How do I check whether a trial's registered primary outcome matches the published outcome?

Compare the registered primary outcome's measure, timepoint, and analysis with the published primary outcome, then record any switch in an outcome-matching table that states which record you will cite.

ClinicalTrials.gov defines structured registration and results fields for study identification, status, design, eligibility, locations, outcomes, and reported results, so the registered outcome entry is the reference point for this comparison. Use the search interface to filter records by condition, intervention, location, status, study type, and dates to locate the correct registration.

Build the outcome-matching table

1. Retrieve the registration record and copy the primary outcome exactly as registered: the outcome measure, the timepoint, and any stated analysis description. Record the registration number and the date you accessed it.
2. Retrieve the published article and extract the primary outcome as reported: the measure, the timepoint, and the analysis actually used. Note the page or table where each element appears.
3. Place the registered and published versions side by side in one row per element and mark each row as matching, differing, or absent from one source. A difference in measure, timepoint, or analysis is a switch that the table should flag explicitly.
4. State in the table which record you will cite for the primary outcome and why, for example the registration when the published article reports a different measure or timepoint.

Use the table in bias assessment

RoB 2 assesses a specific randomized-trial result across bias domains using signalling questions and written justifications, producing low risk, some concerns, or high risk judgements. The outcome-matching table supplies the documented comparison that a bias assessment of that result can draw on, and it keeps the registered and published versions visible together.

Handle unresolved cases

When a registration record is missing, incomplete, or cannot be matched to the article, mark the element as unresolved in the table and record the next verification step, such as checking the registration history or contacting the authors. Treat an unmatched element as an open question rather than as evidence of a switch.