In 1998, a total of 424 sows had sera collected in the Mekong delta in Vietnam. Of these, 283 sows were from 151 small-scale family farms in 19 villages, and 141 from seven large-scale state farms. The sera were subjected to the microscopic agglutination test (MAT) for antibodies to 13 Leptospira serovars. The overall leptospiral seroprevalence for titres > or =1:100 and > or =1:400, was 73 and 29%, respectively, and was higher (P=0.001) at small- than at large-scale farms. The highest seroprevalence was recorded for Leptospira interrogans serovar (sv) bratislava (52%). At small-scale farms, higher prevalences were found to serovars L. interrogans sv icterohaemorrhagiae (P=0.04) and L. interrogans sv pomona (P=0.02). Epidemiological information (at the individual-animal and herd-levels) was collected with a questionnaire. The data were analysed using logistic multiple regression. At the animal-level, sows seropositive for L. interrogans sv australis and sv autumnalis had less direct contact with sows in neighbouring pens (odds ratio (OR)=0.3 and 0.4, respectively) and sows seronegative for L. interrogans sv bratislava were of lower age (OR=0.1 for seropositivity). Also, sows seropositive for L. interrogans sv icterohaemorrhagiae had higher odds (OR=5.8) if they had not been born on the farm (had been introduced to it as gilts). Herds seropositive for sv javanica showed association with farms not taking measures to control the local rodent population (OR=7.8). Serovar pomona was also linked to the use of artificial insemination (AI), as opposed to natural-breeding services (OR=11.2). These results indicate that housing and management could affect the seroprevalence of Leptospira infection in pigs.