PII-036 - PRECLINICAL TO CLINICAL TRANSLATION OF ADVERSE EVENTS FOR SMALL MOLECULE ONCOLOGY DRUGS

K. Collins¹, L. Wollenberg², A. Topletz-Erickson³, R. Mittapalli⁴; ¹Pfizer, La Jolla, CA, USA, ²Pfizer, Boulder, CO, USA, ³Pfizer, Bothell, WA, USA, ⁴Pfizer, San Diego, CA, USA.

Background: The starting dose for oncology First-in-Patient (FIP) studies is typically established via a toxicity-based approach, utilizing the highest non-severely toxic dose (HNSTD) or 1/10th the severely toxic dose (STD10) from preclinical GLP toxicity studies. Understanding how preclinical adverse events (AEs) translate clinically is key for the appropriate selection of the FIP starting dose that would be safe, but also reduce the number of patients exposed to subtherapeutic doses. We investigated how preclinical GLP toxicity data predicts high grade AEs (Gr ≥3) / dose-limited toxicities (DLT) in FIP oncology small molecule studies.

Methods: Relevant preclinical and clinical data for 25 Pfizer small molecules were extracted to evaluate translation of AEs from 28-day GLP preclinical toxicity studies to clinical ≥3 Gr AEs or DLTs. AEs were categorized according to CTCAE v4.0. Translational likelihood of AE categories was determined by calculating preclinical to clinical positive predictive values (PPV) and negative predictive values (NPV) [PMID: 32491423], corresponding to the probability that the toxicity category would or would not be detected in FIP clinical studies if observed preclinically.

Results: AEs compared between 28-day GLP studies and corresponding ≥3 Gr incidence in human FIP studies had mean PPV of 0.53 and NPV of 0.85. Positive predictive correlations appeared to be organ-dependent, such as in hematological and cutaneous systems (≥0.67 PPV) . Negative prediction in all organ classes were >0.67 NPV except for gastrointestinal (0.58 NPV). Comparison by certain factors such as preclinical species (non-human primates vs rats), number of preclinical species used (2 vs 3), or drug class (tyrosine kinase inhibitor [TKI] vs. non-TKI) did not result in a notable difference in PPV or NPV breakdown. However, PPV trended higher in TKIs compared to non-TKI drugs and NPV trended higher for drugs tested in 3 vs 2 preclinical species.

Conclusion: Overall, for the 25 Pfizer molecule programs in the data set the translation of AEs observed in 28-day GLP studies to those observed in clinical FIP studies showed limited correlation based on PPV (0.53). However, the high NPV (0.85) indicates that toxicities not observed preclinically are unlikely to be observed in clinical FIP studies.