PII-038 - STRATEGIC CONSIDERATIONS FOR TUMOR DYNAMICS MODELING IN EARLY DEVELOPMENT: LEARNINGS FROM AN ANTI-HER2 ANTIBODY-DRUG CONJUGATE

K. Collins¹, M. Reddy², E. Hahn³, L. Wollenberg², L. Zhou¹, J. Li¹; ¹Pfizer, La Jolla, CA, USA, ²Pfizer, Boulder, CO, USA, ³Pfizer Inc., Boulder, CO, USA.

Background: The FDA has recently emphasized the value of tumor dynamics models (TDM) in dose optimization [PMID: 36095296]. While many published examples have been with use of ongoing Phase 3 or retrospective data, TDM can have significant utility earlier in clinical development if key considerations are met. Here we highlight important considerations for developing TDMs to support phase 1 milestones using a previously presented example with the next generation anti-HER2 antibody-drug conjugate, PF-06804103.

Methods: TDM development for PF-06804103 was initiated at the end of first-in-human dose escalation, using exposure and tumor size data from a total of 23 participants with metastatic breast cancer (mBC) across 7 dose levels. The model supported selection of a low-dose backfill regimen to test in expansion based on a simulated tumor shrinkage. During the expansion phase the TGI model was updated (n = 63 participants) and linked with a time-to-event model characterizing risk of Grade ≥2 peripheral neuropathy (PN) to account for PN-driven dose modifications. The updated model was used to simulate ORR of different dosing regimens vs. the standard of care in a larger phase 3 trial.

Results: Early initiation of TDM is feasible if key conditions are met, such as study design and extent of activity observed during dose escalation/expansion. PF-06804103 provided a strong case to begin TDM early in dose escalation given anti-tumor activity was already observed in a substantial portion of participants in dose escalation. Early development and use of the TDM supported further testing of 3 and 4 mg/kg dose levels, and then allowed for a model update to support dose optimization with additional efficacy and safety data. This continuous model development strategy provided confidence in model utility for selection of the RP2D regimen and support of key decisions during early clinical development.

Conclusion: TDM in early dose optimization can include uses like selecting backfill or expansion dose levels to better explore dose- or exposure-response relationships, supporting recommended dose(s) for expansion cohorts, and the dose modification strategy, which can be done in the presence or absence of a safety model. Here we present the real-time applications of TDM to inform PF-06804103 dose finding decisions, highlighting its utility at two early development milestones.