BACKGROUND The low dose dexamethasone suppression test (DST) has been used to detect subtle variations in the feedback suppression of the hypothalamic-pituitary-adrenal axis, which may contribute to the pathogenesis of several diseases including depression, the metabolic syndrome and coronary artery disease. Little is known about the reproducibility of this test, or whether the test can be combined with analysis of salivary cortisol which would offer a significant advantage over plasma in population studies. METHODS A low dose DST was carried out in 29 healthy subjects (14 men, 15 women), aged 24-54 (mean 35.1) years, on two separate occasions 1-10 weeks apart. Following the administration of 0.25 mg dexamethasone (DXM) at 2200 h, plasma and saliva were sampled at 0830 h the next day. Cortisol was measured by radioimmunoassay in plasma and time-resolved immunofluorescent assay ('DELFIA') in saliva. Bland-Altman plots were produced for post-DXM plasma and salivary cortisol measures and used to derive a coefficient of repeatability for each measure, which describes the range of cortisol measurements within which 95% of repeated measurements will fall. RESULTS The baseline, pre-DXM cortisol concentrations were far more variable for saliva (mean 16.5, range 4.4-34 nmol/l) than for plasma (mean 407.5; range 232-958 nmol/l). Following DXM both measurements showed an approximately 30% suppression from baseline but the variability of salivary cortisol was much greater. From the Bland-Altman plots the 95% range for the differences about their mean was calculated and used as an indication of repeatability. For plasma 95% of differences were within 0.78 log units, indicating that a repeated measurement was approximately half as small or twice as large as the first. For saliva 95% of differences were within 1.64, indicating that a repeated measurement was approximately five times as small or five times as large as the first. CONCLUSIONS Assessment of dexamethasone suppression by salivary cortisol measurement is far less repeatable than the use of plasma cortisol. In the context of field studies of dexamethasone suppression, salivary cortisol measurements may only be appropriate for large numbers of subjects.