From the University of Michigan, 1500 E Medical Center Dr, Room B2 A209A, Ann Arbor, MI 48109-5030.
Radiology. 2022 Feb;302(2):357-366. doi: 10.1148/radiol.2021210853. Epub 2021 Nov 2.
Background The Bosniak classification system for cystic renal masses was updated in 2019 in part to improve agreement compared with the 2005 version. Purpose To compare and investigate interrater agreement of Bosniak version 2019 and Bosniak version 2005 at CT and MRI. Materials and Methods In this retrospective single-center study, a blinded eight-reader assessment was performed in which 195 renal masses prospectively considered Bosniak IIF-IV (95 at CT, 100 at MRI, from 2006 to 2019 with version 2005) were re-evaluated with Bosniak versions 2019 and 2005. Radiologists (four faculty members, four residents) who were blinded to the initial clinical reading and histopathologic findings assessed all feature components and reported the overall Bosniak class for each system independently. Agreement was assessed with Gwet agreement coefficients. Uni- and multivariable linear regression models were developed to identify predictors of dispersion in the final Bosniak class assignment that could inform system refinement. Results A total of 185 patients were included (mean age, 63 years ± 13 [standard deviation]; 118 men). Overall interrater agreement was similar between Bosniak version 2019 and version 2005 (Gwet agreement coefficient: 0.51 [95% CI: 0.45, 0.57] vs 0.46 [95% CI: 0.42, 0.51]). This was true for experts (0.54 vs 0.49) and novices (0.50 vs 0.47) and at CT (0.56 vs 0.51) and MRI (0.52 vs 0.43). Nine percent of masses prospectively considered cystic using Bosniak version 2005 criteria were considered solid using version 2019 criteria. In general, masses were more commonly classified in lower categories when radiologists used Bosniak version 2019 criteria compared with version 2005 criteria. The sole predictor of dispersion in Bosniak version 2019 class assignment was dispersion in septa or wall quality (ie, smooth vs irregular thickening vs nodule; 72% [MRI] and 60% [CT] overall model variance explained; multivariable < .001). Conclusion Overall interrater agreement was similar between Bosniak version 2019 and version 2005; disagreements in septa or wall quality were common and strongly predictive of variation in Bosniak class assignment. © RSNA, 2021 . See also the editorial by Eberhardt in this issue.
背景 囊性肾肿块的 Bosniak 分类系统于 2019 年进行了更新，部分目的是提高与 2005 年版本相比的一致性。目的 比较 CT 和 MRI 中使用 2019 年版和 2005 年版 Bosniak 分类的观察者间一致性。材料与方法 在这项回顾性单中心研究中，进行了八位观察者的盲法评估，前瞻性地考虑了 195 个 Bosniak IIF-IV 级（95 个 CT，100 个 MRI，来自 2006 年至 2019 年，使用 2005 年版本）的肾肿块，使用 2019 年和 2005 年版本重新评估 Bosniak 分类。八位观察者（四位教员，四位住院医师）对初始临床阅读和组织病理学结果均不了解，评估了所有特征成分，并分别报告了每个系统的总体 Bosniak 分类。使用 Gwet 一致性系数评估一致性。建立了单变量和多变量线性回归模型，以确定可能影响最终 Bosniak 分类分配的离散性的预测因素，从而为系统改进提供信息。结果 共纳入 185 例患者（平均年龄，63 岁±13[标准差]；118 例男性）。总体而言，2019 年版和 2005 年版 Bosniak 的观察者间一致性相似（Gwet 一致性系数：0.51[95%CI：0.45，0.57] vs 0.46[95%CI：0.42，0.51]）。专家（0.54 对 0.49）和新手（0.50 对 0.47）以及 CT（0.56 对 0.51）和 MRI（0.52 对 0.43）的结果均如此。根据 Bosniak 2005 年标准前瞻性地认为囊性的 9%的肿块，根据 2019 年标准认为是实性的。通常情况下，与使用 Bosniak 2005 年标准相比，使用 Bosniak 2019 年标准时，肿块更常被归入较低的类别。Bosniak 2019 年分类分配中离散性的唯一预测因子是隔室或壁质量的离散性（即，平滑对不规则增厚对结节；72%[MRI]和 60%[CT]总模型方差解释；多变量<.001）。结论 2019 年版和 2005 年版 Bosniak 之间的观察者间总体一致性相似；隔室或壁质量的差异很常见，并且是 Bosniak 分类分配变化的强烈预测因子。©RSNA，2021。也请参见本期 Eberhardt 的社论。
