Search strategies for retrieving complementary and alternative medicine clinical trials in oncology

Purpose: Several Medline search strategies exist to retrieve complementary and alternative medicine (CAM) literature related to oncology. The objective of this study was to compare different search methods to ascertain the most optimal strategy. Methods: All clinical trials with CAM interventions in patients with cancer, published from 1965 to 2004, were abstracted using 4 different approaches. In the CAM filter search, the PubMed complementary medicine filter was used. The Ovid search was performed using a complex search strategy with the Ovid search engine. CAM keyword and not phytogenic searches involved the CAM filter search with the addition of the search terms "AND (complementary OR alternative)" and "NOT (antineoplastic agents, phytogenic), respectively. Articles were evaluated by 3 reviewers to ascertain whether they were clinical trials, the study intervention was related to CAM, and the condition prevented/treated was cancer related (inclusion criteria). Results: The CAM filter search retrieved 10 718 citations, Ovid retrieved 1190, CAM keyword retrieved 2895, and not phytogenic retrieved 1806. Compared to the CAM filter search, all other methods had significantly lower sensitivity (Ovid 48.3% +/- 3.2%, CAM keys word 5.8% +/- 1.5%, and not phytogenic 77.9% +/- 2.7%, P <.001). The specificity of Ovid (38.4% +/- 2.8%) and not phytogenic (40.8% +/- 2.3%) searches was significantly higher (P <.001) compared to CAM filter (8.8% +/- 0.5%) and CAM keys word searches (1.9% +/- 0.5%). Conclusion: The search strategy using PubMed's complementary medicine filter, although comprehensive, lacks specificity; other methods, although more specific, lack sensitivity. Future indexing of all CAM clinical trials with a common medical subject heading term complementary medicine would enhance efficient retrieval of relevant citations.