Purpose: Late-night salivary cortisol (LNSC) is reportedly highly accurate for the diagnosis of Cushing's Syndrome (CS). However, diagnostic thresholds for abnormal results are based on healthy, young populations and limited data are available on its use in elderly populations and among those with chronic medical conditions The purpose of this study was to evaluate LNSC levels among a population of elderly patients with and without diabetes.

Methods: We prospectively evaluated outpatient LNSC levels (obtained at 2300h) from 154 male veterans with type 2 diabetes and 52 male veterans without diabetes. Participants with LNSC levels in excess of the published threshold (4.3nmol/l or above) underwent secondary testing with 24-hour urine free cortisol (24UFC) and overnight dexamethasone suppression testing (DST) to rule out CS. Participants with positive secondary testing underwent ACTH testing and imaging. We used the Kolmogorov-Smirnov test to evaluate the distribution of LNSC results and the Mann-Whitney U test to compare LNSC levels between diabetic and non-diabetic participants. We performed logistic regression to assess those clinical parameters associated with abnormal LNSC results.

Results: 141 diabetics and 46 controls (mean age, 61) returned samples (91% overall). LNSC levels were not normally distributed, but were skewed to the right. Average LNSC levels (nmol/l) were significantly higher for diabetics than for non-diabetics (median [interquartile range: 2.6[1.8 – 4.1] vs 1.6[1.0 - 2.0]), and for those aged 60 or above compared to those aged < 60 (2.7[2.0 - 4.3] vs 1.9[1.4 – 2.9]) (p<0.001 for each). Thirty-one participants required secondary testing. 79% of participants who underwent secondary testing had normal 24UFC and DST. No cases of CS have been diagnosed to date. Increasing age (OR, 1.9 per decade), a current diagnosis of type 2 diabetes mellitus (OR, 3.7), and elevated blood pressure (OR, 1.3 per 10 mmHg increase in systolic blood pressure) were associated with abnormal LNSC results (p < 0.05 for each), while body mass index, history of psychiatric illness, and history of substance use/abuse were not. At the current diagnostic threshold of 4.3 nmol/l, only 58% of elderly, hypertensive diabetics had normal LNSC levels; at a threshold of 10 nmol/l, this increased to approximately 95%

Conclusions: As LNSC becomes more widely used for the diagnosis of CS, the development of age- and comorbidity-adjusted thresholds may be warranted.