Plasmodium vivax relapses caused by reactivating hypnozoites are a major barrier for elimination and control of this form of malaria. Radical cure is a form of therapy capable of addressing this problem. Recent clinical trials of radical cure have yielded efficacy estimates ranging from 65% to 94%, with substantial variation across trial sites. We performed an analysis of simulated trial data using a transmission model to demonstrate that variation in efficacy estimates across trial sites can arise from differences in the conditions under which trials are conducted. Our analysis revealed that differences in transmission intensity, heterogeneous exposure, and relapse rate can yield efficacy estimates ranging as wide as 12-78%, despite simulating trial data under the uniform assumption that treatment had a 75% chance of clearing hypnozoites. A longer duration of prophylaxis leads to a greater measured efficacy, particularly at higher transmission intensities, making the comparison of the protection of different radical cure treatment regimens against relapse more challenging. We show that vector control and parasite genotyping offer two potential means to yield more standardized efficacy estimates that better reflect protection against relapse. We predict that site-specific biases are likely to contribute to variation in efficacy estimates both within and across clinical trials. Future clinical trials can reduce site-specific biases by conducting trials in low-transmission settings where reinfections from mosquito biting are less common, by preventing reinfections using vector control measures, or by identifying and excluding likely reinfections that occur during follow-up using parasite genotyping methods.