An optimal posttreatment surveillance strategy for cancer survivors based on an individualized risk-based approach

The optimal post-treatment surveillance strategy that can detect early recurrence of a cancer within limited visits remains unexplored. Here we adopt nasopharyngeal carcinoma as the study model to establish an approach to surveillance that balances the effectiveness of disease detection versus costs. A total of 7,043 newly-diagnosed patients are grouped according to a clinic-molecular risk grouping system. We use a random survival forest model to simulate the monthly probability of disease recurrence, and thereby establish risk-based surveillance arrangements that can maximize the efficacy of recurrence detection per visit. Markov decision-analytic models further validate that the risk-based surveillance outperforms the control strategies and is the most cost-effective. These results are confirmed in an external validation cohort. Finally, we recommend the risk-based surveillance arrangement which requires 10, 11, 13 and 14 visits for group I to IV. Our surveillance strategies might pave the way for individualized and economic surveillance for cancer survivors.