Assessing the role of model choice in parameter identifiability of cancer treatment efficacy

Several mathematical models are commonly used to describe cancer growth dynamics. Fitting of these models to experimental data has not yet determined which particular model best describes cancer growth. Unfortunately, choice of cancer growth model is known to drastically alter the predictions of both future tumor growth and the effectiveness of applied treatment. Since there is growing interest in using mathematical models to help predict the effectiveness of chemotherapy, we need to determine if the choice of cancer growth model affects estimates of chemotherapy efficacy. Here, we simulate an in vitro study by creating synthetic treatment data using each of seven commonly used cancer growth models and fit the data sets using the other (“wrong”) cancer growth models. We estimate both the ε max (the maximum efficacy of the drug) and the IC 50 (the drug concentration at which half the maximum effect is achieved) in an effort to determine whether the use of an incorrect growth model changes the estimates of chemotherapy efficacy parameters. We find that IC 50 is largely weakly practically identifiable no matter which growth model is used to generate or fit the data. The ε max is more likely to be practically identifiable, but is sensitive to choice of growth model, showing poor identifiability when the Bertalanffy model is used to either generate or fit the data.