In Brazil, zoonotic visceral leishmaniasis (ZVL) control programs based on the mass elimination of seropositive dogs have failed to reduce the number of leishmaniasis cases. However, these programs have been done under sub-optimal conditions. We studied a cohort of dogs in an urban area in Brazil to determine, whether a dog-culling program optimized with: (i) replacement of a relatively low-sensitivity indirect immune-fluorescent test on blood eluate by a more sensitive enzyme-linked immunosorbent assay on serum blood samples; (ii) shortening of the time interval from serodiagnosis to removal of dogs; (iii) screening a high proportion of the dog population could reduce the incidence of canine Leishmania infection (CLI). The study ran from December 1997 to July 2000, with four follow-up assessments performed at approximately 8-month intervals. All dogs seropositive for anti-Leishmania antibodies were promptly eliminated. A large number of new dogs immigrated to the study area throughout the study period. They comprised 43.8–49.8% of the cohort at each follow-up assessment, and upto 15% of them already had Leishmania infection. Overall, 42 news cases of CLI were identified, for a crude incidence rate of 11.8 cases per 100 dog-years (95% CI 8.6–15.6). In the first, second, third and fourth follow-up assessments the incidence rates were 8.2 (95% CI 3.0–17.9), 12.2 (95% CI 6.3–21.2), 16.4 (95% CI 8.5–28.6) and 13.6