on the background incidence of AEs, which is helpful in understanding the significance of findings in passive-surveillance systems. The disadvantages include missing data and misclassification of key data on outcomes (Hripcsak et al., 2003), drug use, or potential confounding factors. Information on severity of illness or functional status is often uniformly missing (Jackson et al., 2006) and selection bias cannot be prevented from influencing results. The large samples in administrative databases can provide considerable power to assess associations; however, precise but biased estimates of risk are not generally useful. Other disadvantages are difficulties in gaining access to primary medical records (and access to patients themselves), either entirely or on more than just a sample basis; dependence on diagnostic coding systems, which can be problematic for some conditions or topics; and drug-formulary restrictions in some health plans that limit the ability to study newer drugs if they are not on the formulary. Finally, much of the useful clinical information, such as descriptions of adverse reactions, exists only in narrative form, which makes automated analysis difficult (Jollis et al., 1993). There are strategies for correcting for some of the limitations of the databases (such as chart review to find missing data or to improve the accuracy of information), but they are sometimes resource-intensive. Consideration must be given to the strengths and limitations of the data in setting priorities within the program and between research methods for addressing a specific safety problem.
In some instances, active surveillance to generate safety signals and resolve other knowledge gaps is useful. Active surveillance is the regular, periodic collection of case reports from health care providers or facilities. CDER has been involved in developing a limited number of active-surveillance strategies. One example is an emergency room-based surveillance project for drug-induced injury, the National Electronic Injury Surveillance System– Cooperative Drug Adverse Event Surveillance System (NEISS–CADES), jointly funded by the Consumer Product Safety Commission, CDC, and FDA. FDA recently issued a request for information that stated its interest in this regard. In addition, FDA has cosponsored pilot development of a drug-based surveillance system that explores the feasibility of using data-mining techniques to identify safety signals in automated claims databases (DHHS, 2005). NEISS-CADES was used very recently to document AEs associated with stimulant medications used for ADHD (Cohen et al., 2006).
4.2: The committee recommends that in order to facilitate the formulation and testing of drug safety hypotheses, CDER (a) increase their intramural and extramural programs that access and study data from large automated healthcare databases and (b) include in these programs studies on drug utilization patterns and background incidence rates for adverse events of interest, and (c) develop and