Laurent Claret1,2, Christina Pentafragka3, Sanja Karovic4,5, Binsheng Zhao6, Lawrence H Schwartz6, Michael L Maitland4,5, Rene Bruno7,8. 1. Pharsight Consulting Services, Pharsight, a Certara™ Company, Marseille, France. 2. Genentech/Roche, Marseille, France. 3. Pharmacokinetics Unit, Aix-Marseille University, SMARTc, INSERM CRO2 UMR_S 911, Marseille, France. 4. University of Chicago Medicine, Chicago, IL, USA. 5. Inova Schar Cancer Institute and Center for Personalized Health, Falls Church, VA, USA. 6. Department of Radiology, Columbia University, New York, NY, USA. 7. Pharsight Consulting Services, Pharsight, a Certara™ Company, Marseille, France. rbruno100@gmail.com. 8. Genentech/Roche, Marseille, France. rbruno100@gmail.com.
Abstract
PURPOSE: To compare lesion-level and volumetric measures of tumor burden with sum of the longest dimensions (SLD) of target lesions on overall survival (OS) predictions using time-to-growth (TTG) as predictor. METHODS: Tumor burden and OS data from a phase 3 randomized study of second-line FOLFIRI ± aflibercept in metastatic colorectal cancer were available for 918 patients out of 1216 treated (75%). A TGI model that estimates TTG was fit to the longitudinal tumor size data (nonlinear mixed effect modeling) to estimate TTG with: SLD, sum of the measured lesion volumes (SV), individual lesion diameters (ILD), or individual lesion volumes (ILV). A parametric OS model was built with TTG estimates and assessed for prediction of the hazard ratio (HR) for survival. RESULTS: Individual lesions had consistent dynamics within individuals. Between-lesion variability in rate constants was lower (typically < 27% CV) than inter-patient variability (typically > 50% CV). Estimates of TTG were consistent (around 12 weeks) across tumor size assessments. TTG was highly significant in a log-logistic parametric model of OS (median over 12 months). When individual lesions were considered, TTG of the fastest progressing lesions best predicted OS. TTG obtained from the lesion-level analyses were slightly better predictors of OS than estimates from the sums, with ILV marginally better than ILD. All models predicted VELOUR HR equally well and all predicted study success. CONCLUSION: This analysis revealed consistent TGI profiles across all tumor size assessments considered. TTG predicted VELOUR HR when based on any of the tumor size measures.
RCT Entities:
PURPOSE: To compare lesion-level and volumetric measures of tumor burden with sum of the longest dimensions (SLD) of target lesions on overall survival (OS) predictions using time-to-growth (TTG) as predictor. METHODS:Tumor burden and OS data from a phase 3 randomized study of second-line FOLFIRI ± aflibercept in metastatic colorectal cancer were available for 918 patients out of 1216 treated (75%). A TGI model that estimates TTG was fit to the longitudinal tumor size data (nonlinear mixed effect modeling) to estimate TTG with: SLD, sum of the measured lesion volumes (SV), individual lesion diameters (ILD), or individual lesion volumes (ILV). A parametric OS model was built with TTG estimates and assessed for prediction of the hazard ratio (HR) for survival. RESULTS: Individual lesions had consistent dynamics within individuals. Between-lesion variability in rate constants was lower (typically < 27% CV) than inter-patient variability (typically > 50% CV). Estimates of TTG were consistent (around 12 weeks) across tumor size assessments. TTG was highly significant in a log-logistic parametric model of OS (median over 12 months). When individual lesions were considered, TTG of the fastest progressing lesions best predicted OS. TTG obtained from the lesion-level analyses were slightly better predictors of OS than estimates from the sums, with ILV marginally better than ILD. All models predicted VELOUR HR equally well and all predicted study success. CONCLUSION: This analysis revealed consistent TGI profiles across all tumor size assessments considered. TTG predicted VELOUR HR when based on any of the tumor size measures.
Authors: Sreenath M Krishnan; Sofiene S Laarif; Brendan C Bender; Angelica L Quartino; Lena E Friberg Journal: CPT Pharmacometrics Syst Pharmacol Date: 2021-05-02