A high prevalence of antibodies to double-stranded DNA (AbDNAds) has been recently reported in serum of patients with autoimmune thyroid disorders, but the specificity of this finding has been questioned. For this reason, the prevalence of several antibodies to DNA-related nuclear antigens (AbDRENA) has been evaluated in sera of patients with autoimmune and non-autoimmune thyroid disease. The study group included: 46 Graves’disease patients, 28 Hashimoto's thyroiditis patients, 25 patients with toxic nodular goitre and 11 with non-toxic nodular goitre. Twenty-eight Graves' patients were retested during methimazole (MMI) therapy, and 5 after radioiodine administration. Twenty-two patients with systemic lupus erythematosus and 28 normal subjects served as positive and negative controls, respectively. AbDRENA included: AbDNAds by RIA or immunofluorescence (IF); antibodies to single-stranded DNA (AbDNAss) and antibodies to histone (AbHist) by ELISA methods; antibodies to nuclear antigens (ANA) by immunofluorescence. RIA values were considered to be abnormal when 2 SD above the mean of normal controls. In our study 13% of Graves' patients were positive for AbDNAds by RIA: all of them had negative tests by IF; 11% were positive for AbDNAss. 2% for AbHist and 7% for ANA. A comparable prevalence of positive results for AbDNAds by RIA. with negative IF tests, was found in Hashimoto's thyroiditis patients. No significant changes of antibody levels were observed in Graves' patients during MMI treatment or after radioiodine administration. A positivity for AbDNAds or AbDNAss was found in 8% of patients with toxic nodular goitre, but in none of those with non-toxic goitre. All patients with systemic lupus erythematosus had positive tests for AbDNAds by both RIA and IF.In conclusion, the results of the present study indicate that: (i) the prevalence of anti-DNA antibodies is low and does not differ in the different groups of thyroid patients, irrespective of the presence of thyroid auto-immunity and/or thyroid hyperfunction; and (ii) different techniques (e.g. RIA vs. IF) may provide discrepant results, thus at least partially explaining differences among various series.