To examine confounding on the risk assessment of the radiation workers at nuclear facilities in Japan, a questionnaire survey of their characteristics such as life-style and occupational history was performed for 54,369 male and 470 female workers who were currently engaged in the job and valid answers were obtained from 48,281 males and 428 females. In order to know whether these characteristics were different among different dose groups, the Mantel extension statistical test was performed only for male respondents, with cumulative radiation doses stratified into 5 classes. Increasing trend according to the increasing doses was statistically significant for the percentages of tobacco smokers and of heavy smokers. It was also the case for heavy alcohol drinkers. Percentages of workers who were engaged in jobs dealing with specific toxic materials were also increasing in the higher dose groups. On the other hand, percentage of workers who underwent the X-ray examination of the upper digestive tracts or other radiological examinations tended to be lower in higher dose groups. These results indicate that characteristics of radiation workers such as life-style are different among dose groups and thus may play a role as a confounding factor in the statistical relation between the radiation doses and cancer mortalities.