Applied Clinical Trials
Post-approval methods for monitoring the safety of drug exposures in expectant mothers.
Scientifically rigorous efforts to monitor the potential negative effects of biopharmaceutical products are an important component of post-approval product safety. Such monitoring is critical. While the consequences of birth defects for children, their families, and society are potentially serious, the uncertainty and fear surrounding these risks can lead physicians and patients to withhold therapeutic interventions that are necessary to treat severe chronic conditions, and can also lead to unnecessary pregnancy terminations.
During the development process, regulatory agencies require all medicinal products to undergo pre-clinical testing to determine reproductive effects, but the results of these studies are not necessarily predictive of the human experience. Clinical trials exclude pregnant women; thus, little is known about the human teratogenic effect of a product at the time it first enters marketing. Once a drug is on the market, a spontaneous safety reporting system is in place to monitor all adverse drug reactions, including both normal and abnormal outcomes of pregnancy drug exposures. Healthcare providers and patients report adverse drug reactions to the manufacturer of the drug, and the manufacturer is required by law to report these events to regulatory agencies. This spontaneous reporting system is useful for identifying severe and unusual events. However, it is inadequate for monitoring the range of reproductive events that pregnant women exposed to the drug may experience.
ANDERSEN ROSS/GETTY IMAGES
As a result, both the US Food and Drug Administration (FDA) and European Medicines Agency (EMA) have, with increasing frequency, required that application holders make post-marketing commitments to conduct studies of the effects of products when used during pregnancy, especially for biopharmaceutical products that have a high likelihood of use in women of child-bearing potential. Both the FDA and the EMA published guidance on monitoring exposures to medicinal products during pregnancy.1,2 The FDA guidance document focuses on establishing prospective pregnancy exposure registries, while the EMA guidance reviews various study design options that may be used to monitor the safety of product exposures in pregnancy. These guidance documents are useful tools, yet selecting the design that is most appropriate for a particular product or patient population remains challenging.
Typical strategies used to assess the teratogenic potential of various products include prospective pregnancy exposure registries, longitudinal database studies, and case-control studies. The purpose of this article is to describe these three approaches, to evaluate their strengths and weaknesses, and to examine the utility of these approaches in monitoring product safety in pregnancy.
A pregnancy registry is a prospective, observational study with an active data collection system spanning the time from exposure through a follow-up period. Pregnant women are enrolled as soon after the exposure as possible and followed throughout their pregnancy. At enrollment, the registry captures data on the exposure, as well as demographic and background information. At outcome, the registry captures data on additional exposures during pregnancy, risk factors, and details regarding the pregnancy outcome, including the presence or absence of birth defects and other perinatal complications. If a live birth occurs, the infant may be followed over a period of time and monitored for the detection of birth defects and other functional or developmental deficits.
Pregnancy exposure registries offer a unique means for collecting data on pregnancy exposures early in the lifecycle of a product, at a time when interest in the product and its safety is keen and other means of safety monitoring are limited. When animal or clinical studies indicate possible risk, a registry may be useful for providing active surveillance. Pregnancy exposure registries are particularly useful for detecting major teratogenic effects, but have limited ability to detect increases in the risk of specific defects or small increases in the overall risk of defects. They have power-limiting sample size constraints due both to the infrequency of pregnancy drug exposures and the impracticality of enrolling numbers large enough to detect these differences. They may also be useful in assessing other adverse events, such as prematurity, maternal complications, or developmental deficits, although the observational nature of these studies typically limits data collection to details that are recorded in the course of usual care rather than formal standardized assessments. Pregnancy exposure registries tend to be hypothesis-generating tools rather than hypothesis-testing tools. They are used to identify potential safety signals for further study.3
Pregnancy registries offer important advantages. Their active data collection system and prospective orientation offer advantages over the passive, retrospective methods of other post-marketing surveillance techniques.4 Furthermore, because the exposure clearly occurs before the outcome, this design avoids the recall bias that studies with a retrospective design may introduce, and enables the registry to promptly capture critical data directly from the source.5 The pregnant woman can provide detailed data on exposure both to the drug of interest and other drugs, as well as to risk factors associated with the outcome, in order to delineate potential influences during critical phases of fetal development. Her physicians—the prescriber and/or her obstetrician—can easily verify this information, and the obstetrician or pediatrician can provide or verify pregnancy outcome data; thus the data on all critical data points are rich and detailed. The collection of data throughout the pregnancy and during infancy contributes to the accuracy and detail of the registry data. A final advantage of pregnancy registries is that they are well suited to studying newly marketed products with low market penetration, which are commonly not well represented in automated databases. Signal detection in a pregnancy registry can be more robust than in spontaneous reporting systems, which are the default source of event reporting at initial marketing.
Table 1. A quick summary of the methods of monitoring pregnancy drug exposures.
One disadvantage of pregnancy registries is their voluntary nature. This design may lead to selection bias—women who choose to participate may be different from those who do not. For example, women who choose to participate may be motivated to participate because they are more educated and knowledgeable about the dangers of drug exposure in pregnancy, or because they may be at high risk for a poor pregnancy outcome. Losses to follow-up can also introduce bias if, for example, women who have a good outcome lose interest in participating and fail to provide follow-up data.
Identifying an adequate comparator group may be one of the most challenging aspects of designing a pregnancy registry. There are two commonly used sources of comparator data: an actively enrolled group of pregnant women with the disease of interest who were not exposed to the drug of interest, and background rates from external surveillance sources or published literature. There are limitations to both approaches. It may be more logistically challenging and more costly to use internal comparators rather than external comparators. In addition, while use of prospectively enrolled internal comparators may appear to be a scientifically superior approach, registry participants are not randomized, and therefore the unexposed population may differ significantly from the exposed women. Relying on background rates may be the least desirable approach from a scientific standpoint but may be the only feasible option. A recent survey of pregnancy registries found that background rates, such as the Metropolitan Atlanta Congenital Birth Defects Program,6 are the most commonly used comparator.7
A multi-product design can simplify identification of a comparator because one product can be compared to another. Comparison with a population that is exposed to a different drug rather than completely unexposed can create obvious methodological problems. However, the two comparator populations are recruited and managed in the same way, and typically both have the disease of interest; thus, they are more comparable with one another than either is with published rates. When choosing a comparator, it may be necessary to strike a balance between scientific rigor, feasibility, and practicality. Using more than one comparator can improve the validity of the registry, and indeed, almost three-fourths of pregnancy registries use more than one comparator group.7
Registries are typically more costly than database studies, primarily because registries have a prospective design, requiring set-up, regulatory and ethics approval, and follow-up. Enrollment rates are often unpredictable, and eligible patients are not always concentrated at specific sites, so recruitment must be sought from the population at large.8 Consequently, the enrollment period for pregnancy registries is often lengthy, running for seven to 12 years, accruing in that time an average of 400 to 700 pregnancies exposed during the critical period of organogenesis in the first trimester. Sample size is limited; even when registries continue to enroll for a number of years, the sample size is generally sufficient to rule out only a major teratogenic effect, not specific defects.9
The information collected by pregnancy registries has been used in product labeling and also used to support changes in the pregnancy risk categorization used in labeling. As an example, Acyclovir was initially designated as a category C drug (indicating that animal studies suggest risk but human data are lacking). After the Acyclovir pregnancy registry determined that birth defect rates in 749 pregnancies exposed in the first trimester were similar to those of the general population, the risk category was changed to B (indicating that animal studies suggest risk but human data are reassuring).10 The FDA has proposed a major overhaul to the pregnancy labeling system. Pregnancy registries promise to play an even greater role in the proposed new labeling system, which is aimed at better describing the benefits and risks of drug use for the mother and fetus, and emphasizes the importance of human data, including data collected in pregnancy registries.
Pregnancy registries are well suited to multinational monitoring of the reproductive safety effects of pregnancy exposures in pharmaceutical products. Differences in countries' regulatory and ethics committee approval requirements vary, which can prove challenging to any global study. However, once country-specific approval has been obtained, a multinational pregnancy registry can capture consistent data by using the same methodology in each country.
Large computerized databases show great promise for monitoring the effects of drug exposure in pregnancy. Computerized database studies are derived from existing databases, such as insurance claims databases (e.g., United Health), health maintenance organization (HMO) databases (e.g., Kaiser Permanente), or national databases (e.g., Danish Medical Birth Registry). These studies require availability of data on drug exposure, prenatal and obstetric outcome, and infant outcome. Such data are not always within the same database, and often multiple databases, such as a prescription database, an obstetric database, and a pediatric database, must be linked. Linkage may be difficult. Furthermore, the databases may not contain all of the necessary data. Additional chart reviews or supplemental questionnaires may be required to obtain sufficient detail to characterize the safety of drug exposure in pregnancy.
Database studies offer a number of advantages over registries. For example, sample size is usually not a constraint, because the populations are very large. Database studies can also provide results more quickly than a registry, and they are typically less costly. Comparator groups are more easily identified and usually may be found within the same database. The comparator may be women who have the same underlying disease but were not exposed to the drug of interest; they may be unexposed or exposed to another drug.
One of the main disadvantages of a database study is the difficulty pinpointing the time of exposure. Assessing drug safety during pregnancy requires establishing a temporal association between the exposure and the formation of the birth defect or other outcome of interest. In a database study, exposure is usually presumed from the date the prescription was filled, which does not necessarily indicate when the drug was taken. A woman may fill the prescription and decide not to take the drug. Conversely, a woman could be misclassified as not exposed if she takes the drug after obtaining it either with an old prescription not captured in the database or from a friend's prescription. Databases often lack much-needed detail on the birth outcome, necessitating use of chart reviews or questionnaires to supplement data with detail on outcomes or risk factors, potentially increasing costs and prolonging the study. Finally, database studies may not be well suited for newly marketed drugs with low market penetration, which are usually better studied with a pregnancy registry design.
Another strategy to monitor safety of drug exposure in pregnancy is a case-control study. Case-control studies have a retrospective design, beginning with the outcome of interest. The cases are infants who have a particular birth defect or other outcome of interest, and controls are carefully selected from infants who do not have the outcome of interest but share other characteristics with the cases. The two groups are then compared with regard to exposure to the drug of interest, with controls for risk factors. While the typical approach focuses on a specific birth defect, multiple case-control studies conducted in parallel can broaden the focus.11
Case-control studies are particularly well suited to studying rare events such as birth defects. This approach offers substantial statistical power to identify teratogens among relatively rare exposures.11 However, case-control studies are limited in their ability to identify a pattern of minor and major malformations characteristic of prenatal exposure to drugs with teratogenic potential.5 Furthermore, case-control studies are retrospective, assessing drug exposure and other risk factors after the fact. Case-control studies that rely on maternal interview for exposure data can introduce recall bias, because a mother of an affected infant may be able to recall medication use in early pregnancy more readily than a mother of an unaffected infant. Medical records may be used to verify or obtain data for case-control studies, but may lack detailed data on the exposure, such as timing, dose, and duration. It may be difficult to select an appropriate comparator group that is comparable to the cases in terms of the potential for exposure to the drug of interest during the time period of risk under consideration. In general, case-control studies may be less efficient and timely than database studies or pregnancy registries.11 Finally, rates of birth defects in the exposed and unexposed populations cannot be determined in case-control studies.
Case-control and database studies may also be conducted multinationally. However, it may be challenging to identify data sources across countries that use the same methods and data collection strategies to assure consistency of data.
All studies designed to monitor the impact of drug exposures in pregnancy require regulatory and/or ethics committee approval. Studies that solicit data directly from patients and their health care providers, such as pregnancy registries and case-control studies, require patient consent. Database studies, however, do not typically require patient consent, because patient identifying information is typically stripped from the analysis data set provided to researchers.
In conclusion, pregnancy registries, database studies, and case-control studies are all viable methods for monitoring product safety in pregnancy. There are definite advantages and disadvantages to the three types of studies. The choice between a registry, database, and case-control design will depend on the drug of interest and the primary objectives of the study. It is important to carefully weigh the strengths and weaknesses of each type of design in relation to the drug, the objectives, and other imperatives, such as the timing of study conduct, outcomes of interest, and available resources.
While all three approaches are useful for monitoring drug safety in pregnancy, no single methodology is sufficient to assess completely the potential negative effects of drugs during pregnancy; it may be necessary to adopt multiple strategies. In some situations, when resources are available, it may be beneficial to conduct both a pregnancy registry and a longitudinal database study in parallel. This strategy may be useful if investigators suspect that infrequent exposure will constrain a pregnancy registry. The database study can document the exposure frequency and place pregnancy registry enrollment metrics into perspective for sponsors or regulatory agencies. Additionally, it may be necessary to use a database study or a case-control study to explore a signal identified in a registry. All three types of studies are valuable tools for monitoring the safety of biopharmaceutical product exposure during pregnancy.
Deborah Covington,* DrPH, is Global Head of Observational Studies and Pregnancy Registries, e-mail: deborah.covington@ppdi.com, and Laura McKain, MD, is Medical Director, Pharmacovigilance, at PPD, Inc., 929 North Front St., Wilmington, NC.
*To whom all correspondence should be addressed.
1. Food and Drug Administration (FDA), Guidance for Industry: Establishing Pregnancy Exposure Registries, (US Department of Health and Human Services, Center for Drug Evaluation and Research and Center for Biologics Evaluation and Research, Rockville, MD, 2002).
2. European Medicines Agency (EMA), Guideline on the Exposure to Medicinal Products During Pregnancy: Need for Post-Authorization Data, (Committee for Medicinal Products for Human Use, 2005).
3. D. L. Covington, J. D. Albano, S. S. Roberts, and L. F. McKain, "Signal Detection Rules Across Pregnancy Registries: Do they Make Sense?" Birth Defects Research (Part A), 85 (5) 462 (2009).
4. D. L. Kennedy, K. Uhl, and S. L. Kweder, "Pregnancy Exposure Registries," Drug Safety, 27 (4) 215–228 (2004).
5. C. D. Chambers, S. R. Braddock, G. G. Briggs, A. Einarson, Y. R. Johnson, R. K. Miller, J. E. Polifka, L. K. Robinson, K. Stepanuk, and K. L. Jones, "Postmarketing Surveillance for Human Teratogenicity: A Model Approach," Teratology, 64 (5) 252–261 (2001).
6. A. Correa-Villasenor, J. Cragan, J. Kucik, L. O'Leary, C. Siffel, and L. Williams, "The Metropolitan Atlanta Congenital Defects Program: 35 Years of Birth Defects Surveillance at the Centers for Disease Control and Prevention," Birth Defects Research (Part A), 67 (9) 617–624 (2003).
7. D. L. Covington, S. S. Roberts, J. Albano, and L. F. McKain. "Pregnancy Exposure Registries: Challenges in Identifying Comparison Groups," Pharmacoepidemiology and Drug Safety, 18 (Supplement 1) S11 (2009).
8. L. F. McKain, J. D. Albano, S. S. Roberts, and D. L. Covington, "Variability Across Current Pregnancy Registry Designs: Impact on Interpretation of Results," Birth Defects Research (Part A), 85 (5) 460 (2009).
9. D. L. Covington, H. Tilson, J. Elder, and P. A. Doi, "Assessing Teratogenicity of Antiretroviral Drugs: Monitoring and Analysis Plan of the Antiretroviral Pregnancy Registry," Pharmacoepidemiology and Drug Safety, 13 (8) 537–545 (2004).
10. K. M. Stone, R. Reiff-Eldridge, A. D. White, J. F. Cordero, Z. Brown, E. R. Alexander, and E. B. Andrews, "Pregnancy Outcomes Following Systemic Prenatal Acyclovir Exposure: Conclusions from the International Acyclovir Pregnancy Registry, 1984-1999," Birth Defects Research (Part A), 70 (4) 201–207 (2004).
11. A. A. Mitchell, "Systematic Identification of Drugs that Cause Birth Defects—A New Opportunity," New England Journal of Medicine, 349 (26) 2556–2559 (2003).