Summary

  • ClinicalTrials.gov launched on 29 February 2000; its early search expanded everyday terms and showed users the query it actually ran.
  • Recruitment defaults shaped the visible list, while later query links let other websites send visitors to an updated search.

A public database has a rather demanding idea of hospitality if visitors must learn its vocabulary before asking a question. ClinicalTrials.gov gave some of that homework to the search box. Type “heart attack”, and the early system also searched for “myocardial infarction”. The visitor could bring ordinary words. The machinery supplied more of the specialist ones.

The service opened on 29 February 2000. Its first collection contained more than 4,000 studies, primarily supported by NIH. The launch announcement and questions and answers described records containing study purpose, participation criteria, locations and recruiting contacts. Access was free, with no registration or password. A visitor could find a study's details and contact its team.

Making those records findable required another layer of work. Tony Tse's September 2000 operating guide described automatic synonyms and suggested spelling corrections. It also offered a way to examine the actual expanded query. The search engine could explain which words it had added. That mattered when the user's short phrase had become a longer instruction backstage.

The result list had its own opinion about usefulness. By default, searches retrieved studies recruiting or not yet recruiting participants. Completed, suspended, terminated and no-longer-recruiting studies could be included through a checkbox. Their records remained in the collection. The default selected a particular view of it, aimed at people looking for possible participation.

Two years later, the interface was still learning how people asked questions. An August 2002 bulletin described additions made that May: searching within an existing result set, editing query details and receiving alternative combinations of search terms. Its example of a full everyday question produced no initial results, then suggestions for reformulating the search. A sentence could be perfectly sensible to a person and still need help crossing the search box.

The same bulletin showed librarians and website developers how to encode a search in a link. A community website could send visitors directly to matching records. The query ran against the maintained collection when someone followed the link, saving the referring site from nursing a copied list as it aged. Public access was taking shape in small decisions: accept familiar words, expose the interpretation, make the filters visible and give the next website a useful doorway.