April 20, 2026

Pediatric pharmacovigilance: why children are not small adults, statistically or clinically

There is a phrase that comes up often in discussions about pediatric drug safety: children are not small adults. It has become a cliché, but it keeps being repeated because the field keeps rediscovering how true it is.

The problem is not just that children weigh less. It is that nearly every aspect of how a drug moves through the body, how it acts on its target, and how adverse reactions manifest is shaped by developmental biology. And the systems we have built for detecting drug safety problems — spontaneous reporting, disproportionality analysis, reference sets, regulatory workflows — were largely designed with adult populations in mind.¹²

That mismatch creates real gaps. Some of those gaps are clinical. Others are statistical. Both matter.

The pharmacokinetic problem

The most obvious difference between children and adults is developmental pharmacokinetics. In neonates and infants, drug absorption, distribution, metabolism, and excretion are all changing rapidly as organ systems mature. Gastric pH, body composition, protein binding, hepatic enzyme activity, renal clearance — none of these are constant during the first years of life, and many do not reach adult levels until late childhood or adolescence.³⁴

This means that a dose calculated by simple weight-based scaling from an adult dose may produce a completely different drug exposure in a child. A neonate may have a fraction of the enzyme activity needed to metabolise a drug, leading to accumulation. An older child may actually clear some drugs faster than an adult, requiring a higher weight-adjusted dose. The variability across pediatric age subgroups — neonates, infants, toddlers, school-age children, adolescents — is enormous.³

For pharmacovigilance, this is important because it means that the adverse reaction profile of a drug in children may be genuinely different from the profile in adults. Not just scaled-down versions of the same reactions, but different reactions entirely. Doxorubicin-induced cardiotoxicity, for example, follows a different pattern in children than in adults. Methylphenidate can produce psychiatric effects in paediatric populations that do not have clear adult equivalents. These are not edge cases. They are examples of how developmental biology can change the safety story of a drug.⁵

Off-label use and the evidence gap

A second problem is the sheer extent of off-label drug use in children. Estimates vary, but published reviews suggest that a large proportion of drugs administered to paediatric patients are used in ways not supported by the approved labelling. In some hospital settings, off-label rates exceed 50%. In neonatal intensive care, the numbers are even higher.⁶⁴

Off-label use is not automatically unsafe, but it does mean that many paediatric drug exposures occur without the safety data that would normally come from controlled clinical trials. Children have historically been excluded from most drug development programs, a situation that led to the term “therapeutic orphans.” Legislation in both the United States and the European Union has tried to address this, but the legacy of sparse paediatric safety data persists.⁶

For pharmacovigilance, off-label use creates a double challenge. First, the baseline expectation about a drug’s safety profile is derived from adult data, which may not apply. Second, when adverse events do occur in children, there is often no clinical trial denominator against which to interpret them. The only data available may be whatever accumulates in spontaneous reporting databases — which brings us to the next problem.

Underreporting is especially severe in children

Underreporting is a fundamental limitation of all spontaneous reporting systems, but the evidence suggests it is particularly pronounced in paediatric populations. One long-term hospital-based study found that paediatric reports represented only about 6% of all spontaneously reported adverse drug reactions over an 11-year period, despite children accounting for a significant share of patient visits.⁷

Several factors contribute to this. Clinicians may be less likely to attribute symptoms to a drug in a child, especially when the drug is being used off-label and the adverse reaction profile is poorly characterised. Parents and caregivers may not know how to report, or may not realise that a symptom could be drug-related. And in very young children, the patients themselves obviously cannot describe what they are experiencing.¹

This means that the safety signal landscape in paediatric pharmacovigilance is thinner and noisier than in adult surveillance. A signal that might emerge clearly from thousands of adult reports may remain invisible in a paediatric database with far fewer cases.

Disproportionality analysis was not designed for this

Most of the standard disproportionality methods — PRR, ROR, IC, EBGM — were developed and validated using adult-dominated databases. The reference sets used to benchmark their performance, such as the OMOP and EU-ADR gold standards, consist largely of drugs and events relevant to adult medicine. Neither was designed to capture the kinds of drug–event associations that are specific to children.⁸

This is not a trivial gap. A study by de Groot and colleagues specifically addressed this by creating a paediatric-specific reference set for signal detection performance testing. They argued that existing reference sets contained many drugs rarely prescribed in children and events that seldom occur in paediatric populations, making them unsuitable for evaluating signal detection methods in this age group.⁸

The implication is that when we report performance metrics for signal detection algorithms, those metrics may not reflect how well the same algorithms would perform on paediatric data. A method with good sensitivity for adult drug–event pairs may miss paediatric-specific signals, or it may generate different false positive profiles when applied to a database where reporting patterns, drug exposure distributions, and event frequencies are all different.

The age subgroup problem

Even within paediatrics, the population is not homogeneous. A neonate is not an infant, an infant is not a toddler, and an adolescent is not a young child. Each subgroup has different pharmacokinetic properties, different prescribing patterns, different disease prevalences, and different adverse reaction susceptibilities.⁴

This creates a stratification challenge for pharmacovigilance. Aggregating all paediatric reports into a single group can mask signals that are specific to one age subgroup. But stratifying into narrow age bands reduces the already small case counts even further, making statistical detection harder.

Some researchers have proposed age-stratified disproportionality analyses within paediatric databases, but the practical feasibility depends on having enough data, and many paediatric drug–event combinations simply do not have enough reports to support stable estimates in any single age band.

What would better paediatric pharmacovigilance look like?

There is no single fix, but several directions are worth highlighting.

First, paediatric-specific reference sets for benchmarking signal detection need to become standard. The work by de Groot and colleagues is a starting point, but it needs to be expanded and maintained as an ongoing resource.⁸

Second, disproportionality analyses applied to paediatric data should routinely report results by age subgroup where feasible, and should flag when counts are too low for reliable estimation rather than defaulting to pooled paediatric analyses.

Third, the integration of electronic health record data alongside spontaneous reports could help address both the underreporting problem and the denominator problem. If we know how many children were exposed to a drug and can link that to outcome data, we move from passive signal detection toward active surveillance.

And fourth, the pharmacovigilance community needs to be honest about how much of the paediatric safety landscape remains poorly mapped. For many drugs, especially older generics used off-label in children, the safety evidence base is thin, the reporting is sparse, and the methods have not been validated in this population.

That honesty is not a weakness. It is a prerequisite for making things better.

1. Li Y, Wu Y, Jiang T, et al. Opportunities and challenges of pharmacovigilance in special populations: a narrative review of the literature. Therapeutic Advances in Drug Safety. 2023;14:20420986231200746. ↩ ↩²

2. Clavenna A, Bonati M. Adverse drug reactions in childhood: a review of prospective studies and safety alerts. Archives of Disease in Childhood. 2009;94(9):724–728. ↩

3. Lu H, Rosenbaum S. Developmental pharmacokinetics in pediatric populations. Journal of Pediatric Pharmacology and Therapeutics. 2014;19(4):262–276. ↩ ↩²

4. Kearns GL, Abdel-Rahman SM, Alander SW, Blowey DL, Leeder JS, Kauffman RE. Developmental pharmacology — drug disposition, action, and therapy in infants and children. New England Journal of Medicine. 2003;349(12):1157–1167. ↩ ↩² ↩³

5. Fernandez E, Perez R, Hernandez A, Tejada P, Arteta M, Ramos JT. Factors and mechanisms for pharmacokinetic differences between pediatric population and adults. Pharmaceutics. 2011;3(1):53–72. ↩

6. Czarniak P, Bint L, Favié L, et al. Off-label medication use in children, more common than we think: a systematic review of the literature. Journal of Pediatric Pharmacology and Therapeutics. 2019;24(4):292–303. ↩ ↩²

7. Aagaard L, Christensen A, Holme Hansen E. Information about adverse drug reactions reported in children: a qualitative review of empirical studies. British Journal of Clinical Pharmacology. 2010;70(4):481–491. ↩

8. de Groot MCH, van Puijenbroek EP, van Eijk ME, et al. Pediatric drug safety signal detection: a new drug–event reference set for performance testing of data-mining methods and systems. Drug Safety. 2015;38(2):207–217. ↩ ↩² ↩³

Back to blog