Computational model of treatment resistance heterogeneity

In metastatic cancer patients, diverse levels of resistance, which are the result of genetic heterogeneity, lead to treatment response (TR) heterogeneity. To evaluate the role of resistance on TR, we constructed a population model, simulating cellular dynamics in individual metastasis. The model was benchmarked with imaging metrics extracted from the ¹⁸F-NaF PET/CT scans of 39 metastatic prostate cancer patients, received at baseline and after 3 cycles of therapy. Patients were treated with chemotherapy or hormonal therapy. Two model settings were evaluated: one considering only inter-patient and one considering both inter- and intra-patient heterogeneity in the proportion of intrinsically resistant cells (IR). Model performance, considering both settings, was evaluated using the Akaike information criterion (AIC). TR after 6, 9, and 12 months was predicted and compared using the Wilcoxon rank sum test. Considering both inter- and intra-patient heterogeneity in IR resulted in significantly better model performance (AIC=-250) than considering only inter-patient heterogeneity (AIC=6). Differences in predicted TR were not significant between treatment groups (p>0.15). The model has identified IR as an important factor influencing on inter- and intra-patient TR heterogeneity.