2023 - A Coruņa - Spain

PAGE 2023: Drug/Disease Modelling - Oncology
Anna Kondic

The combined use of propensity score matching and a joint tumor growth dynamics (TGD) - Overall Survival (OS) model to benchmark the efficacy of new treatments for advanced renal cell carcinoma (RCC)

Mayu Osawa1, Martin Winiger2, Ramon Garcia3, Jonathan French3, Anna Kondic1, Bauke Stegenga2, and Amit Roy1

1 Clinical Pharmacology and Pharmacometrics, Bristol Myers Squibb, Princeton, NJ, United States 2 Worldwide Strategic Collaborations, Global Medical, Bristol-Myers Squibb, Princeton, NJ, United States 3 Metrum Research Group, Tariffville, CT, United States

Objectives: 

Modeling tumor growth dynamics in different tumor types and under various therapeutic interventions has been an active area of research and has been used in a variety of applications [1,2]. While multiple approaches exist, the general modeling framework consists of a 1) tumor dynamic model (usually fitted based on longitudinal biomarker data such as the sum of longest diameters (SLD), 2) association to a clinical outcome (typically OS) via a predictive TGD metric, either in two stage or joint modeling framework to create dynamic predictions.The published applications span a few tumor types and operate under tumor-specific, treatment-independent paradigm [2]. Advanced RCC is a tumor type of interest where there is a need to test the feasibility of a model that encompasses different treatment regimen and modalities. Schindler et al created a detailed specific modeling framework to quantify the relationship between masitinib exposure, biomarkers linked to the pharmacology of the drug, SLD and OS [3].

 

This work was motivated by the need to enhance the interpretability of efficacy data generated in non-registrational data generation (NRDG) and investigator sponsored research (ISR) studies of anti-cancer agents, which are commonly relatively small non-randomized studies without a control arm, and limited OS data. Our approach to increase the interpretability of data from NRDG/ISR studies was to use a TGD-OS model to predict OS, and then benchmark the predicted OS against an established therapy using weights derived from a propensity score model (PSM).

Methods: 

A joint TGD-OS modeling framework for advanced RCC was adapted to describe the data from CA209-214 (NCT02231749). This study served as the pivotal study leading to the approval of nivolumab (NIVO) + ipilimumab (IPI) for previously untreated advanced RCC with intermediate or poor risk. The NIVO + IPI treatment in this study was NIVO 3 mg/kg and IPI 1 mg/kg administered intravenously Q3W for four doses, followed by NIVO 3 mg/kg monotherapy Q2W [4]. This previously developed TGD-OS RCC model uses a mixture-Wang framework [5] for the TGD part; it was fit to the CA209-214 data using longitudinal SLD, and Karnofsky score, albumin, line of therapy, number of lesions and PD-L1 status as baseline covariates. The overall survival of each subject was assumed to follow a log-logistic distribution.

The fitted joint TGD-OS model was then used to predict OS of RCC patients in the NRDG study CA209-920 (Cohort 1, NCT02982954), a single-arm study of NIVO combined with IPI (N=106). Subjects in CA209-920 (Cohort 1) received NIVO 6 mg/kg Q6W in combination with IPI 1 mg/kg Q6W, which is different from the approved NIVO + IPI treatment [6]. OS was predicted utilizing only immature tumor size data available approximately 2-years prior to the availability of the mature OS data.

The predicted OS for CA209-920 (Cohort 1) was benchmarked against the approved NIVO + IPI treatment regimen (investigated in CA209-214) by utilizing a PSM to compute weights to adjust for the differences in markers for disease severity and other covariates.

Results: 

Cross-study comparison of the data revealed that subjects in study CA209-920 were more likely to have poor MSKCC or intermediate IMDC score, higher baseline tumor size and lower albumin compared to CA209-214. A random forest-based PSM yielded weights that could be used to balance the covariates, resulting in a synthetic control arm comprised of patients with comparable baseline characteristics to those of CA209-920. The refined TGD-OS model, including key baseline covariates, was then used to predict OS in CA209-920 using immature TGD data. These model predictions were then shown to be in close agreement with final OS data (observed OS was well within the 95% prediction interval).

Conclusions: 

This work established the feasibility of predicting OS in advanced RCC with longitudinal TGD and immature OS data together with baseline covariates, and benchmarking the predicted OS with that of an established therapy using PSM weighting.



References:
[1] Wang Y, Sung C, Dartois C, Ramchandani R, Booth BP, Rock E, Gobburu J. Elucidation of relationship between tumor size and survival in non-small-cell lung cancer patients can aid early decision making in clinical drug development. Clin Pharmacol Ther. 2009 Aug;86(2):167-74. 
[2] Bruno R, Mercier F, Claret L. Evaluation of tumor size response metrics to predict survival in oncology clinical trials. Clin Pharmacol Ther. 2014 Apr;95(4):386-93.
[3] Schindler E, Amantea MA, Karlsson MO, Friberg LE. A Pharmacometric Framework for Axitinib Exposure, Efficacy and Safety in Metastatic renal Cell Carcinoma Patients,
CPT Pharmacometrics Syst Pharmacol. 2017 Jun;6(6):373-382.
[4] Motzer RJ, Tannir NM, McDermott DF, Arén Frontera O, Melichar B, Choueiri TK, Plimack ER, Barthélémy P, Porta C, George S, Powles T, Donskov F, Neiman V, Kollmannsberger CK, Salman P, Gurney H, Hawkins R, Ravaud A, Grimm MO, Bracarda S, Barrios CH, Tomita Y, Castellano D, Rini BI, Chen AC, Mekan S, McHenry MB, Wind-Rotolo M, Doan J, Sharma P, Hammers HJ, Escudier B; CheckMate 214 Investigators. Nivolumab plus Ipilimumab versus Sunitinib in Advanced Renal-Cell Carcinoma. N Engl J Med. 2018 Apr 5;378(14):1277-1290. 
[5] Feng, Y.,Wang, X., Suryawanshi, S., Bello, A. and Roy, A. Linking TumorGrowthDynamics to Survival in Ipilimumab-Treated PatientsWith AdvancedMelanoma UsingMixture Tumor Growth DynamicModeling. CPT Pharmacometrics Syst Pharmacol 8 (2019):825–834.
[6] George DJ, Spigel DR, Gordan LN, Kochuparambil ST, Molina AM, Yorio J, Rezazadeh Kalebasty A, McKean H, Tchekmedyian N, Tykodi SS, Zhang J, Askelson M, Johansen JL, Hutson TE. Safety and efficacy of first-line nivolumab plus ipilimumab alternating with nivolumab monotherapy in patients with advanced renal cell carcinoma: the non-randomised, open-label, phase IIIb/IV CheckMate 920 trial. BMJ Open. 2022 Sep 14;12(9). 


Reference: PAGE 31 (2023) Abstr 10608 [www.page-meeting.org/?abstract=10608]
Oral: Drug/Disease Modelling - Oncology
Top