An interesting finding can be valuable even when it was not the original question. The problem is not exploration itself; it is presenting an analysis chosen after seeing the data as though it had been fixed beforehand. Clear labels let readers assign appropriate weight to each result.
Find the dated analysis record
The ICH statistical-principles material and its estimand addendum emphasise alignment between the research question, planning and interpretation. For a reader, the protocol and statistical analysis plan help show which decisions were made before the relevant results were available. ICH E9 materials.
Record the document version, date and what it specifies. A broad phrase such as “subgroup analyses will be performed” is less informative than a record defining the subgroup, outcome, comparison and method.
Distinguish the analysis from its timing
An exploratory analysis can use sound statistics. A prespecified analysis can still have weaknesses. Timing is one important part of the evidence, not a replacement for evaluating the method itself.
If the plan changed, look for the reason and whether decision-makers had access to comparative outcomes. A transparent amendment is more informative than an unexplained difference between the dated plan and the final paper.
Consider a fictional search across results
Imagine researchers inspect ten outcomes at three time points in four subgroups. That creates many opportunities to notice an unusually striking result. Selecting the strongest-looking combination after inspection and reporting it alone would hide the search that produced it.
The appropriate response is to document the exploration, use a justified analysis strategy and seek independent confirmation where needed. It is not to pretend that every inspected combination was a separate prospectively planned hypothesis.
Keep confirmatory and exploratory tables readable
A results summary can distinguish the primary planned comparison, other prespecified analyses and post hoc exploration. Include the effect estimates, uncertainty and multiplicity approach where relevant. Do not upgrade a result merely because its p value is smaller than those in the planned analysis.
Make the next question explicit
An exploratory observation may suggest a useful future study. A careful conclusion states the observation, why it warrants investigation and what would be needed to test it independently. It does not promise replication. Our primary-and-secondary-endpoints guide explains why the original outcome hierarchy should remain visible in the published report.
Published by Clova. Sources, editorial responsibility & corrections · Contact the editorial team
Top comments (0)