To estimate the prevalence of Escherichia coli O157 on Dutch dairy herds, faecal samples were collected once from 678 randomly selected dairy farms in the period October 1996-December 2000. Samples were cultured for E. coli O157. Thirty-eight isolates were tested for virulence genes (eae, VT1 and VT2). A questionnaire about farm characteristics was taken from the farm manager, resulting in variables that could be analysed to identify and quantify factors associated with presence of E. coli O157. In total, 49 of the 678 herds (7.2%) showed at least one positive pooled sample. E. coli O157 was not isolated from herds sampled in December-April in consecutive years (except for one isolate found in March, 2000). VT- and eae-genes were found in 37 and 38 isolates, respectively. Logistic regression was performed on variables obtained from the questionnaire, comparing E. coli O157-positive herds to negative herds. To account for season, a sine function was included in the logistic regression as an offset variable. In the final model, the presence of at least one pig at the farm (OR = 3.4), purchase of animals within the last 2 years before sampling (OR = 1.9), supply of maize (OR = 0.29) to the cows, and sampling a herd in the year 1999 or 2000 (compared to sampling in 1998; OR = 2.1 and 2.9, respectively) had associations with the presence of E. coli O157.