Opportunities for Enhancing the FDA Guidance on Pharmacovigilance

The Food and Drug Administration (FDA) has produced a large number of extremely useful guidances for industry. These guidances include recommendations about the development, review, and approval of various disease-specific drugs, as well as general matters related to drug metabolism/drug interactions, premarketing risk assessment, pharmacogenetics, and pharmacovigilance. In May 2008, the FDA held a public workshop to aid in the development of a draft guidance about how to use large electronic health care databases for drug safety studies.¹ The draft guidance, which will be completed in fiscal year 2010 and finalized in 2011, aims ambitiously at being a major part of a national strategy to monitor medical product safety.²

Several textbooks and journals are devoted to pharmacoepidemiology, a highly complex field rooted in pharmacology and observational epidemiology. The effort to create a single guidance about so large a subject is therefore formidable. Most of the research experience with large electronic health care data sets has occurred in the setting of specific health care delivery systems such as health maintenance organizations or the Veterans Administration. The recent appearance of alternative models, based on personalized health records, will provide new challenges for the conduct of high-quality studies. With the development and implementation of this guidance, the FDA has an opportunity to improve public health and public trust.

Selected findings from preapproval studies are often the source of ideas for additional studies in the postmarket setting. In the preapproval evaluation of rosiglitazone for the treatment of diabetes, for instance, Misbin³ noted both the improvement in glycemic control and the adverse effects on weight and lipids. “Heart disease due to atherosclerosis is a major cause of morbidity and mortality in patients with type 2 diabetes, and it cannot be assumed that treatment with RSG [rosiglitazone] will decrease the risk. . . . A postmarketing study to address these issues needs to be a condition of approval.”³ Through a process of negotiation with the FDA, the sponsor agreed to conduct 2 large clinical trials.

In the ADOPT (A Diabetes Outcome Progression Trial) study,⁴ patients with new-onset diabetes were randomized to receive rosiglitazone, metformin, or glyburide as monotherapy, and the primary outcome was length of time before the first-line treatment failed to provide adequate glycemic control. Cardiovascular events were not included or evaluated as a primary or secondary outcome, although serious adverse events for heart failure were rereviewed retrospectively. In the DREAM (Diabetes Reduction Assessment with Ramipril and Rosiglitazone Medication) trial,⁵ patients with impaired glucose tolerance or impaired fasting glucose were randomized to receive rosiglitazone or placebo, and the primary outcome was the composite outcome of the development of diabetes, defined by a plasma glucose concentration greater than or equal to 7 mmol/L (126 mg/dL), or death. In this study of low-risk patients, the evaluation of cardiovascular events, a secondary outcome, was underpowered. Both trials targeted marketing issues and, at the same time, conspicuously avoided answering the cardiovascular risk-benefit question posed by Misbin.³ Before the RECORD (Rosiglitazone Evaluated for Cardiac Outcomes and Regulation of Glycemia in Diabetes) trial,⁶ which was required by the European regulatory authorities, could be completed, a meta-analysis of these trials and other small studies suggested that rosiglitazone unexpectedly increased rather than decreased the risk of myocardial infarction.⁷

Even if observational studies of intended beneficial drug effects such as the prevention of heart disease by antidiabetic treatments could be designed and conducted, their results would not generally be credible or convincing.⁸ Even the meta-analysis,⁷ which is itself an observational study of trial results, has been controversial. Large long-term well-designed randomized trials are essential for the evaluation of potential drug benefits. Whenever feasible, these trials should also address adverse effects. The industry-FDA negotiations that transformed a trial to evaluate cardiovascular risk into a trial that evaluated time to monotherapy failure⁴ represent a missed opportunity. The judicious use of well-designed phase 4 trials that address important public health questions, as the FDA is contemplating for antidiabetic therapy,⁹ should help to simplify the burden placed on pharmacovigilance.

Despite the conduct of phase 4 randomized trials, safety signals of uncertain significance will nonetheless arise from a variety of sources during the prolonged use of new drugs in large numbers of people, many of whom might not have met the limited eligibility criteria of the preapproval clinical trials. In contrast to the evaluation of intended beneficial effects, observational studies are often quite well positioned to evaluate the association between drug use and unintended effects.⁸ A key requirement for the conduct of high-quality observational studies is a genuine scientific curiosity that can be pursued without fear or favor. Scientists employed by many sponsors are well qualified in this regard. But the pharmaceutical industry's fiduciary duty to shareholders introduces a major conflict of interest. Marketing has often dominated science in the representation of research findings in the published literature.¹⁰ It is difficult to imagine how the current industry focus on short-term returns will permit sponsors to pursue aggressively and credibly the conduct of observational studies that may identify important harms associated with their drugs. Indeed, the attention to marketing rather than science, the concealment of safety findings, and payments to physicians and researchers have eroded public trust in the industry, the FDA, and the medical profession.

With the Reagan Udall Foundation,¹¹ the FDA Amendments Act has provided a potential mechanism for design, conduct, and funding of selected safety studies. When an important drug safety question arises, the FDA can put out a request for proposals through the foundation, which can receive funding from industry. A peer-review process can apply National Institutes of Health (NIH)–like criteria to the evaluation of the proposed studies. Are the right questions being asked? Are the design and methods optimal for addressing the primary hypotheses? Do the analysis plans match the aims and the methods? Does the study have adequate power? Have the investigators assembled a team that can effectively carry out the proposed project? Is the research setting and environment appropriate for conducting the proposed study?

The use of peer-reviewed safety studies conducted by independent scientists will vastly simplify not only the proposed guidance for industry but also the later effort to design and conduct pharmacovigilance studies. Under the Reagan-Udall Foundation model, the guidance can focus on general principles and avoid trying to specify rules for all the contingencies that may arise in safety studies. The specific details for each study are left to peer review. This approach also obviates the need for protracted negotiations between the FDA and sponsors about the details of the study design for every safety study. And by turning the conduct of the study over to an independent group, the FDA does not need to manage the potential conflict of interest that occurs when the sponsor implements, oversees, or contracts for its own safety studies.

These observational studies, because they are commissioned to address a particular safety signal or to resolve an area of uncertainty or controversy, should be registered and their results should be posted on a public Web site in a timely fashion. The protocol should also be publicly available so that when manuscripts are submitted to journals, editors and reviewers can evaluate whether the protocol was followed in the conduct and analysis of the study. This requirement, which is unusual for observational studies and may be counterproductive in the discovery phases of research,⁸ is important for pharmacoepidemiologic studies that are commissioned to address a regulatory problem. In other words, the provisions of the FDA Amendments Act for full registration and the posting of results and protocols of clinical trials¹¹ should apply to these observational studies as well. The reporting of these observational studies can also take advantage of recent guidelines.¹²

With several important exceptions,^{13 - 15} the NIH has largely turned the evaluation of drug treatments over to industry. At the same time, industry lacks a symmetric interest in safety and efficacy. In the absence of another independent national source for identifying the major treatment questions of public health importance, the FDA has an opportunity to take a lead role in identifying the key efficacy and safety questions that need to be asked and answered so that useful information about the full spectrum of risks and benefits of new drug therapies is acquired and disseminated in a timely fashion. The judicious use of randomized trials will help to ensure solid evidence about the benefits of drug treatment; and a combination of randomized-trial and observational-study evidence will provide the best information about safety questions. Finally, the use of peer review for selecting observational studies to be conducted by independent scientists would likely help restore trust in industry-funded studies, in academic researchers, and in the FDA.