脓毒症并发危重病相关皮质类固醇不足的危险因素及列线图预测模型构建
该病例对照研究纳入河南省人民医院2018年1月至2023年12月脓毒症并发危重病相关皮质类固醇不足(CIRCI)患者50例为CIRCI组,同期随机选取脓毒症无CIRCI患者150例为对照组。多因素Logistic回归和IPTW回归分析显示,2型糖尿病史、PCT升高、血钾升高、APACHEⅡ评分升高和SOFA评分升高是CIRCI的独立危险因素,据此构建列线图预测模型。模型预测CIRCI发生的AUC为0.965(95%CI 0.942-0.989),最佳截断值0.201,敏感度92.0%、特异度87.3%;校准曲线C-index为0.966,Dxy为0.933。作者认为该模型可用于CIRCI的早期识别和干预,有利于改善脓毒症患者预后。
要点
- 脓毒症患者中,2型糖尿病史(OR=1.574,95%CI 0.558-4.437,P=0.022)、PCT升高(OR=4.271,95%CI 1.637-11.139,P=0.003)、血钾升高(OR=2.115,95%CI 0.909-4.921,P=0.044)、APACHEⅡ评分升高(OR=1.258,95%CI 1.071-1.478,P=0.006)和SOFA评分升高(OR=1.236,
- 基于上述5项指标构建的列线图模型预测脓毒症患者发生CIRCI的AUC为0.965(95%CI 0.942-0.989),最佳截断值0.201,敏感度92.0%、特异度87.3%。
- CIRCI与更长的ICU住院时间、更长的机械通气时间以及更高的28天死亡率和ICU死亡率相关(均P<0.05)。
- 单因素分析显示,年龄较大、WBC升高、APACHEⅡ评分升高和SOFA评分升高与CIRCI风险增加相关,而血钠升高与CIRCI风险降低相关(均P<0.05)。
- 决策曲线分析显示,在0.05-0.50的预设阈值概率范围内模型提供正向净获益,阈值概率超过事件率(约0.25)时模型优于全治和全不治策略。
适用人群与场景
- 2018年1月至2023年12月河南省人民医院收治的脓毒症患者,其中并发CIRCI者50例为CIRCI组,同期随机选取脓毒症无CIRCI者150例为对照组。
诊断标准与关键阈值
- 列线图模型预测CIRCI发生的最佳截断值为0.201,对应敏感度92.0%、特异度87.3%。
研究设计与关键结果
- 研究类型为病例对照研究,CIRCI组50例,对照组150例。
- 多因素Logistic回归和IPTW回归分析显示,2型糖尿病史(OR=1.574,95%CI 0.558-4.437,P=0.022)、PCT升高(OR=4.271,95%CI 1.637-11.139,P=0.003)、血钾升高(OR=2.115,95%CI 0.909-4.921,P=0.044)、APACHEⅡ评分升高(OR=1.258,95%CI 1.071-1.478,P=0.006)和SOFA评分升高(OR=1.236,95%CI 1.049-1.456,P=0.012)是CIRCI的独立危险因素。
- 基于上述5项指标构建的列线图模型预测CIRCI发生的AUC为0.965(95%CI 0.942-0.989),最佳截断值0.201,敏感度92.0%、特异度87.3%。
- 校准曲线显示训练队列C-index为0.966,Dxy统计量为0.933,C-index标准误为0.028。
- 决策曲线分析显示,在0.05-0.50的预设阈值概率范围内模型提供正向净获益(基于1000次Bootstrap重采样的95%CI不跨越零),阈值概率超过事件率(约0.25)时模型优于全治和全不治策略。
- CIRCI与更长的ICU住院时间、更长的机械通气时间以及更高的28天死亡率和ICU死亡率相关(均P<0.05)。
证据类型、等级与局限
- 证据类型为病例对照研究。
- 局限:原文未报告局限性。
对床旁实践的改变
- 作者认为,2型糖尿病史、PCT升高、血钾升高、APACHEⅡ评分升高和SOFA评分升高可能是脓毒症患者CIRCI的独立危险因素。
- 作者认为,所构建的列线图预测模型具有较高的准确性和有效性,可用于CIRCI的早期识别和干预,有利于改善脓毒症患者的预后。
提炼说明
- 本条依据论文摘要整理;摘要未涉及「处置流程与药物剂量」「红旗征与不能漏的情况」「指南或共识出处」,故未列出这些栏目;全文请点「阅读原文」。
英文原文(摘要)· 点击展开
OBJECTIVE: To identify the risk factors of critical illness-related corticosteroid insufficiency (CIRCI) in patients with sepsis and to construct a nomogram model for predicting the occurrence of CIRCI in septic patients.
METHODS: A case-control study was conducted. A total of 50 patients with sepsis complicated by CIRCI who were admitted to Henan Provincial People's Hospital from January 2018 to December 2023 were enrolled as the CIRCI group. Meanwhile, 150 patients with sepsis but without CIRCI admitted during the same period were randomly selected as the control group. Baseline characteristics, etiologies of sepsis, vital signs, adrenal function indicators, serological parameters, intervention measures, disease severity scores, and prognosis-related outcomes were collected and compared between the two groups. Multivariate Logistic regression and inverse probability of treatment weighting (IPTW) regression analyses were performed to evaluate the correlation between various factors and the occurrence of CIRCI, screen out the influencing factors of CIRCI in septic patients, and construct a nomogram prediction model. The receiver operator characteristic curve (ROC curve) and calibration curve were used to assess the discrimination of the model. The Bootstrap method (with 1 000 repeated samplings) was adopted for internal validation, and the calibration curve was plotted to evaluate the calibration of the model. Decision curve analysis (DCA) was performed to assess the clinical validity of the model.
RESULTS: After adjusting for confounding factors, univariate Logistic regression analysis showed that older age, history of type 2 diabetes mellitus, higher white blood cell count (WBC), higher procalcitonin (PCT) level, higher serum K+ level, higher Acute Physiology and Chronic Health Evaluation II (APACHE II), and higher Sequential Organ Failure Assessment (SOFA) were associated with an increased risk of CIRCI, whereas higher serum sodium (Na+) level was associated with a decreased risk of CIRCI (all P<0.05). In addition, CIRCI was associated with longer ICU length of stay, longer duration of mechanical ventilation, as well as higher rates of 28-day mortality and ICU mortality (all P<0.05). Multivariate Logistic regression and IPTW regression analyses revealed that history of type 2 diabetes mellitus [odds ratio (OR)=1.574, 95% confidence interval (95%CI) was 0.558-4.437, P=0.022], elevated PCT level (OR=4.271, 95%CI was 1.637-11.139, P=0.003), elevated serum K+ level (OR=2.115, 95%CI was 0.909-4.921, P=0.044), elevated APACHE II score (OR=1.258, 95%CI was 1.071-1.478, P=0.006), and elevated SOFA score (OR=1.236, 95%CI was 1.049-1.456, P=0.012) were independent risk factors for CIRCI. A nomogram prediction model was constructed based on the above 5 indicators. The ROC curve demonstrated that the model achieved an area under the curve (AUC) of 0.965 (95%CI was 0.942-0.989) for predicting the occurrence of CIRCI. The optimal cut-off value was 0.201, yielding a sensitivity of 92.0% and a specificity of 87.3%, indicating excellent discriminative performance. The calibration curve revealed a C-index of 0.966 and a Dxy statistic of 0.933 in the training cohort, reflecting very high predictive accuracy and strong discriminative ability. The standard error of the C-index was 0.028, supporting the reliability of this estimate. Furthermore, the calibration curve showed excellent agreement with the ideal reference line. DCA curve indicated that the model provided positive net benefit across a predefined threshold probability range of 0.05-0.50 (with the 95%CI based on 1 000 Bootstrap resamples not crossing zero). Notably, at threshold probabilities exceeding the event rate (approximately 0.25), the model outperformed both the treat-all and treat-none strategies.
CONCLUSIONS: History of type 2 diabetes mellitus, elevated PCT, elevated K+, elevated APACHE II score and elevated SOFA score may be independent risk factors for CIRCI in septic patients. The constructed nomogram prediction model has high accuracy and effectiveness, which can be used for early identification and intervention of CIRCI, and is conducive to improving the prognosis of sepsis patients.
参考来源
- Zhao E, Hu C, Zhang T, et al. [Risk factors for critical illness-related corticosteroid insufficiency in sepsis and construction of nomogram prediction model]. Zhonghua Wei Zhong Bing Ji Jiu Yi Xue. 2026. PMID: 41958126. doi:10.3760/cma.j.cn121430-20250616-00587
供执业医师抢救与诊疗参考,不替代临床判断;剂量请结合体重、肝肾功能和本院药品规格复核,以指南原文为准。