Although it is well established that highly penetrant genes explain less than 5% of all cancers, it is much less clear what proportion is attributable to low penetrant genes and their interactions with environmental exposures. It was possible to estimate indirectly the fraction of lung and bladder cancers attributable to known genetic polymorphisms, on the basis of an extensive review of the literature, and of the pooled analyses of approximately 15000 healthy subjects in three continents. The emerging picture is one of considerable homogeneity within ethnic group, and heterogeneity among ethnic groups. By combining relative risks and genotype frequencies we have computed theoretical attributable risks for lung and bladder cancers and the CYP1A1 Msp1, CYP1A1 Exon 7, GSTM1 and NAT2*5 genotypes, among Caucasians and among Asians. Such attributable risks are probably overestimated, since: (a) they include the interaction with carcinogenic exposures, in the absence of which genetic polymorphisms per se are not effective; (b) they do not reflect the combination of different genotypes. However, our calculations suggest that attributable risks due to genetic susceptibility are much lower than those related to smoking or other environmental risk factors, and that Asians seem to carry higher risks than Caucasians. In addition, a theoretical approach to the problem of gene-environment interactions at low levels of exposure is offered.