Time to Act on Drug Safety

LAST SEPTEMBER, the diet drugs fenfluramine and dexfenfluramine were withdrawn after the Food and Drug Administration (FDA) reviewed echocardiographic studies indicating heart valve abnormalities in 31% of the patients tested.¹ While the FDA acted within days of receiving the heart valve data, fenfluramine had been approved 24 years earlier.² In December 1997, the popular nonsedating antihistamine terfenadine was withdrawn because a safer alternative existed without terfenadine's risk of potentially lethal cardiac arrhythmia.³ Terfenadine had been on the market for 12 years.

Discovering new dangers of drugs after marketing is common. Overall, 51% of approved drugs have serious adverse effects not detected prior to approval.⁴ Merely discovering adverse effects is not by itself sufficient to protect the public. Each year prescription drugs injure 1.5 million people so severely they require hospitalization and 100000 die, making prescription drugs a leading cause of death in the United States.⁵ At a time when the Congress recently passed legislation intended to make new drugs available more rapidly, it is important to reexamine the adequacy of the existing system to monitor and enhance the safety of approved drugs, both new and old. A more active and effective safety program for marketed drugs is essential to protect the public health.

While the FDA has more than 1400 employees with principal duties related to approving new drugs, a full-time staff of only 52 monitors the safety of approximately 5000 brand name, generic, and over-the-counter drugs already in the marketplace (Ralph Lillie, RPh, MPH, written communication, April 1998). This task is performed primarily in the Division of Pharmacovigilance and Epidemiology (DPE), which includes only 8 individuals with an MD degree and 1 with a PhD in epidemiology (Vincent Guinee, MD, MPH, written communication, May 1997). Some FDA staff assigned to evaluate new drugs also spend part of their time monitoring marketed drugs. In addition, the FDA currently has annual external research contracts totaling $2.4 million (Ralph Lillie, RPh, MPH, written communication, April 1998). Also, the FDA MedWatch Office, with 4 employees, is responsible for promoting the reporting of serious adverse reactions and product defects, not only for prescription and over-the-counter drugs but also for medical devices, biological products, and special nutritional products. During 1998, the FDA plans a reorganization. The DPE will get a new name—Office of Post-Marketing Drug Risk Assessment—and will advance 1 level on the FDA organizational chart. In addition, staff may increase to 65 positions, and a more capable computer system to facilitate analyzing adverse event data is expected to become operational (Murray Lumpkin, MD, written communication, January 1998). Although all FDA employees working in drug regulation have a role in drug safety, the FDA will continue to focus primarily on evaluating new drugs expeditiously, with only modest resources and staff to monitor the safety of drugs once they enter the marketplace.

A precise assessment of whether the existing safety monitoring program for marketed drugs is adequate requires considering the 4 specific tasks that are or should be involved in the DPE's broader mission of "protecting the American public from the health risks of marketed drugs." These tasks are as follows: (1) estimating the number and cause of serious injuries and deaths, (2) identifying new serious adverse reactions, (3) monitoring the effect of previous safety alerts, and (4) operating an early warning system. Each makes an important contribution to drug safety.

First, a rational program to monitor the risks of marketed drugs ought to begin with reliable annual estimates of deaths and serious injuries from prescription drugs and information about the likely causes. Without such data, it is impossible to determine whether serious injuries associated with prescription drug adverse effects are declining, or whether an epidemic of drug-induced injury may be occurring. In addition, it is difficult to know which drugs, or families of drugs, may require greater vigilance in their use or the development of safer new agents. This information is also essential to identify the priority targets for prevention efforts. Are preventable drug-related injuries occurring because of inadequate initial testing, a breakdown in postmarket surveillance, physician-prescribing errors, or a lack of patient compliance? Neither the DPE nor any other office at the FDA collects, analyzes, or publishes any such information on a regular basis. It is not routinely collected anywhere else in the US government or the larger medical research establishments. It makes no more sense to monitor drug safety without knowing the extent of serious injuries than to have a National Highway Transportation Safety Administration operating without information about automobile accidents, or a Federal Aviation Adminstration not knowing how many airplane crashes have occurred.

Most of the DPE's staff and resources are now devoted to a second important safety monitoring task—the identification of new serious adverse effects of recently marketed drugs. This effort is based primarily on spontaneous anecdotal reports. Reports of adverse reactions from consumers, physicians, pharmacists, nurses, and other health professionals are received both directly and through the manufacturers, who are required by law to forward information they receive. In 1994, the FDA received just 3863 (5.2%) of 73887 reports of adverse reactions directly from physicians,⁶ although additional physician reports were likely submitted to the manufacturers, and in some hospitals pharmacists submit adverse event reports. The reports are collected in a computer database. About two thirds of the DPE staff are assigned to data entry, verification, or review of these data. However, once the DPE staff identify a possible serious unlabeled effect, its powers are limited. The DPE staff may seek confirming evidence through data from external clinical databases. They report their findings to the FDA division that approved the drug. That division, in turn, may take action, question the findings, or request additional studies from the manufacturers or external researchers with cooperative agreements. The most common corrective action is a change in the product disclosure label or package insert.

The data from adverse reaction reporting make a valuable contribution to drug safety. For example, the DPE identified the cases of liver damage from troglitazone in 1997, the risk of seizures and drug dependency for tramadol in 1996, a 10-fold elevated risk of aplastic anemia for felbamate in 1993, and blood disorders from temafloxacin in 1992. If serious unlabeled effects are occurring for any prescribed drug and are reported, this information will become known rapidly without the delays inherent in designing, conducting, analyzing, and reporting a new clinical investigation.

A major weakness of spontaneous anecdotal reporting is that it is difficult or impossible to estimate reliably how often adverse events might be occurring since, according to FDA estimates, only about 1% of adverse events are ever reported.⁷ For example, toxic effects of digoxin, including a particularly serious arrhythmia, are well documented. The average of 82 adverse reaction reports received by the FDA each year for digoxin suggests that this known risk does not pose a problem. However, a systematic study of Medicare records disclosed 202211 hospitalizations for digoxin adverse effects in a 7-year period.⁸ Also, the spontaneous reporting of the same adverse effect of the same drug with stable sales may vary greatly from year to year.⁹ Presumably, because the system is not capable of providing such incidence data in convincing form, the FDA rarely attempts to track known adverse events even though 1.5 million people are hospitalized each year for adverse reactions that, by definition, are already identified.⁵

The monitoring system based on spontaneous reports is also incapable of detecting many important potential dangers of approved drugs. For example, if a drug causes an event that might be expected as part of the natural history of the disease being treated, the spontaneous detection system fails.¹⁰ It is not capable of detecting that flosequinan increases mortality in congestive heart failure, or that flecainide and encainide can cause cardiac arrest. A spontaneous reporting system also cannot capture adverse effects that manifest themselves as a disease with high prevalence or with a long delay between exposure and clinical manifestation. Cancer is the classic example. While spontaneous reporting makes a valuable contribution, it provides only a fraction of information required to develop programs to protect the public from the health risks of marketed drugs.

The third task in safety monitoring is to assure that once new warnings or recommended restrictions are included in the product labeling, these messages are effectively communicated to prescribing physicians and to the general public. While the FDA has access to up-to-date prescription volume data, it has no organized program to check whether the most important warning messages are being heeded. The limited information available, however, suggests that some important safety information—such as boxed warnings on drug disclosure labels—either was not received or had little effect. For example, one outcome of the protracted debate over the safety of the sedative triazolam was a new drug label warning that it should be prescribed for only 7 to 10 days. Several years later an FDA task force reported that 85% of the prescriptions were being written for longer periods.¹¹ Similarly, new warnings about the addictive properties of propoxyphene had no effect on either prescription volume or the number of overdose deaths.¹² After the FDA's warnings against using class I antiarrhythmic drugs for suppression of premature ventricular contractions and other mild ventricular arrhythmias, 81% of cardiologists reduced the number of new prescriptions, but a similar percentage continued to prescribe these drugs for all or most of their existing patients on drug therapy.¹³

While few want the FDA to regulate the practice of medicine, safety monitoring should include surveillance of the most important known and expected risks, and regular public reporting on the response of the public and prescribing physicians. Also, it is important to discover why important safety warnings are not being heeded. Are "Dear Doctor" letters and label changes being lost in the avalanche of pharmaceutical marketing and advertising information being sent to the physicians' office? Or do physicians disbelieve the message? Neither the FDA nor any other agency has an organized program to find out whether the important warning messages are achieving their intended purpose of protecting the public and, if not, discovering the cause.

The fourth element in postmarketing surveillance involves sentinel programs that focus on the most common drug-induced medical problems—rather than on specific suspected drugs. Instead of waiting passively for spontaneous reports, active surveillance involves looking for sensitive indicators of possible problems. In particular, regular monitoring is needed to check for drug involvement in reported cases of birth defects, agranulocytosis, a disorder primarily caused by drugs, and aplastic anemia, which is frequently caused by drugs. So many drugs—including tricyclic antidepressants, agents for migraine headaches, diet drugs, class I antiarrhythmics, nonsedating antihistamines, and cardiac glycosides—have been linked to arrhythmia and other cardiac effects that a cardiac early-warning system may also be warranted. In the last 2 decades the FDA has funded a handful of specific studies on drug-induced disorders,^{14 - 15} but no regular monitoring now takes place. Just as we maintain a global sentinel system to detect new strains of influenza, we need early-warning systems designed to detect unsuspected drug effects.

Effective safety monitoring also requires independence and the authority to act on drug hazards that are identified. The discovery of serious adverse effects not detected in preapproval testing can be unwelcome news to both the drug's sponsor and the FDA division that evaluated and approved the drug. At present, the DPE is limited to reporting its information and conclusions back to the FDA review divisions, even if these findings call into question the original decision to approve the drug.¹⁶ The need for more independent safety monitoring has been long recognized. In 1970, a National Academy of Sciences panel recommended establishing a National Drug Surveillance Center reporting directly to the FDA commissioner.¹⁷ In 1980 the Joint Commission on Prescription Drug Use recommended to Congress the creation of Center for Drug Surveillance completely separate from the FDA.¹⁸ In 1993, the FDA conducted its own lengthy review of postmarket safety monitoring because of senior management's "concern the epidemiology program was deteriorating."¹⁹ (appendix 6) The 5 outside expert consultants all recommended greater independence for the DPE, most of them advocating an independent office reporting directly to the FDA commissioner.¹⁹ (p2)

In the United Kingdom, safety monitoring of marketed drugs is separated from initial approval even more completely than any US report has recommended. In the British system, one major office reviews and recommends approval of new drugs and an entirely separate unit monitors the safety of approved drugs and may order changes in product labeling and even may order outright drug withdrawal.¹⁹ (appendix 1) France has a different organizational plan, with a well-developed network of 30 regional pharmacovigilance centers, a national database that practitioners can query, and an important drug safety journal.²⁰

With limited resources and continuing pressures to act rapidly on new drug applications, the FDA focuses on improving its performance on the narrowly defined mission of identifying unlabeled serious adverse effects of newly marketed drugs. Without additional funds and a new mandate, the other 3 vital safety monitoring tasks will continue to be largely neglected. The nation will continue to operate a drug safety program without reliable data on how many people are seriously injured each year, and without knowing which drugs or families of drugs may require greater vigilance. There will be no program to assure that important safety warnings are communicated effectively to physicians and the public, and no early-warning system focusing on the most common drug-induced disorders.

The nation needs an office of drug safety with the authority, independence, funds, and legal mandate to undertake all 4 of the major tasks that define a basic drug safety monitoring program. The funds to pay for this expanded safety program should come from user fees collected from pharmaceutical companies. The drug industry should be no less liable for the costs of its safety regulation than the nuclear industry, which pays for the full costs of the Nuclear Regulatory Commission. Even if the pharmaceutical industry, already the nation's most profitable as measured by return on investment, passes on the whole burden to consumers, the added costs would be small. Just 1 penny per prescription would yield approximately $24 million for an enhanced drug safety monitoring program. Additional sums—raised through the same mechanism—could finance other needed initiatives in long-term drug testing, safety research, and education at the nation's academic medical centers. This investment in drug safety has the potential to save thousands of lives and prevent tens of thousands of serious injuries every year.