To ascertain the strength of the association between thyroid autoimmunity and miscarriage, we performed a meta-analysis of both case-control and longitudinal studies performed since 1990 when this association was first described. A clear association between the presence of thyroid antibodies and miscarriage was found with an odds ratio (OR) of 2.73 (95 % confidence interval (CI), 2.20-3.40) in eight case-control and ten longitudinal (OR, 2.30; 95 % CI, 1.80-2.95) studies. This association may be explained by a heightened autoimmune state affecting the fetal allograft, of which thyroid antibodies are just a marker. Alternatively, the association can be partly explained by the slightly higher age of women with antibodies compared with those without (mean+/-S.D. age difference, 0.7+/-1.0 years; P<0.001). A third possibility is mild thyroid failure, as thyroid-stimulating hormone (TSH) levels in antibody-positive but euthyroid women are higher than in antibody-negative women: difference 0.81+/-0.58 mU/l (P=0.005). Randomized clinical trials with l-thyroxine (aiming at TSH values between 0.4 and 2.0 mU/l) and with selenium (to decrease antibodies against thyroid peroxidase) are clearly needed to elucidate further the nature of this association.