Objectives: Visceral leishmaniasis (VL) is a public health threat for several tropical countries, including Brazil. Therapy failures and relapses aggravate VL morbidity and mortality. Our study aimed at identifying predictors of relapse and thus contributes to directing therapeutic options and patient followup. Methods: A nonconcurrent cohort of 571 subjects who completed successful therapy for VL in the city of Bauru, São Paulo State, Brazil, was followed for 24 months in order to identify the incidence and predictors of relapse. Extensive review of medical charts and laboratory files was conducted. Univariate and multivariable Cox regression models were used to identify predictors for the outcome of interest. A hierarchical strategy was used for variable selection in multivariable models. Results: Relapses occurred in 6.8% of treated subjects, after a median of 6 months (interquartile range, 4-9). In a comprehensive multivariable model, relapse was associated with: HIV-coinfection (hazard ratio [HR], 7.47; 95% confidence interval [CI], 2.58-21.55); the presence of lower limb edema (HR, 6.06; 95%CI, 1.38-26.77) and low platelet count upon admission (HR for platelet count  1000, 0.99; 95%CI, 0.98-0.99) ; and secondary pneumonia (HR, 5.49; 95%CI, 1.49-20.18). On the other hand, therapy with Liposomal Amphotericin (as opposed to Antimoniate) was not independently associated with relapse (HR, 5.97; 95% CI, 0.63-56.29). Conclusion: Besides reinforcing the impact of HIV coinfection on the outcome of VL, our study points to clinical and laboratory findings that characterize patients who were more likely to relapse. Those groups should be more closely followed, and possibly could benefit from novel therapeutic options.