To study the incidence of and mortality from cancer among sewage workers a retrospective analysis was performed on a cohort of 656 men employed for at least one year at any one of 17 Swedish sewage plants during the years 1965-86. Assessment of exposure was done by classification of work tasks. Lower than expected total mortality (standardised mortality ratio (SMR) = 0.75, 95% confidence interval (95% CI) 0.58-0.97) and cardiovascular mortality (SMR = 0.61, 95% CI 0.39-0.91) was found. This was interpreted as a result of the healthy worker effect. For all cancers combined the mortality (SMR = 1.08, 95% CI 0.68-1.67) and morbidity (SMR = 1.02, 95% CI 0.72-1.38) were comparable with those of the general population. There were increased incidences for brain tumours (SMR = 2.19, 95% CI 0.45-6.39), gastric cancers (SMR = 2.73, 95% CI, 1.00-5.94), and renal cancers (SMR = 1.68, 95% CI = 0.35-4.90). For lung cancer the risk was reduced (SMR = 0.70, 95% CI 0.15-2.05). Allowance for a latency period of 10 years from the start of exposure did not change the pattern. Logistic modelling was used to search for exposure-response relations. In a logistic model with the confounder age forced in, renal cancer had a significant positive relation with a weighted sum of employment times, where the weights describe the classification of exposure. No exposure-response relations were found for brain tumors or gastric cancers. The increased risks are based on small numbers of cases. A future follow up will add more conclusive power to the study. Specific exposures need to be identified to allow for a better dose-response analysis.