Late-night salivary cortisol and cortisone should be the initial screening test for Cushing’s syndrome

Introduction

Cushing’s syndrome (CS) is caused by prolonged and inappropriate exposure of tissues to glucocorticoids (1). Endogenous CS often poses considerable diagnostic challenges. Most guidelines recommend two different tests to screen for CS including late-night salivary cortisol (LNSC), overnight dexamethasone suppression test (ODST), low-dose dexamethasone suppression test (LDDST) and 24-h urinary free cortisol (UFC) (1, 2, 3). A hallmark of CS is the disruption of the circadian rhythm of cortisol secretion whereby, in the evening, cortisol secretion is higher in CS patients than in normal subjects (4). Therefore, assessment of LNSC, without the impracticality of hospitalisation and disruption of a normal routine, would be the ideal screening test for CS. Collection of LNSC is convenient for patients, it is non-invasive and avoids the need to obtain unstressed late-night blood samples, which is impractical in most circumstances (4). Despite the fact that LNSC has been shown to have high diagnostic sensitivity and specificity and that it has been shown to be cost-effective, it still remains the least widely used biochemical screening tool for CS (5, 6).

Salivary glands express 11-β-hydroxysteroid dehydrogenase (11B-HSD2) which converts salivary cortisol to cortisone. The amount of salivary cortisone is significantly greater than salivary cortisol and an increased cortisol/cortisone ratio reflects exposure to both endogenous and exogenous cortisol. Salivary cortisone can be reliably measured by liquid chromatography-tandem mass spectrometry (LC-MS/MS) (7). There is no evidence of 11B-HSD2 activity in collected saliva. Therefore, measurement of salivary cortisone is particularly useful when contamination with topical or oral corticosteroids is suspected (8). In this study, we compare the diagnostic accuracy of LNSC and LNS cortisone to the diagnostic accuracy of other commonly used tests for the diagnosis of CS.

The aim of this retrospective study was to assess the reliability of LNSC and LNS cortisone in diagnosing CS compared to other commonly used methods in a single tertiary referral centre.

Methodology

This study was approved by the Imperial College Healthcare NHS Trust governance team who confirmed that we are reporting on routinely collected non-identifiable clinical audit data, therefore, no approval from a research ethics committee was required under the UK policy framework for Health and Social Care. Data were collected from patients who underwent LNSC testing between 2017 to March 2021 at Imperial College Healthcare NHS Trust which is a regional tertiary referral centre. Patients with a high pre-test probability for CS were included in the study. Patients who were suspected to have cyclical CS, subclinical CS, pseudo-CS due to disrupted circadian rhythm or patients who were known non-compliant with the collection protocol were excluded from the study. The diagnosis of pituitary Cushing’s disease (CD) was confirmed by positive histology. The diagnosis of adrenal CS was established by suppressed adrenocorticotrophic hormone (ACTH) levels in the context of hypercortisolaemia and histology in keeping with a cortisol-producing adenoma/adrenocortical carcinoma.

Patients were advised to collect saliva between 23 h and midnight. Saliva was collected with Salivette® tubes (Sarstedt, Germany). Patients were asked to collect saliva before brushing their teeth and no earlier than 30 min after eating or drinking. Patients were advised to avoid smoking, drinking alcohol or coffee for 2 h before sample collection. Steroid-containing skin creams were to be avoided, and patients were advised to wash their hands before collection. Some patients gave their saliva samples while having other tests (such as cortisol day curve, overnight dexamethasone suppression test (ODST) and LDDST) as an inpatient. Collected samples were stored in the patient’s refrigerator and subsequently delivered to the pathology department. Salivary cortisol and cortisone analysis was carried out by liquid chromatography and tandem mass spectroscopy (LC-MS/MS).

Results are presented as mean ± s.d. for quantitative variables. Sensitivity, specificity, area under the curve (AUC), positive likelihood ratio (LR), negative LR, positive predictive value (PPV), negative predictive value (NPV), of LNSC, LNS cortisone, ODST, LDDST and UFC were calculated. Cut-off values of <2.6 nmol/L for LNSC, <18 nmol/L for LNS cortisone and less than the upper limit of normal for 24-h UFC (270 nmol/24 h) were considered normal based on assay-specific reference ranges. Serum cortisol of less than 50 nmol/L for ODST and LDDST was considered normal (3).

Results

Out of 76 patients, 54 were eligible for the study (45 were females and nine were males). The mean age was 44.8 years. CS was diagnosed in 21 patients and CS was excluded in 33 patients (Table 1). Ten patients were diagnosed with CD first presentation, six with recurrent CD, two with CS due to adrenal adenomas, two with CS due to adrenocortical cancers and one patient had CS due to an ectopic ACTH source. The latter patient underwent inferior petrosal sinus sampling to exclude a central source of ACTH with a confirmed ACTH source from diffuse idiopathic pulmonary neuroendocrine cell hyperplasia. Table 2 summarises the investigations performed.

LNSC had a sensitivity of 95%, a specificity of 91%, a PPV of 87% and a NPV of 97%. The AUC was 0.931, the positive LR was 10.5 and the negative LR was 0.05. LNSC had a specificity of 100%, with a PPV of100%. However, sensitivity (86%) and NPV (92%) were lower than for LNSC. 36 patients underwent ODST. Although the sensitivity and the NPV of ODST were 100%, the specificity and the PPV were 77 and 74% respectively. Like ODST, LDDST had a sensitivity and a NPV of 100% but specificity (83%) and PPV (87%) were higher than with ODST. Fifteen patients underwent UFC as a screening test for CS. UFC was 100% specific with a PPV of 100% but had showed the lowest sensitivity (60%) and NPV (56%) of all investigations (Table 3). Receiver operating characteristic (ROC) curves were generated for LNSC, LNS cortisone, ODST, LDDST and UFC levels and showed area under the curve of 0.931, 0.929, 0.886, 0.917 and 0.800, respectively. Figure 1 shows optimal cut-offs derived by ROC analysis for LNSC, LNS cortisone, ODST, LDDST and UFC. LNSC value of >3.3 nmol/L provided a sensitivity of 95.2% and a specificity of 97% and LNSC cortisone value of >14.5 nmol/L provided a sensitivity and specificity of 95.2 and 100%, respectively, in diagnosing CS. Moreover, ODST value of 47 nmol/L provided a sensitivity of 100% and a specificity of 77.3% while LDDST value of >28 nmol/L provided a sensitivity and specificity of 100 and 83.3%, respectively. AUC for LNSC, LNS cortisone, ODST, LDDST and UFC levels with optimal cut-offs by ROC analysis showed 0.990, 0.997, 0.948, 0.881 and 0.900, respectively.

View Full Size

ROC curve for (A) LNSC, (B) LNSC cortisone), (C) ODST, (D) 48-h LDDST and (E) 24-h UFC. AUC, area under the curve; LNSC, late-night salivary cortisol; LDDST, low-dose dexamethasone suppression test; ODST, overnight dexamethasone suppression test; ROC, receiver operating characteristics; UFC, urinary free cortisol.

Citation: Endocrine Connections 11, 7; 10.1530/EC-22-0050

• Download Figure
• Download figure as PowerPoint slide

Sub-analysis of LNSC for CD with first presentation (n = 10 patients) and recurrent CD (n = 6) had same sensitivity (100%), specificity (91%), NPV (100%), positive LR (10), negative LR (0) and AUC (0.960) except PPV with 77 and 67%, respectively. LNS cortisone for CD had sensitivity of 90%, NPV 97%, negative LR 0.1 and AUC of 0.950. Patients with recurrent CD had sensitivity of 83%, NPV 97%, negative LR 0.17 and AUC of 0.920 without affecting specificity (100%) or PPV (Table 4). Figure 2 also shows the same data.

View Full Size

Box and whisker plot demonstrating (A) late-night salivary cortisol in Cushing’s disease first presentation; (B) late-night salivary cortisol in recurrent Cushing’s disease; (C) late-night salivary cortisone in Cushing’s disease first presentation; (D) late-night salivary cortisone in recurrent Cushing’s disease; median value showed by horizontal line; box indicates 25–75th percentile, whisker indicates 2.5–97.5th percentiles of the data. P values are from Mann–Whitney rank sum tests.

Citation: Endocrine Connections 11, 7; 10.1530/EC-22-0050

• Download Figure
• Download figure as PowerPoint slide

Discussion

Tandem mass spectrometry reliably measures salivary cortisol without cross-reactivity between cortisol and synthetic steroids (9, 10). Jones et al. showed LC-MS/MS method demonstrated excellent imprecision and accuracy for salivary cortisol and salivary cortisone. The assay was shown to be very specific for both cortisol and cortisone following the analysis of 29 structurally related steroids at supraphysiologic doses. Only 0.3% interference from prednisolone was observed for salivary cortisol assay and 1% interference from prednisone for salivary cortisone. Salivary cortisol assay was linear up to a concentration of 2293 nmol/L and salivary cortisone was linear up to 3676 nmol/L. The lower limit of quantification was found to be 0.75 nmol/L for cortisol and 0.5 nmol/L for cortisone (11).

ERCUSYN data suggest that LNSC is not frequently used in the diagnostic workup for suspected CS despite having been shown to have excellent sensitivity and specificity (6). Our data confirm that LNSC has high sensitivity and high NPV when used in a tertiary referral setting. By contrast, UFC had very low sensitivity and NPV, thus making it unreliable as a screening tool for CS. A recent guideline update on CS diagnosis and management suggests LNSC has 97% sensitivity and 97.5% specificity in diagnosing CS and 75–90% sensitivity and 93–95% specificity in diagnosing recurrent CD. Moreover, UFC has 91% sensitivity and 81.5% specificity in diagnosing CS and 68% sensitivity and 100% specificity in diagnosing recurrent CD (3). Despite that, UFC remains the first-line screening test in many countries which can be explained by its historic use and therefore wider availability. We strongly recommend that ERCUSYN should reassess whether there has been any change in the use of diagnostic tests for CS across Europe over the past 5 years.

LNSC is not affected by BMI or cortisol binding globulin and therefore accurately reflects serum cortisol concentrations. There is conflicting evidence regarding LNSC levels in polycystic ovarian syndrome. Ozkaya et al. found cortisol not to be affected by a PCOS phenotype (12). However, Basu et al. reported higher LNSC levels in PCOS patients compared to controls (13). Several studies found that there appears to be an age-related increase in LNSC (14, 15). It also appears that LNSC detects recurrent CD earlier than UFC thus making it the ideal test to follow up patients after successful pituitary surgery for CD (6, 11, 13). Our results are comparable to data published by others which confirm high sensitivity and specificity of LNSC (16, 17). A meta-analysis by Carroll et al. concluded that LNSC is a robust and convenient test to screen and diagnose CS. Similar to our own analysis, the authors found that LNSC had a diagnostic sensitivity of 92% and a specificity of 96% (18). Despite the strong evidence base and the practical advantages of the test, various reasons such as experience of a given centre, availability of adequate assays, individual diagnostic choices or delayed adaptation of evidence-based medicine into clinical practice may be responsible that LNSC is not yet the preferred first-line investigation for the evaluation of CS (6).

More than 90% of the variability in salivary cortisone can be accounted for by changes in serum cortisol (19). Recently published data show that when an 23 h upper reference limit of salivary cortisol and cortisone of 3.6 nmol/L and 13.5 nmol/L were used, the sensitivities and specificities of LNSC were 90 and 96%, and for LNS cortisone 100 and 95%, respectively (20). In our study, we used 18 nmol/L as the upper reference range for cortisone and calculated a sensitivity of 86% and specificity of 100%. When the upper reference limit was 14.5 nmol/L based on optimal cut-off by ROC analysis, sensitivity rose to 95.2% without affecting specificity. Normal LNSC: cortisone ratio is 0.2 with LC-MS/MS (6). Contamination of the salivary sample with topical hydrocortisone should be suspected when LNSC results are markedly increased, particularly when they are out of proportion with other biochemical results or clinical findings. In this instance, contamination with topical hydrocortisone would be suggested by a normal salivary cortisone concentration and high cortisol to cortisone ratio (21). LNSC and LNS cortisone can conveniently be analysed together with LC-MS/MS from a single saliva sample which appears to be one of the advantages compared to the immunoassay. However, many centres do not have access to LC-MS/MS. There is also the option of measuring LNSC by immunoassay which is much easier and cheaper with similarly excellent results (22).

A recently published study examining 1453 LNSC samples from 705 patients received by a diagnostic laboratory requested by both primary care physicians and specialists as part of clinical care showed that a majority of patients who had at least one positive LNSC sample did not have CS (22). A majority of samples in this study which was carried out in the United States were sent from primary care physicians following the publication of clinical practice guidelines. It is not known whether sampling conditions were accurately followed and there is no information about pre-test probability. In our study, the majority of the patients who had a positive LNSC had CS. There are several reasons for this discrepancy. Our centre is a tertiary referral centre, and, therefore, many of the patients investigated at our centre have a high pre-test probability. In keeping with our own data, Raff et al. confirmed that LNSC has excellent sensitivity and specificity in a tertiary referral setting when used in combination with a high pre-test probability (4). In addition, we excluded patients with so-called ‘pseudo-Cushing’s’ and those with subclinical CS in our study. Crucially, an experienced clinician’s assessment is the most important factor to correctly diagnose CS. For example, conditions like obstructive sleep apnoea result in hypoxia-induced stress and subsequent ACTH-driven hypercortisolaemia thus mimicking ACTH-dependent CS biochemically. However, results of investigations for hypercortisolism often fully normalise once obstructive sleep apnoea has been treated. Multiple sequential LNSC measurements are very useful to investigate diagnostically challenging patients with cyclic Cushing’s disease. However, the clinical assessment of an experienced endocrinologist remains key to make a correct diagnosis (23).

Sandouk et al. demonstrated that there can be significant fluctuations in salivary cortisol levels in CD. In newly diagnosed CD this seems to only rarely impair the diagnostic ability of LNSC, whereas in patients with recurrent or persistent disease after pituitary surgery, LNSC was frequently within the reference range, with the potential to cause false-negative results (20). In contrast with those findings, in our cohort, patients with proven recurrent CD all had elevated LNSC. In this study, only LNSC was measured by LC-MS/MS and measuring both LNSC and LNS cortisone using this method might improve its sensitivity as cortisol elevations maybe modest in recurrent disease. Garrahy et al. showed that a combination of LNSC and LNS cortisone had a sensitivity of 94% compared to LNSC and LNS cortisone alone which had a sensitivity of 92 and 87%, respectively. The previously mentioned guideline update on CS diagnosis and management recommends only using LNSC if cortisone concentrations can also be reported if an adrenal tumour is suspected (3). We fully agree that LNS cortisone should be routinely included in CS workup where available.

Limitations to our study were its retrospective nature, the small number of patients with CS and the fact that the selection of screening tests other than LNSC was based on the clinicians’ preferences. Nevertheless, our study adds further evidence to the existing literature confirming that LNCS in combination with LNS cortisone represents the ideal screening test for patients with suspected CS referred to a tertiary centre following a thorough clinical assessment by an experienced physician.

Declaration of interest

The authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the research reported.

Funding

This work did not receive any specific grant from any funding agency in the public, commercial, or not-for-profit sector.