Diagnostic Value of MR Imaging in the Diagnosis of Adnexal Torsion 1

Transcription

1 This copy is for personal use only. To order printed copies, contact Sophie Béranger-Gibert, MD Hajer Sakly, MD Marcos Ballester, MD, PhD Andrea Rockall, MRCP, FRCR Marie Bornes, MD Marc Bazot, MD, PhD Emile Daraï, MD, PhD Isabelle Thomassin-Naggara, MD, PhD Diagnostic Value of MR Imaging in the Diagnosis of Adnexal Torsion 1 Purpose: Materials and Methods: To retrospectively evaluate the diagnostic performance of magnetic resonance (MR) imaging for the diagnosis of adnexal torsion in a series of patients with an equivocal adnexal mass at ultrasonography (US) in the context of acute or subacute pelvic pain. The institutional ethics committee approved the study and waived informed consent. All patients with acute or subacute pelvic pain who were undergoing MR examination for the exploration of an equivocal adnexal mass (January 2007 to December 2012) with surgical exploration or clinical and radiologic follow-up of at least 3 months were retrospectively included (n = 58). The prospective interpretations were recorded. Additionally, three radiologists who were blinded to the clinical, US, and surgical data retrospectively and independently reviewed MR images. Features associated with adnexal torsion were identified by using univariate and recursive partitioning multivariate analysis. Original Research n Genitourinary Imaging 1 From the Department of Radiology (S.B.G., H.S., M. Bazot, I.T.N.) and Department of Gynaecology and Obstetrics (M. Ballester, I.T.N.), AP-HP, Hôpital Tenon, 4 rue de la Chine, Paris, France; Department of Gynaecology- Obstetrics, Sorbonne Universités, UPMC Université Paris 06, Paris, France (M. Ballester, M. Bornes, E.D.); Department of Radiology, Imperial College of London, London, England (A.R.); and INSERM, UMR970, Equipe 2, Imagerie de l angiogenèse, Paris, France (I.T.N.). Received February 2, 2015; revision requested March 23; revision received May 6; accepted June 8; final version accepted August 18. Address correspondence to I.T.N. ( isabelle. thomassin@tnn.aphp.fr). Results: Conclusion: Twenty-two patients (38%) had a diagnosis of adnexal torsion. The accuracy of MR imaging at the time of prospective interpretation was 80.6% (25 of 31 patients) and 85.1% (23 of 27 patients) in acute and subacute torsion, respectively. The accuracy of image interpretation by each retrospective reader was 83.9% (26 of 31 patients), 90.3% (28 of 31 patients), and 83.9% (26 of 31 patients) in the context of acute pelvic pain and 92.6% (25 of 27 patients), 88.9% (24 of 27 patients), and 81.5% (22 of 27 patients) in the context of subacute pelvic pain for readers 1, 2, and 3, respectively. At multivariate analysis, the whirlpool sign (odds ratio = 6.5 [95% confidence interval: 1.36, 31.0], P =.01) and a thickened tube (.10 mm) (odds ratio = 8.2 [95% confidence interval: 1.2, 56.8], P =.03) were associated with adnexal torsion, with substantial interreader agreement (k = and , respectively). The presence of adnexal hemorrhagic content was associated with nonviable ovaries in seven of 10 patients (70%) and with viable ovaries in 12 of 45 patients (27%) (P =.009). MR imaging is an accurate technique for the diagnosis of adnexal torsion in patients who have an adnexal mass with acute or subacute pelvic pain. q RSNA, 2015 q RSNA, 2015 Radiology: Volume 279: Number 2 May 2016 n radiology.rsna.org 461

2 Adnexal torsion is a rare condition (accounting for 2.7% of female acute pelvic pain) (1) that is responsible for acute or subacute congestion and ischemia of the ovary. In most cases, torsion involves both the ovary and the tube and, rarely, the tube alone (2,3). There is usually a preexisting adnexal abnormality (1,4,5). Early diagnosis is crucial because delayed identification of torsion may lead to ovarian necrosis and affect a woman s fertility (6,7). Therefore, surgical exploration is mandatory when torsion is highly suspected on the basis of clinical and imaging findings. The main clinical feature is sudden and marked pelvic pain (8,9), but intermittent vascular compression with varying severity may lead to more subacute or chronic clinical presentation, with frequent spontaneous resolution and recurrence of pelvic pain. Thus, clinical diagnosis may be difficult, since torsion is suspected preoperatively in only 23% 66% of cases, while around half of the patients undergoing surgery for suspicion of adnexal torsion have a different final diagnosis (10,11). In this setting, improving the preoperative diagnosis of adnexal torsion is Advances in Knowledge nn In a retrospective review of patients who underwent MR imaging after nondiagnostic US examination and had a clinical history of pain, the diagnosis of torsion was assigned prospectively in 77.3% of patients (17 of 22) and retrospectively in 86.4% of patients (19 of 22). nn The whirlpool sign (odds ratio = 6.5 [95% confidence interval: 1.36, 31.0], P =.01) and a thickened tube of more than 10 mm (odds ratio = 8.2 [95% confidence interval: 1.2, 56.8], P =.03) are associated with adnexal torsion. nn The presence of high T1-weighted signal intensity that demonstrates blood products within the lesion was seen in seven of 10 patients with nonviable ovaries (70%) and in 12 of 45 patients (27%) with viable ovaries. a priority issue (12). Ultrasonography (US) is the first-line imaging technique in the investigation of lower abdominal or pelvic pain in the premenopausal woman. When an adnexal mass is discovered in this context, US is helpful in most cases for the diagnosis of functional abnormalities. For the diagnosis of adnexal torsion, US sensitivities range from 46% to 74% (13). Thus, in case of difficult or doubtful diagnosis, magnetic resonance (MR) imaging may be helpful. In some studies, investigators have evaluated the value of computed tomography (CT) in the diagnosis of adnexal torsion (14 16) because of its accessibility in the context of emergency. However, exposure to pelvic irradiation makes this a less favorable imaging method in women of reproductive age. Additionally, in the presence of subacute or more chronic clinical presentation, the diagnosis of adnexal torsion may be overlooked (17). As a consequence, patients may undergo pelvic MR examination for pain evaluation without obvious clinical suspicion of torsion. In this specific setting, MR findings may lead to the proper diagnosis. Nevertheless, published reports that address the diagnostic performance of MR are scarce (13,16,18 24). The purpose of this study was to evaluate retrospectively the diagnostic performance of MR imaging for the diagnosis of adnexal torsion in a series of patients with an indeterminate adnexal mass at US in the context of acute or subacute pelvic pain. Materials and Methods Our institutional ethics committee approved the study and granted a waiver of informed consent. Population Our MR imaging database was retrospectively queried to identify women Implication for Patient Care nn MR imaging is an accurate second-line technique to diagnose adnexal torsion without any pelvic irradiation. who underwent pelvic MR examination to evaluate adnexal masses in the context of acute pelvic pain (of less than 4 hours duration) or subacute pelvic pain (defined as recurrent intermittent acute pelvic pain for no more than 1 month) between January 1, 2007, and December 31, In our center, an endovaginal US examination is systematically performed to evaluate any adnexal mass associated with pelvic pain. The examination is performed by a gynecological sonographer (a radiologist or gynecologist) with at least 5 years of experience. MR imaging is a secondline technique performed only in cases of doubtful or difficult diagnosis at pelvic US. Demographic and clinical data were extracted from our institutional database, with additional retrospective medical record review. Hormonal status (ie, pregnancy, ovarian stimulation, or menopause) was noted. The following clinical findings, which were typically described to characterize acute or subacute pelvic pain, were also noted: (a) painful vagina at examination, (b) persistent pain despite analgesic use, (c) abdominal guarding, (d) nausea, (e) medical history of torsion, (f) elevated temperature, and (g) leukorrhea. Finally, the delay between the beginning of the pain and the MR imaging examination was recorded. A total of 65 patients were identified. Women who were lost to follow-up (n = 6) or who had technical problems related to the picture archiving and Published online before print /radiol Content codes: Radiology 2016; 279: Author contributions: Guarantors of integrity of entire study, M. Bornes, I.T.N.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; agrees to ensure any questions related to the work are appropriately resolved, all authors; literature research, H.S., A.R., M. Bornes, M. Bazot, E.D., I.T.N.; clinical studies, H.S., M. Bornes, M. Bazot, I.T.N.; experimental studies, S.B.G.; statistical analysis, H.S., M. Ballester, I.T.N.; and manuscript editing, A.R., M. Bazot, E.D., I.T.N. Conflicts of interest are listed at the end of this article. 462 radiology.rsna.org n Radiology: Volume 279: Number 2 May 2016

3 communication system (n = 1) were subsequently excluded. The final cohort consisted of 58 women (mean age, 38 years; range, years), including 13 of 58 menopausal women (22%) and six of 58 pregnant women (10%). Among these patients, eight underwent CT, which was performed before US. MR Acquisition Protocol MR examinations were performed with a 1.5-T clinical MR imaging unit (Sonata, Siemens, Forchhiem, Germany; or HDXT, GE, Milwaukee, Wis) by using a phased-array pelvic coil. The acquisition protocols, including sequences and parameters, are detailed in Table 1. In the absence of contraindications, patients received an antispasmodic drug intravenously (Glucagen; Novo Nordisk, Bagsvaerd, Denmark) immediately before MR imaging to reduce bowel peristalsis. Patients underwent variable MR protocols, including at least the following common sequences: sagittal and axial turbo spin-echo T2-weighted sequences and axial gradient-echo T1-weighted sequences, with and without fat suppression. No respiratory gating was used. Delayed contrast material enhanced axial and sagittal T1-weighted gradientecho images with breath hold were acquired in 46 patients (79%) after gadolinium chelate injection (Dotarem; Guerbet, Aulnay-sous-Bois, France). MR Data Analysis All MR images were reviewed retrospectively and independently on a picture archiving and communication system workstation (Agfa Healthcare, Mortsel, Belgium) by three radiologists (I.T.N. [reader 1], H.S. [reader 2], and S.B.G. [reader 3], with 11 years, 2 years, and 1 year of experience in pelvic MR imaging, respectively), who were blinded to the clinical and surgical data. None of the readers were involved in patient care. The following tubal and ovarian features were noted: direct adnexal signs of torsion with (a) presence of a whirlpool sign (a twisted ovarian pedicle or twisted fallopian tube), (b) tubal wall thickening (more than 10 mm), (c) dilated fallopian tube, (d) symmetrical or asymmetrical thickening of the wall of an ovarian cyst, (e) ovarian stromal edema, (f) presence of normal ovarian parenchyma, and (g) prominent follicles in the periphery of the ovarian parenchyma. The other following features were also recorded: (a) presence, size, and side of the adnexal mass (when masses involved both sides, the painful side was considered for analysis); (b) adnexal position in the pelvic cavity; (c) uterine displacement toward the involved side or opposed side; (d) ovarian vascular pedicle enlargement; (e) presence of pelvic fluid or ascites (ie, peritoneal fluid cranial to the level of the cecum); and (f) presence of fat stranding. Finally, the presence of ovarian or tubal hemorrhage was noted on T1-weighted images, and ovarian parenchymal enhancement after contrast material administration was noted. Reference Standard The final diagnosis was established either by means of surgery in 46 of 58 patients (79.3%) or by means of clinical and radiologic follow-up for at least 3 months in 12 of 58 patients (20.7%). Surgical evaluation findings corresponded to laparoscopic findings (n = 40) and laparotomic findings (n = 6). The median delay between MR imaging and surgery was 2 days (range, 0 15 days), including 22 women with adnexal torsion (with a median delay of 0 days [range, 0 13 days]) and 24 women without adnexal torsion (with a median delay of 11 days [range, 0 15 days]). Adnexal torsion was always proven at surgery. The other 12 patients underwent clinical and/or imaging follow-up for 3 months. In this group, the absence of adnexal torsion was finally diagnosed. Table 1 MR Sequence Parameters Sonata HDXT Parameter Sagittal Turbo Spin-Echo T2-weighted Sequence Axial Turbo Spin-Echo T2-weighted Sequence Axial Gradient-Echo T1-weighted Sequence Sagittal Turbo Spin-Echo T2-weighted Sequence Axial Turbo Spin-Echo T2-weighted Sequence Axial Gradient-Echo T1-weighted Sequence Repetition time (msec) Echo time (msec) Echo train length Flip angle (degrees) Section thickness (mm) Gap (mm) Field of view Matrix No. of signals acquired No. of sections Time (sec) Note. The Sonata is manufactured by Siemens; the HDXT is manufactured by GE. Radiology: Volume 279: Number 2 May 2016 n radiology.rsna.org 463

4 Statistical Analysis Quantitative data were analyzed by using the nonparametric Mann-Whitney test, and categorical data were compared by using the Fisher exact test or the x 2 test, as appropriate. Interreader agreement was evaluated by using the Cohen k statistic, according to the following scale: , slight agreement; , fair agreement; , moderate agreement; , substantial agreement; and , almost perfect agreement (25). Clinical data and imaging features associated with adnexal torsion were identified by using univariate analysis. Multivariate analysis was based on the interpretation of the most experienced reader. Factors that demonstrated correlation with adnexal torsion with a P value less than.1 were then included in a multivariate logistic analysis model. P values less than.05 were considered to indicate a statistically significant difference. Statistical analyses were performed by using MedCalc version software (MedCalc, Ostend, Belgium) and R version 2.15 software (R Foundation, Vienna Austria), available online for multivariate analysis. Results Final Diagnosis Overall, 22 of 58 patients (38%) had a final diagnosis of adnexal torsion, and 36 of 58 patients (62%) had no adnexal torsion (Table 2). Ovarian masses were identified in 46 patients and included 41 benign masses, four borderline (mucinous subtype) masses, and two invasive malignancies (one mucinous ovarian carcinoma and one immature ovarian teratoma). Ovarian masses were present in a similar percentage of cases with torsion and without (19 of 22 cases [86.4%] with torsion and 27 of 36 cases [75%] without, P =.34). All cases of pelvic inflammatory disease (n = 8) were exclusively found in the group of women without torsion (P,.001). The start of pain was within 4 hours in 31 women. Acute pelvic pain was correlated with adnexal torsion in 35.5% of patients (11 of 31). In three Table 2 Final Diagnosis Diagnosis Torsion (n = 22) No Torsion (n = 36) P Value Functional abnormality Functional cyst 4 (18.2) 12 (33.3)*.24 Ovarian hyperstimulation syndrome 1 (4.5) Ovarian tumor.3 Epithelial 13 (59.1) 15 (41.7) Benign 2 (9.1) 3 (8.3) Borderline 2 (9.1) 2 (5.6) Malignant 1 (4.5) Germ cell tumors Mature cystic teratoma 3 (13.6) 6 (16.7) Immature teratoma 1 (2.8) Sex cord stromal tumor Fibroma 3 (13.6) 2 (5.6) Other 1 (4.5) Endometrioma 1 (4.5) 1 (2.8) Isolated tubal torsion 2 (9.1) Pelvic inflammatory disease 8 (22.2),.0001 Other (unknown, uterine pedunculated leiomyoma) 2 (9.1) 1 (2.8).55 Note. Numbers in parentheses are percentages. In our population, there were six pregnant women at 5, 6, 7, 12, 14, and 17 weeks of amenorrhea. Three women had adnexal torsion complicated ovarian hyperstimulation (n = 1), luteinized pregnancy cyst (n = 1), or luteal cyst (n = 1); and three women without adnexal torsion had luteal cyst (n = 2) or borderline mucinous cystadenoma (n = 1). * Two of 12 women underwent surgery. Six of eight women underwent surgery. of these cases, the ovary was nonviable at the time of surgery (mean delay from MR imaging to surgery, 2 days; range, 0 5 days). In 27 cases with subacute pelvic pain for 4 hours to 30 days with a mean delay of 9 days, adnexal torsion was present in 40.7% of patients (11 of 27). In seven of these cases, the ovary was nonviable at the time of surgery (mean delay from MR imaging to surgery, 5 days; range, 2 15 days). US findings suggested adnexal torsion in 16 of 58 patients (28%), with a correct diagnosis in 10 patients and an incorrect diagnosis in six patients. MR imaging was indicated after US because of atypical clinical presentation of adnexal torsion. In 42 of 58 patients, US findings did not suggest the diagnosis of torsion. MR imaging was indicated after US in the context of pelvic pain because of the presence of complex adnexal masses (n = 23); atypical unilateral pelvic inflammatory disease (n = 8), including seven nontubular masses that were larger than 6.5 cm in diameter; atypical hemorrhagic cysts (n = 6); or ovarian edema without ovarian mass detected (n = 5). Specific Diagnostic Value of Clinical and Imaging Findings to Diagnose Adnexal Torsion No single clinical finding was significantly associated with the diagnosis of adnexal torsion (Table 3). The mean long-axis diameter of ovarian masses was 92.6 mm (range, mm), with no difference between the group with torsion and the group without torsion (P =.19). However, the size of the ipsilateral ovarian parenchyma was higher in the torsion group (mean size 6 standard deviation, 11.3 cm 2 6 8; range, cm 2 ) than in the group without torsion (mean size, 5.3 cm ; range, cm 2 ; P =.01). At univariate analysis, the following MR imaging findings were significantly more frequent in the adnexal torsion group: uterine 464 radiology.rsna.org n Radiology: Volume 279: Number 2 May 2016

5 Table 3 Clinical and Imaging Features Associated with the Diagnosis of Adnexal Torsion Finding Torsion (n = 22) No Torsion (n = 36) P Value Derived from Univariate Analysis Odds Ratio MR finding Long-axis diameter of mass within the ovary (mm) (31 212) (21 270).19 Size of ipsilateral ovarian parenchyma (cm 2 ) (3.8 30) (0.8 23).01 Uterine deviation toward involved side 9 (40.9) 3 (8.6)* [1.47, 47.25] Uterine deviation toward opposite side 2 (9.1) 12 (34.3)*.056 Whirlpool sign 15 (68.2) 4 (11.1), [3.72, 88.07] Tubal wall thickening 20 (90.9) 14 (38.9), [2.92, 151.5] Dilated tube 2 (9.1) 8 (22.2).29 Cystic wall thickening 7 (31.8) 10 (27.8).77 Asymmetrical cystic wall thickening 6 (27.3) 4 (11.1).57 Stromal edema 7 (31.8) 6 (16.7).21 Normal ovarian parenchyma 8 (36.4) 19 (52.8).28 Ovarian vascular pedicle enlargement 13 (59.1) 7 (19.4) [1.59, 23.24] Peripheral follicles 6 (27.3) 4 (11.1).15 Peritoneal fluid 12 (54.5) 15 (41.7).42 Ascites 5 (22.7) 4 (11.1).28 Fat stranding 0 (0) 7 (19.4).04 Hemorrhagic content 11 (50) 8 (22.2) [0.96, 12.89] Contrast material enhancement 10 (58.8) 27 (93.1) [0.01, 0.72] Clinical findings Pregnancy 3 (13.6) 3 (8.3).66 Cervical motion tenderness 7 (31.8) 13 (36.1).78 Persistent pain 12 (54.5) 12 (33.3).19 Abdominal guarding 0 (0) 2 (5.6).52 Subacute pain 19 (86.4) 29 (80.6).83 Nausea 5 (22.7) 7 (19.4)..99 History of torsion 0 (0) 1 (2.8)..99 Ovarian stimulation 4 (18.2) 4 (11.1).64 Menopause 6 (27.3) 7 (19.4).52 Fever 2 (9.1) 8 (22.2).29 Leukorrhea 0 (0) 6 (16.7).07 Note. Quantitative values are expressed as means 6 standard deviations, with ranges in parentheses. Categorical values are expressed as number of cases, with percentages in parentheses. Numbers in brackets are 95% confidence intervals. * Only 57 patients had a uterus at the time of the analysis. Only 46 patients underwent contrast agent injection. displacement toward the involved side (40.9% [nine of 22 patients] vs 13.6% [three of 22 patients], P =.006), whirlpool sign (Fig 1) (68.2% [15 of 22 patients] vs 11.1% [four of 36 patients], P,.0001), tubal wall thickening (90.9% [20 of 22 patients] vs 36.1% [13 of 36 patients], P,.0001) (Figs 2, 3), ovarian vascular pedicle enlargement (59.1% [13 of 22 patients] vs 19.4% [seven of 36 patients], P =.004) (Fig 4), and ovarian hemorrhagic content (50% [11 of 22 patients] vs 22.2% [eight of 36 patients], P =.04). Fat stranding (0% vs 19.4% [seven of 36 patients], P =.04) and contrast material enhancement (58.8% [10 of 17 patients] vs 93.1% [27 of 29 patients], P =.007) were less frequent in the adnexal torsion group. No difference was found between the two groups for the following criteria: uterine deviation toward the opposite side, dilated tube, cystic wall thickening, ovarian stromal edema, peritoneal fluid, peripheral follicles, and remaining normal ovarian parenchyma. At multivariate analysis (including the whirlpool sign, tubal thickening, ovarian vascular pedicle enlargement, size of the ipsilateral ovarian parenchyma, and adnexal hemorrhagic content), the whirlpool sign (odds ratio = 6.5 [95% confidence interval: 1.36, 31.0], P =.01) and presence of a thickened tube (odds ratio = 8.2 [95% confidence interval: 1.2, 56.8], P =.03) were associated with adnexal torsion. No correlation was found between any of the clinical features and the viability of the ipsilateral ovary. Among Radiology: Volume 279: Number 2 May 2016 n radiology.rsna.org 465

6 Figure 2 Figure 1 Figure 1: Transverse T2-weighted fast spin-echo MR image demonstrates the whirlpool sign. A right twisted fallopian tube (arrow) is seen in a 57-year-old woman with subacute pain (of 9 hours duration). Image shows a serous borderline cystadenoma on the right ovary and a uterine leiomyoma. Intraoperative findings were a double twist and a 74-mm mass, which appeared to be partially necrotic at histopathologic examination. Figure 2: Sagittal T2-weighted fast spin-echo MR image demonstrates tubal wall thickening. A thickened right fallopian tube q with a serous borderline cystadenoma on the right ovary and a left ovarian serous borderline cystadenoma (arrow) are shown in a 57-year-old patient. Right adnexal torsion was found at surgery. MR features, the presence of hemorrhagic content was associated with nonviable ovaries in seven of 10 patients, whereas it was found in only 12 of 45 Figure 3 Figure 3: Sagittal T2-weighted fast spin-echo MR image demonstrates tubal wall thickening. A thickened fimbrial end of the left adnexal tube (arrows) in front of an ovarian serous benign cystadenoma is shown in a 37-year-old woman, with a single twist found at surgery. women (27%) with viable ovaries (P =.0009). In contrast, no correlation was found with either ovarian contrast material enhancement, which was found in 60% of the group of women (six of 10) with nonviable ovaries versus 64.6% in the group of women (31 of 48) with viable ovaries. Interobserver Agreement and Diagnostic Performance of the Three Readers Interobserver agreement was substantial or good for all criteria but one: The presence of a dilated tube (k = between reader 1 and reader 2) showed moderate and fair agreement (Table 4). The accuracy of MR imaging was 83.9% (26 of 31 patients), 90.3% (28 of 31 patients), 83.9% (26 of 31 patients), and 80.6% (25 of 31 patients) in the presence of acute pelvic pain and 92.6% (25 of 27 patients), 88.9% (24 of 27 patients), 81.5% (22 of 27 patients), and 85.2% (23 of 27 patients) in the presence of subacute pelvic pain for readers 1, 2, and 3 and prospective analysis, respectively (Table 5). Five patients received a false-negative diagnosis from a least one of the three readers. Each reader missed three cases of adnexal torsion, including one case missed by all readers. This false-negative case corresponds to a benign serous cystadenoma with a maximal diameter of 56 mm in a 44-year-old woman who experienced subacute pelvic pain for 8 days. All readers misdiagnosed adnexal torsion as a hemorrhagic cyst (Fig 5). Prospectively, MR findings suggested an adnexal torsion in 24 of 58 patients, with a correct diagnosis in 17 patients (sensitivity, 77.2% [17 of 22 patients; 95% confidence interval: 56.7, 89.8]) and an incorrect diagnosis in five cases (specificity, 86.1%; 95% confidence interval: 71.3, 93.9). At the time of patient care, adnexal torsion was incorrectly diagnosed in three cases (misdiagnosed as an indeterminate adnexal mass without torsion [two with acute pain and one with subacute pain]) and missed in two cases (misdiagnosed as functional disease [rupture of a luteal cyst] in one patient with acute pain and one with subacute pain. Among the 16 patients with suspicion of adnexal torsion at US, the diagnosis was changed correctly in 466 radiology.rsna.org n Radiology: Volume 279: Number 2 May 2016

7 Figure 4 three patients without adnexal torsion and incorrectly changed in one patient with adnexal torsion. Table 4 Figure 4: Transverse T2-weighted fast spin-echo MR image in a 37-year-old woman shows a left vascular pedicle enlargement (arrow) on a left adnexal torsion. Interobserver Agreement for the Different MR Features Finding Agreement between Readers 1 and 2 Agreement between Readers 2 and 3 Torsion Uterine attraction toward the involved side Uterine deviation toward the opposed side Whirlpool sign Tubal thickening Dilated tube Cystic wall thickening Asymmetrical cystic wall thickening Stromal edema Ovarian vascular pedicle enlargement Fat stranding Hemorrhagic content Contrast material enhancement Note. Data are k statistics. Agreement between Readers 1 and 3 Discussion The results of this study demonstrate that MR imaging is accurate for the diagnosis of adnexal torsion in patients with an adnexal mass with acute and subacute pelvic pain at presentation, with a negative predictive value higher than 90% in our three readers with 1 11 years of experience in pelvic MR imaging. Prospective MR analysis has a sensitivity of 77.2% (17 of 22 patients) and a specificity of 86.1% for the diagnosis of adnexal torsion. The probability of adnexal torsion was sixfold and eightfold higher in cases of visualization of the whirlpool sign and tubal thickening, respectively, and both of these signs were found to be identified reproducibly between readers. Moreover, the demonstration of blood products within the ovarian mass on T1-weighted images may suggest nonviable ovary. Currently, the management of suspected adnexal torsion is mainly based on clinical and US findings, with a highly variable rate of misdiagnoses (10,11) that is mainly dependent on the experience of the sonologist (22,26 28). Our findings suggest that MR imaging is accurate in the diagnosis of adnexal torsion in the context of acute pelvic pain that presents rapidly over less than 4 hours, with accuracy higher than 80%. In the group of women with subacute pelvic pain, adnexal torsion was not initially suspected with US, and MR Table 5 Diagnostic Performance according to Reader Parameter Reader 1 (%) Reader 2 (%) Reader 3 (%) Prospective MR Interpretation Sensitivity 86.4 (19/22) [65.8, 96.1] 86.4 (19/22) [65.8, 96.1] 86.4 (19/22) [65.8, 96.1] 77.3 (17/22) [56.7, 89.8] Specificity 88.9 (32/36) [74, 96.2] 91.7 (33/36) [77.4, 97.8] 80.6 (29/36) [64.6, 90.5] 86.1 (31/36) [71.3, 93.9] Positive predictive value 82.6 (19/23) [62.2, 93.6] 86.4 (19/22) [65.8, 96.1] 70.4 (19/27) [51.3, 84.3] 77.3 (17/22) [56.7, 89.8] Negative predictive value 91.4 (32/35) [76.9, 97.8] 91.7 (33/36) [77.4, 97.8] 90.6 (29/32) [75, 97.5] 86.1 (31/36) [71.3, 93.9] Accuracy 87.9 (51/58) [76.8, 94.3] 89.6 (52/58) [78.9, 95.5] 82.8 (48/58) [70.9, 90.5] 82.8 (48/58) [70.9, 90.5] Note. Numbers in parentheses are the data used to calculate the percentages. Numbers in brackets are 95% confidence intervals. Radiology: Volume 279: Number 2 May 2016 n radiology.rsna.org 467

8 Figure 5 Figure 5: MR images demonstrate a false-negative case in a 44-year-old woman with right ovarian torsion of a benign 56-mm serous cystadenoma. (a) Transverse T2-weighted fast spin-echo MR image (repetition time [msec]/echo time [msec], 89/6790) shows a thin (,10-mm) left adnexal tube (arrow) but a uterine deviation toward the involved side (white line shows the middle of the pelvic cavity). (b) Transverse T2-weighted fast spin-echo MR image (89/6790) shows a normal left vascular pedicle (arrow). (c) Transverse T1-weighted fast spin-echo MR image with fat suppression (2.6/232) shows high intensity of the content of the serous cystadenoma (arrow), suggesting hemorrhagic content. (d) Transverse T1-weighted contrast-enhanced fast spin-echo MR image (12/498) shows adnexal tube enhancement (arrow). All three readers diagnosed this as a hemorrhagic cyst. imaging was performed with a mean delay of 9 days. Our findings also demonstrate that MR imaging is accurate in this setting. This is an important clinical issue because the percentage of women with adnexal torsion in both the acute (40.7%, 11 of 27 patients) and subacute (35.5%, 11 of 31 patients) pain groups was similar. In our series, the multivariate analysis showed that the whirlpool sign and a thickened tube were the two main imaging features associated with the diagnosis of adnexal torsion. These MR findings should therefore be actively searched for in any female patient undergoing MR imaging for acute or subacute pelvic pain. Interestingly, both features have been reported to be the most specific imaging findings of adnexal torsion by using other imaging techniques (US and CT) (14 16,28). The tube is considered normal if it is no wider than 10 mm in diameter at any part of its length (29). Thickening of the tube wider than this diameter indicates congestion and edema in the context of torsion. In our study, this sign was found in 90.9% of women (20 of 22) with adnexal torsion as reported in previous CT and MR series (16,23). The whirlpool sign is the direct sign that represents the twisted pedicle also described in US (28) and CT (16). In our study, this feature was accurate with only four false-positive cases, which included two patients with subacute pelvic pain. Since our reference standard was the detection of at least one complete twist around the ligamentous support of the tube, it is possible that the torsion may have resolved prior to the time of surgery. The design of our study helps to evaluate the specificity of features described previously in smaller MR series of patients with proven adnexal torsion (16,24). Uterine displacement toward 468 radiology.rsna.org n Radiology: Volume 279: Number 2 May 2016

9 the involved side and ovarian vascular pedicle enlargement have a high specificity (91.7% [33 of 36 patients]) and 80.5% [29 of 36 patients], respectively). However, other findings, including cystic wall thickening or asymmetrical cystic wall thickening, have a low specificity and could not be used to differentiate women with adnexal torsion from those without. Regarding ovarian features typically described in adnexal torsion, including peripheral follicles or stromal edema (30), our results underline the fact that these findings are also found in the group of women without torsion and were not discriminant for the diagnosis of adnexal torsion. The assessment of the viability of the ipsilateral ovary is another important issue. In our study, 10 women were found to have nonviable ovaries at surgery. However, we found no correlation between viability and streaky changes in adjacent fat, as demonstrated at CT (18), because this sign is not specific and may also be found in pelvic inflammatory disease. Finally, in line with our second-line recruitment method in the acute and subacute period, no correlation was found with either clinical findings or the presence of an underlying mass, as demonstrated by using US (18). Aside from its retrospective singlecenter design, our study has several limitations. First, the readers were biased because they knew that the topic of this study was torsion. Second, we did not evaluate the value of diffusionweighted imaging (19 21) or dynamic contrast-enhanced imaging. Both were performed in a subgroup of patients included in the present study, but the limited number of these cases (n = 33 and n = 31) did not allow us to perform valid statistical analysis. In our institution, diffusion-weighted and dynamic contrast-enhanced MR sequences are routinely performed for the characterization of complex or indeterminate adnexal masses. Moreover, in the present setting of acute pelvic pain, our goal was to quickly evaluate all the patients, especially when they were pregnant (n = 6) at the time of diagnosis. Third, all patients underwent prior US examination, and MR imaging was performed because a definite diagnosis of adnexal torsion had not been assigned. As a consequence, our recruitment bias may limit extrapolation to all patients who experience acute or subacute pelvic pain. Fourth, we did not compare the diagnostic performance of MR imaging to that of CT, and we evaluated the performance of US on the basis of variable expertise (gynecologist or radiologist sonographer). Another limitation of our study is the small size of the population, which required a calculation of sensitivity and specificity values on the basis of a simple stratified report of the data and not on thresholds issued from the statistical model. Finally, not all patients underwent surgery, and some cases of adnexal torsion might have been missed initially. In conclusion, MR imaging is an accurate second-line technique for the diagnosis of adnexal torsion in the setting of patients with an equivocal adnexal mass at US who have acute or subacute pelvic pain. A whirlpool sign and a thickened tube should therefore be actively searched for in this context. Disclosures of Conflicts of Interest: S.B.G. disclosed no relevant relationships. H.S. disclosed no relevant relationships. M. Ballester disclosed no relevant relationships. A.R. disclosed no relevant relationships. M. Bornes disclosed no relevant relationships. M. Bazot disclosed no relevant relationships. E.D. disclosed no relevant relationships. I.T.N. Activities related to the present article: disclosed no relevant relationships. Activities not related to the present article: author received payment and nonfinancial support from General Electric for lectures and travel for a congress. Other relationships: disclosed no relevant relationships. References 1. Hibbard LT. Adnexal torsion. Am J Obstet Gynecol 1985;152(4): Gross M, Blumstein SL, Chow LC. Isolated fallopian tube torsion: a rare twist on a common theme. AJR Am J Roentgenol 2005;185(6): Sasaki KJ, Miller CE. Adnexal torsion: review of the literature. J Minim Invasive Gynecol 2014;21(2): Warner MA, Fleischer AC, Edell SL, et al. Uterine adnexal torsion: sonographic findings. Radiology 1985;154(3): Nichols DH, Julian PJ. Torsion of the adnexa. Clin Obstet Gynecol 1985;28(2): Bayer AI, Wiskind AK. Adnexal torsion: can the adnexa be saved? Am J Obstet Gynecol 1994;171(6): ; discussion Celik A, Ergün O, Aldemir H, et al. Longterm results of conservative management of adnexal torsion in children. J Pediatr Surg 2005;40(4): Huchon C, Panel P, Kayem G, Schmitz T, Nguyen T, Fauconnier A. Does this woman have adnexal torsion? Hum Reprod 2012; 27(8): Lomano JM, Trelford JD, Ullery JC. Torsion of the uterine adnexa causing an acute abdomen. Obstet Gynecol 1970;35(2): Anders JF, Powell EC. Urgency of evaluation and outcome of acute ovarian torsion in pediatric patients. Arch Pediatr Adolesc Med 2005;159(6): Chung JC, Cho GS, Shin EJ, Kim HC, Song OP. Clinical outcomes compared between laparoscopic and open appendectomy in pregnant women. Can J Surg 2013;56(5): Graif M, Shalev J, Strauss S, Engelberg S, Mashiach S, Itzchak Y. Torsion of the ovary: sonographic features. AJR Am J Roentgenol 1984;143(6): Wilkinson C, Sanderson A. Adnexal torsion a multimodality imaging review. Clin Radiol 2012;67(5): Ghossain MA, Buy JN, Bazot M, et al. CT in adnexal torsion with emphasis on tubal findings: correlation with US. J Comput Assist Tomogr 1994;18(4): Ghossain MA, Buy JN, Sciot C, Jacob D, Hugol D, Vadrot D. CT findings before and after adnexal torsion: rotation of a focal solid element of a cystic adjunctive sign in diagnosis. AJR Am J Roentgenol 1997;169(5): Rha SE, Byun JY, Jung SE, et al. CT and MR imaging features of adnexal torsion. Radio- Graphics 2002;22(2): Padovan RS, Kralik M, Prutki M, Hrabak M, Oberman B, Potocki K. Cross-sectional imaging of the pelvic tumors and tumor-like lesions in gynecologic patients misinterpretation points and differential diagnosis. Clin Imaging 2008;32(4): Chiou SY, Lev-Toaff AS, Masuda E, Feld RI, Bergin D. Adnexal torsion: new clinical and imaging observations by sonography, computed tomography, and magnetic resonance imaging. J Ultrasound Med 2007;26(10): Fujii S, Kaneda S, Kakite S, et al. Diffusion-weighted imaging findings of adnexal torsion: initial results. Eur J Radiol 2011; 77(2): Radiology: Volume 279: Number 2 May 2016 n radiology.rsna.org 469

10 20. Kilickesmez O, Tasdelen N, Yetimoglu B, Kayhan A, Cihangiroglu M, Gurmen N. Diffusion-weighted imaging of adnexal torsion. Emerg Radiol 2009;16(5): Kato H, Kanematsu M, Uchiyama M, Yano R, Furui T, Morishige K. Diffusion-weighted imaging of ovarian torsion: usefulness of apparent diffusion coefficient (ADC) values for the detection of hemorrhagic infarction. Magn Reson Med Sci 2014;13(1): Lourenco AP, Swenson D, Tubbs RJ, Lazarus E. Ovarian and tubal torsion: imaging findings on US, CT, and MRI. Emerg Radiol 2014;21(2): Hiei K, Takagi H, Matsunami K, Imai A. Ovarian torsion; early diagnosis by MRI to prevent irreversible damage. Clin Exp Obstet Gynecol 2010;37(3): Kimura I, Togashi K, Kawakami S, Takakura K, Mori T, Konishi J. Ovarian torsion: CT and MR imaging appearances. Radiology 1994;190(2): Landis JR, Koch GG. An application of hierarchical kappa-type statistics in the assessment of majority agreement among multiple observers. Biometrics 1977;33(2): Auslender R, Shen O, Kaufman Y, et al. Doppler and gray-scale sonographic classification of adnexal torsion. Ultrasound Obstet Gynecol 2009;34(2): Auslender R, Lavie O, Kaufman Y, Bardicef M, Lissak A, Abramovici H. Coiling of the ovarian vessels: a color Doppler sign for adnexal torsion without strangulation. Ultrasound Obstet Gynecol 2002;20(1): Valsky DV, Cohen SM, Hamani Y, Lipschuetz M, Yagel S, Esh-Broder E. Whirlpool sign in the diagnosis of adnexal torsion with atypical clinical presentation. Ultrasound Obstet Gynecol 2009;34(2): Rouvière H. Anatomie humaine descriptive et topographique. Paris, France: Masson, Ghossain MA, Hachem K, Buy JN, et al. Adnexal torsion: magnetic resonance findings in the viable adnexa with emphasis on stromal ovarian appearance. J Magn Reson Imaging 2004;20(3): radiology.rsna.org n Radiology: Volume 279: Number 2 May 2016