Division of Endocrinology and Metabolism, Department of Medicine, Faculty of Medicine, Chiang Mai University Hospital, 110 Intrawarorot Road Soi 2, Si Phum, Amphoe Mueang Chiang Mai, Chiang Mai, 50200, Thailand.
Department of Orthopaedics, Faculty of Medicine, Chiang Mai University Hospital, Chiang Mai, Thailand.
BMC Endocr Disord. 2020 Feb 19;20(1):24. doi: 10.1186/s12902-020-0508-7.
Adrenal insufficiency (AI) in hospitalized patients is a fatal condition if left undiagnosed. Most patients may require an adrenocorticotropic hormone (ACTH) stimulation test to facilitate AI diagnosis. We aim to identify simple biochemical and clinical factors and derive a predictive model to help identify hospitalized patients with biochemical AI who have indeterminate 0800 h serum cortisol levels.
A seven-year retrospective study was performed in a tertiary care medical center. We identified 128 inpatients who had undergone low-dose or high-dose ACTH stimulation tests. The association between biochemical AI and other factors was evaluated using a logistic regression model clustering by ACTH dose. Stepwise regression analysis was used to demonstrate the predictive model. Diagnostic performance was evaluated using ROC analysis.
Of the 128 patients, 28.1% had biochemical AI. The factors associated with biochemical AI were serum random cortisol < 10 μg/dL (OR = 8.69, p < 0.001), cholesterol < 150 mg/dL (OR = 2.64, p = 0.003), sodium < 140 mmol/L (OR = 1.73, p = 0.004)). Among clinical factors, cirrhosis (OR = 9.05, p < 0.001), Cushingoid appearance in those with exogenous steroid use (OR = 8.56, p < 0.001), and chronic kidney disease (OR = 2.21, p < 0.001) were significantly linked to biochemical AI. The AUC-ROC of the final model incorporating all factors was 83%.
These easy-to-perform biochemical tests and easy-to-assess clinical factors could help predict biochemical AI in hospitalized patients with high accuracy. The physician should therefore have a high index of suspicion to perform dynamic tests for AI diagnosis in those who meet the proposed model criteria.
住院患者的肾上腺功能不全（AI）如果未得到诊断则可能致命。大多数患者可能需要进行促肾上腺皮质激素（ACTH）刺激试验以协助 AI 诊断。我们旨在确定简单的生化和临床因素，并得出预测模型以帮助识别生化 AI 但 0800h 血清皮质醇水平不确定的住院患者。
对一家三级护理医疗中心进行了一项为期 7 年的回顾性研究。我们确定了 128 名接受了低剂量或高剂量 ACTH 刺激试验的住院患者。使用逻辑回归模型，按 ACTH 剂量聚类评估生化 AI 与其他因素之间的关联。逐步回归分析用于展示预测模型。使用 ROC 分析评估诊断性能。
在 128 名患者中，有 28.1%存在生化 AI。与生化 AI 相关的因素包括血清随机皮质醇 <10μg/dL（OR=8.69，p<0.001）、胆固醇 <150mg/dL（OR=2.64，p=0.003）、钠 <140mmol/L（OR=1.73，p=0.004））。在临床因素中，肝硬化（OR=9.05，p<0.001）、外源性类固醇使用的库欣样外观（OR=8.56，p<0.001）和慢性肾脏病（OR=2.21，p<0.001）与生化 AI 显著相关。纳入所有因素的最终模型的 AUC-ROC 为 83%。
这些易于进行的生化检查和易于评估的临床因素可帮助高准确度预测住院患者的生化 AI。因此，医生应该高度怀疑那些符合所提出模型标准的患者存在 AI，并进行动态测试以进行 AI 诊断。
