Cancer of unknown primary site (CUP) may be called an "orphan" disease, as it is diagnosed when metastases are detected while the primary tumor typically remains undetected, and because little research has been done on its primary causes. So far, few epidemiological studies, if any, have addressed possible risk factors for CUP. We analyzed data from the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort (N=476,940). During prospective follow-up, a total of 651 cases of incident cases of CUP were detected (ICD-O-2 code C809). Proportional hazards models were conducted to examine the associations of lifetime history of smoking habits, alcohol consumption, levels of education and anthropometric indices of adiposity with risk of being diagnosed with CUP. Risk of being diagnosed with CUP was strongly related to smoking, with a relative risk of 3.66 [95% C.I., 2.24-5.97] for current, heavy smokers (26+ cigarettes/day) compared to never smokers (adjusted for alcohol consumption, body mass index, waist circumference and level of education) and a relative risk of 5.12 [3.09-8.47] for cases with CUP who died within 12 months. For alcohol consumption and level of education, weaker associations were observed but attenuated and no longer statistically significant after adjusting for smoking and indices of obesity. Finally, risk of CUP was increased by approximately 30 per cent for subjects in the highest versus lowest quartiles of waist circumference. Our analyses provide further documentation, in addition to autopsy studies, that a substantial proportion of cancers of unknown primary site may have their origin in smoking-related tumors, in particular.