We retrospectively reviewed dynamic ACTH and cortisol responses to insulin hypoglycemia in 193 subjects with suspected ACTH deficiency to ascertain the predictive values of various diagnostic criteria. Based on the achievement of a peak cortisol level of 18 micrograms/dL or above, 133 subjects were classified as having an intact hypothalamic-pituitary-adrenal (HPA) axis, and 60 subjects were determined to have ACTH deficiency. Baseline and peak cortisol concentrations were strongly correlated (r = 0.63; P < 0.0001). Peak cortisol increased in parallel to ACTH increments, but plateaued at approximately 22 micrograms/dL at peak ACTH levels above approximately 75 pg/mL (r = 0.61; P < 0.0001). Basal cortisol values above 17 micrograms/dL or below 4 micrograms/dL were highly predictive of an intact or impaired HPA axis, respectively, but intermediate values had only limited sensitivity and specificity. The criteria of HPA axis integrity, defined as an increment in plasma cortisol of more than 7 micrograms/dL above the baseline or as a doubling of the baseline cortisol value, were associated with high false positive and false negative rates. We conclude that 1) the baseline morning serum cortisol concentration has very limited predictive power in differentiating between normal and impaired HPA function; 2) the use of criteria based on incremental changes in serum cortisol from baseline leads to unacceptably high false positive and false negative rates; and 3) insulin hypoglycemia is still the best indicator of the integrity of the response of the HPA axis to stress.