Abstract
Approval of a novel drug is oftentimes seen as the end of the development pathway. However, the appearance of rare but serious side effects in patients taking approved drugs has led to increased attention to phase 4, or postmarketing, research. Traditionally, postmarketing research relied on reports from clinicians to monitor for unexpected toxicity. However, such reporting will produce a biased assessment of risk due to underreporting of toxic effects in older medications. The availability of large, representative databases and more flexible analysis tools has led to comprehensive and near “real-time” surveillance programs. These programs have been used by the US Food and Drug Administration to explore toxicities of approved medications. The need for effective tools with which to monitor clinically relevant outcome events has been further increased by the development of accelerated pathways to drug approval. In areas without effective treatments, such pathways lead to licensure without rigorous clinical efficacy data. Continued approval is frequently made contingent on the availability of postmarketing surveillance data demonstrating improvements in clinical end points and these data can also be used to monitor for unexpected toxicity.
Introduction
Regulatory approval of new drugs, and ongoing review of marketed drugs, is a complex process that has become more rigorous. Rigor has been increased as a result of a more keen focus on adverse events that might offset therapeutic benefit, increased awareness of clinical trial design and trial design flaws, and a focus on ensuring that drugs produce a clinical benefit to patients rather than simply improving a surrogate marker or outcome. The increased focus on safety, coupled with the interests of the pharmaceutical industry to obtain additional data beyond that obtained from phase 3 studies leading to an approved indication, has led to the increased use of studies that collect data after drug approval. In some cases, continuing regulatory approval may be contingent upon the collection and provision of data in the postapproval period.
Clinical trials research has traditionally been divided into 4 phases1,2 and these phases are oftentimes presented as if they are immutable and required for drugs to obtain approval. Depending on clinical circumstances, the nature of the disease being studied, and the characteristics of the medication, a traditional 4-phase research program may be inappropriate or impossible. In general, preclinical research is that research that does not involve humans and for which the characteristics of the product are determined both chemically and in a variety of animal models. Phase 1 research usually involves testing the product in healthy volunteers, often over a wide variety of dose ranges, to assess both for very frequent toxicities and to establish a dose range across which the desired therapeutic effect is likely to be seen.3 Phase 2 research is classically that within which a group of patients with the disease of interest are exposed to the drug, sometimes over a small range of doses, to determine whether there is any signal of therapeutic effect and to assess for less frequent toxicities. A variety of phase 2 designs have been used, including phase 2 dose escalation designs and active drug comparator designs. Most phase 2 studies will use response of a surrogate marker of efficacy, rather than a clinically relevant end point, as their primary outcome to determine whether there is likely to be a treatment effect at “tolerable” doses. At the end of phase 2, an estimate of the “best” dose is made to guide dosing in a phase 3 study. The phase 2 project is pivotal because it will determine whether a reduction in clinical events is seen—and such a reduction (with acceptable toxicity) is required for regulatory approval.
Traditionally, phase 4 research was performed to determine whether a product had unanticipated toxicity that occurred sufficiently infrequently that it was not detected in the pre-approval research program. These “pharmacovigilance” programs could be informal (eg, based on summation of individual case reports), formal but not specific to a product (eg, using the US Food and Drug Administration's [FDA's] Mini-Sentinel program that monitors a variety of data sources for toxicity but is collecting data on many products at the same time4 ), or formal and specific (eg, mandated by drug regulators as a condition of licensure or to monitor for a specific anticipated adverse effect). For example, the recent “accelerated” approval of pomalidomide is contingent on the completion and submission of an ongoing clinical trial powered to detect differences in clinically relevant outcomes5 (www.clinicaltrials.gov identifier NCT01734928).
Example of toxicity detected during postmarketing surveillance
Examples of drugs withdrawn as a result of toxicity detected during postmarketing surveillance are widely known. Review of a typical case is informative. Cerivastatin, a cholesterol-lowering agent, was withdrawn from the market as a result of reports of deaths due to rhabdomyolysis occurring at a frequency higher than that seen with other statin medications. Further, the frequency of toxicity appeared to be increased in patients receiving concomitant gemfibrozil.6 Although muscle injury is likely a class effect of statin medications, the increased frequency of this complication with cerivastatin only became apparent once the drug was used in the general population.
Limitations of traditional postmarketing surveillance
Traditional postmarketing surveillance has been criticized. Fontanarosa et al, in an editorial published in JAMA in 2004 identified a list of concerns with “informal” postmarketing surveillance. These concerns included: the unreliable nature of voluntary reporting of adverse events; the poor quality of reports submitted by frontline clinicians; the likelihood that only a small fraction of adverse events would be reported, resulting in difficulty in calculating rates of adverse events because neither the numerator nor denominator can be known; and the inability of reporting to establish causal relationships and to determine whether the drug or the disease caused the adverse event.7 Despite these limitations, postmarketing surveillance remains the only tool that will be able to detect very infrequent toxicity because studies of sufficient size to detect all clinically relevant side effects are infeasible, particularly for less common diseases. This suggests that improving the quality of postmarketing surveillance is required.
Innovative methods to capture adverse events
The scope of phase 4 research has expanded beyond simple systems (oftentimes based on informal reporting) that capture adverse events. Event detection and monitoring is now used by regulators to support novel approaches to drug review and approval; these pathways are designed to provide more rapid access to drugs in therapeutic areas where there are no effective therapies or where current therapies have significant limitations. The FDA has programs that take advantage of “postlicensure” monitoring to accelerate approval of drugs. In a broad sense, these programs allow approval of a drug based on either response of a surrogate outcome or preliminary evidence of effect on a clinical outcome in disease states for which there are either no or significantly limited alternative therapies. Approval is granted for marketing contingent on the approval and completion of studies designed to provide evidence that the preliminary findings are valid.
The first program is the “breakthrough therapies” program, which was designed to accelerate the availability of drugs that may “…[provide] substantial improvement on at least one clinically significant endpoint over available therapy.”8 An example of a drug achieving breakthrough status is daratumumab (a CD38 monoclonal antibody); this drug was accorded a breakthrough therapy designation from the FDA for the treatment of patients with multiple myeloma who have received at least 3 prior lines of therapy that have included a proteasome inhibitor and an immunomodulatory agent or who are double refractory to a proteasome and an immunomodulatory drug. This status was achieved in March 2012.9 A second program is the “accelerated approval” program, which allows approval based on observed changes in a surrogate end point felt to predict strongly a response in a clinically relevant end point.10,11 An example of an agent achieving accelerated status is pomalidomide, which is used for the treatment of patients with multiple myeloma demonstrating disease progression despite having received at least 2 prior therapies, including lenalidomide and bortezomib.5 These programs are designed to accelerate approval of innovative therapy; however, because approval is based on surrogate outcomes, they are dependent on their postmarketing research program to both confirm their efficacy and to identify and quantify toxicities associated with their use.
In selected circumstances, the risk to benefit ratio of a drug may not be apparent in all clinically relevant scenarios at the time of approval. To reduce this risk, the “Risk Evaluation and Mitigation Strategies (REMS)” program was created by the FDA to “ensure that the benefits of a drug outweigh the risks of the drug.”12 REMS may take a variety of forms, but all are designed to ensure that appropriate data on safety and efficacy have been collected and that this information is disseminated to health care practitioners.
A relatively recent, and powerful, method to increase both the quality and quantity of data collected after marketing, the Sentinel project, was initiated in 2008 as a national system to monitor FDA-regulated medical products. The project will use multisource and existing electronic health care data to monitor for complications of approved medical products.13 The initial phases of this project are known as Mini-Sentinel. The key features of the Mini-Sentinel project are: (1) active surveillance using ongoing assessment of a variety of electronic databases, (2) the ability to monitor prescribing changes in response to FDA regulatory actions, (3) the ability to respond rapidly to specific queries, and (4) public disclosure of results while maintaining privacy (through use of data without identifiers). As a pilot, Mini-Sentinel is designed to allow the FDA to pilot various strategies for precise data extraction and compilation.4 Using Mini-Sentinel data, investigators were able to query data from “real-world data” that is less prone to bias than spontaneously reported data.
Phase 4 research may also be carried out by pharmaceutical companies or investigators to provide additional safety or efficacy data in areas that were not fully explored in the phase 3 program or in novel areas that were not previously tested.
Case example of dabigatran etexilate
To further explore the successes and limitations of phase 4/postmarketing research, it is illustrative to examine the case of dabigatran etexilate, an oral prodrug that has been studied for the primary and secondary prevention of thromboembolism in a variety of patient types. The active molecule (dabigatran) is linked with etexilate, which allows absorption from the gastrointestinal tract. After absorption, the inactive etexilate molecule is cleaved off by the liver and the dabigatran molecule then acts as a direct acting anticoagulant through inhibition of thrombin. In an extensive phase 1 through 3 development program, its safety and efficacy profile was thoroughly examined in comparison with other anticoagulants.14-16 Based on the results of these studies, dabigatran etexilate has been approved for the prevention of stroke and systemic embolism in patients with atrial fibrillation in the United States.17 In other jurisdictions, current or pending additional approvals include prevention of thromboembolism in patients undergoing major orthopedic surgery and secondary prevention of venous thromboembolism.
Dabigatran is useful as a contemporary example of various types of postmarketing or phase 4 research. With respect to the traditional role of phase 4 research (exploring the safety of an intervention in a broad population), it is important to note that dabigatran is an anticoagulant. As such, patients taking this medication will experience bleeding. Neither dabigatran nor the other “novel anticoagulants” have an effective antidote to their anticoagulant effect. Therefore, major or life-threatening bleeding has been identified as a major concern with this agent given that its “usual” comparator, warfarin, has effective reversal strategies.18-20 Identification of bleeding led to reports in the lay press that dabigatran was causing excess bleeding, compounded by the lack of a reversal agent.21 In assessing whether dabigatran was associated with unexpected bleeding, the rates of bleeding with dabigatran need to be compared with those seen with warfarin (the “usual anticoagulant agent”). In informal reporting systems, rates of bleeding with dabigatran appeared higher than with warfarin; however, it was suspected this was due, at least in part, to significant underreporting of bleeding events in patients taking warfarin. This hypothesis was supported by the observation that rates of bleeding in large studies were similar between these 2 agents. However, the relative risk of bleeding was unknown in these drugs once they were being used in the unselected patients in whom risk factors for bleeding (such as renal insufficiency) would be more frequent than was seen in the studies that led to dabigatran licensure. Compounding this problem was the acknowledged and systematic underreporting of warfarin-associated bleeding and a widespread belief that warfarin's bleeding complications were easily treatable despite the lack of a truly effective warfarin antidote in the United States. To complete an analysis of bleeding with dabigatran, the FDA used the Mini-Sentinel database.22 This analysis demonstrated gastrointestinal bleeding at a rate of 1.6 per 100 000 patient days for dabigatran-treated patients compared with 3.5 per 100 000 patient days for warfarin-treated patients. Rates for intracranial hemorrhage were 0.9 and 1.9, respectively. Based on this analysis, the investigators concluded that: (1) bleeding rates with dabigatran were similar to those seen in clinical trials and (2) there was no evidence of excess bleeding in patients allocated to dabigatran.22 Further analyses are under way; however, based on this analysis, the FDA issued a statement that the FDA has not changed its recommendations with respect to dabigatran.23
The manufacturer of dabigatran and various investigator groups have also used the postmarketing period to undertake research exploring clinical issues with insufficient data provided in large studies and to explore new therapeutic areas. After approval of dabigatran, a randomized trial comparing warfarin with dabigatran for the prevention of stroke and systemic embolism in patients with mechanical heart valves was initiated (www.clinicaltrials.gov identifier NCT01452347). This study was stopped after 249 patients were enrolled due to an excess of strokes in patients allocated to dabigatran.23 Without this postmarketing study, it is possible that off-label use of dabigatran would have occurred and the toxicity signal would not have been detected.
Similarly, a prospective study of the use of dabigatran after noncardiac surgery in patients with unexpected troponin elevations is under way (www.clinicaltrials.gov identifier NCT01661101). This study randomizes such patients to a short course of dabigatran (110 mg twice daily) or placebo and will provide important information on the increased risk of myocardial infarction seen in dabigatran-treated patients in several large, active comparator studies.24,25
Finally, the manufacturers of dabigatran are undertaking the GLORIA AF registry (www.clinicaltrials.gov identifier NCT01468701 and others). This 48 000-patient prospective cohort study is designed to monitor “real-world patients” for rates of both expected and unexpected outcomes associated with anticoagulant treatment for recently diagnosed atrial fibrillation. This study should provide reliable data on the frequency of both common and uncommon complications in patients treated with anticoagulants for atrial fibrillation.
Summary
Phase 4 or “postmarketing” research has evolved from research performed to detect rare toxic side effects that are oftentimes gathered using spontaneous reporting and other unreliable detection methods into a much more rigorous research methodology. In some cases, such research may be used to support a licensing application in the setting of promising surrogate of clinical data from underpowered early phase research (thus providing patients with no other options access to such medications). Simultaneously, to provide more robust and reliable data, traditional postmarketing surveillance has evolved to use large datasets with obligatory and uniform reporting that can be queried in near real time to provide information from large numbers of patients on real-world efficacy and toxicity data.
Disclosures
Conflict-of-interest disclosure: The author is on the board of directors or an advisory committee for Sanofi-Aventis, Octapharma, Leo Pharma, Boehringer Ingelheim, Baxter, Asahi Kasai, and Viropharma; has received research funding from Sanofi-Aventis, Pfizer, and Leo Pharma; has consulted for Sanofi-Aventis, Pfizer, Octapharma, Merck, Leo Pharma, and Boehringer Ingelheim; has received honoraria from Sanofi-Aventis, Pfizer, and Leo Pharma; and has been affiliated with the speakers' bureau for Leo Pharma, CSL Behring, and Baxter. Off-label drug use: None disclosed.
Correspondence
Mark Crowther, St Joseph's Hospital, 50 Charlton Ave East, Room L208, Hamilton, ON L8N 4A6 Canada; Phone: 905-521-6024; Fax: 905-540-6568; e-mail: [email protected].