Karen Yeo, Senior Vice President of Client and Regulatory Strategy at Certara and a leader in pharmacokinetic modeling, has spent her career advancing simulation approaches to support drug development and dosing. In this interview, she discusses how pharmacokinetic modeling is helping address the long-standing gaps in drug use during pregnancy and highlights its role in supporting dosing, research, and global health.
Can you tell us about your journey into biopharmaceuticals?
I grew up in Zimbabwe and initially studied physics and mathematics at university, developing a strong interest in analytical and quantitative thinking. During this time, I also became interested in natural products through a chemistry professor. We would collect plant materials and study their potential biological effects, which introduced me to the idea of linking science with real-world applications.
I later moved to the UK, still unsure of my exact career path, and applied for various opportunities. I was offered a PhD position in drug metabolism at the University of Sheffield, despite having no formal background in the field. This marked the start of my career in pharmacology. I then completed a postdoctoral position in childhood acute lymphoblastic leukemia and pharmacogenetics.
Following this, I became a lecturer in clinical pharmacology and therapeutics at the University of Sheffield, where I taught medical students and continued my research.
Around this time – approximately 25 years ago – colleagues at Sheffield were collaborating with pharmaceutical companies and working with large datasets generated during drug development. These companies were interested in whether this data could be used to simulate drug behavior in virtual populations.
This led to the development of a modeling platform. We began by collecting detailed physiological data from healthy volunteers, including liver size, blood flow, and cardiovascular parameters, to build a virtual representation of the human body.
The initial focus was on predicting drug–drug interactions. Pharmaceutical companies used the platform to assess how new drugs might interact with others. At the same time, we worked closely with regulators, particularly the FDA, to build confidence in simulation-based approaches. Around 15 years ago, regulatory agencies began accepting simulation data alongside clinical data, which significantly advanced the field.
Over time, the platform expanded to include a wider range of populations, including patients with organ impairment, pediatric groups, and different ethnic populations. More recently, the focus has shifted toward underrepresented groups, such as pregnant and breastfeeding women.
About a decade ago, funding from the Gates Foundation supported work on modeling antimalarial drugs, particularly for use in regions such as sub-Saharan Africa.
Today, my work focuses on continuing to engage with global regulators to expand acceptance of simulation approaches, particularly in pregnancy and breastfeeding, where clinical data are often limited.
Pregnant women are often excluded from clinical research – what are the risks of this, and why is better data needed?
A key issue is that drug labels are meant to guide dosing, but because pregnant women are often excluded from clinical trials, there is limited data to inform treatment decisions. As a result, clinicians are left to decide whether to treat – and how – without h4 evidence, which creates uncertainty for both physicians and patients.
If treatment is given, it may also be suboptimal. Pregnancy involves physiological changes that can alter how drugs are processed. For example, metoprolol is metabolized by CYP2D6, and drug exposure can be significantly lower in pregnant women due to increased enzyme activity. Without proper data, dosing may be inadequate.
At the same time, not treating can have serious consequences for both mother and fetus. Conditions such as preeclampsia require timely intervention, and even commonly used treatments, like aspirin in high-risk pregnancies, rely on evidence-based guidance.
The core challenge is the lack of data needed to make informed decisions – both about whether to treat and at what dose. This is where approaches such as modeling can help fill gaps.
A similar issue exists with other conditions, such as depression, which affects a significant proportion of pregnant and postnatal women. Without clear dosing guidance, patients may be under- or overtreated, which can impact both maternal health and fetal outcomes.
You mentioned changes in enzyme activity – what other physiological changes during pregnancy can affect drug response, beyond the presence of the fetus itself?
Key factors include changes in plasma protein binding, blood flow, body water distribution (both intracellular and extracellular), and overall body weight. Understanding how these physiological changes affect a drug’s absorption, distribution, metabolism, and elimination is essential for predicting its behavior during pregnancy.
Modeling and simulation can then integrate these changes to estimate how drug exposure may differ. However, the most important factors influencing exposure are enzyme activity, blood flow, and protein binding.
Where are the biggest research gaps, and what key unknowns remain?
We have a good understanding of how some of the major liver enzymes – such as key CYP enzymes – change during pregnancy and affect drug metabolism. However, there are still important gaps. Less is known about other pathways, including UGT enzymes and drug transporters, which play a growing role in how drugs are taken up and processed in the liver. As more drugs rely on these pathways, understanding their behavior during pregnancy becomes increasingly important.
One advantage of physiologically based pharmacokinetic (PBPK) modeling is that it can highlight where data are missing. This has led to more collaboration between industry and academic groups to generate the data needed to improve these models.
How willing are pregnant women to participate in clinical research, and how does this impact efforts to close existing data gaps?
Interestingly, my experience suggests that more women may be willing to participate in research than we often assume. For example, I was recently speaking with someone who had experienced preeclampsia, and she told me there was little discussion about treatment – she was simply advised to take aspirin.
From my work, including collaborations with researchers in Uganda and elsewhere, I’ve found that many women are open to taking part in studies, provided they are well informed. The key factor is clear communication between clinicians and patients.
Ultimately, it comes back to data. Clinicians can only guide patients based on the evidence available, and improving that evidence base is essential to support informed decision-making.
How is simulation changing the conversation around research in pregnant women, and what key advances have you seen in this area?
Modeling can play an important role even before clinical studies begin. Using existing drug models, we can simulate how pregnancy might affect dosing and use this to help design more informed clinical trials.
One challenge with pregnancy studies is that they often include small numbers of participants, making it difficult to capture variability in drug exposure. In addition, physiological changes differ across trimesters, and these are not always fully represented in clinical data. Modeling can help address these gaps by simulating changes over time and across different scenarios.
Recent regulatory developments are also encouraging this approach. Guidance such as ICH E21 highlights the inclusion of pregnant and breastfeeding women in research and supports the use of modeling to inform decision-making. This is important because industry often looks to regulators for direction, and endorsement of these methods is likely to increase their use.
Historically, pregnancy studies have been conducted after a drug is approved, which can limit their priority. However, as regulators begin to require earlier planning – similar to what has been done in pediatric drug development – this may shift how companies approach research in pregnant populations.
Overall, modeling can support dose selection, complement limited clinical data, and help explore scenarios that are not easily studied in trials. This is where it offers particular value.
How have modeling technologies improved, and how can we ensure they avoid reinforcing existing biases in research – particularly those affecting women?
These models are not developed in isolation. They are used and evaluated by a large consortium of pharmaceutical companies, as well as many independent academic groups. In fact, they are already being applied in clinical practice – for example, to inform dosing recommendations in pregnancy within national formularies. This broad use helps ensure that the models are continually tested and improved.
Feedback from independent users is critical. When models do not perform as expected, this is openly addressed and used to refine them further. Ultimately, however, clinical decisions remain with physicians, and models are intended to support – not replace – their judgment.
As modeling becomes more widely used, data from various sources – including clinical studies, pharmacovigilance, and case reports – are being integrated into shared databases. Some of these resources are now also incorporating modeling outputs, which helps expand the available evidence base.
Building trust is key. To support this, training and workshops are increasingly being offered to clinicians, helping them understand how models work and how to interpret their results. This hands-on approach can improve confidence and transparency.
Importantly, modeling is best viewed as a complement to clinical data. It can help explore scenarios that are difficult to study directly and provide additional context, but it is not a replacement for clinical research.
With so many simulation and AI-based solutions now available, what distinguishes a robust, reliable model from a less credible one, and what should drug developers look for when evaluating them?
I think one of the key things companies should look for is regulatory support. Guidance such as ICH E21, which endorses the use of PBPK modeling, reflects growing confidence from regulators and is an important signal of credibility.
Another important factor is the strength of the evidence base. An increasing number of peer-reviewed publications and real-world case studies – across both industry and clinical settings – helps demonstrate how these models perform in practice.
However, having data is not enough; effective communication and dissemination of that information are essential so that both clinicians and developers understand how to use these tools appropriately.
In terms of AI, there is a lot of interest, but its role is still evolving. One clear area where it can help is in bringing together large volumes of data more efficiently. That said, expert curation remains essential to ensure data quality and reliability. AI can accelerate the process, but it cannot replace scientific oversight.
Beyond pregnancy, where else do you see the greatest opportunities for simulation in drug development?
At present, PBPK modeling is used from the early stages of drug discovery and is applied extensively to assess drug–drug interactions, as well as increasingly to specific patient populations.
However, there is potential to go much further. As more individualized data become available, there is growing interest in using these models to support personalized dosing. This concept is often linked to “digital twins,” where patient-specific physiological data are used to simulate how a drug may behave in an individual.
If these approaches continue to develop, they could allow dosing to be tailored more precisely – for example, based on factors such as genotype or stage of pregnancy. In some settings, clinicians are already beginning to explore this type of individualized approach.
Overall, the field is evolving rapidly, and modeling – alongside advances in AI – may play an increasingly important role in shaping more personalized and data-driven drug development.
Beyond industry needs, what do you see as the broader public health importance of this work?
One important point to emphasize is the global health perspective. While much of this discussion focuses on drug development, a key driver for me is the impact on underserved populations.
Having grown up in Southern Africa, I’ve seen how conditions such as malaria, HIV, and tuberculosis affect large numbers of women, including those who are pregnant. In many cases, these women lack access to appropriate treatment or evidence-based dosing information.
This highlights the importance of making clinical research more inclusive. Initiatives such as the WHO task force on including pregnant women in clinical trials are an important step forward.
Ultimately, while industry and drug development are important, improving access to safe and effective treatment for these populations remains a central priority.
Newsletters
Receive the latest pharmaceutical news, personalities, education, and career development – weekly to your inbox.
