医学
接收机工作特性
逻辑回归
优势比
内科学
曲线下面积
死亡率
统计
C反应蛋白
预测值
置信区间
缺少数据
病因学
聚类分析
急性胰腺炎
子群分析
可能性
作者
Samir Vanani,Mipasha Patel,Dhvani Adhvaryu,N. Nandwani,Atul K. Desai,Dhiraaj Desai,Rajiv Mehta
标识
DOI:10.14309/01.ajg.0001126552.93065.34
摘要
Introduction: C-reactive protein (CRP) is commonly used in acute pancreatitis (AP) to assess severity, but the value of early serial CRP trends for mortality prediction is unclear. We evaluated whether changes over the first 4-5 days can improve risk stratification and guide clinical decision-making. Methods: We ambispectively analyzed 372 AP patients (median age: 39 [IQR 30-54]; 80% men) with serial CRP values recorded from days 2 to 5 of symptom onset (CRP2-CRP5). Missing values were imputed using constrained interpolation or iterative regression. Post-imputation analyses showed consistent trends and predictive performance compared to complete-case data. K-means clustering identified CRP trajectory phenotypes. Logistic regression, receiver operating characteristic analysis, and decision curve analysis assessed mortality prediction and net clinical benefit. Subgroup analysis was performed by CRP2 level and BISAP score. Results: Overall mortality was 6.2% (n = 23). Mean CRP declined from 250 mg/L (day 2) to 134 mg/L (day 5). Trajectory clustering revealed 2 phenotypes, resolving CRP (mortality 2.2%) and persistently elevated CRP (mortality 12.4%; odds ratio [OR] = 6.29). A model using CRP2-CRP4 yielded strong discrimination for mortality (AUC = 0.809; 95% CI: 0.689–0.911). Overall, a CRP4/CRP2 ratio ≥0.90 predicted mortality with AUC of 0.64. In patients with CRP2 >300 mg/L (n = 134), predictive performance improved (AUC = 0.820, NPV = 93.9%), highlighting its utility in high-baseline inflammation states. Predictive trends remained consistent across etiologies and robust across sensitivity analyses. Decision curve analysis showed that adding CRP5 to the model of CRP2-4 provided minimal incremental benefit (AUC gain of +0.002), with a maximum net benefit of 0.0064 at a threshold probability of 0.27, suggesting most prognostic information is captured by Day 4. Early CRP trends can inform ICU triage, escalation of care, and early discharge planning, supporting more efficient resource allocation. Conclusion: CRP trajectories through Day 4 define meaningful phenotypes in AP. The CRP4/CRP2 ratio provides modest overall discrimination but is highly predictive in patients with elevated baseline CRP. The limited added value of day 5 CRP supports more efficient care pathways, especially in settings with constrained resources. Early CRP trends offer a practical, scalable tool for personalized and resource-conscious AP care.
科研通智能强力驱动
Strongly Powered by AbleSci AI