Objective:The aim of the study was to examine the association of tested TSH with age, gender, and diabetes in a large population-based cohort without evidence of thyroid disease.Design:Record-linkage technology was used retrospectively to identify people without evidence of thyroid disease in the general population of Tayside, Scotland, from July 1, 2003, to December 31, 2009.Cohort:All Tayside residents who had thyroid function tests performed were identified. Using a unique patient identifier, data linkage enabled a cohort without thyroid disease to be identified by excluding anyone with thyroid or antithyroid prescription, thyroid-related admission or surgery, treatment with radioactive iodine and/or positive thyroid antibodies. Cases below 18 years of age were also excluded.Outcome Measures:We measured TSH distribution among different age groups and by gender.Results:We identified the latest TSH measurements in 153127 people from the reference population after applying the exclusion criteria. There was a significant increase in median TSH (1.58 mU/L at 31-40 y to 1.86 mU/L at >90 y; P <.001) and 97.5th centile TSH (3.98 to 5.94 mU/L, respectively) with increasing age. The 2.5th centile decreased with age (0.51 to 0.31 mU/L). Patients with diabetes had marginally higher TSH concentration (1.80 vs 1.70 mU/L; P <.001).Conclusion:The use of these age-specific reference intervals for TSH, especially in those over 70 years old, would result in the reclassification of many TSH results from "abnormal" to "normal" (within the 95th centile reference interval) and avoid unnecessary treatment.