First, to evaluate the performance level of a new TSH binding inhibitory antibody assay using human recombinant TSH receptors (h-TBII) in comparison with a thyroid stimulating antibody (TSAb) bioassay performed before, at the end of treatment (18 months) and after antithyroid drug withdrawal in Graves' disease patients; second, to assess the accuracy with which h-TBII levels could predict relapse and remission. Retrospective study on serum samples of Graves' disease patients treated by antithyroid drugs for 18 months. Serum samples from 140 patients (27 men and 113 women; median age 42 years) with recent onset hyperthyroidism due to Graves' disease were retrospectively tested for h-TBII at diagnosis, at 18 months and for 76 of them 6, 12, 24 and 36 months after drug withdrawal or at relapse. TSAb were also evaluated at each time. Thyroid blocking antibodies (TBAb) were measured in sera positive for h-TBII and negative for TSAb. h-TBII levels were measured with a radioreceptor assay using the human recombinant TSH receptor (DYNOtest TRAK human from B.R.A.H.M.S. Diagnostica, Berlin, Germany). TSAb and TBAb levels were assayed in thyrocyte cultures. At diagnosis, high levels of h-TBII were found in 138 of 140 patients with Graves' disease (98.6%). High TSAb values were also detected in the same 138 patients. The h-TBII and TSAb values were significantly correlated (r = 0.582, p < 0.0001). At 18 months, h-TBII were found in 48 of the 140 patients (34%) and TSAb in 43 patients (31%). The h-TBII and TSAb values were significantly correlated (r = 0.618, p < 0.0001) and fell significantly between diagnosis and 18 months (Z = - 9.84, p < 0.0001 and Z = - 9.19, p < 0.0001). TBAb were found in two of the 11 sera positive for h-TBII and negative for TSAb. At diagnosis, the median levels of h-TBII and TSAb were not significantly different between the patients who relapsed within the 3 years after the withdrawal of treatment (n = 60) and those who did not (n = 80) (Z = - 1.1, ns; Z = - 0.216, ns). At 18 months, the prevalences of h-TBII and TSAb were significantly increased in patients who relapsed compared to those who remained in remission at 3 years (60% vs. 15%, chi2 = 30.8, p < 0.0001 and 55% vs. 13%, chi2 = 29.1, p < 0.0001, respectively). h-TBII and TSAb median levels were also significantly increased in patients who relapsed compared to the others (Z = - 4,8, p < 0.0001; Z = - 3,01, P < 0.005). Among the 60 patients who relapsed, 36 (60%) displayed h-TBII and 33 (55%) TSAb at 18 months. The majority of patients who relapsed during the 2 years following the end of treatment, in contrast to those who relapsed later, were positive for h-TBII (67% vs. 27%, chi2 = 6.01, p = 0.035). Seventeen of the 18 tested patients who relapsed were negative for h-TBII at 18 months, were then positive for h-TBII at the time of relapse, whereas 15 of them were still negative, 6-12 months before the relapse. Among the 80 patients who remained in remission at 3 years, only 10 (13%) displayed TSAb and 12 (15%) h-TBII at 18 months. In 10 of these 12 patients who were further evaluated for h-TBII positivity, h-TBII fell to control levels during the 3 years following the end of treatment. The new h-TBII assay is a simple and rapid method with a performance level similar to that of TSAb determination. Its sensitivity is close to 100% at diagnosis. In the whole group, TBII level analysis is relevant as a predictor for short time relapse. However, some of the patients are "misclassified", due to the inter-individual variability in the time course of h-TBII activity. Our results confirm that, in addition to h-TBII, even when measured with a highly sensitive test, the concomitant analysis of other clinical and/or biological parameters is necessary to improve the prediction of treatment outcome.