Predicting tumor volume doubling time and progression-free survival curves in cancer patients from patient-derived-xenograft (PDX) models: a translational model-based population approach
Elena Maria Tosca1, Davide Ronchi1, Marzio Cossali1, Martina Zavettieri1, Maurizio Rocchetti2, Paolo Magni1
1) Department of Electrical, Computer and Biomedical Engineering, University of Pavia, Pavia (Italy); 2) Consultant, Rho (Italy)
Introduction: Tumor volume doubling time (TVDT), i.e. the time required by a tumor to doubling its volume, is typically used in clinics to summarize tumor growth (TG) dynamics. Its computation is challenging as requires at least 2 measurements of untreated tumor volume in 2 subsequent time instants. However, TVDT knowledge is essential in clinical cancer management for a multitude of medical decisions. TVDT is also strongly correlated with tumor response to therapy and, thus, to tumor progression and patient survival. Progression-free-survival (PFS) is defined as the timeframe between study randomization and tumor progression or death, whichever occurs first. Accordingly, it is inherently linked to TG dynamics. PFS is often used as primary endpoint in early clinical studies where therapeutic benefits of investigated treatments are quantified in terms of PFS timespan increment with respect to standard therapies, as a placebo-controlled comparison is rarely available.
Objective: Here, a translational modeling framework to predict TVDTs in an untreated population from TG data collected in patient-derived xenograft (PDX) mice experiments was developed. In addition, exploiting the intrinsic relationship between TG and PFS, the expected PFS curves in absence of anticancer treatment were generated. Finally, the translational population approach was extended to predict TG dynamics and PFS curves in pancreatic (PA) and liver (LI) cancer patients treated with Gemcitabine and Sorafenib, respectively.
Methods: 11 types of solid cancers were considered. For each of them, the following multistep procedure was applied. i) A panel of TG studies in PDX mice [1] was analyzed through a population modeling approach. The Simeoni model [2] was used to characterize the distribution of the exponential tumor growth rate in mice (TGRmice) and anticancer potency (K2) of Gemcitabine and Sorafenib in PA and LI PDXs, respectively. ii) Assuming an exponential tumor growth in humans, TGRmice were scaled up to humans (TGRhuman) based on allometric scaling. K2 was scaled with the same strategy. iii) TVDT distributions in untreated cancer patients were derived from TGRhuman as TVDT=ln(2)/TGRhuman. iv) Kaplan-Meier (KM) curves for PFS in untreated patients were computed from TGRhuman considering a single spherical tumor mass and only progressive disease events (20% increase in diameter). v) TG profiles in PA and LI cancer patients treated with Gemcitabine and Sorafenib, respectively, were simulated linking patient PK models with the scaled TGI model (drug affecting the exponential TGR though K2 potency) and used to build the corresponding PFS curves.
Results: TVDT predictions were compared to the literature values in untreated cancer patients, resulting in an outstanding agreement. The predicted and observed TVDT medians were very closed, with predictions within 2-fold of observations for almost all the cancer types (10/11). In addition, 76% of individual TVDTs, derived from the PDX models, fell within the ranges reported in clinical studies (80% CI). Simulated KM-PFS curves in absence of treatment were overlaid to published PFS data from different clinical trials including patient cohorts receiving placebo or active treatments exerting little to no effect on TG. In most of the cases the observed and simulated curves were closed throughout the entire time course. These results were confirmed by the agreement between the observed and predicted median PFS (i.e. time at which the PFS curve crosses the 50% point). Similarly, the simulated KM-PFS curves in PA and LI patients treated with Gemcitabine and Sorafenib, respectively, were compared to the appropriate clinical PFS data. A very good agreement was found in both the case studies. In addition, predicted TG dynamics in PA cancer patients under Gemcitabine treatment were successfully validated against longitudinal data of tumor size [3].
Conclusions: The proposed model-based translational framework is able to predict TVDTs in untreated cancer patients from TG data in PDX mice, providing a powerful tool to increase the knowledge on TVDT. In addition, the expected PFS curves in absence of treatment can be predicted, potentially filling the lack of placebo-controlled arms against which to compare experimental agents during clinical trials. Finally, the potential of early predicting the treatment effect on PFS starting from a panel of TGI studies in PDX mice has been successfully proved in two cases studies.
References:
[1] HuBase database, Crownbio Bioscience Inc., https://www.crownbio.com/).
[2] Simeoni M, Magni P, Cammia C, Nicolao G De, Croci V, Pesenti E, et al. Predictive pharmacokinetic- pharmacodynamic modeling of tumor growth kinetics in xenograft model after administration of anticancer agents. Cancer Res. 2004; 64:1094–101.
[3] Garcia-Cremades, M., Pitou, C., Iversen, P. W., & Troconiz, I. F. Predicting tumour growth and its impact on survival in gemcitabine-treated patients with advanced pancreatic cancer. European Journal of Pharmaceutical Sciences. 2018; 115, 296-303.