Clinical Outcome Instruments: From Registry Names to CDISC QRS and Study-Build Decisions

A name match starts the specification work

A named outcome instrument can help a study team find terminology, representation guidance and relevant measurement evidence. Version, usage rights and fitness of the electronic implementation each need separate confirmation before an instrument becomes a controlled study-build specification.

Our independently recomputed analysis found at least one selected CDISC terminology name in the registered outcome text of 50,119 of 299,607 interventional studies, or 16.7%. Primary-outcome text matched in 18,579 studies, or 6.2%. Another 84,489 studies, or 28.2%, had generic assessment wording in at least one outcome title and no dictionary match anywhere in the scanned outcome text. These are results of a restricted phrase screen. They are not estimates of how many trials use standardized instruments, have unnamed endpoints or fail submission requirements. [1][2]

The practical opportunity lies in reuse with verification. Twenty names accounted for 68,910 of the analysis's 106,117 matched instrument-outcome-row pairs, or 64.9%. A team can prioritize those frequently encountered names for reusable specifications. It still needs to distinguish measurement evidence, permission to use the instrument, data representation and electronic implementation. Record the evidence for each question separately.

The report compares three evidence surfaces: public trial wording, CDISC's QRS catalogue and NIH's CDE repository. It then checks the FDA standards timetable and proposes four routes for resolving instrument requirements. All seven figures' values and sources are available in the chart-data workbook.

What the 299,607-study screen includes

The source was the 1 August 2026 AACT snapshot of ClinicalTrials.gov. The cohort included interventional studies with registered start years 2015 through 2026, with no restriction to recruitment status. It contained 237,497 actual, 44,339 estimated and 17,771 unspecified start-date types. In 2,959 records, the stated start date was later than the snapshot. The population therefore consists of registered studies assigned those start years, not 299,607 studies verified to have already begun. [1][3]

We searched 2,308,502 outcome rows against selected names from the 27 March 2026 NCI/CDISC terminology release. The dictionary had 1,501 patterns representing 531 selected terminology terms, including questionnaires, functional tests, clinical classifications and oncology response criteria. Several spellings or aliases can refer to one term, and many other clinical measurements fall outside the selected lists. [1][2]

The publication analysis uses complete outcome measures and descriptions, with the longest overlapping name preferred at the same text position. That prevents a mention of MDS-UPDRS from automatically contributing a second UPDRS hit, while allowing genuinely separate mentions in a row. Selected short acronyms remain case-sensitive, and ambiguous short expressions are excluded. Generic wording is identified in outcome titles with bounded expressions, including a word-boundary check around rating. [1]

Precision and recall against a human-labelled sample remain unmeasured. It can still miss lowercase abbreviations, differently expressed versions, versionless names and synonyms. For example, an entry that says RECIST v1.1 may behave differently from one using the selected exact name, while a versionless instrument title may fail a version-specific pattern. A negative result is consequently a discovery gap under this search method, not a finding that the protocol lacks an appropriate measure.

The analysis keeps three units separate:

There were 102,839 outcome rows with at least one match, or 4.5% of rows. That is different from the 16.7% study-level result. Neither percentage uses participants as its denominator. It would be incorrect to turn the pair count into a number of trials administering a scale, or to treat the difference between rows and pairs as duplication that should always be removed. [1]

Phase and disease patterns are discovery signals

The phase comparison shows where the selected dictionary is more or less visible. Phase 2 records had matches in 7,709 of 34,847 studies, or 22.1%; phase 3 records in 3,964 of 19,367, or 20.5%; phase 1 records in 2,738 of 27,133, or 10.1%. Those patterns reflect the measures and language represented by this dictionary as well as the mix of studies in each group. Study conduct would need to be assessed through other evidence. [1]

The “generic wording, no match” series needs particular care. It means that at least one outcome title used generic assessment wording and none of the scanned outcomes matched the dictionary. Other outcomes in the same study may be specific, and the protocol may supply further instrument details. Survival, laboratory measurements, imaging measures and many other outcomes appropriately use terms outside these selected lists.

The year plot is a cross-section of the August 2026 registry. It includes amendments made after a study's start year and, for 2026, estimated starts later in the year. The figure can help identify records for closer inspection. A claim about a policy effect or standards compliance would require additional evidence. [1][3]

MeSH-indexed disease groups show another form of concentration. The dictionary matched 10,833 of 26,665 mental-disorder-indexed studies, or 40.6%, and 12,573 of 38,531 nervous-system-indexed studies, or 32.6%. The Neoplasms group had 10,805 matches among 49,275 studies, or 21.9%. A study can belong to several groups, so these denominators overlap. [1]

The operational use is to prioritize investigation of instrument families in the team's actual portfolio. A neuroscience group may encounter a different mix of functional tests and clinician assessments from an oncology group handling tumour response. Estimate training, licensing and implementation effort from the instrument-specific requirements in each portfolio.

Where reuse is concentrated

RECIST 1.1 led the instrument-row ranking with 14,217 pairs, followed by EQ-5D-5L with 6,361, HADS with 5,279 and PHQ-9 with 5,199. The top 20 together accounted for 64.9% of all 106,117 pairs. This denominator includes the full matched list rather than an arbitrarily truncated top-120 total. [1]

A reusable library can begin with frequently encountered names, but popularity is not a licence or a fit-for-purpose assessment. A useful library entry identifies the exact instrument version, source materials, permitted modes and languages, scoring specification, representation guidance and the checks already performed. When a new study requests the entry, the team can compare those facts with its protocol rather than assuming that a familiar acronym means the same implementation.

The category distribution reinforces that point. The dictionary produced 56,539 questionnaire pairs, 21,345 oncology-response pairs, 15,519 functional-test pairs and 12,714 clinical-classification pairs. Those four groups sum to the complete pair count and describe the terminology lists used for matching. Dataset representation is a separate decision. [1][2]

In particular, the clinical-classification terminology category is not evidence of an SDTM “CC” domain. CDISC's QRS material and SDTMIG overview describe clinical classification and disease-response content in relation to RS; the actual representation depends on the assessment and applicable guidance. Other oncology data may require related structures. Assigning a domain by taking the first two letters of a dictionary category would skip the modelling decision. [4][5]

Even familiar names can carry unexpected catalogue details. MoCA sits in the functional-test category in the pinned terminology release used here. Administration, permission and endpoint definition remain separate specification fields. Keeping the source category visible helps data teams avoid silently rewriting the terminology to match their intuition. [2]

QRS tells you about representation and supplement status

CDISC's QRS programme covers questionnaires, ratings and scales, with supplements describing representation for particular assessments. Its public main table contained 324 catalogue entries when checked on 6 September 2026, corresponding to 305 distinct display names. Repeated instruments, versions and dataset-related entries account for the difference between catalogue rows and distinct names. [4]

The main-table statuses were 125 Granted, 103 Public Domain, 52 No Response Received, 23 Denied, 12 Author Permission Required, seven Exempt from Copyright and two without a specified status. We counted the main table only, without adding repeated entries from the separate update table. These are dated catalogue counts and may change.

The most important interpretive boundary is the status's subject. Denied or No Response Received concerns CDISC's permission process for developing a supplement. Sponsors should establish their own applicable administration, translation and electronic-use rights, alongside the instrument's measurement evidence. [4]

We also checked the literal top 44 terms in the corrected registry ranking against the public QRS table. Twenty-five had a published or author-obtainable supplement, seven had Denied or No Response Received status, and twelve had no matched main-table entry. That is a selected crosswalk, not a catalogue-wide availability rate. The crosswalk uses publicly visible version and alias metadata; licensed supplement content remains outside the review. [1][4]

The worked examples preserve both all-outcome and primary-outcome pair counts. DLQI, PCWG 2016, ESS, HAM-A, GDS and Wolchok 2009 each have nonzero primary-outcome matches. Usage counts and catalogue availability are separate fields, each populated from its own source.

For a build team, the next step after finding a supplement is to confirm that it corresponds to the requested instrument and version, determine its applicable representation, and compare that guidance with the study's capture and analysis requirements. A supplement can reduce interpretation work. The team still needs its own endpoint rationale, implementation checks and scoring verification.

NIH searches illustrate the difference between discovery and reuse

The NIH/NLM CDE repository provides another route to metadata and forms, but phrase-search results are highly sensitive to names and aliases. We reproduced the selected queries through the repository's documented public search API and inspected relevant record metadata. Quoted phrases were retained so an unquoted broad search would not silently produce a different comparison. [6]

The clearest example is the six-minute walk test. The phrase “6 Minute Walk” returned no forms or data elements. “Six minute walk” returned one form and eight CDEs. A directly inspected NHLBI record describes total distance walked in metres; the linked form is a broader physical-exam form, not a complete stand-alone instrument package. A zero for the first query was therefore a naming result, not proof that NIH had no relevant content. [6][7]

Counts are dated search results from this review, not distinct-instrument coverage percentages. The same caution applies to UPDRS: a search can retrieve both UPDRS and MDS-UPDRS forms. Several returned records are copyright placeholders with the form content withheld. Metadata discoverability is different from access to a reusable form or permission to administer it. [6][8]

The repository's contributor mix also affects what a search is likely to find. In the separately archived 22,308-element index, NINDS contributed 13,545 entries, or 60.7%. NCI contributed 92. This particular index has a strong neurological contributor base. The archived counts retain their own denominator, separately from live search totals. [6]

A study team should keep the query, retrieval date, record identifier, steward, instrument version and access restriction with any candidate it selects. It should then inspect the actual content and its relationship to the planned measurement. Assess the record's designation and applicability to the intended study use individually.

FDA's catalogue is a version window, not one mandatory number

The FDA Data Standards Catalog v11.0 lists a requirement-start date of 15 March 2025 for SDTMIG 3.4. It also retains SDTMIG 3.3 without a support or requirement end date. FDA's standardized-study-data guidance explains that where more than one version is supported, a sponsor may choose among eligible versions; it encourages the latest supported version. Version 3.3 remains an option within its applicable support window. [9][10]

The decision also depends on scope. The standardized-data requirements address specified submission types and centers, including applicable commercial INDs and marketing applications, with exemptions and qualifications described in the guidance. Applicability should be established for the intended submission. The FDA definition of clinical study start for this purpose is tied to the earliest informed consent, so the team should verify that event directly. [9][10]

A standards decision record can therefore be short and specific: identify the intended submission context, the relevant study-start evidence, the catalogue version checked, the supported versions considered and the version selected. Record the controlled-terminology release separately. This is a suggested way to retain the rationale, not a new FDA form.

Instrument suitability is another independent question. FDA's final October 2025 PFDD Guidance 3 addresses selecting, developing or modifying a fit-for-purpose clinical outcome assessment. The evidence should fit the concept of interest, target population and context of use. Catalogue presence and absence should both lead to the same context-specific assessment. [11]

Formal FDA COA qualification is optional. FDA describes qualification as a voluntary pathway for use across development programmes, and the COA Compendium is a non-exhaustive starting point rather than an endorsement of every listed assessment for every endpoint. A newly developed measure's endpoint role depends on its supporting evidence and the intended context. [12][13]

Four routes from an instrument request to a build specification

The following routes are an editorial work process derived from the distinctions above. Use them to assign the next evidence-gathering task, with scope and timing set by the study.

The routes can overlap. A study may find a QRS supplement and still need to settle translation rights or an electronic-mode change. Another may find no QRS entry but have a well-established instrument with an appropriate scientific rationale and licensor-supplied materials. Resolve the missing evidence while preserving the study's scientific requirements.

For an established instrument, separate the approved source material from the build representation. The source defines the content the study is entitled and intends to use; the implementation specifies how it appears, behaves and produces data. If the study changes wording, response options, recall period, mode or scoring, the team should assess the implications rather than describing the change as a cosmetic migration. FDA's fit-for-purpose framework makes the required evidence depend on the nature and extent of the change. [11]

For a generic request such as “quality-of-life score,” the immediate task is to clarify which concept and instrument the protocol intends. There may be several legitimate choices with different evidence and rights. The scientific team should settle the instrument choice before programming begins. If the protocol intentionally uses another kind of outcome, a questionnaire need not be introduced simply to create a match.

A study-build checklist that keeps the decisions connected

A reusable specification earns its value when another team can see what was decided and which assumptions still need confirmation. The checklist below is a proposed operating aid. Related items can share a controlled specification, with regulatory review remaining a separate process.

The distinction between scoring and representation deserves its own check. Score derivation and dataset structure need separate checks. Conversely, a correct score can lose meaning if the dataset does not identify the instrument version, timing or appropriate qualifiers. Reviewing both paths helps the team trace a result from the original response to the analysis variable.

Verification cases should come from the instrument and intended workflow. Depending on the applicable scoring rules, useful cases may include allowable response boundaries, missing items, reverse-scored items and a change in an answer before completion. These are examples for a team to consider, not universal test cases imposed by the sources. The expected result must come from the authoritative specification rather than the implementation being tested.

The same principle applies to an electronic migration. A visual review can catch wording, line breaks and response-layout problems. Measurement equivalence, licensing rights and score correctness require their corresponding evidence. The scientific, rights, interface and data questions should be assigned to the people able to answer them and linked through controlled versions.

EClinCloud describes scale management as a service area, with licensing due diligence, linguistic validation where applicable and delivery/change-management support. Its RM product is Resource Management, a different offering. The relevant service boundary should be stated accurately when discussing instrument work. [14]

Its EDC and eCOA offerings may support the resulting capture workflows. Instrument validation and usage permission need their own evidence. The specification, scientific rationale and applicable rights remain study-specific decisions. [15][16]

What to preserve when the analysis is refreshed

A future update should retain the registry snapshot, selected terminology release, search rules and full denominator. Report changes in the dictionary separately from changes in trial wording. An increase caused by a new alias should be attributed to the search change. If a QRS entry changes status, verify the particular version and access path before treating it as newly reusable.

NIH searches should retain the exact query and record identifiers, with follow-up for relevant aliases and version differences. A zero should remain a search result until further evidence supports a broader absence claim. Catalogue membership, formal qualification, licensing and build verification should continue to be recorded as separate fields.

The public evidence is useful because it makes repeated specification questions visible. It becomes misleading when a naming percentage is promoted into a compliance rate. For a study team, the useful endpoint is an instrument request whose identity, scientific role, rights, representation and implementation can each be reviewed and traced.

Sources

1. EClinCloud original analysis of CTTI AACT downloads, 1 August 2026 snapshot, independently recomputed 6 September 2026. Interventional studies with registered start years 2015-2026; full-description, longest-overlap phrase screen and bounded generic-word matching.

2. NCI Enterprise Vocabulary Services, CDISC SDTM Controlled Terminology, pinned 27 March 2026 release. Selected names from questionnaire, functional-test, clinical-classification and oncology-response lists; not a complete universe of measurements.

3. ClinicalTrials.gov, Protocol Registration Data Element Definitions, updated 24 April 2025, accessed 6 September 2026.

4. CDISC, QRS programme and catalogue, public main table and status descriptions accessed 6 September 2026. Licensed/authenticated supplement files were not reviewed.

5. CDISC, SDTMIG v3.3, published 20 November 2018. Public overview, including functional tests and expanded disease-response representation.

6. NIH/NLM, Common Data Elements Repository and documented public API, exact quoted-phrase searches checked 6 September 2026. Archived 22,308-element index used separately for contributor counts; live queries and archived totals are not combined.

7. NIH/NLM, NHLBI, Record the total distance walked during the six-minute walk test, inspected record, accessed 6 September 2026.

8. NIH/NLM, NINDS, UPDRS form record and MDS-UPDRS form record, metadata and copyright restrictions checked 6 September 2026.

9. U.S. FDA, Data Standards Catalog v11.0, 20 March 2025. XLSX version rows and footnotes directly inspected.

10. U.S. FDA, Providing Regulatory Submissions in Electronic Format: Standardized Study Data, final guidance, June 2021, particularly section II.E on applicable dates and version updates.

11. U.S. FDA, Patient-Focused Drug Development: Selecting, Developing, or Modifying Fit-for-Purpose Clinical Outcome Assessments, final Guidance 3, October 2025.

12. U.S. FDA, Clinical Outcome Assessment Qualification Program, current programme description, accessed 6 September 2026.

13. U.S. FDA, Clinical Outcome Assessment Compendium, scope and interpretation notice, accessed 6 September 2026.

14. EClinCloud, Scale Management service and Resource Management product, current pages checked 6 September 2026. Company descriptions, not independent outcome evidence.

15. EClinCloud, Electronic Data Capture, current product page, accessed 6 September 2026.

16. EClinCloud, Electronic Clinical Outcome Assessment, current product page, accessed 6 September 2026.