Introduction

Endometriosis is a chronic, benign, inflammatory gynecologic disorder caused by endometrial tissue implanted outside the endometrial cavity [1]. Endometriosis affects 6 to 10% of reproductive age women and commonly causes chronic pelvic pain, dyspareunia, and infertility [1]. Endometriosis is most commonly found in pelvis, where it can affect the pelvic peritoneum, ovaries, and/or the rectovaginal septum [1]. There are three clinically distinct forms of pelvic endometriosis: endometriotic implants on the peritoneum (superficial endometriosis), ovarian cysts lined by endometrioid mucosa (endometriomas), and deeply infiltrating endometriosis (DIE) [2]. DIE is defined by > 5 mm subperitoneal invasion by endometriotic lesions [3].

The gold standard for the staging of endometriosis is by laparoscopy [4]. The most widely accepted staging system for endometriosis is the revised American Society for Reproductive Medicine (rASRM) classification [5] (Fig. 1). The rASRM classification uses a scoring system based on intra-operative findings to stage endometriosis as no endometriosis (0), minimal (I), mild (II), moderate (III), or severe (IV) [6].

Fig. 1
figure 1figure 1

Reprinted from [6], with permission from Elsevier

Scoring system (a) and examples (b) for the revised American Society for Reproductive Medicine (rASRM) classification.

Full size image

For women with symptomatic endometriosis, treatment options include medical therapy or surgery. Surgical management of endometriosis-associated pelvic pain and infertility has been shown to improve pain symptoms and pregnancy rates, respectively [7]. The standard treatment of deep infiltrating endometriosis is complete excision of the lesions [8].

Transvaginal ultrasound is the initial imaging modality of choice to evaluate pelvic endometriosis, specifically to discern endometriomas from other ovarian cysts [7]. With its high spatial resolution, multiplanar evaluation, and good tissue characterization, magnetic resonance imaging (MRI) provides added value in the pre-operative work-up in pelvic endometriosis, specifically for the evaluation of deep infiltrating endometriosis. MRI is highly accurate for diagnosis and localization of DIE and useful for improved presurgical planning and preparation [3, 8, 9]. MRI findings likely to result in a change in operative management include features of malignancy, involvement of the bowel, bladder, ureters, sciatic nerve, cesarean-section scar/abdominal wall, and upper abdomen, including the liver and diaphragm.

The ovaries are anchored in position by three structures. The mesovarium anchors the ovary to the posterior surface of the broad ligament. The utero-ovarian ligament anchors the ovary to the uterus. The suspensory ligament anchors the ovary to the pelvic side wall. Although these structures provide anchoring, they all have a degree of laxity, allowing for a certain amount of ovarian mobility. In nulliparous women, the ovaries are typically found in the ovarian fossae, shallow peritoneal depressions in the lateral pelvic side walls. During the first pregnancy, there is stretching of the anchoring structures of the ovary and the ovaries are pushed out of the pelvis. Following the first pregnancy, the ovaries can assume variable positions in the pelvis [10].

Prior work has demonstrated that the sonographic sign of “kissing ovaries” is strongly associated with moderate to severe endometriosis [11]. This study aims to investigate the association between ovarian position and endometriosis stage, specifically the association between kissing and retropositioned ovaries and stage of endometriosis.

Materials and methods

After Institutional Review Board approval, cases were gathered (a) by query of a prospectively maintained database of patients with suspected endometriosis and bowel symptoms or endometriomas who underwent pre-operative MRI and (b) by query of a web-based search application for radiology reports from our institution. Via the web-based search application, all MRI reports from December 2015 to June 2017 were queried with the term “endometriosis.” All MRI reports with positive or equivocal findings of endometriosis were included in the study. Equivocal findings of endometriosis included findings that were suggested in the report that may be related to endometriosis, but were not specific to endometriosis. The date range was selected to match that of the prospectively maintained database of patients with suspected endometriosis and bowel symptoms or endometriomas. Patients that were of status post hysterectomy or oophorectomy were excluded from the study. All MRI examinations were performed using 1.5- or 3.0-Tesla superconducting magnets (GE, Waukesha, WI).

All MRI examinations were independently reviewed by a trained senior radiology resident and board-certified abdominal radiologist experienced in endometriosis imaging for ovarian positioning and the presence of an endometrioma. In a minority of cases in which the designation of ovarian position differed between the resident and radiologist, a second board-certified abdominal radiologist experienced in endometriosis imaging provided a consensus read. The position of the ovaries was classified as (a) kissing when they were both posterior to the cornua of the uterus and in contact (Fig. 2), (b) retropositioned when they were both posterior to the cornua of the uterus and appeared adherent to the uterine serosa, but not in contact with each other (Fig. 3), or (c) normal when they were not retropositioned or kissing (Fig. 4). The radiology report was reviewed to obtain data on whether or not there was any additional MR evidence of endometriosis. Equivocal findings of endometriosis were marked as positive for the purpose of this study.

Fig. 2
figure 2

Coronal T2 weighted (a), axial T2 weighted with fat-saturation (b), and axial T1 weighted with fat-saturation (c) images demonstrating kissing ovaries (arrows) with bilateral ovarian endometriomas. This patient had stage 4 endometriosis intra-operatively

Full size image
Fig. 3
figure 3

Coronal T2 weighted (a), axial T2 weighted with fat-saturation (b), and axial T1 weighted with fat-saturation (c) images demonstrating retropositioned ovaries (arrows) with bilateral ovarian endometriomas. This patient had stage 4 endometriosis intra-operatively

Full size image
Fig. 4
figure 4

Coronal T2 weighted (a), axial T2 weighted with fat-saturation (b), and axial T1 weighted with fat-saturation (c) images demonstrating normally positioned ovaries (arrows), without evidence of an ovarian endometrioma. This patient had stage 0 endometriosis intra-operatively. Incidentally noted IUD in the uterine cavity

Full size image

A retrospective review of the electronic medical record was performed to identify the patients who had undergone subsequent surgery and obtain data on the intra-operative staging of endometriosis, presence of DIE, patient age, and history of prior surgery for endometriosis. Intra-operative staging of endometriosis was determined using the revised American Society for Reproductive Medicine classification system (rASRM) by a surgeon with expertise in endometriosis surgery (Fig. 5). The presence of DIE was obtained from the operative report. All cases of positive endometriosis were confirmed by pathology.

Fig. 5
figure 5

Intra-operative photo of kissing ovaries (same case as Fig. 1). Intra-operative findings consistent with rASRM stage 4 endometriosis

Full size image

Statistical analysis of the correlation between ovarian positioning and endometriosis staging was evaluated with a logistical regression analysis. Correlation between ovarian positioning and the presence of an ovarian endometrioma, history of prior surgery for endometriosis, and the presence of deep invasive endometriosis were evaluated with Chi-square analysis. Sensitivity, specificity, and accuracy of kissing or retropositioned ovaries were also calculated.

Results

Between December 2015 and June 2017, 65 women with suspected endometriosis underwent pre-operative MRI and subsequent intra-operative staging of endometriosis. Nine additional patients who underwent pre-operative MRI and subsequent surgery for endometriosis were excluded from the study due to prior history of hysterectomy or oophorectomy. The median age at the time of the MRI was 36 years (range 17–59 years); 34 women (52%) had undergone a prior surgery for endometriosis at the time of MRI examination.

MR images revealed kissing ovaries in 12 women (18.5%), retropositioned ovaries in 17 women (26.1%), and normally positioned ovaries in 36 women (55.4%). On MRI, endometriosis was found in 46 patients and no evidence of endometriosis was observed in 17 patients. Thirty two patients (49%) had endometrioma on MRI.

The median time between the MRI examination and surgery was 45 days (range 1–432 days). Intra-operative staging of endometriosis revealed 13 women with stage 0 endometriosis (20%), 15 women with stage I endometriosis (23%), 6 women with stage II endometriosis (9%), 9 women with stage III endometriosis (14%), and 22 women with stage IV endometriosis (34%). According to the operative reports, DIE was observed in 13 women (20%). DIE was most commonly observed involving the posterior cul-de-sac and rectosigmoid bowel, with other sites including the anterior cul-de-sac, vagina, cervix, ureter, pelvic sidewall, and broad ligament.

Chi-square analysis demonstrated a strong association between ovarian positioning and stage of endometriosis (p < 0.0001) (Fig. 6). An association between ovarian positioning and the presence of an endometrioma was also observed (p < 0.0001) (Fig. 7). The presence of an endometrioma was 83% (CI 67–94%) specific and 90% (CI 67–90%) sensitive for retropositioned or kissing ovaries with a positive predictive value of 81% (CI 67–90%). Additionally, an association between ovarian positioning and DIE was demonstrated (p = 0.01) (Fig. 8). The presence of retropositioned or kissing ovaries was 68% (CI 51–82%) specific and 67% (CI 45–84%) sensitive for DIE with a positive predictive value of 55% (CI 42–68%). No statistically significant association between ovarian positioning and prior surgery for endometriosis was demonstrated (p = 0.53), with 38% of the women with prior surgery having retropositioned or kissing ovaries, compared to 51% of the women without prior surgery.

Fig. 6
figure 6

Bar graph demonstrating the stage of endometriosis by ovarian position

Full size image
Fig. 7
figure 7

Bar graph demonstrating the presence of an ovarian endometrioma by ovarian position

Full size image
Fig. 8
figure 8

Bar graph demonstrating the presence of deep infiltrating endometriosis (DIE) by ovarian position

Full size image

Kissing ovaries demonstrated a sensitivity of 41% (95% CI 21–64%) and a specificity of 95% (95% CI 84–99%) for stage IV endometriosis. Kissing or retropositioned ovaries demonstrated a combined sensitivity of 86% (95% CI 65–97%) and a specificity of 79% (95% CI 64–90%) for stage IV endometriosis. All cases with kissing ovaries had stage III or IV endometriosis.

A correlative analysis with logistic regression demonstrated that the odds of stage IV endometriosis were eight times higher given kissing or retropositioned compared to normal ovaries, regardless of the presence of an endometrioma (p = 0.01).

Discussion

Our study demonstrates a strong association between retropositioned or kissing ovaries on MR imaging and stage III/IV of endometriosis, in women with suspected endometriosis. Kissing ovaries alone have an even stronger association with the most severe stage of disease. The association between retropositioned or kissing ovaries and stage IV endometriosis was independent of the presence of an endometrioma. Retropositioned or kissing ovaries were associated with deep infiltrating endometriosis at surgery. Thus in women suspected of having endometriosis, a finding of retropositioned or kissing ovaries on the MRI is suggestive of severe pelvic endometriosis and should prompt a search for additional findings of endometriosis. The ability of retropositioned or kissing ovaries to indicate higher stages of endometriosis can allow for improved pre-operative planning for the surgeon.

Retropositioned or kissing ovaries are likely caused by an inflammation stimulated by the endometrial tissue in the posterior cul-de-sac, on the uterosacral ligaments, and on the ovaries [12]. Chronic cyclical inflammation likely leads to the formation of adhesions that tether the ovaries to the posterior uterus, displacing them medially and posteriorly.

Slightly over half of the women in our study had prior history of surgery for endometriosis. As any intra-abdominal surgery can cause adhesions, we investigated whether retropositioned or kissing ovaries were associated with prior surgery, a potential confounding variable. Our results do not support an association between retropositioned or kissing ovaries and prior endometriosis surgery. Therefore, retropositioned or kissing ovaries are concerning for the highest stage of endometriosis regardless of surgical history.

Several strengths and limitations of this study warrant review. Our study substantiates the work of Ghezzi et al. [11] in which the finding of kissing ovaries on ultrasound was associated with severe endometriosis. To the best of the authors’ knowledge, this is the first study to examine the relationship between ovarian position on MRI and stage of endometriosis. Furthermore, this is probably the first study to explore the relationship between retropositioned ovaries and endometriosis staging and the relationship between ovarian position and DIE.

The primary limitations include the retrospective design and sample size. Our findings should be confirmed in larger series. Future studies could include prospectively referring women with retropositioned or kissing ovaries on ultrasound for further evaluation with MRI for evidence of endometriosis. Pre-operative detection of kissing or retropositioned ovaries could have a significant impact on operative planning and reduce the number of patients with inadequate primary surgery for endometriosis due to unexpected moderate or severe disease.