Patients with serum thyroid-stimulating hormone (TSH) levels outside the reference range and levels of free thyroxine (FT4) and triiodothyronine (T3) within the reference range are common in clinical practice. The necessity for further evaluation, possible treatment, and the urgency of treatment have not been clearly established. To define subclinical thyroid disease, review its epidemiology, recommend an appropriate evaluation, explore the risks and benefits of treatment and consequences of nontreatment, and determine whether population-based screening is warranted. MEDLINE, EMBASE, Biosis, the Agency for Healthcare Research and Quality, National Guideline Clearing House, the Cochrane Database of Systematic Reviews and Controlled Trials Register, and several National Health Services (UK) databases were searched for articles on subclinical thyroid disease published between 1995 and 2002. Articles published before 1995 were recommended by expert consultants. A total of 195 English-language or translated papers were reviewed. Editorials, individual case studies, studies enrolling fewer than 10 patients, and nonsystematic reviews were excluded. Information related to authorship, year of publication, number of subjects, study design, and results were extracted and formed the basis for an evidence report, consisting of tables and summaries of each subject area. The strength of the evidence that untreated subclinical thyroid disease is associated with clinical symptoms and adverse clinical outcomes was assessed and recommendations for clinical practice developed. Data relating the progression of subclinical to overt hypothyroidism were rated as good, but data relating treatment to prevention of progression were inadequate to determine a treatment benefit. Data relating a serum TSH level higher than 10 mIU/L to elevations in serum cholesterol were rated as fair but data relating to benefits of treatment were rated as insufficient. All other associations of symptoms and benefit of treatment were rated as insufficient or absent. Data relating a serum TSH concentration lower than 0.1 mIU/L to the presence of atrial fibrillation and progression to overt hyperthyroidism were rated as good, but no data supported treatment to prevent these outcomes. Data relating restoration of the TSH level to within the reference range with improvements in bone mineral density were rated as fair. Data addressing all other associations of subclinical hyperthyroid disease and adverse clinical outcomes or treatment benefits were rated as insufficient or absent. Subclinical hypothyroid disease in pregnancy is a special case and aggressive case finding and treatment in pregnant women can be justified. Data supporting associations of subclinical thyroid disease with symptoms or adverse clinical outcomes or benefits of treatment are few. The consequences of subclinical thyroid disease (serum TSH 0.1-0.45 mIU/L or 4.5-10.0 mIU/L) are minimal and we recommend against routine treatment of patients with TSH levels in these ranges. There is insufficient evidence to support population-based screening. Aggressive case finding is appropriate in pregnant women, women older than 60 years, and others at high risk for thyroid dysfunction.