Using a Markov process to simulate and assess surveillance policies for stage I nonseminomatous germ cell tumors, the authors analyzed costs and effectiveness for the various policies that have been used clinically. They found no real differences in effectiveness, defined as probability of surviving two years, between the policies. There were significant differences in costs. The model is used to suggest schedules that might be more cost-effective than those currently employed. The model suggests that computed tomographic scanning of the abdomen may not be essential to the effectiveness of surveillance protocols.