Introduction: Identifying all relevant randomised controlled trials (RCTs) is of paramount importance to systematic reviews. However, searching bibliographic databases for randomised controlled trials (RCTs) can be time consuming and expensive. Using multiple databases can result in many duplications and irrelevant citations.

Objective: Our aim was to investigate which of 3 commonly used databases was most efficient in identifying RCTs relevant to a new meta-analysis of chemotherapy in high grade glioma.

Methods: Cancerlit, Medline and Embase were accessed through DataStar (using Knight-Ridder Information) and were first searched using a simple strategy for identifying RCTs in glioma. This was compared with the results of the highly sensitive Cochrane strategy for RCTs. Finally, we looked at all the trials identified for the meta-analysis and where we found them.

Results:

Simple Strategy Cochrane Strategy
Citations RCTs Citations RCTs
Medline 116 72 144 17
Cancer Lit 115 65 152 18
Embase 61 21 107 4
Total 183 96 296 27

For the simple strategy, 6 of the 96 RCTs were in all three databases and 50 were in Medline and Cancerlit. For the Cochrane strategy, 1 of the 27 RCTs was in all three and 10 were in Medline and Cancerlit. 21 RCTs were eligible for the meta-analysis, 9 located by the simple strategy (3 in Medline alone, 6 in Medline & Cancerlit) compared to 3 (all in Medline & Cancerlit) using the Cochrane strategy. The remainder were found by hand searching. The Embase searches did not yield any additional RCTs to the Medline/Cancerlit searches.

Discussion: Where time and money are limited, Medline seems to be the most important database to search for RCTs in cancer. We feel that in this particular area, Embase is underproductive and expensive. These results also emphasise the importance of hand searching as the majority of RCTs were found this way and suggest that resource may be better spent hand searching than conducting supplementary electronic searching.