The European Medicines Agency (EMA) defines extrapolation as:
‘extending information and conclusions available from studies in one or more subgroups of the patient population (source population(s)), or in related conditions or with related medicinal products, in order to make inferences for another subgroup of the population (target popula- tion), or condition or product, thus reducing the amount of, or general need for, additional information (types of studies, design modifications, number of patients required) needed to reach conclusions’.34, 35
Examples of extrapolation in practice include extrapolating from historical data to predict drug effects in contemporary patients, extrapolating from one geographic region to another to predict clinical benefits, or extrapolating from adults to sup- port licensing decisions in the paediatric population. In order for extrapolations to be appropriate, strong assumptions on similarity are required between the source and target populations; a big challenge is deciding whether and to what extent extrapolation is appropriate. The US FDA1 and International Council for Har- monisation of Technical Requirements for Pharmaceuticals for Human Use (ICH) E1123 guidelines outline an algorithmic, assumptions based approach to decid-
ing on the level of extrapolation appropriate in paediatric medicine development and for determining which data are needed to support paediatric licensing of a medicine. For a particular treatment and indication, the choice of the level of ex- trapolation depends on whether it is reasonable to make the following assumptions between adults and children: similar disease progression; similar response to the intervention; similar PK-PD relationships; and whether there is a PD response that can be used to predict efficacy. Depending on which of these assumptions
are deemed reasonable, extrapolations can range from no extrapolation, through partial extrapolation, to complete extrapolation. According to Dunne et al.,36 for each level of extrapolation the evidence required in the paediatric population is as follows: for no extrapolation a full development programme is required; for partial extrapolation the evidence needed ranges from a single adequate, well-controlled trial to confirm efficacy to a PK-PD study to confirm response; for complete ex- trapolation only PK studies, to establish dosing, and safety studies are needed (in certain cases only safety data may be needed). Hampson et al.33 describe this type of extrapolation, where uncertainty regarding extrapolation assumptions is not accommodated, as ‘deterministic extrapolation’.
An alternative framework for the use of extrapolation in the development of medicines for paediatrics has more recently been proposed by the EMA.34, 35 This framework supports the use of quantitative methodology to help understand (in terms of disease, drug pharmacology and clinical response) how relevant existing source population (e.g. adults) information is to the target paediatric popula- tion and identify any important assumptions and uncertainties about the relation between dose, PK, PD and clinical efficacy which should be documented as an ‘ex- trapolation concept’. Conditional on these assumptions, the question of whether clinical efficacy can be predicted in the paediatric population, from the source population, can be assessed and a specific extrapolation plan can be developed to address any identified gaps in knowledge (and also identify where large confi- dence exists, to avoid unnecessary duplication of information). This extrapolation plan would detail any trials and study objectives needed to fill in knowledge gaps and provide evidence which could validate the extrapolation concept. After any planned studies, if the extrapolation concept is deemed to be valid, relevant evi- dence from the source population and evidence generated in the paediatric popu- lation could contribute to regulatory decision making for marketing authorisation.
tional post-authorisation data may need to be gathered.35 Such an approach to extrapolation, where knowledge regarding extrapolation assumptions can be up- dated and verified, is referred to by Hampson et al.33 as ‘stochastic extrapolation’. A benefit of extrapolation in the setting of paediatric trials is the potential to reduce the number and size of studies required to demonstrate efficacy of a new medicine in children. This is important as it may be difficult to recruit children into clinical trials, there may not be many children meeting inclusion criteria for recruitment (e.g. having already been exposed to the drug off-label) and there may be constraints on the number and type of clinical measurements which can be taken from children. Additionally, there is a tradeoff between risk and benefit for testing medicines in children: if a highly effective treatment already exists there may be a reluctance to expose children to a potentially harmful new treatment; however, if there is currently an unmet medical need in children, it would be con- sidered ethical to trial a drug without proven efficacy, even with the potential for adverse events. Conversely, the consequences of extrapolating when the assump- tions do not hold could include exposing children to an ineffective medicine or exposing children to an unacceptably toxic dose.
Examples of extrapolation in practice can be seen in the cases of Zmax (azithromycin extended-release) for treating Community-Acquired Pneumonia, where use in the paediatric population (6 months and older) is based on extrapolation of adult efficacy data with additional safety and PK data in paediatric patients;37 and
oxcarbazepine (an anti epileptic drug), where the efficacy data from trials of ad- junctive therapy (the test treatment in conjunction with another treatment) were used to inform the approval of oxcarbazepine as monotherapy.38 In patients with focal epilepsies, adjunctive therapy data from adults and paediatrics were used to support the extrapolation of efficacy data from adults on oxcarbazepine monother- apy to paediatric patients in order to gain FDA approval. However, it should be
noted that this extrapolation was made with the knowledge that oxcarbazepine was already approved as monotherapy for paediatric patients in the EU.