Identifying clinical trials in the medical literature with electronic databases: MEDLINE alone is not enough

The objective of this study was to compare the performance of MEDLINE and EMBASE for the identification of articles regarding controlled clinical trials (CCTs) published in English and related to selected topics: rheumatoid arthritis (RA), osteoporosis (OP), and low back pain (LBP). MEDLINE and EMBASE were searched for literature published in 1988 and 1994. The initial selection of papers was then reviewed to confirm that the articles were about CCTs and to assess the quality of the studies. Selected journals were also hand searched to identify CCTs not retrieved by either database. Overall, 4111 different references were retrieved (2253 for RA, 978 for OP, and 880 for LBP); 3418 (83%) of the papers were in English. EMBASE retrieved 78% more references than MEDLINE (2895 versus 1625). Overall, 1217 (30%) of the papers were retrieved by both databases. Two hundred forty-three papers were about CCTs. Two-thirds of these were retrieved by both databases, and one-third by only one. An additional 16 CCTs not retrieved by either database were identified through hand searching. Taking these into account, EMBASE retrieved 16% more CCTs than MEDLINE (220 versus 188); the EMBASE search identified 85% of the CCTs compared to 73% by MEDLINE. No significant differences were observed in the mean quality scores and sample size of the CCTs missed by MEDLINE compared to those missed by EMBASE. Our findings suggest that the use of MEDLINE alone to identify CCTs is inadequate. The use of two or more databases and hand searching of selected journals are needed to perform a comprehensive search.