An encouraging subgroup result should not replace the trial's overall comparison. This is particularly important when the subgroup analysis is exploratory. Primary study
What was reported
The trial randomized 111 patients to growth hormone or placebo after hip fracture, with six weeks of treatment and follow-up through 24 weeks. Overall changes in the Modified Barthel Index did not differ significantly. An exploratory analysis of older participants favored treatment for return to prior independent living. Baseline living situations were imbalanced. Primary report
Our appraisal: a restricted analysis changes the question
The report restricted living-situation comparisons to participants who had been independent before fracture, and the relevant table concerned those completing follow-up. Those restrictions make the analysis more specific than a comparison of everyone randomized.
That specificity can be clinically sensible, but it also reduces the sample and changes the population represented by the result. Missing follow-up can matter if completion is related to health or recovery.
The overall daily-function finding should therefore remain prominent. It is not accurate to describe the trial simply as proving better functional recovery because one age-defined analysis was favorable.
Age subgroup interpretation also requires a direct assessment of whether treatment effects differ across ages. A favorable result within one group alone is not enough. Exploratory findings are most useful for designing a subsequent confirmatory study.
The paper illustrates why outcome definitions, baseline balance, and analysis populations matter. It provides a hypothesis about selected older patients, not a demonstrated general recovery benefit or a basis for choosing treatment solely from an age cutoff.
Frequently asked questions
Was the overall Barthel comparison significant?
No.
Was an older subgroup finding favorable?
Yes, for a living-situation analysis.
Was that analysis exploratory?
Yes.
Were baseline living situations balanced?
No.
Did the living analysis represent everyone randomized?
No.
Does subgroup significance prove an age interaction?
No.
Is general functional recovery benefit established?
Not by this trial.