Researchers have developed a data-driven model of how organs develop in order to more quickly determine safe drug doses for children.
Paediatric drug development programs are essential in the process of making any drug available for children. However, determining safe yet effective drug dosages for children is a continuous challenge for pharmaceutical companies.
There are numerous factors that limit children’s participation in drug trials. Patient recruitment is a substantial challenge due to ethical and legal constraints. Also, some diseases are rarer in children than adults, making datasets very sparse. Therefore, during drug development, an adult population is often studied first. This is then followed by the knowledge generated being subsequently transferred to the design of a paediatric trial.
‘Children are merely small adults’
The typical underlying assumption is that children are the same as adults, but just smaller. However, this view may overlook differences that arise from the fact that children’s organs are still developing. In truth, functional maturation of the body continues from birth until adolescence. Also, rates of functional development vary from system to system, organ to organ, and tissue to tissue during this time.
Adjusted drug doses for children are often derived from adult pharmacokinetic (PK) models, whereby the data is simply extrapolated to fit patients with a smaller body size, without accounting for differences in organ sizes. Thus, organ maturation in children can be overlooked when selecting adequate doses for paediatric trials. Allometric scaling may not take certain physiological effects of the drug into account when designing a trial for children. Therefore, additional organ maturation may need to be accounted for.
Organ modelling for drug development
Researchers at Aalto University in Finland and Novartis, the pharmaceutical company, have created a model that improves the understanding of how organs develop, in the hope to make drugs and their development safer for children.
The team proposed a model based on a non-parametric regression method called Gaussian Process (GP). In order to explore whether such maturation effects are being detected, the GP model data-driven approach was used to estimate any potential deviations from an adult PK model adjusted for children.
The approach used four simulation studies based on realistically sparse datasets. Extreme data sparsity can present a problem to non-parametric approaches like GPs. To overcome this, virtual data was used to add prior assumptions like monotonicity to the model, making sparsity less of an issue. The results showed that, if present, the approach was capable of correctly detecting any unexpected maturation function.
The future of drug modelling
The GP approach showed that it could reliably detect whether dosing adjustments are necessary. Compared to model checking approaches, the data-driven GP approach is less subjective and avoids bias. The regression was also better at handling small sample sizes, as uncertainties are accounted for. Therefore, the GP approach could lead to better decisions in paediatric drug development.
The possible benefits of the approach are far reaching – it could be particularly useful in an array of situations. For example, when a drug needs testing on a completely new and small sample size. Also, the method could be more effective than existing practises in drug repurposing.
Aki Vehtari is an associate professor at Aalto University and the Finnish Centre for Artificial Intelligence. He said: “It works for any drug whose concentration we want to examine. This is a method that could help determine safe drug doses more quickly and with less observations than before”.
Image credit: FreePik rawpixel.com
