Patients, prescribers, and policymakers often assume that the most serious short-term adverse effects of a drug are identified in pre-marketing studies, so recognition of unexpected harm after widespread use raises concern about "failures of the system." For drug-induced birth defects, however, the situation is reversed: the unfortunate reality is that we learn about virtually all teratogenic effects only after a drug has already received marketing approval, and of course, only after it has been used by pregnant women. Indeed, some teratogenic effects may be identified only after the drug has been in use for many decades. In these situations, the problem is due not to a failure of the system but, rather, to the fact that there is no system that can effectively identify teratogens before the drug be-comes available on the market. Teratogenic effects in humans cannot be predicted reliably from the class of a drug or from what is known about its pharmacology and toxicology. 1 Studies in animals are seriously limited in their ability to predict human teratogenesis because of considerable variations in species-specific effects, even among mammalian species. 2 Furthermore, teratogens are not detected in the human trials that are conducted before approval for marketing is granted, since such studies are too small, and more important, they usually — and appropriately — exclude women who might become pregnant, particularly if there is any suspicion generated from animal studies that a drug might be teratogenic.
Since most human teratogens are identified only in the post-marketing setting, one would expect there to be sophisticated approaches in place to accomplish this identification systematically, rapidly, and rigorously. Though a number of individual re-search programs have been established to generate and test selected hypotheses regarding human teratogens, these programs have not had the mission or the resources for the systematic and routine study of the risks and safety either of drugs newly introduced to the market or of those most commonly used in pregnancy. Operating without overarching guidance, direction, or support, these programs do not act as a coherent system to ensure the early and effective detection of birth defects caused by drugs in current use. A comprehensive surveillance system for teratogens is critically needed.
What do we need to know?
Defining the specific objectives of such a system requires that we understand what information is needed regarding the teratogenic risk of a given drug. Women who are or who could become pregnant and prescribers need three kinds of information. First, before a given medication is taken or prescribed, it is important to know whether the drug dramatically increases the rate of birth defects among exposed fetuses. Drugs that pose such large risks might be labeled "high-risk teratogens"; these include isotretinoin 3 and the "classic" teratogen, thalidomide, 4 and they typically affect at least one of every four exposed fetuses. But it is not enough to identify only high-risk teratogens, since some drugs cause birth defects in smaller proportions of pregnancies. Therefore, the second kind of information we need is whether a drug should be classified as a "moderate-risk teratogen." Such an agent may in-crease the risk of specific birth defects by a factor of anywhere from 2 to 10 or somewhat greater, so that the base-line risk of 0.1 percent for a neural-tube defect might be increased 10 times, to 1 percent. Although high-risk teratogens fortunately appear to be few, even our limited state of knowledge indicates that proven and putative moderate-risk teratogens are more numerous and include not only prescription drugs (e.g., carbamazepine 5 and valproic acid 6 ), but also over-the-counter drugs (e.g., pseudoephedrine 7 ). Third, since we are unlikely ever to have enough information to identify or rule out definitively all potential teratogenic effects of every drug that might be taken in pregnancy, we need some understanding of the current state of knowledge or uncertainty for any given drug.
From the standpoint of public health, those with regulatory, legislative, and financial responsibilities require the same kinds of information to determine a given drug’s "teratogenic burden." This cost to society derives from a combination of the prevalence of the drug’s use and the magnitude and severity of its effects. Public attention tends to focus on the few high-risk teratogens, but the more common moderate-risk teratogens can, by virtue of wider use, pro-duce a substantially greater burden.
These questions define the objectives of a teratogen surveillance system, but the system must also determine which drugs to monitor. Two principles can be helpful in guiding priorities. The first is simple: newly introduced drugs need urgent attention, since we cannot be sure whether such drugs are high- or moderate-risk teratogens until they are used by pregnant women. However, the vast majority of medications currently in use have not been studied in a way that would reveal moderate teratogenic risks. Though most concern is directed toward prescription products, over-the-counter medications are used far more frequently. 8 Therefore, the second priority is to identify moderate-risk teratogens among prescription and over-the-counter drugs that are most commonly used by women who are pregnant or who might become pregnant while taking those drugs.
What study designs can provide this information?
Two main study approaches have been developed for the purpose of identifying teratogens in the post-marketing setting: follow-up studies and case–control surveillance. Follow-up studies identify women exposed to specific drugs and may involve the creation of cohorts of women who have been exposed to a particular drug in pregnancy and are then followed to identify the outcomes of their pregnancies. They may also involve the "mining" of extant data bases (such as claims data collected for Medicaid or health plans). The former, often called "pregnancy registries," are typically developed by the pharmaceutical industry or by research groups based in universities. 9 The design of pregnancy registries permits investigators to efficiently identify drugs that are high-risk teratogens (or drugs that put a child at high risk for other common adverse pregnancy outcomes, such as developmental delay). For these purposes, small numbers suffice. For example, only 36 pregnancies that were prospectively followed were needed to identify the teratogenicity of isotretinoin. 3 It is important to recognize that usually the effects of high-risk teratogens are so great that they outweigh analytic concern about control groups and confounding variables. However, the relatively small samples recruited in such studies provide insufficient statistical power to identify moderate-risk teratogens because the base-line risk of a specific defect may be 1 in 1000 or less and the risk among exposed fetuses may be in the range of 2 to 10 per 1000. It is a matter of concern that pregnancy registries for specific drugs, often operated by the manufacturers, seek to recruit patients receiving different drugs into separate registries and a patient receiving multiple drugs into multiple registries. Such competition for patients and practitioners can compromise recruitment for individual registries and may dilute or fragment the efforts of pregnancy registries that focus on a wide range of drugs.
Data bases developed for administrative purposes can be used to identify both women who were exposed to a specific drug while pregnant and the outcomes of their pregnancies. These data can be used to identify high-risk teratogens, although for newly introduced drugs they are likely to provide results more slowly than pregnancy registries. Be-cause such data bases may include more pregnancies in which the fetus was exposed to a given drug than pregnancy registries, one might think these data bases could also effectively identify moderate-risk teratogens. However, the valid identification of moderate-risk teratogens (unlike high-risk teratogens) requires rigorous consideration of potential confounding variables, such as other drugs (including over-the-counter drugs), smoking, alcohol, and the indication for which the drug was taken. Such information is rarely available in data bases developed for administrative purposes.
Case–control surveillance was introduced in the mid-1970s as a system for identifying serious illnesses that result from medications used in an ambulatory population. 10 Whereas a typical case–control study is designed to obtain information on a specific type of exposure among both subjects with and those without a specific outcome, case–control surveillance broadens the focus such that multiple case–control studies are conducted within a single infrastructure. This approach was applied to birth defects in 1976 by my group of university researchers, 11 and it has now been expanded to four geographic centers. Another case–control surveillance study of birth defects, involving eight states, began in 1998 under the sponsorship of the Centers for Disease Control and Prevention (CDC). 12 By including information on infants with any of a wide range of specific birth defects and interviews with women that focus on details of their antenatal exposure to all prescription and over-the-counter medications (including herbal products), case–control surveillance can provide opportunities to examine large numbers of specific defects in relation to the wide range of medications taken by pregnant women.
A major strength of such studies is their substantial statistical power, which permits the efficient identification of moderate-risk teratogens not only among commonly used drugs, but also among drugs that are used relatively infrequently. For ex-ample, moderately increased risks (an odds ratio of 3 or 4) were identified for gastroschisis in relation to pseudoephedrine 7 and for cardiac defects in relation to trimethoprim. 13,14 Furthermore, case–control surveillance studies include appropriate control groups and information on potential confounding variables. However, a potential limitation of this de-sign is that recall of medication use may be more complete among mothers of affected infants than among mothers of unaffected infants. This approach is also less efficient and timely than follow-up studies for identifying high-risk teratogens.
Even when the accumulated data are not yet sufficient to provide definitive answers regarding the relation of a given drug to a given defect, the large samples in case–control surveillance studies can provide preliminary estimates of safety and risk for all drugs in relation to all birth defects. The stability of these estimates is reflected in their 95 percent confidence intervals. Thus, for a drug for which no increased risk is identified (i.e., one with an odds ratio of 1.0), knowing, for example, that there is a 95 percent chance that the true risk estimate falls between 0.8 and 1.5 provides far greater reassurance than would the same odds ratio with a safety estimate of 0.2 to 50. These estimates are similarly useful in interpreting odds ratios that are greater than 1.0. Of course, preliminary analyses of these data bases must be carried out with the recognition that statistical stability does not reflect scientific validity and that repeated forays into the data increase the likelihood that associations will be observed by chance.
How can a teratogen surveillance system be developed?
Current follow-up and case–control surveillance studies provide important information about teratogens, but these efforts form a patchwork at best. The United States does not have a coherent, organized system of teratogen surveillance. This is so because no agency has taken the responsibility for the broad direction of data collection and analysis that is needed to provide information on the teratogenic effects of newly marketed prescription drugs, as well as established prescription drugs and over-the-counter drugs. In keeping with its regulatory mission and obligations, the Food and Drug Administration (FDA) is the logical agency to establish such a system. This objective is now readily achievable.
Specifically, the FDA should create a teratogen surveillance system based on a consortium of groups conducting follow-up and case–control surveillance studies, which could take advantage of the complementary strengths of the various post-marketing epidemiologic studies and infrastructures that are already in place. Follow-up studies can quickly and efficiently identify high-risk teratogens, and case–control surveillance can efficiently identify moderate-risk teratogens and provide, for the remaining drugs, safety and risk estimates that be-come more precise as data accumulate. An advisory committee should set overall goals and provide guidance for the research activities, and should serve as an authoritative forum for interpreting conflicting and often controversial data relating to the teratogenic risks of drugs. This advisory committee should include members who represent the FDA, the CDC, the pharmaceutical industry, the study investigators, and independent experts.
Until recently, the FDA might have been con-strained in its ability to provide even the modest sup-port needed to develop and maintain such a system. However, the situation has changed dramatically in the past year, providing the FDA with a unique opportunity. The Prescription Drug User Fee Act (PDUFA), first passed by Congress in 1992, was de-signed to enhance the drug-review process through the use of fees collected from the pharmaceutical industry to assist the FDA. The third iteration of this legislation, PDUFA III, approved by Congress in 2002, directs the FDA to spend considerable portions of the fees collected for specific post-marketing activities and requires that additional funds from the total appropriation to the FDA be reserved for the Office of Drug Safety, which is responsible for post-marketing safety efforts. 15
The FDA should use these resources to initiate development of the proposed teratogen surveillance system by providing basic support for the data-collection infrastructure and the advisory committee. As part of their post-marketing commitments, drug manufacturers currently support numerous studies designed to evaluate the safety of specific drugs for use during pregnancy. These studies vary in their level of independence from the sponsor, and some sponsors may initially be reluctant to yield control. However, redirecting their support to a common system would offer manufacturers many advantages, including both cost savings and standardization of what is now an inconsistent process. As a public–private partnership, the teratogen surveillance system would provide an organized mechanism to develop critical information on high and moderate teratogenic risks of new drugs. At the same time, this system would be efficiently positioned to obtain information on all drugs taken by pregnant women, including older prescription drugs, the wide range of over-the-counter drugs, and even herbal products.
In the late 1950s and early 1960s, more than 10,000 children outside the United States were born with devastating birth defects because their mothers had taken thalidomide during pregnancy. As a result, Congress expanded the mandate of the FDA to ensure that similar tragedies would not take place in the United States. Over the intervening years, we have seen increased efforts directed at ensuring the safety of drugs for the general population, but the mandate to identify teratogenic drugs has gone largely unfulfilled. Forty years after the U.S. public became aware of the dangers of thalidomide, new federal legislation provides a historic opportunity to correct this situation at last. We can and must develop an efficient, comprehensive, ongoing system that will identify and quantify the risk of birth defects posed by the large number of prescription and over-the-counter medications that are available now and that will be available in the future. Pregnant women, their babies, care providers, and the public health require nothing less.
From the Slone Epidemiology Center, Boston University, Boston. Address reprint requests to Dr. Mitchell at the Slone Epidemiology Center, Boston University, 1010 Commonwealth Ave., Boston, MA 02215.
|
To
send us your comments, questions, and suggestions click
here |