BACKGROUND Infection with Ehrlichia canis causes a highly variable, multisystemic disease in dogs. Nevertheless, many clinicians in Rio de Janeiro, Brazil, use the presence of only thrombocytopenia to make a presumptive diagnosis of E canis infection. OBJECTIVE The objective of this study was to determine the prevalence of E canis in thrombocytopenic dogs from Rio de Janeiro, Brazil, using polymerase chain reaction (PCR). METHODS Following DNA extraction of whole blood samples from 226 dogs, PCR assays were done using primers for rickettsial DNA (including Ehrlichia spp, Anaplasma platys and A phagocytophilum) and using E canis-specific primers (16S rRNA gene). Dogs were grouped as thrombocytopenic and nonthrombocytopenic based on platelet counts. The null hypothesis that there was no difference in the prevalence of E canis in these groups was rejected at P<.05. RESULTS Thirty-six (32.1%) of the thrombocytopenic dogs and 4 (3.5%) of the nonthrombocytopenic dogs were positive for rickettsial gene sequences (P<.0001). Further, 30 (26.8%) of thrombocytopenic dogs and 4 (3.5%) nonthrombocytopenic dogs were positive for E canis-specific gene sequences (P<.0001). CONCLUSIONS Although the prevalence of E canis infection was higher in thrombocytopenic dogs, less than one third of these dogs had demonstrable E canis infection. Thus, thrombocytopenia is not specific for the detection of E canis infection and should not be used solely to establish a diagnosis of canine ehrlichiosis, even in a geographic area with relatively high disease prevalence.