In contrast to bibliographic databases, trial registries do not offer the option of formulating complex search queries, thus making targeted searches more difficult. However, ClinicalTrials.gov (CT.gov) and the International Clinical Trials Registry Platform (ICTRP) offer different search features that may help compensate this limitation. Our aim was to determine the importance of search features (for example, searches using synonyms or, additionally in CT.gov, automatic inclusion of further search fields) for trial registry searches.

We conducted a project called “Trial registry searches for studies of newly approved drugs” (1). One analysis investigated the question as to whether searches for different health conditions and interventions (new drugs) directly identified registry entries with the search terms entered or whether certain search features were responsible for this. We searched CT.gov and ICTRP for different conditions and interventions using the advanced search interface. For each search, we documented the synonyms listed in the two registries. We imported the registry entries into EndNote and evaluated whether the search terms used were available in the corresponding search fields (condition; intervention).

For CT.gov, 96 registry entries on 18 interventions and 190 entries on 12 conditions were analysed. Of these, twenty-three (24 percent) entries for interventions and thirty-eight (20 percent) for conditions were identified by search features, not by search terms. For ICTRP, 32 entries on 10 interventions and 100 entries on 9 conditions were analysed. Of these, five (16 percent) entries for interventions and eight (8 percent) for conditions were identified by search features.

Trial registry search features have an important impact on the sensitivity of searches. Many studies are not identified by the search terms entered, but by searches using synonyms and, additionally in CT.gov, by automatic inclusion of further search fields. Moreover, search features in CT.gov are more effective than in ICTRP – even though the same search terms are used, they consistently yield higher sensitivities.