Case series imputation evaluates the aggregate case diagnostic evidence in a case series to create an argument supporting a potential relationship between a product and an AE. The concept of case series imputation, coming after and based on report-based signalling, is inherent in AE surveillance schemes, and has occasionally been noted by experts in AE surveillance methods (Venulet, 1984; Begaud, 1994). In recent years this distinction has been emphasized by Meyboom, who has pointed out the difference between using imputation to identify "interesting" adverse events, versus using it to review case series of suspected product-AE pairs (Meyboom et al., 1997b; Meyboom, 1998). Case series imputation methods are characterized by four main attributes: (1) use of a case definition that specifies which reports are to be included in the case series, (2) analytic orientation and time intensive-ness, (3) focus on summarization of evidence derived from (usually) multiple cases, and (4) use of pre-epidemiologic methodology to generate or strengthen hypotheses pertaining to specific pro-duct-AE relationships (see Tables 18.2 and 18.4).
Like all case series formation methods, case series imputation requires that the evaluator devise a case definition that is defensible from a medico-epidemiologic perspective (Clark et al., 1990; Blum et al., 1994; Meyboom et al., 1997b). Although accepting reports at face value may sometimes be reasonable, the importance of careful definition of cases is underlined by the relatively low rate of case validation frequently observed at the time of follow-up (Funch et al., 1993; Sicker et al.,
Table 18.4. Basic features of case series imputation instruments.
• Based on cases (i.e. reports that meet a case definition)
• Time intensive
• Represents a summarization of diagnostic causal evidence of all cases in a case series
• Uses pre-epidemiologic methodology to address product-AE relationships
1998). Following the application of a case definition to candidate reports, the resulting case series is examined for the presence of tests for causality, and the aggregate testing evidence is summarized as formally as the data permit. Case series imputation procedures often serve as the focal point around which a variety of relevant observations can become incorporated into a comprehensive safety evaluation, in particular the number of cases, data consistency, relationships pertaining to time and dose, biologic plausability, analogous observations involving other products, and the nature and quality of the data (Begaud et al., 1994; Meyboom, 1998).
From the time AE surveillance systems were first established, the most important safety-related actions that have resulted from health care product surveillance, including product withdrawals, have come from an examination of reported case series. Many of these analyses incorporated some type of description of case causal diagnostic features (In-man, 1993; McEwen, 1999). Reliance by safety monitors on the aggregate testing for causality contained in a case series has undoubtedly been influenced by the substantial under-reporting that is typical of passive (spontaneous and literature) AE reporting systems. In the presence of a non- or marginally quantitative signalling argument, aggregate diagnostic evidence from multiple cases offers a complementary rationale by which AE surveillance can be systematically performed. For example, the halothane-hepatitis experience demonstrated early on how case series containing instances of positive tests for causality (the time-to-onset patterns of repeat positive rechallenge) can justify important public health warnings, even in the absence of a clear-cut quantitative argument (Inman and Mushin, 1974). Despite such contributions, relatively few product-AE case series imputation methods (as opposed to single case imputation methods) have ever been published, and literature reviews devoted to this subject are virtually non-existent. One of the challenges facing researchers within the AE surveillance community is to develop a better understanding of how case series imputation has contributed to public health actions, and to incorporate these lessons into an improved and more expansive case series imputation methodology.
Case series imputation methods can be classified as: subjective, generalized rule-based, retrospective Bayesian, and AE-specific rule-based. Subjectively imputed case series (no written imputation assessment rules) have been used as long as case series of safety outcomes have been utilized for decision-making by the medical community (McEwen, 1999). In the 1980s, rule-based methods were first proposed as a basis for defining product-AE case series criteria (Food and Drug Administration, 1987), while, in the 1990s, the first Bayesian method for analyzing a product-AE case series was published (Naranjo et àl, 1990, 1990b). The slow rate of developmental evolution for imputation-oriented procedures probably derives to a great extent from the difficulties that are confronted when sensitivity and specificity data for tests for causality must be combined across multiple cases (Clark et al., 2000). At present, case series imputation continues to consist mostly of descriptions of the prevalence of causality testing in a case series using either descriptive statistical or non-statistical (narrative) means. Widely accepted methods for mathematically combining imputation-based evidence in favor of product-AE causality in a multiple product-AE case series have not yet been published.
Was this article helpful?