用于评估急性胰腺炎患者结局的计算机断层扫描指数「SMART-CT」的建立与验证
这项回顾性研究纳入2017年6月至2019年3月间连续接受增强CT的急性胰腺炎患者,将123例CT随机分为训练组(n=103)和验证组(n=20),由两名放射科医生分析改良CT严重指数及13项探索性指标,筛选出7项有统计学意义的指标,结合SMOTE技术建立基于列线图的SMART-CT指数,用于预测死亡、ICU入住、住院时长等临床结局,AUC为0.64–0.83,验证组与训练组相当,结论为该列线图指数预测临床结局具有中等准确度。
要点
- SMART-CT指数由胰腺坏死、积液数目、积液大小、腹水、胸腔积液、腹腔动脉受累及肝脂肪变7项CT指标构成。
- 该指数对死亡、ICU入住、住院时长、ICU时长、再入院次数及严重度的AUC分别为0.79、0.66、0.75、0.83、0.70和0.64。
- 验证组AUC与训练组相当,提示该列线图指数预测临床结局具有中等准确度。
核心结论
- 基于列线图的新型指数预测临床结局的发生具有中等准确度。
研究设计与样本
- 回顾性研究,纳入2017年6月至2019年3月间连续接受增强CT的急性胰腺炎患者,共123例CT扫描随机分为训练组(n=103)和验证组(n=20)。
- 由两名放射科医生分析改良CT严重指数所描述的征象及13项额外探索性指标,采用二项逻辑回归建立预测算法并制作列线图。
- 采用合成少数类过采样技术(SMOTE)平衡少数类别的代表性,故该指数命名为「SMOTE Application for Reading CT in AcuTe Pancreatitis(SMART-CT指数)」。
关键影像表现/征象
- 筛选出7项有统计学意义的CT指标用于建立指数:胰腺坏死、积液数目、积液大小、腹水、胸腔积液、腹腔动脉受累及肝脂肪变。
诊断标准与关键数值
- SMART-CT指数对死亡的AUC为0.79(95%CI 0.65–0.93)。
- 对ICU入住的AUC为0.66(95%CI 0.54–0.77)。
- 对住院时长的AUC为0.75(95%CI 0.65–0.85)。
- 对ICU时长的AUC为0.83(95%CI 0.69–0.96)。
- 对再入院次数的AUC为0.70(95%CI 0.60–0.81),对严重度的AUC为0.64(95%CI 0.53–0.75)。
临床应用与阅片要点
- 阅片时可关注胰腺坏死、积液数目与大小、腹水、胸腔积液、腹腔动脉受累及肝脂肪变这7项指标,作为SMART-CT指数的构成要素。
- 该指数基于列线图预测死亡、ICU入住、住院时长、ICU时长、再入院次数及严重度,验证组AUC与训练组相当。
局限性
- 原文未报告局限性。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
英文原文(摘要)· 点击展开
PURPOSE: The existing CT indices do not allow quantitative prediction of clinical outcomes in acute pancreatitis (AP). The aim of this study was to develop and validate a revised CT index using a nomogram-based approach.
METHODS: This retrospective study comprised consecutive patients with AP who underwent contrast-enhanced CT between June 2017 and March 2019. 123 CT scans were randomly divided into training (n = 103) and validation groups (n = 20). Two radiologists analyzed CT scans for findings described in modified CT severity index and additional exploratory items (13 items). Seven items (pancreatic necrosis, number of collections, size of collections, ascites, pleural effusion, celiac artery involvement, and liver steatosis) found to be statistically significant were used for development of index. Synthetic minority oversampling technique (SMOTE) was employed to balance representation of minority classes and hence this index was named "SMOTE Application for Reading CT in AcuTe Pancreatitis (SMART-CT index)". Binomial logistic regression was used for development of prediction algorithm. Nomograms were then created and validated for each outcome.
RESULTS: The new CT index had area under the curve (AUC) of 0.79 [95% CI 0.65-0.93], 0.66 (95% CI 0.54-0.77), 0.75 (95% CI 0.65-0.85), 0.83 (95% CI 0.69-0.96), 0.70 (95% CI 0.60-0.81), and 0.64 (95% CI 0.53-0.75) for mortality, intensive care unit (ICU) stay, length of hospitalization, length of ICU stay, number of admissions, and severity, respectively. The AUC of validation cohort was comparable to the training cohort.
CONCLUSION: The novel nomogram-based index predicts occurrence of clinical outcome with moderate accuracy.
引用
- Gupta P, Kumar-M P, Verma M, et al. Development and validation of a computed tomography index for assessing outcomes in patients with acute pancreatitis: "SMART-CT" index. Abdom Radiol (NY). 2020. PMID: 32936420. doi:10.1007/s00261-020-02740-y
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。