01 · The Question
If a Study Measures Several Outcomes, Can Researchers Change Which One Matters Most?
A study begins with one outcome designated as primary. After data are collected, that outcome shows little evidence of an effect, but a secondary measure produces a striking result. The published paper then emphasizes the successful measure as the primary outcome.
That is the kind of situation that raises concern about outcome switching .
Yet research protocols sometimes need legitimate amendments. Outcomes can become impossible to measure, new evidence may change what is clinically or scientifically meaningful, or unforeseen circumstances may require a different assessment. The issue is therefore not simply whether an outcome ever changed. It is what changed, when, why, whether the researchers knew the emerging results, and whether the change was transparently reported .
02 · The Short Answer
Outcome Switching Changes the Outcomes a Study Appears to Have Planned
In Brief
Outcome switching occurs when prespecified outcomes are changed, introduced, omitted, or reclassified during or after a study, particularly when the final report differs from the protocol, registration, or original analysis plan.
An outcome change is not automatically improper. The serious concern is undisclosed or result-driven switching, especially when an unfavorable prespecified primary outcome is demoted or omitted while a favorable secondary or newly introduced outcome is promoted, because this can give readers a biased impression of what the study was designed to test.
03 · What You Need to Know
How Outcome Switching Changes the Meaning of Research Findings
The Primary Outcome Is Supposed to Be Identified in Advance
Many studies collect several outcomes, but they do not necessarily have equal status. In randomized trials, the primary outcome is the prespecified outcome considered most important for evaluating the intervention and is commonly linked to the study objective and sample-size calculation. Other prespecified outcomes may be secondary.
Current SPIRIT and CONSORT guidance requires primary and secondary outcomes to be clearly defined, including the measurement variable, analysis metric, method of aggregation, and relevant time point.
This specificity matters. Saying “depression” is not necessarily a complete outcome definition if researchers could later choose among several scales, scoring methods, or time points. Each additional degree of freedom can create another opportunity to favor the version producing the most attractive result.
Outcome Switching Can Take Several Forms
Change
Illustrative example
Why it matters
Primary outcome is demoted
The original primary outcome becomes secondary in the publication
The main planned test receives less emphasis after its result is known
Secondary outcome is promoted
A favorable secondary outcome becomes the reported primary outcome
The study may appear to have succeeded on its main objective when that was not the original test
Prespecified outcome is omitted
An unfavorable planned outcome disappears from the publication
Readers cannot see part of the evidence the study intended to collect
New outcome is introduced
An unplanned measure becomes a major reported endpoint
A post hoc finding may appear prespecified
Measurement time point changes
The protocol specifies six months, but the paper emphasizes three months
Different time points can produce different conclusions
Outcome definition changes
A continuous scale is replaced by a favorable threshold-based definition
The operational definition itself may be selected after results are known
Not every discrepancy necessarily represents deliberate manipulation. Some may result from legitimate protocol amendments, reporting errors, or practical problems. That is precisely why researchers should explain changes rather than leave readers to guess.
Why Switching the Primary Outcome Is Especially Consequential
Imagine a trial with one prespecified primary outcome and eight secondary outcomes. The primary result is nonsignificant, while one secondary outcome produces p =.02.
If the paper simply reports the secondary finding as the primary endpoint, the evidential history has changed. Readers may reasonably assume that this was the main outcome selected before results were available and that the study's design and inferential strategy were built around it.
They are no longer seeing the same research question the study originally set out to answer.
CONSORT notes that having multiple outcomes creates problems involving multiplicity and selective reporting, and it recommends consistency between reported primary and secondary outcomes and those prespecified in protocols and registries. Changes should be identified and explained.
Outcome Switching Is Closely Related to Selective Outcome Reporting
The concepts overlap but are useful to distinguish.
Outcome switching
Changes the identity, definition, timing, or status of outcomes relative to what was prespecified.
Selective outcome reporting
Reports outcomes selectively so that the published account does not adequately represent the outcomes measured or planned.
If a prespecified primary outcome is silently omitted and a favorable secondary outcome is promoted, both problems may occur. The outcome has been switched, and the resulting report selectively presents the evidence.
The broader issue of selective outcome reporting therefore extends beyond merely changing which outcome carries the “primary” label.
Outcome Switching Can Favor Statistically Significant Findings
This concern is not merely hypothetical. CONSORT's evidence review notes discrepancies between outcomes reported in protocols or registries and those appearing in final publications, with discrepancies frequently favoring statistically significant outcomes.
CONSORT also cites empirical work in which protocol-publication comparisons identified primary outcomes that had been changed, introduced, or omitted, often without those changes being explained in the published reports.
The implication should be interpreted carefully. A discrepancy does not by itself prove that researchers deliberately changed an outcome to obtain a publishable result. But systematic changes favoring statistically significant findings can distort the published evidence base.
Changing an Outcome Is Not Automatically Wrong
Research occurs in the real world, where circumstances change.
A measurement instrument may become unavailable. New external evidence may show that the originally selected time point is unsuitable. Recruitment or event rates may differ substantially from assumptions. A better-validated measure may become necessary. Formal adaptive designs may also allow prospectively specified modifications.
CONSORT explicitly recognizes that outcome changes can sometimes be necessary. Its recommendation is not “never change an outcome.” Instead, important changes after a trial begins, including outcomes that were not prespecified, should be reported together with the reasons for those changes. Changes made using unblinded outcome data are substantially more concerning.
The timing and information available at the time of the decision therefore matter enormously.
Changing an Outcome Before Knowing the Results Is Different From Changing It Afterward
Suppose investigators learn midway through a blinded trial that an external study has established a substantially better measurement instrument. They amend their protocol before examining outcome differences between groups, document the change, update relevant records, and explain it in the final report.
Now compare that with investigators who examine the group results, discover that the primary outcome is unfavorable, and then promote a favorable secondary endpoint.
Both studies changed an outcome. Their risks of bias are very different.
This is why the question of whether you can change a primary outcome after data collection begins cannot be answered simply by saying yes or no. What information was available when the decision was made is central.
Registries and Protocols Make Outcome Switching Detectable
If the only public record is the final paper, readers may have no way to know that an outcome changed.
Prospective protocols, trial registrations, and statistical analysis plans create records of what researchers intended before the results were known. CONSORT notes that access to the protocol provides context for interpretation and can help constrain undeclared post hoc methodological changes and selective outcome reporting.
These records do not prevent every problematic change. Their value is partly forensic: they allow readers, reviewers, editors, and systematic reviewers to compare what was planned with what was eventually reported.
Watch Out
Simply updating a registry after results are known does not make an outcome retrospectively prespecified. When an outcome changes, preserve the chronology. Readers need to know what was planned originally, what changed, when it changed, and why.
04 · A Practical Example
When a Secondary Outcome Quietly Becomes the Main Result
Hypothetical Example
A Trial of a Stress-Reduction Program
A randomized trial prespecifies perceived stress at 12 weeks as its primary outcome. Sleep quality, anxiety, wellbeing, and program satisfaction are secondary outcomes.
Primary outcome
The difference in perceived stress at 12 weeks is small and not statistically significant.
Secondary outcomes
Most secondary outcomes are also inconclusive, but sleep quality shows a statistically significant difference favoring the intervention.
The switch
The manuscript identifies sleep quality as the primary outcome and gives perceived stress little attention. The change from the registered protocol is not mentioned.
The impression
A reader who sees only the paper may reasonably conclude that the study was designed primarily to test whether the program improves sleep and that its primary hypothesis was supported.
The sleep result itself has not become false merely because it was originally secondary. It may be scientifically interesting.
The problem is that its evidential status has been rewritten. The publication makes a secondary finding appear to be the study's prespecified main test while obscuring the unsuccessful primary outcome.
A transparent paper could instead report that the prespecified primary outcome did not provide clear evidence of an effect, present the sleep finding with its correct original status, discuss relevant multiplicity, and explain any legitimate changes to the protocol.
06 · What This Means for You
What to Do If an Outcome Really Needs to Change
The safest approach is not to treat the original protocol as immutable. It is to preserve the history of the study.
A simple decision framework
If the change is proposed before relevant outcome data are available
Document the rationale, timing, and amendment through the appropriate protocol, registry, ethics, or governance procedures for your study.
If the outcome must change for a methodological or practical reason
Explain why the original outcome is no longer appropriate or feasible and preserve the original specification in the study record.
If you have already examined the outcome results
Treat a proposed switch with much greater caution and make the data-dependent timing explicit rather than presenting the replacement as prespecified.
If an unexpected secondary outcome is compelling
Report and interpret it in its actual prespecified or exploratory status instead of silently promoting it to primary.
If the final report differs from the protocol or registry
Identify the consequential differences and explain why they occurred.
For randomized trials, CONSORT 2025 specifically requires important changes after commencement, including outcomes or analyses that were not prespecified, to be reported with reasons. Results should also be reported for planned primary and secondary outcomes rather than only those that are statistically significant or interesting.
Requirements differ across study designs, disciplines, registries, journals, and funders, so researchers should also verify the standards that apply to their own work.
07 · A Quick Checklist
Before Changing or Reporting an Outcome
If an outcome has changed, check:
Was the original primary outcome clearly specified in the protocol, registration, or analysis plan?
What exactly changed: the outcome, its primary or secondary status, measurement instrument, analysis metric, or time point?
When was the decision made relative to data collection and access to outcome results?
Was the change motivated by external scientific or methodological information rather than by a favorable observed result?
Have the original outcome and the amendment both been preserved in the research record?
Have applicable registry, protocol, ethics, funder, or governance records been updated appropriately?
Does the final manuscript explicitly disclose and explain consequential outcome changes?
Are planned outcomes reported even when their results are nonsignificant or less interesting?
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
