📄 pharmacology and toxicology

A novel workflow integrating whole-body PET microdosing data and therapeutic-dose pharmacokinetics across species to inform first-in-human dose selection

This study demonstrates that integrating whole-body PET microdosing data with therapeutic-dose pharmacokinetics in non-human primates within a PBPK framework successfully predicts first-in-human exposure and tissue distribution for dolutegravir, validating a novel workflow to inform dose selection.

Original authors: Nguyen, B. T., Barrail-Tran, A., Ursino, M., Goutal, S., Caille, F., Naninck, T., Le Grand, R., Lambotte, O., Tournier, N., Comets, E.

Published 2026-09-08

📖 4 min read☕ Coffee break read

CC0 1.0

Original authors: Nguyen, B. T., Barrail-Tran, A., Ursino, M., Goutal, S., Caille, F., Naninck, T., Le Grand, R., Lambotte, O., Tournier, N., Comets, E.

Original paper dedicated to the public domain under CC0 1.0 (https://creativecommons.org/publicdomain/zero/1.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Developing a new medicine is a high-stakes gamble. Before a drug can ever reach a patient, scientists must first guess how much of it to give a human for the first time. This guesswork is necessary because testing high doses directly on people is dangerous, and the body's reaction to a substance often changes drastically between a mouse, a monkey, and a human. Traditionally, researchers rely on animal studies to predict human safety, but these predictions are often imperfect. They struggle to see how a drug moves through the deep tissues of the body or how the body's own machinery might handle a large dose differently than a tiny one. To bridge this gap, regulators have begun to allow "microdosing," where a tiny, harmless amount of a drug is given to a human to see how it behaves. However, getting this data early in the process is difficult, and scientists need better ways to translate what they see in animals into reliable predictions for people.

A team of researchers has proposed a new way to solve this puzzle by combining two different types of animal data into a single, powerful computer model. They focused on dolutegravir, a widely used medication for treating HIV, to test their method. The researchers worked with non-human primates, giving them two different kinds of doses. First, they injected a microscopic amount of the drug that had been tagged with a radioactive tracer. This allowed them to use a special camera, known as a PET scanner, to watch the drug flow through the animal's body in real time, tracking its journey through organs like the liver and kidneys over a few hours. This provided a detailed map of where the drug went, but only for a short window of time. Next, the team gave the same animals a full, therapeutic dose of the drug. They then measured the drug levels in the blood and tissues over a longer period. This second set of data showed how the body handled a large amount of the medicine, revealing how the body's filtering systems might get saturated or slow down when overwhelmed.

The scientists built a virtual model of the primate's body inside a computer, feeding it the data from both the short, radioactive scan and the longer, full-dose study. They used the quick, detailed images from the PET scan to set the initial rules for how the drug moved between organs. Then, they used the data from the full dose to adjust those rules, teaching the model to account for the fact that the body behaves differently when it is processing a large amount of a substance versus a tiny one. This created a refined digital twin of the animal's physiology that understood both the drug's path and its limits. Once the model was trained on the animal data, the researchers swapped the animal's body for a human one within the simulation. They ran the model forward to predict what would happen if a human took the standard daily dose of the medication.

The results showed that using only the quick, radioactive scan data was not enough; the model predicted that the drug would disappear from the body much faster than it actually does in humans. However, when the model was updated with the data from the full-dose animal study, the predictions changed. The refined model successfully forecasted the levels of the drug in human blood, matching the real-world data from clinical trials with remarkable accuracy. It correctly estimated the peak amount of drug in the blood and the total exposure over a day. While the model slightly underestimated the lowest levels of the drug just before the next dose, it captured the overall behavior of the medicine far better than previous methods. This work demonstrates that by layering a quick, high-resolution look at drug distribution with a longer look at how the body processes a full dose, scientists can build a much more reliable bridge between animal studies and human safety. The approach offers a promising path to selecting safer starting doses for new medicines, potentially reducing the need for extensive animal testing and speeding up the journey to finding effective treatments.

Technical Summary: A Novel Workflow Integrating Whole-Body PET Microdosing and Therapeutic-Dose PK for First-in-Human Dose Selection

Problem Statement
Drug development faces high failure rates at early proof-of-concept stages, largely due to the limited predictive value of preclinical models for extrapolating human pharmacokinetics (PK) and pharmacodynamics. Standard Phase I trials are costly, time-consuming, and offer restricted insights into tissue distribution and mechanistic biomarkers. While regulatory frameworks (FDA, ICH) support "Phase 0" microdosing studies to generate early PK and target engagement data, integrating these subtherapeutic, dynamic imaging data with conventional therapeutic-dose data into a unified modeling framework for first-in-human (FIH) dose selection remains a challenge. Specifically, there is a need to address non-linearities (e.g., saturable ADME processes) and bridge the gap between animal microdosing and human therapeutic exposure.

Methodology
The authors propose a structured, four-step PET-informed PBPK modeling framework to integrate microdosing and conventional dose data across species:

Step A (Preclinical Microdosing): Utilizing dynamic whole-body PET imaging in non-human primates (NHPs) following radiolabeled microdose injection to generate time-activity curves (TACs) in key organs. These data are used to inform initial PBPK model parameters, specifically tissue partition coefficients and distribution kinetics.
Step B (Preclinical Dose Extrapolation): Bridging microdoses to therapeutic doses within the same preclinical species. This step incorporates conventional PK data (plasma, urine, and tissue concentrations) from therapeutic-dose studies to identify and model non-linearities (e.g., saturable metabolism or transport) and refine clearance parameters.
Step C (Human Translation - Conceptual): The framework envisions using human PET microdosing data to refine interspecies scaling and tissue distribution parameters, though this step was not executed in the current case study due to data unavailability.
Step D (Human PK Prediction): Combining optimized parameters from Steps A and B to simulate human PK profiles for FIH dose selection, utilizing population physiology distributions.

Case Study Application
The framework was applied to dolutegravir, an antiretroviral drug, using data from the RHIVIERA project:

Data Sources: PET microdosing data from 4 NHPs receiving IV [18F]dolutegravir (with or without a pre-injection of unlabeled drug) and conventional PK data from NHPs receiving a 2.5 mg/kg IV therapeutic dose.
Modeling Tool: A whole-body PBPK model was constructed in PK-Sim/MoBi (v12.1).
Workflow:
- Sensitivity analyses identified estimable parameters (e.g., tissue partition coefficients, unbound fraction, metabolic $V_{max}$ ).
- The model was fitted sequentially: first to the 3-hour PET microdosing data, then to the conventional-dose plasma and tissue data.
- Parameters derived from the NHP models were extrapolated to simulate steady-state PK in a virtual population of 100 healthy Caucasian adults receiving 50 mg oral dolutegravir once daily.

Key Results

Microdose Modeling: The PBPK model adequately described radioactivity TACs in NHPs over 3 hours, capturing early distribution dynamics. However, microdose-only parameter estimates systematically underpredicted human exposure metrics.
Therapeutic Dose Calibration: Re-estimating parameters using conventional-dose data revealed dose-dependent non-linearities, specifically higher metabolic $V_{max}$ and lower tissue partition coefficients compared to microdose estimates, consistent with saturation effects at therapeutic levels.
Human Prediction:
- Microdose-only parameters ( $\theta_P$ ): Systematically underpredicted human steady-state exposure (AUC $_{0-24}$ by 45%, $C_{max}$ by 29%, $C_{trough}$ by 82%).
- Combined Micro/Therapeutic parameters ( $\theta_C$ ): Successfully predicted steady-state geometric mean AUC $_{0-24}$ and $C_{max}$ closely matching reported human reference values. While $C_{trough}$ was slightly underpredicted (36% lower), the prediction intervals overlapped with literature data.
- The model provided dynamic insights into tissue distribution (liver, kidney, spleen, brain) that are not feasible with conventional plasma-only methods, though some discrepancies remained in organs like the brain and bile due to methodological differences and saturation effects.

Significance and Claims
The authors present this work as a proof-of-concept study demonstrating the feasibility of a sequential workflow that integrates NHP PET microdosing data with conventional therapeutic PK data to inform FIH dose selection.

Primary Contribution: This is the first reported integration of PET microdosing data from NHPs into a whole-body PBPK model to quantitatively extrapolate dolutegravir PK to humans.
Clinical Relevance: The workflow supports the use of early PET data to build robust priors for human dose-finding studies, potentially refining dose selection and assessing early tissue exposure.
Limitations and Modesty: The authors acknowledge that the study was limited by the absence of human PET data (Step C of the framework), sparse sampling in the conventional NHP study, and the inability to distinguish parent drug from metabolites in PET data. They note that tissue prediction quality varied and that further methodological refinements (e.g., global sensitivity analysis, Bayesian approaches for parameter uncertainty) are required before such workflows can be routinely deployed for translational decision-making. The study does not claim to replace Phase I trials but rather to complement them for safer and more informed dose selection.

More like this