects’—including the typical PK parameters, described by the PK structural model, and significant covariates—and ‘random effects’, which quantify variability at the parameter and observation levels.17 PK modelling confers many benefits (see box 1) and has many applications, including (1) model-based dose optimisation, (2) extrapolation for paediatric dose regimen selection, (3) clinical trial simulations, (4) model-based TDM algorithms and (5) enhancing understanding of developmental pharmacology.17 19

Scaling and extrapolation

Scaling refers to normalising data either between species or between humans of different ages and sizes. PK models usually aim to estimate drug exposure (AUC), which under steady-state conditions depends only on dose and clearance (CL). In 1950, Crawford, a paediatrician, recognised CL scales approximately with BSA, and described the apparent parallel between CL scaling and that of basal metabolic rate with the so-called allometric model.20 Allometric scaling of CL a priori is now common in PK modelling and is an appropriate way to scale for size in children over 2 years of age; box 2 gives the standard allometric model.21

Delineation of the correlated effects of weight, age and organ function is challenging, generating debate regarding the appropriate value of the allometric exponent (b). Knowledge from interspecies and biological process scaling, and anthropomorphic measurements (eg, organ sizes22), has identified 0.75 as a reasonable value for b when scaling CL, and 1 when scaling the volume of distribution.21 Weight raised to the power of 0.75 describes clearance well in children over 2 years old but may not address preceding age-related maturational effects.23 This can be overcome using a standard parameterisation, incorporating allometric weight scaling and a sigmoid maturation function.6 Germovsek et al 23 compared this model with distinct clearance covariate models systematically identified in the literature to fit the same data set. No published model gave a superior fit to the standard model, providing a powerful argument for researchers to adopt a standard parameterisation, facilitating comparison of studies of the same drug and model-based meta-analysis.

Renal function (as related to clearance) is a common covariate in PK studies. The Food and Drug Administration (FDA) draft guidance24 recommends use of the Schwartz formula in children under 12 years and the Cockroft-Gault equation in adolescents. This recommendation leads to two fundamental problems: first, the methods use demographics (eg, height, weight, age) to estimate glomerular filtration rate (GFR), which may already have entered the model (eg, via allometric scaling with a maturation function); and second, the methods have different units: mL/min/1.73 m² for Schwartz and mL/min for Cockroft-Gault, respectively. One sensible solution would be to avoid GFR, and instead use serum creatinine alongside allometric scaling and a maturation function. Since serum creatinine levels change with age (high at birth, falling rapidly, then rising again), this must be scaled by the age-expected median serum creatinine; an elegant method is available.25 However, the contribution of muscle mass and maternal creatinine transfer in preterm neonates must be considered.26

Extrapolation is defined as ‘Extending information and conclusions available from studies in one or more subgroups of the patient population (source population) to make inferences for another subgroup of the population (target population)’.27 For drugs with an established linear PK profile in adults and where sufficient knowledge is available regarding disease aetiology in children, often a single-dose study generates sufficient evidence for a paediatric PK assessment.24 The FDA provides an algorithm (‘decision tree’) to guide appropriate extrapolation for paediatric medicines.28 Evidence of similar disease progression, concentration–response relationships and comparable endpoints allows the extrapolation of PK and safety studies to support paediatric dosing rationale.28 29 New adverse effects may subsequently be identified in children relating to developmental PD, mandating ongoing safety monitoring throughout each agent’s life-cycle.

Dose selection for first-in-child studies: innovative methods

Often there is limited safety information to support dose selection in an initial paediatric PK study. The initial dose requires attention to (1) relative bioavailability, (2) age group and weight of study participants, (3) therapeutic index and (4) PK data from other populations and preclinical studies.24 Advantages of various dose-finding study designs (parallel, staggered, intrapatient dose titration) are summarised elsewhere.29 Current guidelines stipulate that the initial dose should be given as a fraction of the dose calculated from adult exposure based on the aforementioned factors and any additional paediatric expertise.24 However, the role of modelling and simulation in dose selection is increasing, to help avoid children being exposed to subtherapeutic doses using current approaches. Extrapolation can involve ‘top-down’ scaling of adult PK (described above), or ‘bottom-up’ physiology-based PK modelling, combining extensive information on the drug and the system (‘virtual child’)30 to estimate PK parameters. Ideally one should compare results from both approaches, with thorough investigation using sensitivity analysis to explain any differences.

To obviate potential toxicity and uncertainty surrounding first-in-child doses, some advocate microdosing (giving a very low dose), the feasibility of which has been demonstrated in infants.31 There are potential caveats, including limitations with assay performance and ethical concerns regarding subtherapeutic dosing. An important prerequisite is dose linearity across the range from microdose to therapeutic dose: adult studies may be required first to establish this, or a thorough theoretical evaluation. Following the first dose, Bayesian (adaptive) study designs allow dose titration to quickly reach the desired therapeutic window, within either individual or sequential patients.

Realising an optimal design for PK studies

Optimal design refers to identification of the most informative study design, in order to optimise various aspects of the PK protocol, as summarised in box 3; results should be reviewed with consideration of practical and financial constraints.32

Many available software tools can help identify optimal study design.33 For paediatric PK studies, optimal design-based approaches could be viewed as an ethical requirement to minimise the number of children and the number of samples required.32 One disadvantage is that optimal design relies heavily on the proposed model being useful for the target population. Since many PK studies are exploratory, it may not be possible to define an adequate model a priori. Therefore, study design should include techniques such as simulation re-estimation studies (to test whether a proposed design can recapture known parameters) and ensure samples covering the whole dosing interval are included at the population level, if not in every patient. Solely opportunistic sampling can give poor model parameter estimates, and when feasible obtaining samples at rational time points should be coordinated with routine blood tests.34 When samples cannot be optimally timed, the so-called ‘scavenged’ sampling methods may be adopted, provided the analytes being studied are suitable.35

Recommendations for research in practice

We recommend incorporation of the steps summarised in table 3 when designing paediatric and neonatal PK studies to maximise study quality and the feasibility of developing new evidence-based dosing recommendations. Incomplete reporting of PK studies hinders the utility and comparability of study results, as highlighted by the 2015 consensus-based ClinPK statement.36 The ClinPK minimum reporting criteria will most likely benefit from adaptation to address the unique challenges of paediatric/neonatal PK study design, warranting future research.

Changing clinical practice: implementing evidence-based formularies

Even when new PK studies are completed in children, these new data may be ignored by formulary committees and not used to inform future dosing recommendations. There are currently few transparent evidence-based paediatric and neonatal formularies, where the end users (ie, physicians or pharmacists) can easily access the evidence on which the dosing recommendations are based. The Dutch have reported their experience of developing and publishing an evidence-based formulary, with clear descriptions of the timelines and funding required.37 This example demonstrates the feasibility of such endeavours and paves the way for other national and international formularies to follow suit. It is our belief that incorporating the latest PK evidence into formularies should become routine to inform dosing at the bedside. Our vision for the life-cycle of paediatric therapeutics is summarised in figure 1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Paediatric product life-cycle from development to implementation. PD, pharmacodynamic; PK, pharmacokinetic.

Current PK research needs