AIMS AND SUBJECTS: We investigated the prospective relationship between thyroid status and the risk of dementia and Alzheimer's disease among a random sample of 1843 participants, aged 55 years and over, from the population-based prospective Rotterdam Study. METHODS: Thyroid status was measured at baseline (1990-93), through assessment of serum antibodies to thyroid peroxidase (TPO-Abs, positive: >10 IU/ml), serum TSH levels, and when TSH was abnormal (<0.4 or >4.0 mU/l), serum thyroxin levels (T4). At baseline and at follow up, after on average 2 years, participants were screened for dementia. Diagnoses were based on international criteria. RESULTS: Persons with reduced TSH levels at baseline had a more than threefold increased risk of dementia (RR = 3.5, 95%CI: 1.2-10.0) and of Alzheimer's disease (RR = 3.5, 95%CI: 1.1-11.5), after adjustment for age and sex. Among persons with reduced TSH levels, T4 levels appeared to be positively related to the risk of dementia (RR per SD increase = 2.9, 95%CI: 0.7-12.2), although none of those who became demented had a T4 level above the normal range (>140 nmol/l). The risk of dementia was especially increased in subjects with low TSH who were positive for TPO-Abs (RR = 23.7, 95%CI: 4.0-140). CONCLUSIONS: This is the first prospective study to suggest that subclinical hyperthyroidism in the elderly increases the risk of dementia and Alzheimer's disease.