OBJECTIVE Stimulation with corticotropin-releasing hormone (CRH) is one of principal tools for the differential diagnosis of ACTH-dependent Cushing's syndrome. However, different dosages and species of CRH may be employed; further, ACTH levels can be measured by radioimmunoassay (RIA) or immunoradiometric assay (IRMA). The aims of the present study were to perform a reappraisal of the diagnostic accuracy of the CRH test taking these different testing modalities into consideration and to study the correlation between basal ACTH and cortisol levels and their responses to CRH in patients with Cushing's disease. METHODS The study population comprised 148 patients with Cushing's disease and 12 patients with ectopic ACTH secretion collected through an Italian multicentre study. METHODS Patients were submitted to stimulation with 100 microg human or ovine CRH (36% and 64% of subjects, respectively) and ACTH measured either by RIA or IRMA (28% and 72%, respectively). A 50% increase in ACTH and cortisol levels was considered indicative of Cushing's disease. RESULTS Mean peak ACTH levels measured by RIA and IRMA were comparable, as was the diagnostic accuracy of the test with the two assays (87% for IRMA and 84% for RIA, ns). In patients with Cushing's disease, stimulation with ovine CRH induced greater hormonal responses compared to testing with human CRH although only the cortisol response reached statistical significance (ACTH: 247.5 +/- 28.0% vs. 168.5 +/- 21.3% over baseline, P = 0.06; cortisol: 89.3 +/- 8.5% vs. 60.8 +/- 9.6% over baseline, P < 0.05 for ovine and human CRH, respectively). No appreciable rise in ACTH and cortisol levels was registered among patients with ectopic ACTH secretion. Diagnostic accuracy of the cortisol response was significantly greater with the ovine than with human peptide (71% vs. 49%, P < 0.05, for ovine and human CRH, respectively) while the ACTH response yielded equal diagnostic accuracy (86% vs. 87%, ns, for the ovine and human peptide, respectively). Interestingly, the correlation between ACTH and cortisol peak responses in patients with Cushing's disease was significantly greater for human than for ovine CRH (r = 0.68 vs. r = 0.41, P < 0.01, respectively). In addition, baseline cortisol levels exhibited a significant negative correlation with both the ACTH and cortisol response to CRH suggesting the persistence of the negative cortisol feedback in patients with Cushing's disease. CONCLUSIONS (A) Both RIA and IRMA can be used indifferently for the assessment of the ACTH response to CRH. (B) Human and ovine CRH provide the same diagnostic accuracy as regards the ACTH response which, incidentally, represents the most accurate criterion for the evaluation of the CRH test; ovine CRH is superior to the human peptide in the evaluation of the cortisol response. (C) In patients with Cushing's disease, endogenous cortisol maintains the ability to negatively modulate CRH-stimulated corticotropin secretion.