利用CT预测胃癌上皮-间质转化分子亚型
本研究回顾性分析451例胃癌根治性切除并接受分子亚组分析的患者,基于年龄、Lauren分型和CT壁分层构建上皮-间质转化(EMT)亚型预测模型并绘制列线图,模型AUC为0.865,bootstrap验证AUC为0.860,最佳截断值94.622对应敏感度90.9%、特异度67.2%、准确率71.8%。
要点
- CT壁分层是预测胃癌EMT亚型的独立影像变量,与年龄、Lauren分型共同构建模型。
- 列线图总 prognostic 分≥94.622预测EMT亚型,敏感度90.9%、特异度67.2%。
- 模型AUC 0.865,bootstrap内部验证AUC 0.860,具有较高诊断准确率。
核心结论
- 基于患者年龄、Lauren分型和CT壁分层的预测模型可用于个体化估计胃癌EMT分子亚型,列线图可作为有用的筛查工具。
研究设计与样本
- 回顾性研究,纳入451例接受胃癌原发灶切除并完成分子亚组分析的患者,采用逐步变量选择的多变量分析构建预测模型并绘制列线图,通过bootstrap重采样进行内部验证。
关键影像表现/征象
- CT壁分层(mural stratification)是预测EMT亚型的入选影像变量之一。
- EMT亚型88例,非EMT亚型363例,基于转录组分析确定。
诊断标准与关键数值
- 预测模型纳入年龄、Lauren分型和CT壁分层,AUC为0.865。
- bootstrap验证AUC为0.860。
- 列线图总预后评分最佳截断值为94.622,敏感度90.9%、特异度67.2%、准确率71.8%。
临床应用与阅片要点
- 原文结论认为列线图可作为个体化估计EMT亚型的有用筛查工具。
- 原文指出将CT表现加入临床病理变量可提高预测模型的准确率,优于仅使用临床病理变量。
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
OBJECTIVES: To develop a prediction model with computed tomography (CT) images and to build a nomogram incorporating known clinicopathologic variables for individualized estimation of epithelial-to-mesenchymal transition (EMT) subtype gastric cancer.
METHODS: Patients who underwent primary resection of gastric cancer (GC) and molecular subgroup analysis (n = 451) were reviewed. Multivariable analysis using a stepwise variable selection method was performed to build a predictive model for EMT subtype GC. A nomogram using the results of the multivariable analysis was constructed. An optimal cutoff value of total prognostic points of the nomogram for the prediction of EMT subtype was determined. The predictive model for the EMT subtype was internally validated by bootstrap resampling method.
RESULTS: There were 88 patients with EMT subtype and 363 patients with non-EMT subtype based on transcriptome analysis. The patient's age, Lauren classification, and mural stratification on CT were variables selected for the predictive model. The area under the curve (AUC) of the model was 0.865, and the validated AUC of the bootstrap sample was 0.860. The optimal cutoff value of total prognostic points for the prediction of EMT subtype was 94.622, with 90.9% sensitivity, 67.2% specificity, and 71.8% accuracy.
CONCLUSION: A predictive model using patient's age, Lauren classification, and mural stratification on CT for EMT molecular subtype GC was made. A nomogram was built which would serve as a useful screening tool for an individualized estimate of EMT subtype.
KEY POINTS: • A predictive model for epithelial-to-mesenchymal transition (EMT) subtype incorporating patient's age, Lauren classification, and mural stratification on CT was built. • The predictive model had high diagnostic accuracy (area under the curve (AUC) = 0.865) and was validated (bootstrap AUC = 0.860). • Adding CT findings to clinicopathologic variables increases the accuracy of the predictive model than using only.
引用
- Cha DI, Lee J, Jeong WK, et al. Prediction of epithelial-to-mesenchymal transition molecular subtype using CT in gastric cancer. Eur Radiol. 2021. PMID: 34120231. doi:10.1007/s00330-021-08094-3
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。