Twenty-seven patients with Cushing's disease underwent transsphenoidal pituitary surgery. After the operation 16 patients were cured, and 11 remained hypercortisolemic. In the cured patients a significantly lower incidence of paradoxical responsiveness to TRH or to LRH was found preoperatively (1 of 16) than in the failures (6 of 11; P less than 0.02, by Fisher's chi 2 test). Furthermore, responsiveness of cortisol to CRH administration was significantly lower in the failures [maximum, 0.30 +/- 0.10 mumol/L (10.8 +/- 3.6 micrograms/dL) vs. 0.50 +/- 0.07 mumol/L (18.1 +/- 2.5 micrograms/dL) in the cured patients; P less than 0.05]. There were no differences in basal plasma cortisol levels, ACTH levels, or suppressibility by dexamethasone between the 2 groups. In the 16 patients who were cured initially, 4 patients relapsed after a mean period of 4 yr. These 4 patients had significantly higher basal cortisol levels postoperatively than those who remained in remission [0.14 +/- 0.03 mumol/L (5.1 +/- 0.8 micrograms/dL) vs. 0.04 +/- 0.01 mumol/L (1.4 +/- 0.3 micrograms/dL); P less than 0.01]. Cortisol responses to CRH after the operation positively correlated with the basal cortisol levels at that time (P less than 0.05, by Spearman's rank correlation test; r = 0.64), thus, the relapsing patients had higher cortisol responses to CRH than patients who stayed in remission [maximum response, 0.31 +/- 0.07 (11.2 +/- 2.5 micrograms/dL) vs. 0.12 +/- 0.03 mumol/L (4.3 +/- 1.1 microgram/dL), respectively; P less than 0.05]. We conclude that 1) patients responding paradoxically to TRH and/or LRH have a lower chance of being cured after pituitary surgery; and 2) patients with higher cortisol levels (greater than 0.10 mumol/L; 3.6 micrograms/dL) after being cured initially have a higher chance of recurrence of their disease.