Résumé : Five different ultrasensitive thyrotropin (TSH) assay kits (Boots-Celltech, Immunotech, ORIS-CIS, Travenol and Boehringer) have been used for TSH measurements in various conditions. All the kits were based on an immunometric method but differed with regard to components and procedure. The sensitivity appeared essentially the same for the five kits (0.10 microU/ml) as well as the intraassay precision (coefficient of variation less than 12%). In contrast, the interassay coefficients of variation in the low TSH range varied from 12.8 to 21.3%. Discrepancies from kit to kit were observed and accounted for by differences in the components and procedure of the kits. Basal serum TSH was determined in normal subjects (n = 261) and in patients with thyroid dysfunction (n = 392). No overlap was shown between normals and patients with overt hypothyroidism. In contrast, an overlap existed between normals and hyperthyroids for all the kits but one. Measurements in patients with nontoxic goiter showed that TSH may be undetectable in clinically euthyroid patients, whatever the kit used. After TRH stimulation, 95% of the 375 patients tested associated either an absence of response to TRH with undetectable basal TSH values, or a blunted response with low basal TSH levels or normal response with normal basal TSH concentrations. However, 9 patients with suppressed TSH showed a response to TRH and 7 patients with normal basal TSH levels presented an exaggerated response to TRH. Taken together, these results demonstrate that even though ultrasensitive measurements of TSH do not meet the expectation of completely discriminating euthyroid from hyperthyroid patients, ultrasensitive TSH assay kits represent a powerful tool in the diagnosis of thyroid dysfunction, which would eliminate, in most instances, the need for TRH test and diminish thyroid hormone assay requests.