The Value and Future of Pharmacovigilance in the United States

THE VALUE AND FUTURE OF PHARMACOVIGILANCE IN THE UNITED STATES

Pharmacovigilance is the cornerstone of postmarketing drug safety activities in the United States and will likely remain so for the foreseeable future. Nearly all postmarketing labeling changes related to drug toxic-ity are based on spontaneous case reports. The same holds true for drug withdrawals. Since 1980, there have been 22 major prescription drug withdrawals in the United States (Wysowski and Swartz, 2005). Of these, spontaneous case reports and their analysis were a critical informational component contributing to the withdrawal decision in 20. The two exceptions were encainide and flosequinan, where randomized clinical trials identified the increased mortality risk conferred by these approved drugs (Echt et al., 1991; Massie et al., 1993). This should not be a surprise because patients with cardiac arrhythmias under treatment of those arrhythmias will sometimes experience sudden death due to arrhythmias, and death is not infrequent among patients with congestive heart failure. In situ-ations where the underlying disease being treated and the ADR resulting from treatment are the same, only a well-conducted randomized trial will convincingly establish the drug–ADR association.

As mentioned earlier, data mining is emerging as a potential means of generating new safety signals using existing ADR databases (Lindquist et al., 2000; Almenoff et al., 2005). A variety of methods have been developed, each of which compares every poten-tial drug–ADR combination in the database for statis-tical evidence of a discrepancy from an ‘expected’ number derived from all case reports in the database. The hope, as yet unrealized, is that such screening will help pharmacovigilance practitioners to identify and respond to previously unrecognized safety problems, and to do so with a shorter lag time. The integra-tion of data mining into routine business practices is beginning to occur at a number of national and international pharmacovigilance centers. However, to data, data mining has not been shown prospectively to improve overall signal detection.

In another recent development, ‘active’ surveillance for ADRs has emerged as a potential comple-ment to the traditional so-called ‘passive’ surveil-lance afforded by voluntary case reports, such as those collected by national pharmacovigilance centers (Food and Drug Administration, 2005). The inten-tion here is to search prospectively and proactively for ADR signals. Suggested methods include setting-, drug- and outcome-based approaches. Setting-based active surveillance has been pilot tested in emergency rooms and blood banks (Bennett et al., 2000; Budnitz et al., 2005) while outcome-based surveillance has been applied to the problem of drug-induced acute liver failure (Ostapowicz et al., 2002; Larson et al., 2005). The principle of drug-based active surveillance was demonstrated recently in a study of the associ-ation between rotavirus vaccine and intussusception (Davis et al., 2005). These authors applied sophisti-cated statistical methods to automated, longitudinal claims data from a large healthcare organization and showed that it might be possible to detect impor-tant safety signals prospectively in real time. The use of longitudinal healthcare data for active surveillance requires much additional work but offers the prospect of a significant advance for pharmacovigilance.

While the utility of case reports is undeniable, there is much that might be done to improve and expand their value. Strategies to improve the level of reporting of serious ADRs need to be developed The proverb about ‘strength in numbers’ also applies to pharmacovigilance. A few reports may provide a sufficient basis upon which to modify a prod-uct’s label. However, important information regarding the magnitude and duration of risk as well as risk factors for ADR occurrence is more easily and reli-ably discovered through careful analysis of a larger series of cases. Hand in hand with the value of a larger number of serious case reports is improved quality and completeness of those reports. The more clinically detailed a series of reports is, the greater the range of analytic possibilities. The value of this for regulatory decision-making and risk management efforts cannot be overstated. How to achieve these goals in an envi-ronment of immense time constraints and litigation fear is an important challenge for the future.

Another area of potentially great public health value is the expansion of current pharmacovigilance prac-tice to include other venues and types of ADRs. In the United States, the focus of pharmacovigilance has been on the rapid identification of serious unlabeled events. Many, if not most of these, fall into the cate-gory of ‘unexpected’ or ‘idiosyncratic’ and have been referred to as type B reactions (Meyboom et al., 1997). This is an important endeavor but from a population perspective, the bulk of drug-related morbidity and mortality is because of type A reactions, i.e. those that represent an extension of the drug’s pharmacol-ogy. The problem is great enough to represent one of the leading causes of mortality in the United States (Lazarou, Pomeranz and Corey, 1998). Pharmacovig-ilance strategies in this arena might lead to the iden-tification of ‘problem areas’ and provide the basis for more effective intervention and prevention.

Finally, advances in technology in the 1990s and the advent of the ICH process have created an environ-ment where global pharmacovigilance is now conceiv-able. A remaining challenge is to make this a reality.