Blogs

This Session, “Integrating In Vitro Data and PBK Modeling for Exposure-Led Risk Assessment,” chaired by Nynke Kramer, Wageningen University, The Netherlands, was not centered around the well-trodden “hazard” arm of risk assessment, but instead started with the equally critical “exposure” arm. The four talks brought together speakers from academia and industry representing major companies and European initiatives. Importantly, it examined integration of realistic human exposures with physiologically based pharmacokinetic (PBPK) models and in vitro to in vivo extrapolation (IVIVE). The combination of these methods, it asserted, was the key to efficient and exposure-led risk assessment in a world understandably hesitant about performing novel in vivo studies.

Building a Bridge from First Principles

Dr. Kramer opened with a grounding question: “How do we parameterise human-relevant PBPK models without relying on animal data?” Her answer was pragmatic. We already have many of the pieces; we just need to connect them better. She walked through how in vitro ADME data on protein binding, permeability, metabolic clearance, and transporter activity can empower in silico predictions from the ground up. Not fitted to animal data but constructed mechanistically.

There was particular focus on distribution, describing where a chemical accumulates not only in a human body but also in vitro. In this respect, her case studies from PFAS to cisplatin covered both plastic and protein binding, accumulation in cells and medium, and transport into target tissues. Although she did not shy away from the remaining uncertainty in these models, her talk outlined that biologically interpretable approaches are no longer theoretical and can already approximate real-world toxicity surprisingly well.

Pushing the Boundaries of Assumptions

If Dr. Kramer’s talk was about building the framework, Marjory Moreau, ESQlabs, propelled it into one of the hardest contexts: pregnancy. She outlined enormous physiological changes even in the first trimester: organ volumes shift, enzyme activity fluctuates, and protein binding adjusts. Furthermore, beyond alterations in maternal modeling, the fetus was also introduced as a second, interconnected system essential to consider if we are to predict developmental and reproductive toxicity (DART).

The presentation revolved around a tiered QIVIVE strategy, which was able to incorporate available data as needed. Using a carbaryl case study, Dr. Moreau showed how moving up tiers reduces uncertainty and refines risk prediction. One particularly striking point was that, for developmental neurotoxicity (DNT), timing matters as much as dose. A narrow window in pregnancy can define risk, and Cmax at the “wrong” moment can be far more consequential that a higher average exposure for pregnant women and their babies.

Not All Toxicants Are Small Molecules

It was not just unique contexts like pregnancy considered in this session but also unique exposures. Zhoumeng Lin, University of Florida, presented a highly engaging outline of nanoparticles in the context of exposure-led risk assessment. Current PBPK modeling assumptions, such as simple partition coefficients, endocytosis-driven uptake, and strong dependence on particle size, need to be reevaluated in this setting, and the talk outlined equilibrium-based models and time-dependent uptake functions.

Dr. Lin showed that integrating in vitro uptake and retention data into PBPK frameworks is essential but still emerging. Compared to small molecules, this space is relatively immature, more uncertain, and dependent on bespoke data. If we are to continue to integrate in vitro data and in silico approaches for risk assessment, we need to also consider niche situations and places where current assumptions break down.

What If We Have No Data at All?

Ans Punt, Unilever, closed the talks with an uncomfortable but highly relevant question: “What if we have no in vivo data at all with which to validate our predictions?” Her answer was simple; we need to stop treating in vivo experiments as a gold standard and instead build confidence from the ground up using multiple lines of evidence. To outline this approach, sunscreen ingredient benzophenone-4 (BP-4) was used as an example.

Even with minimal human data, Dr. Punt showed that you can still arrive at a protective, decision-relevant conclusion. She emphasized that confidence is not the product of a single model but instead comes from multiple sensitivity analyses, comparison across platforms, and read-across using mechanistic analogues. The latter was exemplified particularly elegantly using an antiviral drug that was not structurally similar but shared transporter biology with BP-4 and could be used to validate renal clearance behavior.

Panel Discussion

The discussion revolved around a deceptively simple question: “How complex should a model be?” The consensus was that, although more complexity can improve realism, it can also introduce uncertainty and unnecessary data demands. “As simple as possible, but as complex as necessary” was the key take-home message.

Crucially, you don’t always need to model everything. Plasma concentrations, while imperfect, are often sufficiently protective for risk assessment. Moving to tissue-level predictions, especially brain, often adds difficulty without proportionally increasing decision value. One striking moment that highlighted this came when discussing uncertainty factors, which “always end up as the familiar factor of 10.”

So, where does this leave the field of exposure-led risk assessment? Walking out of the session, I felt that although not everything is solved, realistic and relevant PBPK modeling is entirely possible without animal data. Progress now rests on ensuring that we build trust in these approaches, and this trust will not come from a single validation study. Instead, the field needs transparency, consistency, and a growing body of evidence across chemicals and diverse contexts.

This blog reports on the Workshop Session titled “Integrating In Vitro Data and PBK Modeling for Exposure-Led Risk Assessment” that was held during the 2026 SOT Annual Meeting and ToxExpo. An on-demand recording of this session is available for meeting registrants on the SOT Online Planner and SOT Event App.

This blog was prepared by an SOT Reporter and represents the views of the author. SOT Reporters are SOT members who volunteer to write about sessions and events in which they participate during the SOT Annual Meeting and ToxExpo. SOT does not propose or endorse any position by posting this article. If you are interested in participating in the SOT Reporter program in the future, please email SOT Headquarters.