Can real-world evidence fill the paediatric data gap?
Health records, registries and claims data contain years of experience with medicines in children. Turning that into regulatory-grade evidence is harder than it looks.
In brief
Can health records and registries fill the paediatric evidence gap? How real-world evidence supports dosing, safety and extrapolation, and its limits.
Children take medicines every day that have never been formally studied in them. Every one of those prescriptions leaves a trace in a health record. For years, researchers have asked whether that data could answer the questions trials have not.
The regulatory opening
The FDA published a framework for real-world evidence in 2018, following the 21st Century Cures Act, setting out how data from outside traditional trials could support regulatory decisions. In Europe, the EMA and national regulators established DARWIN EU, a network giving regulators access to real-world data from across the region.
Paediatrics is an obvious use case. Real-world data can describe how a drug is actually used in children, at what doses and with what outcomes, and can identify safety signals in groups too small for trials to detect.
What it is good for
- Dosing. Records of drug levels taken in routine care can feed pharmacokinetic models.
- Natural history. Registries describe how rare paediatric diseases progress untreated, forming the basis for external control arms.
- Safety. Large databases can detect rare adverse events across thousands of children.
- Utilisation. Data reveal where off-label use is widespread and paediatric studies are most needed.
Where it falls short
Real-world data were not collected to answer research questions. Doses and weights may be missing, outcomes inconsistently recorded, and children who receive a drug may differ systematically from those who do not. Those biases can be reduced with careful design but rarely eliminated.
That is why real-world evidence has so far been most persuasive as a complement to trials, particularly for safety and for supporting extrapolation, rather than as a replacement for them.
