Sorting procedures are data management screens that create report sets of regulatory and/or epidemiologic interest. Historically, they have been considered a form of signalling (Begaud et al., 1994a), although their results are almost always re-examined through the application of additional techniques. Sorting methods are the basic data management tools of spontaneous signalling, and are probably the most common signalling procedures used in health care product surveillance. The careful inclusion of report sorting in a product-specific surveillance program is generally believed to be important in enhancing its effectiveness (European Agency for the Evaluation of Medicinal Products, 1997).
Qualitative Sorting (Key Content) Methods
Qualitative sorting by key content selects subsets of reports for further examination that contain content elements of interest from a regulatory, medical, or report quality perspective (Hartmann et al., 1997). Qualitative sorting allows an evalua-tor to break down a large reporting universe into smaller, more meaningful parcels, which can then be considered separately by safety professionals. Perhaps the most important qualitative sorting methodology is the well-known alert (or 15-day) report, which depends on the AE seriousness and local labelling data fields that are added to spontaneous reports by manufacturers (Venulet, 1988). Key content methods often focus as well on data fields believed to possess diagnostic value (e.g. positive dechallenge or rechallenge), important outcomes (Finney, 1971a; Venulet, 1988) (e.g. fatal, life-threatening), previously unreported AE types (Royall, 1971; Royall and Venulet, 1972), or AE types of high interest (Hartmann et al., 1997), such as those included in the Food and Drug Administration's (FDA) partial listing of designated medical events (Food and Drug Administration, 2000).
Quantitative Sorting Methods
The first published quantitative sorting method was based on a subjective impression of excessive report numbers (the "pigeon hole'' signal of Napke) (Napke, 1968). Although a formal evaluation of subjective quantitative signalling has not been published, the same limitations that are seen with subjective methods in general (i.e. increased intra- and inter-observer variability) would be expected to apply (see the section Imputation Screening (Causality) Assessments below).
Cutoff signalling methods refer to procedures that detect an arbitrary number of reports of a product-AE pair. While unrefined, cutoff criteria have long been regarded as useful mechanisms for directing resources (Finney, 1971a; Royall, 1971; Venulet, 1973; Hartmann et al., 1997; Lindquist et al., 1999). Simple approaches such as cutoff methods may be the only reasonable quantitative approach for drug-specific AE surveillance programs that accumulate few reports, since, in this setting, more elaborate procedures are often not feasible.
Was this article helpful?