Objective: Medications are frequently prescribed outside their approved indications (off-label), particularly when appropriate therapies are not available. However, the risk/benefit ratio of drugs in off-label use needs to be critically appraised because it may differ substantially from approved on-label usage. Therefore, an extensive exploration of current evidence is well-advised. The objective of this study was to develop two search strategies that facilitate detection of off-label drug use documents in MEDLINE and EMBASE via OvidSP.
Study Design and Setting: We compiled a gold standard reference set of reports classified as "relevant" or "not relevant" to off-label drug use. Search queries, including search words and strings, were conceived based on a definition of off-label use of medications as well as text analysis of 500 randomly selected relevant documents. The selected terms were searched in MEDLINE (from 1948 to 2011) and EMBASE (from 1988 to 2011). In comparison with the gold standard, we determined sensitivity and precision of search queries and their combinations. We developed a sensitivity-maximizing, and a sensitivity- and precision-maximizing search strategy for each bibliographic database.
Results: Our gold standard set contained 4,067 relevant documents overall out of 6,785 records. The most sensitive single term was "off label*.af." in both MEDLINE (sensitivity 40.9%, precision 84.4%) and EMBASE (sensitivity: 77.5%, precision 88.1%). The highest sensitive search strategy in MEDLINE was achieved by combining 31 search queries (sensitivity 53.3%, precision 60.3%) and 36 search queries in EMBASE (sensitivity 94.0%, precision 69.5%). Two optimal sensitive and precise search strategies yielded a precision of 84.0% in MEDLINE and 87.4% in EMBASE at the expense of decreasing sensitivity to 49.0% and 89.4%, respectively.
Conclusions: We empirically developed two versions of optimized sensitive search strategies which can achieve reasonable performance for retrieving off-label drug use documents in OvidSP MEDLINE and OvidSP EMBASE.