Concepts to be considered when appraising applicability of evidence to older people - sub-group analyses

Definition

Subgroup analyses - examining whether a treatment effect differs across pre-specified patient subgroups (e.g. by age band) within a larger trial - a common but statistically hazardous way trial data is used to make claims about specific populations like older adults
Subgroup findings (including age-based subgroups relevant to geriatric practice) are frequently reported and cited, but are much more prone to false-positive or misleading conclusions than the trial's primary overall result

Multiple subgroup comparisons increase the risk of a chance finding appearing statistically significant (multiple testing problem); subgroups are also typically smaller and therefore underpowered compared with the overall trial population, widening the true uncertainty around any subgroup-specific estimate

Pre-specified subgroup analyses (planned before the trial, stated in the protocol) are more credible than post-hoc subgroup analyses (identified after looking at the data), which carry a much higher risk of being a spurious/chance finding dressed up as a real effect
A formal statistical test for interaction (not just comparing whether one subgroup's result is "significant" and another's is not) is the correct way to determine whether a treatment effect genuinely differs between subgroups

Treat subgroup-specific claims (e.g. "this treatment doesn't work in patients over 80") with appropriate caution unless pre-specified, adequately powered, and supported by a formal interaction test - the overall trial result generally remains the best available estimate for an individual patient unless there is strong subgroup evidence otherwise
Use subgroup data to generate hypotheses for further dedicated study, not as definitive practice-changing evidence on its own

Age-based subgroup claims in cardiovascular/oncology trials, statistical multiple testing, evidence-based medicine appraisal

Many "this doesn't work in the elderly" claims in clinical folklore trace back to underpowered, non-pre-specified subgroup analyses rather than dedicated, adequately powered evidence in that population - critical appraisal skill is needed to distinguish a genuine effect-modification finding from statistical noise

5 more sections, plus exam facts

Premium unlocks every note across every specialty, and the full exam fact library behind it.