- Published
- Updated
When several hypotheses support confirmatory claims, test order and error control belong in the design. A long endpoint list does not itself explain which findings can be considered confirmatory under which combinations of results.
Define the question and assumptions
Map hypotheses by clinical priority and decision relationships. Describe how one test affects access to the next, which hypotheses belong to the family and how interim analyses connect. The method needs specialist statistical justification; calling a retrospectively chosen order a hierarchy does not provide valid prospective error control.
Hypothetical study example
In a hypothetical study, a secondary endpoint shows a strong difference after the primary hypothesis is not met. The planned sequence does not permit confirmatory interpretation of that finding. The team reports the estimate without claiming a level of certainty that the testing strategy cannot support.
Design review checks
- Define the hypothesis family and separately identify exploratory evaluations outside it.
- Test transition conditions and stopping points using example combinations of results.
- Reconcile ordering across the protocol, analysis plan and planned report tables.
Record the decision and its limits
An accepted map yields consistent decisions for different result combinations. Apply the prespecified strategy rather than rearranging the sequence to accommodate favourable findings. Clinically interesting exploratory observations may inform future research. Multiplicity control cannot make an unsuitable comparison or poor-quality measurement scientifically valid simply by adjusting a statistical threshold.
Frequently asked questions
- Are all secondary endpoints exploratory?
- No. Their roles and positions in a testing strategy can be prespecified.
- Can test order change after results are known?
- Explain changes and consequences; the original confirmatory status is not automatically preserved.