Model-informed antenatal doses, as outlined in this study, referred to medication doses tailored for pregnant women and/or their unborn children, derived from pharmacokinetic (either pop-PK or PBPK) simulations, alongside clinical and/or animal data. The latter doses and underlying evidence should be reviewed by a multidisciplinary committee of experts and patients before endorsement for clinical use (Koldeweij et al., 2024a).

This cross-sectional study was aimed towards the two main groups of stakeholders involved in shared decision-making on antenatal dosing, namely, HCPs and pregnant women. Two questionnaires, one for each targeted group, were designed (Supplementary Material S1). Both questionnaires comprised demographic questions followed by statements drawn from focus groups and interviews on the perceived barriers and facilitators for model-informed dosing in pregnancy previously conducted among HCPs and pregnant women in the Netherlands (Koldeweij et al., 2024). The proposed model-informed dosing approach was introduced briefly on the survey’s opening page, with more details provided later in the survey. Explored themes in both surveys included respondents’ knowledge, information needs and search behaviours with regards to medication dosing in pregnancy, their views on important considerations for dosing, their willingness-to-use model-informed antenatal doses and their information needs and preferred features in this regard. HCPs were additionally asked about their current dosing practices. The questionnaire for HCPs comprised 39 questions and the questionnaire for pregnant women 27 questions. Each statement was associated with either a four-point Likert scale (strongly agree, somewhat agree, somewhat disagree, fully disagree) or a dichotomized response (agree, disagree). Some statements additionally included a ‘not applicable’ response. Optional text boxes were included for respondents to explain their answers. The questionnaires were designed for access on laptops and smartphones and were available in English and Dutch. Questionnaire content and language was tailored in complexity for each group, drawing on experience gained from the focus groups that informed this survey. Castor EDC (https://www.castoredc.com/) was used for questionnaire design and administration.

Each questionnaire was user-tested for aspects including content, length (aiming to not exceed 10 minutes), user-friendliness, intelligibility and technical functioning. User-testing was performed by at least three clinicians (for HCP surveys) and three pregnant or recently pregnant women (for surveys targeting pregnant women) who were fluent in the survey language (Supplementary Material S2). The surveys were additionally tested by three experts in online surveys on related topics. Several iterations of the questionnaires were made to address user feedback.

HCPs were eligible if they were practicing in a European country (OECD). Women were eligible if they were currently pregnant or had been pregnant in the last 3 years, if they were more than 18 years old and resided in a European country. All participants had to be proficient in English or Dutch. The surveys were disseminated through open hyperlinks shared with professional societies of HCPs involved in making decisions or providing information on dosing in pregnancy, including physicians across various specialties, pharmacists, pharmacologists and midwives. Additionally, they were shared on websites and social media for pregnant women across European countries, primarily the Netherlands and Belgium. The links to both surveys were additionally placed on the websites of several national teratology information services (TIS). Posters and flyers with QR codes were left at a small number of outpatient clinics in the Netherlands. No incentive for participation was offered. More information on dissemination channels can be found in Supplementary Material S3. Other than the TIS websites, the links to each survey were disseminated through separate channels so that it was unlikely that a respondent would participate in both surveys. The surveys remained active between 8 September and 30 November 2022.

No sample size was pre-emptively determined. Given the open survey dissemination a response rate could not be calculated, and the lack of prior cross-sectional studies on a similar topic made it difficult to estimate the distribution of values for each survey item a priori. All surveys for which demographic information was available and at least one statement had been rated were included. Survey item ratings were analysed individually. The ratings were dichotomized, combining “strongly agree” and “somewhat agree” into the category “agree” and employing a similar method for “disagree.” Statements with a “non-applicable” response were trichotomized. For relevant questions, we broke down the results by five subgroups of HCPs: community pharmacists, hospital pharmacists or clinical pharmacologists, obstetricians-gynecologists, general practitioners and other medical specialists. The analysis was descriptive, using Excel 2013 version 16.68. “Survey participation” describes the ratio of participants who filled in any information divided by the number of survey visitors. “Survey completion” corresponds to the ratio of users finishing the survey divided by those agreeing to participate.

The study protocol was assessed by the Medical Ethics committee of the Radboud University Medical Centre (2021-13417) and was not subject to the Medical Research Involving Human Subjects Act. Information on the survey and underlying goals were shared with potential respondents in the introductory pages of the survey. Potential respondents were then asked if they wished to participate and agreed with the use of their anonymous data (yes or no button with exit option, Supplementary Material S1). Participation beyond this point implied informed consent. Responses were not identifiable, IP addresses were not recorded.

It is recognized that not all women were pregnant at the time of completing the survey and that not all pregnant people identify as women (Rioux et al., 2022); however for brevity we use the term ‘pregnant women’ to describe our target population. Reporting of the study was guided by the CHERRIES checklist (Supplementary Material S4) (Eysenbach, 2004).

In total, 608 HCPs from 15 countries and 794 women from 12 countries participated. Among HCPs (Table 1), 56% worked in the Netherlands and 25% in Belgium. HCPs comprised 37% pharmacists or clinical pharmacologists, 26% obstetricians-gynecologists, 11% midwives and 26% HCPs from other specialties. HCPs had a median of 13 years of professional practice and worked across a variety of settings, ranging from primary care to academic hospitals. Among pregnant women (Table 2), 49% resided in the Netherlands and 32% in Belgium. A large majority (96%) described themselves as white. Forty-three percent of women had obtained a university degree and 31% were healthcare professionals. 33% were currently pregnant. Eighty-five percent of women had used medication during their pregnancy, and 38% had a chronic or recurring condition that required medication. The survey participation rates were 14% among HCPs and 35% among pregnant women. Survey completion rates were 70% for HCPs and 68% for pregnant women (Supplementary Material S5).

Just over half of HCPs (56%) indicated regularly adjusting doses for pregnant patients (Table 3). This proportion varied by medical specialty, from 48% of obstetrician-gynecologists to 74% among other medical specialists (Supplementary Material S6). Most HCPs (73%) reported discussing medication doses with pregnant patients. While 83% of HCPs indicated concerns about suboptimal pharmacological care for pregnant women, two-thirds prioritized fetal safety over maternal efficacy when prescribing antenatal medications (Figure 1). Virtually all HCPs (97%) indicated that better antenatal dosing information was needed. The same proportion of HCPs believed that an evidence-based resource with model-informed doses would greatly enhance the quality of antenatal pharmacotherapy. On the other hand, HCPs reported limited comprehension of pharmacokinetics in pregnancy. Only 34% of obstetrician-gynecologists, 33% of other medical specialists, 39% of general practitioners and 51% of clinical pharmacists and pharmacologists considered their understanding of pharmacokinetics in pregnancy and the impact on medication safety and efficacy sufficient. Additionally, HCPs reported limited knowledge of pharmacokinetic models. While two-thirds of HCPs agreed that they knew what such models entailed, that only 21% strongly agreed with this statement (Supplementary Material S5).

Almost all women (99%) wanted to know which dose to use during pregnancy (Table 4). While sixty-five percent of women had searched information about medication doses while pregnant, only 40% stated that they could easily find clear and helpful information in this regard. Nearly two-thirds of women reported being aware of the potential need for dose adjustments in pregnancy based on their changing bodies. However, only 52% knew that medication doses had generally not been researched in pregnant women. Virtually all women indicated a desire not only to receive information on antenatal dosing (96%) but also to be involved in antenatal dosing decisions by their HCP (96%). Almost half of women (49%) described maternal efficacy and fetal safety as equally important considerations in this regard, with 50% assigning more weight to fetal safety. The remaining 1% assigned more weight to maternal efficacy (Figure 1). Eighty-two percent were open to using a higher dose upon their HCP’s recommendation. Of the remaining 18% who were reluctant to use increased doses, virtually all (98%) cited fetal safety concerns. However, 87% of women would be willing to reconsider if their HCP provided a rationale for altered dosing.

When initially asked, 92% of HCPs stated that they would be willing to follow model-informed antenatal doses. Respondents reiterated this view after receiving more insights on the proposed approach, with 85% of them indicating that they would be willing to advise a higher dose to a pregnant woman based on evidence from pharmacokinetic models (Figure 2). This high willingness-to-use model-informed antenatal doses was found across HCP specialties (Supplementary Material S6). Endorsement of the proposed resource and methods by recognized institutions was perceived as helpful in this view by 96% of HCPs.

Among women, 77% would be willing to use doses issued based on evidence from computer models (Figure 2). When asked about potential advantages of such models, 96% of pregnant women acknowledged the ability of pharmacokinetic models to guide the selection of more effective maternal doses. A similar proportion welcomed such models’ ability to provide information on fetal exposure. Potential concerns included the lack of individual applicability of model-informed doses, computational errors and insufficient evidence on the predictive ability of models.

HCPs’ key information needs included: information on fetal safety (98%) and on the maternal and fetal risks and benefits of a dose (97%), guidance on considerations for individual dose adjustments (97%) and on how to detect underdosing or toxicity (94%), alongside an overview of the physiological and pharmacokinetic changes that may justify dose alterations (92%). Over half of HCPs (51%) appeared unwilling to follow model-informed antenatal doses without information on fetal exposure. In addition, 88% of HCPs indicated that they would consult evidence on model-informed antenatal doses, starting with information on the quality of the available evidence (97%). If the evidence for a dose stemmed from a pharmacokinetic model, 62% of HCPs would like to access information about the model. Details that HCPs wanted to access in this regard included the general assumptions of the model (94%), information on how fetal exposure was determined (93%), and information on model validation (89%). Just half (52%) of HCPs reported wanting access to the model itself. Lastly, eighty-seven percent of HCPs agreed that information for patients should be provided. Key considerations for integrating medications in the proposed antenatal dosing resource according to HCPs included the frequency of use of the medication in pregnancy (74%) and the consequences of underdosing or overdosing (78%).

Eighty-six percent of pregnant women desired to know more about how computer models could be used to guide the establishment of pregnancy-adjusted doses. Additionally, 97% of them would like to access information about the relationship between changes in their bodies and their required amount of medication during pregnancy. Pregnant women wanted to access information on model-informed doses and the underlying approach on a website (95%) as well as through their HCP (97%).

This international, cross-sectional study explored the perspectives of HCPs and pregnant women on the acceptability and desirable features of model-informed antenatal doses. Firstly, it revealed that most HCPs and pregnant women deemed the existing information on antenatal dosing inadequate and believed that enhanced information, including from pharmacokinetic models, may greatly improve the quality of maternofetal pharmacotherapy. Secondly, while almost two-thirds of HCPs reported advising or performing antenatal dose adjustments, less than half felt that they had sufficient guidance or knowledge to do so. Thirdly, despite their limited familiarity with pharmacokinetic models, over 90% of HCPs and three-quarters of pregnant women appeared willing to follow model-informed antenatal doses. This contrasted with initial concerns expressed by certain HCPs participants to the focus groups that informed this study regarding the credibility and feasibility of model-informed doses for clinical use. While their information needs differed, both HCPs and pregnant women expressed a desire to better understand dosing rationales, with most participants in both groups wanting to access information on pharmacokinetic models. Other facilitators for the adoption of model-informed antenatal doses according to HCPs included endorsement of the methods by recognized institutions, guidance on applying model-informed antenatal doses for individual patients, and various levels of access to the evidence. A potential barrier related to both groups’ unmet information needs regarding fetal safety, a knowledge gap that may only partly be addressed by pharmacokinetic models according to participants. Lastly, this study highlighted diverging views among HCPs and pregnant women regarding shared decision-making on antenatal dosing. Most pregnant women desired more information and active participation in antenatal dosing decisions. In addition, they tended to value maternal efficacy and fetal safety equally. In contrast, fewer HCPs reported discussing antenatal doses with pregnant women, and most HCPs prioritized fetal safety over maternal efficacy.

By focusing on the two main groups of stakeholders for a resource with model-informed antenatal doses, this study delivered valuable insights on the acceptability and preferred features for such a resource among potential end-users. Through its broad thematic scope, including HCPs’ dosing practices and HCPs and women’ views on available dosing information, this study was the first to explore the perceived relevance of model-informed antenatal doses against the status quo. Although selection bias may be present, as discussed below, the design of the survey questionnaires, drawing on insights from previous focus groups among pregnant women and HCPs in the Netherlands, and the large number of participants likely increased the generalizability of the obtained results.

However, several limitations apply. First, the survey was mainly disseminated in countries where Dutch or English were widely spoken and our findings primarily originated from the Netherlands and the Dutch-speaking part of Belgium. Second, despite the general scope of the dissemination channels used, study participants may have had an above-average degree of knowledge and/or affinity with (model-informed) dosing in pregnancy. For pregnant women especially, limited ethnic diversity, higher-than-average educational levels and an elevated proportion of medical professionals (Centraal Bureau voor de Statistiek, 2023; Hoeveel inwoners hebben een herkomst, 2023; Ministerie van Onderwijs Cultuur en Wetenschap, 2023), may have influenced the insights gathered. However, the incidence of medication use and chronic conditions among pregnant participants aligned well with external data (Lupattelli et al., 2014; Houben et al., 2020; Laroche et al., 2020). Participants’ potentially above-average interest in antenatal dosing and/or pharmacokinetic modelling due to self-selection could either have manifested as a more favorable stance towards model-informed antenatal doses, or conversely, as more skepticism. A cross-sectional study conducted by Nordeng et al. among 1,793 women in Norway found that higher education levels were associated with a significantly increased risk perception regarding antenatal drug use and choosing not to use a drug during pregnancy (Nordeng et al., 2010). The latter pattern was observed in the focus groups that informed this study: individuals with a deeper grasp of pharmacokinetic models, such as clinical pharmacologists, tended to raise more questions about the modelling approach. The lack of questions, in this survey, on challenges associated with data collection for model development and validation (Kluwe et al., 2021), may also have resulted in limited awareness, among participants less familiar with pharmacokinetic modelling, of this concern for model credibility. Questionnaire content may thus have influenced participants’ readiness-to-use model-informed doses, especially among those less knowledgeable on this subject. Other limitations included the exclusion of partners of pregnant women and the absence of formal mechanism to prevent double entries in the survey.

The limited reports on the clinical implementation of model-informed dosing were expert reviews that examined the use of this approach in non-pregnant populations (Darwich et al., 2017; Hartman et al., 2020; Kluwe et al., 2021). These studies generally focused on technical requirements rather than end-users’ preferences, as explored in this study. Beyond examining the acceptability and preferred features of model-informed antenatal doses, this study clarified stakeholders’ perspectives on antenatal dosing, a poorly studied area. While previous work noted HCPs’ limited awareness of specific antenatal dosing needs (Westin et al., 2018), most participants in our study believed that there was a lack of high-quality information on antenatal doses and felt that their knowledge was insufficient to make informed antenatal dosing decisions. To our knowledge, pregnant women’s information needs and preferences with regards to antenatal dosing have not previously been explored.

Prior research into stakeholder views regarding antenatal pharmacotherapy focused on medication use rather than dosing. A cross-sectional study in Belgium found that many pregnant women sought pharmacological information online but that fewer discussed their findings with a HCP (Ceulemans et al., 2019). An international review reported that most pregnant women searched health information, including on medication, online, although the aspects of medication researched were not described (Sayakhot and Carolan-Olah, 2016). An earlier investigation into HCPs and pregnant women’s perceptions of risks from antenatal medication use found that both groups were highly sensitive to potential risks in the face of scarce evidence (Widnes and Schjøtt, 2017). The importance each group placed on those risks compared to potential benefits from antenatal medication use was not quantified. However, both HCPs and pregnant women in this study were deemed to place undue emphasis on teratogenicity compared to the potential maternal and fetal benefits from medication use according to the study authors, a finding aligning with previous research (Nordeng et al., 2010; Lynch et al., 2018). By contrast, other researchers reported that pregnant women, despite concerns about fetal safety, also cared about the maternal effects of medication (Lynch et al., 2018). This viewpoint was also conveyed by several pregnant participants to the focus groups informing this study. The present study confirmed and quantified the differing views of HCPs and pregnant women regarding the importance of fetal safety compared to potential maternal benefits when making antenatal dosing decisions, with two-thirds of HCPs prioritising fetal safety while nearly half of pregnant women valued both equally. Despite previous anecdotal accounts of a tendency among HCPs to prioritise fetal safety (Widnes and Schjøtt, 2017), this quantified assessment among both groups represents a new finding within the limited research on HCPs’ and pregnant women’s decision-making on antenatal medication, warranting further exploration. The need to better align HCPs’ dosing practices with pregnant women’s preferences was further underlined by both groups’ diverging views regarding shared decision-making on antenatal dosing, another topic that has received little scrutiny.

Model-informed antenatal doses established through the MADAM project will be published and made accessible on the websites of the Dutch Teratology Information Service and the Dutch National Formulary so they can inform clinical practice in the Netherlands (Koldeweij et al., 2024b). Next steps as part of this proof-of-concept will focus on disseminating established doses internationally, for example, through international TISes. Tailored information for HCPs and patients on the approach used and underlying premises will be made available on a separate website (https://melinda-dosing.com/).

Although model-informed dosing may help deliver better-evidenced doses in pregnancy, this study also suggests that it may not fully address the perceived challenges surrounding antenatal medication. Improving awareness of potentially altered antenatal dosing needs among HCPs and pregnant women, while expanding the evidence on the pharmacokinetics, efficacy and safety of antenatal medications, are also essential. Addressing these gaps may require multiple interventions including more systematic pharmacokinetic studies in pregnancy (Quinney et al., 2023), broader enrolment of pregnant women in clinical trials (Eke et al., 2019; Eke et al., 2021; Ren et al., 2021), education of HCPs and women on pharmacokinetics in pregnancy (Westin et al., 2018), alongside the deployment of evidence-based information resources and shared decision-making aids on antenatal medication use and dosing (Eke et al., 2019). To enhance these efforts, future research may involve exploring the characteristics of healthcare practitioners and pregnant women more or less inclined to follow model-informed doses in pregnancy, and for pregnant women to be involved in antenatal dosing decisions. Insights from this study will be incorporated in the design of an international, model-informed antenatal dosing resource. Lastly, this study highlighted a need to alter risk perspectives associated with antenatal drug use among HCPs and pregnant women. This may entail further exploring how these views are influenced by individual features potentially related to understanding of these risks, particularly among HCPs, and how these views may differ from those of pregnant women. These insights could inform tailored knowledge dissemination efforts including dedicated trainings on risk evaluation and communication for HCPs, both as part of medical school curriculums and continuous education. Insights from this study may also guide the development of enhanced tools for risk communication and shared decision-making integrating pregnant women’s values (Eke et al., 2019; Kennedy et al., 2020)). Such efforts may also help more broadly address a potential disconnect between the expectations and behaviors of HCPs and pregnant women regarding antenatal pharmacotherapy, as revealed by this study.