01 · The Question
Is It Too Late to Change a Primary Outcome Once Data Collection Has Started?
You begin a study with a clearly specified primary outcome. Then something changes. Perhaps the outcome proves difficult to measure reliably. New evidence suggests that another endpoint is more meaningful. An event occurs far less frequently than anticipated. Or an unforeseen methodological problem makes the original outcome unsuitable.
Can you change it?
Potentially, yes. Beginning data collection does not make every element of a study permanently immutable. But changing a primary outcome is a consequential amendment, and the circumstances matter greatly. In particular, a change made without knowledge of comparative outcome results is very different from replacing an unfavorable primary outcome after seeing that another outcome produces a better result.
02 · The Short Answer
A Primary Outcome Can Change, but the Change Must Remain Visible
In Brief
Yes, a primary outcome can sometimes be changed after data collection begins when there is a defensible scientific, methodological, or practical reason, but the original outcome, timing of the amendment, rationale, and relevant access to outcome data should be transparently documented and reported.
A change becomes particularly problematic when researchers already know the results and promote a more favorable outcome while concealing the original primary outcome. Applicable protocol, registration, ethics, regulatory, funder, and journal requirements should also be followed.
03 · What You Need to Know
The Timing of a Primary Outcome Change Is Only Part of the Story
Starting Data Collection Is Not a Magical Point of No Return
It is tempting to use a simple rule: a primary outcome can be changed before the first participant is enrolled but never afterward. Research practice is more complicated.
Protocols can require amendments after a study begins. SPIRIT 2025 describes the protocol for a randomized trial as a living document that may be formally amended during the trial and recommends a transparent audit trail of changes. Important amendments should be reported to relevant ethics bodies and trial registries as they occur and subsequently described in reports of the completed trial.
CONSORT 2025 likewise requires reports of randomized trials to identify important changes made after the trial commenced, including outcomes or analyses that were not prespecified, and to give the reason for those changes.
The relevant question is therefore not simply, “Had data collection started?” You also need to ask what prompted the change and what information was available when the decision was made.
Access to Outcome Results Changes the Risk of Bias
Consider three points at which researchers might amend a primary outcome:
Timing
What researchers know
Main concern
Before data collection
No study outcome data exist
The change may still require documentation, but it cannot have been selected because of observed study results
After data collection begins but before relevant comparative results are known
Some data exist, but investigators remain appropriately masked to the outcome comparison relevant to the decision
The rationale and chronology still need documentation, but result-driven switching may be less plausible
After relevant results are examined
Researchers know how the original and alternative outcomes behave
A change can be influenced by which outcome produces the more favorable result
This is why “after data collection begins” is too crude a boundary. In a long randomized trial, thousands of observations might already exist while investigators remain masked to treatment-specific outcome results. Conversely, in a small unblinded study, researchers might see the relevant results almost immediately.
There Are Legitimate Reasons to Amend a Primary Outcome
A primary outcome might need reconsideration because circumstances emerge that were not reasonably anticipated when the protocol was written.
For example, an outcome may prove infeasible to collect, an instrument may become unavailable, new scientific evidence may establish a better endpoint, or event rates may differ so substantially from assumptions that the original design becomes untenable.
CONSORT 2025 provides an illustrative example in which a trial originally used all-cause mortality as its primary endpoint. During a masked analysis, unexpectedly low mortality made the originally planned study impractical, and the investigators adopted co-primary endpoints that retained the original endpoint and added a prespecified secondary endpoint. The important features are not merely that the endpoint changed, but that the reason and circumstances of the amendment were reported.
That is quite different from discovering after analysis that the primary outcome is nonsignificant and deciding that a statistically significant secondary outcome would make a more appealing primary endpoint.
A New Scientific Rationale Does Not Erase the Original Outcome
Suppose strong external evidence published midway through your study shows that another measure is substantially more valid than the one you selected. You may have a defensible reason to amend the outcome.
The amendment should not rewrite history. The research record should continue to show what was originally primary, what became primary, when the decision occurred, and why.
This distinction is central to understanding outcome switching . A transparent amendment allows readers to reconstruct the chronology. An undisclosed switch makes the final outcome appear to have been the plan from the beginning.
Changing the Primary Outcome Can Affect More Than the Label
The primary outcome often has methodological consequences. In randomized trials, CONSORT describes it as the prespecified outcome considered most important and indicates that it should be the outcome used in the sample-size calculation.
Changing it may therefore affect questions about statistical power, multiplicity, interpretation, analysis, and whether the existing sample-size justification still makes sense.
A new primary outcome should not simply be substituted into the manuscript while everything else is assumed to remain valid. Researchers need to consider what methodological consequences follow from the amendment.
Changing the Time Point or Definition Can Also Change the Primary Outcome
A primary outcome is more specific than a broad concept such as “depression,” “achievement,” or “mortality.”
CONSORT 2025 asks randomized-trial reports to specify the measurement variable, analysis metric, method of aggregation, and time point for each primary and secondary outcome. SPIRIT 2025 similarly requires detailed outcome specification in protocols.
Suppose your protocol specifies mean depression score at six months, but the paper reports the proportion of participants below a diagnostic threshold at three months as the primary outcome. You have not preserved the original primary outcome merely because both involve depression.
Changing how, when, or in what form an outcome is evaluated can materially change the research question.
Knowing That the Original Outcome Is Nonsignificant Is a Major Warning Sign
Suppose the primary outcome produces p =.21, while a secondary outcome produces p =.01. You then decide that the secondary outcome is theoretically more meaningful and redesignate it as primary.
Perhaps there really is a compelling theory. The problem is that you developed or applied that rationale while already knowing which outcome produced the more favorable result.
The data can now influence the decision.
Empirical evidence shows why this deserves scrutiny. In a comparison of randomized-trial protocols and publications, Chan and colleagues found that statistically significant outcomes were more likely to be fully reported than nonsignificant outcomes, and 62% of the trials examined had at least one primary outcome that was changed, introduced, or omitted between the protocol and publication.
This does not mean every discrepancy was intentionally manipulated. It demonstrates why prospective records and transparent explanations matter.
Changing the Registry Afterward Does Not Make the Outcome Prospectively Specified
A registry record may be editable, but editing history matters.
If you change the registered primary outcome after examining the results, the new entry does not become a genuinely prespecified outcome merely because it now appears in the registry. Prospective specification concerns chronology, not just what the current webpage says.
Watch Out
Never erase the chronology of an outcome amendment. Preserve dated versions of protocols, registrations, and analysis plans where applicable. Retrospectively changing a record so that the final outcome appears to have been specified earlier defeats much of the transparency those records are intended to provide.
04 · A Practical Example
Changing an Outcome Mid-Study Without Rewriting the Study
Hypothetical Example
An Intervention With an Unexpected Measurement Problem
A randomized intervention study prespecifies a particular standardized assessment at six months as its primary outcome. After recruitment begins, the assessment provider announces that the instrument will no longer be supported, creating a serious feasibility problem for later participants.
Identify the problem
The research team determines that consistent administration of the original primary measure will not be feasible across the completed study.
Protect the decision from outcome knowledge
The amendment is considered without examining comparative treatment results for the candidate replacement outcome.
Select and justify the replacement
The researchers identify a validated alternative and document why it better preserves the intended construct.
Amend the research record
Applicable protocol, registration, ethics, statistical analysis, and governance records are updated while retaining the original specification and date of the change.
Report the amendment
The final manuscript tells readers what the original primary outcome was, what replaced it, when the decision occurred, and why.
This study has changed its primary outcome after data collection began. That fact alone does not make the amendment deceptive.
Now change one detail. Imagine the researchers first analyze both outcomes, discover that the original endpoint is unfavorable while the alternative is statistically significant, and only then make the alternative primary. The same nominal change now carries a much greater risk of outcome-driven selection.
The difference is not cosmetic. It concerns what the researchers knew when they chose which outcome would carry the greatest inferential weight.
06 · What This Means for You
What to Do When Your Primary Outcome Really Needs to Change
If you encounter a genuine reason to amend the primary outcome, treat it as a documented methodological decision rather than quietly editing the study into a cleaner story.
A simple decision framework
If the change can be made before examining relevant outcome results
Make and document the decision at that point rather than waiting until the comparative results are known.
If an external scientific, measurement, or feasibility problem motivates the change
Record the evidence and rationale supporting the amendment.
If the new outcome changes the statistical design
Reconsider implications for sample size, analysis, multiplicity, and interpretation rather than merely changing the outcome label.
If you already know the relevant outcome results
Do not present a newly favored outcome as though it had always been primary. Disclose the timing and data-dependent context explicitly.
If the study is governed by a protocol, registry, ethics approval, statistical analysis plan, or regulatory process
Follow the applicable amendment and reporting requirements rather than assuming that changing the manuscript alone is sufficient.
Requirements vary across study designs and institutions. The SPIRIT and CONSORT recommendations discussed here specifically concern randomized trials. Observational studies, laboratory research, qualitative research, and other designs may operate under different reporting frameworks.
The transferable principle is simpler: do not make the final report erase the sequence of decisions that actually occurred.
07 · A Quick Checklist
Before Changing a Primary Outcome After Data Collection Begins
Before making the amendment, check:
What specific scientific, methodological, ethical, or practical problem requires the primary outcome to change?
Have you or the decision-makers already seen comparative results for the original or proposed replacement outcome?
Can the amendment be decided while relevant outcome information remains masked?
Have you preserved the original outcome specification and dated the amendment?
Have you documented why the replacement outcome is appropriate?
Does the change affect sample-size assumptions, multiplicity, statistical analysis, or interpretation?
Have applicable protocol, registry, ethics, funder, regulatory, and statistical analysis records been amended as required?
Will the final report tell readers what changed, when it changed, and why?
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
