BACKGROUND: Wastewater use in urban agriculture is common as a result of rapid urbanisation, and increasing competition for good quality water. In order to minimize risks to farmers and consumers of wastewater irrigated produce the World Health Organization (WHO) has developed guidelines for the safe use of wastewater in agriculture. These guidelines are based on a Quantitative Microbial Risk Assessment (QMRA) model, though the reliability of this model has been questioned due to a lack of primary data. This study aimed to assess the ability of the WHO guidelines to protect farmers' health, by identifying and quantifying key exposures associated with the transmission of faecal pathogens in wastewater irrigated agriculture. METHODS: Eighty farmers were observed and interviewed during the dry and wet seasons, and water and soil samples were analysed for the presence of E. coli. STATA 12 was used for descriptive analyses of farmers' exposure and risk practices, and also to determine risk factors for soil and irrigation water contamination, while the WHO QMRA model and @Risk 6 were used to model farmers' infection risk to pathogens. RESULTS: The results showed that although irrigation water was highly contaminated (5.6 Log E. coli/100 ml), exposure to farm soil (2.3 Log E. coli/g) was found to be the key risk pathway due to soil-to-mouth events. During the observations 93 % of farmers worked barefoot, 86 % experienced hand-to-soil contact, while 53 % experienced 'soil'-to-mouth events, while no 'water' to mouth contacts were observed. On average, farmers were found to have 10 hand-to-mouth events per day. From the indicator based QMRA model the estimated norovirus infection risk to farmers was found to be higher than guidelines set by the WHO. CONCLUSIONS: This study found exposure to soil as the critical pathway of pathogen risk in wastewater farmers, and that this risk exceeded recommended health targets. The study recommends the incorporation of hand-to-mouth events, the use of actual pathogen concentrations, and the use of direct exposure frequencies in order to improve the reliability of risk estimates from QMRA models.