Article type
Year
Abstract
Background: MeSH terms are designed to categorise articles into topic areas which allows us to search with some degree of specificity. In 1999, we analysed whether the MeSH term 'Breast Neoplasms' was correctly assigned to records in MEDLINE. We found that the MeSH term was accurately assigned to over 85% of the retrieved records. As our group’s search strategy has since been revised, the accuracy of the MeSH term 'Breast Neoplasms' needed to be reassessed.
Objectives: To evaluate the accuracy of the MeSH term 'Breast Neoplasms'.
Methods: We applied our search strategy to MEDLINE for trials published in 2009/10. The retrieved records were imported into Reference Manager 11 and coded by two authors. Records were categorised as whether they were related to breast cancer, indexed with the MeSH term 'Breast Neoplasms' and if the trial was randomised. Full manuscripts were retrieved, if necessary. Any uncertainty while coding was resolved through discussion.
Results: 2000 records were retrieved (87 were duplicates). 1473 records were tagged with the MeSH term 'Breast Neoplasms' of which 164 records were not relevant to breast cancer but on cancer in general (121) or incorrectly tagged (43). Of the 440 records without the MeSH term 'Breast Neoplasms', 36 were relevant to breast cancer and on reconstructive surgery, risk factors, diagnostics and endocrine therapies. There were 501 randomised trials on breast cancer, 21 of which were not assigned the 'Breast Neoplasms' MeSH term. These included trials on reconstructive surgery, mammographic density and drug therapies. This indicated that the MeSH term was relatively sensitive with minor (<5%) errors.
Conclusions: MeSH terms are used to identify relevant publications when searching databases. Our analysis shows that the term 'Breast Neoplasms' is not always accurately assigned. Systematic reviewers should be aware of the limited accuracy of MeSH terms when searching for studies.
Objectives: To evaluate the accuracy of the MeSH term 'Breast Neoplasms'.
Methods: We applied our search strategy to MEDLINE for trials published in 2009/10. The retrieved records were imported into Reference Manager 11 and coded by two authors. Records were categorised as whether they were related to breast cancer, indexed with the MeSH term 'Breast Neoplasms' and if the trial was randomised. Full manuscripts were retrieved, if necessary. Any uncertainty while coding was resolved through discussion.
Results: 2000 records were retrieved (87 were duplicates). 1473 records were tagged with the MeSH term 'Breast Neoplasms' of which 164 records were not relevant to breast cancer but on cancer in general (121) or incorrectly tagged (43). Of the 440 records without the MeSH term 'Breast Neoplasms', 36 were relevant to breast cancer and on reconstructive surgery, risk factors, diagnostics and endocrine therapies. There were 501 randomised trials on breast cancer, 21 of which were not assigned the 'Breast Neoplasms' MeSH term. These included trials on reconstructive surgery, mammographic density and drug therapies. This indicated that the MeSH term was relatively sensitive with minor (<5%) errors.
Conclusions: MeSH terms are used to identify relevant publications when searching databases. Our analysis shows that the term 'Breast Neoplasms' is not always accurately assigned. Systematic reviewers should be aware of the limited accuracy of MeSH terms when searching for studies.