Salivary cortisol on ROCHE Elecsys immunoassay system: pilot biological variation studies

Salivary analysis is a noninvasive alternative that may be more acceptable to patients, especially children. As such, it has the potential for incorporation into comprehensive, dynamic investigations of metabolic dysfunctions - a significant advancement over a single time point serum analysis. In this study, we wanted to determine if the serum cortisol assay on our routine immunoassay analyzer could reliably measure salivary cortisol concentrations. Because of potential fluctuations in salivary concentrations, we included a biologic variation study as a main facet of our preliminary method evaluation.

Twenty-eight healthy individuals (12 males, 16 females) volunteered to provide 5 nonconsecutive first morning saliva samples over a two-week period. Samples were stored frozen at home until the completion of the study. Following thawing and centrifugation, cortisol was measured in batch mode for each set of participant samples on the ROCHE Elecsys. Biologic variation was determined following removal of outliers. A method comparison was performed with the DPC Coat-A-Count Cortisol assay following the recommended modifications for salivary analysis, and with the Salimetrics HS-Cortisol two-site monoclonal assay optimized for salivary cortisol.

Mean salivary cortisol concentration was 20.4 nmol/L. Analytical variation (CV(A) = 3.8%), within-subject variation (CV(I) = 6.3%), between-subject variation (CV(G) = 20.5%), index of individuality (II = 0.36) and reference change value (RCV = 20.4%) were determined. A negative 40% proportional bias was observed on the Elecsys compared to the two methods that have already been optimized for salivary analysis.

This study demonstrated that salivary cortisol can be reliably measured on a routine automated immunoassay analyzer such as the ROCHE Elecsys. This particular assay needs to be optimized at the low end of the standard curve for routine use with salivary samples. Based on the relatively small intra-individual variation and low index of individuality, reference change values are preferable to the use of population reference intervals for this assay.