The aim of the study was to evaluate the diagnostic value of the human CRH test and the basal morning serum cortisol for the diagnosis of adrenal insufficiency. Putative peak cortisol cut points for the CRH test and basal cortisol cut points were determined by receiver operating characteristic (ROC) analysis with the insulin tolerance test as reference test. Fifty-four patients with suspected hypothalamic-pituitary-adrenal disease were tested. In 20 healthy controls, CRH led to a mean peak cortisol of 594.8 +/- 21.7 nmol/liter. The lower limit of a normal response was calculated as 400 nmol/liter. ROC analysis of peak cortisol levels during CRH testing of patients with suspected hypothalamic-pituitary-adrenal disease suggested an optimal peak cortisol cut point of < or 377 nmol/liter for the diagnosis of adrenal insufficiency and a 96% specificity but poor sensitivity of 76%. The baseline cortisol in the healthy control group showed a mean of 439.3 +/- 24.9 nmol/liter, resulting in a lower limit of 267 nmol/liter. ROC analysis of patients suggested the highest accuracy for basal cortisol levels of 285 nmol/liter or more for the diagnosis of adrenal insufficiency (100% sensitivity and 61% specificity). Within this patient group, a cortisol of more than 98 nmol/liter excluded adrenal insufficiency among those without the disorder, yielding 100% specificity. Using these criteria of upper (285 nmol/liter) and lower (98 nmol/liter) cut-off points with high sensitivity and specificity can reduce the number of individuals who need provocative tests. Basal cortisol is less expensive, and we therefore suggest to use it as a first-line test of adrenal insufficiency. Because of the low sensitivity of the human CRH test, we do not recommend it as a second test.