Much of what is known about the deleterious effects of tobacco use on health was learned from epidemiologic studies over the last half century. These studies establish unequivocally that tobacco use, particularly manufactured cigarette smoking, causes most cancers of the lung, oropharynx, larynx, and esophagus in the USA, and approximately one-third of all cancers of the pancreas, kidney, urinary bladder and uterine cervix. More recent evidence also implicates smoking with cancers of the stomach, liver and colorectum. While over half of the estimated 440 000 smoking-attributable deaths that occur annually in the USA involve non-malignant cardiovascular and respiratory conditions, smoking-attributable cancers are more recognized and feared. Geneticists increasingly study tobacco use as a model for environmental carcinogenicity. Tobacco-exposed populations provide opportunities to characterize the somatic mutations that give rise to specific cancers and to identify the inherited genetic traits that confer susceptibility or resistance. Studies to identify the genetic determinants of addiction may be particularly important. Future research to identify other susceptibility factors, such as genes that modify carcinogen metabolism or DNA repair, will need to be substantially larger and to quantify lifetime tobacco exposure with more precision than have past studies in order to distinguish gradations in risk due to exposure from those caused by genetic susceptibility. This review considers: (a) the epidemiology of tobacco use; (b) cancers presently classified as smoking-attributable by the US Surgeon General; (c) the magnitude of the epidemic of cancers and other diseases caused by tobacco use; (d) selected issues in the epidemiology of lung cancer; and (e) the interface of genetics and epidemiology in understanding, preventing, and treating tobacco-attributable disease.