Multiple Endpoints: Which Question Was the Study Designed to Answer?
One experiment can ask many questions
A peptide study might measure several proteins, assess multiple time points and compare several groups. Each analysis can contribute information. The difficulty arises when the most striking result is presented as though it were the single question the experiment was designed to answer.
Distinguish outcomes from analyses. One outcome measured at several times can generate multiple comparisons, as can separate analyses in subgroups. Counting only the number of named endpoints may therefore underestimate the range of opportunities for an apparently unusual result to emerge.
Find the question specified in advance
A primary outcome identifies the principal endpoint used to address the main study question. Secondary outcomes provide additional information, while exploratory analyses may generate hypotheses for future work. These labels are most informative when documented before results were known.
The CONSORT 2025 explanation and elaboration: updated guideline for reporting randomised trials asks authors to define outcomes and report important changes. Compare the published paper with its protocol or registration where available. If a different endpoint becomes prominent in the abstract, look for an explanation rather than assuming that the change was either harmless or improper.
Understand the multiplicity problem
Running many statistical tests creates more opportunities for a small P value to appear by chance when the relevant null hypotheses hold. The details depend on the number and relationships of the tests, so a universal numerical correction is not appropriate for every design.
The Reporting animal research: Explanation and elaboration for the ARRIVE guidelines 2.0 discusses prespecified outcome measures and statistical reporting in animal experiments. For a hypothetical peptide project, a broad screen of signalling markers can be valuable discovery work. A highlighted marker from that screen should be described as a candidate finding rather than retrospectively promoted to a prespecified confirmatory result.
Preserve the whole pattern
Read tables as well as the abstract. Were related outcomes consistent? Were negative findings shown? Did the study distinguish planned analyses from those added after seeing data? An isolated difference can become more interpretable when its place within the complete set of measurements is visible.
A practical note for our research resources records the primary endpoint, its time point, the analysis plan and any multiplicity method. Then describe exploratory results in their own terms. This protects useful discovery findings from being oversold and helps identify the focused question that an independent follow-up experiment could test.
Related research reading
Explore the research blog and original batch-document library for further context.