Volume 58, Issue 2 p. 190-200
Systematic Review
Free Access

Optimal imaging modality for detection of rectosigmoid deep endometriosis: systematic review and meta-analysis

B. Gerges

Corresponding Author

B. Gerges

Sydney Medical School Nepean, University of Sydney Nepean Hospital, Kingswood, NSW, Australia

Sydney West Advanced Pelvic Surgery (SWAPS), Blacktown Hospital, Blacktown, NSW, Australia

Correspondence to: Dr B. Gerges, Acute Gynaecology, Early Pregnancy and Advanced Endosurgery Unit, Nepean Hospital, Penrith, NSW 2750, Australia (e-mail: [email protected])Search for more papers by this author
W. Li

W. Li

Department of Obstetrics and Gynaecology, Monash University, Clayton, VIC, Australia

Search for more papers by this author
M. Leonardi

M. Leonardi

Sydney Medical School Nepean, University of Sydney Nepean Hospital, Kingswood, NSW, Australia

Department of Obstetrics and Gynecology, McMaster University, Hamilton, Canada

Search for more papers by this author
B. W. Mol

B. W. Mol

Department of Obstetrics and Gynaecology, Monash University, Clayton, VIC, Australia

Aberdeen Centre for Women's Health Research, University of Aberdeen, Aberdeen, UK

Search for more papers by this author
G. Condous

G. Condous

Sydney Medical School Nepean, University of Sydney Nepean Hospital, Kingswood, NSW, Australia

Search for more papers by this author
First published: 10 October 2020
Citations: 27

ABSTRACT

en

Objectives

To review the accuracy of different imaging modalities for the detection of rectosigmoid deep endometriosis (DE) in women with clinical suspicion of endometriosis, and to determine the optimal modality.

Methods

A search was conducted using PubMed, MEDLINE, Scopus, EMBASE and Google Scholar to identify studies using imaging to evaluate women with suspected DE, published from inception to May 2020. Studies were considered eligible if they were prospective and used any imaging modality to assess preoperatively for the presence of DE in the rectum/rectosigmoid, which was then correlated with the surgical diagnosis as the reference standard. Eligibility was restricted to studies including at least 10 affected and 10 unaffected women. The QUADAS-2 tool was used to assess the quality of the included studies. Mixed-effects diagnostic meta-analysis was used to determine the overall pooled sensitivity and specificity of each imaging modality for rectal/rectosigmoid DE, which were used to calculate the likelihood ratio of a positive (LR+) and negative (LR–) test and diagnostic odds ratio (DOR).

Results

Of the 1979 records identified, 30 studies (3374 women) were included in the analysis. The overall pooled sensitivity and specificity, LR+, LR– and DOR for the detection of rectal/rectosigmoid DE using transvaginal sonography (TVS) were, respectively, 89% (95% CI, 83–92%), 97% (95% CI, 95–98%), 30.8 (95% CI, 17.6–54.1), 0.12 (95% CI, 0.08–0.17) and 264 (95% CI, 113–614). For magnetic resonance imaging (MRI), the respective values were 86% (95% CI, 79–91%), 96% (95% CI, 94–97%), 21.0 (95% CI, 13.4–33.1), 0.15 (95% CI, 0.09–0.23) and 144 (95% CI, 70–297). For computed tomography, the respective values were 93% (95% CI, 84–97%), 95% (95% CI, 81–99%), 20.3 (95% CI, 4.3–94.9), 0.07 (95% CI, 0.03–0.19) and 280 (95% CI, 28–2826). For rectal endoscopic sonography (RES), the respective values were 92% (95% CI, 87–95%), 98% (95% CI, 96–99%), 37.1 (95% CI, 21.1–65.4), 0.08 (95% CI, 0.05–0.14) and 455 (95% CI, 196–1054). There was significant heterogeneity and the studies were considered methodologically poor according to the QUADAS-2 tool.

Conclusions

The sensitivity of TVS for the detection of rectal/rectosigmoid DE seems to be slightly better than that of MRI, although RES was superior to both. The specificity of both TVS and MRI was excellent. As TVS is simpler, faster and more readily available than the other methods, we believe that it should be the first-line diagnostic tool for women with suspected DE. © 2020 International Society of Ultrasound in Obstetrics and Gynecology

RESUMEN

es

Modalidad de imagen óptima para la detección de la endometriosis profunda rectosigmoidea: revisión sistemática y metaanálisis

Objetivos

Revisar la precisión de diferentes modalidades de técnicas basadas en imágenes para la detección de la endometriosis profunda (EP) rectosigmoidea en mujeres con indicios clínicos de endometriosis, y determinar la modalidad óptima.

Métodos

Se realizó una búsqueda en PubMed, MEDLINE, Scopus, EMBASE y Google Scholar para identificar estudios que hubieran utilizado imágenes con las que evaluar a mujeres con indicios de EP, publicados desde su comienzo hasta mayo de 2020. Los estudios se consideraron elegibles si eran prospectivos y utilizaban cualquier modalidad de imagen para evaluar de forma preoperatoria la presencia de EP en el recto/rectosigmoide, que luego se correlacionó con el diagnóstico quirúrgico como estándar de referencia. La elegibilidad se limitó a los estudios que incluían al menos 10 mujeres afectadas y 10 no afectadas. Para evaluar la calidad de los estudios incluidos se utilizó la herramienta QUADAS-2. Se utilizó un metaanálisis para el diagnóstico de efectos mixtos con el que determinar la sensibilidad y la especificidad globales conjuntas de cada modalidad de imagen para la EP rectal/rectosigmoidea, que se emplearon para calcular el cociente de verosimilitud de una prueba positiva (LR+) y negativa (LR-) y la razón de momios del diagnóstico (RMD).

Resultados

De los 1979 registros identificados, se incluyeron en el análisis 30 estudios (3374 mujeres). La sensibilidad y la especificidad globales conjuntas, las LR+ y LR- y la RMD para la detección de la EP rectal/rectosigmoidea mediante ecografía transvaginal (ETV) fueron, respectivamente, del 89% (IC 95%, 83–92%), del 97% (IC 95%, 95–98%), del 30,8 (IC 95%, 17,6–54,1), del 0,12 (IC 95%, 0,08–0,17) y del 264 (IC 95%, 113–614). Para las imágenes por resonancia magnética (IRM), los valores respectivos fueron del 86% (IC 95%, 79–91%), 96% (IC 95%, 94–97,0%), 21.0 (IC 95%, 13,4–33,1), 0,15 (IC 95%, 0,09–0,23) y 144 (IC 95%, 70–297). Para la tomografía axial computerizada, los valores correspondientes fueron del 93% (IC 95%, 84–97,0%), 95% (IC 95%, 81–99,0%), 20,3 (IC 95%, 4,3–94,9%), 0,07 (95%, 0,03–0,19) y 280 (IC 95%, 28–2826). Respecto a la ecografía endoscópica rectal (EER), los valores respectivos fueron del 92% (IC 95%, 87–95%), 98% (IC 95% 96–99%), 37,1 (IC 95%, 21,1–65,4), 0,08 (IC 95%, 0,05–0,14) y 455 (IC 95%, 196–1054). La heterogeneidad fue significativa y los estudios se consideraron metodológicamente de mala calidad, de acuerdo con la herramienta QUADAS-2.

Conclusiones

La sensibilidad de la ETV para la detección de la EP rectal/rectosigmoidea parece ser ligeramente mejor que la de la IRM, aunque la EER fue superior a ambas. La especificidad fue excelente tanto para la ETV como para la IRM. Dado que la ETV es más sencilla, rápida y accesible que los otros métodos, creemos que debería ser la herramienta de diagnóstico de primera línea para las mujeres con indicios de EP. © 2020 International Society of Ultrasound in Obstetrics and Gynecology.

摘要

zh

检测直肠乙状结肠深部子宫内膜异位症的最佳影像学手段:系统评价和meta分析

目的

评述对临床疑似子宫内膜异位症的妇女检测直肠乙状结肠深部子宫内膜异位症(DE)的不同影像学手段的准确性,以确定最佳方法。

方法

使用PubMed、MEDLINE、Scopus、EMBASE数据库及谷歌学术展开了一次检索,以识别出使用影像来评估疑似DE妇女的研究,从最开始的发表文献一直到2020年5月。前瞻性研究和使用了任何影像学手段在术前评估直肠/直肠乙状结肠内DE的研究都被认为是符合条件的,再以相关的外科诊断作为参考标准。将条件限制在至少含有10名患病妇女和10未患病妇女的研究。使用了QUADAS-2工具来评估所含研究的质量。用混合效应诊断meta分析确定用于评估直肠/直肠乙状结肠内DE的每种影像学手段的总敏感性与特异性,并以此计算出阳性(LR+)和阴性(LR–)测试结果的概率比,以及诊断比值比。

结果

在被识别的1979个记录中,本分析包含了30项研究(涉及3374名妇女)。采用经阴道超声检测直肠/直肠乙状结肠DE的总敏感性与特异性、LR+、LR–和DOR,分别为89%(95% CI,83–92%),97%(95% CI,95–98%),30.8(95%CI,17.6–54.1),0.12(95%CI,0.08–0.17)和264(95% CI,113–614)。采用磁共振成像(MRI)时,各个值分别为86%(95% CI,79–91%),96%(95% CI,94–97%),21.0(95%CI,13.4–33.1),0.15(95% CI,0.09–0.23)和144(95% CI,70–297)。采用计算机断层扫描时,各个值分别为93%(95% CI,84–97%),95%(95% CI,81–99%),20.3 (95%CI,4.3–94.9),0.07(95%CI,0.03–0.19)和280(95% CI,28–2826)。采用直肠内窥镜超声(RES)时,各个值分别为92%(95% CI,87–95%),98%(95% CI,96–99%),37.1(95% CI,21.1–65.4),0.08(95%CI,0.05–0.14)和455(95% CI,196–1054)。这里存在明显的异质性,而且根据QUADAS-2工具这些研究被认为是在方法上较差。

结论

用TVS检测直肠/直肠乙状结肠DE的敏感度似乎比用MRI检测的略好一些,而用RES检测的敏感度比前两者都要高。用TVS和TVS检测的特异性都很好。与其他方法相比,由于TVS更简单、更快速且更容易获得,我们认为它应当是对疑似DE妇女的首选诊断工具。© 2020年国际妇产科超声学会。

CONTRIBUTION

What are the novel findings of this work?

Transvaginal sonography (TVS) was marginally superior to magnetic resonance imaging for the diagnosis of rectal/rectosigmoid deep endometriosis (DE) and, while computed tomography also performed well, rectal endoscopic sonography outperformed all other imaging techniques.

What are the clinical implications of this work?

The findings confirm the importance of the use of imaging for the preoperative diagnosis of rectal/rectosigmoid DE, with all imaging modalities performing well, although there are key differences in their applicability. As TVS is simpler, faster and more readily available than the other methods, we believe it should be the first-line diagnostic tool.

INTRODUCTION

Endometriosis is typically characterized by the deposition of endometrial-like cells outside the uterine cavity, resulting in lesions ranging from peritoneal implants to deep endometriosis (DE). DE is defined as endometrial implants or nodules with ≥ 5 mm subperitoneal infiltration that can involve multiple organs in the pelvis, including the bladder, vagina, uterosacral ligaments, bowel and peritoneum1, 2. Resultant distortion in anatomy can occur, including obliteration of the pouch of Douglas (POD) or cul-de-sac. Symptoms of endometriosis can include chronic pelvic pain, dyspareunia, urinary and/or bowel symptoms and infertility3-5.

Many imaging modalities have been used in an attempt to diagnose accurately the presence and severity of endometriosis, specifically DE, preoperatively. These include, but are not limited to, transvaginal sonography (TVS), rectal endoscopic sonography (RES), magnetic resonance imaging (MRI) and computed tomography (CT). Although TVS and MRI have been shown to be comparable with regard to diagnostic accuracy6-8, any differences being in cost and expertise needed, laparoscopy remains the gold standard for the diagnosis of endometriosis2. In 2016, the International Deep Endometriosis Analysis (IDEA) group published a consensus statement to standardize the terms, definitions and sonographic evaluation of the pelvis in women with suspected endometriosis and, in particular, identified five locations for examination, namely the bladder, rectovaginal septum, posterior vaginal fornix, uterosacral ligaments/torus uterinus and rectum/rectosigmoid/sigmoid9.

There have been a number of systematic reviews assessing separately the accuracy of TVS7, 10, 11 and MRI12 for the detection of DE, including those of Guerriero et al.10 and Pereira et al.8, in which head-to-head analyses of MRI and TVS were performed to assess the accuracy of TVS for the detection of rectosigmoid DE, as well as the most recent Cochrane systematic review, by Nisenblat et al.13, in 2016, which compared multiple imaging technologies.

The purpose of this systematic review was to assess the accuracy of different imaging modalities for the preoperative detection of DE in the rectum/rectosigmoid, as defined by the IDEA group, in women of reproductive age, as compared with surgical diagnosis.

METHODS

Protocol and registration

This systematic review and meta-analysis was designed according to the PRISMA statement14 and the Synthesizing Evidence from Diagnostic Accuracy Tests (SEDATE) guidelines on how to conduct a systematic review on diagnostic test accuracy15. All inclusion/exclusion criteria, data extraction and quality assessment were specified prior to commencement, and the protocol was registered prospectively with PROSPERO (CRD42017059872).

Eligibility criteria

Published, peer-reviewed studies that compared one or more imaging modality/ies, considered the ‘index test’, for the preoperative evaluation of the presence of DE, with surgical/histological diagnosis as the reference standard, were included. The gold-standard surgical diagnosis of endometriosis was made if any of the following criteria were satisfied: (1) histological confirmation of endometriosis in at least one resected subperitoneal nodule; (2) visualization and palpation of a subperitoneal nodule without biopsy and another histologically proven location of endometriosis; and (3) visualization of complete obliteration of the POD16, although it is important to note that this is not always due to rectosigmoid DE, as adhesions may be secondary to adhesions from uterosacral ligament DE to the rectum, superficial bowel endometriosis to the uterus or previous surgeries, among other causes. Studies were included if they were prospective cohort studies including women of reproductive age presenting with clinical suspicion of DE based on symptoms and/or physical examination from any healthcare-center setting.

All types of imaging modalities for DE of the rectum/rectosigmoid, including variations of conventional techniques, such as bowel preparation (BP) or rectal water or gel contrast media, were included, the outcome being the presence and position of DE. Imaging modalities were assessed overall and separately for different variations of the technique. Studies were included only if sufficient data were reported to construct a 2 × 2 contingency table. Only studies including at least 10 affected and 10 unaffected women, as determined by the reference standard, were considered eligible in order to reduce the risk of selection bias. There were no language restrictions.

Information sources

Searches were conducted using PubMed, MEDLINE, Scopus, EMBASE and Google Scholar to identify studies published from inception until May 2020. No filters were applied to the databases in order to reduce the possibility of omitting relevant studies17. Additionally, the reference lists of included studies and relevant reviews were searched manually to identify additional studies.

Search

The databases were searched using the following terms: (deep AND endometriosis) AND (imaging OR ultrasound OR ultrasonography OR sonography OR (magnetic resonance) OR (shift imaging) OR (proton spin) OR (spin echo) OR MRI OR NMR OR MR OR (computed tomography) OR (computer assisted tomography) OR (beam tomography) OR (Computerized Axial Tomography) OR CT OR CAT) (Appendix S1). The identified records were then screened for studies that specifically assessed rectal/rectosigmoid DE to ensure that studies that may have adopted outdated or inaccurate DE descriptions were not excluded.

Study selection

The records were screened based on their title and abstract by one author (B.G.), who obtained the full texts of those considered to be potentially eligible. Two authors (B.G. and G.C.), blinded to the other's assessments, independently examined the full texts for compliance with the inclusion criteria and selected eligible studies. Study investigators of potentially eligible articles were contacted as required in order to clarify study eligibility. When a study updated a previous publication, which was performed in either part or all of the same study population, the most recent and complete study was used to avoid duplicate data, assuming that some, if not all, of the previous study participants would be included. Similarly, for interobserver agreement studies in which two independent reviewers performed evaluations, only the most senior and accurate reviewer's results were included. We excluded conference abstracts, case reports/series, descriptive or systematic reviews and studies that were retrospective in nature, whether this applied to the study design, review of the case notes or the index test(s) being performed following the reference test. Disagreements were resolved by consulting a third author (M.L.). The selection process was documented using a PRISMA flowchart14.

Data items

Data were extracted by one author (B.G.), including the name of the first author, year of publication, country, period of enrolment, setting, age (mean and SD), presenting symptoms (dysmenorrhea, chronic pelvic pain, dyspareunia, bowel symptoms, urinary symptoms and infertility), imaging method(s), location of DE and the results of the 2 × 2 table (true positives, false negatives, true negatives, false positives).

Risk of bias and quality assessment

Risk of bias, applicability and methodological quality of each study were evaluated independently by two authors (B.G. and M.L.), as suggested by QUADAS-218, including the following four domains: (1) patient selection; (2) index text; (3) reference standard; and (4) flow and timing (risk of bias only) (Appendix S2). A summary score to calculate the overall quality for each study was not performed19.

In view of the difficulty of detecting and correcting for publication and other reporting biases, we tried to minimize their potential impact by ensuring a comprehensive search for eligible studies and by being alert to duplication of data.

Synthesis of results

We performed mixed-effects diagnostic meta-analysis to determine the overall pooled sensitivity and specificity of each imaging modality for rectal/rectosigmoid DE, from which we calculated the likelihood ratio of a positive (LR+) and negative (LR–) test20, diagnostic odds ratios (DOR) and areas under summary receiver-operating-characteristics (sROC) curves (AUC) with their respective 95% CIs. This method requires at least four studies for each meta-analysis15. LR+ and LR– characterize the utility of a diagnostic modality based on the pretest probability of disease. The DOR measures test accuracy by combining both likelihood ratios (DOR = LR+/LR–), and represents how much greater the odds are of having rectal/rectosigmoid DE in a patient with an abnormal diagnostic test than they are in a patient with a normal test. The AUC is a measure of the ability of the imaging technique to classify correctly those with and those without rectal/rectosigmoid DE. Forest plots of sensitivity and specificity were produced for diagnostic modalities with a sufficient number of studies for analysis. sROC curves were plotted to illustrate the AUC and the relationship between sensitivity and specificity. Subgroup analysis, where possible, was performed using the same methods.

We evaluated the presence and magnitude of heterogeneity for sensitivity and specificity using Cochran's Q and the I2 index. P < 0.1 on Cochran's Q-test suggests the presence of heterogeneity. The I2 index describes the percentage of the total variation across studies that can be explained by heterogeneity rather than chance. I2 values of 25%, 50% and 75% were considered to indicate low, moderate and high heterogeneity, respectively21. Heterogeneity was also visualized by the 95% prediction interval on the sROC plot. Subgroup analysis was planned for two-dimensional (2D)-TVS and 2D-MRI if there was a sufficient number of studies.

We assessed potential publication bias using the Deeks funnel plot asymmetry test, which computes a regression of diagnostic log odds ratio against 1/root (effective sample size), weighted by the effective sample size. P < 0.10 for the slope coefficient suggests significant asymmetry and possible publication bias22. All analyses were performed using Stata version 16.1 for Windows (StataCorp., College Station, TX, USA).

RESULTS

Search results

The literature search identified 1979 references. The study selection is summarized in Figure 1. There were 1061 duplicates, and a further 591 studies were excluded as they were identified as not relevant by screening the titles. A further 194 studies were excluded after screening the abstracts, as either they were retrospective or they did not answer the research question. The full texts of the remaining 133 articles were reviewed, of which 88 were excluded as they did not meet the inclusion criteria or had overlapping study populations. Of the remaining 45 eligible studies, 3023-52 specifically assessed the rectum/rectosigmoid and were included in the analysis. A list of excluded studies is provided in Appendix S3.

Details are in the caption following the image
Flowchart of studies included in systematic review on different imaging modalities for the preoperative diagnosis of rectal/rectosigmoid deep endometriosis (DE).

The 30 studies included a total of 3374 patients with a median of 92 women per study (range, 23–376; interquartile range, 56.5–138.25). Of the 30 studies, 22 were conducted in Europe, five in South America, one in Australia, one in Asia and one in the Middle East. A total of 21 (2857 patients, with 3055 examinations included in the analysis due to some studies comparing more than one TVS technique in the same patients37, 50) studies assessed TVS, of which 13 (2002 patients) used standard 2D-TVS24, 25, 28, 36-38, 40-42, 44, 45, 48, 50, five (502 patients) assessed TVS with rectal water contrast (RWC)27, 30, 34, 51, 52, one used TVS with gel contrast43, one assessed three dimensional (3D)-TVS37 and three assessed TVS with BP23, 49, 50. A total of seven studies assessed MRI (852 patients), of which one compared standard 2D- and 3D-MRI29, one compared MRI with and without rectal ultrasound gel39 and the remainder assessed standard 2D-MRI only (with standard 2D-MRI assessed in 813 patients overall)23, 24, 28, 32, 37. Six studies (402 patients) assessed CT, of which three assessed standard CT31, 47, 51 and three assessed CT colonography25, 27, 34. Eight (850 patients) studies assessed RES24, 26, 28, 30, 33, 35, 44, 46 and three used double-contrast barium enema (DCBE)32, 46, 52. The pretest probabilities of disease in patients undergoing TVS, MRI, CT and RES were 36%, 40%, 62% and 36%, respectively. The study characteristics are shown in Table S1 and the summary findings are shown in Table 1.

Table 1. Summary pooled results of preoperative diagnostic accuracy of different imaging modalities for detection of rectal/rectosigmoid deep endometriosis
Imaging technique Studies (n) Patients (n) Sensitivity (%) (95% CI) Specificity (%) (95% CI) DOR (95% CI) AUC (95% CI)
TVS
Overall 21 2857* 89 (83–92) 97 (95–98) 264 (113–614) 0.98 (0.96–0.99)
Standard 2D-TVS 13 2002 84 (76–90) 97 (94–99) 168 (56–501) 0.97 (0.95–0.98)
TVS-RWC 5 502 88 (80–93) 97 (93–98) 214 (80–572) 0.98 (0.96–0.99)
MRI
Overall 7 852 86 (79–91) 96 (94–97) 144 (70–297) 0.97 (0.95–0.98)
Standard 2D-MRI 7 813 85 (77–90) 96 (93–97) 125 (64–243) 0.96 (0.94–0.98)
CT 6 402 93 (84–97) 95 (81–99) 280 (28–2826) 0.98 (0.96–0.99)
RES 8 850 92 (87–95) 98 (96–99) 455 (196–1054) 0.99 (0.97–0.99)
  • * Corresponding to 3055 examinations owing to some studies performing more than one transvaginal ultrasound (TVS) technique in the same patients37, 50.
  • 2D, two-dimensional; AUC, area under receiver-operating-characteristics curve; CT, computed tomography; DOR, diagnostic odds ratio; MRI, magnetic resonance imaging; RES, rectal endoscopic sonography; RWC, rectal water contrast.

Methodological quality of included studies

The quality of the studies according to QUADAS-2 is presented in Figures 2 and S1. The majority of studies were of poor methodological quality.

Details are in the caption following the image
QUADAS-2 (Quality Assessment of Diagnostic Accuracy Studies-2) quality evaluation of 30 included studies. image, low risk; image, unclear risk; image, high risk.

Nineteen studies were assessed to be at low risk of patient selection bias23, 24, 26, 30, 32, 33, 36-42, 45-50, eight were high risk25, 28, 29, 31, 34, 35, 43, 44 and three had an unclear risk27, 51, 52. For the index test domain, regarding how the index test (i.e. preoperative imaging) was conducted and interpreted, 20 studies were deemed low risk23, 26-29, 31-37, 39-42, 44, 46, 48, 49, eight were high risk24, 25, 30, 38, 45, 50-52 and two had an unclear risk43, 47. All studies were considered to be at high risk of bias for the reference standard domain, as surgeons were not blinded to the results of the index test. With regard to the flow and timing domain, three studies were deemed to be high risk26, 30, 49, eight had an unclear risk24, 28, 29, 33, 35, 40, 43, 48 and the remaining 19 were low risk23, 25, 27, 31, 32, 34, 36-39, 41, 42, 44-47, 50-52. All the studies were considered to be at low risk for biases concerning applicability, given that the studies were included only if they had (1) a clinically relevant population that would have undergone the index test in real practice, (2) sufficient information on the index test, with all imaging modalities included and (3) laparoscopy/laparotomy as a reference test (Appendix S2).

Performance of TVS for detection of rectal/rectosigmoid DE

The overall pooled sensitivity, specificity, LR+, LR– and DOR for the detection of rectal/rectosigmoid DE using TVS were, respectively, 89% (95% CI, 83–92%), 97% (95% CI, 95–98%), 30.8 (95% CI, 17.6–54.1), 0.12 (95% CI, 0.08–0.17) and 264 (95% CI, 113–614). There was high heterogeneity for sensitivity (I2, 86.1%; Cochran's Q, 158.0; P < 0.01) and specificity (I2, 82.8%; Cochran's Q, 127.7; P < 0.01) (Figure 3). The sROC curve is displayed in Figure 4a, with an AUC of 0.98 (95% CI, 0.96–0.99). There was no evidence of publication bias (P = 0.86) (Figure S2).

Details are in the caption following the image
Forest plots of sensitivity and specificity of transvaginal ultrasound for detection of rectal/rectosigmoid deep endometriosis.
Details are in the caption following the image

Summary receiver-operating-characteristics curves (image) for transvaginal ultrasound (a), magnetic resonance imaging (b), computed tomography (c) and rectal endoscopic sonography (d) in the detection of rectal/rectosigmoid deep endometriosis. Circles indicate observed data. image, Summary operating point; image, 95% confidence contour; image, 95% prediction contour. (a) Summary operating point: sensitivity, 0.89 (95% CI, 0.83–0.92); specificity, 0.97 (95% CI, 0.95–0.98); area under curve (AUC), 0.98 (95% CI, 0.96–0.99). (b) Summary operating point: sensitivity, 0.86 (95% CI, 0.79–0.91); specificity, 0.96 (95% CI, 0.94–0.97); AUC, 0.97 (95% CI, 0.95–0.98). (c) Summary operating point: sensitivity, 0.93 (95% CI, 0.84–0.97); specificity, 0.95 (95% CI, 0.81–0.99); AUC, 0.98 (95% CI, 0.96–0.99). (d) Summary operating point: sensitivity, 0.92 (95% CI, 0.87–0.95); specificity, 0.98 (95% CI, 0.96–0.99); AUC, 0.99 (95% CI, 0.97–0.99).

Subgroup analyses of TVS

The pooled sensitivity, specificity, LR+, LR– and DOR for the detection of rectal/rectosigmoid DE using standard 2D-TVS were, respectively, 84% (95% CI, 76–90%), 97% (95% CI, 94–99%), 27.2 (95% CI, 12.7–58.5), 0.16 (95% CI, 0.10–0.26) and 168 (95% CI, 56–501). There was high heterogeneity for sensitivity (I2, 86.2%; Cochran's Q, 87.0; P < 0.01) and specificity (I2, 86.1%; Cochran's Q, 86.4; P < 0.01) (Figure S3). The sROC curve is displayed in Figure S4, with an AUC of 0.97 (95% CI, 0.95–0.98).

The pooled sensitivity, specificity, LR+, LR– and DOR for the detection of rectal/rectosigmoid DE using TVS-RWC were, respectively, 88% (95% CI, 80–93%), 97% (95% CI, 93–98%), 27.0 (95% CI, 12.8–56.7), 0.13 (95% CI, 0.08–0.21) and 214 (95% CI, 80–572). There was moderate heterogeneity for sensitivity (I2, 68.7%; Cochran's Q, 12.8; P = 0.01) and specificity (I2, 58.3%; Cochran's Q, 9.6; P = 0.05) (Figure S3). The sROC curve is displayed in Figure S4, with an AUC of 0.98 (95% CI, 0.96–0.99).

There was no evidence of publication bias for the analysis of standard 2D-TVS or TVS-RWC (P = 0.68 and P = 0.23, respectively) (Figure S2).

Given the low number of studies, it was not possible to perform subanalyses for TVS-BP or gel-contrast TVS. However, the results of TVS for the detection of rectal/rectosigmoid DE were improved with BP, with a sensitivity and specificity of 97–100% and 96–100%, respectively23, 49, 50. Based on the only study assessing gel-contrast TVS, it outperformed standard 2D-TVS for the detection of rectosigmoid DE, with a sensitivity and specificity of 100% and 93%, respectively; however, the performance was similar between the two modalities with regard to the rectum, which appears to be included in the study's definition of the retrocervix, with sensitivity of 84% and specificity of 96%43.

Performance of MRI for detection of rectal/rectosigmoid DE

The overall pooled sensitivity, specificity, LR+, LR– and DOR for the detection of rectal/rectosigmoid DE using MRI were, respectively, 86% (95% CI, 79–91%), 96% (95% CI, 94–97%), 21.0 (95% CI, 13.4–33.1), 0.15 (95% CI, 0.09–0.23) and 144 (95% CI, 70–297). There was moderate heterogeneity for sensitivity (I2, 61.0%; Cochran's Q, 20.5; P = 0.01) and low heterogeneity for specificity (I2, 34.0%; Cochran's Q, 12.1; P = 0.15) (Figure 5). The sROC curve is displayed in Figure 4b, with an AUC of 0.97 (95% CI, 0.95–0.98). There was no evidence of publication bias (P = 0.98) (Figure S5).

Details are in the caption following the image
Forest plots of sensitivity and specificity of magnetic resonance imaging for detection of rectal/rectosigmoid deep endometriosis.

Subgroup analyses of MRI

The pooled sensitivity, specificity, LR+, LR– and DOR for the detection of rectal/rectosigmoid DE using standard 2D-MRI were, respectively, 85% (95% CI, 77–90%), 96% (95% CI, 93–97%), 19.9 (95% CI, 12.9–30.7), 0.16 (95% CI, 0.11–0.24) and 125 (95% CI, 64–243). There was moderate heterogeneity for sensitivity (I2, 62.7%; Cochran's Q, 16.1; P = 0.01) and low heterogeneity for specificity (I2, 35.8%; Cochran's Q, 9.3; P = 0.16) (Figure S6). The sROC curve is displayed in Figure S7, with an AUC of 0.96 (95% CI, 0.94–0.98). There was no evidence of publication bias (P = 0.74) (Figure S5).

Given the low number of studies, it was not possible to perform subanalyses for 3D-MRI or MRI with rectal ultrasound gel. In a study assessing interobserver agreement, 3D-MRI performed similarly to standard 2D-MRI for the detection of rectosigmoid DE, with sensitivity and specificity between radiologists ranging from 89% to 100% and 94% to 100%, respectively29. 2D-MRI with rectal ultrasound gel outperformed standard 2D-MRI, with a sensitivity of 99% and specificity of 96%39.

Performance of CT for detection of rectal/rectosigmoid DE

The overall pooled sensitivity, specificity, LR+, LR– and DOR for the detection of rectal/rectosigmoid DE using CT were, respectively, 93% (95% CI, 84–97%), 95% (95% CI, 81–99%), 20.3 (95% CI, 4.3–94.9), 0.07 (95% CI, 0.03–0.19) and 280 (95% CI, 28–2826). There was high heterogeneity for sensitivity (I2, 84.3%; Cochran's Q, 31.9; P < 0.01) and specificity (I2, 80.2%; Cochran's Q, 25.2; P < 0.01) (Figure 6). The sROC curve is displayed in Figure 4c, with an AUC of 0.98 (95% CI, 0.96–0.99). There was no evidence of publication bias (P = 0.10) (Figure S5).

Details are in the caption following the image
Forest plots of sensitivity and specificity of computed tomography for detection of rectal/rectosigmoid deep endometriosis.

It was not possible to perform subanalyses of CT colonography, although the results ranged widely, with poor performance in one study25, reporting a sensitivity and specificity of 68% and 67%, respectively, compared with values ranging from 93% to 95% and 87% to 93%, respectively, in two other studies27, 34.

Performance of RES for detection of rectal/rectosigmoid DE

The overall pooled sensitivity, specificity, LR+, LR– and DOR for the detection of rectal/rectosigmoid DE using RES were, respectively, 92% (95% CI, 87–95%), 98% (95% CI, 96–99%), 37.1 (95% CI, 21.1–65.4), 0.08 (95% CI, 0.05–0.14) and 455 (95% CI, 196–1054). There was low heterogeneity for sensitivity (I2, 42.8%; Cochran's Q, 12.2; P = 0.09) and specificity (I2, 13.0%; Cochran's Q, 8.0; P = 0.33) (Figure 7). The sROC curve is displayed in Figure 4d, with an AUC of 0.99 (95% CI, 0.97–0.99). There was no evidence of publication bias (P = 0.37) (Figure S5).

Details are in the caption following the image
Forest plots of sensitivity and specificity of rectal endoscopic sonography for detection of rectal/rectosigmoid deep endometriosis.

Performance of DCBE for detection of rectal/rectosigmoid DE

As there were only three studies assessing DCBE, it was not possible to perform a meta-analysis, although the reported sensitivity, specificity and positive and negative predictive values were, respectively, 84.7%, 93.7%, 98.0% and 62.5% in one study32, 88%, 54%, 78% and 70% in another study46 and 96.4%, 100%, 100% and 98% in the third study52.

DISCUSSION

Summary of evidence

In 2004, Bazot et al.53 correlated the ultrasound and surgical findings of deep pelvic endometriosis and, since then, there has been a considerable number of studies published assessing preoperatively the presence and various characteristics of DE. Of all the imaging modalities used for the preoperative diagnosis of DE, TVS is the most widely studied and is often used as the first-line imaging technique, given its accessibility and relatively low cost and invasiveness, although adequate training is required54. MRI, while more expensive, has the advantage of being able to assess peritoneal and extrapelvic lesions, as well as adhesions12. The results of this meta-analysis on the detection of rectal/rectosigmoid DE revealed a superiority, albeit marginal, of TVS over MRI, with an overall pooled sensitivity of 89%, pooled specificity of 97%, DOR of 264 and AUC of 0.98, while the respective values for MRI were 86%, 96%, 144 and 0.97. The results of subgroup analyses for standard 2D-TVS and, in particular, TVS-RWC, were comparable with those of TVS overall. Interestingly, while CT also performed well, with sensitivity, specificity, DOR and AUC of 93%, 95%, 280 and 0.98, respectively, RES outperformed all the techniques, with sensitivity, specificity, DOR and AUC of 92%, 98%, 455 and 0.99, respectively. However, while RES might be the preferred imaging modality for detecting rectal DE, it is more invasive and not as widely available or used in this context. Furthermore, it is not possible to assess the remainder of the pelvis for DE with RES, and women would therefore require multiple invasive investigations for a complete assessment if this technique were used.

Interpretation of results

Our results differ slightly from those of other meta-analyses. Nisenblat et al.13 assessed a total of 21 studies (2222 participants), in which the pooled overall sensitivity and specificity for rectosigmoid DE were, respectively, 90% and 96% for TVS (14 studies), 92% and 96% for MRI (six studies) and 91% and 96% for RES (four studies). Although similar to our results, the differences are probably due to our current meta-analysis including more studies, limited to only those with at least 10 affected and 10 unaffected patients. More recently, in 2019, Noventa et al.55 published a meta-analysis of only head-to-head studies and found TVS to be superior to MRI for the detection of rectosigmoid DE, with sensitivities of 85% and 83%, respectively, as well as RES, with sensitivities of 89% and 88%, respectively, while RES was superior to MRI, with sensitivities of 91% and 84%, respectively. However, they included retrospective studies, limited to those determined to be well-conducted. Moura et al.56 performed a meta-analysis of studies comparing MRI and TVS in the diagnosis of rectosigmoid DE in the same population and found that both modalities had similar sensitivity, specificity and AUC of, respectively, 90%/96%/0.95 for TVS and 90%/96%/0.90 for MRI. This was similar to the findings of the more recent meta-analysis by Pereira et al.8 on MRI and TVS for the detection of rectosigmoid DE, including comparisons between enhanced techniques, with a sensitivity and specificity of, respectively, 80%/94% for MRI and 82%/94% for TVS. The limitation of the latter two meta-analyses is the small number of included studies, being eight and 11, respectively, although they were both well-conducted and confirmed the high diagnostic accuracy of both TVS and MRI.

Strengths and limitations

A strength of the current systematic review is the inclusion of only prospective studies, thereby reducing any potential selection bias that may be associated with retrospective studies, regardless of their quality. In addition, we believe that ensuring in advance that all studies had at least 10 affected and 10 non-affected participants further reduced the risk of selection bias, making the results more applicable to the general population. Furthermore, the initial search was purposely broad to ensure that all potential studies were included, especially given the inconsistencies in the descriptions of rectal and rectosigmoid DE. Prior to publication of the IDEA consensus9, rectal/rectosigmoid DE had also been referred to as retrocervical, POD, posterior cul-de-sac or bowel DE. Therefore, by including all studies with any mention of ‘deep’ and ‘endometriosis’, the risk of any such studies being missed was minimized, although the potential for relevant studies being missed remains a limitation of this and any other similar systematic review. Moving forward, the introduction of the IDEA terms and definitions into ultrasound practice may help to unify internationally the anatomical classification of rectal and sigmoid endometriosis, which may in turn result in more meaningful meta-analyses of pooled data.

While we attempted to obtain answers with regard to whether different contrast media or BPs improve the performance of the assessed imaging modalities, subanalyses were not possible owing to the limited number of studies. There was also substantial heterogeneity and a high risk of bias in all the included studies, which decreases the quality of the evidence. The lack of blinding of surgeons to the preoperative imaging findings and the risk of misdiagnosis due to lack of expertise or histopathological findings are potential causes of bias. Furthermore, there is a risk of patient selection bias, given that the pretest probability of disease in this analysis ranged from 36% to 62%, which is significantly higher than the reported incidence of rectal/rectosigmoid DE, which ranges from 5.3% to 12%57, 58. This is probably due to the inevitable self-selection of women referred to specialized units in which they are examined by experienced clinicians, who would be more attuned and consequently likely to make the correct diagnosis of rectal/rectosigmoid DE. All of these are potential sources of bias that need to be considered. Furthermore, the included studies span from 1998 to 2020, during which time there have been significant advances in technology as well as improvements in techniques and expertise, thus it is possible that the accuracy of the assessed imaging modalities may currently be higher than that determined in our analysis.

Conclusions

All the assessed imaging modalities perform well for the preoperative diagnosis of rectal/rectosigmoid DE, although there are key differences with respect to their applicability. The sensitivity of TVS for the detection of rectal/rectosigmoid DE seems to be slightly better than that of MRI and inferior to that of RES; however, it is not possible to assess the remainder of the pelvis using RES. The specificity of both MRI and TVS was excellent. As TVS is simpler, faster and more readily available, with a robust guideline (IDEA consensus) to standardize test performance that increases its reproducibility worldwide, we believe that it should be the first-line diagnostic tool for women with suspected DE. More studies are required using standardized definitions such as those proposed by the IDEA group, as well as comparing the various contrast media that may improve the accuracy of the different imaging modalities.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.