Pregnancy registries: Differences, similarities, and possible harmonization
Address correspondence to Torbjörn Tomson, Department of Clinical Neuroscience, Karolinska Institutet, Department of Neurology, Karolinska University Hospital, SE 171 76 Stockholm, Sweden. E-mail: firstname.lastname@example.org
Epilepsy and pregnancy registries have been operational for more than 10 years, have accrued considerable experience, and collected an impressive amount of data. As findings have been published, it has become important to understand how observations from the different registries are comparable, especially since their methodologies differ somewhat. In September 2008, representatives of the UK Epilepsy and Pregnancy Register, the North American AED Pregnancy Registry (NAAPR), and the European and International Registry of Antiepileptic Drugs in Pregnancy (EURAP) met for a workshop. Their objectives were to exchange information on their methodologies and to discuss areas where harmonization might be possible. This report summarizes these discussions.
In an attempt to assess pregnancy outcomes in women with epilepsy, different groups established epilepsy and pregnancy registries in the late 1990s with the aim of enrolling prospectively large numbers of pregnancies with exposure to antiepileptic drugs (AEDs) (Beghi et al., 2001). Some were set up by pharmaceutical companies and collect data on the manufacturers’ own product (e.g., GSK International Lamotrigine Registry, UCB Keppra registry). Others have been established by independent research groups and include information on all AED exposures. These are national, regional (e.g., Australia, United Kingdom, North America, Kerala, India), or broadly international (European and International Registry of Antiepileptic Drugs in Pregnancy, EURAP). Many of the registries have been operational for approximately 10 years. As some registries begin to release results, the question as to whether results can be compared across studies has arisen (Tomson et al., 2007).
The International League Against Epilepsy (ILAE) Commission on Therapeutic Strategies thought it useful to promote an in-depth discussion and so initiated a workshop (sponsored by the Milken Foundation) to include representatives of the three largest independent registries: the UK Epilepsy and Pregnancy Register, the North American AED Pregnancy Registry (NAAPR), and EURAP. The primary objective was to provide a forum for discussion, allowing representatives of the registries to exchange information on their methodologies and to gain from the experience of others. The discussions at the workshop are summarized in this report. At the end of each section, potential harmonizations discussed at the meeting are described.
The overall objectives of the major registries are similar, namely, to assess the risk of major congenital malformations after prenatal AED exposure. Nonetheless there are differences in the methodologies of each, which may have an impact on the results and affect the conclusions drawn, as discussed below.
Methods for Enrollment
Often, registries must balance between a comprehensive approach, which may limit the number of pregnancies that will be enrolled, versus a more simplistic, “user-friendly” evaluation system that will capture more pregnancies but provide less comprehensive data. One of the major differences between pregnancy registries is whether women self refer or are referred by physicians, and, if so, whether the physicians are generalists or specialists, and of which kind. Increased specialist input will likely lead to evaluation of more severely afflicted women, but probably also to more accurate data on epilepsy classification.
The UK registry enrolls pregnancies from the United Kingdom and Ireland and has given priority to a high ascertainment rate and external validity. Among the three registries this is the one enrolling the highest proportion of eligible pregnancies within its region of interest (estimated 25–33%). To achieve this, reporting is made simple with one report on enrollment and a second at follow-up 3 months after birth. Approximately 50% of the pregnancies are enrolled through direct self-referral of the pregnant women, and the rest through general practitioners, midwives, or other health care personnel. General practitioners are also the source of follow-up information. Because the amount of information collected has been kept to a minimum, assessing the role of potential confounders is more difficult. The NAAPR enrolls pregnancies mainly from the United States and Canada. Recruitment is by self-enrollment of pregnant women who must call a toll-free number to register. There are two follow-ups, at 7 months of pregnancy and 8–12 weeks after estimated date of delivery. This is done by a telephone call from the staff of the registry with a phone interview with the mother. The information obtained in the interview is supplemented with data from medical records, with the written consent from the woman (which is granted by about 60% of the enrolled women). Only a small percentage of pregnancies of women with epilepsy on AED treatment in North America is captured by the registry, and there are regional differences in ascertainment rate.
EURAP relies on enrollment through regional and national networks of collaborating physicians in more than 40 countries in Europe, Asia, Oceania, Australia, and South America. There are separate follow-up reports after each trimester and at 1 year after birth. These are submitted by the reporting physician and channeled through a national coordinator to the central registry in Milan Italy. Pregnant women cannot self-enroll. The ascertainment rate varies markedly between regions ranging from a few percent up to 20–30%. A few percent of the pregnancies enrolled in the UK registry are also included in EURAP. Likewise EURAP collaborates with the Australian and Kerala registries, which are similar in methodology. Pregnancies from these registries that are enrolled and assessed according to the EURAP methodology also contribute to the EURAP database.
Hence, the NAAPR and EURAP cohorts are recruited differently, and both differ from the UK system. The requirement of self-enrollment in North America is likely to introduce a bias toward more motivated and probably better informed women. In EURAP there may be a bias toward more severe epilepsy. Women enrolled in the UK register are likely to be representative of the UK epilepsy population, but data are somewhat limited in detail compared to the other two registries. Information related to the woman’s epilepsy is probably most reliable in EURAP, as the source in general is her treating physician (epileptologist). On the other hand, unlike the North American Registry, EURAP lacks direct access to additional information on the patients to allow for more detailed follow-up questions.
Inclusion Criteria and Classification of Pregnancies
The UK registry includes pregnant women with epilepsy with or without ongoing AED treatment. NAAPR includes all women on AEDs at any time during pregnancy irrespective of the diagnosis. Since 2000 the NAAPR has enrolled an internal control group of friends and family members of eligible and enrolled participants who are not taking AEDs. The criterion for inclusion in EURAP is AED use at time of conception regardless of the indication. The proportion of pregnancies with AEDs for indications other than epilepsy is low in both NAAPR and EURAP, estimated at approximately 10% and 2%, respectively.
The criteria for women included in the analysis may have a significant impact on results. For the primary assessment of teratogenic risks, all registries limit their analysis to prospectively enrolled pregnancies to reduce the risk of selection bias. However, the definitions of a prospective enrollment vary between registries. In the UK registry, subjects are considered prospective if they are referred to the registry before the outcome of the pregnancy is known. Excluded are cases for which any prenatal test has shown an abnormality, and cases resulting in a pregnancy loss in which an abnormality has been identified before referral to the registry (Morrow et al., 2006). NAAPR includes women at any time during pregnancy. Women are classified as “purely prospective” if they, at the time of enrollment, could not know whether the fetus has a malformation, that is, too early in pregnancy or before any prenatal test. The “traditional prospective” enrollees have some knowledge of the status of the fetus, typically after having prenatal screening by ultrasound at 16–20 weeks of gestation, but are considered prospective if still pregnant at time of enrollment (Holmes et al., 2008a,b). Sixty percent of the enrolled pregnancies have been “pure” prospective. EURAP applies the strictest criteria for prospective pregnancies, including only pregnancies enrolled before outcome is known, and no later than week 16 of gestation. Retrospective pregnancies are also enrolled and followed but not included in the primary analysis.
These differences in criteria for prospective pregnancies could have an effect on the results. Selective exclusion of pregnancies, whether by exclusion or by referral bias, where prenatal screening before enrollment has revealed abnormal outcome would reduce rates for malformations that can be detected by these tests. Exclusion of all with informative prenatal screening, as in EURAP and pure prospective NAAPR, does not introduce the same bias, but will make women who for any reason have early tests less likely to be included in the registry. Including “traditional prospective” pregnancies as defined in NAAPR, on the other hand, could introduce a selection and reporting bias.
Information on AED Exposures
In some cases, women start, stop, add, or subtract AEDs during their pregnancy. Even more commonly, AED doses will be adjusted. How to classify exposure may not be obvious. A clear definition of type and level of exposure to AEDs is critical. The registries differ slightly in the level of details on AED doses and dose changes during pregnancy, and how women are classified. However, all include fairly detailed information on AED treatment during the first trimester. EURAP excludes pregnancies in which changes occur during the first trimester. NAAPR includes as monotherapy exposure patients who started after conception, or stopped before end of trimester, but excludes those switching between AEDs in the first trimester. In the UK registry, women who started an AED after conception are excluded. Those with AED exposure from onset but where AEDs are withdrawn during the first trimester are classified according to their original treatment. Those changed from monotherapy to polytherapy are classified as polytherapies, as are those changed from poly to monotherapy. In case of dose changes, the highest dose is considered.
The work group suggested that for the analysis of association with teratogenic outcome, polytherapy should be defined as use of two or more AEDs throughout the first trimester and that the monotherapy exposures should be defined as those pregnancies exposed to a single AED from conception to the end of first trimester. For analyses of dose–effect relationships the average dose over the first 14 weeks was recommended. Evaluation of the highest dose/administration was also considered to be of value in assessments of potentially dose-related teratogenic effects.
Pregnancy registries are observational studies. Because patients have not been randomized to their treatment, it is particularly important to analyze the impact of heterogeneous characteristics across treatment groups on the observed associations between AED exposure and teratogenic outcome. All registries collect information on such potential confounders, with most detail being included in EURAP and NAAPR. The methods for obtaining this information also vary. In the UK registry, the information is provided from multiple sources including the patient, neurologist, epilepsy nurse, midwife, general practitioner, and/or any specialists involved in the care of the offspring. In NAAPR, information is obtained through the telephone interview with the mother and copies of medical records from the woman’s neurologist and her infant’s pediatrician and other consultants, as needed. In EURAP the information is obtained and provided by the reporting physician on three occasions throughout pregnancy.
Family history of birth defects is considered to be one of the most important confounding factors. Although all registries record this information, it is not defined by all registries in the same way. Educational level or socioeconomic status of the parents, type of epilepsy, and occurrence of convulsive seizures during pregnancy were other factors considered important to include as covariates in the analysis.
Environmental factors, such as smoking and use of alcohol, use of other concomitant medications, and comorbidities may affect rates of birth defects in otherwise healthy pregnant women. Information on smoking and alcohol intake are obtained in NAAPR and EURAP only, with other medications and comorbidities being obtained in all registries. Methodologies do differ, however. Although NAAPR includes a checklist in the direct interview of the pregnant woman, these are open questions to be answered by the reporting physicians in EURAP, which could lead to less complete reporting.
The three registries decided to restrict family history to first-degree relatives in the analysis. It was recommended that the presence of environmental factors should be accounted for descriptively in tables along with lists of pregnancies with adverse outcome when these are reported.
Methods for Follow-Up and Assessment of Outcome
Major congenital malformation is the outcome of primary interest in all registries, and minor anomalies are not reported. A commonly used definition of a major malformation is a structural abnormality with surgical, medical, or cosmetic importance. Although the definition of a major congenital malformation is similar across registries, there are borderline categories of abnormalities and it appears that EURAP is more inclined to include such cases. Chromosomal or monogenic abnormalities are excluded or analyzed separately by all registries. There are also borderline cases for which it is difficult to make a clear distinction and for which the registries may handle individual cases differently.
Information on outcome is obtained in the United Kingdom initially through the general practitioner 3 months after estimated date of delivery and then through the relevant specialists as necessary. In NAAPR, this information is collected by a telephone interview of the participant 3–4 months after delivery, and the outcome can be confirmed by review of medical records in approximately 60% of participating women who grant such access. EURAP relies on written information from the reporting physicians 2 and 12 months after delivery. The obtained data in all registries are reviewed centrally by an expert blinded to the type of exposure.
These differences could have an impact on rates and types of malformations. A major difference between the registries is the time window for assessment of teratogenic outcome. In EURAP >30% of the malformations are not reported until the 12-month follow-up after birth, and not reported at 2 months after delivery, which is close to the time frame for reporting abnormalities in the other two registries.
To facilitate comparison between the registries it was considered essential to record and report when and how abnormalities were first detected. If a comparison between registries were to be undertaken post hoc, a time cut-off for the included abnormalities could then be used to bring the results into alignment.
EURAP and the UK registry rely on internal comparisons of teratogenic outcome between different AED treatments. In contrast, NAAPR utilizes internal as well as external unexposed comparison groups. In the first 380 enrolled unexposed internal controls, the malformation rate at birth was 1.8% (Smith & Holmes, 2008). A second comparison group for NAAPR comes from the Active Malformations Surveillance Program at Brigham and Women’s Hospital in Boston, where the malformation rate was 1.62% after excluding infants with genetic disorders and chromosomal abnormalities (Nelson & Holmes, 1989; Peller et al., 2004).
The registries differ widely in publication policy. The UK registry has released data as it has become available without preset criteria and have thus published on malformation rates with the most common therapies (Morrow et al., 2006), with levetiracetam (Hunt et al., 2006), and with topiramate (Hunt et al., 2008).
NAAPR has used predefined release criteria. The criterion for release of results for a positive association (relative risk >1) is met when the lower of the 95% confidence interval (CI) is greater than or equal to 2.0. The release criterion for no associated increase in the frequency of all major malformations is met when the upper of the 95% confidence limits does not exceed 2.0. Publications have been released on outcomes with phenobarbital (Holmes et al., 2004) and valproate (Wyszynski et al., 2005). Data on lamotrigine were published based on additional release criteria for an increased frequency of specific malformations (Holmes et al., 2008a,b). Recently, the NAAPR Scientific Advisory Committee decided to release the findings for all drugs used as monotherapy, restricted to those drugs with a sample size of at least 50. These findings were published in 2008 (Holmes et al., 2008a,b). Separate analyses have been carried out on the correlation of the type of epilepsy in enrolled women taking valproate with the rate of malformations in their infants (Bromfield et al., 2008).
EURAP does not release outcome data for individual drugs or drug combinations until sufficient data are available for a meaningful multivariate analysis with inclusion of relevant confounding factors, for example, family history of birth defects, type of epilepsy, seizure control during pregnancy, and maternal educational level. The number of pregnancies needed has not been defined in advance but has been estimated to be at least 5,000 prospectively ascertained cases with completed 1-year follow-up after birth. This has not yet been accomplished and hence EURAP has not released any data on malformation rates with individual AEDs. The only major publications so far from EURAP have described seizure control during pregnancy (EURAP study group, 2006) and AED utilization in different countries (EURAP study group, 2009).
Are there Possibilities of Harmonization/Collaboration?
It is clear that the registries are too different in many respects to make general pooling of data meaningful. Some areas where harmonization would be appropriate and practical have, however, been identified above.
Some objectives can also be identified for which it is unrealistic to expect that an individual registry itself can provide conclusive data and for which a closer collaboration or pooling of data will be the only way to find answers within a reasonable time. Examples might be assessment of outcomes associated with the less frequently used monotherapies, or combination therapies; analysis of dose–effect relationships; and studies of uncommon specific malformations.
The present report has focused on the three major registries that were represented at the workshop. Other registries, such as the Kerala Registry of Epilepsy and Pregnancy, KREP (Thomas et al., 2008) and the Australian Register of AEDs in Pregnancy (Vajda et al., 2003) have also made significant independent contributions, and also contribute 40–80% of their pregnancies to EURAP. Characteristics of all five registries are summarized in Table 1. The pregnancy registries have so far gathered essential prospective data on pregnancy outcome that would never have been available without their existence. The three registries discussed in detail here are slightly different in their scope, differ significantly in their methodologies, and their results are not readily suited for pooling. Instead, they can be regarded as complementary, and the existence of different distinct registries should be seen as a major asset allowing observations from one study to be confirmed or refuted by others. Any differences in results should, however, be interpreted against the background of known variations in methodology. Continued functioning of the registries as separate independent studies is of vital importance to provide answers to the many remaining questions and to the new questions that will come with the introduction of new therapies. However, a more formalized collaboration between the registries is planned for some specific issues that cannot be resolved without combined efforts and resources.
Table 1. Summary of study design and other methodologic issues in five AED and pregnancy registries
|Enrolled pregnancies (n)||7,175||7,312||1,510||13,999||1,306|
|Setting and methods for enrollment||U.S. and Canada self enrollment by the pregnant woman||UK through physicians, nurses, and patient self-enrollment||Australia (contributes to EURAP)|
Self-enrollment by eligible women
|International (42 countries)|
Through network of reporting physicians
|Kerala, (contributes to EURAP)|
Referred by physicians, gynecologist and patients self referral
|Inclusion criteria||Women taking AEDs for any reason during pregnancy||Women with epilepsy with/without AEDs 1st trimester||Women with epilepsy with/without AED 1st trimester|
Women on AEDs for other indications 1st trimester
|Pregnancies with AED exposure at time of conception||Women with active epilepsy with/without AEDs.|
Women on AEDs in 1st trimester for other indications
|Exclusion criteria|| ||Prenatal tests with abnormality before referral. Change in AED in 1st trimester|| ||Change in AED 1st trimester. Outcome unclassifiable|| |
|Criteria for prospective||“Pure” prospective enrolled before results from prenatal screening||Enrollment before pregnancy outcome is known||Prospective enrolled before screening results known||Enrolled before outcome is known (prenatal screening) and within week 16||Enrolment before pregnancy outcome is known|
|AEDs exposure||AED dose, regimen, and brand||AED dose, regimen||AED dose, regimen||AED, dose, regimen||AED, dose, regimen|
|AED drug levels||Not systematically||Not recorded||Not recorded||Not recorded||Not recorded|
|Data collection and methods for follow-up||3 contacts, Telephone interviews with subject, supplemented by medical records in 60%||2 contacts with patient’s physician||4 telephone interviews with patient supplemented by contact with physician||4–5 contacts, mainly personal visits with reporting physician supplemented by medical records||4-5 contacts with physician supplemented by medical records|
|Diagnosis of epilepsy||Self-report; records from neurologist of enrollee||Patient’s physician||Patient’s physician||Patient’s physician||Patient’s physician|
|Outcome ascertainment and classification|
|Exclusion criteria||Genetic/chromosomal, minor anomalies, positional deformations||Genetic/chromosomal, minor and major anomalies analyzed separately malformations||Genetic/chromosomal abnormalities analyzed separately||Genetic/chromosomal abnormalities analyzed separately|
As those with screening before enrolment
|Minor anomalies/chromosomal abnormalities|
|Assessment||Review of medical records by blinded teratologist, direct communication with mother/physician when needed||Abnormal outcomes classified by one clinical geneticist based on medical records||Based on review of medical records||Central classification by blinded teratologists based on reports from physicians||Direct evaluation by reporting physician supplemented by echocardiography and ultrasonography|
|Time window of assessment||Malformations detected at two time points: i) within first 5 days of life; ii) at postpartum call at 8–12 weeks of age.||Within 3 months after birth||Within 12 months after birth||Within 12 months after birth|| Within 12 months after birth|
|Classification of outcome||Major malformations||Major malformations; according to EUROCAT criteria||Major malformations (Birth defects as defined by Victorian Birth Register)||Major malformations according to EUROCAT criteria||Major malformations|
|Comparator/control population||1. External comparison group|
2. AEDs for nonepilepsy
3. Internal unexposed control group is being recruited
|Internal comparison between different AED groups and untreated epilepsy||1. Untreated women with epilepsy|
2. AEDs for nonepilepsy
3. Internal comparison between different AED
|Internal comparison between different AED treatments||External comparison with a small cohort of healthy women recruited prospectively in 1st trimester from a maternity hospital.|
The financial support for the workshop provided by the Milken Family Foundation is gratefully acknowledged. We thank Dr. Frank Vajda for contributing information from the Australian register and Sanjeev Thomas for information on the Kerala registry as summarized in Table 1.
We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.
SH-D has received unrestricted training and research grants from Pfizer, AstraZeneca. Novartis, and Wyeth. LBH received in 2008 salary support from the six companies that sponsored the NAAPR: Abbott, Eisai, GlaxoSmithKline, Ortho-McNeil, Novartis, and Pfizer. TT, DB, JC, and JM have received speakers’ honoraria or research grants (for their respective pregnancy registries) from Eisai, GlaxoSmithKline, Janssen-Cilag, Novartis, Pfizer, Sanofi-Aventis, and UCB. JF discloses consulting through the Epilepsy Study Consortium: GlaxoSmithKline, Pfizer, UCB, Johnson & Johnson, Cyberonics, Ortho-McNeil, Eisai, Jazz Pharmaceuticals, Ovation Pharmaceuticals, Bial Pharmaceuticals, Neurovista, Valeant Pharmaceuticals, Icagen, Supernus, Ikano, SK Pharmaceuticals, Taro Pharmaceuticals, Neurotherapeutics, Sepracor, and Novartis. DL has no conflicts of interest to declare.