ClinicalTrials.gov Outcome Measures: Aligning Registry and Protocol Wording

For a trial subject to the registration requirements of 42 CFR Part 11, the primary and secondary outcome measures in its ClinicalTrials.gov record are required registration information, not an administrative summary to tidy up later. Each one must give the name of the specific measure, a description of the metric used to characterize it, and the time point(s) at which it is assessed (42 CFR 11.10(b)(19) and (20)). Write those three fields from the protocol's endpoint definition, not from the synopsis. For an applicable clinical trial that must be registered, submission is due no later than 21 calendar days after the first participant is enrolled (42 CFR 11.24). Later updates follow an at-least-annual baseline and specified shorter deadlines. One trigger matters directly for endpoints: when amendment changes are communicated to participants, the relevant registry elements must be updated within 30 calendar days of the IRB's approval of the amendment (42 CFR 11.64(a)(1)(ii)(O), for trials initiated on or after January 18, 2017).

The same wording is tested again at results submission, whose standard deadline is one year after the primary completion date, subject to the permitted delays and extensions in 42 CFR 11.44. Section 11.44(d) separately addresses later-collected secondary outcomes and adverse-event information, and 11.44(e) permits extensions for good cause. Results must include a data table for each primary and secondary outcome measure (42 CFR 11.48(a)(3)), plus the protocol and statistical analysis plan (SAP, if separate). The upload includes applicable IRB-approved amendments made before submission that apply to all trial facilities (42 CFR 11.48(a)(5)). In practice the record has to survive three comparisons: registry versus protocol, registry versus results, and registry versus publication. This guide covers the field-by-field wording, which update clock applies when an endpoint changes, what results quality control (QC) actually reviews, how journals use the record, and the enforcement record to date.

Two different regulations share the informal label "Part 11." FDA's 21 CFR Part 11 governs electronic records and signatures in regulated computerized systems, which our 21 CFR Part 11 evidence guide for eCOA covers. 42 CFR Part 11 is the HHS rule that implements the clinical trial registry and results requirements of section 402(j) of the Public Health Service Act, as expanded by section 801 of the Food and Drug Administration Amendments Act of 2007 (FDAAA 801). It controls what the public can see about your trial's design and endpoints, and it took effect on January 18, 2017. This guide addresses applicable clinical trials initiated on or after that date and subject to the cited registration and results duties. A ClinicalTrials.gov listing by itself does not determine whether a trial is an applicable clinical trial; other requirements, including NIH policy and journal policy, have their own scope. The ClinicalTrials.gov reporting requirements page is the starting point for separating those obligations.

For an applicable clinical trial initiated on or after January 18, 2017, 42 CFR 11.28(a)(2)(i) lists the required descriptive elements. Primary Outcome Measure Information is item (W) and Secondary Outcome Measure Information is item (X). The definitions in 42 CFR 11.10(b)(19) and (20) give every primary and secondary outcome measure the same three-part structure:

• Name of the specific outcome measure. What is measured, stated precisely enough to distinguish it from every other assessment in the trial.

• Description of the metric. How the measure is characterized: the instrument, the scoring, and how the reported value is calculated.

• Time point(s) at which the measurement is assessed. When the metric is assessed for each participant.

An objective alone is not an outcome definition. A synopsis sentence such as "to evaluate the effect of compound X on cognition" has no metric and no time point. Check the complete endpoint definition in the current protocol rather than trying to infer those details from that sentence. If you work from an ICH M11-structured protocol, use its endpoint and estimand text; see writing executable endpoint definitions under ICH M11. A synopsis that reproduces the full definition may help, but confirm it against the current protocol and SAP before transcribing the registry fields.

Choosing a category is a reporting decision, not just a label. The ClinicalTrials.gov data element definitions describe the primary outcome measure as the outcome "of greatest importance specified in the protocol, usually the one(s) used in the power calculation." A secondary outcome measure is of lesser importance but is "part of a pre-specified analysis plan" and "is not specified as an exploratory or other measure." Exploratory endpoints belong under Other Pre-specified Outcome Measures.

At results time, the rule requires a data table for each primary and secondary outcome measure (42 CFR 11.48(a)(3)). The results QC criteria expect all collected data for pre-specified primary and secondary measures, or a valid explanation; reporting other pre-specified and post-hoc measures is optional under those criteria. Match the registry classification to the protocol and pre-specified analysis plan. An inaccurate label in PRS does not redefine the endpoint or remove an applicable reporting duty. Calling a true secondary endpoint "other" can conceal an obligation rather than resolve it. Results QC criteria

In the Protocol Registration and Results System (PRS), the three regulatory elements become three fields. The Protocol Registration Data Element Definitions set the character limits. The Results Quality Control Review Criteria describe what reviewers expect when the same measures come back with data:

• Title (limit 254 characters). The name of the specific measure. The QC criteria ask for titles that are "outcome neutral" when possible, especially for change in a continuous variable: "Change in Systolic Blood Pressure," not "Decrease in Systolic Blood Pressure." Name what was measured, not the hoped-for result.

• Description (limit 999 characters). The metric, if the title does not already contain it. For scale, grading, or staging measures, the QC criteria require the range of possible scores, whether higher numbers indicate better or worse outcomes, and brief descriptions of any categories reported. A responder measure needs a brief, informative description of the criteria for "response."

• Time Frame (limit 254 characters). The data element definition says the time frame must be specific to the outcome measure and "is generally the specific duration of time over which each participant is assessed (not the overall duration of the study)." The results criteria add that it should be expressed from the participants' perspective, for example "from enrollment to end of treatment at 8 weeks" rather than "until end of study." Calendar dates and whole-study durations fail that test.

The definitions page lists these limits for primary outcome measures. They are system constraints, not regulatory limits, and can change with PRS releases. Two formatting rules from the QC criteria are easy to miss: acronyms are spelled out on first use in each of the Protocol and Results sections, and outcome measure titles do not end with a period. The table below maps each part of a protocol endpoint definition to the PRS fields, using an invented cognitive score solely to illustrate field structure.

Consider NEURO-301, a hypothetical Phase 2, randomized, double-blind, placebo-controlled trial of an investigational oral agent in mild cognitive impairment. Assume it is an applicable clinical trial initiated after January 18, 2017, and subject to these registration and results requirements. The trial, sponsor, dates, and both example scores are invented. No real instrument, validation evidence, or recommended endpoint is being described. Protocol version 1.0 defines the primary endpoint as follows:

The team registers before the 21-day deadline in 42 CFR 11.24. If it plans to publish in a journal that follows the ICMJE Recommendations, it registers earlier still, because ICMJE asks editors to require registration "at or before the time of first patient consent for enrollment." That is stricter than the federal deadline. The protocol endpoint becomes three fields:

• Title: Change From Baseline in Example Cognitive Score at Week 24 (58 characters; limit 254).

• Description: For this invented example only, the Example Cognitive Score ranges from 0 to 40; higher scores represent greater impairment. Change from baseline is the Week 24 score minus the baseline score, so negative changes represent improvement on this illustrative scale. (262 characters; limit 999).

• Time Frame: Baseline and Week 24 (20 characters; limit 254).

The entry is outcome-neutral, names the illustrative measure, states the score range and direction, and gives participant-level time points. The pre-specified analysis population is reported later in the results module's analysis population information (42 CFR 11.48(a)(3)(ii)), so it does not need to crowd the registered title. Whether a late Week 24 visit still counts is a protocol window question, not a registry one; see recording assessment windows. Later comparisons need the initial entry and the subsequent record history, rather than treating either version as the whole trial record.

The diagram shows the comparisons the registered outcome measure has to survive over the life of the trial. The dashed path is the amendment trigger discussed in the next section.

flowchart TD
    A["Protocol endpoint definition"] -->|Submit required registration within 21 days of first enrollment| B["Registry outcome measure:<br/>Title, Description, Time Frame"]
    A -->|Operational reconciliation before unblinding| C["Statistical analysis plan"]
    A -.->|Amendment communicated to participants| D{"Update within 30 days of IRB approval<br/>42 CFR 11.64(a)(1)(ii)(O)"}
    D --> B
    B -->|Standard results deadline: 1 year after primary completion| E["Results tables plus uploaded protocol and SAP<br/>42 CFR 11.48(a)(3) and (a)(5)"]
    C --> E
    E -->|NLM quality control review| F["Posted results and public record history"]
    F -->|Registry compared with manuscript| G["Journal review under ICMJE policy"]

Protocols change. 42 CFR 11.64 sets the clocks for reflecting those changes in the registry. The paragraph numbers below apply to trials initiated on or after January 18, 2017. Trials initiated earlier follow 11.64(a)(1)(i), which has a shorter list of update triggers.

• The general clock, 11.64(a)(1)(ii)(A). Changes to registration information must be updated "not less than once every 12 months."

• Record Verification Date, 11.64(a)(1)(ii)(N). Update it whenever the responsible party reviews the complete record for accuracy, and at least every 12 months even if nothing else changed.

• The amendment trigger, 11.64(a)(1)(ii)(O). The provision that decides most endpoint changes, quoted below.

• At results submission, 11.64(a)(1)(iii). Registration information must also be updated when results information is first submitted.

The amendment trigger turns on whether changes are communicated to participants, rather than simply on their perceived importance. If an endpoint amendment entails a revised consent form, new diary instructions, or an added visit communicated to participants, calendar the 30 days from IRB approval, rather than from the communication date. If this trigger does not apply, check whether another specific update or correction deadline applies before using the 12-month baseline. In particular, correcting an error is governed by 11.64(b), rather than being deferred to an annual review. As an operational control, coordinate the registry decision with any affected assessment build; eCOA change control after a protocol amendment covers that separate workflow.

Updating the registry does not amend the protocol, does not replace IRB or IND amendment obligations, and does not mean FDA has accepted the new endpoint. It is a disclosure duty that runs alongside those processes.

Corrections have their own clocks. When NLM sends electronic notice of apparent errors, deficiencies, or inconsistencies found in QC review, the responsible party must correct or address them within 15 calendar days for registration information and 25 calendar days for results information (11.64(b)(1)). The same 15- and 25-day limits apply to errors the responsible party finds on its own (11.64(b)(2)). Under 11.64(b)(3), complying with the QC process "does not constitute a legal defense to enforcement" under FDCA sections 301(jj) or 303(f)(3).

One more provision explains why a late update cannot be hidden. Under 11.64(a)(4)(ii), prior registration and results information is retained and made publicly available through ClinicalTrials.gov "so that updates do not result in the removal of any information from the original submission or any preceding update." Anyone can see what the record said before and when it changed.

Now let NEURO-301 go wrong. Every event below is hypothetical.

1. The amendment. Twelve months into enrollment, the hypothetical team adds a functional assessment. Protocol Amendment 2 makes change from baseline in Example Function Score at Week 24 a co-primary endpoint alongside the cognitive score. For this invented score, the range is 0 to 20 and higher values represent greater functional difficulty. The team must reconcile the co-primary decision with its analysis plan; see multiple endpoints and multiplicity. The central IRB approves the amendment and a revised consent form on March 15, and re-consent of enrolled participants begins.

2. The missed clock. Internal procedures treat ClinicalTrials.gov as an annual task, so the update is scheduled for the November record review. The 30-day deadline under 11.64(a)(1)(ii)(O), April 14, passes. The public record still lists a single primary outcome.

3. Results time. Assume no permitted delay or extension applies: results are due no later than one year after the primary completion date (11.44(a)). The biostatistics team analyzes both co-primary endpoints under the final SAP, built with the controls described in what a SAP must lock before unblinding.

4. The uploads. The submission must include a data table for each primary and secondary outcome measure (11.48(a)(3)) and the protocol and separate SAP with applicable IRB-approved amendments made before submission and applying to all trial facilities (11.48(a)(5)). Each uploaded document carries its version date on the cover page, and uploaded documents are posted publicly after QC review. A document version date is not necessarily its IRB approval date.

What follows is less dramatic than a "rejection," and more durable. According to the Results Quality Control Review Criteria, NLM's review is intended to help identify apparent errors, deficiencies, and inconsistencies. It "does not assess the appropriateness of the scientific design and analytic approach" and "cannot ensure that the information is 'truthful and non-misleading.'" Reviewers return Major Issues, which must be corrected or addressed, and Advisory Issues, which are suggestions. The NEURO-301 package is exposed in predictable places:

• Completeness. All collected data for pre-specified primary and secondary outcome measures must be reported, or a valid explanation given. Completeness and interpretability are separate checks: report both endpoints, and give each a self-contained title, description with score range and direction, and time frame consistent with the data table.

• Consistency. The QC criteria expect information to be consistent between data elements, and 11.64(a)(1)(iii) requires registration information to be updated when results are first submitted. The team can make the registration section list both co-primary endpoints at that point. It cannot change when that update happened: the record history shows the primary outcome changing months after March 15, and the posted amendment provides a dated version to compare. Establishing the legal timing failure also requires the actual IRB approval and participant-communication records.

• Scale information. If the hastily added Example Function Score description omits the score range or the direction of better and worse, that is a separate defect against the scale criterion.

QC review ends when every Major Issue is corrected or addressed, but posting does not establish legal compliance. In NEURO-301, the stated approval and communication facts make April 14 the update deadline. A later correction cannot turn that submission into a timely one. Public registry history and dated documents can expose a mismatch; evaluating the trigger requires the approval and communication evidence too. The operational repair is to assign an owner to the registry update when the amendment is approved, and retain the submission date with that decision. QC scope and limitations

The third comparison happens outside the regulatory system. The ICMJE Recommendations state that the purpose of registration includes preventing "selective publication and selective reporting of research outcomes." ICMJE accepts registration in ClinicalTrials.gov or a WHO primary registry that includes the minimum 24-item data set. It considers registrations inadequate if they are missing any of those fields or have fields that "contain uninformative information." It also recommends that journals publish the trial registration number at the end of the abstract, which makes the record easy for any reader to find.

This policy concerns differences between results posted in a registry and results reported in a paper. For changed endpoints, keep the amendment rationale, IRB approval date, and registry submission date available when preparing an explanation. Journal acceptance of a late registration is exceptional: ICMJE asks authors to explain the delay, and any editorial exception should be stated in the publication. It is not a substitute for meeting a regulatory deadline.

The meta-research shows why reviewers have reason to look:

• Discrepancies are common. Across 27 studies in a systematic review by Jones et al. (2015), the median proportion of trials with a discrepancy between the registered and published primary outcome was 31% (IQR 17% to 45%; range 0% to 100%). Among studies that compared prospectively registered outcomes with publications, the median was 41%. In studies that described the discrepancies, a median of 13% of trials introduced a new, unregistered outcome in the publication.

• Changed outcomes travel with larger reported effects. In a cross-sectional study of 389 trials with prospectively registered primary outcomes, Chen et al. (2019) found that 130 (33.4%) had at least one primary outcome change between registration and publication. Most of those changes (66 of 130) were registered primary outcomes that were not reported or were omitted. Trials with a change reported an intervention effect about 16% larger (pooled ratio of odds ratios 0.84, 95% CI 0.73 to 0.96). The design shows an association, not proof that outcome changes cause inflated effects.

• Many changes are visible only in the record history. Holst et al. (2023) analyzed the full registry history of 1,746 randomized trials completed at German university medical centers between 2009 and 2017. They found that 23% changed a primary outcome between trial start and the latest registry version, and 8% made a major change (adding, dropping, or reclassifying a primary outcome). In a sample of 292 trials, 41% of publications differed from the latest registry entry. Another 14% had changes visible only in earlier registry versions, and just 2 of 161 trials with a change reported it in the publication.

Citing earlier work, Holst and colleagues note that peer reviewers check for transparent reporting of protocol changes in only a minority of cases. So do not assume a journal will catch a discrepancy, and do not assume it will not. Adding Example Function Score as a co-primary endpoint, as NEURO-301 did, would count as a major change in Holst's classification. A timely registry update, a methods paragraph that states the amendment and its date, and a record history that matches both help readers assess the documented decision behind an outcome change.

Accuracy is a legal requirement, not a QC preference. Under 42 CFR 11.6, clinical trial information "shall not be false or misleading in any particular." 42 CFR 11.66 maps non-compliance to prohibited acts under section 301(jj) of the Federal Food, Drug, and Cosmetic Act (FDCA). These are failing to submit a required certification or knowingly submitting a false one (301(jj)(1)), failing to submit required clinical trial information (301(jj)(2)), and submitting information that is false or misleading in any particular (301(jj)(3)). How those provisions apply to a specific late or outdated entry is a question for regulatory counsel. The operational point is that the record history anyone would examine is public.

FDCA section 303(f)(3), codified at 21 U.S.C. 333(f)(3), provides a civil monetary penalty for violations adjudicated in a single proceeding and an additional daily penalty when a notified violation remains uncorrected after the 30-day period. This describes the statutory mechanism; it does not calculate an assessed penalty or current monetary exposure.

Keep that enforcement correction window separate from the registry-update clock. The 30 days following notification under the penalty provision do not extend the original registration, amendment-update, or results-submission deadline. The FDA notice page describes how corrective action is considered in its enforcement process.

For grant-funded trials, 42 CFR 11.66(c) adds a separate lever. When an applicable clinical trial is funded in whole or part by HHS, grant and progress reports must certify that required submissions were made. If that cannot be verified, remaining or future grant funding will not be released. After notice of non-compliance, the grantee has 30 days to correct it.

FDA's page on ClinicalTrials.gov Notices of Noncompliance and Civil Money Penalty Actions, current as of February 5, 2026, lists eight Notices of Noncompliance. They run from April 2021 (Acceleron Pharma) to August 2025 (an individual investigator), and include Mezzion Pharma (March 2025) and the FADOI Foundation (December 2024). No civil money penalty amount is listed for any of them. FDA describes a Pre-Notice for voluntary corrective action before a Notice. A Notice conveys FDA's determination that a responsible party failed to register, failed to submit required information, or knowingly submitted false or misleading information. Penalties may follow if adequate corrective action is not taken within 30 calendar days, and FDA notes that these violations could also lead to injunctions or criminal prosecution. The record is sparse and should not be read as a forecast in either direction; the absence of posted penalties does not mean none will be assessed.

One recent change to the rule is easy to over-read. A December 9, 2024 technical amendment (89 FR 97559) replaced the web address "https://prsinfo.clinicaltrials.gov" in the regulation with "https://clinicaltrials.gov or successor site," as the ClinicalTrials.gov reporting requirements page describes. It is not a source of new registration deadlines, outcome-measure definitions, or results duties.

Teams that want the registry, protocol, SAP, results, and manuscript to describe the same endpoint can build five controls into existing procedures:

1. Decompose each endpoint from the protocol body. For every primary and secondary endpoint, draft the Title (neutral, within 254 characters), Description (metric and calculation, plus instrument, score range, direction, and response criteria where relevant, within 999 characters), and Time Frame (participant-level time points, within 254 characters). Spell out acronyms on first use in each of the Protocol and Results sections. Register exploratory endpoints as Other Pre-specified Outcome Measures, not as secondary.

2. Add a registry decision to amendment change control. For every protocol amendment, record whether it changes a registered data element and whether the change is communicated to participants. If both, calendar the 30-day deadline from IRB approval under 11.64(a)(1)(ii)(O). If that trigger does not apply, check the other specific update and correction deadlines; use the 12-month baseline only where no shorter requirement applies. When the change came from patient input, documenting patient input that changes an endpoint covers the decision record behind it.

3. Reconcile the SAP with the registry before unblinding. Confirm that each registered primary and secondary outcome measure has a matching SAP analysis, and that planned results tables use the registered time points.

4. File the registry trail with the decision. Keep PRS release records, QC review comments and responses, and the submission date and a dated copy of the record after each update in the trial master file, next to the amendment approval each update implements. Our TMF Reference Model guide covers where decision evidence belongs.

5. Check the manuscript against the record history, not only the latest version. Before submission, compare each manuscript primary outcome with the registry history. Where they differ, the methods section should state what changed, why, the IRB approval date, and the registry update date.

This procedure has limits. Registry alignment is not protocol fidelity or scientific validity: a perfectly synchronized record can still describe a poorly chosen endpoint. In Holst and colleagues' data, minor changes to wording or detail outnumbered major ones, so not every registry revision is outcome switching. Discrepancy rates ranged from 0% to 100% across the studies Jones and colleagues reviewed, and all three studies draw on trials that are older, academic, or both, so no single rate generalizes. The regulatory text was checked against the eCFR rendering, which is authoritative but unofficial, and the U.S. Code citation uses an unofficial mirror. PRS field limits and QC criteria reflect ClinicalTrials.gov pages as of October 2026 and can change. The FDA enforcement table is current only to February 5, 2026. NEURO-301 and its example scores are wholly hypothetical; their ranges demonstrate registry wording and do not establish measurement validity or recommend an endpoint. This is regulatory-operations guidance, not legal advice.