Ideally, additional information would also be available about other patient characteristics (e.g., age, gender, severity of illness) that might affect the likelihood of the event. Adverse event databases suffer deficiencies in all these areas.
Further, because underreporting, incomplete reporting, and biased reporting are such severe problems with passive adverse event reporting systems, it is expected that reported instances of a particular device-associated event will typically be only a fraction of all instances of the event. Underreporting and the lack of comparative population information necessary to construct event rates are important reasons for the interest in surveillance based on large automated population databases such as those of big HMOs
For some devices, it may be possible to make estimates for some missing variables using information from the manufacturer, for example, device tracking registries or registries created as a condition of marketing approval. Other registries such as those created by professional groups or academic medical centers may likewise be useful in making estimates. FDA analysts may also seek data on adverse events from public health databases, as in the investigation of meningitis cases among recipients of cochlear implants cited above and discussed in Chapter 6.
In addition, data from premarket clinical studies may be evaluated. A case in point involves reports of subacute thromboses and possible hypersensitivity reactions following introduction of a drug-eluting stent. In October 2003, FDA issued a public health advisory on the topic. A month later it issued another notice stating that the agency’s review of pre-approval clinical trial data indicated that the rate of subacute thromboses was the expected rate for such stents and that many of the hypersensitivity reactions may be related to the drug therapy associated with the stenting procedure (Alonge, 2004).
Improved epidemiologic research capacity would help the agency tap alternative information sources. FDA funding restricts the time that analysts can spend both reviewing reports of serious adverse events and investigating those reports that are suggestive of a problem with a device or its use.
As described in several vignettes above, the most immediate response to an adverse event may be rescue interventions undertaken by health care professionals (or parents at home). In some cases, the event may prompt no investigation. In other cases, an affected institution may identify a problem with a device and take action to change or restrict the use of the device, perhaps without reporting the problem to either the manufacturer or FDA. Such isolated responses deprive other patients and physicians of potentially valuable safety information, although professional communication net-