The Safety of Newly Approved Medicines: Title and subTitle BreakDo Recent Market Removals Mean There Is a Problem?

Between September 1997 and September 1998, 5 prescription medications were removed from the market because of unexpected adverse reactions. These actions raised questions about whether unsafe products were reaching the US market because the drug approval process had been expedited under the Prescription Drug User Fee Act of 1992.^{1 - 6} The US Food and Drug Administration's (FDA's) system for detecting postapproval toxicity includes the Adverse Event Reporting System, which gathers and analyzes spontaneous reports from companies, those sent directly to the FDA through the MedWatch program, and from ongoing clinical trials. When a problem is detected, follow-up analysis is performed. The FDA examined the events surrounding each of 5 recently withdrawn drugs, including an analysis of market withdrawals over the last 2 decades, to evaluate any possible relationship between the speed of review and the need to withdraw an approved drug. This article reports the review's findings and assesses the current rate of marketing withdrawals in the context of historical rates. Finally, the performance of the US system of ensuring drug safety is considered, taking into account premarketing clinical testing as well as postmarketing safety surveillance, especially as it may lead to a market withdrawal.

In September 1997, Wyeth-Ayerst Laboratories, St Davids, Pa, and Interneuron Pharmaceuticals, Lexington, Mass, removed from the market 2 diet drugs, fenfluramine (Pondimin) and dexfenfluramine (Redux), respectively, because of reports of cardiac valvular disease.⁷ Fenfluramine was first approved by the FDA in 1973 but was not extensively used until the early 1990s, when a combination of fenfluramine tablets and phentermine tablets, known as fen-phen, became extremely popular as a weight loss regimen.⁸ This was an off-label use of the drugs since the FDA had approved neither the combination program nor the long-term use of fenfluramine. Prior to knowledge of the valvulopathy, however, the FDA did approve dexfenfluramine for use as a single agent for up to 1 year.

In July 1997, Mayo Clinic researchers reported 24 cases of an unusual form of cardiac valvular disease characterized by specific valvular pathology and aortic and mitral regurgitation in patients taking fen-phen.^{9 - 10} Although the Mayo Clinic case reports did not clearly demonstrate a causal connection, the reports led the FDA to issue a public health advisory, initiate further epidemiological studies, and encourage echocardiographic prevalence studies of valvular disease in patients taking fenfluramine or dexfenfluramine alone or in combination with phentermine.

The initial, uncontrolled echocardiographic prevalence studies suggested that approximately 30% of people taking fen-phen had aortic and/or mitral regurgitation whereas only 1% of the general population exhibited these valve abnormalities.^{11 - 14} Neither the FDA, the manufacturers, nor the medical community had expected valvulopathy to be a toxic result of fenfluramine or dexfenfluramine. The increased magnitude and duration of exposure after the fen-phen regimen had become popular may have increased the prevalence, and thereby facilitated the detection, of valvulopathy.

After the link between fen-phen use and valve disease was brought to the FDA's attention by Mayo Clinic investigators, the early uncontrolled echocardiographic prevalence studies, coordinated by the agency's postmarket surveillance program, gathered sufficient data to establish the association, even though the pathophysiologic mechanism was unclear. These investigations led to the conclusion that the diet drugs needed to be removed from the market.

The third drug recently removed from the market is the prescription antihistamine terfenadine, which was marketed as Seldane. Approved in 1985, terfenadine, which is rapidly converted into the active metabolite fexofenadine, appeared to be a safe drug when used alone. However, accumulation of terfenadine associated with potentially fatal cardiac arrhythmias can occur when it is coadministered with an inhibitor of the hepatic cytochrome p450 enzyme system, such as erythromycin or ketoconazole.^{15 - 16} This interaction was not known at the time of its approval. After recognition of the potential interaction, the FDA performed critical laboratory investigations to explain and demonstrate the interaction conclusively. Physician education reduced concomitant use significantly.

Terfenadine's manufacturer, Hoechst Marion Roussel, Kansas City, Mo, developed a new product that contains only fexofenadine (Allegra), the active metabolite of terfenadine. Fexofenadine provides terfenadine's therapeutic effects and can be taken safely with other commonly prescribed medications.^{17 - 19} Therefore, the FDA proposed removing terfenadine from the market because an identical therapeutic molecular substitute without the serious risk profile existed. The manufacturers of both brand name and generic forms of terfenadine eventually agreed and removed their products from the market.²⁰

In June 1998, 2 drugs were withdrawn from the market after the manufacturers and the FDA discovered adverse events of a magnitude and seriousness not detected in the premarket clinical trials. One drug, mibefradil (Posicor), is a calcium-channel blocker approved for the treatment of patients with hypertension or chronic stable angina. Hoffman LaRoche Laboratories, Nutley, NJ, withdrew mibefradil because of potentially harmful interactions with an unmanageably large number of other drugs.²¹ Mibefradil itself did not cause toxic effects.²² Rather, mibefradil reduced the activity of several isoenzymes, most notably 3A4, in the cytochrome p450 enzyme system.^{23 - 25} This catabolic hepatic liver enzyme is important for the elimination of a variety of drugs from the body. In the presence of mibefradil, such coadministered drugs could accumulate to toxic levels. Mibefradil's inhibition of p450 isoenzymes and its interaction with certain drugs was known from both in vitro and clinical data prior to approval. Thus, mibefradil's original labeling described hepatic enzyme inhibition and listed the types of drugs with potentially harmful interactions. The FDA strengthened the labeling and issued a public warning in December 1997 after several patients experienced serious adverse events while taking mibefradil with other medications. The FDA also issued a public warning and the company distributed a letter to health care professionals.^{26 - 27}

With continued postmarket surveillance, the agency recognized that, unlike other strong cytochrome 3A4 inhibitors (eg, ketoconazole), mibefradil was being used commonly with interacting drugs. Because mibefradil is not a negative ionotrope, unlike certain other calcium channel blockers, the drug also was being studied in the treatment of patients with congestive heart failure. An unpublished international 3-year study of mibefradil in such patients, however, demonstrated no detectable clinical benefit over placebo (Hoffman-LaRoche, unpublished, May 1998). Because mibefradil could demonstrate no definite additional benefit over existing treatments and in light of the increasing number of drug interactions, the FDA concluded that the risks of the drug outweighed its benefits and should be removed from the market. Requesting withdrawal of a drug because it caused reactions to other medicines was an unprecedented step for the FDA. At the time of its withdrawal, mibefradil was known to interact with 26 different drugs, a number and diversity that could not, in practical terms, be addressed by standard labeling instructions or additional public warnings.

In June 1998, Wyeth-Ayerst Laboratories withdrew bromfenac sodium (Duract), a nonsteroidal anti-inflammatory drug, which had been approved in July 1997 for the short-term (generally 10 days or less) management of acute pain.^{28 - 29} The FDA had received 20 reports of serious adverse hepatotoxic effects. Four of these patients died of liver failure; 8 required liver transplants.

In the premarketing clinical trials, no patients developed overt hepatic failure, even in the 6- or 12-month exposure studies in patients with osteoarthritis and rheumatoid arthritis. Clinical trial findings clearly showed that some patients who received bromfenac developed elevated hepatic transaminases after a minimum of 35 days' use. Therefore, the FDA had required that the labeling emphasize the short-term use, warn of potential adverse hepatotoxic effects, and recommend monitoring hepatic enzymes if providers decided nonetheless to prescribe bromfenac for more than 4 weeks.

In February 1998, after receiving reports of liver failure in patients taking the drug for more than 1 month and in whom liver monitoring had not been performed, the FDA and Wyeth-Ayerst issued a letter to health care professionals and changed the labeling both to strengthen the warning of potential adverse hepatotoxic effects and to emphasize bromfenac's indication for short-term (less than 10 days) use.^{30 - 31} This labeling revision and the mailing to providers reduced the number of long-term prescriptions to 15% of the total prescriptions, but it did not eliminate use of bromfenac beyond 10 days. Given the availability of other analgesics with a wider margin of safety than bromfenac, the FDA believed that the risk from this product outweighed its benefits. In addition, the FDA and the company concluded that further attempts to restrict use of bromfenac to 10 days or less would not prove effective. Therefore, the company withdrew the product.

Each of these 5 cases presented the agency with dramatically different problems. Although they occurred at different times, they coincidentally reached a climax within the same 12-month period. There is no obvious thread linking these 5 drug withdrawals to any fundamental problem with the FDA review process itself, including increased review speeds.

Furthermore, the agency's postmarket surveillance system worked as designed for detecting the unexpected problems with mibefradil and bromfenac. For the diet drugs, the strongest indication of a problem came from Mayo Clinic case reports, but the FDA's postmarket system was able to gather enough additional cases to make a statistically significant linkage that led to market removal.

This cluster of withdrawals in a 12-month period led some to suggest that increased FDA review speed is to blame.^{1 - 6} They suggested that this increased speed of review could result in FDA staff missing key findings, which could increase the chance that a dangerous drug would reach the market.

In 1992, the US Congress passed the Prescription Drug User Fee Act, legislation that provides the FDA additional resources needed to expedite the new drug application (NDA) review process and improve its efficiency. With these resources, which come from user fees paid by pharmaceutical companies, the FDA hired additional drug reviewers and support staff, upgraded its computer infrastructure, and modernized its drug review management practices. As a result of the Prescription Drug User Fee Act, the review process for drug applications has accelerated dramatically.³² Currently, it takes, on average, about a year for the FDA to review a new drug. That is much faster than a decade ago, when it required nearly 3 years on average to complete the review process.

Since 1980, a total of 13 products have been withdrawn from the US market for safety reasons. Table 1 compares the review times for recently withdrawn drugs to those previously withdrawn and demonstrates that for the 5 drugs taken off the market in the recent 12-month period, the length of the FDA review was not abbreviated and 2 were approved well before the Prescription Drug User Fee Act was passed.

The shortest total time to approval among the 5 withdrawn drugs was for mibefradil, which took 15 months. (Total time to approval begins when the sponsor submits the initial NDA and ends at the approval date, including both the FDA review time and the time a company takes to answer questions and deficiencies posed by the agency.) The 2 other drugs reviewed under the Prescription Drug User Fee Act program, dexfenfluramine and bromfenac, were reviewed in 35 and 28 months, respectively. Compared with previously withdrawn drugs or with historical FDA review times, these particular products did not undergo notably fast reviews.

In addition, when each withdrawn pharmaceutical is examined with reference to the year in which that pharmaceutical was approved—rather than the year in which it was withdrawn— the pattern suggests no correlation between the Prescription Drug User Fee Act program and any hypothesized increase in drug withdrawals. Figure 1 demonstrates an apparent increase in the number of withdrawals in 1997 and 1998 when the drugs are analyzed by the year of withdrawal. But when the same data are displayed by the year in which the drug was approved, rather than the date of removal (Figure 2), no such clustering is apparent.

These data do not suggest a relationship between the time required to review an application properly and the likelihood of product withdrawals because of adverse events. These charts suggest no dramatic rise in the number of withdrawals per year, based on the year of approval.

It is true, however, that soon after implementation of the Prescription Drug User Fee Act, the number of NDAs submitted to the agency, which had been stable for decades, began to rise steeply, at a rate of about 12% a year. Not surprisingly, the number of NDAs approved also has risen (Figure 3).

With the steady increase in NDAs, there also has been a marked rise in applications for new molecular entities (NMEs). New molecular entities are chemically unique pharmaceuticals that have never before been marketed in the United States in any form. This rise in the number of newly approved drugs, especially NMEs, increases the number of products potentially subject to recall. Compared with the 1980s, we estimate the number of NMEs approved in the 1990s will increase by approximately 40%. This is significant because the first several years after the introduction of an NME is, historically, the time when unexpected and serious adverse events leading to market withdrawal are most likely to occur.

If evaluated by the year of approval, the rate at which NMEs are withdrawn has decreased steadily since the mid 1980s. Figure 4 shows the trend in product withdrawals by the year of NME approval, from 3.2% of NMEs marketed between 1979 and 1983 to the 1.2% withdrawal rate between 1994 and 1998.

It is possible that with longer follow-up other recently approved NMEs will need to be removed. This would increase the withdrawal rate for these latter cohorts. However, most withdrawals occur within a couple of years of approval, so many of the approvals since the passage of the Prescription Drug User Fee Act have been on the market long enough to accumulate new safety data. In any case, the available data do not support a higher rate of withdrawals of products reviewed and approved under the Prescription Drug User Fee Act.

The steady, if not declining, rate of withdrawals is particularly reassuring because the result of the improved review speed is that drugs are less likely to be approved in other countries first. In the past, if a drug was approved outside the United States and the approving country had an effective postmarketing safety system, rare, serious, unexpected adverse events would be discovered in that country once large numbers of people were exposed, the drug would be removed, and the US NDA would be withdrawn prior to approval. There are many examples of this. Sertindole was sold in 13 countries, including the United Kingdom, Germany, and Italy, for the treatment of schizophrenia for up to 29 months. These countries removed it from the market in December 1998, because it caused cardiac arrhythmias and sudden cardiac death. Prior to the market withdrawal, Abbott Laboratories, Abbott Park, Ill, withdrew its US NDA in January 1998 because of safety concerns. The drug sorivudine was under review in the United States as a treatment for herpes zoster but was removed from the Japanese market in 1993 after 1 month because of interactions with 5-fluorouracil. After 41 months on the market in about 13 countries, where it was approved while the United States was considering the NDA, remoxipride was withdrawn worldwide in 1994 because of an association with aplastic anemia. After 20 months on the market in some of the countries where it had been approved, nebacumab was dropped in 1993 during US review because the FDA-required studies showed excess mortality compared with placebo.

The present efficiency in the FDA review times, combined with a trend toward pharmaceutical firms submitting marketing applications simultaneously in several countries, has made these scenarios less likely. The number of NMEs marketed in the United States before anywhere else in the world has risen steadily over the last several years (Figure 5). In the early 1980s, only 2% to 3% of the new drugs were first commercially introduced in the United States. By 1998, more than 60% of the new drugs have appeared first on the US market.³³ It is conceivable that as the number of NMEs reaching the US marketplace before anywhere else in the world grows, there may be proportionately more US withdrawals.

In summary, the historical data show that a lower percentage of drugs approved in the 1990s have been withdrawn from the market than previously. Neither the increased speed of the FDA review process, nor the increased volume of the FDA-approved drugs reaching the market, appears to have adversely affected the overall safety of recently approved products as measured by drug withdrawals. Of course, assessing the safety of approved drugs requires a broader perspective than a simple tally of withdrawals. After all, it is the safety (and efficacy) of medicines that remain on the market that most profoundly affect people's health.

Health professionals understand that no pharmaceutical is completely safe. However, few may be aware of the overall impact of a drug's toxic effects on the health of a patient. Expected toxic effects from marketed drugs, even when used appropriately, is estimated to rank among the top 10 causes of death in the United States and is estimated to cost more than $30 billion annually.^{34 - 35} These findings lend special urgency to the goal of safer therapeutics.

In evaluating drugs for approval, the FDA uses a pragmatic standard: do the demonstrated benefits outweigh the known risks? This judgment takes into account situations for which much is at stake and severe risk can be accepted if benefit is present (eg, life-threatening illnesses), as well as cases for which only minor risks are acceptable (eg, over-the-counter drugs). However, as the above withdrawals illustrate, drug approvals are made on the basis of limited information and more is inevitably learned as a drug becomes widely used. The new information may alter the benefit-to-risk balance, sometimes precipitating withdrawal but usually leading to changes in the use of the product by health care professionals. For this reason, postmarket safety surveillance is an integral part of drug regulation.

Evaluation of a drug's safety is begun using animal and in vitro testing prior to clinical studies and continues as long as the drug is marketed. Many groups play a role in the process—pharmaceutical firms, health professionals, epidemiologists, and health care delivery systems—as well as the FDA staff. Some of the following critical steps in the process, and their weaknesses, have been highlighted in issues raised by the recent withdrawals.

Routinely, extensive safety testing is performed during premarketing clinical trials. The 1994 international drug safety standard^{36 - 37} for a chronically administered drug used long-term to treat a nonlife-threatening condition recommends that 1500 patients be exposed overall with 600 treated for 6 months and 300 for 1 year. This is considered adequate to detect an adverse event occurring in 1 of every 300 to 500 patients. Although many modern NDAs contain data on even more patients—usually several thousands of patients—current drug development programs are not large enough for reliable detection and recognition of very rare, serious events (Table 2).

For example, the incidence of death due to liver toxicity associated with bromfenac was below the practical detection limit of most clinical trial programs. Serious hepatotoxic effects occurred in approximately 1 in 20,000 patients who took the drug for longer than 10 days. To reliably detect this toxic effect, the NDA clinical databases would have had to include some 100,000 patients.

To deal with this uncertainty, the FDA reviewers search the clinical data set for markers, such as transaminase elevations, that might signal future hepatic problems. However, many drugs not associated with hepatic necrosis (eg, aspirin) cause enzyme elevations. Therefore, unless frank hyperbilirubinema in association with elevated liver enzymes is observed, reliable predictions of serious hepatic problems cannot be made. There is an ongoing need for better markers for rare, serious toxic effects.

Although the pharmaceutical industry considers US requirements for clinical testing generally stringent, a number of limitations deserve attention. Clinical trials are not real life. To assess efficacy in as unconfounded a manner as possible, trials sometimes exclude certain patients (eg, the elderly, the very young, those too sick, or those taking certain other medications). Any special vulnerability to adverse events in those groups will be missed. Although drug labeling can warn about the lack of information in such patients, a physician may nevertheless decide that treatment is in the patient's best interest and thus is warranted. Many other real-life factors may affect a drug's safety. Ultimately, it is not always possible to predict, based on protocol-driven clinical trials that are designed primarily to demonstrate whether the product is effective, how all physicians will actually use medications in practice.

When an NDA is submitted, FDA staff perform a detailed review of the toxicologic data as well as all clinical efficacy and safety data. The product safety profile is often an issue discussed with the FDA advisory committees, which consist of outside experts. If after review, the drug's demonstrated benefits are judged to outweigh its recognized risks, the FDA staff review the proposed labeling to help ensure that the data are communicated accurately and that instructions for use are adequate. Proposed drug promotion is also reviewed to see that there is a balanced presentation of risks and benefits. Postmarketing (phase 4) trials may be requested to further define risks, eg, to subgroups, with long-term use, or when used with other medications.

A drug that is tested in a few thousand people may be administered to several million within the first years of marketing. During this critical time, new information about the drug will accumulate. Pharmaceutical sponsors frequently conduct additional trials. Health professionals and consumers report suspected adverse reactions to the sponsor (which must then report to the FDA) or health care professionals may report directly to the FDA through its MedWatch system. The FDA enters the reported adverse event data into its Adverse Event Reporting System database. Serious events are individually scrutinized and epidemiological analyses are performed. Labeling changes, mailings to health professionals, and other information updates are provided to help manage the newly found risks.

The country's postmarketing surveillance system is becoming an increasingly crucial component of drug safety assurance. With scientific advances driving drug development, an ever-growing number of pharmaceuticals are available on the US market. Some of these pharmaceuticals—those intended for serious or life-threatening diseases that lack satisfactory alternatives—are approved with a much smaller safety data set than is traditional, with more information gathering deferred into the postmarket period. In these cases, too, the FDA and the community are willing to take greater safety risks due to the serious nature of the illness being treated.

The FDA currently receives more than 250,000 potential adverse reaction reports yearly from all sources. Over the last 3 years, the agency has implemented a new computer system designed to permit electronic submission of reports and to facilitate analysis and signal recognition. However, much more needs to be done to help ensure that providers and patients have the up-to-date complete information they need to prescribe and use medications as safely as possible. This will require continued efforts by health care professional organizations, academia, provider groups, and government groups in addition to the FDA.

Although the current system seems to be adequately protecting the public from harmful medications that should be withdrawn, it is possible that larger numbers of withdrawals will occur in the future. The number of prescription drugs on the market has increased dramatically over the last 2 decades, and Americans are using a greater number of pharmaceuticals than they have before. According to a recent report by the National Association of Chain Drug Stores, consumers are projected to fill 2.8 billion prescriptions in 1998, averaging more than 11 prescriptions for every person in the United States.³⁸ This trend toward greater medication use is likely to continue as the population ages. In addition to this rise in the use of prescription drugs, the US population consumes an increasing number of over-the-counter drugs and a growing number of dietary supplements, some of which may interact in unknown ways with FDA-approved pharmaceuticals. The increased rate at which drugs are entering the market, the higher consumption of medicines by the population, and the use of pharmacologically active "alternative medicines" all increase the probability of misprescribing and adverse reactions. Whether this results in drug withdrawals remains to be seen.

Existing evidence does not support the hypothesis that the recent number of market withdrawals is related to the improved speed of the FDA review and approval. Although a cluster of drug marketing withdrawals occurred recently, when the withdrawals are analyzed by year of approval rather than time of withdrawal, there is no indication that the improved speed in the FDA review process was a factor in the removals.

Although new medicines have not become less safe, adverse reactions to marketed drugs continue to be a serious problem worldwide. The aging US population, the availability of many new drugs, and changes in medication consumption patterns threaten to exacerbate problems such as misprescribing, medication errors, and undetected interactions. Given these realities, the FDA's postmarketing surveillance system will become even more active and essential in ensuring drug safety in the coming decade.