PII-153 - PREDICTION OF PFS BASED ON TUMOR GROWTH INHIBITION METRICS IN HER2 POSITIVE BREAST CANCER PATIENTS FOLLOWING HER2 ADC FS-1502 TREATMENT

L. Diao¹, C. Zhou², Y. Sun², X. Xiang², C. Li³, H. Hua², X. Wang², Y. Zheng², Z. Wu², J. Qiu²; ¹Shanghai Fosun Pharmaceutical Development Co., Ltd, ²Shanghai Fosun Pharmaceutical Development Co., Ltd., ³Anheart Therapeutics.

Background: FS-1502 is an HER2 ADC bearing a novel cleavable linker and payload of monomethyl auristatin F (MMAF) and currently being investigated in patients with HER2 expressed advanced solid tumors. Based on the longitudinal tumor size and immature progression free survival (PFS) data, a tumor growth inhibition (TGI)-PFS model was developed to predict the median PFS of FS-1502 in HER2 positive breast cancer patients for early decision making.

Methods: Longitudinal tumor size data from 35 patients who received 2.3 mg/kg FS-1502 in the Phase I study were fitted using a bi-exponential TGI model by NONMEM (version 7.5.0). The impact of TGI metrics (shrinkage rate and growth rate) and baseline prognostic factors on PFS were evaluated by univariate Cox regression first using R (version 4.2.1). Then these significant variables (p < 0.05) were included in a multivariate parametric survival regression model. The Akaike information criterion (AIC) was used to select the best model (among logistic, log-logistic, lognormal, Weibull, Exponential, Gaussian distribution) to describe the PFS distribution.

Results: The bi-exponential TGI model was best tumor dynamic model to explain the tumor size data based on goodness of fit and precision of parameters estimation. The results shown that log transformed tumor growth rate (logKG) was the top predictor for PFS in both univariate Cox and multivariate parametric models, and the Weibull distribution provided the best fit to the survival data. The effect size of logKG to PFS was -1.374 (p=0.0005) which is the most adverse prognostic factor in the model. The model-predicted median PFS of 2.3 mg/kg FS-1502 in HER2 positive breast cancer patients was 12.98 months (80% Confidence Interval = [7.23, 22.89]) for patients with median level of logKG.

Conclusion: A survival model has been developed that uses model-based estimates of tumor growth rate to predict PFS for HER2-positive breast cancer. The results show that the efficacy of FS-1502 is promising and the model can be used to support early decision making in the development of FS-1502 in this indication. The study is continuing and updated model with the inclusion of more PFS data will be presented at the meeting.