The purpose was to investigate the association between serum osteoprotegerin (OPG) and risk of incident cancer and cancer mortality in a general population. OPG was measured in serum collected from 6,279 subjects without prior cancer recruited from a general population. Incident cancer and cancer-related mortality were registered from inclusion in 1994-95 until end of follow-up December 31, 2008. Cox regression models were used to estimate crude and adjusted (for age, sex and other confounders) hazard ratios and 95% confidence intervals (HR 95% CI). There were 948 incident cancers and 387 deaths in the cohort during 71,902 person-years of follow up (median 13.5 years). Subjects with serum OPG in the upper tertile had 79% higher risk of incident gastrointestinal cancer than those in the lowest tertile (HR 1.79, 95% CI 1.19-2.67). In women <60 years, serum OPG (per SD 0.81 ng/ml) was associated with reduced risk of incident cancer (all cancers merged; 0.73; 0.57-0.94) and breast cancer (0.51; 0.31-0.83) after adjustment. Subjects in the upper tertile of OPG had higher risk of cancer-related mortality (1.63; 1.16-2.28), particularly mortality from cancer in the gastrointestinal system (2.28; 1.21-4.28) compared to those in the lowest OPG tertile. No significant association was detected between OPG and risk of death from cancer in the respiratory system or death from prostatic cancer. Our findings from a large population based cohort study suggest that serum OPG was associated with increased risk of incident gastrointestinal cancer, inversely associated with breast cancer, and predicts cancer-related mortality.