对术前评估附件肿块良恶性的常用超声诊断模型的诊断效能进行比较分析。

收集2017年10月至2021年9月在解放军总医院第一医学中心超声诊断科经阴道或经直肠超声检查诊断为附件肿块并行手术治疗的患者资料，采用3种常用的附件肿块良恶性鉴别的超声模型，即简单法则（SR）、逻辑回归模型2（LR2）和卵巢-附件影像报告和数据系统（O-RADS），分别对附件肿块超声特征进行评估，判断肿块良恶性，以病理结果作为金标准，绘制受试者工作特征（ROC）曲线，计算并比较3种方法的诊断效能：采用DeLong方法比较ROC曲线AUC的差异；采用配对χ²检验比较敏感度、特异度的差异，采用χ²检验及Fisher精确检验比较阳性预测值和阴性预测值的差异。

本研究纳入151例肿块，其中良性肿块106例（70.20%），恶性肿块45例（29.80%）。SR、LR2和O-RADS准确诊断附件肿块良恶性的AUC分别为0.901（0.842~0.944）、0.940（0.890~0.972）和0.926（0.872~0.963），SR的AUC低于LR2，差异具有统计学意义（P=0.034），SR与O-RADS、LR2与O-RADS之间的AUC比较差异均无统计学意义（P均＞0.05）。SR的敏感度为100%，特异度为80.19%，阳性预测值为68.18%，阴性预测值为100%；LR2诊断恶性肿块的最佳阈值为＞0.125，此时敏感度为95.56%，特异度为85.85%，阳性预测值为74.14%，阴性预测值为97.85%；O-RADS诊断恶性肿块的最佳阈值为＞3类，此时敏感度为100%，特异度为78.30%，阳性预测值为66.18%，阴性预测值为100%。O-RADS、LR2和SR 3种超声模型的敏感度、阳性预测值、阴性预测值比较，差异均无统计学意义（P均＞0.05）。SR与LR2、SR与O-RADS特异度比较，差异均无统计学意义（P均＞0.05），O-RADS的特异度低于LR2，差异具有统计学意义（P=0.021）。O-RADS误诊的肿块中，有3例肿块后方伴有声影，这3例肿块SR全部诊断正确。本研究中26例有后方声影的附件肿块，病理结果全部为良性病变。3种超声模型均未能正确评估良恶性的附件肿块有12例，其中3例经超声造影检查诊断为良性，与病理结果相符。另有6例（炎症4例，子宫内膜异位囊肿破裂2例）患者，均有典型临床表现及化验结果，结合临床资料，可避免误诊。

SR、LR2和O-RADS这3种超声模型对于诊断附件恶性肿块具有相似且较高的敏感度，但O-RADS的诊断特异度较低。肿块后方声影作为良性指标纳入O-RADS超声模型中有可能提高其判断肿块良恶性的准确性。超声造影检查可作为鉴别附件肿块良恶性的补充方法。结合病史、临床症状及化验指标有助于附件肿块良恶性的诊断。

To compare the value of different models in the diagnosis of the adnexal masses.

We retrospectively collected patients with adnexal masses detected by transvaginal or transrectal ultrasound at the department of ultrasound of First Center in Chinese PLA General Hospital from October 2017 to September 2021. Simple rules (SR), Logistic regression model 2 (LR2), and ovarian-adnexal reporting and data system (O-RADS) were used to evaluate the ultrasound features of the cases, and the pathological results were used as the reference standards. Receiver operating characteristic (ROC) curve analysis was performed to compare the diagnostic efficacy of the three models. The area under the ROC curve (AUC) values were compared by the DeLong method. The paired chi-square test was used to compare sensitivities and specificities. The chi-square test and Fisher's exact test were used to compare positive predictive values (PPV) and negative predictive values (NPV).

Among the 151 masses, 106 (70.20%) were benign and 45 (29.80%) were malignant according to the pathological results. The AUC values of SR, LR2, and O-RADS were 0.901 (0.842-0.944), 0.940 (0.890-0.972), and 0.926 (0.872-0.963), respectively. The AUC of SR was significantly lower than that of LR2 (P=0.034), but there was no significant difference between SR and O-RADS, or between LR2 and O-RADS (P＞0.05 for both). The sensitivity, specificity, PPV, and NPV of SR were 100%, 80.19%, 68.18%, and 100%, respectively. The cutoff for LR2 was＞0.125, and the sensitivity, specificity, PPV and NPV were 95.56%, 85.85%, 74.14%, and 97.85%, respectively. The cutoff for O-RADS was＞3, and the sensitivity, specificity, PPV, and NPV were 100%, 78.30%, 66.18%, and 100%, respectively. There was no significant difference in sensitivity, PPV, or NPV between O-RADS, LR, and SR. No significant difference was detected in specificity between SR and LR2, SR and O-RADS (P＞0.05). The specificity of O-RADS was significantly lower than that of LR2 (P=0.021). There were three adnexal masses that showed acoustic shadows among masses misdiagnosed by O-RADS, all of which were diagnosed correctly by SR. All of the 26 masses that presented with acoustic shadows in our study were pathologically confirmed to be benign. There were 12 adnexal masses misdiagnosed by all the three models, three of which were diagnosed to be benign masses by contrast enhanced ultrasound and confirmed by pathology. Another six patients (4 patients with inflammation and 2 patients with rupture of endometriosis masses) showed typical clinical symptoms and laboratory examination results.

SR, LR, and O-RADS have similar sensitivity, but the specificity of O-RADS is the lowest. Using acoustic shadows of adnexal masses as one of the principles of O-RADS to recognize benign masses may improve the accuracy of that model. Contrast enhanced ultrasound, clinical symptoms, and laboratory examination can help to differentiate benign and malignant adnexal masses.