{"paper_id":"366e421d-a466-41e9-b365-e4c09b12418a","body_text":"R E S E A R C H A R T I C L E Open Access\nUltrasound mapping of pelvic endometriosis:\ndoes the location and number of lesions affect\nthe diagnostic accuracy? a multicentre diagnostic\naccuracy study\nTom K Holland 1*, Alfred Cutner 2, Ertan Saridogan 2, Dimitrios Mavrelos 1, Kate Pateman 2 and Davor Jurkovic 2\nAbstract\nBackground: Endometriosis is a common condition which causes pain and reduced fertility. Treatment can be\ndifficult, especially for severe disease, and an accurate preoperative assessment would greatly help in the\nmanagment of these patients. The objective of this study is to assess the accuracy of pre-operative transvaginal\nultrasound scanning (TVS) in identifying the specific features of pelvic endometriosis and pelvic adhesions in\ncomparison with laparoscopy.\nMethods: Consecutive women with clinically suspected or proven pelvic endometriosis, who were booked for\nlaparoscopy, were invited to join the study. They all underwent a systematic transvaginal ultrasound examination in\norder to identify discrete endometriotic lesions and pelvic adhesions. The accuracy of ultrasound diagnosis was\ndetermined by comparing pre-operative ultrasound to laparoscopy findings.\nResults: 198 women who underwent preoperative TVS and laparoscopy were included in the final analysis. At\nlaparoscopy 126/198 (63.6%) women had evidence of pelvic endometriosis. 28/126 (22.8%) of them had\nendometriosis in a single location whilst the remaining 98/126 (77.2%) had endometriosis in two or more locations.\nPositive likelihood ratios (LR+) for the ultrasound diagnosis of ovarian endometriomas, moderate or severe ovarian\nadhesions, pouch of Douglas adhesions, and bladder deeply infiltrating endometriosis (DIE), recto-sigmoid colon\nDIE, rectovaginal DIE, uterovesical fold DIE and uterosacral ligament DIE were >10, whilst for pelvic side wall DIE\nand any ovarian adhesions the + LH was 8.421 and 9.81 respectively.\nThe negative likelihood ratio (LR-) was: <0.1 for bladder DIE; 0.1-0.2 for ovarian endometriomas, moderate or severe\novarian adhesions, and pouch of Douglas adhesions; 0.5-1 for rectovaginal, uterovesical fold, pelvic side wall and\nuterosacral ligament DIE. The accuracy of TVS for the diagnosis of both total number of endometriotic lesions and\nDIE lesions significantly improved with increasing total number of lesions.\nConclusions: Our study has shown that the TVS diagnosis of endometriotic lesion is very specific and false positive\nresults are rare. Negative findings are less reliable and women with significant symptoms may still benefit from\nfurther investigation even if TVS findings are normal. The accuracy of ultrasound diagnosis is significantly affected\nby the location and number of endometriotic lesions.\n* Correspondence: tomkholland@yahoo.com\n1Early Pregnancy and Gynaecology Assessment Unit, Department of\nObstetrics and Gynaecology, Suite 8, Golden Jubilee Wing, King ’s College\nHospital, London SE5 8RX, UK\nFull list of author information is available at the end of the article\n© 2013 Holland et al.; licensee BioMed Central Ltd. This is an open access article distributed under the terms of the Creative\nCommons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and\nreproduction in any medium, provided the original work is properly cited.\nHolland et al. BMC Women's Health 2013, 13:43\nhttp://www.biomedcentral.com/1472-6874/13/43\n\nBackground\nEndometriosis is a common gynaecological condition,\ndefined as the presence of endometrial-like tissue out-\nside the uterus, which impairs quality of life. In more se-\nvere cases it forms cysts in the ovaries and deeply\ninfiltrates pelvic organs.\nIn women with pelvic endometriosis ultrasound exam-\nination has been shown to be able to diagnose ovarian\nendometriomas with a high degree of accuracy, but\nother endometriotic lesions were considered to be un-\ndetectable [1]. Several recent studies have shown that a\ntargeted ultrasound examination using high resolution\nequipment can also detect deep infiltrating endometrio-\ntic (DIE) lesions, which are affecting other organs within\nthe lesser pelvis. The reported accuracy of the ultra-\nsound diagnosis of DIE varies between different studies,\nwhich may reflect the variations in the examination\ntechnique, quality of ultrasound equipment and experi-\nence of the operators. The prevalence of disease is also\nvariable in different studies, which may bias the findings.\nOnly a few studies have attempted to assess the ability\nof ultrasound examination to detect presence of pelvic\nadhesions in women with pelvic endometriosis and to\nassess their severity [2,3].\nThe detection of all endometriotic lesions within the\npelvis and assessment of the severity of adhesions are re-\nquired in order to assess the severity of endometriosis\nusing the standard revised ASRM classification and tri-\nage women for surgical treatment. Women with severe\ndisease and extensive adhesions could be thus referred\nto centres of excellence to ensure complete surgical\nexcision [4].\nThe objective of this study is to assess the accuracy of\npre-operative transvaginal ultrasound scanning (TVS) in\nidentifying the specific features of pelvic endometriosis\nand pelvic adhesions in comparison to laparoscopy.\nMethods\nThis was a prospective, observational, multicentre study,\nwhich was conducted at King ’s College Hospital and\nUniversity College Hospital in London. These are both\nmajor teaching hospitals and the latter has a specialist\ntertiary referral endometriosis centre. Consecutive women\nwith clinically suspected or proven pelvic endometriosis\nwere invited to join the study. The inclusion criteria\nwere: pre-menopausal women with a clinical suspicion\nof endometriosis awaiting diagnostic laparoscopy or\nwomen diagnosed with pelvic endometriosis at diagnos-\ntic laparoscopy awaiting operative treatment. Other cri-\nteria included age 16 or over and the ability to provide\ninformed consent. Women who could not undergo a\ntransvaginal ultrasound scan and those who became\npregnant whilst awaiting surgery were excluded from the\nstudy.\nThe study was ethically approved and an information\nleaflet was given to all eligible women before assessment.\nInformed consent was obtained from all women who\nagreed to take part in the study.\nProcedures\nAll women were assessed by the attending clinicians\nwho obtained a detailed history, which was recorded on\na dedicated clinical database (ViewPoint, GE Healthcare,\nFairfield, Connecticut, USA). Women were specifically\nasked about symptoms associated with endometriosis\nsuch as dysmenorrhoea, chronic pelvic pain, dyspar-\neunia, subfertility, dyschezia and cyclic rectal bleeding.\nTransvaginal ultrasound examination was performed\nby two ultrasound operators who were both gynaecolo-\ngists with a high level of expertise in gynaecological\nultrasonography. The ultrasound operators were blinded\nto any previous surgical findings. All women were oper-\nated on by four different laparoscopic surgeons with a\nhigh level of expertise in laparoscopic surgery. The find-\nings were recorded using the revised ASRM classifica-\ntion of the severity of endometriosis. When moderate,\nsevere or deeply invasive disease (DIE) was present a\ncomplete surgical exploration of the pelvis was per-\nformed. This involved dissection of the pouch of\nDouglas when obliterated and resection of any DIE in-\ncluding the RVS, so as not to miss any disease. The oper-\nating surgeons were blinded to the detailed transvaginal\nultrasound findings.\nTransvaginal ultrasound assessment of pelvic\nendometriosis\nAll women were examined in the dorsal lithotomy pos-\nition using a high resolution transvaginal ultrasound\nprobe. The examinations were performed in a standar-\ndised and systematic way. Firstly the uterus was assessed\nin the transverse and sagittal planes. Next the ovaries\nwere found and their size was measured in three orthog-\nonal planes.\nOvarian cysts were diagnosed as endometriomas when\nthey appeared as well circumscribed thick walled cysts\nwhich contained homogenous low level internal echos\n(“ground glass ”) [5]. Measurements were recorded from\nthe inside of the cyst wall in three orthogonal planes.\nThe average of the 3 diameters (D1 + D2 + D3)/3 was.\nThe adnexa were also systematically examined for the\npresence of tubal dilatation.\nOvarian mobility was assessed by a combination of\ngentle pressure with the vaginal probe and abdominal\npressure with the examiner ’s free hand as in a bimanual\nexamination. The ovary was deemed to be completely\nfree when all of its borders could be seen sliding across\nthe surrounding structures. Minimal adhesions were\nconsidered to be present when some of the surrounding\nHolland et al. BMC Women's Health 2013, 13:43 Page 2 of 9\nhttp://www.biomedcentral.com/1472-6874/13/43\n\nstructures could not be separated from the ovary with\ngentle pressure but the ovary could be mobilised from\nthe majority (approximately >2/3) of the surrounding\nstructures. Moderate adhesions were thought to be\npresent when the ovarian mobility was reduced due to\nadhesions with the surrounding structures but the struc-\ntures on 2/3-1/3 of the surface of the ovary were sliding\nacross it on gentle pressure. Fixed ovaries could not be\nmobilised at all with gentle pressure nor separated from\nthe surrounding structures. If the tubes were dilated, the\nmobility of the dilated tubes was documented in a simi-\nlar fashion. Normal fallopian tubes are difficult to iden-\ntify in the absence of background fluid in the pelvis and\ntherefore it was not possible to score non dilated tubes\nfor adhesions. It is difficult to see filmy adhesions on\nTVS unless there is fluid entrapped within the adhe-\nsions, giving rise to the “flapping sail sign ” [6], or unless\nthe mobility of the affected organs is reduced and there-\nfore these features were not scored separately at TVS.\nThe presence of adhesions in the pouch of Douglas\nwas assessed next. The uterus was gently mobilised by a\ncombination of pressure on the cervix with the ultra-\nsound probe alternating with pressure on the fundus\nfrom the examiners free hand on the abdominal wall.\nThe aim was to watch the interface of the posterior uter-\nine serosa and the bowel behind to ensure that the two\nstructures were sliding easily across one another. If these\ntwo surfaces were completely free of one another this\nwas assessed as no adhesions present. Complete obliter-\nation was assessed as the absence of any sliding between\nthe serosa on the posterior surface of the cervix or\nuterus and the bowel behind. Partial obliteration of the\npouch of Douglas was present if there were some adhe-\nsions between the bowel and the uterus but some free\nsliding was seen. Partial obliteration was also present\nwhen adnexal structures were firmly adherent to the\nposterior aspect of the uterus but the bowel appeared to\nbe free.\nEndometriotic nodules or deeply invasive endometri-\nosis (DIE) were typically visualised as stellate hypoechoic\nor isoechogenic solid masses with irregular outer mar-\ngins [7,8], which were tender on palpation and fixed to\nthe surrounding pelvic structures. They were usually\nlocated in the uterosacral ligaments, adnexa, rectovagi-\nnum, and urinary bladder. Endometriotic nodules lo-\ncated in the wall of the rectosigmoid colon tend to\nappear as hypoechoic thickenings of bowel muscularis\npropria, which sometimes protrude into the lumen of\nthe bowel [9]. Rectovaginal endometriosis is defined as\ndisease affecting the posterior pelvic compartment with\nevidence of endometriotic nodules which are located be-\ntween the rectum and posterior fornix of the vagina\nand/or posterior aspect of the cervix. The presence and\nlargest diameter of any deep lesions were documented.\nAll these findings were recorded on a database file\nusing a Microsoft Excel for Windows spreadsheet to\nfacilitate data entry and retrieval. The severity of endo-\nmetriosis as assessed by TVS was compared with laparo-\nscopic findings using the rASRM classification [10].\nStatistical analysis\nAll statistical analyses were carried out using Medcalc\nversion 9.2.0.2 (Medcalc Software, Mariakerke, Belgium).\nThe diagnostic accuracy of the tests was assessed using\nsensitivity, specificity, positive (PPV) and negative (NPV)\npredictive value, and positive (LR+) and negative (LR −)\nlikelihood ratio measures. Overall levels of agreement\nfor non binary data was calculated using Cohen ’s quad-\nratic weighted Kappa coefficient. Kappa values of 0.81-\n1.0 indicated very good agreement, Kappa values of\n0.61-0.80 good agreement, Kappa values of 0.41-0.60\nmoderate agreement, Kappa values of 0.21-0.40 fair agree-\nment and Kappa values <0.20 poor agreement [11,12].\nThe Kruskal-Wallis one-way analysis of variance was used\nto assess for statistical difference between rank sum of the\ngroups as the data was not normally distributed.\nResults\nFrom July 2006 to September 2009 we recruited 237\nwomen into this study. 39 women were excluded from\nthe final analysis: twenty nine because they were not\nassessed by one of the two designated ultrasound opera-\ntors, five became pregnant whilst awaiting surgery, one\ncancelled her operation, one laparoscopy was unsuccess-\nful and three women were lost to follow up.\nTable 1 The prevalence of endometriotic lesions at\ndifferent anatomical locations at laparoscopy\nSite of disease N (%)\nEndometrioma on either ovary 51/198 (25.7%)\nUnilateral 27/198 (13.6%)\nBilateral 24/198 (12.1%)\nModerate/severe adhesions on either ovary 78/198 (39.4%)\nUnilateral 30/198 (15.2%)\nBilateral 48/198 (24.2%)\nDIE of USL unilateral 8/198 (4.0%)\nDIE of USL bilateral 12/198 (6.1%)\nComplete obilteration of POD 30/198 (15.2%)\nPartial obilteration of POD 24/198 (12.1%)\nDIE of Rectum/Sigmoid 11/198 (5.6%)\nDIE of RVS 32/198 (16.2%)\nDIE of bladder 5/198 (2.5%)\nDIE of utero vesical fold (separate from bladder) 6/198 (3.0%)\nDIE of PSW unilateral 7/198 (3.5%)\nDIE of PSW bilateral 3/198 (1.5%)\nDIE is deeply infiltrating endometriosis, USL is uterosacral ligaments, POD is\npouch of Douglas, RVS is rectovaginal septum, PSW is pelvic side wall.\nHolland et al. BMC Women's Health 2013, 13:43 Page 3 of 9\nhttp://www.biomedcentral.com/1472-6874/13/43\n\n198 women were included in the final analysis. The\nmean age was 35.0 (95% CI 33.98 – 35.97, SD 7.10)\n(range 19 –50) years. The presenting symptoms were\ndysmenorrhoea for 143/198 (72.2%), chronic pelvic pain\nfor 98/198 (49.5%), dyspareunia for 91/198 (45.9%), in-\nfertility for 42/198 (21.2%), dyschezia for 19/198 (9.6%)\nand cyclic rectal bleeding for 3/198 (1.5%) women. A\nsingle presenting symptom was present in 72/198\n(36.4%) women, two presenting symptoms in 66/198\n(33.3%), three presenting symptoms in 39/198 (19.7%),\nfour or more symptoms in 19/198 (9.6%) women.\nAt laparoscopy 126/198 (63.6%) women had endomet-\nriosis. Of these women 30 /126 (23.8%) had stage 1\nendometriosis by the rASRM classification, 24/126\n(19.0%) had stage 2, 21/126 (16.7%) had stage 3 and\n51/126 (40%) had stage 4 disease. Of the 104 women with\nfocal lesions (excluding women with only diffuse superfi-\ncial peritoneal disease) 28/104 (26.9%) women had endo-\nmetriosis in a single location whilst the remaining 73.1%\nhad endometriosis in two or more locations.\nThe ultrasound examinations were performed by two\nexaminers: examiner A performed 104 (52.5%), examiner\nB 94 (47.5%). All women were operated on by one of\nfour laparoscopic surgeons: surgeon A operated on 79\n(39.9%), surgeon B on 54 (27.3%), surgeon C on 35\n(17.7%) and surgeon D on 30 (15.2%) women. The mean\ninterval between TVS and operation was 36.8 days (95%\nCI 33.4 – 41.1, SD 22.9) (range 0 –87 days).\nTable 1 shows the prevalence of the individual features\nof pelvic endometriosis at laparoscopy. Table 2 gives the\ndetails of the individual locations of endometriosis in\nrelation to whether they were isolated lesions or multifocal\nlesions. Of the 104 women with focal lesions (excluding\nwomen with only diffuse superficial peritoneal disease)\nTable 2 Isolated and multiple endometriotic lesions in\nrespect to their locations\nSite of disease Endometriosis of\na single location\nEndometriosis\nmultiple locations\nN (%) N (%)\nOvarian endometrioma n = 51 2/51 (3.9%) 49/51 (96.1%)\nOvarian adhesions n = 85 16/85 (18.8%) 69/85 (81.2%)\nAdhesions in POD n = 54 1/54 (1.9%) 53/54 (98.1%)\nUSL DIE n = 23 5/23 (21.7%) 18/23 (88.3%)\nRV or POD DIE n = 32 1/32 (3.1%) 31/32 (96.9%)\nDIE of rectum or sigmoid n = 9 0/9 (0%) 9/9 (100%)\nDIE of bladder n = 5 1/5 (20%) 4/5 (80%)\nDIE of UVF n = 6 1/6 (16.7%) 5/6 (83.3%)\nDIE of PSW n = 9 1/9 (11.1%) 8/9 (88.9%)\nTotal 28/104 (26.9%) 76/104 (73.1%)\nDIE is deeply infiltrating endometriosis, USL is uterosacral ligaments, POD is\npouch of Douglas, RVS is rectovaginal septum, UVF is utero vesical fold, PSW is\npelvic side wall. All bilateral structures were considered as one location\n(ovaries, USL, PSW).\nTable 3 Accuracy of pre-operative ultrasound diagnosis of endometriotic lesions affecting different pelvic organs\nSite of disease Sensitivity Specificity PPV NPV LR+ LR- Area under\nROC curve\nOvarian\nendometrioma\nN=7 5\n84.0 (95% CI\n73.7 – 91.4)\n95.6(95% CI\n92.8 – 97.6)\n81.8 (95% CI\n71.8 – 90.6)\n96.2 (95% CI\n93.5 – 97.8)\n19.26 (95% CI\n11.431 – 32.451)\n0.167 (95% CI\n0.10 – 0.281)\n0.898 (95% CI\n0.864 – 0.926)\nP = 0.0001\nDIE of bladder\nN=5\n100 (95% CI\n48.0 – 100)\n100 (95% CI\n98.1 – 100)\n100 (95% CI\n48.0 – 100)\n100 (95% CI\n98.1 – 100)\n∞ (95% CI 0- ∞) 0.00 (95% CI 0- ∞) 1.00 (95% CI\n0.981 – 1.00)\nP = 0.000\nDIE Rectum/\nSigmoid N = 9\n33.3 (95% CI\n12.1 – 64.6)\n98.9 (95% CI\n96.2 – 99.7)\n60 (95% CI\n23.1 – 88.2)\n96.9 (95% CI\n93.4 – 98.6)\n31.5 (95% CI\n5.992 – 165.6)\n0.674 (95% CI\n0.424 – 1.07)\n0.661 (95% CI\n0.591 – 0.727)\nP = 0.111\nRV DIE N = 32 50.0 (95% CI\n33.6 – 66.4)\n100 (95% CI\n97.7 – 100)\n100 (95% CI\n80.6 – 100)\n96.9 (95% CI\n93.4 – 98.6)\n∞ (95% CI 0- ∞) 0.50 (95% CI\n0.354 – 0.707)\n0.758 (95% CI\n0.692 – 0.816)\nP = 0.0001\nDIE of UVF\nN=6\n16.7 (95% CI\n2.8 – 63.9)\n99.0 (95% CI\n96.3 – 99.8)\n33.3 (95% CI\n6.1 – 79.2)\n97.4 (95% CI\n94.2 – 98.9)\n16.0 (95% CI\n1.68 – 153.94)\n0.84 (95% CI\n0.589 – 1.205)\n0.578 (95% CI\n0.506 – 0.648)\nP = 0.528\nDIE of PSW\nN=1 3\n15.4 (95% CI\n2.4 – 45.5)\n98.17 (95% CI\n96.3 – 99.3)\n22.2 (95% CI\n0.063 – 0.547)\n97.2 (95% CI\n95.0 – 98.4)\n8.421 (95% CI\n1.933 – 36.65)\n0.862 (95% CI\n0.683 – 1.087)\n0.568 (95% CI\n0.517 – 0.617)\nP = 0.419\nDIE of USL\nN=4 0\n10.0 (95% CI\n2.9 – 23.7)\n99.16(95% CI\n97.6 – 99.8)\n57.1 (95% CI\n25.0 – 84.2)\n90.7 (95% CI\n87.5 – 93.2)\n11.867 (95% CI\n2.754 – 51.14)\n0.908 (95% CI\n0.818 – 1.007)\n0.546 (95% CI\n0.495 – 0.596)\nP = 0.351\nPPV is positive predictive value, NPV is negative predictive value, +ve LH is positive likelihood ratio, -ve LH is negative likelihood ratio, ROC is re ceiver operating\ncharacteristics, DIE is deeply infiltrating endometriosis, RV is rectovaginal, UVF is utero vesical fold (separate from bladder), PSW is pelvic sid e wall, USL is\nuterosacral ligaments. (All bilateral anatomical locations were treated separately).\nHolland et al. BMC Women's Health 2013, 13:43 Page 4 of 9\nhttp://www.biomedcentral.com/1472-6874/13/43\n\n28/104 (26.9%) of these women had endometriosis in a\nsingle location whilst the remaining 73.1% had endometri-\nosis in two or more locations.\nOvarian endometriomas were rarely isolated lesions as\novarian adhesions were also present in 48/51 (94%) of\ncases. 27/51 (52.9%) women with endometriomas had uni-\nlateral and 24/51 (47.1%) had bilateral lesions. There was\nno significant difference in the frequency of endometrio-\nmas located in the right or left ovary (Chi-square =0.327\np = 0.51). Women with bilateral endometriomas were no\nmore likely to have associated DIE 16/24 (66.6%) com-\npared to women with unilateral endometriomas 14/27\n(51.8%) (Chi-square =0.621 p = 0.431 stat).\nDiagnostic accuracy of pre-operative TVS for each of\nthe specific anatomical locations of endometriosis is\nshown in Table 3. There was a significant difference be-\ntween the sensitivities for the different locations (Chi\nsquared = 74.97, P < 0.0001) while the specificities were\nsimilar (p > 0.05). The positive likelihood ratio (LR+) was\nvery useful (>10) for the TVS diagnosis of endometriosis\nof the following anatomical locations: ovarian endome-\ntriomas; moderate or severe ovarian adhesions; pouch of\nDouglas adhesions; and deeply infiltrating endometriosis\n(DIE) of the bladder; rectum or sigmoid; rectovaginum;\nuterovesical fold; and the uterosacral ligaments. Only for\npelvic side wall DIE and mild ovarian adhesions was the\nLR + moderately useful (5 –10). The negative likelihood\nratio (LR-) was very useful (<0.1) for bladder DIE and mod-\nerately useful (0.1-0.2) for ovarian endometriomas, moder-\nate or severe ovarian adhesions, and pouch of Douglas\nadhesions. The sensitivity was highest for bladder and\novarian endometriomas and lowest for DIE of the uterova-\nsical fold, pelvic side wall and uterosacral ligaments.\nThe LR + and –LR for all adhesions on the ovaries\nwere moderately and somewhat useful respectively.\nHowever for the assessment of moderate or severe adhe-\nsions on the ovary the LR + and –LR was very and mod-\nerately useful respectively as detailed in Table 4. When\nthe diagnosis of ovarian adhesions was stratified accord-\ning to the ASRM classification into mild, moderate and\nsevere the overall level of agreement between scan and\nlaparoscopy was very good (Table 5). The LR + and –LR\nfor adhesions in the pouch of Douglas were very and\nmoderately useful respectively as detailed in Table 4.\nWhen pouch of Douglas obliteration was assessed ac-\ncording to the ASRM classification into partial and\nTable 4 Accuracy of pre-operative ultrasound diagnosis of pelvic adhesions in women with suspected endometriosis\nSite of disease Sensitivity Specificity PPV NPV LR+ LR- Area under\nROC curve\nAny adhesions on\novary N = 130\n79.6 (95% CI\n72.0 – 85.5)\n91.9 (95% CI\n87.9 – 94.6)\n83.8 (95% CI\n76.6 – 89.2)\n89.5 (95% CI\n85.2 – 92.6)\n9.81 (95% CI\n6.456 – 14.92)\n0.222 (95% CI\n0.160 – 0.310)\n0.865 (95% CI\n0.827 – 0.897)\nP = 0.0001\nMod/Severe adhesions\non ovary N = 123\n83.7 (95% CI\n76.2 – 89.2)\n94.1 (95% CI\n90.7 – 96.4)\n86.6 (95% CI\n79.3 – 91.6)\n92.8 (95% CI\n89.1 – 95.3)\n14.288 (95% CI\n8.826 – 23.131)\n0.173 (95% CI\n0.116 – 0.258)\n0.889 (95% CI\n0.854 – 0.919)\nP = 0.0001\nSevere adhesions on\novary N = 103\n83.5 (95% CI\n75.1 – 89.4)\n93.5 (95% CI\n90.1 – 95.8)\n81.9 (95% CI\n73.5)\n94.2 (95% CI\n90.8 – 96.3)\n12.876 (95% CI\n8.266 – 20.057)\n0.176 (95% CI\n0.114 – 0.273)\n0.867 (95% CI\n0.830 – 0.899)\nP = 0.0001\nAny adhesions in\nPOD N = 54\n83.3 (95% CI\n71.3 – 91.0 )\n95.1 (95% CI\n90.3 – 97.6)\n86.5 (95% CI\n74.7 – 93.3)\n93.8 (95% CI\n88.7 – 96.7)\n17.143 (95% CI\n8.242 – 35.656)\n0.175 (95% CI\n0.096 – 0.318)\n0.892 (95% CI\n0.841 – 0.932)\nP = 0.0001\nComplete obliteration\nof POD N = 30\n83.3 (95% CI\n66.4 – 0.927)\n97.0 (95% CI\n93.2 – 98.7)\n83.3 (95% CI\n66.4 – 92.7)\n97.0 (95% CI\n93.2 – 98.7)\n28.0 (95% CI\n11.636 – 67.376)\n0.172 (95% CI\n0.077 – 0.383)\n0.902 (95% CI\n0.852 – 0.939)\nP = 0.0001\nPPV is positive predictive value, NPV is negative predictive value, +ve LH is positive likelihood ratio, -ve LH is negative likelihood ratio, ROC is re ceiver operating\ncharacteristics, POD is pouch of Douglas.\nTable 5 Comparison of ultrasound and laparoscopy for the assessment of severity of ovarian adhesions\nTVS assessment of ovarian adhesions\nLaparoscopic assessment of ovarian adhesions Absent Minimal Moderate Severe Total\nAbsent 238 6 5 10 259 (65.4%)\nMinimal 10 3 0 1 14 (3.5%)\nModerate 7 1 4 8 20 (5.1%)\nSevere 11 1 5 86 103 (26.0%)\nTotal 266 (67.2%) 11 (2.8%) 14 (3.5%) 105 (26.5%) 396\nWeighted Kappa = 0.801 (standard error (Kw ’ = 0) 0.050 and (Kw ’#0) 0.031).\nHolland et al. BMC Women's Health 2013, 13:43 Page 5 of 9\nhttp://www.biomedcentral.com/1472-6874/13/43\n\ncomplete obliteration the overall level of agreement be-\ntween scan and laparoscopy was very good (Table 6).\nTable 7 shows that the accuracy of the diagnosis of DIE\nincreases significantly with the total number of endome-\ntriotic lesions present. This data is represented graphic-\nally in Figure 1. Table 8 shows that although the number\nof endometriotic lesions seen on scan significantly in-\ncreases with the number of lesions present (Figure 2) the\nproportion of the total lesions correctly diagnosed in-\ncreases to a maximum at three lesions present at lapar-\noscopy then declines (Figure 3).\nDiscussion\nOur study has shown that pre-operative transvaginal\nultrasound examination can be used to diagnose pelvic\nendometriosis and to assess its severity. The total num-\nber of endometriotic lesions found at laparoscopy has\nstatistically significant positive effect on the accuracy of\nultrasound diagnosis of deeply infiltrating lesions. The\nsensitivity of the ultrasound diagnosis was significantly\naffected by the location of the endometriotic lesions but\nthe specificity remained high throughout. We have\nshown for the first time that ultrasound enables detec-\ntion and assessment of severity of adhesions affecting\nthe ovaries and pouch of Douglas.\nThe accuracy of TVS was highest in the diagnosis of\novarian adhesions, pouch of Douglas obliteration and\nbladder lesions. The accuracy for these features was simi-\nlar to the accuracy for ovarian endometriomas which were\npreviously thought of as the only feature of pelvic endo-\nmetriosis which it is possible to diagnose on ultrasound\n[1]. Previous studies have stated that left sided endome-\ntriomas are more common than right [13] but there was\nno statistically significant difference in our data set. Our\nstudy has also shown that only 26.9% of women with focal\nendometriosis will have disease in only one location and\ntherefore in all cases the examiner should perform a de-\ntailed search for lesions in other typical locations.\nThere are few studies on the accuracy of TVS for the\ndiagnosis of ovarian adhesions. Our study has shown a\nhigh level of accuracy for this diagnosis with a kappa value\nof 0.801. No study has previously assessed severity of ovar-\nian adhesions classified as either minimal, moderate or se-\nvere in accordance with the rASRM classification [14].\nGuerriero et al., [15] used the combination of three fea-\ntures as suggestive of ovarian adhesions: blurring of the\novarian margin, the inability to mobilise the ovary on pal-\npation (fixation) and an increased distance from the probe.\nThey found that these tests either combined or individu-\nally gave a kappa value of between 0.25 and 0.51. Okaro\net al., [2] examined women with chronic pelvic pain prior\nto laparoscopy for the presence of ovarian adhesions and\nclassified them as either mobile or fixed. They found a\nhigh degree of agreement between TVS and laparoscopy\nat identifying ovarian adhesions (0.81 kappa). This com-\npares with the results of our study of an area under the\nTable 6 Comparison of ultrasound and laparoscopy for the assessment of severity of adhesions in the pouch of Douglas\nPouch of Douglas obliteration at TVS\nPouch of Douglas obliteration at laparoscopy No adhesions Partial obliteration Complete obliteration Total\nNo adhesions 137 4 3 144 (72.7%)\nPartial obliteration 9 13 2 24 (12.1%)\nComplete obliteration 0 5 25 30 (15.2%)\nTotal 146 (73.7%) 22 (11.1%) 30 (15.2%) 198\nWeighted Kappa kappa = 0.852 (standard error (Kw ’ = 0) 0.071 and (Kw'#0) 0.038).\nTable 7 Women with DIE separated into groups by total\nnumber of endometriotic lesions compared with the\naccuracy of diagnosis of DIE in each group\nTotal number of\nendometriotic lesions\nNumber of\nwomen (n = 61)\nNumber correctly\ndiagnosed with DIE (n,%)\nSingle lesions 10 1 (10.0%)\n2 lesions 8 3 (37.5%)\n3 lesions 16 9 (56.3%)\n4 lesions 16 11 (68.8%)\n5 lesions or more 11 8 (72.7%)\nKruskal-Wallis test of correlation between total number of lesions and% of\nwomen correctly identified with DIE (P = 0.0228).\n Means  (error bars: 95% CI for mean)\n1.2\n1.0\n0.8\n0.6\n0.4\n0.2\n0.0\n-0.2\nTotal number of lesions\n1 2 3 4 5\nPercentage of women correctly diagnosed with DIE\nFigure 1 Bar chart of total number of endometriotic lesions at\nlaparoscopy against percentage of women correctly diagnosed\nwith DIE in each group.\nHolland et al. BMC Women's Health 2013, 13:43 Page 6 of 9\nhttp://www.biomedcentral.com/1472-6874/13/43\n\nROC of 0.889 for the presence of either moderate or se-\nvere adhesions and a kappa of 0.801 for the three stages of\nseverity. Yazbek et al., [16] examined the role of ultra-\nsound for the preoperative assessment of adnexal masses.\nThey found a sensitivity of 44% and a specificity of 98% in\nthe diagnosis of severe pelvic adhesions. The technique\nfor examination of adhesions was similar to that used in\nthis paper but they do not state ovarian adhesions separ-\nately. Guerriero et al., [3] used a technique of applying\npressure between the uterus and ovary. If they remained\nlinked then this was suggestive of adhesions. This gave a\nsensitivity and specificity of 89% and 90% respectively for\nfixation of the ovaries to the uterus.\nThe preoperative diagnosis of partial or complete ob-\nliteration of the pouch of Douglas has not been reported\non directly before. Our study shows a high accuracy of\nthis diagnosis. Hudelist [17] gave a high accuracy for the\ndiagnosis of pouch of Douglas endometriosis but did not\nreport obliteration separately. Yazbek [16] described the\ntechnique for diagnosing POD obliteration but did not re-\nport this finding separately from severe pelvic adhesions.\nThe high level of accuracy for the diagnosis of bladder\nendometriosis is concordant with previous studies,\nwhich showed a high level of accuracy in the TVS diag-\nnosis of bladder endometriosis [7,8].\nThere were poor levels of sensitivity for the diagnosis\nof endometriosis affecting the uterosacral ligaments and\npelvic side walls. The low accuracy of TVS for diagnos-\ning endometriosis of the uterosacral ligaments and pelvic\nside walls has also been previously reported [18,19].\nHudelist et al., [20] report higher levels of sensitivity for\nthe diagnosis of uterosacral disease however these levels\nwere lower than for almost all of the other locations of\nDIE. The preoperative diagnosis of endometriosis in\nthese locations is not critical for the management as\nthese are rarely missed at laparoscopy and surgical exci-\nsion can usually be achieved without involvement of\nother surgical specialists.\nOur study showed a high specificity of the diagnosis of\nrectovaginal disease and a lower sensitivity. This agrees\nwith the results of a recent review by Hudelist [21]\nencompassing 10 studies on the diagnostic accuracy of\nTable 8 Shows the mean number and mean proportion of lesions diagnosed on scan for all women with endometriotic\nlesions grouped by total number of lesions\nTotal number of lesions Number of women N = 104 Mean number of lesions diagnosed\non scan\nMean proportion of total lesions\ndiagnosed on scan\nSingle lesions 28 0.429 (95% CI 0.207 to 0.651) 0.3929 (95% CI 0.2000 to 0.5857)\n2 lesions 25 1.800 (95% CI 1.4232 to 2.1768) 0.8000 (95% CI 0.6541 to 0.9459)\n3 lesions 24 2.8750 (95% CI 2.4562 to 3.2938) 0.8750 (95% CI 0.7749 to 0.9751)\n4 lesions 16 3.5625 (95% CI 3.0871 to 4.0379) 0.8594 (95% CI 0.7756 to 0.9432)\n5 lesions or more 11 3.5455 (95% CI 2.4471 to 4.6438) 0.6450 (95% CI 0.4584 to 0.8316)\nP < 0.0001* P = 0.0008*\n*Kruskal-Wallis test of correlation between total number of lesions at laparoscopy and mean number of lesions diagnosed on scan and mean proportion o f total\nlesions diagnosed on scan respectively.\n Means  (error bars: 95% CI for mean)\n5\n4\n3\n2\n1\n0\nNumber of lesions at laparoscopy\n1 2 3 4 5\nMean number of lesions on scan\nFigure 2 Bar chart of total number of endometriotic lesions at\nlaparoscopy against the mean number of lesions seen on scan\nin each group.\n Means  (error bars: 95% CI for mean)\n1.0\n0.8\n0.6\n0.4\n0.2\n0.0\nNumber of lesions at laparoscopy\n1 2 3 4 5\nMean proportion of lesions diagnosed on scan\nFigure 3 Bar chart of total number of endometriotic lesions\nseen at laparoscopy against mean proportion of lesions\ndiagnosed on scan in each group.\nHolland et al. BMC Women's Health 2013, 13:43 Page 7 of 9\nhttp://www.biomedcentral.com/1472-6874/13/43\n\nTVS for intestinal endometriosis. He found sensitivities\nranging from 67- 98% and specificites of 92-100%.\nThe effect of the number of lesions on the sensitivity\nof ultrasound diagnosis of specific endometriotic lesions\nin different locations has not been assessed before. Our\ndata shows that the accuracy of the diagnosis of individ-\nual specific lesions increases with their absolute number\nup to a maximum of three lesions. With increasing\nnumber of lesions above that level the sensitivity de-\nclines. A possible reason for this could be that in more\nsevere disease the adhesions tend to obscure other small\nlesions further away from the ultrasound probe. There is\nalso a possibility of operator bias as in women with evi-\ndence of severe disease documentation of the presence\nof small lesions such as those located at utero-sacral lig-\naments becomes less clinically relevant.\nOur study could be criticised for not more accurately\ndifferentiating between DIE of the rectum and sigmoid\nor between rectovaginal and vaginal disease. We could\nalso be criticised for including subjective assessments\nsuch as ovarian and pouch of Douglas mobility which\ncannot be recorded with ease. However we diagnosed\novarian and pouch of Douglas disease with greater ac-\ncuracy than other features of endometriosis which indi-\ncates that subjective assessment is accurate enough to be\nused in routine practice. Reproducibility of these find-\nings however needs to be externally validated before we\ncan reach a consensus about the value of subjective as-\nsessment for the diagnosis of ovarian and pouch of\nDouglas adhesions. Scanning for endometriosis is diffi-\ncult and we believe that the use of palpation is of critical\nimportance to achieve good diagnostic accuracy. Gynae-\ncologists use palpation routinely as part of pelvic exam-\nination and they can incorporate it more easily into\nultrasound examination than sonographers or radiolo-\ngists. For this reason it remains to be seen whether these\nresults can be extrapolated to units with different levels\nof experience and expertise.\nThe benefit of an accurate diagnosis of individual fea-\ntures of endometriosis is that it provides a better overall\nassessment of the severity of the disease and aids in\ncounselling and planning of treatment. If surgery is re-\nquired, then women with severe disease may be referred\nto a tertiary centre with expertise in treating bladder and\nbowel disease. Prior knowledge of the extent of the dis-\nease facilitates comparisons of clinical symptoms with\nanatomical locations of endometriotic lesions. This im-\nproves pre-operative counselling of women and helps to\ntailor treatment in a way which will ensure excision of\nsymptomatic lesions and avoid complex procedures to\nremove asymptomatic lesions from difficult anatomical\nlocations. It also aids the surgeon in planning the oper-\nation and ensuring that the necessary staff are available,\nsuch as colorectal surgeons, when treatment of the\ndisease involving bowel is required. Preoperative under-\nestimation of the severity of DIE lesions increases the\nrisk of incomplete surgical excision, further progression\nof the residual disease and the need for multiple surgical\nprocedures [22,23].\nConclusions\nOur study has shown that the specificity of the ultrasound\ndiagnosis of pelvic endometriotic lesions is high with low\nfalse positive rates. The negative diagnostic rate was less\nhigh especially in the diagnosis of bowel, rectovaginal, uter-\nosacral ligament, pelvic side wall and uterosacral ligament\nlesions. Therefore women with significant symptoms and a\nnegative diagnosis still require further investigation. The\naccuracy of ultrasound diagnosis is significantly affected by\nthe location and number of endometriotic lesions.\nDetails of ethics approval\nKings College Hospital, London, Research Ethics Commit-\ntee reference number 06/Q0703/119. Full title of study.\nThe accuracy of gynaecological ultrasound examination\nfor the diagnosis of severe pelvic endometriosis.\nAbbreviations\nASRM: American society of reproductive medicine; CI: Confidence interval;\nDIE: Deeply infiltrating endometriosis; PPV: Positive predictive value;\nNPV: Negative predictive value; LR+: Positive likelihood ratio; LR-: Negative\nlikelihood ratio; SD: Standard deviation.\nCompeting interests\nES received honoraria from Ethicon for provision of training to healthcare\nprofessionals and consultancy fees from Bayer. AC is on the advisory board\nfor surgical innovations for which he receives an annual honorarium. AC also\nreceived support for courses and education from Storz and Johnson and\nJohnson and support for clinical nursing from Covidien and Lotus. The other\nauthors declared no competing interests.\nAuthors’ contributions\nTH designed the study protocol, wrote the ethics committee application,\nrecruited and scanned approximately half the patients, collected data,\nanalysed the data and drafted the manuscript. AC and ES operated on many\npatients and collected data. DM and KP collected data. DJ conceived of the\nstudy, and participated in its design and coordination, recruited and scanned\napproximately half the patients and helped with data analysis. All authors\nrevised the manuscript and read and approved the final manuscript.\nAuthor details\n1Early Pregnancy and Gynaecology Assessment Unit, Department of\nObstetrics and Gynaecology, Suite 8, Golden Jubilee Wing, King ’s College\nHospital, London SE5 8RX, UK. 2Department of Obstetrics and Gynaecology,\nUniversity College Hospital, 235 Euston Road, London NW1 2BU, UK.\nReceived: 5 April 2013 Accepted: 8 October 2013\nPublished: 29 October 2013\nReferences\n1. Moore J, Copley S, Morris J, Lindsell D, Golding S, Kennedy S: A systematic\nreview of the accuracy of ultrasound in the diagnosis of endometriosis.\nUltrasound Obstet Gynecol 2002, 20:630–634.\n2. Okaro E, Condous G, Khalid A, Timmerman D, Ameye L, Huffel SV, Bourne T:\nThe use of ultrasound-based ‘soft markers ’ for the prediction of pelvic\npathology in women with chronic pelvic pain –can we reduce the need\nfor laparoscopy? BJOG 2006, 113:251–256.\nHolland et al. BMC Women's Health 2013, 13:43 Page 8 of 9\nhttp://www.biomedcentral.com/1472-6874/13/43\n\n3. Guerriero S, Ajossa S, Garau N, Alcazar JL, Mais V, Melis GB: Diagnosis of\npelvic adhesions in patients with endometrioma: the role of transvaginal\nultrasonography. Fertil Steril 2010, 94:742–746.\n4. RCOG guideline: The investigation and management of endometriosis.Green\ntop Guidel 2006, 24 [http://www.rcog.org.uk/files/rcog-corp/GTG2410022011.pdf]\n5. Van Holsbeke C, Van Calster B, Guerriero S, Savelli L, Paladini D, Lissoni AA,\nCzekierdowski A, Fischerova D, Zhang J, Mestdagh G, Testa AC, Bourne T,\nValentin L, Timmerman D: Endometriomas: their ultrasound\ncharacteristics. Ultrasound Obstet Gynecol 2010, 35:730–740.\n6. Savelli L, de Iaco P, Ghi T, Bovicelli L, Rosati F, Cacciatore B: Transvaginal\nsonographic appearance of peritoneal pseudocysts. Ultrasound Obstet\nGynecol 2004, 23:284–288.\n7. Fedele L, Bianchi S, Raffaelli R, Portuese A: Pre-operative assessment of\nbladder endometriosis. Hum Reprod 1997, 12:2519–2522.\n8. Bazot M, Malzy P, Cortez A, Roseau G, Amouyal P, Daraï E: Accuracy of\ntransvaginal sonography and rectal endoscopic sonography in the diagnosis\nof deep infiltrating endometriosis.Ultrasound Obstet Gynecol2007, 30:994–1001.\n9. Koga K, Osuga Y, Yano T, Momoeda M, Yoshino O, Hirota Y, Kugu K, Nishii\nO, Tsutsumi O, Taketani Y: Characteristic images of deeply infiltrating\nrectosigmoid endometriosis on transvaginal and transrectal\nultrasonography. Hum Reprod 2003, 18:1328–1333.\n10. Holland TK, Yazbek J, Cutner A, Saridogan E, Hoo WL, Jurkovic D: The value\nof transvaginal ultrasound in assessing the severity of pelvic\nendometriosis. Ultrasound Obstet Gynecol 2010, 36:241–248.\n11. Bland JM, Altman DG: Statistical methods for assessing agreement\nbetween two methods of clinical measurement. Lancet 1986, 1:307–310.\n12. Bland JM, Altman DG: Measuring agreement in method comparison\nstudies. Stat Methods Med Res 1999, 8:135–160.\n13. Chapron C, Chopin N, Borghese B, Foulot H, Dousset B, Vacher-Lavenu MC,\nVieira M, Hasan W, Bricou A: Deeply infiltrating endometriosis: pathogenetic\nimplications of the anatomical distribution.Hum Reprod 2006, 21:1839–1845.\n14. Society AF: Revised American fertility society classification of\nendometriosis. Fertil Steril 1985, 43:351–352.\n15. Guerriero S, Ajossa S, Lai MP, Mais V, Paoletti AM, Melis GB: Transvaginal\nultrasonography in the diagnosis of pelvic adhesions. Hum Reprod 1997,\n12:2649–2653.\n16. Yazbek J, Helmy S, Ben-Nagi J, Holland T, Sawyer E, Jurkovic D: Value of\npreoperative ultrasound examination in the selection of women with adnexal\nmasses for laparoscopic surgery.Ultrasound Obstet Gynecol2007, 30:883–888.\n17. Hudelist G, Oberwinkler KH, Singer CF, Tuttlies F, Rauter G, Ritter O,\nKeckstein J: Combination of transvaginal sonography and clinical\nexamination for preoperative diagnosis of pelvic endometriosis.\nHum Reprod 2009, 24:1018–1024.\n18. Savelli L: Transvaginal sonography for the assessment of ovarian and\npelvic endometriosis: how deep is our understanding? Ultrasound Obstet\nGyne 2009, 33:497–501.\n19. Bazot M, Thomassin I, Hourani R, Cortez A, Darai E: Diagnostic accuracy of\ntransvaginal sonography for deep pelvic endometriosis. Ultrasound Obstet\nGynecol 2004, 24:180–185.\n20. Hudelist G, Ballard K, English J, Wright J, Banerjee S, Mastoroudes H, Thomas\nA, Singer CF, Keckstein J: Transvaginal sonography vs. clinical examination\nin the preoperative diagnosis of deep infiltrating endometriosis.\nUltrasound Obstet Gynecol 2011, 37:480–487.\n21. Hudelist G, English J, Thomas AE, Tinelli A, Singer CF, Keckstein J: Diagnostic\naccuracy of transvaginal ultrasound for non-invasive diagnosis of bowel\nendometriosis: systematic review and meta-analysis. Ultrasound Obstet\nGynecol 2011, 37:257–263.\n22. Chapron C, Pietin-Vialle C, Borghese B, Davy C, Foulot H, Chopin N:\nAssociated ovarian endometriomas is a marker for greater severity of\ndeeply infiltrating endometriosis. Fertil Steril 2008, 92:453–457.\n23. Fedele L, Bianchi S, Zanconato G, Berlanda N, Borruto F, Frontino G:\nTailoring radicality in demolitive surgery for deeply infiltrating\nendometriosis. Am J Obstet Gynecol 2005, 193:114–117.\ndoi:10.1186/1472-6874-13-43\nCite this article as: Holland et al.: Ultrasound mapping of pelvic\nendometriosis: does the location and number of lesions affect the\ndiagnostic accuracy? a multicentre diagnostic accuracy study. BMC\nWomen's Health 2013 13:43.\nSubmit your next manuscript to BioMed Central\nand take full advantage of: \n• Convenient online submission\n• Thorough peer review\n• No space constraints or color figure charges\n• Immediate publication on acceptance\n• Inclusion in PubMed, CAS, Scopus and Google Scholar\n• Research which is freely available for redistribution\nSubmit your manuscript at \nwww.biomedcentral.com/submit\nHolland et al. BMC Women's Health 2013, 13:43 Page 9 of 9\nhttp://www.biomedcentral.com/1472-6874/13/43","source_license":"CC0","license_restricted":false}