The Reflex Dexamethasone Suppression Test: Development and Assessment of Reflexed Serum Dexamethasone Measurement for the Diagnosis of Cushing Syndrome

Posted on August 23, 2025 by MaryO

Abstract

Background

Screening for Cushing syndrome (CS; endogenous overproduction of ACTH or cortisol) is performed by the low-dose overnight serum dexamethasone suppression test (oDST) with the measurement of serum dexamethasone concentration to assure an effective dose.

Objective

We evaluated the utility of only measuring serum dexamethasone in samples with nonsuppressed serum cortisol using a conservative serum cortisol cutoff.

Methods

This retrospective study included 261 oDSTs completed before Reflex implementation (Pre-Reflex-oDST) and 281 oDSTs completed after (Post-Reflex-oDST). Serum cortisol and serum dexamethasone data were paired to the diagnosis and analyzed with comparative statistical tests and receiver operating characteristic curve (ROC) analysis.

Results

Endogenous hypercortisolism was diagnosed in 38 of 261 Pre-Reflex-oDSTs (14%) and 40 of 281 (14%) Post-Reflex-oDSTs. In oDSTs with SerCort >1.8 mcg/dL, there were 9% and 6% false positives in the Pre-vs Post-Reflex-oDST group, respectively. In the Pre-Reflex-oDST group, the median SerCort was 1.1 mcg/dL (95% CI: 0.8–1.5) in patients without CS and 3.9 mcg/dL (95% CI: 2.6–7.9) in those with CS (P < 0.001). The optimal ROC cutoff of SerCort in the Pre-Reflex-oDST group was 2.1 mcg/dL (sensitivity 92%, specificity 93%). In the Post-Reflex-oDST group, the median SerCort was 1.1 mcg/dL (95% CI: 0.8–1.5) in patients without CS and 2.9 mcg/dL (95% CI: 2.6–7.9) in those with CS (P < 0.001). The optimal ROC cutoff of SerCort in the Post-Reflex-oDST group was 2.1 mcg/dL (sensitivity 95%, specificity 93%; not different from Pre-Reflex-oDST group).

Conclusion

Reflex measurement of the serum dexamethasone did not affect oDST test performance while reducing costs.

Abbreviations

CS

Cushing syndrome
oDST

low-dose overnight serum dexamethasone suppression test
ROC

receiver operating characteristic
NH

neoplastic hypercortisolism
NNH

non-neoplastic hypercortisolism
HPA

hypothalamic pituitary adrenal
SerCort

serum cortisol
SerDex

serum dexamethasone
WDL

Wisconsin Diagnostic Laboratories
LOQ

limit of quantification
UFC

urine free cortisol

Highlights

  • •
    Reflexing only nonsuppressed serum cortisol samples for the measurement of serum dexamethasone does not negatively affect the performance of the overnight low-dose DST (oDST)
  • •
    Reflex implementation greatly reduced the number of serum dexamethasone measurements thereby decreasing unnecessary costs
  • •
    The oDST appeared to be valid as long as there was a measurable serum dexamethasone result (>50 ng/dL)

Clinical Relevance

We report a novel Reflex overnight dexamethasone suppression test (oDST) serum dexamethasone measurement protocol with the benefit of greatly lowering costs without loss of oDST performance supporting its implementation in screening for Cushing syndrome.

Introduction

Endogenous Cushing syndrome (CS) includes neoplastic hypercortisolism either due to autonomous cortisol production or excessive ACTH secretion.1 Non-neoplastic hypercortisolism (NNH) is also an important clinical entity that is characterized as bona fide cortisol excess caused by conditions such as depression, chronic kidney disease, and poorly controlled diabetes.2, 3, 4 Chronically elevated cortisol levels contribute to significant morbidity and mortality due to cardiovascular, metabolic, musculoskeletal, and immunologic effects, including hypertension, diabetes, osteoporosis, and increased susceptibility to infections5,6 leading to prolonged disease burden and worsening clinical outcomes.2,3
Biochemical evaluation is indicated for individuals presenting with features of hypercortisolism, as well as for those with adrenal incidentalomas, regardless of symptoms, in accordance with current guidelines.7 However, the workup is complicated by varying severities of hypercortisolism and diurnal rhythm of endogenous cortisol requiring screening tests to take advantage of predictable nadirs and negative feedback.1 One of the current first-line screening tests for CS is the 1 mg (low dose) overnight dexamethasone suppression test (oDST).8 Positive oDST results require additional testing such as late-night salivary cortisol and 24 h urine free cortisol measurements.8, 9, 10
The oDST takes advantage of decreased HPA negative feedback sensitivity in CS.11 An oral 1 mg dexamethasone dose is given between 2300 hours and 2400 hours and serum cortisol (SerCort) is then measured at 0800 h the following morning with levels <1.8 mcg/dL (50 nmol/L) representing normal suppression.8 A clinical sensitivity of 95% highlights the oDST as a useful screening tool; however, a specificity of ∼80% indicates a greater potential for false-positive results.10,12 A higher burden of false-positive results complicates diagnosis, leading to further testing, delays in diagnosis and treatment, and increased costs and resource utilization for both patients and the health care system. Therefore, the measurement of a serum dexamethasone (SerDex) in the next morning SerCort sample should identify insufficient SerDex levels that may result from factors such as mistiming of, or altogether missing the dexamethasone dose, differences in dexamethasone metabolism, variation in gastrointestinal absorption, increased cortisol binding globulin (eg, due to oral contraceptives), and medications that alter CYP3A4 activity.13, 14, 15, 16, 17
To reduce unnecessary and costly SerDex measurements, our institution implemented in 2023 a Reflex protocol in which SerDex is only measured if post-oDST SerCort is ≥ 1.6 mcg/dL (ie nonsuppressed). This approach was suggested but not evaluated by Genere et al.18 Since the purpose of this study was not to validate the concept of the oDST, we chose 1.6 mcg/dL as a conservative cutoff so that borderline SerCort with small variations around the accepted 1.8 mcg/dL cutoff would not bias the results. This study assessed equivalency in test performance between Pre- and Post-Reflex-oDST implementation and estimated the associated cost savings. As a secondary outcome, we evaluated therapeutic SerDex levels necessary for valid testing.

Materials and Methods

Study Design

A retrospective cohort study was performed on all oDSTs completed at Froedtert & the Medical College of Wisconsin and the affiliated Wisconsin Diagnostic Laboratories between May 2023 and April 2024. This study was approved by the Medical College of Wisconsin Institutional Review Board as a Quality Improvement Project under PRO00050802. All data in the database were de-identified and coded.

Study Population

The study population included patients aged ≥18 years who completed a oDST in the outpatient setting through the system with serum samples processed through Wisconsin Diagnostic Laboratories. Cohorts were divided into Pre- and Post-Reflex-oDST groups based on the date of Reflex implementation described below (October 31, 2023), with the Pre-Reflex-oDST group consisting of tests performed in the 6 months prior and the Post-Reflex-oDST group including tests from the 6 months following implementation. The ordering clinician still had the option to choose oDST AM cortisol alone without ordering a SerDex measurement. Exclusion criteria included repeat oDST for the same patient within the same 6-month period (Pre- or Post-Reflex-oDST groups); however, results from patients who underwent oDST during both the Pre- and Post-Reflex-oDST periods were included. Patients with SerDex <50 ng/dL were also excluded. At this time, all patients included in the analysis had either a confirmed hypercortisolism diagnosis or were determined not to have pathophysiological hypercortisolism based on clinical and biochemical evaluation.8 A combination of biochemical tests (24-hour free urine cortisol, late-night salivary cortisol, oDST), imaging (adrenal and pituitary CT/MRI), and other testing (DDVAP testing, surgical outcome, biopsy) were used to confirm the diagnosis of CS per the current guidelines.8 Hypercortisolism included both neoplastic causes (CS) and non-neoplastic causes (NNH).2 In our study, NNH included patients with chronic nausea and weight loss, chronic kidney disease, poorly controlled diabetes, excess alcohol intake, obesity with physiological stress, chronic pain, and opioid withdrawal. For simplicity, all hypercortisolism patients are abbreviated “CS” whether neoplastic or non-neoplastic.

Procedures

Patients were instructed to take 1 mg of dexamethasone at 11:00 pm and then had their blood sampled the next morning between 8:00 and 9:00 am SerCort level was measured using the Roche Elecsys Cortisol II Electrochemiluminescence immunoassay performed on a Cobas e801 module.19 In the Pre-Reflex-oDST group, individual orders for serum cortisol and dexamethasone were placed by the provider. Both tests were performed, regardless of the subsequently reported oDST serum cortisol value. In the Post-Reflex-oDST group, a single “Dexamethasone Suppression Cortisol Reflex” order was placed, that prompted a SerCort measurement. If SerCort was ≥1.6 mcg/dL (ie, nonsuppressed), an electronic order for a SerDex send-out measurement was automatically placed by the laboratory information system. The laboratory’s automated processing line removed the serum sample from storage, created an aliquot and placed it in a queue for samples to be sent to ARUP Laboratories. ARUP Laboratories measured SerDex by liquid chromatography-mass spectrometry (limit of quantitation [LOQ] = 50 ng/dL; reference interval: 140 – 295 ng/dL; https://ltd.aruplab.com/Tests/Pub/2003248).

Data Collection

Data acquired from each oDST, including date and time of collection, SerCort, and SerDex, were extracted from the laboratory information system. Additional data, including demographics, CS diagnosis, and treatment were collected from the electronic health record and securely stored in RedCAP (version 15.0.2; Nashville, TN). Patients with inconclusive test results were followed for several months after oDST until a diagnosis was established through additional testing and/or clinical evaluation.

Outcomes Assessment

The primary outcome of this study was to quantify the number of SerDex tests avoided while assessing whether implementation of the Reflex oDST affected overall test performance in the screening for endogenous hypercortisolism. Equivalency between Pre- and Post-Reflex-oDSTs was defined by the similarities in prevalence of a new diagnosis, average SerCort by diagnosis, and SerCort cutoffs by receiver operating characteristic (ROC) curve AUC, and optimal cutoff values. Secondary outcomes included quantification of avoided SerDex measurements to estimate the cost savings associated with Reflex implementation using the US Medicare Reimbursement Rate [20 Accessed 4/1/2025; CPT code 80 299)]. We also evaluated suppressed oDST’s SerDex concentrations to determine if there was a correlation between SerDex level achieved and the degree of suppression of SerCort.

Statistical Analysis

All statistical analyses were performed using Sigmaplot 15.0 (RRID:SCR_003210; https://scicrunch.org/resolver/SCR_003210; Systat Software, Inc, Inpixon, Palo Alto, CA). Continuous variables that were not normally distributed are presented as the median and interquartile ranges. Demographic data were analyzed by two-way analysis of variance and chi-square. Differences in SerCort and SerDex between Pre- and Post-Reflex-oDST groups and CS diagnosis were tested by Mann-Whitney U test and t-test. Optimal cutoff values for each group were determined by ROC analysis using Youden’s index and AUCs were compared with a DeLong test. A P value <0.05 was considered statistically significant. Post-hoc power analyses results are provided where appropriate.

Results

Study Population Characteristics

A total of 616 oDSTs (308 in the Pre-Reflex-oDST and 308 in the Post-Reflex-oDST groups) were screened for eligibility (Fig. 1). After excluding oDSTs with SerDex below the LOQ (n = 12) and repeated oDSTs in the same patient (n = 62), a total of 542 oDSTs were included for analysis. Demographic data are presented in Table 1. The without CS group was younger than the patients with CS group in both Pre- and Post-Reflex-oDST groups. There were no differences in the distribution of race between the groups with and without CS groups and the Pre-vs Post-Reflex-oDST groups. There were more females regardless of diagnosis, but the sex distribution was not different Pre vs Post-Reflex-oDST implementation.

  1. Download: Download high-res image (445KB)
  2. Download: Download full-size image

Figure 1. Flowchart of participants selection from 616 completed oDSTs completed 6 months before (n = 308) and after (n = 308) Reflex implementation. Subsequent oDSTs for the same patient and unmeasurable post-oDST serum dexamethasone (SerDex) (<50 ng/dL [Lower quantifiable limit]) were excluded from analysis. A total of 542 oDSTs were included for analysis and breakdown of CS diagnosis and etiology are shown. ACTH-dependent CS is further broken down to differentiate neoplastic (NH) versus non-neoplastic (NNH) etiologies. CS = Cushing Syndrome; NH = neoplastic hypercortisolism; NNH = nonneoplastic hypercortisolism; oDST = overnight dexamethasone suppression test; SerDex = serum dexamethasone.

Table 1. Demographic Characteristics of Patients Who Underwent oDST Before and After Reflex Implementation

Empty Cell Pre-Reflex Post-Reflex
Yes CS with NNH Yes CS without NNH No CS Yes CS with NNH Yes CS without NNH No CS
N 38 34 223 40 38 241
Age
 Mean (SD) 63.6 (13.8) 63.8 (14.4) 56.0 (15.1)a 63.8 (13.2) 63.1 (13.1) 55.3 (15.5)b
Sex
 Male (%) 6 (15.8) 4 (11.8) 57 (25.6) 11 (27.5) 11 (28.9) 67 (27.8)
 Female (%) 32 (84.2) 30 (88.2) 166 (74.4) 29 (72.5) 27 (71.1) 174 (72.2)
Race
 American Indian or Alaskan Native (%) 0 0 1 (0.4) 0 0 1 (0.4)
 Asian (%) 0 0 3 (1.3) 0 0 1 (0.4)
 Black or African American (%) 5 (13.2) 5 (14.7) 27 (12.2) 6 (15.0) 6 (15.8) 23 (9.5)
 Other (%) 2 (5.3) 1 (2.9) 8 (3.6) 1 (2.5) 1 (2.2) 9 (3.8)
 White (%) 31 (81.5) 28 (82.4) 184 (82.5) 33 (82.5) 31 (82.0) 207 (85.9)
Data are further stratified by Cushing syndrome (CS) diagnosis. Age is presented as mean (SD); sex and race as counts (percentages).
a
Age different from group with CS within Pre-Reflex-oDST (P = 0.005).
b
Age different from group with CS within Post-Reflex-oDST (P < 0.001) regardless of whether NNH cases are included. Male vs female distribution NS (χ2 = 2.533, 3 df, P = 0.469). Race distribution NS (χ2 = 4.37733, 12 df, P = 0.976).
Within the Pre-Reflex-oDST group, 38 patients (14%) were diagnosed with endogenous hypercortisolism (CS) with 13 being ACTH-dependent (9 pituitary and 4 non-neoplastic hypercortisolism) and 25 being ACTH-independent. Within the Post-Reflex-oDST group, 40 patients (14%) were diagnosed with CS with 11 being ACTH-dependent (8 pituitary, 2 non-neoplastic, and 1 ectopic ACTH) and 25 being ACTH-independent.

Prereflex-oDST vs Post-reflex-oDST Analysis

In the Pre-Reflex-oDST group, out of the 261 included subjects, 172 oDSTs (65%) suppressed to <1.6 mcg/dL, meaning that only 89 tests would have undergone reflex SerDex measurements after implementation of Reflex testing. Among these, 51 were determined not to have CS. In the Post-Reflex-oDST group, out of 281 subjects, 191 oDSTs (68%) suppressed to <1.6 mcg/dL, resulting in 90 reflexed SerDex measurements, of which 50 did not have CS. Within the Pre-Reflex-oDST group, there were 38 patients who had CS and in the Post-Reflex-oDST group 40 patients were found to have CS. There was no difference in CS prevalence between the Pre- and Post-Reflex-oDST groups (P = 0.52). Among oDSTs with SerCort levels >1.8 mcg/dL [the conventional cutoff 8], 24/262 (9% false positive) in the Pre-Reflex-oDST group and 21/281 (7% false positive) in the Post-Reflex-oDST group were later determined not to have CS by standard guidelines criteria.8 Reflex implementation resulted in a reduction of the number of SerDex measurements by 68% resulting in cost savings of at least $18.64 per ordered oDST.
In the Pre-Reflex-oDST group, the median SerCort was 1.1 mcg/dL (95% CI: 0.8–1.5) in patients who did not have CS and 3.9 mcg/dL (95% CI: 2.6–7.9) in those who had CS (P < 0.001). In the Post-Reflex-oDST group, the median SerCort was also 1.1 mcg/dL (95% CI: 0.8–1.5) in patients who did not have CS and 2.9 mcg/dL (95% CI: 2.6–7.9) in those who had CS (P < 0.001) (Fig. 2). There was no difference comparing the CS diagnosis status between Pre- and Post-Reflex-oDST groups. There was still no difference in median SerCort in patients with CS when comparing the Pre- and Post-Reflex-oDST groups with NNH cases excluded (P = 0.269). Furthermore, the NNH patients were biochemically indistinguishable from patients with neoplastic hypercortisolism (NH). In fact, when we compared post-oDST cortisol between NH (3.6 [2.4-7.7; N = 72) and NNH (3.4 [2.7-8.5; N = 6]), the P value was 0.729. There was also no difference between CS NH with NNH included vs CS NH excluding NH within the Pre-Reflex-oDST group and within the Post-Reflex-oDST group.

  1. Download: Download high-res image (195KB)
  2. Download: Download full-size image

Figure 2. Comparison of oDST serum cortisol (SerCort) levels Pre-vs Post-Reflex-oDST implementation. The medians are further stratified based on whether the patient did not have Cushing Syndrome (No CS – red) and those who had CS (Yes CS – blue). Each box represents the interquartile range and the horizontal line within represents the median. The error bars represent the 10th-90th percentiles and dots represent results outlying the 10th-90th percentiles. a, denotes significant difference of median SerCort levels between no CS vs CS in both Pre- and Post-Reflex-oDST groups (P < 0.001). There was no difference in medians following exclusion of NNH from Yes CS in both Pre- and Post-Reflex-oDST groups (P = 0.269). CS = Cushing Syndrome; NH = neoplastic hypercortisolism; NNH = nonneoplastic hypercortisolism; oDST = overnight dexamethasone suppression test; SerDex = serum dexamethasone.

A power analysis was performed to compare Pre-reflex oDST to Post-Reflex oDST SerCort values in patients with CS with the null hypothesis that there was no effect of implementing the reflex approach. Assuming a clinically significant effect of implementing the reflex test (a difference of oDST serum cortisol of 2 mcg/dL with a conservative SD of the difference of 2.5 mcg/dL) with sample sizes of 38 Pre-Reflex and 40 Post-Reflex in our study, the power was 0.937 with an alpha of 0.050. With an expected difference in serum cortisol of 0.8 mcg/dL and an SD of the difference of 1.0 mcg/dL, the power was 0.937 with an alpha of 0.05. Considering that the patients before and after reflex implementation were from the same population, institution, and ordering clinicians, we are confident that the Reflex testing did not influence the oDST performance and the laboratory data outcomes.
There was also no difference in SerCort between Pre- and Post-Reflex-oDST tests in the predictive performance for CS. The ROC curve AUC of SerCort in both the Pre- and Post-Reflex periods was 0.97. The optimal ROC cutoff of SerCort in the Pre-Reflex-oDST group was 2.1 mcg/dL (sensitivity 92%, specificity 93%). The optimal ROC cutoff of SerCort in the Post-Reflex-oDST group was 2.1 mcg/dL (sensitivity 95%, specificity 93%; not different from Pre-Reflex-oDST group) (Table 2). When NNH cases were excluded and ROC curves were rerun, there was no difference in ROC curve area, optimal SerCort cutoff values, or sensitivity and specificity in Pre- and Post-Reflex-oDST groups.

Table 2. Receiver operating characteristic (ROC) analysis of oDST SerCort results for Pre-vs Post-Reflex-oDST groups. A. Analysis including NNH patients are at the top; B. Analysis excluding NNH patients are at the bottom

Empty Cell Pre-Reflex Post-Reflex
A. Including NNH patients
 ROC Curve Area (SE) 0.97 (0.01) 0.97 (0.01)
 95% confidence interval 0.96-0.99 0.95-0.99
 P value P < 0.0001 P < 0.0001
 Sample size: No CS/Yes CS 223/38 241/40
Cutoff Sensitivity Specificity Cutoff Sensitivity Specificity
 Optimal 8 AM SerCort Cutoff (mcg/dL) 2.1 92% 93% 2.1 95% 93%
Empty Cell Pre-Reflex without NNH Post-Reflex without NNH
B. Excluding NNH patients
 ROC curve area (SE) 0.97 (0.01) 0.97 (0.01)
 95% confidence interval 0.96-0.99 0.95-0.99
 P Value P < 0.0001 P < 0.0001
 Sample size: No CS/Yes CS 223/34 241/38
Cutoff Sensitivity Specificity Cutoff Sensitivity Specificity
 Optimal 8 AM SerCort Cutoff (mcg/dL) 2.1 91% 92% 2.1 95% 93%
Area under the curve (AUC) was calculated and compared with a DeLong test (AUC = 0.97, P < 0.0001, for both). Using Youden’s Index, optimal cutoff values were determined by maximizing sensitivity and specificity. When ROC rerun without NNH, the sensitivity and specificity did not change in both Pre- and Post-oDST-Reflex groups.

Prereflex-oDST Comparison of SerDex vs SerCort

In comparing the Pre-Reflex-oDST group SerDex results of <140 ng/dL versus >140 ng/dL (the lower reference limit of the SerDex assay), median SerCort was 1.2 mcg/dL and 1.1 mcg/dL, respectively (P = 0.621) (Fig. 3A). The scatter regression plot illustrates that there was no relationship between SerDex (ng/dL) and SerCort (mcg/dL) by CS diagnosis (Fig. 3B). Each point represents an individual oDST, with red indicating patients who did not have CS (n = 223) and blue indicating those who had CS (n = 38). In patients who did not have CS, SerDex ranged from 61.5 to 908.9 ng/dL whereas in patients who had CS, SerDex ranged from 96.3 to 646.0 ng/dL. Theoretical linear regression lines are shown. In fact, no significant correlation between SerDex and SerCort was found in the group who did not have CS (r = 0.002; P = 0.972) nor the group who had CS (r = 0.114; P = 0.494) so the regression lines are only provided for visual clarity. When NNH cases were excluded, there was still no correlation between SerDex and SerCort in patients with CS (P = 0.432). Furthermore, analysis of only NNH cases also showed no correlation between SerDex and SerCort (P = 0.871).

  1. Download: Download high-res image (397KB)
  2. Download: Download full-size image

Figure 3. Comparison of post-oDST serum cortisol (SerCort) to serum dexamethasone (SerDex) in Pre-Reflex-oDST group. (A) Comparison of post-oDST SerCort and SerDex for no CS patients in the Pre-Reflex-oDST group. SerCort in the No CS patients stratified by the ARUP Lower limit of the reference range for SerDex (140 ng/dL). There was no significant difference in median SerCort with the SerDex <140 ng/dL (N = 20) and >140 ng/dL (N = 203) groups (1.2 vs 1.1 mcg/dL, respectively, P = 0.621). (B) Comparison of all Pre-Reflex-oDST group oDSTs stratified by patients with (blue) and without (red) CS. The black vertical solid line represents the limit of quantitation (LOQ) of SerDex (50 ng/dL). There was no correlation of SerDex and SerCort achieved in either group (see text for specifics). There was no correlation when NNH cases were removed as well (P = 0.432). CS = Cushing Syndrome; NH = neoplastic hypercortisolism; NNH = nonneoplastic hypercortisolism; oDST = overnight dexamethasone suppression test; SerDex = serum dexamethasone.

Discussion

The purpose of this study was to assess the usefulness of implementing a protocol to only reflex samples for the measurement of SerDex that do not suppress post-oDST SerCort (the “Reflex-oDST”) using a very conservative 8 AM cortisol cutoff. The major findings were as follows: (a) There was no detrimental effect on oDST-suppressed SerCort levels with the implementation of Reflex testing. That is, the SerCort levels in patients without CS were not different from each other Pre- and Post-Reflex-oDST. The same was found in the group who had CS. (b) There were comparable optimal SerCort ROC cutoff values in the Pre- and Post-Reflex-oDST groups, which also demonstrated a lack of a detrimental effect on test performance. (c) The Reflex protocol eliminated the need for SerDex measurement in 68% of oDSTs ordered without reducing the accuracy of the oDST. (d) No correlation was found between SerCort and SerDex indicating that the SerDex concentration achieved may not be an important factor in assessing the accuracy of the oDST; rather the presence of a detectible SerDex may be sufficient (ie, a Boolean function).
When comparing the Pre- and Post-Reflex-oDST groups, we observed no difference in patient population characteristics, including the prevalence of positive oDSTs, CS diagnosis, and CS etiology. Notably, patients who had CS were older than those who did not have CS in both groups. Ueland et al demonstrated a positive correlation between age and oDST SerCort, which may partially explain the significantly higher age in patients with CS.17 Age-related changes in HPA axis dynamics likely contributed to an increased prevalence of unsuppressed results.21 Additionally, older patients often have more comorbidities, which may lead clinicians to have a lower threshold for further evaluation, increasing the likelihood of identifying CS in this population.
To demonstrate that implementation of oDST Reflex protocol did not negatively affect diagnostic performance, we compared SerCort levels and false positive rates defined by unsuppressed SerCort later determined not to have CS through further testing. We found no difference in SerCort levels before and after implementation of Reflex testing (within the patients who did not have CS and within the patients who had CS). We observed a 7% false positive rate in the Pre-Reflex-oDST group that was comparable to 10% to 14% in previous studies.17,22 We found a comparable 9% false positive rate in the Post-Reflex-oDST group demonstrating preservation of test performance with Reflex implementation. These data are similar to those theorized by Genere et al18 and we have now validated the approach.
Implementation of reflex SerDex testing reduced the number of oDST SerDex measurements by ∼68% and resulted in a cost savings of $18.64 per ordered oDST using Medicare Clinical Diagnostic Laboratory Test reimbursement data, though this likely underestimates the true financial burden as patients are often billed at higher rates.20 Therefore, the Reflex protocol not only did not have a detrimental effect on test performance but also improved efficiency by reducing unnecessary costs and resource utilization. It is also important to point out that the approach is likely to save additional costs like avoiding additional analysis such as unnecessary plasma ACTH, salivary cortisols, UFCs, and even MRIs.
For the purposes of diagnosis of CS, we used accepted oDST SerCort cutoff of ≤1.8 mcg/dL (50 nmol/L) that yields a sensitivity of 95% and a specificity of ∼80%.10,12 We, like most clinicians, utilize a ≤1.8 mcg/dL SerCort diagnostic cutoff with concomitant measurement of SerDex to improve specificity by reducing false positives due to subtherapeutic dexamethasone levels. For the current study, we used the cutoff of 1.6 mcg/dL to determine when to reflex samples for SerDex in the Post-Reflex-oDST group to be as conservative as possible particularly when considering biological variability around a value of 1.8 mcg/dL.
To demonstrate equivalency, we found SerCort cutoff values that maximize specificity based on oDSTs performed similarly in both Pre- and Post-Reflex-oDST groups. Our calculated cutoff value for SerCort Pre- and Post-Reflex-oDST validated that a small increase in SerCort oDST cutoffs results in an increase in specificity to >95% (utilizing Youden’s index). However, given that the oDST is a screening test, sensitivity should be prioritized. While higher specificity reduces false positives, it may also increase the risk of missing mild CS cases. Notably, it has been shown that concomitant SerDex measurement reduces false positives by 20%,17 reinforcing the benefit of its inclusion.
To be conservative, we reanalyzed all of the data excluding the 6 patients with NNH. This had no effect on any of the outcomes. In fact, the oDST 8AM cortisol was almost identical when comparing NH to NNH patients further emphasizing the clinical challenge of distinguishing these highly overlapping groups in terms of their laboratory results.
Similar to previous studies, we found a broad range of SerDex levels across all oDSTs with minimal or no correlation to SerCort.9,23,24 To maximize specificity, Ueland et al proposed a SerDex cutoff value of 130 ng/dL [0.130 mcg/dL (3.3 nmol/L)], while Ceccato et al proposed a SerDex cutoff of 180 ng/dL [0.180 mcg/dL (4.5 nmol/L)] prioritizing specificity.17,25 SerCort levels the morning after taking 1 mg of dexamethasone (8 AM–9 AM) reflects delayed glucocorticoid negative feedback at the pituitary and hypothalamus that takes at least 1-2 h to be fully expressed.26 Therefore, differences in measured SerDex at 8 AM the morning after the 1 mg dose ingestion reflect the variability of HPA axis feedback sensitivity and the timing of the pharmacokinetics of dexamethasone metabolism that can be influenced by several factors, such as age, BMI, and concomitant medications.11,13,27,28 Our findings suggest that any detectible SerDex level (>50 ng/dL in our study), even if below the established reference interval (eg,140–295 ng/dL), is sufficient for a valid test and probably does not require repetition. That said, it may still be prudent to repeat the test if the SerCort does not suppress to <1.8 mcg/dL and the SerDex is < 100 ng/dL particularly with a high index of suspicion for CS.24 oDSTs with SerDex values below the laboratory’s detectable limits (LOQ) should still be considered invalid, most likely due to dexamethasone noncompliance or differences in absorption and/or pharmacokinetics as described above. It is also important to point out that the LOQ for some serum dexamethasone assays are higher than the assay we used, which makes this point even more important.18
An important final point is the practicability of the approach. Why not just store the serum cortisol sample and only test it for serum dexamethasone if requested18? This is very challenging for clinicians in practice who work with a variety of reference laboratories and are often not aware of the SerCort oDST results until after the sample has been discarded. By building the reflex approach into the ordering system, this problem is avoided as it does not require the intervention of the ordering clinician, thereby reducing the administrative burden while reducing laboratory costs. The significance of our novel study is confirmed by the fact that, subsequent to our implementation, a major reference laboratory has recently established a similar Reflex oDST test (https://www.labcorp.com/tests/503990/cortisol-dexamethasone-suppression-test-with-reflex-to-dexamethasone). Others are likely to follow. Our study, which is the first of its kind to our knowledge, should give the clinician assurance that this approach is appropriate. At our relatively small laboratory, this results in annual cost savings of $11,500 per year just for the dexamethasone levels not needed. The savings for each institutional and provider would obviously be different depending on their patient mix and test volumes.

Conclusion

We demonstrated that the implementation of the Reflex protocol avoided unnecessary SerDex measurements without affecting test performance, highlighting its utility from both a resource and cost standpoint. Additionally, our findings suggest that any quantifiable SerDex level, even if below the established reference interval, does not invalidate the oDST.

Disclosure

Dr Carroll is an Editorial Board Member of this journal and was not involved in the editorial review or the decision to publish this article. Dr Nerenz receives research funding from Abbott Laboratories.

Acknowledgment

The authors thank the personnel at Wisconsin Diagnostic Laboratories for their work to develop and implement the Reflex Testing protocol. J.D.K. is a recipient of the 2024 Research Experience for Graduate and Medical Students (REGMS) award from the Endocrine Society (US).

References

Filed under: Cushing's, Diagnostic Testing | Tagged: , , , , | Leave a comment »