{
  "abstract": "Background Quantitative systems pharmacology (QSP) models have been used to guide clinical dose selection for T cell engagers (TCEs). These models are typically parameterized using data from in vitro cytotoxicity experiments and describe populations of tumor cells and T cells over time. The formation of trimers (the complex formed between the TCE, the T cell receptor, and the tumor cell antigen) are presumed to drive efficacy and toxicity. No direct bioassays exist for trimer formation, and different assumptions around binding kinetics made across the literature lead to differing inferred trimer densities. This work compared four model assumptions for trimer formation and assessed their impact on in vitro simulations and clinical dose projections.Methods A QSP model for TCE binding and in vitro activity was developed, applying four alternative binding assumptions to investigate trimer formation: well-mixed, cell-surface density-dependent, cellular density-dependent, and combined density-dependent binding. Models with each binding assumption were fit to published in vitro cytotoxicity data from three TCEs (teclistamab, elranatamab, and epcoritamab). Hypothetical experiments were simulated to explore possible in vitro study designs that could distinguish between the different binding assumptions.The QSP model was then translated to clinical scenarios using a minimal physiologically based pharmacokinetic model and plausible assumptions around tumor composition, and tumor responses to TCE dosing were simulated.Results All model variants described cytotoxicity data for each TCE well after parameter optimization; the root mean square error ranged from 3.5% to 6.3% across the model fits. The trimers per tumor cell required for efficacy however varied drastically between model variants, which highlighted the challenge of using an unobservable variable for dose projections. Additional simulations suggested that measuring cytotoxicity across a range of E:T ratios and TCE concentrations, ideally at multiple time points, could distinguish between the binding assumptions. Both model assumptions around binding and T cell infiltration impacted the projected clinical dose responses. The predicted minimally efficacious clinical dose spanned from 8.2e-6 to 25 mg/kg.Conclusions QSP models with different binding assumptions described in vitro experiments for TCEs equally well, but yielded substantially different clinical dose projections. Selection of the appropriate binding assumption may be informed through well-designed in vitro studies. To improve the quality and reliability of the QSP models, and therefore clinical dose projections, TCE in vitro study designs should include evaluation of cytotoxicity across multiple E:T ratios and at multiple time points.",
  "authors": [
    {
      "affiliations": [
        "Metrum Research Group, Boston, MA, USA"
      ],
      "name": "Kiersten Utsey"
    },
    {
      "affiliations": [
        "Metrum Research Group, Boston, MA, USA"
      ],
      "name": "Daniel Kirouac"
    }
  ],
  "title": "1127 Insights from a quantitative systems pharmacology model to guide in vitro study design for T cell engager clinical development",
  "uid": "9b6c3d96-2e0a-5f17-bc53-a34f968ca535"
}