Why Evidence in Children's Medicine Is Often Uncertain

Evidence for treating children can be harder to build than evidence for adults, but that problem is easily overstated. A widely repeated claim that about 90% of medical treatments lack high-quality evidence did not come from a study of pediatric medicine. It came from an analysis of Cochrane reviews covering healthcare interventions across specialties.

Researchers examined 154 updated Cochrane reviews and had comparable evidence ratings for 151. Fifteen, or 9.9%, rated the evidence for the first listed primary outcome as high quality under the GRADE framework. Fifty-six were rated moderate, 47 low and 33 very low. Those results describe confidence in a selected body of evidence. They do not show that 90% of treatments fail, cause harm or lack any scientific support.

Children's medicine has additional challenges because infancy, childhood and adolescence are periods of rapid physical development. Evidence from adults may sometimes inform care, but it may not answer questions about age-specific effects, formulations or outcomes. The correct response is to identify exactly where uncertainty lies, not to portray pediatricians as improvising without standards.

The 90% Figure Is Not a Pediatric Statistic

The 2020 meta-epidemiological study used GRADE to assess updated systematic reviews. GRADE can lower certainty because of risk of bias, inconsistent findings, indirect evidence, imprecise estimates or publication bias. A low rating means confidence in an estimated effect is limited and future research may change it. It is not a count of treatments proved useless.

The study also had limits. Its sample consisted of updates to an earlier group of Cochrane reviews, and the authors said it might not represent all medical interventions. They focused on the first listed primary outcome. An intervention could have stronger evidence for another outcome, and some interventions are difficult to blind even when they are useful.

Applying the 9.9% finding specifically to children changes both the population and the meaning of the result. It turns a measure of evidence certainty across multiple fields into a claim about pediatric treatment coverage. The available research does not support that conversion.

The distinction also prevents a second mistake: describing every uncertain treatment decision as an experiment on a child. Clinical care can draw on regulatory labeling, trials, observational studies, established practice, pharmacology, specialist guidance and the individual patient's circumstances. Those forms of evidence do not all carry the same certainty, but they are not equivalent to having no basis for a decision.

Four Problems Can Weaken Confidence

Andrew Booth, a professor of evidence synthesis and the author of the 2026 commentary, described four recurring reasons that reviewers may judge evidence too weak for a firm conclusion. The first is risk of bias. If participants, families or investigators know which treatment was given, expectations may affect reporting or assessment.

The second is inconsistency. Studies addressing a similar question may produce different results. Counting how many point in one direction is not enough; reviewers must consider differences in age, intervention, comparison group, outcome and study quality.

The third is indirectness. Evidence may concern a different population or a related but not identical clinical question. Adult results can be indirect for children when development changes disease course, drug handling or relevant outcomes. At the same time, adult evidence is not automatically unusable. Whether it can be extrapolated is a scientific question that depends on the medicine, condition and age group.

The fourth is imprecision. Pediatric conditions may involve small eligible populations, and dividing participants into infants, younger children and adolescents can make estimates even less certain. A small study may be compatible with both meaningful benefit and little benefit. The honest conclusion is a range of possibilities, not a confident verdict selected from that range.

Pediatric Research Needs Additional Safeguards

Research involving children is necessary precisely because children cannot always be treated as smaller adults. Some diseases occur only in childhood, and the effects of a condition or treatment may change with growth and development. Study designs may need age-appropriate outcome measures, formulations, sample collection and follow-up.

Children also receive additional ethical and regulatory protections because they cannot provide legal consent for themselves. Parents or guardians provide permission, and children may provide assent when they are able. These safeguards make study design more demanding, but presenting protection itself as an obstacle to be removed creates a false choice between better evidence and participant welfare.

US policy combines requirements and incentives. The Pediatric Research Equity Act gives the Food and Drug Administration authority to require pediatric studies for certain drugs and biological products, using formulations appropriate for each age group. The Best Pharmaceuticals for Children Act offers additional marketing exclusivity when sponsors voluntarily complete studies requested by the FDA.

In the European Union, a pediatric investigation plan sets out how a medicine will be studied in children. The European Medicines Agency can grant a deferral, for example until enough adult safety and effectiveness information exists, while retaining pediatric studies and timelines in the plan. It can also grant a waiver when pediatric development is not needed or appropriate, including for conditions that affect only adults. A deferral or waiver is therefore a documented regulatory decision, not proof that a company simply abandoned children.

Uncertainty Must Be Named, Not Dramatized

Evidence gaps are real. Booth pointed to a review of child and adolescent obesity treatments in which published accounts of adolescents' experiences were limited and accounts from children aged ten or younger were absent. That kind of gap affects how well guidelines can reflect what treatment feels like to the people receiving it, even when other evidence about effectiveness exists.

Better pediatric research may require coordinated multicenter trials, age-appropriate outcomes, careful use of observational data and transparent reporting of harms. Modeling and extrapolation can help answer defined questions, but their assumptions must be visible and they do not remove the need for clinical data when important differences remain.

Uncertainty should not be converted into total ignorance or used to assign a motive. The available evidence does not justify a universal 90% pediatric failure rate or a claim that clinicians treat children by guesswork.

A defensible analysis is harder and more useful. Some pediatric decisions rest on strong evidence, others on evidence that is indirect or incomplete, and the degree of uncertainty differs by treatment, outcome and age. Regulators, researchers and clinicians should make those boundaries auditable. Families need to know what is established, what is inferred and what remains unresolved. Alarmist numbers conceal those distinctions; precise evidence grading exposes them and shows where the next study can matter.