Data

The two files this site is built from, and a description of every column.

Both files are plain comma separated text

They are UTF-8 text with a header row, and they are the files the build reads.

Every figure on the site is computed from these files at build time, so the files and the pages cannot disagree.

There is a machine readable copy of everything here

Nothing on this site is behind JavaScript, and every page has a markdown twin.

registry_et_trials.csv, one row per registered study

Thirteen columns, in this order.

Columns of registry_et_trials.csv.
ColumnMeaning
Columnnct_idMeaningClinicalTrials.gov registration number.
ColumntitleMeaningBrief title as recorded in the registry.
ColumnconditionsMeaningCondition strings as recorded in the registry, separated by semicolons where there is more than one.
ColumnstatusMeaningOverall status from the registry, for example COMPLETED, RECRUITING or TERMINATED.
ColumnphaseMeaningTrial phase, or NA where the study is not phased, for example a device or behavioural study.
Columnstart_dateMeaningStudy start date as recorded in the registry.
Columncompletion_dateMeaningStudy completion date as recorded in the registry. Some records give a year and month only.
Columnmonths_since_completionMeaningWhole months from the completion date to the data date of 18 September 2026. Blank where no completion date is recorded.
ColumnenrollmentMeaningNumber of participants as recorded in the registry, actual where available, otherwise the estimate. Blank where the registry gives none. Spelled as the registry spells it.
Columnsponsor_nameMeaningLead sponsor name, verbatim from the registry.
Columnsponsor_classMeaningSponsor class assigned by the registry, for example INDUSTRY, NIH, FED or OTHER_GOV.
Columnresults_postedMeaningyes where the registry record carries a results section, otherwise no.
Columnregistry_urlMeaningLink to the study record on ClinicalTrials.gov.

publication_status.csv, one row per completed trial without posted results

Sixteen columns, in this order.

Columns of publication_status.csv.
ColumnMeaning
Columnnct_idMeaningRegistration number, the key back to the registry file.
Columncompletion_dateMeaningCompletion date carried over from the registry record.
Columndue_12mMeaningyes where at least twelve months separate the completion date from the data date, no where less. Blank where the registry records no usable completion date.
Columnsponsor_classMeaningSponsor class carried over from the registry record.
ColumnenrollmentMeaningEnrolment carried over from the registry record. Blank where the registry gives none.
ColumnclassificationMeaningOutcome of the literature search: FOUND_RESULTS, MENTIONED_ONLY, NOT_FOUND or UNCERTAIN.
Columnmatch_basisMeaningHow the paper was tied to the trial: NCT_ID_MATCH, SPECIFICS_MATCH, or NONE where no paper was tied to it.
ColumnpmidMeaningPubMed identifier of the paper, where there is one.
ColumndoiMeaningDigital object identifier of the paper, where there is one.
Columnurl_openedMeaningThe record that was opened and read during the search.
Columnpub_titleMeaningTitle of the paper.
Columnpub_yearMeaningYear of publication.
ColumnjournalMeaningJournal, in its indexed abbreviation.
ColumnrationaleMeaningA short note recording why the trial was classified as it was, quoted on the tracker page.
Columnregistry_urlMeaningLink to the study record on ClinicalTrials.gov.
ColumnconditionsMeaningCondition strings carried over from the registry record. The file carries this column twice, with identical values.

Join the two files on nct_id

Every nct_id in publication_status.csv appears in registry_et_trials.csv. The 57 rows of the publication file are exactly the rows of the registry file where status is COMPLETED and results_posted is no. Join on nct_id to recover the title and sponsor.

Reproducing the figures takes one command

Clone the repository and run npm run build. There are no dependencies to install.

The build reads the two files, recomputes every number, writes the site into dist/, then recomputes the headline figures a second time and checks them against the built pages. If the two disagree, the build stops.

Data under CC BY 4.0. Code under the MIT licence. Both are in the repository at github.com/joooord/et-trials-tracker.