O-RADS、GI-RADS、ADNEX模型诊断附件肿瘤良恶性的价值及一致性分析

Comparison of O-RADS, GI-RADS and ADNEX models in the diagnosis of adnexal masses: diagnostic performance, malignancy rate and inter-reviewer agreement

摘要:

目的 分析O-RADS、GI-RADS、ADNEX模型超声诊断分类系统鉴别附件肿瘤良恶性的价值及一致性。方法 回顾性分析2017年1月—2020年12月在常州市第二人民医院行手术治疗的299例患者的324个附件肿瘤，由3位经验丰富的超声医师分别根据O-RADS、GI-RADS、ADNEX模型进行分类，绘制受试者工作特征曲线，获得最佳分界值，以病理检查结果为金标准，分析3种超声分类系统诊断附件肿瘤的有效性。使用κ值评估各超声医师分类的一致性。结果 当取O-RADS >3类的cut-off值时，其诊断附件肿瘤良恶性的曲线下面积（AUC）为0.981，大于GI-RADS（0.934）和ADNEX模型（0.907）（P <0.05）；而GI-RADS与ADNEX模型的AUC比较，差异无统计学意义（P >0.05）。O-RADS诊断附件肿瘤良恶性的敏感性为96.0%，高于GI-RADS（90.0%）和ADNEX模型（88.0%）（P <0.05）；GI-RADS与ADNEX模型的敏感性比较，差异无统计学意义（P >0.05）。O-RADS、GI-RADS、ADNEX模型诊断附件肿瘤良恶性的特异性分别为92.9%、88.4%和93.3%，差异无统计学意义（P >0.05）。3位超声医师采用O-RADS、GI-RADS、ADNEX模型诊断附件肿瘤良恶性的一致性均较好，κ值为0.857～0.937。结论 O-RADS、GI-RADS、ADNEX模型均有较高的诊断价值，且一致性较好，但O-RADS的诊断效能优于GI-RADS和ADNEX模型。

Abstract:

Objective To evaluate the performance and inter-reviewer agreement of the O-RADS and two other well-established ultrasound (US) classification systems for determining the malignancy of adnexal masses (AM).Methods A total of 299 patients with 324 AM who underwent surgical treatments in Changzhou Second People's Hospital from January 2017 to December 2020 were enrolled in this retrospective analysis. Three experienced ultrasonographers independently categorized each AM according to Ovarian-Adnexal Imaging Reporting and Data System (O-RADS), Gynecologic Imaging-Reporting and Data System (GI-RADS), and Assessment of Different NEoplasias in the adneXa (ADNEX) models. The receiver operating characteristic (ROC) curves were plotted to determine the optimal cut-off values, and the efficacy of the three models for diagnosing AM was evaluated with pathological findings as the gold standard. The kappa statistics were used to assess the inter-reviewer agreement (IRA).Results When O-RADS 3 was set as the cut-off value, the area under the ROC curve (AUC) of O-RADS model was 0.981, which was greater than that of GI-RADS (0.934) and ADNEX (0.907) models (P < 0.05). However, there was no difference between the AUC of GI-RADS model and that of ADNEX model (P > 0.05). The specificity of GI-RADS model for diagnosing the malignancy of AM was 88.4%, which was lower than that of O-RADS (92.9%) and ADNEX (93.3%) models (P < 0.05). There was no difference between the specificity of O-RADS model and that of ADNEX model (P > 0.05). The IRA was high in all the three models, with kappa statistics ranged from 0.857 to 0.937.Conclusions The O-RADS, GI-RADS, and ADNEX models are all of high value for diagnosing the malignancy of AM with high IRA. Nevertheless, the diagnostic performance of O-RADS model is even greater than that of GI-RADS and ADNEX models.