been available (Szarfman et al., 2002). Other systematic methods of screening for AEs have also received little attention, although their use appears to be increasing. The committee is aware of the criticisms of AERS, but the committee believes that the planned update known as AERS-2 will useful. The committee supports a focused improvement in how CDER uses passive-surveillance reports as a tool in drug safety research.
4.1: The committee recommends that in order to improve the generation of new safety signals and hypotheses, CDER (a) conduct a systematic, scientific review of the AERS system,4 (b) identify and implement changes in key factors5 that could lead to a more efficient system, and (c) systematically implement statistical-surveillance methods on a regular and routine basis for the automated generation of new safety signals.
The committee does not intend that review of AE reports, whether submitted by manufacturers as mandatory under federal regulation or submitted by patients or their providers through the MedWatch program, be the primary tool used by CDER for postmarket safety analysis. The committee does not support making AE reporting mandatory. Enforcing mandatory reporting is difficult and the committee’s goal is to have better reporting and better use of what is reported, not to increase the workload of CDER safety evaluators with unhelpful information. The passive reporting system in place today is capable of, and has made, important contributions, and the committee hopes that CDER will work to make the current system more efficient. In the next section, the committee offers recommendations for tools that will supplement and complement the AERS system and provide better data for regulatory and public health purposes.
The development and implementation of a lifecycle approach to the evaluation of the risks and benefits related to drugs will require expanded efforts in signal strengthening and signal testing in the postmarket setting. Once safety evaluators in ODS/OSE or clinical reviewers see sufficient numbers of similar case reports, they have to decide whether apparent signals are real—that is indicative of a problem—or just “noise” in the system. That determination should begin with the application of available tools, such as sector maps and empirical Bayes reporting ratios for analyzing spontane-