We studied 18 patients with Cushing's syndrome and 25 patients in whom Cushing's syndrome was excluded on follow-up to evaluate screening tests for Cushing's syndrome in hospitalized patients. Plasma cortisol values (at 8 AM) were found least helpful, yielding 29% false-positive and 60% false-negative values. Diurnal variation of cortisol was present in 30% of patients with Cushing's syndrome and absent in 18% of patients without Cushing's syndrome. When corrected for total urinary creatinine, 24-hour urinary 17-hydroxycorticosteroids were specific (all patients without Cushing's syndrome had normal values) but not very sensitive (2 of 12 patients with Cushing's syndrome had normal values). Similarly, 24-hour 17-ketosteroids were of little help with 17% false-positive and 35% false-negative values. Twenty-four-hour urinary free cortisol was both a sensitive and specific screening test for Cushing's syndrome (no false-positive and no false-negative results). We conclude that urinary free cortisol is the most efficient screening method for Cushing's syndrome in hospitalized patients.