Adaptive and sequential cancer therapies emerge from treatment schedule optimization

Most cancer chemotherapy is delivered at the maximum tolerated dose (MTD) as often as possible, switching to a new combination of drugs upon progression. Clinical trial results have demonstrated that this sequential MTD schedule is not optimal for every patient in every situation; however, there is no systematic program for discovering what the optimal schedule actually is. We developed a simulated-annealing-based algorithm for computing personalized dosing schedules that best achieve treatment objectives. Given a mathematical model of a patient's drug response, the algorithm returns the optimal schedule for either minimizing tumor burden or maximizing progression-free survival. We tested this algorithm on a simple mathematical model of patient treatment response; we found that there are only three optimal one-drug schedules: continuous therapy and two distinct adaptive therapies. For two drugs, each schedule used in the clinic was found to be optimal for some patient parameters, and the algorithm was able to quantitatively map out the boundaries between these schedules. The optimization algorithm developed here can play a key role in upcoming personalized cancer treatment pipelines.