Background: Measurement of late-night salivary cortisol concentrations is increasingly used as a screening test in suspected Cushing's syndrome. Cortisol concentrations are typically extremely low in latenight samples and discordant assay-specific reference ranges have been reported. Therefore, the aim of our study was to assess the analytical performance of the first automated cortisol immunoassay specified for salivary measurements and to establish late-night sampling reference-range data for this test. Methods: Salivary cortisol was measured using the Roche Cobas Cortisol assay (Roche Diagnostics). Five salivary pools in different concentration ranges were used to assess the inter-assay imprecision of this test in a two-centre evaluation protocol including two reagent lots. Linearity was tested by serial dilution. Salivary samples were obtained at 23:00 h from 100 apparently healthy volunteers using a commercially available salivary sampling device (Salivette, Sarstedt). A subset of 20 samples was used for method comparison with isotope dilution liquid chromatography-tandem mass spectrometry. Results: Inter-assay coefficients of variation (ns20) between 11.6% and 40.4% were found for mean cortisol concentrations between 12.9 and 2.6 nmol/L, with an estimated functional sensitivity of approximately 5.0 nmol/L. The test also gave linear results in the lowest concentration range between 1.0 and 8.3 nmol/L. Mean late-night salivary cortisol of 5.0 nmol/L was found for healthy individuals; the absolute range was 1.4-16.7 nmol/L, and the 95th percentile was 8.9 nmol/L. Substantially lower concentrations were found with isotope dilution LC-MS/MS compared to immunoassay results (mean concentrations 1.8 and 4.4 nmol/L, respectively). Conclusions: The automated assay investigated was found to offer acceptable analytical performance in the very low concentration range required for latenight salivary cortisol, despite a very short turnaround time. Using this assay, late-night salivary