Tumor organoids modeling reveals timed responses and interplay of radiotherapy and chemotherapy in pancreatic cancer.

PURPOSE: Model the therapeutic effects of chemotherapy, radiotherapy, and chemoradiotherapy to study the heterogeneous responses of patient-derived tumor organoids (PDTOs).

METHODS

We proposed novel mathematical forecasting frameworks based on logistic growth ordinary differential equations (ODEs) and the characteristics of tumor responses to model the therapeutic effects of chemo- and radiation-induced killing. To validate the models, we cultured PDTOs from different patients and treated them with radiation (4 Gy and 8 Gy), chemotherapy (FOLFIRINOX at a previously determined organoid-specific IC50 dose), and combined regimens, respectively. The diameters of 20-40 organoids per patient were tracked and measured using brightfield images up to 7 or 9 days following each treatment to capture organoid growth dynamics data, which was used for model fitting. The accuracy of the modeling was evaluated by the average normalized mean squared errors (NMSE) of data fitting.

RESULTS

The proposed mathematical modeling frameworks accurately captured the observed growth dynamics of three PDTO samples after chemotherapy, radiotherapy, and chemoradiotherapy, as reflected by the average NMSE of data fitting, which are all close to zero (less than 0.0045). The fitted parameters, including killing strength, effect window, and peak killing timing, quantitatively revealed significant heterogeneities in treatment responses across different PDTOs. Chemotherapy shows pronounced effectiveness in the early stage, while radiotherapy exhibits the effect later in the first week (around day 4), and a significantly stronger secondary response occurs one week after radiation (around day 8). Chemoradiotherapy combines the strengths of both modalities, producing pronounced effects in both response phases, with modeling results suggesting that the radiation-induced killing effect may play a dominant role in the combined interaction.

CONCLUSIONS

Our modeling frameworks demonstrated high accuracy in modeling the heterogeneous therapeutic responses of PDTOs and provided insights into the dynamic killing effects and interplay between chemotherapy and radiotherapy. The modeling of therapeutic responses of PDTOs provides a valuable tool for optimizing treatment regimen and informing clinical trial design, thereby improving the efficacy of personalized medicine.