01 · The Question
What If the Paper's Main Outcome Was Not Originally the Main Outcome?
You read a study reporting a statistically significant result for its primary outcome. Because the outcome is labeled primary, you reasonably assume that it was the study's main test from the beginning.
Then you find the study's earlier registration or protocol. The published primary outcome was originally secondary, or perhaps was not specified at all. Meanwhile, an original primary outcome has disappeared or been relegated to secondary status.
Does that invalidate the paper? Not automatically. But it changes what you need to know before interpreting the finding as confirmatory evidence.
03 · What You Need to Know
Why the History of an Outcome Matters
Primary outcomes carry special interpretive weight
A primary outcome is ordinarily intended to represent a study's main outcome of interest. In confirmatory research, specifying that outcome in advance helps establish what the study was designed principally to test before investigators know whether the result will be favorable.
This matters because a study can measure many things. If researchers are free to inspect all the results first and then decide which outcome deserves primary emphasis, an apparently strong confirmatory result may actually have emerged from a larger set of possibilities.
Prespecification therefore helps separate a planned test from a finding selected after looking at the data.
Outcome switching can take several forms
The phrase is sometimes used as though it means only replacing one primary outcome with another. In practice, discrepancies between prespecified and published outcomes can take several forms.
Change
What happens
Why it may matter
Omission
A prespecified primary outcome is absent from the publication
Readers cannot evaluate one of the study's original main outcomes
Downgrading
A prespecified primary outcome becomes secondary or otherwise loses primary status
An unfavorable main result may receive less emphasis
Upgrading
A prespecified secondary outcome becomes primary
A favorable secondary finding may acquire stronger confirmatory prominence
Introduction
A new primary outcome appears that was not prespecified
Readers may mistake a post hoc choice for the study's original main test
Redefinition
The measurement, aggregation, metric, or time point changes
An apparently identical outcome may actually represent a different result
A recent BMJ meta-epidemiological study illustrates why these distinctions matter. Among 124 registered cohort studies of interventions published from 2015 to 2024, 48% had discrepancies between registered and published primary outcomes. Only two studies with outcome switching provided an explanation. Among studies with switching other than omission, 77% of the changes favored statistically significant findings. Those figures apply to the particular cohort examined and should not be generalized automatically to other research designs or disciplines.
An outcome is more than its label
Comparing only outcome names can miss meaningful changes. "Depression," "achievement," or "quality of life" may each be operationalized in several ways.
An outcome can involve a particular measurement variable, metric, method of aggregation, and time point. Changing from six-month performance to immediate post-test performance, for example, can change the result being evaluated even if both are loosely described as "academic performance."
Cochrane's current risk-of-bias framework similarly asks whether a reported result could have been selected from multiple eligible measurements, definitions, scales, or time points on the basis of the findings.
Outcome switching is concerning when the results drive the switch
The strongest concern arises when investigators know the results and then alter outcome priorities in a way that favors the preferred finding.
Suppose a study prespecifies reading comprehension as its primary outcome and engagement as secondary. Reading comprehension shows little difference, while engagement improves substantially. If the publication silently presents engagement as the primary outcome, the paper gives the favorable finding a confirmatory status it did not originally possess.
The engagement result does not become false. What changes is its evidential role.
Prespecified confirmatory result
The outcome and relevant analysis were established before the researchers knew the result.
Post hoc or exploratory result
The outcome or its priority was determined after relevant data were available and should be interpreted with that exploratory context visible.
Changing an outcome is not automatically wrong
There are legitimate reasons to change outcomes. New external evidence may show that the original endpoint is inappropriate. Measurement problems may emerge. Recruitment or event rates may differ substantially from expectations. An adaptive design may also permit planned modifications.
CONSORT guidance therefore does not prohibit every outcome change. Instead, changes to outcomes after a trial begins should be identified and explained. Changes based on unblinded outcome data are more problematic because investigators may already know which choice produces the preferred result.
Timing is therefore crucial. A documented change made before relevant outcome data were available is quite different from an unexplained change made after investigators examined the findings.
Outcome switching overlaps with selective outcome reporting
Outcome switching and selective outcome reporting are closely related but not identical.
Selective outcome reporting concerns which measured outcomes or results become available. Outcome switching focuses particularly on discrepancies in the status, specification, or identity of outcomes between the prespecified plan and the final report.
A prespecified primary outcome might disappear entirely, creating both an outcome-switching discrepancy and selective non-reporting. Alternatively, the original primary outcome might remain fully reported but be demoted while another outcome is promoted. In that case, nothing necessarily disappears, yet the hierarchy of evidence has changed.
Outcome switching differs from selecting among statistical analyses
A study can retain exactly the same primary outcome while changing how that outcome is analyzed. Choosing the most favorable model from several analytical alternatives is better understood as selective analysis reporting .
The distinction is useful because the remedies differ slightly. Outcome switching requires comparing what was supposed to be measured and prioritized. Analysis switching requires examining how the data were supposed to be analyzed.
Both questions ultimately depend on records created before the relevant results were known.
The paper alone may not reveal that switching occurred
A publication can be internally coherent. The Methods section identifies a primary outcome, the Results section reports it, and the Discussion interprets it. Nothing within the paper necessarily reveals that the designation differs from an earlier plan.
Detecting switching therefore often requires an external reference point such as a prospective registration, protocol, dated statistical analysis plan, ethics submission, or other prespecified documentation.
Watch Out
A publication's statement that an outcome was "primary" tells you how the paper labels it. It does not by itself establish that the outcome was designated as primary before the relevant results were known.
Registration helps only when it is timely and specific
A registration record is most informative when the outcomes were specified prospectively and in enough detail to distinguish the planned result from alternatives.
Vague entries such as "student performance" may leave considerable flexibility about the instrument, metric, aggregation, or time point. Likewise, a registry entry created or substantially changed after investigators already had access to outcome data provides weaker evidence of genuine prespecification.
The practical question is therefore not merely whether a registration exists. Ask whether the relevant version predates knowledge of the results and specifies the outcome precisely enough to support a meaningful comparison.
06 · What This Means for You
Compare the Published Outcome With the Outcome That Was Planned
When a paper makes an important confirmatory claim, particularly one centered on a primary outcome, look for evidence showing what was specified before the result was known.
A simple decision framework
If a prospective registration or protocol is available
Compare the outcome's identity, priority, measurement, metric, aggregation, and time point with the publication.
If the primary outcome changed
Determine when the change occurred and whether the authors disclose and justify it.
If a new favorable outcome became primary after results were available
Interpret that result more cautiously as potentially selected rather than automatically treating it as a prespecified confirmatory finding.
If a change was documented before relevant outcome data were known
Evaluate the scientific rationale rather than assuming that the discrepancy represents result-driven switching.
For your own research, specify outcomes precisely enough that another researcher could later determine whether the reported outcome is the one you originally intended. If a legitimate change becomes necessary, preserve the original record, date the amendment, explain why it was made, and report the deviation in the final paper.
The aim is not methodological rigidity for its own sake. It is to let readers distinguish what the study set out to test from what researchers learned after seeing the data.
07 · A Quick Checklist
How to Check a Paper for Outcome Switching
Before interpreting a reported primary outcome as prespecified, check:
Whether a protocol, prospective registration, or other dated research plan exists.
Whether the published primary outcome was designated as primary in the relevant prespecified record.
Whether any prespecified primary outcomes were omitted or downgraded.
Whether secondary or previously unspecified outcomes were promoted to primary status.
Whether measurement methods, metrics, aggregation methods, definitions, or time points changed.
Whether the timing of amendments shows that changes occurred before the relevant results were known.
Whether the publication explicitly reports and explains important deviations from the original outcome plan.
11 · Cite this Guide
How to Cite This Guide
This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.
Recommended (Field Guide)
APA
MLA
Chicago
Copy Citation
