The shape of the dose-response relationship between carcinogenic exposure and cancer risk is a key issue, both from a theoretical (models of carcinogenesis) and practical (risk assessment) point of view. Human populations exposed to Polycyclic Aromatic Hydrocarbons (PAH) via air pollution showed a non-linear relationship between levels of exposure and WBC-DNA adducts. Among highly exposed subjects, the DNA adduct level per unit of exposure was significantly lower than measured at environmental exposures. The same exposure-dose non-linearity was observed in lung DNA from rats exposed to PAH. We have analyzed 11 case-control studies on bladder cancer (4584 incident cases and 9360 hospital controls) and eight case-control studies on lung cancer (5092 incident cases and 6083 population controls), conducted in Europe in recent years. All the studies collected detailed information on smoking histories with a similar methodology. We have estimated the relationship between the number of cigarettes smoked and the risk of cancer, with and without adjustment by duration of smoking. We have observed a levelling-off of the relationship between the number of cigarettes smoked and the relative risks for lung and bladder cancer, both in men and women. The levelling-off occurred at an odds ratio of about 5 for bladder cancer, while it occurs at about 20 for lung cancer (in men). A potential explanation for such levelling-off involves metabolic pathways and individual susceptibility. It has been suggested that some metabolic polymorphisms exert an effect that is more important at low levels of exposure.