Recent concerns about the unexpected adverse effects of marketed drugs, such as COX-2 (cyclooxygenase-2) inhibitors or specific statins, raise concerns not only about reporting these events during premarket studies, but also about the responsibility for ongoing surveillance of drugs once they are on the market (Psaty and Furberg, 2005). Sometimes serious adverse drug reactions are fully appreciated only after a drug has been on the market for years (Lasser et al., 2002). Therefore, when a drug is approved and released to the market, large numbers of patients will be exposed before all the potential adverse effects have been identified and thoroughly studied. Currently, there is no clearly defined process for addressing safety questions about drugs after premarketing research has occurred (Ray and Stein, 2006).
The problem of adverse drug events (ADEs) is a significant and troubling issue. In a study of Medicare consumers, the incidence of ADEs was determined to be 50.1 per 1,000 person-years of treatment (Gurwitz et al., 2003). The classes of drugs most frequently associated with ADEs were cardiovascular agents, antibiotics, diuretics, nonopioid analgesics, and anticoagulants. Of the reported ADEs, 38.0 percent were categorized as serious, life-threatening, or fatal.
Pre- and postmarket assessments of drugs are important for helping to identify ADEs. Premarket review addresses issues of efficacy and safety of a particular drug, and postmarketing surveillance examines rare adverse reactions, effects that can be appreciated only with long-term use, and off-label uses of FDA-approved drugs.
As Forum member Jeffrey Drazen noted, “Adverse events associated with treatments represent a major hurdle that is very difficult to overcome in the setting of the randomized clinical trials that are done when a drug gets approved for marketing. There is almost no other way … to see how the drug functions in the real world.”
There are significant concerns about the process by which drugs are subject to assessment of their safety. While all drugs must undergo pre-market studies examining effectiveness and safety before they receive FDA approval, these tests are conducted on a limited number of subjects and often do not investigate long-term outcomes (Ray and Griffin, 1993). In addition, drugs are not always used in practice the same way they were used in the studies that led to their approval. For example, patients taking the medication may differ from the population in which it was studied; the treated population may not be as closely monitored as the patients in the clinical trials; or drugs may be prescribed for off-label use, the common practice in which physicians prescribe a medication for a use other than that for which it was tested and approved. In addition, during the initial period after the introduction of a new drug, the likelihood of inappropriate dosing, failure to follow directions, and contraindicated use is high (Smalley et al., 2000; Graham et al., 2001; Griffin et al., 2004).
A rare adverse event, one that occurs in fewer than 1 in 1,000 treated patients (for example, aplastic anemia), may not be identifiable in the premarket data and generally becomes apparent only after the drug is in wide use after its approval by the FDA (Okie, 2005). Other adverse events are first recognized through case reports in the literature, for example, the association of cisapride and torsades de pointes (Ray, 2003). Additionally, if the drug increases the rate of a common condition—for example, if it increases the risk for an adverse vascular event such as heart attack or stroke—this may be identifiable only when many thousands or even millions of people have used the drug.
Postmarketing surveillance can address some of these concerns by studying use by a larger group of patients over a longer period of time than in the premarket phase. Postmarketing research also provides the opportunity for a more comprehensive evaluation of drug utilization in the clinical practice environment.
A limitation in postmarket observational studies is the potential for findings to be due to systematic differences between patients in the treatment and comparison groups, rather than to the drug that is the subject of the study. There are statistical techniques that can be employed to control for confounding factors, but they are only effective when the relationships are known in advance and the data are available (Hunter, 2006). Controlling for confounding factors is also difficult in studies involving causes of mortality. The contribution of medication effects to the overall
death rates may be small because drugs may exhibit a complex combination of effects—some protective and some adverse (Ray, 2005). Another challenge is variation in the actual utilization of medications by patients that may not be captured in administrative data, including poor patient compliance, the use of drug samples given to patients, and interactions with over-the-counter medications and dietary supplements.
The FDA’s program for collecting postmarketing data is known as MedWatch. Information collected through MedWatch is placed in a large database known as the Adverse Event Reporting System (AERS). MedWatch collects ADE information from health-care professionals and consumers through a variety of channels, including mail, Internet, and telephone. The largest source of postmarket information on ADEs is drug companies themselves, which typically submit large numbers of ADE reports in batch form to the FDA.
AERS contains roughly 2.5 million health-care professional or lay-person-derived adverse event reports (Trontell, 2004).1 The FDA searches these reports for serious or life-threatening outcomes that result in death, lead to hospitalization or disability, or require interventions to prevent clinically significant impairment. A larger investigation may be triggered by these reports. The primary purpose of reports sent to AERS is to serve as a platform for hypothesis generation rather than hypothesis testing (Strom, 2004).
Because there are few scientific standards for data collection, the information in the database can be incomplete, and the results derived can be difficult to interpret (Greenhill et al., 2003; Psaty et al., 2004). Detection of signals is limited to the low percentage of actual events that are reported to the database (less than 10 percent), the quality of the reports, and uncertain estimates of exposure (Brewer and Colditz, 1999). Rare events or outcomes not selected as end points for clinical development may not be appreciated until their occurrence in a wider patient population triggers suspicion among health-care professionals. In the United States, clinicians are not required to report suspected drug- or device-related adverse events. Even with the partial reporting that occurs, more than 400,000 ADE reports are submitted to the FDA each year. This, combined with the large number of products on the market, places a large burden on the roughly 100 members of the FDA’s Office of Drug Safety to perform needed analyses of drug safety (Bennett et al., 2005).
In addition to AERS, the FDA requires drug companies to commit to conducting postmarketing studies for new drugs under the Accelerated Approval Program, the Animal Efficacy Rule, and the Pediatric Research
Equity Act. The Accelerated Approval Program allows premarketing studies to use surrogate end points to predict clinical benefit, provided that postmarketing studies confirm the benefit. The Animal Efficacy Rule requires postmarket studies when animal data replace human data for ethical reasons. The Pediatric Research Equity Act enables the FDA to require pediatric studies of drugs already approved for adults before the drugs are labeled for children. However, 60 percent of postmarket studies that are tracked by the FDA are “pending,” meaning that they have not commenced but have also not missed any deadlines (Phase IV Status Report, 2005).
The design and conduct of efficient and accurate postmarket drug surveillance studies must be considered a “cooperative venture between regulator, industry, and prescriber” (Wood, 1999). Some have called for the creation of a completely independent drug safety office (Psaty et al., 2004). The FDA has recently created an independent Drug Safety Oversight Board and a Drug Watch website to inform the public of adverse drug reactions (FDA, 2005a). Postmarketing surveillance data will still be submitted voluntarily, but the office will coordinate data with other government agencies such as the Centers for Medicare and Medicaid Services (CMS) and the Veterans Administration (VA). In addition to the formation of a new board, some also propose changes to the FDA’s existing approval methods, such as “raising its threshold” for the approval of new drugs when effective alternatives already exist (Lasser et al., 2002) or for drugs designed to treat conditions where treatment is more of a convenience than a need (for example, a runny nose or male pattern baldness). However, increasing requirements for drug approval will delay new treatments from reaching the market.
The Medicare Part D benefit, in which millions of elderly Americans will receive reimbursement for prescription drugs, represents an enormous opportunity for engaging in postmarket studies. Through this program, the CMS will collect information on medical and pharmaceutical utilization in a large population that uses a great number of pharmaceuticals and has a high proportion of chronically ill patients (Platt and Ommaya, 2005).
Understanding the risk profile of a new drug requires an integrated and robust system for evaluating the premarket and postmarket risks. A current challenge in our understanding is that patients, health-care providers, and others assume that the most serious adverse effects are identified in premarketing studies; therefore, identification of new risks after widespread use raises concern about system failures (Mitchell, 2003). The true picture is that all drugs will develop new information concerning risks and benefits as they are used in the marketplace. The challenge is
identifying and balancing potential or actual risk with benefit as information develops over time.
The occurrence of ADEs and the role of consumers and health-care professionals in reporting these events prompted the Forum on Drug Discovery, Development, and Translation to convene a workshop to explore issues associated with reporting of ADEs and to identify actions to improve the drug safety system. The workshop addressed the following questions:
How can ADEs be effectively identified, particularly when the adverse effects are rare?
How can the direct, causal effects of drugs be distinguished from simple associations?
How can health-care professionals and their patients aid in the identification of drug-related adverse events?
How can knowledge of ADEs be more effectively used in clinical practice?
The workshop took place on November 3 and 4, 2005, in Washington, D.C., and was chaired by Jeffrey Drazen, MD, Editor-in-Chief of the New England Journal of Medicine and Professor of Medicine at Harvard Medical School. The reader is referred to Appendixes A and B for the workshop agenda and speaker biographies. Section 2 reviews current sources of information on adverse drug events, including the FDA’s MedWatch program and the AERS, institutional review boards, and the CMS. Section 3 describes surveillance systems, surveillance technology, and data quality. Section 4 considers the ways that consumers and advocacy groups can be involved in reporting adverse events. Section 5 discusses drug interactions, problems with current databases for capturing and evaluating interactions, and difficulties in communicating information about adverse drug interactions. Section 6 describes new requirements for information contained on drug labels and how labels can be used to communicate information about risks and drug interactions to consumers and practitioners.