Clinical and biochemical data for the diagnosis of endogenous hypercortisolism: the “Cushingomic” approach

Background: The recommended first-line screening tests for Cushing’s syndrome (CS) are cortisol after dexamethasone suppression test (DS), 24h urinary free cortisol (UFC), and late night salivary cortisol (LNSC). We collected them and evaluated in combination with the clinical presentation of suspected CS.
Methods: We analysed 1-mg DST (with dexamethasone levels), UFC and LNSC (with tandem-mass spectrometry) in patients without CS (263 suspected CS, 319 adrenal and 33 pituitary incidentalomas) and 40 confirmed CS. Non-parametric multivariate methods (principal component analysis, K-means clustering, random forest, and supervised learning algorithm) were used to compute an integrated analysis among screening tests (1-mg DST, UFC, LNSC), cortisol-related comorbidities and signs/symptoms of CS.
Findings: The three tests were able to individuate CS, cortisol after DST and UFC were slightly superior to LNSC. The threshold of 1-mg DST should be adapted to the population considered, especially in adrenal incidentaloma with mild autonomous cortisol secretion, the results of UFC and LNSC were independent of the group or high-risk condition considered. Some cortisol-related comorbidities were more common in patients without CS (diabetes, hypertension, and obesity), their correlation with screening tests was poor: the direction of their vectors was not aligned. A neural network model that combined screening tests and clinical presentation was able to predict the CS diagnosis in the validation cohort with 99% sensitivity, 86% specificity, 99% precision and 86% accuracy.
Interpretation: Screening tests for CS performed adequately. The presence of cortisol-related comorbidities and mild autonomous cortisol secretion in adrenal incidentalomas should be interpreted carefully.

背景：库欣综合征（CS）推荐的一线筛查检测包括地塞米松抑制试验（DS）后的皮质醇水平、24小时尿游离皮质醇（UFC）以及夜间唾液皮质醇（LNSC）。本研究收集并评估了这些指标，并结合疑似CS的临床表现进行综合分析。
方法：我们对无CS的患者（263例疑似CS，319例肾上腺意外瘤，33例垂体意外瘤）和40例确诊CS患者的1毫克DST（地塞米松水平）、UFC和LNSC（串联质谱法）进行了分析。采用非参数多元方法（主成分分析、K-means聚类、随机森林和监督学习算法）对筛查检测（1毫克DST、UFC、LNSC）、皮质醇相关合并症和CS的征兆/症状进行综合分析。
发现：三项检测均能区分CS，地塞米松抑制试验后的皮质醇和UFC略优于LNSC。1毫克DST的阈值应根据考虑的群体进行调整，特别是在肾上腺意外瘤伴轻度自主皮质醇分泌的情况下。UFC和LNSC的结果不受考虑的群体或高风险条件的影响。一些皮质醇相关合并症在无CS的患者中更为常见（如糖尿病、高血压和肥胖），它们与筛查检测的相关性较差：其向量方向不协调。一个结合筛查检测和临床表现的神经网络模型在验证队列中能够以99%的敏感性、86%的特异性、99%的准确率和86%的精确率预测CS的诊断。
解释：CS的筛查检测表现良好。肾上腺意外瘤中皮质醇相关合并症和轻度自主皮质醇分泌的存在应谨慎解读。