A retrospective cohort study of 116 British pig farms was undertaken to investigate the epidemiological risk factors associated with herd breakdowns with postweaning multisystemic wasting syndrome (pmws). Farmers reported the pmws status of their herd (case definition 1) and, where applicable, when the disease was first suspected and what they observed; they described a prolonged increase in mortality in six to 16-week-old pigs that was not attributable to any disease known to be on their farm. There was over 90 per cent agreement on the farmers' pmws status between the farmers and their veterinarians. Approximately 70 per cent of the breakdowns were confirmed at the laboratory (case definition 2) except during the outbreak of foot-and-mouth disease (fmd) in 2001 when it was reduced to 30 per cent. Porcine circovirus type 2 antigen was detected in pigs examined postmortem (case definition 3) in approximately 90 per cent of the farms with increased mortality. The breakdowns occurred initially in the south of England and spread west and north, as well as locally in a radial pattern from the affected farms, and there was strong statistical evidence that there was non-random space-time clustering. The risk of herd breakdowns with pmws was not constant; therefore, for each case definition, three survival models were developed with outcome variable time to breakdown of between January 2000 and January 2001, February 2001 to September 2001 (during fmd) or October 2001 to December 2003. Exposures with a bivariable significance of P<0·20 were tested in three multivariable Cox proportional hazard models. From January 2000 to January 2001 the risk of a herd breakdown with pmws for definitions 1, 2 and 3 was greater for farms with 600 or more breeding sows, and for definitions 1 and 3 there was an increased risk associated with the purchase of replacement gilts rather than using homebred replacements. For definitions 1 and 3 the farms where the nearest pig farm had no breeding pigs were at greater risk of a breakdown than those where the nearest farm had breeding stock, as were the farms where visitors were not requested to avoid pigs for more than three days before visiting the farm during the fmd outbreak. From October 2001, the associated risks were identical for all three case definitions; farms were at greater risk when they had 600 or more breeding sows, if visitors had not avoided contact with pigs for more than three days before visiting the farm, and when there was a farm with pmws less than five miles away. The affected farms were more likely to have disease associated with porcine parvovirus, porcine reproduction and respiratory syndrome virus, erysipelas, Escherichia coli and salmonella. These exposures were positively associated with large herds and the farm being close to other pig farms, but did not remain in the final models for breakdown with pmws, indicating that such farms may be at greater risk of many infectious diseases.