Sex steroids have an important role in bone health, however previous studies on fracture risk have been carried out in older populations. The EPIC-Oxford study is a prospective cohort of men and women living in the UK. Five years after recruitment, participants self-reported previous fractures. Sex steroid concentrations (plasma estradiol, testosterone and sex hormone binding globulin) were measured in 436 cases (155 men, 46 premenopausal women and 235 postmenopausal women) with an incident fracture and 868 matched controls. Fracture risk was inversely related to concentrations of estradiol among men (RR for a doubling of estradiol 0.35, 95% CI 0.44-0.96) but there was no association between fracture risk and testosterone levels. There were no clear associations between fracture risk and hormone levels among postmenopausal women, however there was suggestion of an inverse association for both estradiol and testosterone as the RR in the highest compared with the lowest tertile for estradiol was 0.74 (95% CI 0.46, 1.18) and testosterone was 0.75 (95% CI 0.49, 1.16). Among premenopausal women fracture risk was inversely associated with levels of testosterone (RR for doubling of testosterone 0.46, 95% CI 0.26-0.81), with no association between estradiol and fracture risk. SHBG was not associated with risk of fracture among either men or women. In summary, this study finds evidence of an inverse association between endogenous estradiol and risk of fracture in men, and between endogenous testosterone and risk of fracture in premenopausal women but no clear associations among postmenopausal women.