Urine Free Cortisol in the Diagnosis of Cushing's Syndrome: Is It Worth Doing and, If So, How?

For more than four decades, the measurement of free cortisol excretion in the urine (UFC) has been considered a reliable test for the diagnosis of Cushing's syndrome (1). Over time, assay methodologies have evolved from competitive binding to immunoassay to liquid chromatography (LC) methods, now coupled to tandem mass spectrometry (MS/MS) in many reference laboratories. With the increasing use of more analytically specific assays, the upper limit of the reference range has decreased, but it is not clear that diagnostic utility has improved. The literature regarding test performance is largely based on immunoassays without direct comparison to newer analytic approaches. Others have questioned the diagnostic utility of UFC measurement in the diagnosis of Cushing's syndrome and have suggested the use of different tests as a first step to screen for hypercortisolism (2).

After almost half a century of use and change, we pause to ask: 1) does the measurement of UFC have adequate diagnostic sensitivity and specificity to be a useful test; and, if so, 2) which analytic methodology has the best diagnostic performance? In a recent editorial about the measurement of androgens and estrogens in submissions to the JCEM, Handelsman and Wartofsky (3) remarked: “It is anticipated that this requirement for MS-based assays [in publications in the JCEM] will extend to adrenal steroids… in the near future.” In light of this statement, we will discuss the current status of UFC measurements and hopefully encourage the endocrine community to evaluate this issue comprehensively.

The recognition of unsuspected endogenous hypercortisolism in patients with adrenal nodules, diabetes mellitus, hypertension, and osteoporosis has increased our awareness of this enigmatic disorder and mandates better diagnostic tools (4, 5). The three first-line screening tests recommended for the diagnosis of Cushing's syndrome interrogate different aspects of the pathophysiology of this challenging disease (1, 5). These are: 1) failure to achieve the nadir of the cortisol circadian rhythm—evaluated using the measurement of late-night salivary cortisol (LNSC); 2) attenuated sensitivity to glucocorticoid negative feedback—tested using low-dose dexamethasone suppression testing; and 3) increased production of cortisol—assessed by 24-hour UFC measurement. Although all of these tests have strengths and weaknesses. LNSC has the best diagnostic performance as an initial screening test for patients with suspected hypercortisolism, with the exception of mild ACTH-independent hypercortisolism, for which a low-dose dexamethasone suppression test is most sensitive (6).

Although most direct immunoassays for salivary cortisol show cross-reactivity with similar steroids (in particular, cortisone), there is evidence that immunoassays paradoxically might perform better diagnostically than the more analytically specific LC-MS/MS method (reviewed in Ref. 6). Longitudinal studies in patients after pituitary surgery for Cushing's disease have shown that increases in late-night cortisol secretion are the earliest biochemically detectable abnormality during recurrence and that increased UFC is a later event (6).

Some patients with endogenous mild hypercortisolism do not have Cushing's syndrome, leading to concerns about false-positive results. Newer assays with lower, narrower reference ranges appear to have reduced the infamous false-positive problem. However, many patients with mild Cushing's syndrome have normal UFC with these assays (7). This problem is only partially explained by within-subject variability, which is as high as 50% in patients with proven Cushing's syndrome (8).

Part of the reason for this modern dilemma is that free cortisol only appears in the urine when its concentration exceeds the binding capacity of plasma. A patient with a serum cortisol of 12 μg/dL all day long is likely to have Cushing's syndrome but might have a normal 24-hour UFC measurement. Another important factor is cortisol metabolism. Although the daily cortisol production rate is 6–8 mg/m²/d (∼10 mg/d) (9), the reference range for UFC measured by LC-MS/MS is only < 40 μg/d—so where does the other 9960 μg go? Over 99% of cortisol is metabolized and glucuronidated in the liver, and then it is further metabolized and excreted by the kidney (10). The tetrahydrocortisols and cortisones, cortols, and other metabolites, which are excreted in milligrams per day quantities, have been measured in research laboratories for decades using gas chromatography with mass spectrometry (GC/MS), but these assays are too cumbersome and difficult for routine use in clinical laboratories (10). These metabolites, however, are not heavily protein-bound like cortisol, and many cross-react slightly with cortisol immunoassays. Consequently, as hypercortisolism develops, these integrated cortisol metabolites—if measured—increase before the UFC is convincingly abnormal (11).

This situation begs the question, “Which is the best analytic method to measure UFC?” Wood et al (12) showed that immunoassays have significant positive analytic bias compared to LC-MS/MS due to antibody cross-reactivity with steroid metabolites. In this study, it was clearly demonstrated that LC-MS/MS agreed extremely well with the “gold standard” GC/MS. However, two commercially available immunoassays led to as much as a 225-nmol/L overestimation of the concentration of urinary cortisol. Furthermore, the percentage positive bias of individual samples by immunoassay was not necessarily predictable based on the concentration of cortisol found by GC/MS. Therefore, LC-MS/MS is the best rapid method to measure cortisol, and only cortisol.

It is, however, possible that the cross-reactivity of UFC immunoassays with abundant cortisol metabolites actually improves diagnostic sensitivity compared to the highly specific LC-MS/MS method. Perhaps the best diagnostic approach is to measure cortisol metabolites instead. Even if this approach were feasible for widespread use, what parameters and cutoffs would we use in the absence of the extensive experience we have with UFC?

Given these limitations, is it time to consider the heretical suggestion to retire the venerable UFC? The sobering reality is that we do not have a perfect test to gauge the effects of glucocorticoid exposure on tissues (similar to prostate-specific antigen for androgens), much less to establish the presence or absence of mild hypercortisolism, so we will always have to make compromises among sensitivity, specificity, simplicity, and cost. Laboratory testing complements but does not replace the clinical evaluation.

Before we discard immunoassays for UFC and adopt modern LC-MS/MS methods because the latter is superior from an analytical point of view, it would seem prudent to directly compare the diagnostic sensitivity of the two techniques in a large population of patients with proven Cushing's syndrome of all forms, as well as a large group of patients suspected of but without pathological hypercortisolism. Perhaps we will find that even the newer, lower reference ranges for UFC by LC-MS/MS are too high and we need to further lower the cutoff value for an abnormal screening UFC. Ironically, had LC-MS/MS been the first method used for the measurement of LNSC, this test may not have achieved its widespread use because, in some studies, LC-MS/MS seems to have poorer diagnostic performance compared to less specific immunoassays (13, 14). In cases of hormone measurements, we suggest that editorial policy be guided by evidence, ideally in comparator studies, that one assay technique is superior for specific clinical applications and for publications.