Trial Registration & Outcome Switching

A published trial report shows what investigators chose to present at the end of a study. A registry record, protocol, and statistical analysis plan can show what they intended at earlier points. Reading those records together helps distinguish prespecified questions from later changes and exploratory analyses.

None of these documents is a regulatory approval, an assurance that the methods were followed, or proof that the result is unbiased. Their value is transparency: dated versions make it possible to compare plans, changes, and final reporting.[1][2]

Four records with different jobs

The documents overlap, but they are not interchangeable.

Record	Main role	What to record when appraising it
Trial registration	Public minimum record and unique identifier	Registry, identifier, first-posted or registration date, enrollment date, outcomes, and update history
Full protocol	Planned design and conduct	Version date, objectives, eligibility, interventions, outcomes, sample size, monitoring, and amendment process
Statistical analysis plan	Detailed rules for turning data into estimates	Sign-off date, analysis populations, models, covariates, missing-data methods, multiplicity, and sensitivity analyses
Results report	What was done, found, and interpreted	Participant flow, numerical results, deviations, amendments, prespecified versus post-hoc labels, and limitations

WHO describes registration as making an agreed minimum set of trial information publicly accessible through a qualifying registry.[4] The checked WHO Trial Registration Data Set is version 1.3.1, with 24 items.[5] A full protocol and analysis plan ordinarily contain much more detail. A document labelled “protocol” should also be checked: public summary fields or an intervention procedure are not necessarily the full trial protocol.[2]

Prospective registration is a timing claim

Prospective registration is a policy-specific timing claim. WHO’s ICTRP portal flags a record as prospective when its supplied registration-date field is earlier than its first-enrollment field.[8] ICMJE recommends that medical journal editors require registration at or before first participant consent for enrollment. Importantly, ICMJE uses the date registration materials were first submitted, and says editors may investigate substantial submission-to-posting delays.[7]

Submission, registration, public posting and consent dates are not synonyms. ICMJE also expects an acceptable, publicly accessible registry record with informative required fields; an early but inadequate submission is not automatically proof of full compliance. A careful summary states the dates and named rule rather than treating every “prospective” flag as the same judgment. These are reporting/editorial standards, not a determination of every jurisdiction’s legal deadline.[7][8]

Timeline	What can reasonably be said	What should not be inferred
Qualifying registration meets the named enrollment/consent timing rule	The documented timing meets that rule, subject to record adequacy	The protocol never changed or the trial was well conducted
Registration occurs after the policy’s enrollment/consent deadline but before analysis	The timing requirement was not met	A later public-posting date alone establishes late submission or proves misconduct
Record first appears after results are known	Prespecification is difficult to establish from that record	Every reported analysis is necessarily false
Date is unclear or inconsistent	Registration timing is unresolved	A favorable assumption about timing

A registry entry can be updated. The current text alone may not show what was visible before enrollment or analysis. Compare dated historical versions where accessible. ClinicalTrials.gov also distinguishes an estimated recruitment start from actual first enrollment; a planned date should not be silently substituted for a known participant event.[6] The trial and FDA status guide explains how to keep record fields and authorization questions separate.

Registration is not approval

A registration number identifies a record, not an approval of the intervention or confirmation that its result is scientifically correct. WHO’s ICTRP is a platform, not itself a registry; its portal links records supplied by registries.[4][8]

NLM describes a limited review of submitted information, and says results are checked for clarity and informativeness rather than scientific accuracy. The sponsor or investigator remains responsible for safety, science and accuracy. Quality-control review is therefore real, but it is not a scientific endorsement or finished-drug approval.[9]

Ethics review, registration, authorization to conduct research and marketing approval are separate. ICMJE explicitly says ethics-committee or institutional-review-board approval does not fulfill its registration requirement.[7] The approved-drug versus research-material guide explains the additional product-identity boundary.

The most accurate wording is factual: “The registry lists identifier X, first submitted or registered on date Y, with enrollment beginning on date Z.”

A protocol is the design record

The protocol specifies how the trial is intended to answer its research question. CONSORT 2025 asks for access to the full protocol and recommends protocol sign-off before participant allocation and data collection. This supports appraisal of selective reporting; availability alone does not prove adherence.[2] Useful elements include:

the primary objective and trial framework;
completely defined primary and secondary outcomes;
eligibility criteria and intervention groups;
randomization, masking, and control procedures;
sample-size assumptions;
follow-up and data-collection schedules; and
planned governance, monitoring, and amendment processes.

The phrase “according to protocol” is not enough. Check the version number and date. A protocol finalized after recruitment began cannot, by itself, establish what was planned before the trial started. Later versions remain valuable when their changes, timing, and reasons are visible.

The analysis plan should predate outcome-informed decisions

A statistical analysis plan, or SAP, expands the protocol’s statistical section. It may define analysis populations, effect measures, models, covariates, missing data and intercurrent events, multiplicity, interim analyses and sensitivity analyses.[2]

CONSORT 2025 recommends access to the full plan and says it should be signed off before the dataset is closed for analysis, allowing later changes to be documented transparently.[2] Timing needs closer inspection when anyone drafting or approving the plan had access to unmasked comparative results. “Prespecified” should refer to a dated decision made before the relevant data could influence it, not simply to text published before the journal article.

A later SAP alone cannot establish that a newly introduced outcome was part of the original plan. It may document a justified amendment, but its timing, rationale and relationship to earlier records still need explanation. No document automatically overrides every other source.[2][3]

Outcome switching is a comparison, not a keyword search

ClinicalTrials.gov defines a primary outcome measure as an outcome of greatest importance specified in the protocol, usually one used in the power calculation; it requires a title, description when needed, and a specific time frame.[6] Outcome identity can therefore change even when a short label looks similar.

CONSORT 2025 calls for four elements in each outcome definition.[10] Compare:

Variable: what was measured.
Metric: final value, change, time to event, response threshold, or another participant-level summary.
Aggregation: mean, median, proportion, or another specified group summary.
Time point: when or over what period it was assessed.

Also compare primary, secondary or exploratory status, analysis population, treatment-effect contrast and statistical method. A change from a mean to a responder proportion, or from week 24 to end of study, can alter the question even when the short outcome label is unchanged. Omitting an original outcome or promoting another without explanation can raise selective-reporting concerns.[3][10]

An amendment is not automatically misconduct

Trials sometimes require legitimate changes because of new external evidence, recruitment problems, lower-than-expected event rates, safety information, measurement problems, or a formally planned adaptive design. CONSORT 2025 asks reports to describe important post-commencement changes, including outcomes or analyses that were not prespecified, with their reasons.[3]

The appraisal question is not simply “Did anything change?” It is:

Question	Lower-concern pattern	Higher-concern pattern
When was the decision made?	Before relevant data access, or under a documented safeguarded adaptive process	Unplanned selection after comparative results were available
Was it documented?	Dated registry, protocol, or SAP version	Only apparent in the final article
Was a reason given?	Specific, plausible rationale	No rationale or a vague retrospective statement
Was the original outcome retained?	Original and amended analyses both visible	Original primary outcome omitted
Was multiplicity reconsidered?	Success rule and error control updated prospectively	A new favorable endpoint treated as confirmatory without adjustment

An independent monitoring committee may legitimately see unmasked data under its remit, and a prespecified adaptive design can permit data-dependent changes. That is different from an unplanned favorable endpoint selected after results were examined. Ask who had access, what safeguards applied and whether the adaptation was planned; access alone is not a misconduct finding.[3]

Transparent justification improves interpretability but does not, by itself, remove all bias concerns. Describe the documented discrepancy, chronology and uncertainty without diagnosing intent.

A practical document-comparison sequence

Find every registry identifier and bridge duplicate records for the same trial.
Record the first-submission, registration, first-posted, update, and first-enrollment dates without treating them as synonyms.
Obtain the earliest accessible protocol and available later versions; list missing versions.
Obtain the SAP and record its sign-off date and who could access comparative data.
Build a table of every primary and secondary outcome across the registry, protocol, SAP, results database, and article.
Compare outcome variable, metric, aggregation, time point, and hierarchy.
List amendments with dates and stated reasons.
Distinguish authors’ prespecification claims from independently verified dates; label documented amendments, post-hoc analyses and unresolved chronology.
Check whether omitted outcomes have results elsewhere, including the registry.
Keep interpretation proportional to the documented chronology.

When original plans are unavailable

An absent attachment is not proof that no protocol or SAP existed. Check the registry, lawful journal supplements and named public repositories, then say exactly what was obtained. A document’s upload date is not necessarily its authoring or sign-off date. Do not claim a complete plan review from an abstract, a current registry field, or one later version.[2]

Where chronology remains unresolved, use wording such as “the paper describes this as prespecified, but the original dated plan was not available for this review.” Continue reporting the actual estimate and uncertainty; neither assume misconduct nor silently certify prospective methods. The review-methods page describes the library’s evidence boundaries.

Important limitations

Registration improves transparency but does not validate submitted data or approve a study or intervention.
A registry record contains less detail than a full protocol and SAP.
Current registry fields may differ from what was visible when enrollment began; history must be checked.
Retrospective registration weakens evidence of prespecification but does not prove that every analysis was invented after results were known.
Protocol amendments can be scientifically or ethically justified and are not misconduct by definition.
Undocumented discrepancies can signal selective-reporting risk, but public documents may be incomplete.
Prespecification does not make an outcome clinically meaningful or an analysis unbiased.
Trial documentation cannot establish the identity or quality of a separately sourced material.

Frequently asked questions

Does a ClinicalTrials.gov number mean the study was approved?

No. It identifies a submitted registry record. Registration is distinct from ethics review, authorization to conduct research, and FDA approval of a named finished product and application.

What is prospective trial registration?

It is registration meeting the named policy’s timing rule. WHO’s portal compares registration and first-enrollment fields. ICMJE uses first submission at or before first consent, together with its registry-adequacy requirements. Report the dates and rule, not just the label.

Is a registry record the same as a protocol?

No. A registry supplies a public minimum data set. A full protocol normally gives substantially more detail about design, conduct, analysis, and governance.

What does a statistical analysis plan add?

It provides detailed analysis rules, such as populations, models, covariates, missing-data methods, multiplicity procedures, and sensitivity analyses, together with a dated basis for assessing prespecification.

Is every protocol amendment outcome switching?

No. Amendments can concern many aspects of a trial and may be legitimate. Outcome switching is a specific discrepancy involving outcome definition, hierarchy, timing, or reporting; its meaning depends on chronology, rationale, and data access.

Does changing a primary outcome prove misconduct?

No. A change can be justified. Readers should ask when it occurred, who had access to comparative data, how it was documented, why it was made, and whether the original outcome was reported.

Can registration prove a catalog material matches the studied intervention?

No. Trial documents describe the intervention used in that study. They do not authenticate or establish comparability with a separately sourced material sharing a name.

Primary and official sources

Review and corrections

Sources checked September 2, 2026 (America/Edmonton) / September 3, 2026 (UTC). Serum Research is responsible for this editorial guide. No independent clinical, statistical, scientific or regulatory review is claimed. This is not medical or legal advice, a trial-compliance determination, or a conclusion about any material’s quality or equivalence.

Browse the research library and study records, or contact us with a correction and its source.

Are you over 21?

Trial Registration, Protocols, Analysis Plans, and Outcome Switching