Aims—Epidemiologic, pharmacoepidemiologic and pathophysiologic evidence points consistently to an association between type 2 diabetes and cancer. This asso-ciation could be explained by hyperinsulinemia induced by insulin resistance. We studied the association between fasting serum insulin (FSI) and cancer mortality in a population of non-diabetic individuals. Methods—We followed 3117 healthy workers (50.2% women), included in the TEL-ECOM cohort study, between 1985 and 1987; their median age was 38 years (Q1-Q3 = 30-50). Baseline FSI was measured by radioimmunoassay, the INSI-PR meth-od. People with diabetes or cancer at baseline were excluded. Vital status and caus-es of death were available until December 2013. The association between FSI and cancer deaths was analyzed by sex, using a Cox proportional hazards model with age as the time scale, adjusting for body mass index, smoking habits, alcohol con-sumption, occupational category and ethnic origin. Results—After 28 years follow-up, 330 (10.6%) deaths were reported, among which, 150 were cancer-related (80 men, 70 women). In men, the association between FSI and death by cancer was J-shaped: compared to the average FSI of 7.1 mU/l, men with 5 mU/l and 12.9 mU/l had respectively adjusted Hazard-Ratios (HR) of 1.88 (95% Confidence Interval, 1.00-3.56) and 2.30 (95% CI, 1.34-3.94). Among women, no significant association was found (adjusted HR, 1.03; 95% CI, 0.96-1.11) for an increase of 1 mU/l in FSI. Conclusion—These results strengthen the hypothesis of an independent risk of cancer death associated with extreme values of FSI, mainly the highest, among men, but not among women.