| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Reproductive Biology and Medicine Branch (S.K.B., M.W., L.K.N.), National Institute of Child Health and Human Development, and Biostatistics and Clinical Epidemiology Service (N.S.), Clinical Center, National Institutes of Health, Bethesda, Maryland 20892; and the George Washington University Weight Management Program (D.R.), Washington, D.C., 20037
Address all correspondence and requests for reprints to: Lynnette Nieman, M.D., Building 10, CRC, 1 East, Room 1-3140, 10 Center Drive, MSC 1109, Bethesda, Maryland 20892-1109. E-mail: NiemanL{at}nih.gov.
 |
Abstract |
|---|
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Objective: Our objective was to compare RIA and tandem mass spectrometry (LC-MS/MS) measurement of salivary cortisol in obese subjects and healthy volunteers.
Design and Setting: We conducted a cross-sectional prospective study of outpatients.
Subjects and Methods: We studied 261 obese subjects (186 female) with at least two additional features of Cushing’s syndrome and 60 healthy volunteers (30 female). Subjects provided split bedtime salivary samples for cortisol measurement by commercially available RIA and LC-MS/MS. Results were considered normal or abnormal based on the laboratory reference range. Subjects with abnormal results underwent evaluation for Cushing’s syndrome.
Results: In paired samples, RIA gave a lower specificity than LC-MS/MS in obese subjects (86 vs. 94%, P = 0.008) but not healthy volunteers (86 vs. 82%, P = 0.71). Among subjects with at least one abnormal result, both values were abnormal in 44% (confidence interval 26–62%) of obese and 75% (confidence interval 33–96%) of healthy volunteers. In obese subjects, salivary cortisol concentrations were less than 4.0 to 643 ng/dl (<0.11–17.7 nmol/liter; normal, 
100 ng/dl, 2.80 nmol/liter) by LC-MS/MS and less than 50 to 2800 ng/dl (1.4–77.3 nmol/liter; normal, 170 ng/dl, 4.7 nmol/liter) by RIA. Cushing’s syndrome was not diagnosed in any subject.
Conclusion: Salivary cortisol levels should not be used as the sole test to diagnose Cushing’s syndrome if laboratory-provided reference ranges are used for diagnostic interpretation.
|
Introduction |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
As the measurement of salivary cortisol moves from research to commercial laboratories, the criteria for interpretation of results are not clear. The reported diagnostic cutoff points vary considerably and were determined using different comparison groups, including healthy volunteers, obese subjects, subjects in whom Cushing’s syndrome was excluded, and/or subjects with pseudo-Cushing’s states (1, 2, 3, 4, 5). By contrast, most commercial laboratory reference ranges are developed using healthy volunteer samples and usually do not account for age, gender, or other medical conditions. A recent study suggests that age and medical conditions influence salivary cortisol levels. Liu et al. (7) evaluated salivary cortisol in 206 male veterans with an enzyme immunoassay whose normal range was based on 73 healthy lean subjects. In that report, 20% of all participants and 40% of diabetic, hypertensive subjects at least 60 yr of age had an abnormal 2300-h salivary cortisol level. At the time of publication, no participant had been diagnosed with Cushing’s syndrome.
Inherent differences between assays also may influence the optimal diagnostic criteria. In one study, two different antibody-based techniques (ELISA and RIA) gave different results on the same sample (8). Similar assays are available commercially, as is tandem mass spectrometry (LC-MS/MS) measurement, which has no steroid cross-reactivity and thus a different normal range.
To define the utility of salivary cortisol as a diagnostic test, it is important to know whether assay techniques or normative thresholds affect the rate of false-positive results. The aim of this study was to compare RIA and LC-MS/MS measurements of salivary cortisol and to analyze the accuracy of reference ranges in healthy volunteers and obese subjects with Cushingoid features.
|
Subjects and Methods |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Obese subjects
From October 2003 to April 2006, individuals presenting for weight loss treatment at the George Washington University Weight Management Program were invited to participate in a study evaluating the performance of screening tests for Cushing’s syndrome. Subjects aged 18–75 yr of age were enrolled if they had at least two signs and symptoms of Cushing’s syndrome in addition to weight gain and would return to clinic for 24 months. Exclusion criteria included 1) weight more than 350 pounds, the limit for radiology examination tables; 2) creatinine more than 2.6 mg/dl; 3) pregnancy; 4) serious medical conditions; and 5) recent or anticipated use of glucocorticoids, black licorice, chewing tobacco, phenytoin, barbiturates, loperamide, or opiates.
Each subject underwent screening tests for Cushing’s syndrome, including a 1-mg overnight dexamethasone suppression test (DST), measurement of 24-h urine cortisol excretion (UFC) by LC-MS/MS (Mayo Laboratories, Rochester, MN) (9, 10), and bedtime salivary cortisol measured by RIA (Esoterix Laboratories, Calabasas Hills, CA). Early in the study, RIA salivary cortisol values were abnormal in several subjects. Thereafter, simultaneously obtained split salivary samples were collected to compare RIA and LC-MS/MS results. The RIA sample was measured immediately; the second salivary samples were batched and sent to Mayo later for LC-MS/MS measurement. Samples were not available for additional study to assess synthetic steroid contamination.
Subjects with any abnormal screening result, UFC more than 45 µg/24 h (124 nmol/24 h), salivary cortisol more than 170 ng/dl (4.7 nmol/liter) (RIA) or more than 100 ng/dl (2.8 nmol/liter) (LC-MS/MS), or DST cortisol 1.8 µg/dl (49.6 nmol/liter) or higher, were questioned about the time of saliva collection and were asked to undergo further evaluation at the National Institutes of Health (NIH) Clinical Center. At that visit, each was asked to submit bedtime salivary samples and two additional urine samples for cortisol measurement as follows, based on the hospital’s reference laboratory; saliva was measured by LC-MS/MS immediately, and the second samples were batched and sent for RIA; UFC was measured by Nichols Advantage Immunochemiluminescence Metric Assay at the Department of Laboratory Medicine, NIH, early in the study or later LC-MS/MS by Mayo. Subjects underwent a dexamethasone-CRH test if their screening DST result was abnormal despite an appropriate serum dexamethasone level and/or if any repeat UFC or salivary cortisol result was abnormal (11). Subjects with multiple follow-up normal test results were considered not to have Cushing’s syndrome.
Healthy volunteers
Healthy adults were invited to participate in a protocol evaluating adrenal function. Recruitment strategies included flyers and local advertisements. Each subject provided a medical history; weight and height were measured, and laboratory testing was performed. Exclusion criteria included severe medical or psychiatric illness, abnormal serum hepatic, renal, or electrolyte biochemistry results, pregnancy, lactation, clinical or historical features of adrenal insufficiency, or recent use of antifungal or glucocorticoid-containing medications. Subjects with well-controlled chronic illness (e.g. hypertension) were eligible to participate.
Volunteers provided split bedtime salivary samples for cortisol measurement by RIA and LC-MS/MS and one 24-h urine sample for cortisol measurement as described above. Those with any abnormal result were asked to repeat the test and were questioned about the time of saliva collection.
Saliva collection
Subjects were asked to obtain saliva samples at their usual bedtime and received the following instructions (verbal and written): 1) do not brush teeth, eat, or drink for at least 30 min before collection, 2) avoid watching television or reading material that might incite extreme emotions, 3) reschedule saliva collection if ill, e.g. cold, flu, gastroenteritis, or worse or if recently in an argument or received particularly good or bad news.
To collect saliva, subjects were told to place cotton salivette(s) (Sarstedt, Inc., Newton, NC) inside their mouth and gently chew and/or suck on them for 1–3 min until they became soaked in saliva. Salivettes were stored in home freezers until delivery to the study sites where they were frozen until shipment to the reference laboratories.
Salivary cortisol assays
Salivettes were thawed and spun down immediately before assay of the supernatant. The Esoterix RIA detection limit was 50 ng/dl (1.4 nmol/liter). The intraassay and interassay coefficients of variation (CV) were 14% at 190 ng/dl (5.2 nmol/liter) and 15% at 160 ng/dl (4.4 nmol/liter), respectively. Cross-reactivity with other steroids was less than 0.6% except for corticosterone (7.6%), 21-deoxycortisol (10%), and prednisolone (46%). During the study, the Mayo LC-MS/MS detection limit was 4 ng/dl (0.11 nmol/liter). The intraassay CV was 7.2% at 109 ng/dl (3.0 nmol/liter) and interassay CV was 5.8% at 52 ng/dl (1.4 nmol/liter).
Mayo set their reference range (<100 ng/dl, 2.8 nmol/liter) using the mean ± 2 SD of midnight salivary cortisol levels in 102 adult volunteers (approximately 60% female, ages 18–55 yr except one aged 90 yr). Esoterix set their reference range (50–170 ng/dl (1.4–4.7 nmol/liter) using the minimum and maximum 2300-h salivary cortisol values obtained in a group of 26 adult volunteers (approximately 50% female, ages unknown). The volunteers’ weight, body mass index (BMI), or other medical history is not available.
Statistical analysis
Salivary LC-MS/MS and RIA values below the lower detection limits were analyzed as the numerical values of the limits [<4.0 ng/dl (0.11 nmol/liter) and <50 ng/dl (1.4 nmol/liter), respectively]; data were not used to develop reference ranges.
Clinical characteristics were described using frequency distributions and simple descriptive statistics. Categorical variables were compared using 
2 tests. For comparisons of salivary cortisol levels within each assay, obese subjects were categorized according to age (<60 vs. 
60 yr), diabetes status [current diagnosis, fasting blood glucose 126 mg/dl (7.0 mmol/liter) or 2-h post glucose 200 mg/dl (11.1 mmol/liter)] (12), systolic blood pressure (<140 vs. 140 mm Hg), and BMI (34 vs. >34 kg/m2). Continuous data were not distributed normally; thus, nonparametric tests (i.e. Van der Waerden for independent groups and Wilcoxon signed rank test for paired groups) were used for analysis. Similarly, Spearman rank correlations were used to assess the relation between LC-MS/MS and RIA salivary cortisol results and their correlation with UFC.
For the comparison of specificity, Fisher’s exact test was used if groups were independent (i.e. obese vs. healthy volunteers), and McNemar’s test was used for comparison of paired data (e.g. RIA vs. LC-MS/MS in obese subjects with both results). In addition, a composite exact test was used to compare overall false-positive rates. It combined data from independent and paired groups (e.g. RIA vs. LC-MS/MS in obese subjects with and without paired data). The rationale and form of the test statistic for the composite test is analogous to that derived for a very similar problem, namely Birch’s (13) generalization of Fisher’s exact test to a set of stratified 2 x 2 tables.
Two techniques were used to evaluate the degree of agreement between the absolute value of paired salivary cortisol results from LC-MS/MS and RIA. The first was a visual method using a modification of the Bland-Altman plot (14) in which the difference between LC-MS/MS and RIA values (y-axis) was plotted against LC-MS/MS salivary cortisol values (x-axis) instead of the average of LC-MS/MS and RIA values. This decision was based on our consideration of LC-MS/MS as the gold standard. Good agreement was defined as 1) a mean difference close to zero, 2) 95% or more of data falling within 2 SD of the mean difference, and 3) almost no extreme outliers falling outside 3.5 SD from the mean difference. A second method used the designation of results as normal or abnormal based on each laboratory’s reference range. The percent agreement (both results normal or abnormal) and 
-statistics were calculated. A -statistic of 0.60 or greater indicated good agreement (15). The 95% confidence intervals (CI) about the -statistic also were computed.
To obtain more representative estimates, certain statistics (e.g. Bland-Altman thresholds and clinical characteristics) were computed on a reduced sample that excluded three extreme RIA outliers. These outliers were identified on the scatterplot as those subjects with extreme RIA values but normal LC-MS/MS values.
All statistical analyses were two-tailed, with statistical significance defined as a P value of 0.05. Data were analyzed using SAS system software, release 9.1 (SAS Institute, Inc., Cary, NC).
|
Results |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
displays the clinical characteristics of the subjects. The obese group had higher BMI and a larger proportion of women.
|
Of 310 enrolled, 274 obese subjects completed at least one screening test: 229 completed three tests, 40 completed two tests, and five completed one test. The others did not initiate testing. Most saliva and 24-h urine samples were collected on the same day. None of the subjects have been diagnosed with Cushing’s syndrome.
Of 274 obese subjects, 261 had at least one screening salivary cortisol measurement. Salivary cortisol was measured by RIA in 250 of these subjects and by LC-MS/MS in 169 subjects. The first 77 subjects did not provide split samples. Results from both assays were available in 156 subjects, whose clinical characteristics were not different from those of the entire group. For each assay, approximately half of the missing samples had an insufficient quantity of saliva for assay and the remaining samples were not collected.
Salivary cortisol values ranged from less than 4.0 to 643 ng/dl (<0.11–17.7 nmol/liter) by LC-MS/MS (normal 100 ng/dl, 2.80 nmol/liter) and from less than 50 to 2800 ng/dl (1.4–77.3 nmol/liter) by RIA (normal 170 ng/dl, 4.7 nmol/liter) (Fig. 1 and Table 2).
|
|
Overall, RIA had lower specificity (212 of 250, 85%) than did LC-MS/MS (156 of 169, 92%; P = 0.06), although this difference was not statistically significant. Of 156 subjects with split sample data, the specificity of RIA (n = 22, 86%) was significantly lower than that of LC-MS/MS (n = 10, 94%; P = 0.008). Seven subjects had abnormal salivary cortisol levels in both assays.
The overall agreement between LC-MS/MS and RIA for the assignment of normal or abnormal results in the paired samples was 88%. The -statistic, taking into account agreement by chance alone, was 0.38 (CI 0.16–0.60), indicative of only slight agreement. There was better agreement for normal values than abnormal ones, 94% (CI 90–96%) vs. 44% (CI 26–62%). A Bland-Altman plot (Fig. 2A) (14) showed that RIA values were consistently higher than LC-MS/MS results.
|
. Of these 38 subjects, 26 had normal follow-up testing. Three others had normal follow-up testing except for abnormal salivary cortisol results by RIA (210, 250, and 830 ng/dl). Additional repeat testing was not obtained. Nine subjects did not return for follow-up, including eight subjects with only a RIA salivary cortisol abnormality (three of these had normal LC-MS/MS results, and five had no LC-MS/MS results). The ninth subject had a minimally elevated UFC (46 µg/24 h, 1.7 nmol/liter) and a normal LC-MS/MS salivary cortisol result.
|
Subjects with abnormal results reported bedtimes from 2130 to 0030 h; most went to bed between 2300 and 2330 h. None admitted to recent travel, shift work, or a change in bedtime. One subject, with initially abnormal salivary cortisol and DST results, reported going to bed around 1800 h regularly; repeat testing was normal.
Salivary cortisol levels by LC-MS/MS or RIA were not significantly different by categories of systolic blood pressure or BMI. RIA, but not LC-MS/MS, gave higher salivary cortisol levels in subjects who were older [60 yr of age (n = 32) vs. < 60 yr (n = 218); P = 0.01] or diabetic (n = 52) vs. no diabetes (n = 156) (P = 0.04, Table 4). Excluding the extreme RIA outliers, the same associations were observed, except that salivary cortisol levels by RIA were higher among those with systolic blood pressure of at least 140 mm Hg (P = 0.04).
|
No volunteer had clinical or historical features suggestive of Cushing’s syndrome. Forty-seven of 60 healthy volunteers provided both urine and saliva samples. Fifty had salivary cortisol measurement by RIA, and 57 had LC-MS/MS measurements. Forty-nine volunteers had split sample results. Most missing values represented an inadequate collection. Salivary cortisol values ranged from less than 4.0 to 572 ng/dl (<0.11–16 nmol/liter) by LC-MS/MS and from less than 50 to 940 ng/dl (1.4–26 nmol/liter) by RIA (Fig. 1
and Table 2).
The specificity of LC-MS/MS was 86% (49 of 57), similar to the specificity of RIA, 82% (41 of 50) (P = 0.71). Of the eight volunteers with abnormal results by LC-MS/MS, six also had abnormal results by RIA, the seventh had a normal RIA result, and the eighth subject provided an insufficient sample. Six of these eight volunteers had normal follow-up testing, and two did not return. Three volunteers with only an abnormal salivary cortisol by RIA did not return for follow-up (all had a normal UFC result). Volunteers with abnormal results reported bedtimes between 2130 and 0030 h; most went to bed between 2300 and 2330 h.
In healthy volunteers, there was a 92% overall agreement between the two assays for assignment of the results as normal [95% (CI 82–99%)] or abnormal [75% (CI 33–96%)]. The -statistic was 0.70 (CI 0.43–0.97), indicating substantial agreement. The Bland-Altman plot (Fig. 2B) (14) showed that most values fell within 2 SD of the mean. The one outlier value was less extreme than those in the obese population.
The RIA false-positive rate of 15% in the obese population was not significantly different from the 18% found in the healthy volunteers (P = 0.67). Likewise, the LC-MS/MS false-positive rate of 8% in the obese population was not significantly different from the 14% rate in the healthy volunteer population (P = 0.19).
|
Discussion |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
What could explain this unexpectedly high rate of abnormal results? It is possible that the results are correct and indicate the presence of early or subclinical Cushing’s syndrome in the 14 obese subjects with abnormal results who did not return for additional evaluation. Thirteen of these 14 subjects had only abnormal RIA or LC-MS/MS salivary cortisol levels, whereas the other subject also had a minimally elevated UFC. However, Cushing’s syndrome is very unlikely in the healthy volunteers. Because the obese and healthy volunteers had similar rates of abnormal results, Cushing’s syndrome is probably not the cause.
Another explanation for apparently abnormal results is that subjects obtained the specimens at a time of day when cortisol is normally higher and did not disclose this. Also, not all subjects’ bedtimes matched the collection times of the healthy volunteer populations used to develop the laboratory reference ranges.
Salivary cortisol concentrations increase as a result of physiological activation of the hypothalamic-pituitary-adrenal axis (16). Although subjects were instructed to avoid stimulating activities, we cannot exclude this possibility.
Alternatively, the values may be spuriously elevated. It is not known whether contamination with blood might increase salivary cortisol levels. Although Kivlighan et al. (17) reported that minor to moderate blood leakage as a result of vigorous tooth brushing had no effect, oral sores or injury might be associated with increased amounts of blood contamination.
There are two possible explanations for the discrepancy in absolute values between the two assays. First, RIA measures additional noncortisol steroids that might falsely elevate salivary cortisol values. Over- or underactivity of salivary gland 11ß-hydroxysteroid dehydrogenase type 2 might alter the relative contribution of corticosterone measured as cortisol by RIA (16). There could be direct contamination of the salivette by steroid-containing lotions or oral gels, which are used commonly in the general population. Cross-reactivity of synthetic steroids in the RIA might identify patients with surreptitious steroid use. Although subjects with abnormal results were not known to be using these agents, it is likely that unintentional synthetic steroid contamination caused the three extreme outliers with normal LC-MS/MS results.
Second, the reference ranges may not be appropriate and could be skewed more in one assay than the other. We used normative ranges developed by each laboratory based on the values from healthy individuals. In contrast, most previously reported studies determined the clinical decision limits retrospectively and applied them to the entire study population (1, 2, 3, 4, 5). These clinical decision limits were based on analysis of various groups of normal volunteers, subjects with disease, and/or subjects in whom the disease was excluded (18). The exclusive use of healthy control data for the assignment of a diagnosis may not be ideal, because the definition of normal is then descriptive and not diagnostic (19). The data presented here suggest strongly that use of a reference range instead of a clinical decision limit is not appropriate for the diagnosis of Cushing’s syndrome. However, it is likely that outliers would remain, and causes for these extreme elevations should be evaluated.
Failure to partition the reference range by categories such as age, sex, and confounding comorbidities may contribute to abnormal results. Using a different immunoassay from the one used here, Liu et al. (7) found significantly higher late-night salivary cortisol levels in older, hypertensive, and/or diabetic men. Based on the clinical characteristics used in that study, we found salivary cortisol levels to be higher by RIA but not by LC-MS/MS in subjects 60 yr or older and in those with diabetes; hypertension did not influence the results (Table 4). These results suggest that the influence of age, gender, and confounding comorbidities should be explored further when developing reference ranges.
In this study, the specificity of salivary cortisol ranged from 82–92% depending on the population and assay. Other screening tests such as UFC and DST have specificities reported as 73% (20) and 87.5% (21), respectively. These screening tests evaluate different aspects of the hypothalamic-pituitary-adrenal axis and complement each other in diagnosing Cushing’s syndrome.
In conclusion, this study suggests that although a normal bedtime salivary cortisol result is useful to exclude Cushing’s syndrome, an abnormal level should not be used as the sole tool for the diagnosis of Cushing’s syndrome if laboratory-provided reference ranges are used for interpretation. Further development of optimal diagnostic criteria for Cushing’s syndrome is needed for each type of assay. However, given the higher specificity of the LC-MS/MS salivary cortisol assay, it should be preferred over RIA. Practicing endocrinologists should question patients with abnormal salivary cortisol results regarding collection technique and current medications/lotions and require additional screening tests before concluding that significant hypercortisolism is present.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
Author Disclosure: The authors have nothing to declare.
First Published Online June 5, 2007
Abbreviations: BMI, Body mass index; CI, confidence interval; CV, coefficient of variation; DST, dexamethasone suppression test; LC-MS/MS, tandem mass spectrometry; UFC, 24-h urine cortisol excretion.
Received December 26, 2006.
Accepted May 24, 2007.
|
References |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  S. K. Baid, D. Rubino, N. Sinaii, S. Ramsey, A. Frank, and L. K. Nieman Specificity of Screening Tests for Cushing's Syndrome in an Overweight and Obese Population J. Clin. Endocrinol. Metab., October 1, 2009; 94(10): 3857 - 3864. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
H. Raff Utility of Salivary Cortisol Measurements in Cushing's Syndrome and Adrenal Insufficiency J. Clin. Endocrinol. Metab., October 1, 2009; 94(10): 3647 - 3655. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Boscaro and G. Arnaldi Approach to the Patient with Possible Cushing's Syndrome J. Clin. Endocrinol. Metab., September 1, 2009; 94(9): 3121 - 3131. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. Wood Salivary steroid assays - research or routine? Ann Clin Biochem, May 1, 2009; 46(3): 183 - 196. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. K. Nieman, B. M. K. Biller, J. W. Findling, J. Newell-Price, M. O. Savage, P. M. Stewart, and V. M. Montori The Diagnosis of Cushing's Syndrome: An Endocrine Society Clinical Practice Guideline J. Clin. Endocrinol. Metab., May 1, 2008; 93(5): 1526 - 1540. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. Nieman The Dexamethasone-Suppressed Corticotropin-Releasing Hormone Test for the Diagnosis of Cushing's Syndrome: What Have We Learned in 14 Years? J. Clin. Endocrinol. Metab., August 1, 2007; 92(8): 2876 - 2878. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Endocrinology | Endocrine Reviews | J. Clin. End. & Metab. |
| Molecular Endocrinology | Recent Prog. Horm. Res. | All Endocrine Journals |
) and healthy volunteer subjects (•). Dashed lines represent the upper normal limit for each assay.