Article type
Year
Abstract
Introduction:
Objectives: Systematic reviews addressing drug therapy often lack information about serious adverse drug reactions (ADRs). Such information is typically found in literature other than that pertinent to drug efficacy, and is difficult to locate. Our objective was to develop a generic search strategy for identifying reports of serious ADRs.
Methods: Serious ADRs were defined as those that are fatal, life threatening, permanent, or congenital (United States Food and Drug Administration definition). MEDLINE, EMBASE, and PyschLit were searched to identify serious ADRs for nine antidepressant agents. Initially, a very broad list of search terms, including text words such as "serious" and "adverse" were used. Terms specific for ADRs known to occur with the nine agents were not used. A single physician screened citations (titles and abstracts) from the broad search to identify those that were possible reports of serious ADRs. Full text of the possible reports of serious ADRs were reviewed to confirm true hits of actual serious ADR reports. The MeSH terms of true hits were used to develop various combinations of search terms. These were then tested to identify the search strategy that maximized sensitivity and minimized false positives.
Results: The initial broad search identified 12,374 citations. Of these, the physician reviewed a convenience sample of 3298 citations. Of these, 323 were verified as reports of serious ADRs. A combination of search terms that identified 93% of the 323 true hits was found. The search strategy was tested on the remaining 9076 citations. We found that individual terms were inadequate to ensure sensitivity of retrieval. Nine percent of true hits would have been found with the single textword "adverse", 86% with the subheading "AE" and 86% with the MeSH term "case report". The best combination of terms was "(Ae or co or po)fs. or Case report/ and human/". This strategy captured 99% of ADRs
Discussion: ADRs of drugs can be identified with high sensitivity and without seeking or preselecting specific ADRs. Further work is ongoing to test the generalizability of the strategy to other drugs.
Objectives: Systematic reviews addressing drug therapy often lack information about serious adverse drug reactions (ADRs). Such information is typically found in literature other than that pertinent to drug efficacy, and is difficult to locate. Our objective was to develop a generic search strategy for identifying reports of serious ADRs.
Methods: Serious ADRs were defined as those that are fatal, life threatening, permanent, or congenital (United States Food and Drug Administration definition). MEDLINE, EMBASE, and PyschLit were searched to identify serious ADRs for nine antidepressant agents. Initially, a very broad list of search terms, including text words such as "serious" and "adverse" were used. Terms specific for ADRs known to occur with the nine agents were not used. A single physician screened citations (titles and abstracts) from the broad search to identify those that were possible reports of serious ADRs. Full text of the possible reports of serious ADRs were reviewed to confirm true hits of actual serious ADR reports. The MeSH terms of true hits were used to develop various combinations of search terms. These were then tested to identify the search strategy that maximized sensitivity and minimized false positives.
Results: The initial broad search identified 12,374 citations. Of these, the physician reviewed a convenience sample of 3298 citations. Of these, 323 were verified as reports of serious ADRs. A combination of search terms that identified 93% of the 323 true hits was found. The search strategy was tested on the remaining 9076 citations. We found that individual terms were inadequate to ensure sensitivity of retrieval. Nine percent of true hits would have been found with the single textword "adverse", 86% with the subheading "AE" and 86% with the MeSH term "case report". The best combination of terms was "(Ae or co or po)fs. or Case report/ and human/". This strategy captured 99% of ADRs
Discussion: ADRs of drugs can be identified with high sensitivity and without seeking or preselecting specific ADRs. Further work is ongoing to test the generalizability of the strategy to other drugs.