How Do You Synthesize Evidence Across Age Groups?

Can evidence from children, younger adults, and older adults really be combined?

Your review question concerns a broad population, but the available studies do not divide neatly by age. One study includes adolescents. Another includes adults aged 18 to 65. A third focuses on people over 60. Several report only the mean age of participants.

Should these studies contribute to one synthesis, or does combining them risk hiding age-related differences?

Age can matter substantially, but an age difference does not automatically make evidence incomparable. The important question is whether age could plausibly change the outcome, effect, mechanism, measurement, or applicability relevant to your review.

Combine age groups when they answer the same substantive question

Age can matter in several different ways

Age is often treated as a routine demographic variable, yet its methodological meaning depends on the research question. It may represent biological development, accumulated exposure, disease risk, social roles, educational stage, cognitive development, treatment tolerance, technology use, or numerous other characteristics.

That makes “Does age matter?” too broad a question. You need to ask how age could matter for the specific phenomenon being synthesized.

Age difference does not automatically create indirect evidence

GRADE guidance treats population differences, including age differences, as a potential source of indirectness when researchers apply evidence to a target population. Yet differences between the studied and target populations do not automatically justify lower confidence. The concern becomes important when there is a credible reason to expect those differences to change the effect relevant to the target question.

For example, evidence from adults may sometimes inform decisions about older adults if the underlying mechanism and relative effect are expected to remain similar. In another context, evidence from adults may provide poor guidance for young children because developmental physiology, dosing, measurement, behavior, or implementation differs fundamentally.

The substantive question determines which situation you face.

Distinguish baseline risk from effect modification

An age group can have a different baseline probability of an outcome without necessarily experiencing a different relative intervention effect.

Suppose an adverse event is much more common among older adults. Even if an intervention produces the same relative risk reduction across ages, the absolute benefit may be larger among older adults because their starting risk is higher.

Confusing these concepts can lead researchers to conduct age subgroup analyses simply because age predicts the outcome. A prognostic factor is not necessarily an effect modifier.

Do not create age categories mechanically

Age is fundamentally continuous, even though studies frequently convert it into categories such as children, adolescents, younger adults, middle-aged adults, and older adults. Those categories may be substantively useful, but their boundaries should not be treated as natural laws.

A threshold of 60 or 65 years may make sense for one research question and little sense for another. Developmental research may require much narrower categories, while another phenomenon may change gradually across adulthood.

When possible, base age categories on biological, clinical, developmental, social, or policy relevance rather than selecting cut points merely because they make subgroup analysis convenient.

Broad age ranges can hide within-study variation

A study reporting a mean participant age of 45 years might include people aged 18 to 80. Another study with the same mean could contain almost exclusively participants aged 40 to 50. Their reported means therefore do not establish equivalent age composition.

This becomes particularly important in meta-regression. Cochrane guidance highlights age as an example of aggregation bias: a true relationship between age and treatment effect may exist within individual studies but remain invisible when researchers compare only the mean ages of entire studies.

The reverse problem can also occur. A relationship between study-level mean age and effect estimates does not prove that age modifies the effect among individuals.

Ask whether the intervention operates differently by age

Age deserves greater attention when there is a plausible reason that the intervention or exposure would operate differently across the age range.

In clinical research, pharmacokinetics, comorbidities, organ function, developmental physiology, treatment tolerance, or competing risks may matter. In education, developmental stage, prior knowledge, autonomy, and curriculum can change the meaning of an intervention. In technology research, device familiarity or patterns of use may differ across generations.

None of these differences should simply be presumed. They provide reasons to investigate whether age-related effect modification is plausible.

Check whether the outcome has the same meaning across ages

Measurement comparability can become as important as intervention comparability. The same instrument may not have equivalent validity across children and adults. A behavioral outcome may have different developmental meanings. Functional independence, academic performance, social participation, or quality of life may also be operationalized differently across life stages.

If outcomes are not measuring sufficiently comparable constructs, pooling them because they share a label can produce a deceptively precise answer to an unclear question.

Compare effects directly rather than comparing significance

A familiar subgroup mistake occurs when researchers observe a statistically significant effect among younger participants and a nonsignificant effect among older participants and conclude that the treatment works only in younger people.

That conclusion does not follow. Statistical significance depends partly on sample size and precision. The effect estimates themselves might be very similar.

To evaluate age-related effect modification, the relevant analysis examines evidence for a difference between effects. Even then, subgroup findings should be interpreted alongside their precision, prespecification, biological or substantive plausibility, consistency across studies, and vulnerability to confounding or multiple testing.

Sometimes individual participant data provide a better answer

When age-related modification is central to the research question, aggregate study-level data may be inadequate. Individual participant data meta-analysis can allow age to be modeled at the participant level rather than relying on study means or broad published categories.

This can be particularly useful when age is continuous or when different studies use incompatible age cutoffs. It does not remove every source of bias, but it can avoid some limitations inherent in study-level age comparisons.

Age also affects applicability

Even when a synthesis is statistically coherent, you still need to ask whether it applies to the population of interest. A body of evidence dominated by middle-aged adults may offer less direct evidence for children or very old adults when age-related differences could plausibly alter the finding.

This is part of the broader question of when evidence from another population is directly relevant to yours. The answer depends on the characteristics that matter to the phenomenon, not simply whether the populations have different demographic labels.

Suppose an intervention is studied from adolescence through older adulthood

The result may be less dramatic than announcing that an intervention “works only below age 60,” but it is methodologically more defensible. Age thresholds can look wonderfully crisp in a forest plot while the underlying biology remains stubbornly continuous.

Common mistakes when synthesizing evidence by age

Let the research question determine how much age should structure the synthesis

Before dividing evidence into age groups, specify why age could change the finding. That reasoning should guide extraction, subgroup definitions, synthesis, and interpretation.

Sometimes the most useful conclusion is that the available evidence does not permit a reliable age-specific inference. That is not a failed synthesis. It identifies precisely where the evidence stops supporting confident interpretation.

If the target age group differs substantially from those studied and there is a credible reason the finding could change, it may be necessary to state explicitly that the evidence does not generalize directly.

Before combining evidence across age groups, check:

Questions about synthesizing evidence across ages

Age should structure the synthesis only when it matters to the question

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.