评估4种恶性风险指数（RMI）对鉴别卵巢肿瘤良恶性的诊断价值。

回顾性分析2017年10月至2018年10月在常州市第二人民医院接受治疗的200例卵巢肿瘤患者的术前癌抗原125（CA125）水平、绝经状态、超声分数、肿瘤大小以及术后病理等资料，采用不同的计算方法（RMI1、RMI2、RMI3、RMI4）计算卵巢肿瘤的RMI，并进行良恶性判定，结果与病理对照，评价在不同病理组织类型和临床分期的诊断价值，通过ROC曲线比较4种RMI与单项指标诊断效能的高低，以及各项的阳性预测值、阴性预测值、特异度和敏感度。采用Cochran's Q检验比较RMI1、RMI2、RMI3、RMI4四种方法的异同，Dunn's检验（经Bonferroni法校正）进行4种指数的两两比较。

RMI1、RMI2、RMI3、RMI4的ROC曲线下面积分别为0.859、0.872、0.866、0.878。根据ROC曲线分析，RMI1、RMI2、RMI3的界值为100，RMI4的界值为200时，其敏感度分别为77.5%、82.6%、80.0%、82.5%，特异度分别为94.4%、91.9%、93.1%、93.1%，阳性预测值分别为77.5%、71.7%、74.4%、75.0%，高于原始界值的阳性预测值。4种RMI诊断效能比较差异具有统计学意义（χ²=8.333，P=0.04），进行两两比较，发现RMI1与RMI2两种方法之间差异具有统计学意义（P=0.028）。RMI对卵巢恶性上皮性肿瘤的诊断率较高（71%、77%、71%、74%），其中RMI2的诊断率比其他3种RMI高；RMI对卵巢非上皮性肿瘤的诊断率较低（均为40%）。在卵巢恶性肿瘤中，对I期病变及交界性肿瘤的诊断率低。

4种RMI对卵巢肿瘤的良恶性诊断均有较高的特异度和阳性预测值，但对非上皮性肿瘤、交界性肿瘤、卵巢肿瘤早期患者敏感度较低，仍需进一步改良。

To assess the value of four kinds of risk of malignancy index (RMI) in the differential diagnosis of benign and malignant ovarian tumors.

We retrospectively analyzed the preoperative CA125 levels, menopausal status, ultrasound score, tumor size, and postoperative pathological data of 200 patients with ovarian tumors who were treated at Changzhou Second People's Hospital from October 2017 to October 2018. We used different calculation methods (RMI1, RMI2, RMI3, and RMI4) to calculate the RMI for distinguishing benign and malignant ovarian tumors, and the results were compared with pathology to evaluate their diagnostic value among different pathological tissue types and clinical stages. ROC curve analysis was performed to compare their diagnostic power with individual indicators and calculate their positive predictive value, negative predictive value, specificity, and sensitivity. The Cochran's Q test was used to compare the similarities and differences of the four methods (RMI1, RMI2, RMI3, and RMI4), and the Dunn's test (corrected by Bonferroni method) was used for pairwise comparison of the four methods.

The areas under the ROC curves of RMI1, RMI2, RMI3, and RMI4 were 0.859, 0.872, 0.866, 0.878, respectively. According to the ROC curve analysis, when the cutoff values of RMI1, RMI2, and RMI3 were all 100, and that of RMI4 was 200, the sensitivities were 77.5%, 82.6%, 80.0%, and 82.5%, and the specificities were 94.4%, 91.9%, 93.1%, and 93.1%, respectively. The positive predictive values were 77.5%, 71.7%, 74.4%, and 75.0%, respectively, which were higher than those of the original threshold. There were significant differences among the four RMI methods (χ²=8.333, P=0.04). After pairwise comparison, it was found that only RMI1 and RMI2 had a significant difference (P=0.028). RMI had a higher diagnostic rate for malignant epithelial ovarian tumors (71%, 77%, 71%, and 74%), and RMI2 had a higher diagnostic rate than the other three RMIs. However, RMI had a lower diagnostic rate for non-epithelial ovarian tumors (40% each). Among ovarian malignant tumors, the diagnosis rate for stage I lesions and borderline tumors was low.

The four RMIs have high specificity and positive predictive value for the diagnosis of benign and malignant ovarian tumors, but the sensitivity to non-epithelial tumors, borderline tumors, and early-stage ovarian tumors is low, and further improvement is needed.