How the CIOMS guidelines contribute to fair inclusion of pregnant women in research

Abstract As early as 2002, CIOMS stated that pregnant women should be presumed eligible for participation in research. Despite this position and calls of other well‐recognized organizations, the health needs of pregnant women in research remain grossly under‐researched. Although the presumption of eligibility remains unchanged, the revision of the 2002 CIOMS International ethical guidelines for biomedical research involving human subjects involved a substantive rewrite of the guidance on research with pregnant women and related guidelines, such as those on fair inclusion and vulnerability. However, close reading of the guidelines reveals morally relevant different approaches to fair inclusion of pregnant women and other under‐represented groups, such as children and incompetents. Where CIOMS sets out that children and adolescents must be included unless a good scientific reason justifies their exclusion, no such claim of having to justify exclusion appears in the guideline on pregnant women. Instead, CIOMS claims that research relevant to pregnant women’s health needs must be promoted. This paper analyses how and to what extent the guideline on pregnant women differs from other guidance on fair inclusion in the document. Accordingly, the paper evaluates to what extent the current phrasing may contribute to fair inclusion of pregnant women in research. We will conclude that a system change towards a learning health system is essential to break down the status quo of knowledge generation in the field of medication use during pregnancy and argue that the CIOMS guidelines allow for this system change.

during pregnancy without prior testing in pregnant women, which resulted in unforeseen teratogenic effects with severe birth defects in over 10,000 children. 2 Even though neither tragedy involved clinical research, these events had a great impact on the research community, which at the time was already characterized by a determination to protect allegedly vulnerable populations, including pregnant women, from research participation. The protectionist approach is one of the reasons for the existing precautionary attitude with regard to including pregnant women in clinical research today. 3 Although concerns about fetal well-being are valid, at the same time there is a need for evidence-based information on medications and treatments for pregnant women, because 'pregnant women get sick, and sick women get pregnant'. 4 In the absence of evidence-based knowledge, clinicians may have to prescribe offlabel medications without evidence or based on contradictory evidence, 5 or clinicians or pregnant women themselves may choose to discontinue medically important medications. 6 The result of under-representation of pregnant women in clinical research is a harmful situation, leaving pregnant women at risk for potentially avoidable therapeutic incidents. 7 For example, poorly treated asthma and untreated depression are problematic for pregnant women and fetuses: both are associated with premature birth, low birthweight and fetal growth restriction and, in the case of asthma, a higher risk of hypertension and pre-eclampsia. 8 Physiological changes during pregnancy alter the way that drugs are processed by the body and the ways that drugs act on the body in a fashion difficult to predict from the pharmacokinetics and pharmacodynamics in men and non-pregnant women. Moreover, information about teratology and toxicity is often difficult to extrapolate and interpret from preclinical data or studies in non-pregnant humans. 9 Gathering conclusive data to enable evidence-based thera-peutic decisions for pregnant women therefore requires research in the population of pregnant women in order to develop effective treatments for pregnant women with acute or chronic obstetric and non-obstetric illnesses. 10 Inclusion of pregnant women in research has been promoted in the last decades by, among others, medical ethicists, pharmacolo-  13 At the same time, the number of pregnant women taking medications and the number of medications have both increased.
The total percentage of pregnant women who take medications including off-label medications may currently be as high as 84-99% 14 and the number of pregnant women taking four or more | 379 VAN DER GRAAF et al. medications more than tripled over the last three decades, common ones being antibiotics, asthma medications and anti-nausea medications. 15 Although these data reinforce the need to study safety and efficacy of drugs in pregnant women specifically, pregnant women generally remain excluded from research. To illustrate, exclusion of pregnant women is common practice in industry-sponsored phase IV research. 16 Moreover, a 2014 review demonstrates that between 1960 and 2013, within the pharmacokinetic clinical trials found, less than 1.29% were conducted for pregnant women, and the ones that were undertaken had a strong focus on acute labor and delivery issues. 17 For a population that is under-represented in research, it is at least of relevance to understand what ethical boundaries are set by international ethics guidance documents and how and when these documents impede or promote research with pregnant women. Since this Bioethics Special Issue focuses on the recently revised CIOMS international ethical guidelines, in this paper we will limit our scope to the CIOMS guidelines. 18 We will analyse how these guidelines may help to break down the status quo pertaining to the evidence base and hence contribute to fair inclusion of pregnant women in research.

| FAIR IN CLUS I ON AND THE CI OMS G UIDELINE S
At least three aspects of the revised CIOMS guidelines may affect fair inclusion of pregnant women in research. Shifting the underlying rationale of non-discrimination from the pregnancy guideline to the guideline on vulnerability (Guideline 15) can be explained by the extensive rewrite of the vulnerability guideline itself. The vulnerability guideline no longer classifies entire groups of people as vulnerable, but instead asks researchers and research ethics committees (RECs) to look at characteristics of a person or group that may render the person or group vulnerable and to install special protections tailored to mitigate the risks associated with these characteristics. 19 Pregnancy per se is no such morally relevant characteristic. Recent literature substantiates the non-vulnerability statement. 20 Moreover, even the Common Rule, generally perceived as a more conservative document due to its classification of pregnant women as vulnerable, has changed its position and eliminates the qualification of pregnant women as vulnerable as of July 2018. 21 In sum, CIOMS' explicit claim to not regard pregnant women as vulnerable may help to strengthen research into the health needs of pregnant women.

| Promote research relevant to the health needs of pregnant women
The CIOMS guidelines now claim that 'research designed to obtain knowledge relevant to the health needs of the pregnant and breast- In sum, the claim that research relevant to the health needs of pregnant women must be promoted focuses on a broader group of stakeholders to protect the health interests of pregnant women than protection by RECs alone.

| Clarified levels of acceptable research risks
In the 2002 version of the CIOMS guidelines the level of acceptable research risks for pregnant women and fetuses remained unclear. 28 The guideline left it up to pregnant women themselves to decide whether research risks were acceptable: 'the decision about acceptability of risk to the fetus should be made by the woman as part of the informed consent process'. In 2016, the risk threshold has been clarified and reads as follows: For research interventions or procedures that have the potential to benefit either pregnant or breastfeeding women or their fetus or infant, risks must be minimized and outweighed by the prospect of potential individual benefit. 22  Furthermore, consistent with Guideline 4 on research risks and potential benefits, we argue that the surgery should be able to go forward if: 1. The risks to the fetus are outweighed by the potential benefits to the fetus and the woman. Even if the risks to the woman are high, the net risks for both the woman and the fetus may be regarded as acceptable in light of the direct physical benefits to be expected for both the woman and the fetus.

2.
The risks to the woman are reasonable in relation to the potential benefits for the fetus and the social value of the research.
3. The risks to the woman respect some upper risk limits that researchers and RECs should further specify in concrete cases.
Analogies that RECs may take into account are comparable upper risk thresholds in the case of altruistic organ donation and research with competent consenting participants.
In sum, the now clarified threshold for acceptable research risks may help to prevent pregnant women being unnecessarily excluded from research. Evidently, fair inclusion of pregnant women will not be achieved by clarification of guidance documents alone. For instance, while guidelines such as CIOMS have widened the possibility for research with pregnant women, interpretation and perception of such documents is something different. Mechanisms of over-and underprotection play a role. An example of overprotection is that healthcare professionals may be reluctant to include pregnant women because they perceive a study to be high risk, even though an REC classifies the study as low risk. 31 Underprotection also applies in the sense that research and care have become totally distinct worlds in the field of drug use in pregnancy. Paradoxically, pregnant women are excluded from research for protection, while they run high risks in routine practice, yet without learning from this exposure to protect future medication users. A potential solution for both over-and underprotection of pregnant women taking medications during pregnancy is to transform the field into a so-called learning health system (LHS). As we argued above, since CIOMS Guideline 19 is structured as a guideline that is primarily focused on priority and agenda set-  VAN DER GRAAF et al. In 2006, the U.S. Institute of Medicine called on healthcare leaders to transform their health systems into 'learning health systems', meaning a system that is designed to generate and apply the best evidence for the collaborative healthcare choices of each patient and provider; to drive the process of discovery as a natural outgrowth of patient care; and to ensure innovation, quality, safety, and value in healthcare. 32 The idea of transforming health practices to LHSs increasingly attracts attention, but remains a rather theoretical exercise. It has primarily been discussed in the context of quality improvement and comparative effectiveness research, 33  form. Many argue that these trials can only start after a trial in the non-pregnant population has been performed. Often, this approach results in a situation that no such trial in pregnant women is performed. 35 Furthermore, long-term follow-up of neonates is usually essential to reveal drug-related risks. Also, RCTs are often monodisciplinary, where research outcomes and risks for the mother and future child should be considered multidisciplinary (obstetricians, pediatricians and others, such as rheumatologists and psychiatrists) to optimally determine safety and efficacy of these drugs. Therefore, methods other than RCTs are of utmost importance to speed up the knowledge generation in this field. Despite the availability of promising alternatives, such as registry-based studies, the current research infrastructure does not allow for these alternatives, since solid and comprehensive registries combined with pregnancy-tailored trial methodology are lacking. Moreover, there is limited insight into the advantages and disadvantages of alternative designs in the field of medication use in pregnancy. 36 Another crucial factor to realize transformation to an LHS is multi-stakeholder involvement. Pharma, physicians, patients, funders, regulators and others should all support the idea that we have to leave the current paradigm of knowledge generation in the field of pregnancy. Since an LHS drives on co-creation, meaning that scientists, clinicians, patients, and others collectively generate new knowledge in a dynamic interaction, 37 the LHS seems a promising way forward to realize the transformation. Co-creation implies that stakeholder involvement is essential to further health knowledge that leads to meaningful outcomes, for instance by co-deciding on research questions, methods, endpoints and dissemination of results.

|
In order to transform the field of pregnancy into an LHS, we think that the following features are essential. First, we have to create an, ideally global, meta-registry. Currently, several databases, cohorts and registries exist with unique data regarding pregnancy, adverse drug reactions and the like, such as EUROmediCAT 38 and ENTIS. 39 However, these data sources are not designed to combine data on the drug used, endpoints to determine efficacy and safety of the drug used to treat, birth defects and potential long-term consequences of use of the specific medicines during pregnancy. This meta-registry is essential for clinical practice, since physicians and women are not only interested in teratogenicity, but also in adverse maternal and fetal events, related to either using or not using a drug.
For instance, the risk on teratogenicity should be weighed against risks of serious infections for the neonate when medication is discontinued. Evidently, there is a myriad of challenges regarding the development and implementation of a global meta-registry, such as the linking of data of pregnant women to those of the fetus and child and differences across registries in terms of terminology, formatting, and labelling. For managerial reasons it is therefore reasonable to start small and accordingly publish the results and share them more broadly.
Second, we have to develop appropriate methodology to gather knowledge in the field of pregnancy. Conventional RCTs are typically difficult to perform in pregnant women. Because of additional fetal risks associated with including pregnant women in clinical research and the altered ways in which drugs are processed by the pregnant body, pregnant women cannot be treated as an ordinary subgroup in the various phases of traditional drug development. A meta-registry creates an opportunity to allow for alternative research designs that are able to capture long-term follow-up data, for instance registrybased RCTs and Bayesian designs. However, literature on appropriate methodology for research in pregnant women is scarce and potential research designs profoundly need further study and evaluation. 40 Third, we have to take a co-creationistic approach. One explanation for the continuing status quo is the lack of ownership of the