The clinical features of Cushing's syndrome (such as obesity, hypertension, and diabetes) are commonly encountered in clinical practice. Patients with Cushing's syndrome have been identified by an abnormal low-dose dexamethasone suppression test, elevated urine free cortisol (UFC), an absence of diurnal rhythm of plasma cortisol, or an elevated late-night plasma cortisol. Because the concentration of cortisol in the saliva is in equilibrium with the free (active) cortisol in the plasma, measurement of salivary cortisol in the evening (nadir) and morning (peak) may be a simple and convenient screening test for Cushing's syndrome. The purpose of this study was to evaluate the usefulness of the measurement of late-night and morning salivary cortisol in the diagnosis of Cushing's syndrome. We studied 73 normal subjects and 78 patients referred for the diagnosis of Cushing's syndrome. Salivary cortisol was measured at 2300 h and 0700 h using a simple, commercially-available saliva collection device and a modification of a standard cortisol RIA. In addition, 24-h UFC was measured within 1 month of saliva sampling. Patients with proven Cushing's syndrome (N = 39) had significantly elevated 2300-h salivary cortisol (24.0 +/- 4.5 nmol/L), as compared with normal subjects (1.2 +/- 0.1 nmol/L) or with patients referred with the clinical features of hypercortisolism in whom the diagnosis was excluded or not firmly established (1.6 +/- 0.2 nmol/L; N = 39). Three of 39 patients with proven Cushing's had 2300-h salivary cortisol less than the calculated upper limit of the reference range (3.6 nmol/L), yielding a sensitivity of 92%; one of these 3 patients had intermittent hypercortisolism, and one had an abnormal diurnal rhythm (salivary cortisol 0700-h to 2300-h ratio <2). An elevated 2300-h salivary cortisol and/or an elevated UFC identified all 39 patients with proven Cushing's syndrome (100% sensitivity). Salivary cortisol measured at 0700 h demonstrated significant overlap between groups, even though it was significantly elevated in patients with proven Cushing's syndrome (23.0 +/- 4.2 nmol/L), as compared with normal subjects (14.5 +/- 0.8 nmol/L) or with patients in whom Cushing's was excluded or not firmly established (15.3 +/- 1.5 nmol/L). Late-night salivary cortisol measurement is a simple and reliable screening test for spontaneous Cushing's syndrome. In addition, late-night salivary cortisol measurements may simplify the evaluation of suspected intermittent hypercortisolism, and they may facilitate the screening of large high-risk populations (e.g. patients with diabetes mellitus).