Late-Night Salivary Cortisol Interpreter
Late-Night Salivary Cortisol Interpreter
One of the four first-line tests for Cushing’s syndrome, and the one whose result is most often read against the wrong number. Two things decide whether this result means anything, and neither is the value. The cut-off is assay-specific — immunoassay and LC-MS/MS produce different numbers on the same saliva and are not interchangeable — and the test measures the loss of the normal nocturnal cortisol nadir, so a patient whose sleep-wake cycle is disrupted has no nadir to lose and the test is invalid before it is run.
Is this late-night saliva abnormal?
Value + assay + timing → interpretable or notA patient with new hypertension, weight gain and easy bruising. Late-night salivary cortisol collected at 23:20, conventional sleep pattern, measured by enzyme immunoassay: 3.0 nmol/L.
Three assays, three cut-offs, one patient’s saliva
LC-MS/MS, cortisol — abnormal at or above 2.8 nmol/L (1.02 ng/mL, 0.101 µg/dL). Sensitivity 67.5%, specificity 84.6%.
LC-MS/MS, cortisone — abnormal at or above 8.7 nmol/L (3.15 ng/mL, 0.315 µg/dL). Sensitivity 92.5%, specificity 76.2%.
All three from one prospective study of 705 patients, all derived contemporaneously on the same cohort. They are that study’s numbers, not reference values — use your own laboratory’s.
- assay-specific
- the 2008 Endocrine Society guideline says the cut-offs "vary with different assay methodologies and are highly dependent on the sensitivity of the assays", and the 2021 consensus refers only to an assay-specific reference range. Neither prints a screening number. The three above span a three-fold range on the same saliva, which is why quoting one of them without naming the method is worse than quoting none
- the nadir, not the level
- what the test detects is the LOSS of the normal nocturnal fall in cortisol, which is among the earliest abnormalities in Cushing’s syndrome of any cause. That is why the timing is not a detail and why a disrupted sleep-wake cycle is disqualifying rather than merely awkward: there is no nadir to have lost
- cortisone
- 11-beta-hydroxysteroid dehydrogenase type 2 in the salivary gland converts most cortisol to cortisone as it enters saliva, so cortisone is present at higher concentration and gives a bigger signal. Measured alongside cortisol it also identifies contamination from topical hydrocortisone, which raises salivary cortisol but not cortisone
- 18.8 to 31.3%
- the sensitivity of all three assays for ACTH-INDEPENDENT, adrenal Cushing’s syndrome in the same cohort, with positive predictive values of 2.0 to 4.0%. This is the figure that should change practice and is almost never quoted: late-night salivary cortisol is a good screen for pituitary and ectopic disease and a poor one for adrenal disease. When the reason for testing is an adrenal nodule, use the dexamethasone suppression test
- 0.27 µg/dL (7.5 nmol/L)
- a figure from the 2021 consensus with 75-90% sensitivity, and it is NOT a screening cut-off — it is for detecting recurrence after treatment of Cushing’s disease. It appears here only because it circulates as though it were a screening threshold. Do not apply it to a first presentation
- unit bridge
- cortisol has a molecular weight of 362.46 Da, so 1 ng/mL is 2.75862 nmol/L and 1 µg/dL is 27.5862 nmol/L. ng/mL and µg/L are the same number. The cortisol unit converter does the same arithmetic for serum and urine
Worked example
A patient with new hypertension, weight gain and easy bruising. Late-night salivary cortisol collected at 23:20, conventional sleep pattern, measured by enzyme immunoassay: 3.0 nmol/L.
Sleep-wake cycle conventional and collection at 23:20, so the test is interpretable
The assay is an enzyme immunoassay, so the applicable threshold from Kannankeril 2020 is 3.3 nmol/L
3.0 is below 3.3 → negative on this platform, with a sensitivity of 97.5% and a negative predictive value of 99.8% behind it for ACTH-dependent disease
But the same study's LC-MS/MS cortisol threshold is 2.8 nmol/L, and 3.0 is above that — the identical saliva would have been reported as abnormal by a mass spectrometry laboratory
That is the point of the page. Before accepting either verdict, confirm which method your laboratory used and what cut-off it validated
And if the reason for testing had been an incidental adrenal nodule rather than a Cushingoid phenotype, a negative here would mean very little: sensitivity for adrenal Cushing's syndrome was 18.8 to 31.3% for all three assays
Performance of the three assays in the same 705 patients
| Assay | Cut-off | Sensitivity | Specificity | PPV | NPV |
|---|---|---|---|---|---|
| Enzyme immunoassay, cortisol | 3.3 nmol/L | 97.5% | 69.3% | 16.8% | 99.8% |
| LC-MS/MS, cortisol | 2.8 nmol/L | 67.5% | 84.6% | 21.8% | 97.6% |
| LC-MS/MS, cortisone | 8.7 nmol/L | 92.5% | 76.2% | 19.8% | 99.4% |
What invalidates the test before the number is read
| Problem | Why it matters | What to do |
|---|---|---|
| Night-shift or rotating-shift work | The cortisol rhythm follows sleep, not the clock, so there is no 23:00 nadir | Do not use this test — the 2021 consensus says so explicitly. Use the dexamethasone suppression test or urinary free cortisol |
| Assay not known | The cut-offs span three-fold across methods | Ask the laboratory which method and which validated cut-off |
| Sample outside 23:00 to midnight | Both limbs of the diurnal curve read higher than the nadir | Repeat in the window, time recorded at collection |
| Eating, drinking, smoking or brushing teeth beforehand | Contamination and blood from the gums raise the result | Repeat after 15 minutes of nothing by mouth |
| Topical or oral hydrocortisone | Chemically identical to cortisol; contaminates the sample directly | Measure cortisone as well, which is not raised by contamination |
| Depression, alcohol dependence, poorly controlled diabetes, obstructive sleep apnoea, acute illness | All raise late-night cortisol without pituitary or adrenal disease | Treat or account for these first; consider a different first-line test |
What this test actually measures, and why the number alone is not the result
Cortisol follows a diurnal rhythm with a peak shortly after waking and a nadir around the time a person normally falls asleep. In Cushing’s syndrome that nadir is lost, and the loss is among the earliest detectable abnormalities, present before the clinical phenotype is obvious. Late-night salivary cortisol is a way of sampling that nadir without a needle and without an inpatient stay: salivary cortisol tracks the free, unbound fraction in blood, it is stable at room temperature for days, and the patient can collect it at home on a chewed swab. That combination is why it became one of the four first-line tests in the 2008 Endocrine Society guideline, alongside 24-hour urinary free cortisol and the two dexamethasone protocols.
Two things about it are systematically underappreciated. The first is that there is no universal cut-off. The guideline says the proposed cut-offs vary with assay methodology and depend heavily on assay sensitivity; the 2021 Cushing’s disease consensus refers only to an assay-specific reference range and declines to print a screening number. The scale of the variation is not academic: in a single prospective study of 705 patients, thresholds derived contemporaneously on the same samples were 3.3 nmol/L by enzyme immunoassay, 2.8 nmol/L for cortisol by mass spectrometry and 8.7 nmol/L for cortisone by mass spectrometry. A result of 3.0 nmol/L is therefore negative in one laboratory and positive in another, on the same saliva. The assays also do not rank as expected — the immunoassay was the most sensitive of the three at 97.5%, and mass spectrometry cortisol the least, at 67.5%, with the authors concluding that mass spectrometry did not improve sensitivity.
The second is that the test is invalid when the sleep-wake cycle is disrupted, and this is not a soft caveat. The 2021 consensus states that late-night salivary cortisol should not be done in patients with disruption of the normal day and night cycle, such as night-shift workers. The reason is structural rather than statistical: the cortisol rhythm entrains to the sleep-wake cycle, not to the clock, so a night-shift worker sampled at 23:00 is being sampled at a point on their curve that corresponds to nothing in the derivation cohorts. Shift work is common, and the question is rarely asked. It should be asked before the test is ordered rather than after an abnormal result has generated a work-up.
One further limitation deserves to be better known. In the same 705-patient cohort, sensitivity for ACTH-independent, adrenal Cushing’s syndrome was only 18.8 to 31.3% across all three assays, with positive predictive values of 2.0 to 4.0%. So the test is a good screen for pituitary and ectopic disease and a poor one for adrenal disease — which matters, because the commonest modern reason to ask about cortisol excess is an adrenal nodule found incidentally. In that setting the 1 mg overnight dexamethasone suppression test interpreter is the test the 2023 adrenal incidentaloma guideline actually asks for, and the adrenal incidentaloma interpreter sets out the rest. The urinary free cortisol unit converter is the other first-line option, the cortisol unit converter converts units, and the ACTH-dependent or ACTH-independent interpreter is the next step once hypercortisolism is confirmed.
Frequently asked questions
What is a normal late-night salivary cortisol?
There is no single number, and quoting one without naming the assay is misleading. The 2008 Endocrine Society guideline states that proposed cut-offs "vary with different assay methodologies and are highly dependent on the sensitivity of the assays", and the 2021 Cushing’s disease consensus refers only to an assay-specific reference range. To show the scale: in one prospective study of 705 patients, the abnormal thresholds derived on the same cohort were 3.3 nmol/L by enzyme immunoassay, 2.8 nmol/L for cortisol by LC-MS/MS and 8.7 nmol/L for cortisone by LC-MS/MS. Ask your laboratory which method it uses and what cut-off it has validated, and read the result against that.
Can a shift worker have a late-night salivary cortisol test?
No — or at least, not interpreted against the usual cut-off. The 2021 Cushing’s disease consensus states that the test "should not be done in patients with disruption of the normal day and night cycle, such as night-shift workers", and the 2008 guideline says it may not be appropriate for shift workers or patients with variable sleep patterns. The reason is that the cortisol rhythm entrains to the sleep-wake cycle rather than to the clock, so a night worker has no cortisol nadir at 23:00 for the test to detect the loss of. Use the 1 mg overnight dexamethasone suppression test or a 24-hour urinary free cortisol instead. The same applies after recent long-haul travel, with a newborn at home, or with severe irregular insomnia.
Is LC-MS/MS better than immunoassay for salivary cortisol?
More specific, and — unexpectedly — less sensitive. In Kannankeril and colleagues’ 705-patient prospective study, enzyme immunoassay cortisol had a sensitivity of 97.5% and specificity of 69.3% for ACTH-dependent Cushing’s syndrome, while LC-MS/MS cortisol had a sensitivity of 67.5% and specificity of 84.6%. The authors concluded that mass spectrometry cortisol did not provide an improvement in sensitivity. LC-MS/MS measurement of salivary cortisone performed best overall, at 92.5% sensitivity and 76.2% specificity, because cortisone is present in saliva at higher concentrations. Mass spectrometry has one advantage neither immunoassay offers: measuring cortisol and cortisone together identifies contamination by topical hydrocortisone.
What raises a late-night salivary cortisol apart from Cushing’s syndrome?
Quite a lot, which is why specificity is the weak point of this test. Physiological and pathological causes include depression, alcohol dependence, poorly controlled diabetes, obstructive sleep apnoea, severe obesity, acute illness and any acute stressor; shift work and an irregular sleep pattern invalidate the test outright rather than merely raising it. Collection errors add more: eating or drinking within 15 minutes, smoking or chewing tobacco, brushing the teeth — which can introduce blood, and serum cortisol is far higher than salivary — and any topical or oral hydrocortisone near the mouth. In a referral population, the positive predictive value of an abnormal result in that study was 16.8% to 21.8% depending on assay, so a single abnormal result should always be confirmed.
How many samples are needed?
At least two, on separate nights, in most protocols. The reason is the specificity problem: with a single abnormal sample in a population where most positives are not Cushing’s syndrome, one result carries little weight. The 2008 Endocrine Society guideline additionally requires a second, different first-line test to be abnormal before the diagnosis is accepted — so a raised salivary cortisol is usually paired with a 1 mg overnight dexamethasone suppression test or a 24-hour urinary free cortisol. Where cyclical Cushing’s syndrome is suspected, repeated sampling over weeks to months is more informative than repeated sampling over days.
Related calculators
References
- Kannankeril J, Carroll T, Findling JW, Javorsky B, Gunsolus IL, Phillips J, Raff H. Prospective evaluation of late-night salivary cortisol and cortisone by EIA and LC-MS/MS in suspected Cushing syndrome. J Endocr Soc. 2020;4(10):bvaa107. doi:10.1210/jendso/bvaa107. 705 patients, 1,453 late-night salivary samples, 56 with confirmed Cushing syndrome. Derived abnormal cut-offs: EIA cortisol 3.3 nmol/L, LC-MS/MS cortisol 2.8 nmol/L, LC-MS/MS cortisone 8.7 nmol/L. In ACTH-dependent disease (n = 40): sensitivity/specificity/PPV/NPV of 97.5/69.3/16.8/99.8, 67.5/84.6/21.8/97.6 and 92.5/76.2/19.8/99.4 respectively. In ACTH-independent disease (n = 16): sensitivity 18.8–31.3%, PPV 2.0–4.0%.
- Fleseriu M, Auchus R, Bancos I, Ben-Shlomo A, Bertherat J, Biermasz NR, et al. Consensus on diagnosis and management of Cushing’s disease: a guideline update. Lancet Diabetes Endocrinol. 2021;9(12):847–875. doi:10.1016/S2213-8587(21)00235-7. "LNSC should not be done in patients with disruption of the normal day and night cycle, such as night-shift workers." Also: "Liquid chromatography tandem mass spectrometry can detect both cortisol and cortisone, thereby identifying contamination from topical hydrocortisone … specificity is higher when using mass spectrometry while immunoassay has higher sensitivity." The 0.27 µg/dL (7.5 nmol/L) figure in that paper is for detecting recurrence, not for screening.
- Nieman LK, Biller BMK, Findling JW, Newell-Price J, Savage MO, Stewart PM, Montori VM. The diagnosis of Cushing’s syndrome: an Endocrine Society clinical practice guideline. J Clin Endocrinol Metab. 2008;93(5):1526–1540. doi:10.1210/jc.2008-0125. The four first-line tests, and: "specific cut-off levels are proposed for each test [but] these cut-offs vary with different assay methodologies and are highly dependent on the sensitivity of the assays"; salivary cortisol "may not be appropriate for shift workers or patients with variable sleep patterns".
- data/_factors.json, analyte
cortisol: molecular weight 362.46 Da, PubChem CID 5754. 1 ng/mL = 2.75862 nmol/L; 1 µg/dL = 27.5862 nmol/L. The three cut-offs on this page are therefore 3.3 nmol/L = 1.196 ng/mL = 0.120 µg/dL; 2.8 = 1.015 = 0.101; 8.7 = 3.153 = 0.315.
Medical Disclaimer: The tools and content provided here are for educational and reference purposes only. They are not intended to substitute for professional medical advice, diagnosis, or treatment. Clinical decisions should always be based on the comprehensive assessment of a qualified healthcare professional.