Objective: To prospectively study the relation between TSH and risk of hip and forearm fractures. Design: A population-based cohort study. Methods: In a substudy of the second survey of the Nord Trøndelag Health Study, Norway (HUNT2, 1995-97), linked with a hospital-based fracture registry, we investigated the relation between baseline TSH and risk of hip and/or forearm fractures. Population: A total of 16 610 women and 8595 men aged 40 years or more, without previous self-reported thyroid disease and hip or forearm fractures. Results: During 12.5 years follow-up, a total of 1870 women and 342 men experienced hip or forearm fractures. Overall, there was no relation between baseline TSH and fracture risk. However, there was weak evidence that women with TSH !0.5 and O3.5 mU/l had a slightly increased risk of hip fractures (hazard ratio (HR) 1.30, 95% CI 0.97-1.94 and HR 1.19, 95% CI 0.93-1.52) compared with the reference group with TSH of 1.5-2.4 mU/l. Supplementary analyses showed higher hip fracture risk in women with TSH O4.0 mU/l and negative thyroid peroxidase antibodies (TPOAb) compared with the reference group (HR 1.75, 95% CI 1.24-2.46). Conclusion: We found no statistically significant relation between baseline TSH and subsequent fracture risk, but the data suggest a weak positive association with hip fracture risk among women with both low and high TSH. The latter association was confined to women with negative TPOAb status.