Background: A range of medication options are available for non-metastatic castration-resistant prostate cancer (nmCRPC), with varying efficacy and safety profiles. We used a discrete choice experiment (DCE) to explore how physicians make treatment decisions between adverse events (AEs) and survival gains in nmCRPC in mainland China. Methods: Physicians’ preferences for medication attributes were assessed through an online survey including 14 treatment choice questions, each comparing 2 hypothetical treatment profiles, which varied safety (fatigue, skin rash, cognitive problems, falls, and fractures) in terms of severity and frequency and efficacy (overall survival [OS] and time to pain progression [TPP]) in terms of duration of effect. Preference weights and importance scores for each attribute were estimated using random parameters logit models. The amount of survival gain physicians were willing to trade off for a reduction in specific AEs was also estimated. Results: In total, 80 physicians who had experience in treating nmCRPC patients were involved in this study. Physicians’ duration of clinical practice was about 18.4±7.39 years. In addition, 57.6% of participants were from regions with more developed economics, including North, East and South China. The results indicated that physician preferences were logically ordered, with higher efficacy and lower risks for AEs being favored over lower efficacy and higher risks for AEs. The most important attributes for treatment decision were OS (Relative importance, RI = 29.2%), serious fracture (RI = 24.4%), serious fall (RI =15.4%) and TPP (RI =10.6%). Comparing non-serious AEs including fatigue, skin rash and cognitive problems, physicians emphasized the risk reduction of fatigue the most. The trade-off analysis between OS and adverse events suggested that physicians were willing to trade 7.5 and 4.7 months of OS to reduce the risk of severe fracture from 8% to none, and the risk of serious fall from 8% to none. In an exploratory subgroup analysis, it was found that physicians with more than 15 years of clinical practice put more emphasis on overall survival, risk of serious fracture, and risk of serious fall compared to less experienced physicians (≤15 years). Physicians from regions with more developed economics tended to put more value on reduction of AEs as compared to physicians from the rest regions. Conclusions: OS is the most important attribute for physicians when treating nmCRPC patients. Physicians were willing to trade the duration of OS to reduce the risk of serious adverse events. These results emphasize the importance of carefully balancing therapies’ benefits and risks to ultimately optimize the overall quality of nmCRPC patients’ survival.