Prospective diagnostic test accuracy of transvaginal ultrasound posterior approach for uterosacral ligament and torus uterinus deep endometriosis

To determine the diagnostic test accuracy of transvaginal ultrasound (TVS) using a standardized technique for the diagnosis of deep endometriosis (DE) of the uterosacral ligaments (USLs) and adjacent torus uterinus (TU).

interpreted the ultrasound scans.Accuracy, sensitivity, specificity, positive and negative predictive values (PPV and NPV) and positive and negative likelihood ratios were calculated for the TVS posterior approach for each location using the reference standard.

INTRODUCTION
Endometriosis is a highly prevalent gynecological disease, affecting approximately 10% of women and individuals assigned female at birth.It is characterized by the presence of endometrial-like tissue in areas outside the uterus, resulting in a chronic state of inflammation [1][2][3] , extensive pelvic adhesions and severe morbidity in the form of pelvic pain and/or infertility 3 .Endometriotic lesions are divided into one of three phenotypes: superficial (most common, non-infiltrative), ovarian (OE) (cysts within the ovaries) and deep endometriosis (DE) (infiltrative and most aggressive phenotype, leading to significant distortion of the surrounding anatomical and physiological milieu) 2,4,5 .
Historically, to achieve a diagnosis of endometriosis, a combination of direct visualization of endometriotic lesions at surgery and histopathological analysis was required 6 .However, due to extensive surgical wait times, invasiveness, diagnostic delay and inconsistencies in diagnostic accuracy 7 , the European Society of Human Reproduction and Embryology (2022) guidelines 6 now recommend the use of non-invasive imaging modalities, such as transvaginal ultrasound (TVS) 6 .In 2016, the International Deep Endometriosis Analysis group (IDEA) consensus was developed to describe the sonographic features of endometriosis 8 , improving standardization and characterization of disease phenotypes, including anatomical structures in the posterior compartment, such as the uterosacral ligament (USL) and torus uterinus (TU; junction of the two USLs at the retrocervix).With this development, TVS has become a reliable, non-invasive and rapid diagnostic modality for diagnosing DE, with an overall sensitivity of 88% and specificity of 79% 9 , varying depending on the location of endometriosis.
Despite these recent developments, diagnosis of DE of the USLs remains the most difficult, exhibiting only moderate accuracy 10 , including a sensitivity of 60-67% and specificity of 86-95% 11,12 , even though the USL is the most common location of DE 10,13 .Several research groups have argued that TVS methodology to visualize normal and abnormal USLs and TU is limited, yielding poor diagnostic test accuracy values 11,14 .
Our objective was to assess the diagnostic test accuracy of TVS through a standardized posterior approach for diagnosing DE of the USL and TU, with the probe in the posterior vaginal fornix 14,15 .We hypothesized that this TVS technique would have an improved diagnostic accuracy for DE of the USL and TU relative to what has been described in the literature.

METHODS
This study is reported in accordance with the Standards for Reporting Diagnostic Accuracy (STARD) 2015 guidelines 16 to assist in standardization and transparency of reporting diagnostic test accuracy, taking into consideration the Quality Assessment of the Diagnostic Accuracy of Studies (QUADAS-2) checklist 17 .

Study design
This was a prospective study on diagnostic test accuracy of patients who underwent TVS and laparoscopy (with surgical treatment of endometriosis when it was identified) at McMaster University Medical Center Tertiary Endometriosis Clinic, Hamilton, ON, Canada.Participant recruitment took place from 10 August 2020 to 31 October 2021.A single sonologist (M.L.), considered an expert according to the study co-ordinator and European Federation of Societies for Ultrasound in Medicine and Biology (EFSUMB; Level 3) 18 , was responsible for the final review and reporting of ultrasound findings.All ultrasound scans were performed within 1 year of surgery.Similarly, the reference standard surgical procedure was performed by a single surgeon (M.L.) with minimally invasive gynecological surgery training and working in a center of surgical excellence with available colorectal and urological surgery.The study was approved by the Hamilton Integrative Research Ethics Board (HiREB: 12617).

Participants
Consecutive participants were recruited if they met the inclusion criteria.Inclusion criteria were: age between 18 and 50 years; female sex assigned at birth; postmenarchal and premenopausal patient; history of chronic pelvic pain and/or endometriosis; ability to undergo TVS; and consent to laparoscopic surgery for endometriosis.Patients were excluded if they had a previous diagnosis or current active gynecological malignancy, suspected/confirmed pregnancy or if they ultimately underwent laparoscopy at a different center.Patients on various medical therapies were not excluded.

Study procedure and test methods
All data, including TVS, surgical and histological findings, were collected prospectively in real time.Data were reviewed and entered by two investigators (S.M.F. and V.T.) independently and confirmed and audited for accuracy by the supervising investigator (M.L.).

Index test
The index test (TVS) was performed and the findings were reported in accordance with the IDEA consensus, with the addition of a posterior approach technique described previously 14,15 , which was hypothesized to improve the diagnostic accuracy for the evaluation of DE of the USLs and TU (Appendix S1).A Mindray Zonare machine with a E9-4 (4-9-MHz) transducer was used (Zonare ® Medical Systems Inc., Mountain View, CA, USA).

TVS posterior approach
The technique for TVS assessment of the USL and TU involves insertion of the TVS probe into the posterior vaginal fornix, angled towards the rectum in the Endometriosis, ultrasound and USL 265 midsagittal position.Once the hypoechoic posterior vaginal fornix and the overlying hyperechoic rectouterine peritoneum (thin white line) are visualized, the probe is lateralized and rotated clockwise (right USL) or counterclockwise (left USL), usually by no more than 45 • .Upon lateralization and rotation, there is a progressive thickening of the white line of the peritoneum, revealing the USL at the thickest point.If pelvic vessels are noted in concentration, the probe tip has been lateralized too far.Suspected DE presents as a hypoechoic nodule within the surrounding hyperechoic USL and should be measured in three orthogonal planes.Characterization of DE of the USLs and TU using the standardized posterior approach is illustrated in Figure 1.DE nodules in this space can be regular or irregular and can be confluent with other nodules nearby in the posterior vaginal fornix, bowel, TU or parametrium.Sonographically, the TU is defined as the thin hyperechoic layer of tissue immediately posterior to the cervix.The TU tissue simply continues into the USLs without a clear anatomical border, but the edge of the cervix can be used as an estimate of a border.

Surgical visualization
The index test was compared primarily with direct visualization through laparoscopy as the reference standard.
When tissue was available, the index test was compared with histological confirmation as a secondary reference standard.Direct visualization involved a systematic evaluation of all surfaces of the abdomen and pelvis as per the local standardized protocol (Appendix S1).The reporting of surgical findings mirrored the systematic ultrasound evaluation approach.The surgeon (M.L.) was not blinded to the ultrasound findings, which reflects the real clinical experience of surgeons, who utilize preoperative information to plan and perform surgery appropriately.Complete excision of endometriosis was performed, unless the risk of excision outweighed the benefit as per patient preference and informed consent, in which case occasional intentional incomplete excision was performed.On direct visualization, presence of DE was defined when irregular nodules of varying pigmentations with a fibrotic (or hard) tactile feedback on palpation were noted 19 .OE was noted when an ovarian cyst containing dark brown material (chocolate cyst) was present.Obliteration of the rectouterine pouch was noted when the peritoneum between and inferior to the USLs was not visible or present due to the presence of adhesions.Normal peritoneum and other pelvic surfaces were noted when no evidence of endometriosis (of any subtype) or adhesions was present.If there was uncertainty about whether the appearance of an area was normal or abnormal, it was considered suspicious for endometriosis based on direct visualization,

Histology
Microscopically, DE is characterized by endometrial stroma and/or glands with fibrosis, along with hyperplastic and hypertrophic smooth-muscle fibers 20 .The pathologists performing the histological evaluation were not blinded to the clinical history, which reflects the real clinical experience.However, they were blinded to the index test results.Histology was evaluated only when samples were available and provided as supplementary data.

Other variables
Variables collected for univariate analysis included: patient age; endometriosis phenotype and distribution present on laparoscopy and confirmed histologically; presence of adenomyosis, based on the Morphological Uterus Sonographic Assessment (MUSA) 21 features; presence of fibroids, based on the International Federation of Gynecology and Obstetrics (FIGO) 22 classification features; symptoms (dysmenorrhea, dyspareunia, dyschezia, dysuria and abnormal uterine bleeding) and previous diagnosis and surgery for endometriosis.

Sample size
The sample size for the TVS technique was determined based on the Buderer formula 23 .Using the expected prevalence of surgically confirmed USL DE within our clinical population of 35% and an expected sensitivity and specificity of 90% and 85%, respectively, based on preliminary results of our group (unpubl.data) and a previous diagnostic accuracy study 24 (confidence level of 95%; power of 0.85), a total of 49 participants were required.

Statistical analysis
Data were collected using the REDCap electronic data capture tool (Vanderbilt University, Nashville, TN, USA) and imported into Microsoft Excel for Windows 10 (Microsoft Corp., Redmond, WA, USA).Cleaned data were transferred and analyzed using IBM SPSS statistics v29 software (SPSS Inc., Chicago, IL, USA).
Descriptive statistics were used to summarize all variables.Continuous variables are presented as mean ± SD.Categorical variables are presented as n (%).Additional relevant variables identified during TVS and/or laparoscopy with histological confirmation, including endometriosis phenotype and co-occurrence of benign gynecological disease, are reported as n (%) relative to the total population.Accuracy, sensitivity, specificity, negative and positive predictive values (NPV and PPV), negative likelihood ratios (LR−) and positive likelihood ratios with 95% CI were calculated for the index test relative to laparoscopy as the primary reference standard among all participants.The diagnostic parameters for the index test were then calculated relative to histology as a secondary reference standard.All accuracy parameters were determined using the crosstabulation function in SPSS.

Participant characteristics
Of 160 consecutive patients identified in the clinical setting, 54 met the inclusion criteria and were included in the study (Figure 2).All participants underwent the index test and surgical visualization.Only one participant underwent diagnostic laparoscopy without surgical excision of endometriosis despite its presence; the surgery was canceled due to anesthesiology concerns.
Participant characteristics are summarized in When assessing lesion dimensions in three orthogonal planes sonographically, the mean ± SD length, width and height of the left USL were 9.3 ± 4.5 mm, 5.0 ± 3.0 mm and 10.0 ± 3.9 mm, respectively.The respective values were 10.4 ± 7.0 mm, 6.8 ± 4.8 mm and 8.5 ± 3.8 mm  for the right USL and 7.8 ± 3.6 mm, 5.6 ± 2.7 mm and 10.0 ± 4.2 mm for the TU.
Beyond the nodules of the USL and TU, laparoscopically visualized and histologically confirmed superficial endometriosis was present in 70.4% (38/54) of all patients.Laparoscopically and histologically confirmed OE was present in 29.6% (16/54) of patients, while 11.1% (6/54) of patients were confirmed to have non-OE benign cysts.Beyond the USLs, the most common location for DE was the bowel, which was affected in 18.5% (10/54) of patients, with laparoscopically confirmed rectouterine pouch obliteration in 24.1% (13/54) of patients.

Main findings
In this study on diagnostic test accuracy, we evaluated the accuracy of a standardized TVS posterior approach for identifying and characterizing DE of the USLs and TU proposed by Leonardi et al. 14 in 2020.Our findings suggest that identifying DE in its most common location, the USLs, with the TVS probe placed within the posterior vaginal fornix, may yield improved diagnostic accuracy relative to that reported in the literature 11,12 .When assessing diagnostic performance, our findings suggest variations according to anatomical site and reference test; however, these were subtle, as indicated by the large overlap in 95% CI.

Interpretation and significance
There is a ubiquitous acceptance of the need for USL evaluation as part of TVS examination for endometriosis 11 .Although the USL is the most common location of DE, it remains the most difficult to diagnose due to the small size of nodules and lack of experience of clinicians in evaluating these structures using imaging [10][11][12] .Despite recent advancements made through the development of the IDEA consensus to aid the diagnosis of endometriosis, the diagnostic test performance of TVS for DE in these anatomical areas has not improved 9,11 .A recent study evaluating the diagnostic performance of TVS using the IDEA consensus among all anatomical sites suggested that the accuracy was lowest for the right and left USLs and TU, with accuracy ranging between 65.2% to 74.4%, sensitivity ranging between 44.2% and 58.7% and specificity ranging between 77.8% and 88.2% for the three sites 9 .Our findings suggest that the standardized posterior approach proposed by Leonardi et al. 14 (i.e.approaching the posterior compartment through the posterior vaginal fornix) may improve visualization of normal and abnormal USLs and TU and the diagnostic test performance of TVS.Because the USL is the most common location for DE and the only site of disease in some individuals, the overall diagnostic accuracy of TVS for endometriosis and diagnostic delay in those suffering from endometriosis can be improved by optimizing imaging diagnostic performance of the USLs.Beyond diagnostic improvements, understanding DE of the USL may have clinical utility; the USLs are highly innervated supporting structures, carrying crucial neurological components from the spinal cord, including the inferior hypogastric plexus, supplying pelvic and perineal organs with parasympathetic and sympathetic innervation required for normal physiological function 25 .
Endometriosis of the USL may be clinically linked to symptoms such as dyspareunia and chronic pelvic pain 26 .
Identifying endometriosis of the USLs and respective TU is essential in order to advance our understanding of this enigmatic disease with varying presentations of symptoms.Lastly, it should be noted that scanning the USLs and TU is possible through the anterior vaginal fornix, providing a different perspective from that achieved by the posterior approach, which, on occasion, may provide additional value.

Strengths and limitations
All efforts were made to ensure the robustness of this study through adherence to the STARD guidelines 16 .
The prospective nature of this study involved utilizing standardized reporting forms to ensure consistency among consecutive patients.Furthermore, the study was performed in a specialized center with high-quality ultrasound and surgical equipment, allowing for detailed characterization of the posterior compartment.To ensure generalizability, the technique adhered primarily to the methodology described by Leonardi et al. 14,15 , in addition to the IDEA methodology, which has been shown previously to improve diagnostic accuracy using a generalizable, multisite approach 8,9 .Lastly, an additional strength of this study lies in the novelty of evaluating the diagnostic accuracy of a standardized TVS posterior approach.Several limitations should be noted, particularly in the interpretation of the results.First, the reported accuracy and generalizability may be impacted by the fact that surgery was performed by the same highly trained surgeon sonologist and the use of high-quality equipment.Given the current study design, the person who had developed the technique also conducted all TVS scans and surgeries in this study, precluding a double-blind evaluation.Though this study design is common in diagnostic test accuracy studies on endometriosis, future studies may consider a design that involves blinding the surgeon to ultrasound findings.In addition, one may consider a design comparing non-experienced ultrasound operators (index test) against experienced operators (reference standard), or vice versa, to potentially enhance the generalizability of the evaluated technique.Furthermore, a large proportion of the patient population did not undergo histological evaluation due to the absence of endometriosis on laparoscopy; biopsy of entirely normal tissue was not performed due to ethical considerations and resource limitations.In such cases, it is uncertain whether endometriosis was truly absent in these patients, as samples for histological evaluation were not obtained.The study relied on the accuracy of surgical visualization, which has limitations 7 .Specifically, DE lesions may be missed by the surgeon 27 .Although the diagnosis of endometriosis relies typically on final histological confirmation, histological assessment also has limitations, for example in cases in which stroma and epithelial cells are altered/damaged 28 or those with fibrous obliteration 29 .Similar limitations arise with laparoscopy; studies have suggested that 50% of cases suspected of having endometriosis on laparoscopy were proven histologically 30 and 25% of cases with an atypical appearance of tissue that did not raise suspicion of endometriosis were proven to have endometriosis 31 .These studies illustrate the caveats associated with relying on surgical experience and expertise.Due to the imperfect relationship between laparoscopy and histological findings, the assumption that the surgeon is correct in their diagnosis has been adopted 7 .Additionally, this study had a relatively small sample size, with 44.4% of the participants with a history of surgery for endometriosis, which may create bias, potentially leading to an artificial increase in the diagnostic performance of TVS.Lastly, we did not include a comparison with an anterior TVS approach to elucidate whether the investigated posterior approach is superior in imaging the USLs.

Conclusion
DE of the USLs and TU is highly prevalent, but its detection and characterization using TVS has been poor in previous studies.We have tested a previously proposed technique in which the TVS probe is placed in the posterior vaginal fornix to evaluate the posterior compartment, which we call the posterior approach.Our findings suggest that the posterior approach may yield improved accuracy for the USLs and TU compared with the results of previous studies.External validation and larger studies would be valuable to strengthen or refute these findings.

Figure 2
Figure 2Flowchart summarizing inclusion of participants in study and diagnostic performance of transvaginal ultrasound (TVS) posterior approach in detecting deep endometriosis (DE) in uterosacral ligaments (USLs) and torus uterinus.CPP, chronic pelvic pain.

Ultrasound
Obstet Gynecol 2024; 63: 263-270 Published online in Wiley Online Library (wileyonlinelibrary.com).DOI: 10.1002/uog.27492.This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

Table 1
Characteristics of study population (n = 54) Data are given as mean ± SD or n (%).

Table 2
Diagnostic test accuracy of transvaginal ultrasound posterior approach for left uterosacral ligament (USL), right USL and torus uterinus deep endometriosis relative to laparoscopy as reference standard in 54 patients