Model-based Analysis with Mechanistic Insights to CAR-T Cell Therapy Kinetics: Case Study with Axicabtagene Ciloleucel and Brexucabtagene Autoleucel

Pharmacometrics

MIDD

Methodology

Program strategies and efficiencies

Oncology

Clin Pharmacol Ther. 2026 Jul 21:10.1002/cpt.70395

Mc Laughlin AM, Bergstrand M, Ruiz-Garcia A, Filosto S, Shen R

In CAR-T cell therapy modeling, we often navigate between two extremes. The commonly used piecewise-linear models can accurately describe observed clinical data, but their purely empirical nature makes it difficult to rely on them for extrapolations and simulations. Conversely, highly mechanistic QSP models are invaluable for a detailed understanding of biological mechanisms. During my doctoral studies, I had the privilege of working in this latter space, utilizing rich longitudinal data on individual CAR-T cell phenotypes to develop a QSP model that could be used for early survival prediction.

However, moving from academia to supporting drug development as a model-informed drug development consultant brought a new aspect into my focus. In later-stage drug development, we rarely have access to such granular phenotypic data; instead, we typically rely on measurements of total CAR-T cells. The challenge became clear: how do we bridge the gap between these two modeling extremes? We needed a framework with enough mechanistic depth to support extrapolations and simulations, yet parsimonious enough to be reliably identified from routinely collected clinical data - enabling efficient model development on often tight timelines.

This specific challenge drove our recent work, now published in Clinical Pharmacology & Therapeutics (ASCPT Journal Family).

At Pharmetheus, alongside Martin Bergstrand, we developed a mechanism-based population PK framework designed to sit exactly in this middle ground. Distilled down to just seven structural parameters, the framework remains practical and data-driven. However, unlike piecewise-linear models, it uses a continuous, mechanistically driven approach to characterize PK profiles. This foundation allows us to more confidently perform extrapolations and simulate unobserved scenarios, such as varying doses or covariates.

But building the structural framework is only half the work; it had to be tested against real clinical data. This paper was the result of a close collaboration with our co-authors at Kite Pharma and Gilead Sciences. The clinical and translational expertise of Ana Ruiz, Simone (Simon) F., and Rhine Shen and the comprehensive database of 473 patients across multiple disease and product types were instrumental in validating the theoretical structure and showcasing its use.

Ultimately, the real value of any model lies in its utility. My hope is that this framework provides a reliable platform to support MIDD decisions for future CAR-T cell therapy programs.