Subjective Case Series
With subjective case series an evaluator diagnos-tically interprets case causal information without formally specifying rules. As a result, the development of subjectively imputed case series is strongly influenced by an evaluator's individual medical and epidemiologic judgment. As with individual report imputation, subjective assessments of case series are probably the most common way in which associative evidence for product-AE relationships is summarized.
Rule-based case series are created when either minimum, uniform, or a range of causal diagnostic criteria are applied to candidate reports to create and analyze a case series. A traditional approach to this problem is to include cases using an explicit case definition, following which the cases are examined in aggregate for evidence for causality (Clark et al., 1990). A minimum, uniform approach can also be accomplished by applying a generalized rule-based procedure to reports and only accepting high level ratings that are defined by tests for causality as cases ("definite" and "probable", for example) so that imputational value is included in the case series at the time it is generated (George et al., 1998). These strategies establish per-case minimum standards upon which a subsequent summarization of product-AE causal data is based, regardless of the number of cases that are eventually included.
Two published generalized rule-based case series imputation methods that use both imputation and case number features to define case series are the monitored adverse reaction and the quality criteria grading method of Edwards (Food and Drug Administration, 1987; Edwards et al., 1990). The monitored adverse reaction procedure uses a combination of operational causality ratings (as determined by the FDA's rule-based imputation method) and case number to construct a case series with minimum imputation plus numerical features. The FDA's published rule-based method defines a probable case as the presence of positive dechal-lenge information, a possible case as the presence of "reasonable" temporal sequence, and a remote case as a temporal relationship of any nature (Jones, 1982; Turner, 1984). A monitored adverse reaction is then said to exist if a high-quality case series with any of the following attributes is noted: (1) at least one probable case, (2) at least two possible cases, (3) at least one possible and five remote cases, or (4) at least 10 remote cases (Food and Drug Administration, 1987). The quality criteria grading system of Edwards creates a rule-based case series by specifying imputation, numerical, and/or quality criteria (Edwards et al., 1990). An index case means that either positive rechallenge information is available or the case exhibits no confounding variables, a substantial case means that 11 key variables (which include the six key variables that define feasible cases below) are available for that case, and a feasible case means that six key variables are available for that case. A publishable signal is then said to exist if any combination of at least three index case equivalents are present, where one index case, two substantial cases, or four feasible cases constitute an equivalent. Both the monitored adverse reaction and the Edwards procedure allow the safety evaluator to compensate for decreasing numbers of cases by overweighting key cases with significant diagnostic value, and both are intended as starting points for a case series imputation analysis.
A case series approach based on Bayesian methodology has also been described that is referred to as retrospective Bayesian analysis (Naranjo et al., 1990a, 1990b). In the retrospective Bayesian method, a single estimate for the prior odds for the case series as a whole is estimated from the reported series. Following examination of the included cases, each case is then reassessed in the light of data from all cases, and a calculation of the posterior odds for the entire case series is obtained. A Bayesian-like solution has also been proposed that creates a modified standardized reporting ratio by weighting each possible causal test result sequence for the case series by the number of cases assigned to that stratum (Clark et al, 2000).
Rule-based, AE-specific (also called etiologic-diagnostic (Meyboom et al., 1997b)) case series are generated when AE-specific, rule-based methods, which may contain imputation criteria, are applied to candidate reports to generate a case series. Rule-based, AE-specific case series therefore include the concepts of both consensus-developed case definition and causal diagnostic value, and are thought to be well suited for defining and imputing case series data (Meyboom et al., 1997b; Meyboom, 1998). AE-specific case definitions have typically concentrated on the surveillance of AEs known to be frequent complications of health care product use, such as drug-
induced liver injury or hematologic cytopenias (Benichou, 1990; Benichou and Celigny, 1991; Danan and Benichou, 1993).
Was this article helpful?