Purpose: To investigate the potential in substantially improving Glioblastoma Multiforme (GBM) radiotherapy outcome through personalized spatial dose distributions with 4π radiotherapy and temporal dose fractionation schedule optimization with patient-specific biological models.

Methods: An ordinary differential equation (ODE) model with consideration of cancer stem cell (CSC) dynamics that incorporates the distinct radiosensitivity between CSC and its non-stem counterpart, differentiated cancer cells (DCC) has been developed and shown to be capable of reflecting the definitive treatment failure of GBM was developed. Seven patient-specific models were fitted to match the known times to GBM recurrence of these patients. Recurrence volume of each patient was transferred to generate hypothetical subvolumes with higher tumor aggressiveness on the original clinical plan to receive simultaneous integrated boost (SIB) to study the compound effect in outcome improvement arising from spatial and temporal dose optimization. For each patient, the boost dose is maximized subject to the constraints maintaining acceptable dose to surrounding OARs and coverage to the original planning target volume. With the patient-specific biological models and boost dose, a dose fractionation schedule optimization (FSO) problem with the time interval between fractions and the dose to both the non-boost and boost volumes as variables was formulated and solved with a paired simulated annealing algorithm for boost volumes with a wide range of CSC concentrations.

Results:Simultaneous integrated boost (SIB) dosage of up to 245 Gy within a 60 Gy PTV was shown to be feasible with the 4π SIB optimization formulation. Statistically significant OAR sparing was still achieved with 4π SIB compared with the originally delivered clinical plan with no boost. FSO resulted in high dose fractions in the beginning of the treatment course, followed by relatively constant dose fractions. Scenarios with lower CSC concentration within the boost volume resulted in fractionation schedules with dense once per day fractions in the beginning followed by a long time interval in the end with no treatment. With boost volume CSC concentration increased by 100 fold, maximum recurrence delay of up to 392 days was observed for a patient with the slowest growing disease.

Conclusions: By combining the spatial dose sparing power of 4π radiotherapy and temporal dose fractionation optimization with a CSC dynamics biological model in a personalized manner, significant potential in GBM disease recurrence delay was demonstrated across a cohort with differing disease characteristics. Further investigation is needed to validate the proposed model and resultant dose fractionation schedules to fully realize and translate these substantial clinical benefits.