A new approach to predict PFS in Ovarian Cancer based on tumor growth dynamics.
Jiajie Yu (1) Nina Wang (1) Matts Kagedal (1)
(1) Department of Clinical Pharmacology, Genentech Research and Early Development, South San Francisco, USA.
Objectives:
Progression free survival (PFS) is a surrogate efficacy endpoint that is being used for regulatory approval of investigational drugs in ovarian cancer [1]. The PFS has also been shown to be predictive of the overall survival [1], hence being able to predict PFS for new treatments can be of great value. The PFS is usually analyzed by survival analysis methodology and PFS time is based on events such as target lesion related disease progression and non-target related disease progression whichever occurs first. PFS could be linked to tumor growth dynamics by linking the PFS hazard to target lesion tumor size, similarly to what is commonly done for overall survival [2]. With this approach target lesion tumor size data is included twice in the analysis, first as observations informing the model of tumor growth dynamics (TGD) and second influencing the PFS hazard in the time to event analysis. Based on such model it is possible to simulate unrealistic outcomes with a tumor size profile that meets the definition of target lesion disease progression, but without any PFS event occurring.
The aim of this work was to develop an approach for prediction of PFS that avoids duplicate use of target lesion data. A joint modelling approach including two sub-models is proposed: 1. Model for target lesion tumor growth dynamics. 2. Time-to-event model for non-target related progression. These sub-models combined can be used to predict PFS.
Methods:
A joint TGD-survival model was developed using NONMEM 7.4. The model contained three components to predict PFS. First, the TGD was modeled based on the method developed by Claret et al. [2] where the drug effect was introduced on the tumor size shrinkage parameter (ks). Potency difference between compounds were accounted for. The M3 method was applied to handle the censoring of tumor data when tumor size was below the limit of quantification per RECIST guideline [3]. The TGD was used to determine target lesion disease progression which was defined as more than 20% increase in target lesion tumor size from minimum observed value. Second, the hazard for non-target related disease progression was linked to the individually predicted, time varying target lesion tumor size. Third, a drop-out model was introduced to account for patient discontinuation from the study. Observed Kaplan-Meier curves for PFS were compared to simulated predictions where progression was based on either target lesion progression or non-target related disease progression, whichever occurs first in each patient. The model was developed using a pooled dataset including phase I and II data from platinum resistant ovarian cancer patients. Treatments included Anti-MUC16 ADC, Anti-MUC16 TDC, Anti-NaPi2b ADC and Doxorubicin.
Results:
The model could describe the observed dose response well for different compounds, both in terms of target lesion tumor size over time and in terms of PFS. The first derivative of target lesion tumor with respect to time correlated with the risk for non-target progression and the drug effect on the risk for non-target progression could be predicted based on target lesion tumor size alone.
Conclusion:
A joint model simultaneously estimating TGD based on target lesion tumor size and the risk for non-target related progression was developed. The model could describe the PFS influenced by different drug treatment and may provide a more robust prediction of PFS based on TGD for new treatments in future ovarian cancer clinical trials.
References:
[1] Herzog, Thomas J., et al. "FDA ovarian cancer clinical trial endpoints workshop: A Society of Gynecologic Oncology White Paper." Gynecologic oncology 147.1 (2017): 3-10.
[2] Claret, Laurent, et al. "A Model of Overall Survival Predicts Treatment Outcomes with Atezolizumab versus Chemotherapy in Non–Small Cell Lung Cancer Based on Early Tumor Kinetics." Clinical Cancer Research 24.14 (2018): 3292-3298.
[3] Claret, Laurent, et al. "Model-based prediction of phase III overall survival in colorectal cancer on the basis of phase II tumor dynamics." Journal of Clinical Oncology 27.25 (2009): 4103-4108.
[4] Eisenhauer, Elizabeth A., et al. "New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1)." European journal of cancer 45.2 (2009): 228-247.
