The proportion of lung cancer due to occupational exposure has been estimated by computing the attributable risk in the population (ARp) from various case-control studies. Different criteria have been used in including occupational exposures from published studies: (i) exposure to lung carcinogens according to a job-exposure matrix approach; (ii) occupations in which an increased risk of lung cancer has been established; (iii) occupations in which an increased risk of lung cancer has been suggested; and (iv) occupations associated with a statistically significant increase of lung cancer. Only studies in which the confounding effect of smoking was controlled for have been included. We found a wide variability of ARp estimates, mainly due to the different geographical locations of the studies, while they seem to be only moderately affected by the different criteria of inclusion of the relevant occupational exposures. Studies using job-exposure matrices gave ARps ranging from 0.6% to 35%. When a list of recognized carcinogenic exposures was used for the selection of the relevant occupations, the estimates varied between 2.4% and 40%. From the studies reviewed it also appears that tobacco smoking has a very limited confounding effect. Various limitations of the exercise are discussed.