Article type
Year
Abstract
Introduction: The Cochrane Breast Cancer Group established a specialised register of breast cancer trials in September 1996. The register is based on a thorough 60-line search strategy (consisting of various combinations of keywords and text words) aimed at identifying as many randomised trials as possible in breast cancer. The strategy was developed early in the register's history when it became evident that use of the MeSH term "Breast Neoplasms" was not optimal.
Objectives: To evaluate the adequacy of the MeSH term "Breast Neoplasms".
Methods: The search strategy for randomised trials in breast cancer was applied to the MEDLINE database from 1966 to 1997 (inclusive). Complete MEDLINE records were imported into Reference Manager software. Each reference was then broadly categorised according to disease (eg early breast cancer) and intervention (eg chemotherapy) based on information contained in the abstract. Studies obviously not in breast cancer were coded as such. The full manuscript was referred to if necessary. A 10% random sample of records was re-coded by a second coder.
Results: A total of 2264 references were identified. The MeSH term "Breast Neoplasms" had been applied to 1713, of which 1696 had been assigned correctly. Some of the 17 incorrectly coded references were benign breast conditions but most were in unrelated conditions. The search strategy identified 551 references not assigned the MeSH term "Breast Neoplasms". After excluding studies not in breast cancer (312 studies, mainly in breast feeding) and not randomised trials (16) 223 references remained. Of these 80 were trials in invasive breast cancer, 41 were of early detection methods, and 15 investigated prevention. There were 42 trials of interventions aimed not at treating the cancer, but at managing the side effects of treatment (eg there were 13 antiemetic studies). There did not appear to be any discernible pattern that could explain why trials had not been allocated the MeSH term "Breast Neoplasm" by MEDLINE indexers.
Discussion: When conducting a search of a literature database it is important to be aware of the limitations of keywords in the area of interest. If keywords exist, then the definition applied and the year they commenced usage should be checked. Searches being conducted to identify trials for systematic reviews should be as sensitive as possible to maximise the number of potentially eligible trials identified.
Objectives: To evaluate the adequacy of the MeSH term "Breast Neoplasms".
Methods: The search strategy for randomised trials in breast cancer was applied to the MEDLINE database from 1966 to 1997 (inclusive). Complete MEDLINE records were imported into Reference Manager software. Each reference was then broadly categorised according to disease (eg early breast cancer) and intervention (eg chemotherapy) based on information contained in the abstract. Studies obviously not in breast cancer were coded as such. The full manuscript was referred to if necessary. A 10% random sample of records was re-coded by a second coder.
Results: A total of 2264 references were identified. The MeSH term "Breast Neoplasms" had been applied to 1713, of which 1696 had been assigned correctly. Some of the 17 incorrectly coded references were benign breast conditions but most were in unrelated conditions. The search strategy identified 551 references not assigned the MeSH term "Breast Neoplasms". After excluding studies not in breast cancer (312 studies, mainly in breast feeding) and not randomised trials (16) 223 references remained. Of these 80 were trials in invasive breast cancer, 41 were of early detection methods, and 15 investigated prevention. There were 42 trials of interventions aimed not at treating the cancer, but at managing the side effects of treatment (eg there were 13 antiemetic studies). There did not appear to be any discernible pattern that could explain why trials had not been allocated the MeSH term "Breast Neoplasm" by MEDLINE indexers.
Discussion: When conducting a search of a literature database it is important to be aware of the limitations of keywords in the area of interest. If keywords exist, then the definition applied and the year they commenced usage should be checked. Searches being conducted to identify trials for systematic reviews should be as sensitive as possible to maximise the number of potentially eligible trials identified.