To investigate the relationship between exposure to polycyclic aromatic hydrocarbons (PAHs) and mortality for specifc cancer sites, 1152 men, employed for at least 1 year at a prebake aluminium smelter, were followed-up from 1972 until 31 December 2001. Exposure to PAHs was estimated from a detailed reconstruction of the working history experienced in the plant by each cohort member and from several environmental and personal shift-sampling measurements available, by task and working department, since 1979. Furthermore, information on smoking habits, previous jobs before engagement in the smelter and main clinical findings observed during the follow-up were collected from the personal medical files. This study showed no increased mortality for lung cancer or bladder cancer associated to exposure to PAHs. Mortality for pancreatic cancer, based on 6 observed deaths, was significantly higher than expected in the whole cohort (SMR 2.4; 95%CI 1.1-5.2) and particularly among workers employed in the anodes factory of the plant (SMR 5.0, 95%CI 2.1-12.1), where a relatively consistent exposure to PAHs has been estimated. The nested case-control study planned for pancreatic cancer cases, confirmed that, also after controlling for cigarette smoking, PAH exposure experienced in the anodes factory was associated with a significant increased risk of pancreatic cancer. A pre-existent diabetes mellitus and a potential occupational exposure to pesticides experienced in previous agricultural jobs were found as concurrent significant covariates increasing the risk. In conclusion, the relatively high exposure to PAHs, experienced in the anodes factory and particularly in the green-mill department of this prebake aluminium reduction plant, cannot be ruled out as one of the main factors in the multifactorial aetiology of the pancreatic cancers observed in this study.