Patients with the hyperthyroidism of Graves' disease (GDH) have a higher risk of relapse after antithyroid drug therapy (ATD) therapy when TSH receptor antibodies (TRAb) are positive, but the practical clinical implication of TRAb as a predictor for relapse is still much debated. This study was undertaken to investigate by meta-analysis the results from the literature on the use of TRAb as predictor of long term (i.e. at least 1 yr) relapse after ATD. Eighteen publications from 1975-1991 fulfilled the criteria of 1) availability of TRAb at the end of ATD treatment, 2) at least 1 yr of follow-up after ATD, 3) data presentation in a form suitable for meta-analysis, and 4) no other thyroid-related therapy during the follow-up period. The 10 prospective studies, 5 of which measured TSH binding inhibiting immunoglobulins (total n = 597) and 5 of which measured thyroid-stimulating antibodies (n = 340), were computed together because no significant differences were found. In contrast, retrospective and prospective studies differed. In the prospective studies, the odds reduction of relapse showed 65% less risk of relapse when TRAb were absent compared to that in TRAb-positive patients (P < 0.00001). The present meta-analysis has, thus, confirmed in a large number of patients (n = 1524) that absence of TRAb is significantly protective against relapse of GDH after ATD treatment. However, 25% of the patients are "misclassified," and the main questions arising from the study are, therefore, the following. 1) Is it worthwhile to use TRAb as predictor of relapse? 2) Should patients with GDH continue ATD until TRAb becomes negative, rather than for a fixed period? The available methods for TRAb do not allow sufficiently high prediction of relapse or remission after ATD for the individual patient.