Comparative evaluation of diagnostic performance of the most commonly used screening tests for pathological hypercortisolism: A single centre analysis

Agathoklis Efthymiadis; Helen Loo; Brian Shine; Jeremy Tomlinson; Aparna Pal; Riccardo Pofi

doi:10.1530/endoabs.93.OC22

OC22

Prev Next

Section Contents Cite

Endocrine Abstracts (2023) 93 OC22 | DOI: 10.1530/endoabs.93.OC22

EYES2023 ESE Young Endocrinologists and Scientists (EYES) 2023 Oral communication 4: Pituitary and Neuroendocrinology (8 abstracts)

Comparative evaluation of diagnostic performance of the most commonly used screening tests for pathological hypercortisolism: A single centre analysis

Agathoklis Efthymiadis ¹ , Helen Loo ² , Brian Shine ³ , Jeremy Tomlinson ⁴ , Aparna Pal ² & Riccardo Pofi ⁵

Author affiliations

¹Oxford Centre for Diabetes, Endocrinology and Metabolism, Oxford University Hospitals NHS Trust, Churchill Hospital, Oxford, Endocrinology, Oxford, UK; ²Oxford Centre for Diabetes, Endocrinology and Metabolism, Oxford University Hospitals NHS Trust, Churchill Hospital, Oxford, UK; ³Department of Clinical Biochemistry, Oxford University Hospitals NHS Foundation Trust, Oxford, UK, UK; ⁴Oxford Centre for Diabetes, Endocrinology and Metabolism, Nihr Oxford Biomedical Research Centre, University of Oxford, Churchill Hospital, Oxford, UK., Oxford, UK; ⁵Oxford Centre for Diabetes, Endocrinology and Metabolism, Nihr Oxford Biomedical Research Centre, University of Oxford, Churchill Hospital, Oxford, UK.

Background: To date, there is no consensus as to the gold-standard screening test for diagnosing Cushing’s Syndrome (CS).

Objectives: This study aimed to: a) compare the ability of late-night salivary cortisol (LNSC) against overnight dexamethasone suppression test (ONDT) and urinary free cortisol (UFC) as screening test for pathological hypercortisolism (PH); b) test the performance of those tests in diagnosing Cushing’s disease (CD) or mild autonomous cortisol secretion (MACS); c) refine the screening algorithm by adding clinical symptoms.

Methods: We retrospectively reviewed all consecutive adult patients referred to the Oxford Centre for Diabetes, Endocrinology and Metabolism for evaluation of PH who had LNSC measured from January 2017 to November 2022. A binomial logistic regression (LR) was performed to ascertain the ability of each test in diagnosing PH, and compute receiver-operating-characteristic curve analysis. A stepwise backward LR was run to assess the utility of symptoms to predict PH.

Results: LNSC had the best sensitivity, 100.0% (95%CI 80.5–100.0), specificity, 64.9%, (95%CI 47.5–79.8) in distinguishing CD from the absence of PH (AUC 0.82, 95%CI 0.72–0.93, P<0.001). ODST proved to be the best test in differentiating between MACS and absence of PH (AUC of 0.76, 95%CI 0.66–1.00, P=0.004) with sensitivity of 100.0% (95%CI 82.4–100.0), specificity of 52.2% (95%CI 30.6–73.2). UFC did not reach statistical significance in diagnosing PH (P=0.26). Combining AUCs of pre-test signs, symptoms (hypertension, interscapular fat, facial plethora, striae myopathy, easy bruising) with those of screening tests significantly improved diagnostic performance of LNSC (AUC of 0.83, 95%CI 0.74–0.92, P<0.001), ONDST (0.82, 95% CI 0.71–0.93, P<0.001).

Conclusions: When diagnosing CD, LNSC performed better than ONDST, whereas the opposite was true for patients with MACS. UFCs had the lowest diagnostic accuracy across all PH subgroups. Assessing pre-test clinical probability through the presence of specific symptoms suggesting PH significantly improved the diagnostic accuracy of screening tests.