To the best of our knowledge, this is the first study comparing the clinical characteristics and sonographic appearance between benign and malignant ovarian BTs, and validating the optimal models that offer excellent discrimination between them. The majority of patients were postmenopausal and almost all tumors are unilateral. Most patients with malignant tumors had elevated CA125 serum levels. BBTs were most often small well-defined hypoechoic solid tumors, which contain calcifications with posterior acoustic shadowing. Most BBTs were poorly vascularized and only sparse color Doppler signals were detectable. The most common pattern of MBT was a large multilocular-solid or solid mass with irregular tumor borders, and most were moderately or richly vascularized at color Doppler. These four well-established ultrasound-based diagnostic models could discriminates well between benign and malignant BTs.
These typical clinical and ultrasound features are crucial for distinguishing benign and malignant BTs, and they can improve our knowledge and awareness of this rare tumor. Our results are in agreement with these in the literature. Weinberger et al. [7] demonstrated that eighteen (78%) patients with BTTs were asymptomatic, and only 2 patients (8.7%) manifested raised CA125 level. The majority (82%) of the tumors were unilateral. In a recent population-based analysis, Nasioudis et al. [10] demonstrated that the majority of patients with MBT presented with unilateral tumors with a median size of 10cm, 70.3% had elevated CA125 levels. In previous literature, MBT was extremely rare, comprising < 5% of all BTs [8–13]. However, malignancy rates were 37.5% in our series. This bias may be related to the fact that our hospital is the largest referral center for gynecologic oncology in southern region of our country. BBTs were most often small well-defined hypoechoic solid tumors. The important reason for this sonographic feature is that BTTs contain large amounts of abundant fibrous stroma, which similar to skeletal muscle. Green et al. [4] and Athey et al. [5] described the gray-scale ultrasound findings in nine and four benign Brenner tumors, respectively. BBTs were usually composed of homogeneous solid masses, rarely predominantly cystic, or occasionally multilocular-solid on ultrasound examination. Weinberger et al. [7] described 23 BBTs, of which 8 (35%) were described as solid, 8 (35%) as multilocular-solid and 7 (30%) as unilocular or multilocular. Therefore, the sonographic findings of BBTs are similar to those of subserosal pedunculated leiomyoma of the uterus and other solid ovarian masses, such as ovarian fibromas [1]. Moreover, the BBTs could be misdiagnosed as ovarian mucinous or serous cystadenoma, if it coexists with a multilocular cystic component. In our series, two cases coexisting with ipsilateral mucinous cystadenoma. One case exhibited the mixed multilocular cystic and homogeneous solid components, and cystic component was larger than solid component, which is consistent with existing literature [19]. However, the other case exhibited a mostly solid mass containing small cystic areas, and the proportion of mucinous cystadenoma was only 10%. Calcification with posterior acoustic shadowing is an important clue in diagnosis of BBTs [20]. In our series, the calcifications were located in the solid component and the septum of the tumors, which were detected in all BTTs and 55.6% of MBTs. The calcifications and posterior acoustic shadowing were more commonly seen in benign tumors than malignant ones. The important reason was that the proportion of extensive amorphous calcification was higher in benign lesions, which was accompanied by posterior heavy shadowing. Moreover, the ultrasonographic findings of MBTs were irregular multilocular-solid or solid masses and more richly vascularized on Doppler examination, which were similar to ovarian epithelial cancer.
Recently, the ESGO/ISUOG/IOTA/ESGE demonstrated that the IOTA ADNEX and SR-Risk models were the best models for the characterization of ovarian masses, as they outperform existing morphological scoring systems [14]. In agreement with findings from other studies, these four well-established ultrasound-based diagnostic models had the good performance for distinguishing benign and malignant ovarian BTs in our study (AUCs ranged from 0.892 to 0.913). However, we found a relatively lower specificity at the original cut-off points in four diagnostic models. Especially, O-RADS system and SR-Risk were less effective with the specificity of 40.0%, which may be related to the fact that most (73.3%) BBTs were solid masses. When using the optimal cut-off values in these models, our gynecologists had strong confidence in identifying benign ovarian tumors with higher specificity and avoid unnecessary surgery in benign or probably benign cases, which are preoperatively diagnosed using these model.
The main strengths of our study are that it is the first study comparing the clinical characteristics and sonographic appearance between benign and malignant ovarian BTs, and validating the optimal ultrasound-based diagnostic models that offer excellent preoperative discrimination between them in the same cohort of patients. Our research was limited in two ways. First, the ultrasound information was collected retrospectively from ultrasound reports and images. Therefore, our conclusions on ultrasound features of BTs may be biased and must be interpreted with caution. Second, our data was limited to a single-institution database, the size of population was relatively small, making ROC curve and defining optimal cut-off level may be unstable, so prospective lager sample study is needed to validate the above findings.
In conclusion, our results reveal that calcification with posterior acoustic shadowing is an important clue in diagnosis of BBT. Sonographic appearance of MBT is similar to that of other epithelial ovarian carcinomas with a large and irregular multilocular-solid or solid mass, and most are moderately or richly vascularized at color Doppler. These four well-established ultrasound-based diagnostic models have excellent performance in distinguishing them.