Objective: To compare the performance of MEDLINE and EMBASE for the identification of controlled clinical trials (CCTs) in rheumatoid arthritis (RA), osteoporosis (OP) and low back pain (LBP).

Methods: Two years were chosen for this purpose, 1988 and 1994. The search endorsed by the Cochrane Collaboration was used in MEDLINE. A comparable search was developed by a librarian for EMBASE. After the initial selection of potential trials, English papers were reviewed to confirm that they were CCTs and to assign a quality score to each trial (0 to 5, 5 highest quality).

Results: 2253 papers (81% English) were retrieved by both databases for RA, 978 for OP (80% English), and 880 for LBP (83% English).

Total MEDLINE EMBASE BOTH
RA - Total 2253 403 (18%) 1114 (49%) 736 (33%)
(English) (15%) (50%) (35%)
English CCTs 126 (12%) (20%) (68%)
OP - Total 978 198 (20%) 536 (55%) 244 (25%)
(English) (19%) (55%) (26%)
English CCTs 61 (18%) (30%) (52%)
LBP - Total 880 139 (16%) 517 (59%) 224 (26%)
(English) (3%) (67%) (30%)
English CCTs 51 (14%) (31%) (55%)

Less than 30% of all papers were retrieved by both databases. Overall, EMBASE retrieved more citations. Approximately 79% of the English citations retrieved were CCTs. Searching MEDLINE exclusively would have missed 20 to 31% of English CCTs. Mean quality scores of CCTs retrieved by a single database was similar to the mean score of trials retrieved by both.

Discussion: A comprehensive and systematic search should include both databases.