Screening imputation can be carried out on complete sets of AE reports, or can be made more efficient by the antecedent use of sorting measures that produce important AE report subsets for imputation screening review (e.g. the alert (15-day) or designated medical events sorting methods). In selecting imputation screening instruments for use in an AE surveillance database, the characteristics of a good screening test should be kept in mind (ease of use, generalizability, uniformity of result, and low false positive rate). The selected methods should be adaptable to systematic use with AE surveillance data models, and should allow the rapid insertion of an imputation rating into the records of key reports during typically defined workflows. In modifying existing instruments for this purpose, the safety professional will want to define a screening strategy that increases the operational ratings of reports that contain positive tests for causality. False positive rates can be kept low: by increasing the stringency of criteria needed to obtain a high rating, by periodically assessing the results of imputation screening, and adjusting the "set point'' of the screening imputation instrument as needed. While maintenance of a low false positive rate could increase the number of false negatives that remain undetected in the AE report database, it should be emphasized that imputation screening is only one of several strategies by which periodic AE database screening is performed, and that AE report database screening will be carried out on a repeat basis (Venulet, 1988). In the presence of a product-AE relationship, false negatives that remain undetected using imputation criteria will likely become detectable either through other non-imputation-based means of product-AE identification, or at the time of subsequent screening evaluations.
Likewise, case series imputation is only one of several possible case series formation strategies that can be used by the safety evaluator to summarize evidence in favor of a product-AE relationship. For example, the standardized reporting ratio, which compares the number of reported cases with the number of expected incident cases, is an important complementary numerical approach (Tubert et al., 1991) (see Chapter 19). In creating case definitions, the safety evaluator can use purely AE-associated criteria, or can require as well imputation-related criteria through the use of either generalized or AE-specific rule-based methods. The final construction of a case series may also involve numerical and report quality criteria. When summarizing aggregate tests for causality, it is useful to identify which tests were used, the number of patients to whom each was applied, and the evaluator's impression of test sensitivity and specificity. The results of positive rechallenge tests, tests that provide physical-chemical linkage with the monitored product, and clinical manifestations indicating spatially localized exposures followed by a localized effect will probably have the greatest discriminative ability, and would therefore warrant the greatest emphasis in final interpretation.
Was this article helpful?