P.II.B.4.2. Epidemiological studies
A rationale for the appropriate study design to address safety concerns relating to use of the medicinal product in pregnancy and/or breastfeeding should be provided in the study protocol. Study types by objective include:
• Drug utilization studies: descriptive studies to establish the extent of exposure in women childbearing potential, pregnancy and breastfeeding women, as well as utilization/switching/discontinuation patterns and time trends , including evaluation of user characteristics such as folic acid use, smoking, alcohol intake, other lifestyle factors, body mass index, medical conditions that could lead to adverse embryogenic, foetal or neonatal outcomes, and exposure to known teratogenic or foetotoxic medicines;
• Medicines safety studies: pharmacoepidemiological studies of adverse events of special interest in causal association with a medicine, considering the impact of the underlying maternal condition (i.e. non-exposed disease comparison group) and other potential confounders;
• Studies to evaluate the effectiveness and broader impact of RMM.
Depending on the product characteristics and the context of use, in some cases (e.g. when use in pregnancy is expected and further characterization of associated risks considered necessary) it may be appropriate to initiate a safety study at the time of marketing authorization. In other cases, if a drug utilization study were to show usage in women of childbearing potential or in pregnant women to an extent that studying associated pregnancy outcomes would be warranted, then setting up a PASS with safety endpoints should also be considered. Likewise, a signal (see P.II.B.5.) could lead to a request for a study to examine the extent of use and put the number of spontaneously reported suspected adverse reactions into perspective. The decision on whether and if so, what studies are needed to evaluate specific pregnancy outcomes (see P.II.A.2.) should be guided by reproductive toxicity studies, signals from spontaneous reports or other sources, or the understanding of risk in the pharmacological class. Finally, drug utilization studies can also be designed to show change in use over time with implementation of RMM in specific populations.
Preferably and if feasible, epidemiological studies should be carried out using existing data sources (i.e. secondary data use) and be designed in such a way as to minimize bias and confounding (see P.II.B.4.2.3.). Given the usually limited exposure to medicines in pregnancy and the low incidence of causally related adverse outcomes (see P.II.A.1.3.), it is usually necessary to include participants from more than one country in order to achieve adequate power.
P.II.B.4.2.1. Pregnancy registries
Pregnancy registry actively gathers information on the outcomes associated with exposure to drugs or biological products during pregnancy, which can be used to conduct a prospective observational study (women are enrolled before the pregnancy outcome). If additional pharmacovigilance activities in the form of data collection from a pregnancy registry are justified, the following should be considered:
• Registries that, in principle, aim to capture all pregnant women with the disease are generally more useful than medicinal product-specific registries because they provide for longitudinal study of treatment and effects (including switches between products) throughout pregnancy, comparison between products and pregnancy outcomes in an unexposed population;
• In exceptional cases, a medicinal product-specific pregnancy registry may be appropriate;
• The use of existing (pregnancy) registries or databases should be considered to enhance long-term follow-up, facilitate the inclusion of comparator groups, make use of existing infrastructure for data collection and analysis, to avoid unnecessary duplication of effort and enhance efficiency in general;
• It may therefore be prudent to opt for a hybrid study design in which the product-specific information required from the marketing authorization holder is complemented with public data sources such as birth defects registries, or data captured in electronic health records. Useful information may be acquired and study feasibility may be enhanced by combining existing data sources with de novo data collection regarding use of a specific medicinal product in pregnancy;
• Registries should be inclusive rather than exclusive by means of comprehensive inclusion criteria. Although retrospective enrolment may introduce bias, information entry after the pregnancy outcome is known can still be valuable. Therefore, although prospective enrolment is preferred and should be encouraged, women who wish to enroll retrospectively should not be discouraged to do so and their pregnancy outcomes should be included in the study report. The retrospective nature of such data needs to be accounted for in the analysis;
• Follow-up may include longer-term evaluation of neonates or infants for developmental maturation. In such cases and if the active substance is present in breastmilk, it is considered useful to additionally include information regarding breastfed infants. The healthcare professionals who fill data in the registry should be encouraged to record whether the mother starts to breastfeed and if so, to ask the mother regarding possible adverse reactions in her infant at each visit;
• Information regarding the existence of a pregnancy follow-up activity should be included in any mandated pregnancy-related educational materials.
• The guidance for data collection on pregnancy exposure and outcomes in P.III Appendix 1 should be followed.
P.II.B.4.2.2. Long-term pregnancy outcomes
Assessing the long-term impact of medicine use in pregnancy on the child is challenging, especially as some adverse health outcomes may not become apparent until many years after exposure. Generally, the decision as to whether or not to conduct studies into childhood needs to be based on biological plausibility and/or a combination of information from non-clinical data, clinical data (e.g. malformations, prematurity, growth retardation, foetal and neonatal outcomes), pharmacological properties, and signals regarding adverse long-term outcomes. For evaluating neurodevelopmental outcomes, the time required to develop motor and language skills (from rudimentary skills just after birth to fine motor or language skills later in childhood) mean that different measurements should be used at different ages.
Depending on the outcome of interest, follow-up may be into preschool or school age, and/or adolescence, as appropriate to reflect the neurodevelopmental outcomes mentioned. A complementary approach combining data from existing registries/databases and studies with primary data collection may be needed. A multidisciplinary approach involving epidemiological, pediatric, genetic and neurodevelopmental expertise is crucial.
P.II.B.4.2.3. Handling of bias and confounding
The design and conduct of a PASS in the population of pregnant women should consider the specific characteristics of this population that may lead to confounding. When drug utilization studies are being designed, it is useful to consider including information on such characteristics to aid the design of possible further safety studies. For examples of potential factors of interest include lifestyle factors (e.g. smoking, alcohol intake, folic acid intake, body-mass index (BMI)) or other factors relating to foetal or neonatal development (e.g. maternal pregnancy complication, prior history of negative pregnancy outcomes or pre-term birth, prescription of known teratogenic or foetotoxic medicines, maternal disease likely to cause foetal or neonatal adverse consequences). Additionally, study design should consider misclassification errors that result from incomplete recording of diagnoses or exposure, such as recall bias, as well as limitations regarding identification of competing endpoints (e.g. pregnancy loss, elective termination, miscarriage); this should also be addressed in the protocol and interpretation of the results. Attempts to minimize selection bias should be made for example by ensuring a population-based approach such as through national birth cohorts.
Study design elements that enable less biased results include the use of different comparators, sibling designs, self-controlled designs and positive and negative controls (i.e. exposure before, but not during pregnancy, or exposures in different periods of gestation). These designs may not always be appropriate for the evaluation of medicinal products with a very long half-life.
Based on the guidance in P.II.B.4., for PASS in pregnancy, proposed study designs should specifically address and justify:
• the exposure windows to be studied;
• how gestational age will be determined;
• how challenges with competing endpoints will be handled;
• whether or not, apart from the product of interest, different exposures will be combined (e.g. all products in the same pharmacological class will be treated as one type of exposure, or they will be evaluated as different exposures); and
• which pregnancy outcomes and outcomes in the child will be evaluated;
The PASS protocol should also explain how the bias due to exposure misclassification, missing data, unmeasured confounding and outcome ascertainment as well as co-exposure effects will be handled.
P.II.B.4.3. Clinical lactation studies
In cases where no human data are available on the extent of medicine transfer into breast milk, where use by breastfeeding women is expected to be common, and based on the medicinal product’s pharmacological properties, it is considered plausible that there is a risk to breastfed infants, a PK study amongst breastfeeding women should be considered. This is expected to be the case when a medicinal product is commonly used by women of reproductive age (e.g. antidepressants, anti-infectives, diabetes medications, pain medications), or when there is evidence of use or anticipated use of the medicinal product by lactating women.
Medicine concentration levels in breast milk samples should be measured and a relative infant dose calculated, to obtain information for supporting the risk assessment and provision of advice on timing of medicine intake relative to breastfeeding where this may be feasible (e.g. for short-term or single dose treatments). Moreover, data on the effect of the medicine on milk production or composition should be collected, if potentially clinically relevant.
So far, PASS in breastfed children are very rare. However, in the case of a medicine highly used in women who could breastfeed, with an unknown potential for serious adverse reactions in breastfed children, establishing safety information in the post-authorization phase should be considered as an important source of information. This may include the clinical follow-up of breastfed children whose mothers are treated with a specific medicine. Pregnancy registries in which new-born are further observed could include the collection of information on breastfeeding to allow a comparison of a group of breastfed children to those not breastfed and those breastfed in mothers who are not treated with the product of interest. In case a medicine is used during breastfeeding and questions arise regarding a potential long-term impact on child´s growth, neurodevelopment, or other adverse events with a prolonged latency, it should be considered to carry out long-term follow-up in those children.