While adaptive cancer therapy starts to be a promising approach of building evolutionary dynamics into therapeutic scheduling, the stochastic nature of cancer evolution has rarely been incorporated. In this talk, we show a method that can effectively select optimal adaptive treatment policies under randomly evolving tumor dynamics based on the Stochastic Optimal Control theory. We first construct a stochastic model of cancer dynamics, which is then used to improve the cumulative treatment “cost” - a combination of the total amount of drugs used and the time till recovery. We then develop a numerical method to solve the HJB equation to maximize the probability of recovery at any threshold cost. We will show the recovered threshold-aware optimal policies provide several advantages compared to the policies previously shown to be optimal in the deterministic setting. Both theoretical and numerical results will be presented based on my joint work with Alex Vladimirsky and Jacob Scott.