Hidefumi Kasai and Yusuke Tanigawara
Keio University School of Medicine
Objectives: Tumor growth models have been developed for several chemotherapeutic and molecular targeted drugs and the predicted tumor growth index has been shown to be a promising biomarker for survival. FOLFOX is a combination regimen of oxaliplatin (L-OHP) and 5-fluorouracil (5-FU), and although mFOLFOX with bevacizumab is a standard regimen for colorectal cancer, there have been no reports on modeling and simulation regarding this regimen. Therefore, we aimed to develop a tumor shrinkage model that contains the exposure of chemotherapeutic agents following mFOLFOX regimen and to explore any indices to discriminate the patient response and survival.
Methods: All data used in this analysis were obtained from a post-marketing Phase II multicenter clinical study of first-line mFOLFOX6 with bevacizumab therapy [1]. L-OHP (85 mg/m2) and 5-FU (400 mg/m2 intravenous injection followed by 2400 mg/m2 infusion) were administered with bevacizumab and levofolinate calcium to patients (n=65) of metastatic colorectal cancer. The regimen was repeated every 2 weeks. Tumor growth model was assumed to be composed of two types of tumor cells: sensitive and resistant cells. Both cells were postulated to grow and die with same rate constants Kgrow and Kkill, respectively, but the drugs were assumed to be effective only to the sensitive cells killing. Sensitive cells were converted to resistant cells with a first order rate constant, Kts, which was affected by the drug exposure. Drug concentrations of L-OHP and 5-FU were estimated using the PPK models with their respective covariate effects [2, 3]. The average concentrations over a dosing interval were calculated and then standardized to their standard doses. The sum of the standardized average concentrations of L-OHP and 5-FU was postulated to increase Kkill. Parametric progression-free and overall survival models were constructed using the estimated entire tumor size profiles. All the analyses were performed using the Phoenix NLME software Version 8.1 (Certara, LP, Princeton, NJ).
Results: The sensitive and resistant cells model fitted the tumor size data well. The standardized average concentration of L-OHP and 5-FU had significant effect on increasing Kkill. The fs, ratio of sensitive cells, was estimated to 0.464 at the baseline, and the sensitive cells were decreased through the treatment according to the cumulative AUCs of the drugs. Tumor Response Index (TRI), defined as Kkill/Kgrow, clearly discriminated between the responder (CR and PR, n=34) and non-responder (SD and PD, n=31) (P<0.001). Developed survival models using Weibull hazard described well both PFS and OS, with significant covariates of baseline tumor size and the TRI.
Conclusions: We have developed a tumor shrinkage model for colorectal cancer, with identifying a potential predictive index (Tumor Response Index) to discriminate responder or non-responder for tumor shrinkage. PFS and OS were also reasonably described by the constructed model. Although both the constructed models need to be validated in a future trial for their predictability, they will be useful to evaluate clinical efficacy of the other regimens containing L-OHP and/or 5-FU.
References:
[1] Nishina T et al. Jpn J Clin Oncol (2013) 43:1080-1086.
[2] Nikanjam M et al. Cancer Chemother Pharmacol (2015) 75:495-503.
[3] Booka E et al. Gastric Cancer (2016) 19:876-996.
Reference: PAGE 28 (2019) Abstr 9066 [www.page-meeting.org/?abstract=9066]
Poster: Drug/Disease Modelling - Oncology