Low positive predictive value of midnight salivary cortisol measurement to detect hypercortisolism in type 2 diabetes.

Background: Hypercortisolism is prevalent in type 2 diabetes (T2D), but analytical and functional uncertainties prevail. Measurement of salivary cortisol is considered an expedient screening method for hypercortisolism, but its usefulness in the context of T2D is uncertain.

Aim: To compare late-night salivary cortisol (LNSC) with the 1 mg overnight dexamethasone suppression test (DST), which was considered 'reference standard', in T2D.

Patients And Methods: A total of 382 unselected and recently diagnosed patients with T2D underwent assessment of LNSC and DST, and the test outcome was related to age, gender, body mass index (BMI) and haemoglobin A1c levels. We used the following cut-off values: LNSC ≤ 3·6 nmol/l and DST ≤ 50 nmol/l.

Results: The median (range) levels of LNSC and DST were 6·1 (0·3-46·2) nmol/l and 34 (11-547) nmol/l, respectively. Hypercortisolism was present in 86% based on LNSC values and 22% based on DST. LNSC, as compared to DST, had the following test characteristics: sensitivity: 85% (95% CI: 7-92%), specificity: 14% (95% CI: 10-19%), positive predictive value: 22% (95% CI: 17-27%), negative predictive value: 76% (95% CI: 63-87%), and overall accuracy: 30% (95% CI: 25-34%). LNSC and DST values were not associated with haemoglobin A1c, BMI and age in this cohort of patients with T2D.

Conclusion: The LNSC is characterized by very low specificity and poor positive predictive value as compared to the DST, resulting in an overall low accuracy. Further methodological and clinical studies are needed to substantiate the relevance of cortisol status in T2D.