How Doctors Screen for Cortisol Excess and Why One Abnormal Test Is Not a Diagnosis

Doctors screen for cortisol excess with three first-line tests, the overnight dexamethasone suppression test, late-night salivary cortisol, and 24-hour urinary free cortisol, and the Endocrine Society guideline asks for two of them to agree before the result is called Cushing syndrome. Each test is built to catch nearly everyone who has the disease, which is exactly why each one also flags some people who do not. A single abnormal value is a reason to keep looking, not a diagnosis.

Key points#

• Three first-line tests are used interchangeably as starting points: overnight dexamethasone suppression, late-night salivary cortisol, and 24-hour urinary free cortisol.
• The tests are tuned for sensitivity, so false positives are expected and common.
• Guidelines require two different tests to agree before confirming Cushing syndrome.
• Birth control pills, some anticonvulsants, heavy alcohol use, poorly controlled diabetes, severe obesity, and depression can all raise cortisol without any tumor.
• Who gets tested matters as much as the test itself, because screening a low-risk crowd produces mostly false alarms.

Why this is a hard problem before any blood is drawn#

Cushing syndrome is uncommon, and several everyday conditions push cortisol up without a tumor behind it. That combination is the whole difficulty. A test can be genuinely good and still mislead you when it is aimed at the wrong crowd, because when a disease is rare, even a small false-positive rate outnumbers the true cases it catches. So the screening system is designed around two questions at once: is the test abnormal, and was this the right person to test?

Everything downstream, the choice of test, the insistence on repeating it, the referral thresholds, follows from taking those two questions seriously before you order anything.

The rhythm the tests are probing#

Cortisol is not steady through the day. It surges in the early morning, drifts down through the afternoon, and bottoms out near midnight. It also answers to a feedback loop: dexamethasone, a synthetic steroid, mimics cortisol well enough that a healthy pituitary reads it as "enough already" and turns its own signal down.

Cushing syndrome breaks both patterns. The output becomes autonomous, meaning it no longer respects the clock or the feedback signal. Most cases trace to a pituitary or adrenal tumor, or to cortisol-like medication taken for another illness. Each screening test attacks this from a different angle, and the underlying logic is easier to hold onto than any single cutoff number.

The three first-line tests#

Overnight dexamethasone suppression#

The patient swallows 1 mg of dexamethasone around 11 pm and has blood drawn for cortisol the next morning. A normal axis reads the steroid as a shut-off signal, so morning cortisol lands low. The Endocrine Society guideline (Nieman and colleagues, in the Journal of Clinical Endocrinology and Metabolism) sets the abnormal line at a serum cortisol above 1.8 micrograms per deciliter (50 nmol per liter). Staying above that line means the feedback loop did not respond as expected. It is a flag, not a finding.

Late-night salivary cortisol#

This one targets the midnight low. In cortisol excess, that nighttime trough flattens out. A saliva sample taken between 11 pm and midnight, usually on two different nights, captures the free and biologically active fraction of cortisol with no needle involved. Anything above the testing laboratory's reference range counts as abnormal. That cutoff shifts from lab to lab because it depends on the assay, which is one more reason the guideline leans on repeat sampling rather than a lone number.

24-hour urinary free cortisol#

Collecting every drop of urine across a full day averages cortisol output over the whole period instead of freezing one moment. The guideline asks for at least two separate 24-hour collections, since a single one is easy to spoil by collecting too much or too little. A result above the laboratory's upper limit of normal is abnormal.

The 2008 guideline treats these three as broadly interchangeable opening moves, and the 2021 Pituitary Society consensus on Cushing's disease (Fleseriu and colleagues, in The Lancet Diabetes and Endocrinology) keeps the same menu. Which test to reach for first depends on the patient, and that is where pretest probability comes in.

The mimics that trip a single test#

Screening tests for a serious, treatable disease are deliberately built to miss almost no one. The price of that design is that healthy people, and people with unrelated problems, will sometimes cross the threshold anyway. A handful of these false alarms are well documented.

Estrogen-containing oral contraceptives raise cortisol-binding globulin, the carrier protein for cortisol in blood. Standard serum assays measure total cortisol, so more carrier protein inflates the reading and can produce a false-positive suppression test; the guideline notes this pattern in roughly half of women taking the pill and suggests stopping estrogen weeks before testing. Anticonvulsants such as phenytoin and carbamazepine speed up how fast dexamethasone is cleared, so the dose is gone before it can do its job. Drinking very large fluid volumes inflates urinary cortisol. Smoking or eating licorice before a saliva sample skews that one.

Then there are physiologic states that genuinely rev the stress axis. Poorly controlled diabetes, heavy alcohol use, severe obesity, depression and other psychiatric illness, and acute illness all raise cortisol for real. Once grouped loosely as pseudo-Cushing, this category is now described more precisely as non-neoplastic hypercortisolism, a framing detailed by Findling and Raff in the Journal of the Endocrine Society. The cortisol elevation is real; it comes from an overworked but structurally normal system rather than a tumor. Because these patients can look and test much like true Cushing syndrome, telling them apart takes time and repeat measurement, not a single draw.

Who should be tested, and the two-test rule#

Pretest probability is simply the odds that a person has the disease before any lab comes back, read off the clinical picture you can see. The guideline is picky about who to test at all: people with features that are odd for their age, such as early osteoporosis or hypertension in someone young, and people with several progressive and more specific signs together, including reddish-purple stretch marks, easy bruising, a rounded and flushed face, and weakness in the muscles closest to the trunk. Testing a broad, low-risk group instead almost guarantees that most abnormal results will be false, for the same rarity math that opened this article.

That reasoning is what the two-test structure is built on. One abnormal screen triggers referral and a second, different test. Two abnormal results that agree earn a move to the next stage, working out the cause with ACTH measurement and imaging. Two normal results largely settle the matter. When the two disagree, the answer is more testing or specialist judgment, not a guess. For suspected non-neoplastic hypercortisolism, the 2021 consensus adds a practical move: treat the underlying condition, then reassess and repeat first-line testing three to six months later, since fixing the trigger often lets the axis reset on its own.

None of this is red tape. It is a testing sequence engineered to fail safely, sensitive enough to catch real disease, structured enough to keep a single number, thrown off by a birth control pill or one rough week, from becoming a label you carry for years.