CONICET | Buscador de Institutos y Recursos Humanos

Saliva is a biological fluid easy to obtain with non-invasiveprocedures in a stress-free way compared with the serumcollection for the determination of hormone levels [1].It has been suggested that measurement of salivary cortisollevels is more appropriate for the clinical assessmentof adrenocortical function. In fact, determination oflate-night salivary cortisol concentration is recommendedfor the screening of patients with suspected Cushing?ssyndrome (CS) [2]. Automated systems have a rapidturn-around time on a large number of samples andhave demonstrated high analytical accuracy. Given thatsalivary cortisol levels are a valuable indicator of thehypothalamic-pituitary-adrenocortical axis activity [3], itsdetermination by automated systems may allow a broaderuse of this diagnostic tool [4, 5].The aim of this study was to validate an automatedchemiluminescent immunoassay (CLIA), Siemens Immulite2000® analyzer (Gwynedd, UK), for the measurementof salivary cortisol and to compare it with the electrochemiluminescentimmunoassay (ECLIA), Roche Cobase-411 (Mannheim, Germany), which has been validatedby the manufacturer for the measurement of salivary cortisol.To validate the assay we evaluated limit of blank(LOB), limit of detection (LOD) and limit of quantification(LOQ), precision, linearity, recovery and we performed themethod comparison. We have also determined the cut-offlevel for salivary cortisol by CLIA in CS diagnosis in anadult population.To evaluate the analytical performance of CLIA andcompare it with ECLIA, 50 saliva samples were collectedfrom healthy volunteers at 8 am (27 males and 23 females;body mass index, 25.5 ± 1.5 kg/m2). Analysis of salivary cortisolcut-off values were performed on late-night salivarysamples (11 pm) obtained from 38 healthy subjects (24males and 14 females; body mass index, 24.9 ± 1.9 kg/m2,subjects who presented serious underlying medical conditionswere excluded from the study) and from 21 newlydiagnosed CS patients (6 males and 15 females; body massindex, 28.7 ± 2.2 kg/m2) without medication. All sampleswere collected through passive drool directly into sterileplastic tubes of single use. Saliva donors did not brushtheir teeth, smoke, eat, or drink anything except for waterfor 2 h before testing. Volunteers and patients with freshcuts or abrasions within the oral cavity were excluded.Morning samples were immediately stored at ?20 °C untilanalysis and late-night samples were first preserved at 4 °Cand immediately stored at ?20 °C the next morning. Oncethawed, samples were centrifuged at 1500 g for 10 min atroom temperature and supernatants were used for cortisoldeterminations. Volume samples required for cortisoldeterminations by CLIA and ECLIA were 10 μL and 20 μL,per reaction, respectively. This study was approved by theEthics Committee at the Hospital de Clínicas ?General Joséde San Martín? (according to the Helsinki Declaration formedical studies).Salivary cortisol levels were measured with CLIAmethod with minor modifications. We performed a newcalibration curve with dilutions of cortisol calibrator solutionsin a range of 2.0?100 nmol/L using CLIA?s kit diluentrecommended by the manufacturer. Calibration solutionswere obtained from RIA kit, Immunotech® BeckmanCoulter (catalog number: IM1841), which were calibratedagainst the reference preparation ERM®-DA192 and 193.Logit-log calibration curves were constructed with valuesfrom calibrators in counts per second (cps) and samplevalues were interpolated into this straight line to determinecortisol concentrations in nmol/L.To validate CLIA for salivary cortisol, LOB, LOD, andLOQ were assessed according to the EP-17A protocol [6].LOB was 0.9 nmol/L, established as the 95th percentilefrom repeated non-parametric measurements (60 times)of a blank sample (CLIA?s kit diluent). No significant differenceswere found in the calibration curve when CLIA?sdiluent or saliva were used. The slopes for the assay withCLIA?s diluent and saliva were ?1.6674 (95% CI, ?2.3728 to?0.9620; y = 1.994 ? 1.6674x; r = ?0.990) and ?1.1177 (95%CI, ?1.5498 to ?0.6855; y = 1.794 ? 1.1177x; r = ?0.992),respectively. Data distribution used for LOD calculationwas nonparametric (Shapiro-Wilks p < 0.05), and no significantdifferences were found between variances (F test,p < 0.05). The obtained LOD was 2.0 nmol/L, which wasestablished from repeated measurements of three samplepools of saliva with concentrations between LOB and fourtimes LOB. The LOQ value was 3.4 nmol/L, determinedfrom a precision profile using four saliva samples of differentconcentrations (Figure 1). In the precision assay,the coefficient of variation intra-assay for concentrationsof 11.5 and 20 nmol/L were 8.9% and 6.0%, respectively,and the coefficients of variation inter-assay for the sameconcentrations were 9.9% and 9.4%, respectively.To evaluate linearity in the salivary matrix, serial dilutionsof a cortisol standard solution were performed usingsaliva as diluent (cortisol concentration in the matrix waslower than 2 nmol/L). This assay was linear between 2.0and 100 nmol/L.In the recovery test, three saliva samples, with aknown cortisol concentration, were supplemented with 9,21 or 47 nmol/L of cortisol and the mean analytical recoverywere 107.2%, 103.5% and 97.8%, respectively. Eachdetermination was performed in triplicates.To evaluate method comparison, salivary cortisolfrom 50 volunteers was measured by CLIA and ECLIA.Salivary cortisol concentrations determined by CLIA(median, 6.0 nmol/L; range, 6.0?9.0 nmol/L) were significantlylower compared to ECLIA values (median,8.5 nmol/L; range, 8.5?10.7 nmol/L, p < 0.0001 Mann-Whitney test). The correlation between both methods wassignificant (Spearman?s rank correlation coefficient: 0.656;p < 0.0001). A Bland-Altman plot was performed to evaluatethe concordance of the two immunoassays; the biasbetween both methods (CLIA minus ECLIA values) was?1.9 nmol/L. Although the correlation was not optimal, anappropriate concordance was achieved and the disparitiescould be attributed to differences in the antibodies used ineach immunoassay.In this study, to optimize the diagnostic performanceof the test, CLSI international standards were applied todetermine the LOQ before establishing the cut-off value.This parameter was evaluated using a receiver operatorcharacteristic (ROC) analysis curve (MedCalc StatisticalSoftware v12.7.7, Ostend, Belgium) performed with 59late-night salivary samples. CLIA showed a good diagnosticperformance (area under the curve, 0.990, p < 0.0001,Figure 2A). In addition, the best cut-off value for CLIAobtained from the ROC curve was 5.7 nmol/L, with anoptimal sensitivity and specificity of 90.5% and 97.4%,respectively (Figure 2B).The International Endocrine Society recommendslate-night salivary cortisol determinations as a first linescreening test for CS [2]. Several studies have shownthat it is necessary to establish cut-off values for salivarycortisol measurement in CS diagnosis, according tothe method used [7]. The cut-off point (5.7 nmol/L) wasnot so different to those reported by others in previousstudies with immunoassays [8]. However, the diversityin the cut-off values could be attributed to differences inthe methods used and the population studied, as well asthe source of variability that affect the determination ofcortisol [8, 9].In summary, this study shows the validation of achemiluminescent automated method for the measurementof salivary cortisol that presents convenientanalytical performance in values close to cut-off level.Considering the emerging potential roles of salivary cortisolmeasurement (stress biomarker, congenital adrenalhyperplasia or adrenal insufficiency biomarker) [2], themethod described here can be easily used in clinicallaboratories for diagnostic purposes since it is relativelysimple to perform very sensitive, and has a satisfactoryturn-around time. Moreover, this method offers the possibilityto analyze samples as they are received and helpclinicians to make more timely clinical decisions.