Thirty-nine Greek dairy herds, totalling 6333 cattle, enrolled in a voluntary bovine viral diarrhoea virus (BVDV) eradication programme based on the identification and removal of persistently infected (PI) animals. The aim of this study was to estimate the prevalences of BVD antigen-positive and PI animals, and investigate the significance of the associations between the prevalence estimates and herd size. Initially, all animals were bled and examined for BVDV, using an antigen ELISA. A second sample was collected from the positive animals, after a period of at least three weeks. Animals retested positive were classified as PI. Antigen positive and PI animals were detected in all herds. The respective mean prevalences, adjusted for the test's accuracy and the herd-clustering effect, were 14% (95%CI: 11-18%) and 1.3% (0.8-1.8%), respectively. Herd size was not associated with the prevalence of antigen-positive or PI animals.