In postmenopausal women, endometrial stripe measurements by transvaginal ultrasonography are used to exclude endometrial neoplasia. Limited information exists on the use of endometrial stripe measurements for premenopausal women.1,2 Endometrial thickness varies throughout the menstrual cycle; thus, a thickened stripe may not indicate pathology. A variety of specific endometrial stripe cutoff values ranging from 3 to 14 mm for the detection of pathology in premenopausal women has been evaluated with variable results.1–4 Goldstein et al5 developed an algorithm for the sonographic triage of abnormal uterine bleeding in premenopausal women by using an endometrial stripe cutoff of 5 mm to eliminate pathology. The sensitivity and specificity of the 5-mm cutoff value was not reported in the investigation by Goldstein et al.5
Sonohysterography frequently is used in the evaluation of abnormal uterine bleeding. Sonohysterography, however, is more invasive, causes more discomfort than unenhanced transvaginal sonography, and requires more technical expertise. It would therefore be desirable to limit its use only when indicated. A cutoff value below which benign endometrial pathology could be ruled out would then be clinically useful. We hypothesized that a transvaginal endometrial stripe of less than 5 mm in premenopausal women would exclude endometrial polyps and intracavitary leiomyomata, thereby eliminating the need for sonohysterography.
MATERIALS AND METHODS
Data were retrospectively collected from 206 consecutive, clinically indicated sonohysterographies performed on premenopausal women between August 1998 and October 2001. Patients were referred for abnormal uterine bleeding by university-affiliated obstetrics and gynecology practices or the resident gynecology clinic at the University of Texas Medical Branch and were scheduled for sonographic examination in the proliferative phase of the cycle whenever possible. Medical records were reviewed to obtain subsequent operative findings and surgical pathology results from endometrial sampling, hysteroscopic surgery, or hysterectomy when available. Institutional review board approval was obtained before data collection. All sonographic examinations were performed in an American Institute of Ultrasound in Medicine–accredited referral practice on ATL 3000 equipment (Philips Medical Systems, Andover, MA) by American Registry of Diagnostic Medical Sonographers–credentialed individuals and read by 1 of 2 sonologists. Baseline measurements of endometrial stripe thickness were obtained in the longitudinal axis view, measuring the thickest portion, including intracavitary masses if present. Either an 8F pediatric Foley catheter with a 3-mL balloon or a 5F pediatric feeding tube was used to instill saline into the endometrial cavity. Sonohysterography was performed under real-time visualization of the endometrium.6 Intracavitary masses, if present, were measured in 2 orthogonal planes, and still images were recorded.
Gross endometrial abnormalities can be detected on endovaginal sonography regardless of stripe thickness.1 To identify these cases, baseline endometrial stripe images were reviewed by one of the authors (D.M.B.), who was blinded to the subsequent sonohysterography and/or pathology results. The morphologic appearance of the endometrium was classified as either normal or abnormal. An abnormal endometrial stripe was defined as either irregular in shape or echotexture or consistent with a mass, endometrial polyp,or leiomyoma.
Data were compiled and analyzed in Microsoft Excel (Microsoft, Redmond, WA). Sensitivity, specificity, predictive values, and likelihood ratios were calculated for the diagnostic performance of transvaginal ultrasound measurement of the endometrial stripe in detecting intracavitary masses. The 95% confidence intervals (95% CIs) were estimated for each value.
Likelihood ratios provide a means to evaluate diagnostic test accuracy that is independent of disease prevalence unlike predictive values. Likelihood ratios greater than 10 and less than 0.1 indicate a substantial and clinically significant change from pretest to posttest probability will result. Moderate change in posttest probability will result if the likelihood ratio is between 5 and 10 or 0.1 to 0.2. Likelihood ratios between 1 and 2 or 0.5 to 1 rarely result in significant modification of posttest probability.7
Data were reanalyzed after the exclusion of cases in which review of images identified a morphologically abnormal endometrial stripe (as defined above) on the unenhanced baseline image. Receiver operating characteristic curves were constructed based on sensitivity and specificity derived from 2 × 2 tables for each integer cutoff value for endometrial stripe thickness using Sigma Plot (SPSS Science, Chicago, IL) and are presented in the format used by Dijkhuizen et al.2 The area under the curve for the receiver operating characteristic curves were calculated and compared with the curve obtained from a random test with no discriminatory ability (ie, the line of no information where the area under the curve = 0.5,) using the Mann-Whitney U test in AccuROC (Accumetric Corporation, Montreal, Quebec, Canada). Statistical significance was set at P < .05.
In 6 of the 206 sonohysterograms reviewed, the endometrium could not be measured because of lack of visualization; these were excluded from the analysis of endometrial stripe accuracy (Fig. 1). Study patients were between 18 and 53 years of age (mean, 38.2; standard deviation, 7.5). Median age was 38 years in the subjects with an endometrial stripe thickness less than 5 mm (range 18–52 years) and 39 years in subjects with an endometrial stripe thickness of greater than or equal to 5 mm (range 21–53 years). Median age was not statistically significantly different between the 2 groups (P = .22, Mann–Whitney rank sum test).
Pathologic and/or surgical confirmation of the sonohysterographic findings was available in 97 (47%) of the 206 cases. This subset of patients with proven pathologic and/or surgical findings was analyzed to determine the accuracy of sonohysterography in the detection of polyps and intracavitary myomas in our practice. Intracavitary masses were present in 41 of 97 cases. Sonohysterography detected 40 of the 41 polyps or submucosal fibroids and correctly identified 55 of 56 cases negative for intracavitary masses for a sensitivity of 99% (95% CI 87–100) and a specificity of 98% (95% CI 91–100).
Of the 200 patients who underwent sonohysterography and whose unenhanced endometrial stripe measurement was measured, 40% (n = 80) had an endometrial stripe measurement of less than 5 mm (Table 1). Of these 80 subjects with an endometrial stripe of less than 5 mm, 20% (n = 16) had intracavitary masses on subsequent sonohysterography: 11 had polyps, and 5 had fibroids. The largest diameter of the 11 polyps ranged from 0.41 to 1.61 cm (mean 0.97 cm; standard deviation 0.37 cm). Of the 5 leiomyomas, 2 were completely intracavitary in location on sonohysterography, whereas 3 had an intramural component.
The accuracy of using an unenhanced endometrial stripe thickness as a cutoff value for detection of endometrial polyps and intracavitary leiomyomas was calculated by using sonohysterography as the “gold standard.” The overall prevalence of polyps and intracavitary fibroids in the sample (n = 200) was 31%. When using an endometrial stripe thickness of less than 5 mm, the sensitivity was 74% (95% CI 62–84), the specificity was 46% (95% CI 38–54), the positive predictive value was 37%, and the negative predictive value was 80%. The positive likelihood ratio for detection of intracavitary masses was 1.35 (95% CI 1.09–1.68), whereas the negative likelihood ratio was 0.58 (95% CI 0.37–0.92). Based on the negative likelihood ratio, for every 6 subjects with an endometrial stripe less than 5 mm, 1 had an endometrial polyp or intracavitary myoma not detected by the unenhanced imaging.
Endometrial polyps and fibroids occasionally can be suspected based on unenhanced transvaginal sonography regardless of endometrial stripe thickness. The identification of these abnormalities could theoretically improve the calculated accuracy of unenhanced transvaginal sonography in patients with an endometrial stripe less than 5 mm. To examine the effect of excluding morphologic abnormalities present on baseline transvaginal sonography, we performed a retrospective review of the still-transvaginal images to identify such cases as has been done in previous reports.1,2 Upon blinded review of the unenhanced images, 55 of the 200 cases had endometrial stripes that were abnormal in appearance. Among cases with an endometrial stripe thickness less than 5 mm, the baseline endometrial images were classified as abnormal in 3 of the 11 polyps. Two of the 5 leiomyomas were identified as at least partly intracavitary in location on the presonohysterography images. After exclusion of the 55 cases with abnormal stripe appearance, the sensitivity of an endometrial stripe thickness less than 5 mm was 54% (95% CI 34–74), specificity was 46% (95% CI 37–55), the positive predictive value was 23%, the negative predictive value was 83%, positive likelihood ratio was 1.01 (95% CI 0.68–1.51), and the negative likelihood ratio was 0.83 (95% CI 0.75–0.92).
A receiver operating characteristic curve was constructed to evaluate whether other cutoff values would be more accurate than 5 mm in the detection of benign endometrial masses (Fig. 2). A test is more accurate to the extent the true-positive rate approaches 1.0 and the false-positive rate approaches 0 (ie, the closer the curve approaches the upper left corner of the graph). As the figure demonstrates, the use of endometrial stripe alone does not perform well in the detection of benign endometrial masses in premenopausal women regardless of the cutoff value chosen.
The area under the receiver operating characteristic curve for detection of benign endometrial masses in all cases was 0.66 (95% CI 0.57–0.75) and was statistically significantly different from the line of no information (P < .01). Excluding cases with abnormally appearing endometrial stripes, the area under the curve was 0.49 (95% CI 0.35–0.62) and was not significantly different from the line of no information (P = .82).
In this study, endometrial polyps and leiomyomas were not detected by transvaginal sonography in 1 of 6 women with an endometrial stripe thickness less than 5 mm, for a sensitivity of 74%. Using the negative likelihood ratio, the posttest probability for the presence of an endometrial polyp or submucous myoma was 20% in cases where endometrial stripe thickness was less than 5 mm. Although the receiver operating characteristic curve for all cases was statistically significantly different than the line of no information, clinical significance was not achieved. Test performance as measured by likelihood ratios or sensitivity remained suboptimal regardless of endometrial stripe thickness cutoff value, as demonstrated by the receiver operating characteristic curve. The detection of pathology on the basis of solely an endometrial stripe of less than 5 mm was not improved even when cases with a morphologically abnormal endometrial stripe were excluded.
Transvaginal ultrasonography alone may have missed intracavitary masses detected by sonohysterography for several reasons. The fluid contrast obtained with sonohysterography more clearly outlines the borders of masses, allowing for easier detection. Polyps are often long and narrow in contour, conforming to the shape of the endometrial cavity.1 The endometrial echo may not be enlarged in the longitudinal plane with these narrow intracavitary structures, thus escaping detection by solely relying on endometrial stripe thickness.
Other sonographic markers besides endometrial stripe thickness may be helpful in identifying pathology with unenhanced transvaginal sonography. Baldwin et al8 evaluated the hyperechoic line sign in 42 pre- and postmenopausal women with abnormally appearing endometrial stripes by transvaginal sonography and sonohysterography. All 25 patients with a hyperechoic line sign had a focal intracavitary process as defined as a polyp, fibroid, or focal endometrial thickening. However, 10 of 17 patients without the hyperechoic line sign had focal intracavitary processes. Thus, the authors recommended sonohysterography in patients with thickened endometrial stripes if the hyperechoic line sign is not present. They did not evaluate those women with stripe of normal thickness, nor were cutoff values for stripe thickness defined.
Another sonographic marker for the detection of intracavitary pathology by transvaginal sonography is the pedicle artery sign. This sign, as defined by Timmerman et al,9 involves the detection of a feeding blood vessel reaching the central endometrial echo by color Doppler. Timmerman et al evaluated 3,099 pre- and postmenopausal women for presence of the pedicle artery sign; however, no gold standard was available in 2,230. On the basis of this sample, the sign had sensitivity for detection of endometrial polyps of 76.4% and a specificity of 95.3%. Although the pedicle artery sign shows promise as a tool for identifying endometrial polyps, it does not approach the accuracy of sonohysterography.
This study did not evaluate the pedicle artery sign or the hyperechoic line sign. These markers may improve the performance of transvaginal sonography despite the presence of a thin endometrial stripe. Further studies of these noninvasive sonographic techniques to detect intracavitary pathology are needed to better define their role in the evaluation of abnormal uterine bleeding.
This study uses an unselected large sample from a high-volume clinical sonography referral practice. Trained, experienced personnel performed all exams. Sonohysterography was used as the gold standard for the detection of intracavitary masses. Numerous reports have demonstrated high specificity and sensitivity of sonohysterography for detection of benign endometrial pathology, ranging from 86% to 100%.10–13 Our data were consistent with these findings.
The thickness of the endometrial stripe varies throughout the menstrual cycle, and in this study we evaluated subjects in different menstrual phases because of scheduling constraints. Standardization of the menstrual timing of ultrasound examination to the early proliferative phase could be predicted to lead to fewer false-positive tests. In women with abnormal bleeding, however, it is often difficult to determine the menstrual phase, making the timing of the ultrasound examination difficult to standardize. This would not, however, have affected the calculations of accuracy in this study because the goal was to determine whether a thin endometrial stripe is useful in excluding pathology.
Laifer-Narin et al1 retrospectively reviewed findings from 180 patients undergoing sonohysterography. Of the 180 pre- and postmenopausal patients, 114 had abnormalities on sonohysterography. Fourteen percent of the patients with abnormal sonohysterography findings had a normal-appearing endometrium and an endometrial stripe thickness of less than 14 mm as measured by conventional transvaginal sonography. These authors did not report sensitivity, specificity, predictive values, or likelihood ratios.
Dueholm et al14 measured the endometrial stripe by transvaginal ultrasonography in 329 premenopausal Danish women with abnormal uterine bleeding and indications for endometrial sampling or surgery. Gold standards used included sonohysterography, operative hysteroscopy, or hysterectomy. Endometrial hyperplasia or polyps were found in 8% of 143 patients with an endometrial stripe thickness of less than 7 mm. The performance of endometrial stripe thickness in the detection of these conditions did not improve with lower cutoff values. In those patients with a normal-appearing endometrial stripe of less than 5 mm, 19% had a polyp, submucosal leiomyoma, or hyperplasia.
Dijkhuizen et al2 evaluated the cutoff level of 5 mm in premenopausal women for the diagnosis of intracavitary abnormalities. They prospectively compared transvaginally obtained endometrial stripe measurements with sonohysterography and histologic findings in 50 premenopausal women undergoing hysterectomy for abnormal uterine bleeding. The sensitivity and specificity of an endometrial stripe of 5 mm for detection of polyps and myomas was 61% and 96%, respectively. The negative likelihood ratio was 0.41. A poorly defined endometrium was considered abnormal. Three myomas were missed when using a 5-mm cutoff. Our data are consistent with this report in that endometrial polyps and leiomyomas were present among women with a thin endometrial stripe. Our study, however, included a larger sample size than Dijkhuizen et al2 did. Their study included only subjects who were already scheduled for hysterectomy. Women scheduled for hysterectomy represent a sample that excludes a proportion of patients with polyps and submucous myomas identified and treated by more conservative methods. Our study, in contrast, included women who were referred for the diagnostic evaluation of abnormal uterine bleeding.
We could be criticized that the analysis of measured outcomes (presence or absence of intracavitary masses) did not necessarily equate with actual etiology of symptoms. For instance, some would question whether endometrial polyps less than 1 cm in diameter are clinically significant. This investigation does not address this question and the answer awaits pre- and posttreatment clinical trials in women with known polyps and intracavitary myomas.
These results suggest that endometrial stripe thickness alone should not be used to exclude benign endometrial masses in premenopausal women. Endometrial polyps and intracavitary leiomyomas are common etiologies for abnormal uterine bleeding. Using an endometrial stripe cutoff of 5 mm in premenopausal women would miss significant pathology. Other sonographic markers of intracavitary pathology, if present, may provide additional information but have not been as well studied as sonohysterography. Based on the findings of this investigation, we conclude that sonographic evaluation of abnormal uterine bleeding should include sonohysterography or equivalently accurate testing regardless of endometrial stripe thickness.
1. Laifer-Narin S, Ragavendra N, Parmenter EK, Grant EG. False-normal appearance of the endometrium on conventional transvaginal sonography: comparison with saline hysterosonography. AJR Am J Roentgenol 2002;178:129–33.
2. Dijkhuizen FP, De Vries LD, Mol BW, Brolmann HA, Peters HM, Moret E, et al. Comparison of transvaginal ultrasonography and saline infusion sonography for the detection of intracavitary abnormalities in premenopausal women. Ultrasound Obstet Gynecol 2000;15:372–6.
3. Krampl E, Bourne T, Hurlen-Solbakken H, Istre O. Transvaginal ultrasonography sonohysterography and operative hysteroscopy for the evaluation of abnormal uterine bleeding. Acta Obstet Gynecol Scand 2001;80:616–22.
4. Dijkhuizen FP, Brolmann HA, Potters AE, Bongers MY, Heinz AP. The accuracy of transvaginal ultrasonography in the diagnosis of endometrial abnormalities. Obstet Gynecol 1996;87:345–9.
5. Goldstein SR, Zeltser I, Horan CK, Snyder JR, Schwartz LB. Ultrasonography-based triage for perimenopausal patients with abnormal uterine bleeding. Am J Obstet Gynecol 1997;177:102–8.
6. Breitkopf D, Goldstein SR, Seeds JW, for the ACOG Committee on Gynecologic Practice. ACOG technology assessment in obstetrics and gynecology. Number 3, September 2003. Saline infusion sonohysterography. Obstet Gynecol 2003;102:659–62.
7. Jaeschke R, Guyatt GH, Sackett DL. Users’ guides to the medical literature. III. How to use an article about a diagnostic test. B. What are the results and will they help me in caring for my patients? The Evidence-Based Medicine Working Group. JAMA 1994;271:703–7.
8. Baldwin MT, Dudiak KM, Gorman B, Marks CA. Focal intracavitary masses recognized with the hyperechoic line sign at endovaginal US and characterized with hysterosonography. Radiographics 1999;19:927–35.
9. Timmerman D, Verguts J, Konstantinovic ML, Moerman P, Van Schoubroeck D, Deprest J, et al. The pedicle artery sign based on sonography with color Doppler imaging can replace second-stage tests in women with abnormal vaginal bleeding. Ultrasound Obstet Gynecol 2003;22:166–71.
10. Bernard JP, Lecuru F, Darles C, Robin F, de Bievre P, Taurelle R. Saline contrast sonohysterography as first-line investigation for women with uterine bleeding. Ultrasound Obstet Gynecol 1997;10:121–5.
11. Schwarzler P, Concin H, Bosch H, Berlinger A, Wohlgenannt K, Collins WP, et al. An evaluation of sonohysterography and diagnostic hysteroscopy for the assessment of intrauterine pathology. Ultrasound Obstet Gynecol 1998;11:337–42.
12. Soares SR, Barbosa dos Reis MM, Camargos AF. Diagnostic accuracy of sonohysterography, transvaginal sonography, and hysterosalpingography in patients with uterine cavity diseases. Fertil Steril 2000;73:406–11.
13. Farquhar C, Ekeroma A, Furness S, Arroll B. A systematic review of transvaginal ultrasonography, sonohysterography and hysteroscopy for the investigation of abnormal uterine bleeding in premenopausal women [review]. Acta Obstet Gynecol Scand 2003;82:493–504.
14. Dueholm M, Jensen ML, Laursen H, Kracht P. Can the endometrial thickness as measured by trans-vaginal sonography be used to exclude polyps or hyperplasia in pre-menopausal patients with abnormal uterine bleeding? Acta Obstet Gynecol Scand 2001;80:645–51.