- Published
- Updated
Secondary endpoints can complement the primary question, but they do not all carry the same inferential strength. Listing an endpoint in a protocol does not itself establish a separate confirmatory claim or resolve every multiplicity issue.
Define the question and assumptions
For each secondary endpoint, define its scientific purpose, assessment time, analysis and testing role. Distinguish descriptive summaries, supportive analyses and confirmatory hypotheses. Explain which question determined sample size and how uncertainty will be interpreted for outcomes that were not the basis of the design calculation.
Hypothetical study example
In a hypothetical study, quality-of-life data are collected prospectively but the endpoint is outside the confirmatory testing family. The team reports the difference and confidence interval without substituting it for primary success. Planned collection and confirmatory inference are explicitly treated as separate properties.
Design review checks
- Assign a prospective scientific and statistical role to every secondary endpoint.
- Check measurement, analysis and reporting times against the agreed protocol definitions.
- Track reporting of all planned secondary findings rather than selecting favourable ones.
Record the decision and its limits
Accepted interpretation applies the same role rules to favourable and unfavourable findings. Clearly label an outcome that becomes interesting only after data inspection. An uncertain result is not worthless: present its interval and limitations, then discuss what additional evidence would be needed for a stronger claim.
Frequently asked questions
- Are prospectively collected data always confirmatory?
- No. Collection plans differ from testing and claim strategies.
- Should exploratory results remain unpublished?
- They can be reported, provided their status and uncertainty are explicit.