Model-informed noninferiority of cetuximab biweekly to weekly dosing regimen in metastatic colorectal cancer
Ana-Marija Grišić (1), Siobhán Hayes (2), Colm Farrell (2), Yoshihiro Kuroki (3), Mauro Bertolino (1), Karthik Venkatakrishnan (4), Pascal Girard (5)
(1) Merck Healthcare KGaA, Darmstadt, Germany; (2) ICON plc, Reading, United Kingdom; (3) Merck Biopharma Co. Ltd. (an affiliate of Merck KGaA), Tokyo, Japan; (4) EMD Serono Research & Development Institute, Inc., Billerica, MA, USA, an affiliate of Merck KGaA; (5) Merck Institute for Pharmacometrics (an affiliate of Merck KGaA, Darmstadt, Germany), Lausanne, Switzerland.
Objectives: Cetuximab is a monoclonal antibody that targets epidermal growth factor receptor (EGFR), approved in metastatic colorectal cancer (mCRC). Initially cetuximab was approved at weekly (q1w) dosing regimen (loading dose of 400 mg/m2, followed by 250 mg/m2), however there is increasing evidence of comparability of 500 mg/m2 every-two-weeks (q2w) dosing regimen [1]. Moreover, recent studies identified primary tumor location as a potential prognostic factor in CRC [2]. Thus, the objective of this work was to apply a pharmacokinetics (PK)-tumor size (TS) model linked to overall survival (OS) model in colorectal cancer patients to (1) confirm effects of primary tumor location on tumor growth dynamics and/or efficacy and (2) evaluate noninferiority of q2w relative to q1w dosing regimen.
Methods: Pooled data from five phase I-III clinical trials in patients with KRAS wild-type colorectal cancer treated with cetuximab were analyzed using nonlinear mixed-effects modeling approach for longitudinal PK-TS model development and parametric time-to-event model for OS. As primary tumor location was not reported in one trial, a multiple imputation (n=10) and model averaging strategy [3] was applied. To characterize PK and TS dynamics, a previously developed PK model [4] was used and TS model was developed using Claret et al. model as a starting point. Parametric OS model was evaluated for each of 10 imputed data sets, and the model outputs were averaged using natural method and adjusted standard error. Different distributions (incl. Weibull, exponential, Gompertz, etc) were compared to kernel density estimate to identify the most appropriate assumption for baseline hazard distribution. To test non-inferiority of q2w vs q1w dosing regimen of cetuximab, 10,000 virtual patients randomized 1:1 in 2 arms (q2w and q1w) were simulated based on the averaged OS model and patient characteristics sampled from distributions based on the studies used for model development. Dosing regimen was kept as a covariate in the OS model although it was not significant. Individual probability of survival was determined over 5 years and mean survival probability calculated stratified by regimen. Finally, 1000 trials of 2,000 patients (q1w vs q2w, randomized 1:1) were simulated by resampling the 10,000 simulated patients with replacement and non-inferiority test was performed, with HR <0.8 considered clinically meaningful, corresponding to non-inferiority margin of 1.25 [5]. Software used included NONMEM and R for data processing and model development, and Pharsight Trial Simulator for clinical trial simulations.
Results: In total, data from 852 patients were analyzed. Missing tumor location data were imputed by estimating the likelihood of tumor location using a logistic regression model. For PK-TS model development, individual PK exposures were derived from the PK model, a two-compartment model with linear CL and body surface area as a covariate on CL and V. The TS model was adapted from Claret et al. with an additional effect compartment for exposure; the model assumed an exponential tumor growth, a tumor killing rate proportional to exposure (AUC), exponential appearance of resistance, and Eastern Cooperative Oncology Group (ECOG) status on baseline TS as a covariate (p<0.01 for forward selection and p<0.001 for backward elimination). In the OS model, Weibull distribution appropriately approximated the baseline hazard distribution. For each imputed tumor location set (n=10), OS modeling was performed. Overall, based on average parameter estimates from those 10 OS models, ECOG status (HR=1.588 for ECOG>0), model-predicted early tumor shrinkage at week 8 (HR=0.987) and baseline TS (HR=1.002), and primary tumor location (HR=1.522 for right side) were identified as covariates in the OS model. Simulations demonstrated overlapping survival curves for the two dosing regimens, and across the simulated clinical trials, the vast majority (93%) met non-inferiority at the 0.80 HR for the q2w regimen.
Conclusions: Early tumor shrinkage, baseline TS, ECOG status and primary tumor location were confirmed as significant covariates in the OS model in CRC patients treated with cetuximab. Using simulations, non-inferiority of 500 mg/m2 q2w dosing regimen to 250 mg/m2 q1w with loading dose of 400 mg/m2 was demonstrated.
References:
[1] A.R. Parikh, E. Gonzalez-Gugel, N. Smolyakova, M.-H. Jen, N. Toms, Y. Lin, J.S. Kim, S. Kopetz. Efficacy and safety of cetuximab dosing (biweekly vs weekly) in patients with KRAS wild-type metastatic colorectal cancer: A meta-analysis. Oncologist. DOI: 10.1093/oncolo/oyab030
[2] A. Bahl, V. Talwar, B. Sirohi, P. Mehta, D. Arya, G. Shrivastava, A. Dahiya, K. Pavithran. Primary tumor location as a prognostic and predictive marker in metastatic colorectal cancer (mCRC). Front. Oncol. 10:964 (2020).
[3] Å.M. Johansson, M.O. Karlsson. Comparison of methods for handling missing covariate data. AAPS J. 15:1232-1241 (2013).
[4] Girard P, Kovar A, Brockhaus B, Zuehlsdorf M, Schlichting M, Munafo A. Tumor size model and survival analysis of cetuximab and various cytotoxics in patients treated for metastatic colorectal cancer. Abstract #2902, PAGE 2013.
[5] S. Kasper, C. Foch, D. Messinger, R. Esser, F.-X. Lamy, V. Rothe, W. Chen, A.-L. Cheng, M. Rouyer, T. Brodowicz, C. Zielinski. Noninferiority of cetuximab every-2-weeks versus standard once-weekly administration schedule for the first-line treatment of RAS wild-type metastatic colorectal cancer. Eur. J. Cancer. 144:291-301 (2021).