All Times EDT
Keywords: Subsequent Therapy, Q-Learning method
Patients with cancer or other recurrent diseases may undergo a long process of initial treatment, disease recurrences and salvage treatments. It is important to account for the salvage treatments sequence in this process to maximally prolong patients’ survival. Randomized trial comparing initially randomized treatments based on overall survival must account for the subsequent treatments to facilitate unbiased inference. In most cases the subsequent treatment is not randomly assigned and their choice might depend on a patient’s responses to initial treatment. In this article, we describe a modification of Q-learning method to optimize the subsequent treatment utilizing all the longitudinal data collected during the treatment process, and then compare the initially planned treatments under the optimal subsequent treatment (s). The application of this method is illustrated using data from a study of acute myeloid leukemia.