CEUS Bosniak分级对囊性肾肿块的恶性比例、诊断效能与观察者间可靠性:系统综述与Meta分析
IIF、III、IV级恶性比例分别为5.5%、58.7%、93.7%,合并敏感度97.9%、特异度62.6%。CEUS常导致分级重分类,多较CT或MRI上调,且多数报告未明确采用EFSUMB 2020标准。
要点
- CEUS Bosniak IV级病理证实恶性比例高,III级异质性强,IIF级校准不确定。
- CEUS对恶性囊性肾肿块合并敏感度97.9%、特异度62.6%,但仅基于8个数据集。
- CEUS常致Bosniak分级重分类,多较CT或MRI上调,报告时需注意分类框架差异。
核心结论
- 病理证实的病灶中Bosniak IV级恶性比例高,III级异质,IIF级恶性比例校准不确定。
- CEUS Bosniak分级合并敏感度97.9%、特异度62.6%,但受验证偏倚与谱偏倚限制。
- CEUS常导致分级重分类,多较CT或MRI上调,多数报告未明确使用EFSUMB 2020标准。
研究设计与样本
- 采用Freeman-Tukey/DerSimonian-Laird模型合并病理证实恶性比例,修订时加二项-逻辑广义线性混合模型。
- 双变量模型仅8个信息性数据集,参考标准为病理证实。
诊断标准与关键数值
- CEUS Bosniak IIF级恶性比例5.5%(95%CI 0.0%-25.3%;6个队列,72个病灶)。
- CEUS Bosniak III级恶性比例58.7%(95%CI 41.5%-75.0%;13个队列,199个病灶)。
- CEUS Bosniak IV级恶性比例93.7%(95%CI 90.0%-96.7%;13个队列,313个病灶)。
- GLMM估计IIF级3.3%(0.2%-42.1%)、III级58.9%(42.7%-73.3%)、IV级91.1%(87.2%-93.8%)。
- 合并敏感度97.9%(95%CI 90.7%-99.6%)、特异度62.6%(95%CI 48.2%-75.1%)。
临床应用与阅片要点
- 报告CEUS Bosniak分级时需注明所用分类框架,因多数研究未明确采用EFSUMB 2020标准。
- IIF级恶性比例受手术验证偏倚和模型选择影响,不宜解读为未选择监测人群的癌风险。
- CEUS常致分级重分类且多较CT或MRI上调,联合多模态影像可减少分期偏差。
局限性
- IIF级估计对手术验证偏倚和模型选择高度敏感。
- 双变量模型仅8个信息性数据集,且存在近边界相关、验证与谱效应。
- 读者可靠性因设计和方法学指标异质,范围从中等到几乎完全一致。
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
We evaluated malignancy proportions, diagnostic performance, and reader reliability of contrast-enhanced ultrasound (CEUS)-assigned Bosniak categories across heterogeneous classification frameworks. Five databases were searched through 5 August 2026. Pathology-confirmed proportions were pooled using Freeman-Tukey/DerSimonian-Laird models, with binomial-logit generalized linear mixed models (GLMMs) added at revision. Thirty-eight reports were included. Original-model malignancy proportions were 5.5% (95% CI, 0.0%-25.3%) for IIF (6 cohorts; 72 lesions), 58.7% (41.5%-75.0%) for III (13; 199), and 93.7% (90.0%-96.7%) for IV (13; 313). Corresponding GLMM estimates were 3.3% (0.2%-42.1%), 58.9% (42.7%-73.3%), and 91.1% (87.2%-93.8%). The IIF estimate was highly sensitive to surgical verification bias and model choice; it should not be interpreted as the cancer risk of an unselected surveillance population. Only eight informative datasets contributed to the bivariate model, yielding a summary sensitivity of 97.9% (90.7%-99.6%) and specificity of 62.6% (48.2%-75.1%). A near-boundary correlation and verification and spectrum effects limit these estimates. Reported reliability ranged from moderate to almost perfect across heterogeneous reader designs and agreement metrics and generally improved with reader experience. CEUS frequently resulted in category reclassification, most often upgrading relative to computed tomography (CT) or magnetic resonance imaging (MRI). Most reports did not explicitly use EFSUMB 2020 criteria. Category IV had a high malignancy proportion among verified lesions, whereas III was heterogeneous and IIF calibration remained uncertain. These results describe the studied populations and verification pathways and do not establish population-level risks or validate EFSUMB 2020 as a single standardized system.
引用
- Ge M, Song Y, Wang X, et al. Malignancy Proportions, Diagnostic Performance, and Inter-Observer Reliability of CEUS-Assigned Bosniak Categories for Cystic Renal Masses: A Systematic Review and Meta-Analysis. Ultrasound Med Biol. 2026. PMID: 42963292. doi:10.1016/j.ultrasmedbio.2026.10.005
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。