Accuracy of transvaginal sonography and rectal endoscopic sonography in the diagnosis of deep infiltrating endometriosis

Authors


Abstract

Objectives

To compare the accuracy of transvaginal sonography (TVS) and rectal endoscopic sonography (RES) for the diagnosis of deep infiltrating endometriosis (DIE), with respect to surgical and histological findings.

Methods

This was a longitudinal study of 81 consecutive patients referred for surgical management of DIE, who underwent both TVS and RES preoperatively. The diagnostic criteria were identical for TVS and RES, and were based on visualization of hypoechoic areas in specific locations (uterosacral ligaments, vagina, rectovaginal septum and intestine). We calculated the sensitivity, specificity, predictive values and accuracy of TVS and RES for the diagnosis of DIE.

Results

Endometriosis was confirmed histologically in 80/81 (98.7%) patients. Endometriomas and DIE were present in 43.2% and 97.5% of the women, respectively. For the diagnosis of DIE overall, TVS and RES, respectively, had a sensitivity of 87.3% and 74.7%, a positive predictive value of 98.6% and 98.3%, and an accuracy of 86.4% and 74%. For the diagnosis of uterosacral endometriosis, they had a sensitivity of 80.8% and 46.6%, a specificity of 75% and 50.0%, a positive predictive value of 96.7% and 89.5% and a negative predictive value of 30% and 9.3%. For the diagnosis of intestinal endometriosis, they had a sensitivity of 92.6% and 88.9%, a specificity of 100% and 92.6%, a positive predictive value of 100% and 96% and a negative predictive value of 87% and 80.6%.

Conclusion

TVS is apparently more accurate than is RES for predicting DIE in specific locations, and should thus be the first-line imaging technique in this setting. Copyright © 2007 ISUOG. Published by John Wiley & Sons, Ltd.

Introduction

Endometriosis is defined as the presence of endometrial tissue outside the endometrium and the myometrium1, the most common locations being the ovaries and the pelvic peritoneum, followed by deep infiltrating sites2. Deep infiltrating endometriosis (DIE) is defined by the presence of endometrial implants, fibrosis and muscular hyperplasia under the peritoneum, and can involve, in descending order of frequency, the uterosacral ligaments, the rectosigmoid colon, the vagina and the bladder2, 3. Physical examination has a limited capacity to diagnose and quantify DIE4, 5, and transvaginal (TVS), transrectal (TRS) or rectal endoscopic (RES) sonography as well as magnetic resonance imaging (MRI) have all been recommended for its diagnosis and for determining its location6–14.

TVS has been recommended for the diagnosis of ovarian endometriosis15, 16 and bladder endometriosis17. We recently reported that TVS permits accurate diagnosis of intestinal and bladder endometriosis but that it is less reliable for uterosacral, vaginal and rectovaginal septum involvement11, 12. RES with a high-frequency probe is more widely used compared with TRS because it provides an overview of the rectosigmoid colon. Several reports have suggested RES to be the best imaging technique for identifying rectovaginal, uterosacral and intestinal endometriosis7, 9, 18. However, in a preliminary study, we found that TVS was as efficient as was RES for detecting intestinal and uterosacral endometriosis11.

The aims of this study were to compare the accuracy of TVS and RES for the diagnosis of DIE in a large series of women with surgical and histological documentation.

Methods

This longitudinal study involved 81 consecutive women referred for surgical management of DIE between September 2000 and September 2004. All the women underwent both TVS and RES before surgery. Thirty patients from a previous study were included in this series11. The patients' demographic and clinical characteristics, and their previous history of medical and/or surgical treatment, are shown in Table 1. DIE was removed under laparoscopic guidance in 69 women and via laparotomy in 12.

Table 1. Characteristics of the 81 patients with pelvic endometriosis
Characteristicn (%) or median (range)
  1. GnRH, gonadotropin releasing hormone.

Age (years)31.9 (20–51)
Previous surgery for endometriosis18 (22.2)
Nulliparous62 (76.5)
GnRH analog treatment before surgery39 (48.1)
Infertility16 (19.7)
Dysmenorrhea68 (83.9)
Dyspareunia53 (65.4)
Dyschezia30 (37)

All ultrasound examinations were performed and interpreted in real time and videotaped for review. The sonographers were informed of the women's clinical history and symptoms but were blinded to the results of physical examination and previous imaging examinations. The physicians who interpreted the TVS examination were blinded to the results of the RES examination, and vice versa. Different physicians performed TVS and RES examinations. All potential locations of endometriosis were examined.

Transvaginal sonography

TVS was performed with an Ultramark HDI 5000 (ATL, Bothell, WA, USA) or a Siemens Elegra (Siemens, Erlangen, Netherlands) ultrasound machine, using a wide-band 5–9-MHz transducer. All scans were performed by the same radiologist (M.B.), who has extensive gynecological experience. No bowel preparation was used prior to sonography. In all cases the transducer was first positioned in the posterior cul-de-sac of the vagina and then slowly withdrawn through the vagina to assess the visualization of the posterior subperitoneal space (i.e. uterosacral ligaments, posterior fornix of the vagina, rectovaginal septum). By moving the probe up and down several times from the anal canal to the posterior fornix of the vagina, the bowel wall and the rectovaginal septum were analyzed. Rotation of the probe was also performed to detect posterior endometriotic lesions. Finally, the transducer was positioned in the anterior cul-de-sac of the vagina to analyze the vesicouterine septum.

Pelvic endometriosis was defined as at least one pelvic site of endometriosis (ovarian or deep infiltrating). DIE was diagnosed if at least one structure (uterosacral ligament(s), vagina, rectovaginal septum, rectosigmoid colon, or bladder) was involved. The diagnosis of DIE was based on morphological criteria that varied according to the anatomical location, and included abnormal hypoechoic linear thickening and nodules/masses with or without regular contours.

The uterosacral ligaments were considered to be involved when they were visible and bore a nodule (regular, or with stellate margins) or showed hypoechoic linear thickening with regular or irregular margins (Figure 1). When the involved uterosacral ligament was clearly delineated from adjacent structures, its thickness was measured in the proximal part, near the insertion on the cervix. The vagina was considered to be involved when the posterior vaginal fornix was thickened, with or without a round cystic anechoic area (Figure 2). The rectovaginal septum is a retroperitoneal structure located between the vagina and the rectum. It was considered to be involved when the normal hyperechoic aspect of the rectovaginal septum was replaced by a nodule or mass below the horizontal plane passing through the lower border of the posterior lip of the cervix (Figure 3). The rectum/sigmoid colon was considered to be involved when an irregular hypoechoic mass was found, with or without hypoechoic or hyperechoic foci, penetrating into the intestinal wall. In this case, the normal aspect of the rectum/sigmoid colon muscularis propria, which is hypoechoic and thin (< 3 mm), was replaced by the abnormal tissular mass. No attempt was made to evaluate the depth of endometriotic infiltration within the rectal wall (i.e. submucosa or mucosa). Lesions located on the sigmoid colon or at the rectosigmoid junction replaced the normal adipose tissue lying between the uterus and the rectum/sigmoid colon (Figure 4). Bladder involvement was diagnosed when a hypoechoic nodule and/or a cystic lesion was found within the bladder wall (Figure 5). The most frequently involved site was anterior to the vesicouterine pouch. Bladder involvement was distinguished from superficial vesicouterine endometriotic implants (< 1 cm) in the vesicouterine pouch without bladder wall involvement.

Figure 1.

Uterosacral ligament endometriosis: sagittal transvaginal view showing an irregular endometriotic nodule of the left uterosacral ligament (arrow and calipers) situated below the uterus (star).

Figure 2.

Vaginal endometriosis: sagittal transvaginal view showing a thickening of the posterior vaginal fornix (white arrow and calipers) located just behind the cervix (star), without rectovaginal septum involvement. Note the normal aspect of the posterior vaginal wall (black arrow and smaller calipers) below the vaginal endometriotic lesion.

Figure 3.

Rectovaginal septum endometriosis: sagittal view showing an endometriotic nodule of the upper part of the rectovaginal septum (RVS; arrow). RVS endometriosis is associated with endometriosis of the posterior vaginal fornix (star). Note the presence of fluid in the pouch of Douglas and a normal rectal wall (arrowhead).

Figure 4.

Rectal endometriosis: sagittal transvaginal view showing an irregular hypoechoic mass of the rectosigmoid junction (thick arrow) located just behind the cervix (star). Note the normal aspect of the muscularis propria (thin arrows) of the lower rectum and the sigmoid colon.

Figure 5.

Bladder endometriosis: sagittal transvaginal view showing a hypoechoic mass within the bladder wall (thick arrow) anterior to the vesicouterine pouch (thin arrow). The star indicates the uterus.

The largest diameter of the lesions and infiltration of adjacent organs were recorded. In addition to DIE, we recorded unilateral or bilateral ovarian involvement, defined as the presence of cysts with diffuse low-level internal echoes, multilocularity or hyperechoic wall foci19.

Rectal endoscopic sonography

RES was performed after a simple rectal enema, with an Olympus GF UM 20 Echo endoscope (SCOP Medicine Olympus, Rungis, France) with a diameter of 11.4 mm, operating at 7.5 and 12 MHz. All examinations were performed by two different highly experienced gastroenterologists (G.R. and P.A.). The procedure was performed under general anesthesia in 41 consenting women, and with only local (or topical) anesthetic drugs in 40 women. The transducer was always positioned in the sigmoid and was then slowly withdrawn through the sigmoid and rectum. Studies of the bowel wall and adjacent areas were carried out by moving the probe up and down several times before and after instilling water into the intestinal lumen. The main goal of RES was to evaluate rectal endometriosis. In addition, a systematic involvement of the uterosacral ligaments and the vagina was performed.

Infiltrating endometriosis was defined by the presence of a hypoechoic nodule or mass, with or without regular contours (Figure 6). The largest diameter of the lesions, their location with respect to the anus margins and infiltration of adjacent organs were assessed. In the rectum and/or sigmoid colon, involvement of the muscularis propria (hypoechoic and thin) was distinguished from that of the hyperechoic submucosa and mucosa. When possible, an attempt was made to evaluate the depth of infiltration by endometriosis within the rectal wall (i.e. submucosa or mucosa).

Figure 6.

Rectal endometriosis: rectal endoscopic sonography showing an endometriotic nodule involving the rectal wall (arrow).

Histological and surgical findings

The histological criteria used for the diagnosis of pelvic endometriosis included the presence of ectopic endometrial tissue (glands and stroma)1. DIE was diagnosed in the following circumstances: 1) if endometrial tissue (gland and stroma) was found on histological examination of at least one resected subperitoneal lesion3; 2) if DIE was visualized directly during laparoscopy or laparotomy but only fibrosis and smooth-muscle cells were found on biopsy or the lesion was not biopsied20 (in the latter case, subperitoneal endometriosis was diagnosed if another histologically proven site of endometriosis was found); 3) if complete cul-de-sac obliteration secondary to endometriosis was observed but could not necessarily be cleared surgically. Like Reich et al.21, we considered that deep retrocervical endometriosis was present beneath the peritoneum in such cases.

The largest diameter of the intestinal lesion was measured after colorectal resection. Infiltration of the propria muscularis, submucosa or mucosa of the rectosigmoid colon was recorded. As recommended by Chapron et al.22, a deep endometriotic location was considered isolated (either bladder, uterosacral ligaments, vagina or intestine) when it was not associated with any of the three other deep infiltrating endometriotic locations.

Statistical analysis

For each of the four possible locations of posterior DIE (uterosacral ligament, vagina, rectovaginal septum, and intestine), calculations were carried out with data from the group of women with a given location classified as disease, and with those from the group of women who did not have the location classified as non-disease. The sensitivity, specificity, positive and negative predictive values and accuracy of TVS and RES were evaluated for each site of endometriotic involvement.

Parametric and non-parametric continuous variables were compared using Student's t-test, and categorical variables were compared using the chi-square test, Fisher's exact test, the MacNemar test or the Z-statistic as appropriate. P < 0.05 was considered statistically significant.

Results

Surgical and pathological findings

Pelvic endometriosis found at surgery in at least one site (ovary or a site of deep infiltration) was confirmed by histology in 80 of the 81 (98.8%) patients. Ovarian endometriosis was present in 35/81 (43.2%) patients. Two patients with clinical and surgical signs of uterosacral ligament endometriosis did not fulfil all the histological criteria for DIE, but one of these two women had endometrioma confirmed histologically.

Seventy-nine women were considered to have DIE, as confirmed by surgery plus biopsy in 70/79 (88.6%) cases or by surgery alone in 9/79 (11.4%) cases. DIE was posterior in all 79 (100%) of the women and it was also anterior in three (3.8%) of these women. The distribution of the different sites of DIE is reported in Table 2. DIE was associated with endometriomas in 34 cases (41.9%).

Table 2. Location of deep infiltrating endometriosis diagnosed at surgery and confirmed histologically in 79 patients
Subperitoneal locationNumber of patients with:
Diagnosis at surgerySurgical specimen taken for biopsyHistological confirmationIsolated location of endometriosis
  1. NA, not applicable.

Posterior797270NA
 Uterosacral ligaments75656315
 Vagina2717171
 Rectovaginal septum9990
 Intestine5647453
  Sigmoid colon5552
  Rectosigmoid junction3830281
  Rectum1312120
Anterior3220
 Bladder3220

Among the 56 women with intestinal endometriosis, 47 underwent a bowel resection. Histology confirmed endometriotic involvement of the muscularis propria in 45 cases, the submucosa in 18 and the mucosa in two. Nine women did not undergo a bowel resection. In these cases, diagnosis of bowel involvement was based on both physical examination and imaging analysis.

Isolated DIE was detected in the vagina in one (1.2%) case, the intestine in three (3.8%) cases and the uterosacral ligaments in 15 (18.9%) cases.

Assessment of endometriosis by transvaginal sonography

The sensitivity, specificity, positive and negative predictive values and accuracy of TVS for the diagnosis of the different locations of pelvic endometriosis are shown in Table 3. TVS yielded a diagnosis of uterosacral ligament endometriosis in 61 (75.3%) patients; there were 14 false-negative and two false-positive cases. It yielded a diagnosis of vaginal endometriosis in 15 (18.5%) patients; there were 13 false-negative and two false-positive cases. All but one of the patients with false-negative findings had other deep infiltrating endometriotic locations diagnosed correctly by TVS. TVS yielded a diagnosis of rectovaginal septum endometriosis in one patient (1.2%); there were eight false-negative and no false-positive cases. All patients with false-negative results had other deep infiltrating endometriotic locations diagnosed correctly by TVS. It yielded a diagnosis of intestinal endometriosis in 50 (61.7%) patients; there were four false-negative and no false-positive cases. There was a strong correlation between the sites of involvement identified by TVS and those found at surgery. TVS diagnosed involvement of the rectosigmoid junction, lower rectum and sigmoid colon in 34 (51%), 11 (28.6%) and five (20.4%) cases, respectively. TVS yielded a diagnosis of bladder endometriosis in three (3.8%) patients. The mean size of bladder endometriotic lesions was 30 mm. It yielded a diagnosis of ovarian endometriosis in 40 (49.4%) patients; there were two false-negative and seven false-positive results.

Table 3. Assessment of pelvic endometriosis by transvaginal sonography in comparison to surgical and histological findings in 81 patients
SiteSensitivity (% (n))Specificity (% (n))PPV (% (n))NPV (% (n))Accuracy (% (n))+ LR− LR
  1. + LR, positive likelihood ratio; − LR, negative likelihood ratio; NPV, negative predictive value; PPV, positive predictive value; RVS, rectovaginal septum; USL, uterosacral ligament.

USL80.8 (59/73)75.0 (6/8)96.7 (59/61)30.0 (6/20)80.3 (65/81)3.20.25
Vagina50.0 (13/26)96.4 (53/55)86.7 (13/15)80.3 (53/66)81.5 (66/81)1.10.52
RVS11.1 (1/9)100 (72/72)100 (1/1)90.0 (72/80)90.1 (73/81)0.20.89
Intestine92.6 (50/54)100 (27/27)100 (50/50)87.1 (27/31)95.1 (77/81)0.17
Ovary94.3 (33/35)84.8 (39/46)82.5 (33/40)95.1 (39/41)88.9 (72/81)5.91.4

Assessment of endometriosis by rectal endoscopic sonography

The sensitivity, specificity, positive and negative predictive values and accuracy of RES for the diagnosis of the different locations of pelvic endometriosis are shown in Table 4. RES yielded a diagnosis of uterosacral ligament endometriosis in 38 (46.9%) patients; there were 39 false-negative and four false-positive cases. It yielded a diagnosis of vaginal endometriosis in three (3.75%) patients; there were 24 false negatives and one false-positive case. RES yielded a diagnosis of rectovaginal septum endometriosis in seven (8.6%) patients; there were seven false-negative and five false-positive cases. It yielded a diagnosis of intestinal endometriosis in 50 (61.7%) patients. In women who underwent colorectal resection, RES showed endometriotic involvement of the muscularis propria in 46/46 (100%) cases, the submucosa in 3/18 (16.7%) cases, and the mucosa in 0/2 (0%) cases. The distance between the lower part of the rectal involvement and the anal margin was measured in only 16/48 patients (33.3%) who underwent colorectal resection. RES yielded a diagnosis of ovarian endometriosis in 28 patients (34.6%), and gave 11 false-negative and four false-positive results.

Table 4. Assessment of pelvic endometriosis by rectal endoscopic sonography in comparison to surgical and histological findings in 81 patients
SiteSensitivity (% (n))Specificity (% (n))PPV (% (n))NPV (% (n))Accuracy (% (n))+ LR− LR
  1. + LR, positive likelihood ratio; − LR, negative likelihood ratio; NPV, negative predictive value; PPV, positive predictive value; RVS, rectovaginal septum; USL, uterosacral ligament.

USL46.6 (34/73)50.0 (4/8)89.5 (34/38)9.3 (4/43)46.9 (38/81)0.91.1
Vagina7.7 (2/26)98.2 (54/55)66.7 (2/3)69.2 (54/78)69.1 (56/81)0.9
RVS22.2 (2/9)93.1 (67/72)28.6 (2/7)90.5 (67/74)85.2 (69/81)3.20.8
Intestine88.9 (48/54)92.6 (25/27)96.0 (48/50)80.6 (25/31)90.1 (73/81)12.00.12
Ovary68.6 (24/35)91.3 (42/46)85.7 (24/28)79.2 (42/53)81.5 (66/81)9.00.4

Comparison of TVS and RES for the diagnosis of posterior deep infiltrating endometriosis

TVS and RES correctly diagnosed posterior DIE in 70/81 (86.4%) and 60/81 (74.1%) patients, respectively. The MacNemar test showed a significant difference in accuracy between TVS and RES in the diagnosis of DIE located in the uterosacral ligaments (P < 0.0001), vagina (P = 0.0008) and rectovaginal septum (P = 0.03). When the results of the two techniques were combined, there was one false-positive case and seven false-negative cases, all involving uterosacral endometriosis. TVS gave 11 false-negative diagnoses, of which four (36.4%) were diagnosed correctly by RES. RES gave 20 false-negative diagnoses, of which 13 (65.0%) were diagnosed correctly by TVS.

TVS and RES correctly diagnosed intestinal endometriosis in 50/54 (92.6%) and 48/54 (88.9%) cases, respectively. RES correctly diagnosed submucosal and mucosal rectal infiltration in 3/19 (15.8%) and 0/2 (0%) cases, respectively. Unlike TVS, RES was able to estimate the distance of rectal involvement from the anal margin (6–18 cm).

The largest macroscopic diameter of the endometriotic lesions ranged from 5 to 55 (mean, 31.1 ± 16.2) mm, compared with 5 to 50 (mean, 24.1 ± 11.4) mm with TVS and 4 to 45 (mean, 20.6 ± 8.3) mm for RES. Relative to histological measurements, the size of colorectal endometriotic lesions was estimated correctly by TVS (P < 0.05) and underestimated by RES.

The relative cross-sectional areas (mean ± SD) of endometriotic islands, smooth muscle and fibrotic components in deep endometriotic lesions were 24.2 ± 5.7%, 35.2 ± 15.5% and 40.5 ± 13.5%, respectively.

Discussion

This longitudinal study demonstrates the accuracy of TVS for the diagnosis of DIE. While TVS had certain limitations, especially for the diagnosis of vaginal endometriosis, RES was less accurate for identifying the different locations of DIE, with the exception of intestinal endometriosis.

Because medical treatment of DIE is not very effective22, the recommended treatment is currently full surgical excision of all endometriotic tissue20, 23–25. This treatment improves quality of life26, 27, but surgical complications can occur, especially in women with colorectal involvement. Accurate preoperative documentation of the precise site(s) of DIE is necessary in order to inform patients of the likely outcome.

We have shown that TVS detects DIE accurately (86.4% of cases), suggesting that TVS should be considered as the first-line imaging technique in this setting12. We confirm that TVS accurately diagnoses intestinal and bladder endometriosis11, 12. These results reinforce the fact that TVS should remain the first imaging technique for exploring pelvic disorders, allowing adequate evaluation of the whole pelvic cavity.

We found that TVS was low in sensitivity for vaginal involvement. This could be due in part to the characteristics of the probe. We used a broadband 5–9-MHz transducer, providing good deep ultrasound penetration but poorer visualization of anatomical structures adjacent to the probe. Sonovaginography has been reported to be a reliable and simple method for the assessment of vaginal endometriosis28.

When comparing the accuracy of TVS and RES, one must exclude anterior endometriotic locations that cannot be explored by RES. In addition, RES is not suitable for diagnosing endometrial ovarian cysts. Thus, we compared the accuracy of TVS and RES by focusing specifically on certain endometrial locations (uterosacral ligaments, vagina, rectovaginal septum and colorectum). This explains the difference that we found in sensitivity between TVS and RES for the diagnosis of endometriomas (94.3% and 63.6%, respectively), confirming previous reports showing that TVS is better than RES in this respect12, 29. The low sensitivity of RES for the diagnosis of endometriomas could be related to the characteristics of the probe used; we used one with a frequency range of 7.5–12 MHz, which permits adequate evaluation of only proximal organs and anatomical structures close to the probe.

Our results confirm that the uterosacral ligaments are the most frequent targets of DIE. In contrast to other pelvic locations of DIE, uterosacral ligament endometriosis is often isolated (19% of women in our series). In our study, the sensitivity (80.8%) and specificity (75%) of TVS for uterosacral ligament involvement were in line with our previous work11, 12. It is important to underline the low negative predictive value of TVS for uterosacral ligament involvement. MRI is useful for the diagnosis of uterosacral ligament endometriosis8, 13, but has not been compared with TVS, while a few preliminary studies of TRS followed by RES have assessed the accuracy of these methods. The low sensitivity of RES (46.6%) for uterosacral ligament involvement observed here is in keeping with the value reported by Delpy et al.29 of 42%. In our experience, RES is less accurate compared with TVS for diagnosing uterosacral ligament endometriosis.

Intestinal endometriosis is the most severe form of DIE. We confirm that the most frequent location is the rectosigmoid junction, representing nearly two-thirds of all colorectal sites2, 13. Moreover, in keeping with previous studies11, 27, 30, 31, we found that colorectal endometriosis was rarely isolated. Both TVS and RES readily visualized the rectosigmoid junction, located just behind the cervix. Previous studies have also shown that TVS, RES and TRS are highly accurate for the detection of rectal endometriosis6, 7, 9, 11, 12, 29. Various features of colorectal endometriosis can influence its surgical management, such as the degree of rectal wall infiltration, the size of the rectal lesions, and the distance from the anal margin. Previous studies supported segmental full, deep partial or superficial rectal resection, depending on the degree of endometriotic infiltration of the bowel wall. Only lesions involving at least the muscularis propria should be considered as intestinal endometriosis32. RES appears to be the best technique for evaluating the depth of bowel infiltration by endometriosis9, 18. This study confirms the results of our preliminary study suggesting that TVS and RES are similarly accurate for the diagnosis of rectal endometriosis11, despite two false-positive cases with RES. These two cases corresponded to patients with severe DIE involving torus uterinus, uterosacral ligaments and the pouch of Douglas with impaction of the rectum on the uterus. However, despite a better delineation of intestinal layers by RES, this technique was not more accurate than was TVS for determining which layers of the rectal wall were involved. Moreover, the value of this feature for choosing the surgical technique has recently been challenged33. Indeed, Remorgida et al.33 showed that segmental colorectal resection is the best surgical option for endometriosis, due to the risk of persistent lesions in almost half of the women who undergo full-thickness disc or superficial rectal resection. Both TVS and RES failed to reliably detect endometriosis of the vagina and rectovaginal septum. Again, MRI is useful in this setting13.

Several limitations of our study must be discussed. First, only patients with surgical evidence of DIE were included, so we cannot comment on the accuracy of TVS for diagnosing DIE in the general population. Furthermore, since the incidence of intestinal endometriosis is likely to be higher in our hospital than it is elsewhere (because removal under laparoscopic guidance is our treatment of choice), our results cannot necessarily be extrapolated to the general population. Second, the diagnosis of DIE was confirmed histologically in 88.6% of the patients, and the remaining diagnoses were based on macroscopic findings during surgery. Third, the high prevalence of DIE in our population rendered it difficult to evaluate positive and negative predictive values. In fact, according to Dart et al., the evaluation of positive and negative likelihood ratios34 would be preferable, but, to allow comparison with previous studies, we used predictive values. Finally, TVS and RES were performed by different highly experienced investigators (TVS by a radiologist, RES by a gastroenterologist), so interobserver variability between different levels of expertise was not evaluated.

In conclusion, TVS, contrary to RES, permits extensive exploration of the pelvis in women with suspected DIE. Moreover, TVS is well accepted and widely available. Although the higher probe frequency of RES offers better analysis of the different layers of the bowel wall, RES is no more effective than is TVS for the detection of rectal wall infiltration. Finally, RES sometimes necessitates general anesthesia, with its associated risks. Therefore, in clinical practice, our findings support the first-line use of TVS for the diagnosis of pelvic endometriosis. RES should be used when colorectal involvement is suspected on physical examination but not confirmed by TVS, but its utility relative to MRI needs to be evaluated.

Ancillary