Research methods guide
How to report subgroup analyses and interactions
To compare effects between subgroups, use an interaction term or equivalent formal contrast on the chosen scale. Report subgroup estimates along with the interaction.
In a synthetic example, ORs are 1.25 and 1.10 in subgroups A and B. Even if only one subgroup has P<0.05, estimate the contrast between effects directly.
Separate subgroup estimates from the interaction
The example uses an exposure-by-subgroup interaction on the log-odds scale, A relative to B.
| Item | Worked example | Verification |
|---|---|---|
| Within-subgroup estimates | OR 1.25 in A and 1.10 in B. Report n, events, and intervals for both. | Verify that the event, exposure coding, and adjustments are comparable. |
| Formal comparison | The exponentiated interaction is approximately 1.25/1.10=1.14 under the stated common model. | Obtain the interval from the interaction or planned contrast rather than comparing within-group P values. |
| Interpretation | If the interaction interval is 0.88–1.47, report that the between-subgroup contrast is imprecise. | Label prespecification and assess whether meaningful differences remain compatible with the interval. |
Apply the method to your materials
Select the interaction scale first
Explain whether the target is an absolute risk contrast or a relative scale. Define the interaction, covariates, and reference levels in one model.
Plan for multiple comparisons
Record the number of comparisons, model, exploratory status, and any adjustment. Show intervals and event counts, including imprecise results.
Questions that arise in practice
Does significance in only one subgroup show interaction?
Use the interaction estimate and uncertainty on the chosen scale. Separate P values reflect each subgroup’s data and effect.
Verified sources
The official sources below were checked on 2026-11-12. Access conditions and update dates are rechecked before publication.
RigorMed workspace
Start with your research materials
Bring your research question, available materials, and the steps you want to check.
