Biophysical simulation approach for dose escalation in Phase I clinical trials

Objectives: The FDA recently released the “Project Optimus”1 with the intention to improve the dose selection paradigm in drug development. Inadequate dose selection for a drug affects toxicity and efficacy. Here we performed a proof-of-concept study using trastuzumab emtansine (TDM1), a targeted drug for HER2-enriched breast cancer, to demonstrate how biophysical modeling can support dose selection in phase I clinical trials. Our goal was to compare the simulated dose escalation in our model to the clinical dose escalation completed in the TDM1 phase I clinical trial2.

Methods: To perform the in-silico dose escalation study in virtual patients, we leveraged preclinical pharmacokinetic (PK) and pharmacodynamic (PD) input data to establish the TDM1 model. The PK model parameters are based upon a mouse xenograft study while PD parameters were set using in vitro breast cancer cell line studies. We selected large cohorts of HER2-enriched and HER2-low intend-to-treat patients from our virtual tumor bank of spatially-resolved 3D breast cancer patient tumors. Once the drug model was established and the patients were selected, we used our biophysical simulation platform, TumorScope (TS)3, to simulate the response of TDM1 in the breast tumor of each patient. TS allows tracking of the intratumoral drug concentration and the spatiotemporal tumor response. We were able to estimate individual patient response in addition to the minimal effective dose on a patient- and cohort- levels.

Results: We observed a wide distribution of drug exposure across patients and doses. Across the simulated TDM1 doses (0.3, 0.6, 1.2, 2.4, 3.6, 4.8 mg/kg) drug exposure saturation occurred at a dose similar to the clinical TDM1 dose (3.6 mg/kg) from the phase I clinical trial2. Doses higher than the TDM1 exposure saturating dose offered minimal benefit to the tumor volumetric response rate in our simulations. Of note, TS allows patient-specific dose selection based on predicted drug coverage. We also compared our results to those obtained using three traditional QSP models4. The simulated dose-response curves were inconsistent between these previously published models and the predicted IC50s estimated from each model fell three to four times higher than the clinical dose. TS simulations provide a more accurate estimate of the IC50 than the traditional QSP models.

Conclusions: While further studies in additional drug models are necessary, this proof-of-concept study demonstrates the potential of TS to improve drug dose escalation in phase I clinical trials.

Citations:

  1. Project Optimus | FDA. https://www.fda.gov/about-fda/oncology-center-excellence/project-optimus.
  2. Beeram, M. et al. A phase 1 study of weekly dosing of trastuzumab emtansine (T-DM1) in patients with advanced human epidermal growth factor 2–positive breast cancer. Cancer 118, 5733–5740 (2012).
  3. Cole, J. A. et al. SimBioSys TumorScope: Spatio-temporal modeling of the tumor microenvironment to predict chemotherapeutic response. J. Clin. Oncol. 38, e12650–e12650 (2020).
  4. Haddish-Berhane, N. et al. On translation of antibody drug conjugates efficacy from mouse experimental tumors to the clinic: a PK/PD approach. J. Pharmacokinet. Pharmacodyn. 40, 557–571 (2013).