Transvaginal ultrasonography has been proposed to be the test of first choice in postmenopausal women with vaginal bleeding because of its almost perfect accuracy,1 although others2 report this accuracy to be lower. Because of the fact that the probability of malignancy is strongly reduced in case of an endometrial thickness of 4 mm or less, expectant management may be justified in women with such test results. In case the endometrial thickness is 5 mm or more, endometrial sampling is advised to exclude the possibility of endometrial cancer.3,4 Cost-effectiveness analysis showed that a diagnostic strategy starting with transvaginal ultrasonography followed by endometrial biopsy in case of an increased endometrial thickness was the most cost-effective strategy when the prevalence of endometrial carcinoma was less than 15%.5
Smith-Bindman et al1 performed a meta-analysis summarizing the available evidence on the accuracy of transvaginal ultrasonography in the detection of endometrial cancer. The authors concluded that transvaginal ultrasonography of the endometrium identifies women that are unlikely to have significant endometrial disease and therefore would not benefit from endometrial sampling. Tabor et al6 also performed a meta-analysis on the subject. In contrast to the conclusion of Smith-Bindman, Tabor et al concluded that transvaginal ultrasonography did not reduce the need for invasive diagnostic testing in women with postmenopausal bleeding. The discrepancy between the conclusion of Smith-Bindman and Tabor might be the result of their different meta-analytic approaches; whereas Tabor et al used real distributions of endometrial thickness in diseased and nondiseased patients, Smith-Bindman was likely to use cutoff points at which sensitivity and specificity were optimized, with the latter approach leading to inflated estimates of test accuracy.
An issue that was not addressed in both meta-analyses nor in many other studies on the subject was that test characteristics of endometrial thickness measurement for the detection of endometrial cancer might be affected by particular patient characteristics. It is known in asymptomatic postmenopausal women that endometrial thickness is associated with certain risk factors for endometrial cancer, such as obesity, diabetes, hypertension, as well as current use of hormone replacement therapy (HRT), age, parity, and smoking.7–10 It is likely that the diagnostic accuracy of transvaginal ultrasonography in detecting endometrial cancer is different in obese and nonobese women, in diabetic and nondiabetic women, in hypertensive and normotensive women, and in HRT users versus non-HRT users. Indeed, Smith-Bindman et al1 reported a higher accuracy of endometrium thickness in the detection of endometrial carcinoma among women who were not using HRT as compared with women who were using HRT. Furthermore, these characteristics affect the prior probability for a woman with postmenopausal bleeding to have endometrial cancer.11 We are not aware of any studies that have assessed the association among patient characteristics like diabetes, obesity, and hypertension and the accuracy of transvaginal ultrasonography.
The first aim of the present prospective study among consecutive patients with postmenopausal bleeding was therefore to assess the impact of different patient characteristics (obesity, hypertension, and diabetes) on endometrial thickness measured by transvaginal ultrasonography in relation to the presence or absence of endometrium cancer. The second aim was to compare the accuracy of the endometrial thickness measurement in the diagnosis of endometrial cancer in patients with obesity, diabetes, and hypertension.
MATERIALS AND METHODS
The study was performed in 1 university hospital and 7 teaching hospitals in The Netherlands. Between January 2002 and January 2003, consecutive patients who presented with postmenopausal bleeding were registered prospectively. The study was limited to women not using HRT. For each patient, a case record form was completed containing body mass index (BMI), relevant medical history, and comorbidity. Previously diagnosed diabetes and hypertension, as well as medication use, were recorded as stated by the patients.
Women were evaluated according to the guideline of the Dutch Society of Obstetrics and Gynecology.12 Transvaginal ultrasonography was performed by the gynecologist in the outpatient office or by specially trained ultrasonographists, as is customary, in the hospital where the patient was evaluated. All women underwent a transvaginal ultrasound examination with high-frequency (5–7.5 MHz) transducers from different manufacturers. Endometrial thickness was measured as a double layer measurement at its thickest part in the longitudinal plane. In case it was not possible to measure the endometrial thickness in a reliable way, this limitation also was recorded. When the endometrial layers were separated by intracavitary fluid, both layers were measured and the sum recorded. In case the endometrial thickness was 4 mm or less, the patient was reassured and instructed to contact the doctor if new bleeding occurred. If the endometrial thickness was 5 mm or more, a biopsy was performed using a Pipelle sampler (Pipelle Laboratories CCD, Paris, France), during hysteroscopy, or with dilatation and curettage.
Definite disease state was determined as follows: “no abnormality” was diagnosed in women with an endometrial thickness 4 mm or less with an uneventful follow-up or in women whose specimens showed atrophy, benign polyps, simple hyperplasia, or proliferative endometrium on histology. Hyperplasia with atypia and malignant findings in the histology specimen were combined in the analyses in the diagnostic group “(pre)malignancy” because both diagnostic categories warrant further treatment.
The aim of the analysis was to assess the accuracy of endometrial thickness as measured at transvaginal ultrasonography in the diagnosis of atypical hyperplasia and endometrial cancer in different subgroups of patients with postmenopausal bleeding. First, a receiver operating characteristic curve was constructed that illustrated the capacity of endometrial thickness to discriminate between patients with and without endometrial (pre)malignancy. Subsequently, the area under the receiver operating characteristic curve was calculated. This area under the curve expresses the performance of a diagnostic test, taking values in the range between 0.5 and 1. An area under the receiver operating characteristic curve of 0.5 implies that the diagnostic test under study has a discriminative capacity that does not exceed chance, whereas an area under the curve of 1 implies that the discriminative capacity of the test under study is perfect. Subsequently, the diagnostic accuracy of endometrial thickness measurements was associated with patient characteristics. To do so, we constructed scatter plots for the endometrial thickness in hypertensive versus normotensive women, for diabetic versus nondiabetic women, and for obese versus nonobese women, in patients both with and without a (pre)malignancy of the endometrium. To assess the association between obesity and endometrial thickness, we calculated a correlation coefficient between endometrial thickness and BMI, both in women with a (pre)malignancy of the endometrium and in women with normal findings, separately.
We then compared the distribution of endometrial thickness in subgroups of patients using the Kolmogorov-Smirnov test.13 In all comparisons, P < .05 was considered to indicate statistically significant differences between these distributions. Such differences could have impact on the optimal cutoff level and therefore could be of clinical significance.
Subsequently, receiver operating characteristic analysis was performed in the subgroups of patients with and without the specified characteristic, and likelihood ratios and their 95% confidence intervals were calculated for various endometrial thickness measurements in patients with hypertension, obesity, diabetes, and patients without risk indicators. A likelihood ratio of a particular test result is defined as the probability of that particular test result under condition of the presence of disease, ie, (pre)malignancy, compared with the probability of that particular test result under condition of the absence of disease, ie, nonmalignancy. A likelihood ratio greater than 1 increases the probability of disease, whereas a likelihood ratio less than 1 decreases the probability of disease. A likelihood ratio of 1 indicates that a particular test result has no diagnostic value.14
During the study period, 688 patients were included. In 28 patients, the results of transvaginal ultrasonography were missing. It was not possible to perform transvaginal ultrasonography in 5 of the remaining 660 patients, whereas in 68 patients the ultrasonographist concluded that measurement of endometrial thickness by transvaginal ultrasonography was not reliable at the first attempt. In 7 of these patients, however, a second attempt was successful. Thus, the endometrial thickness as measured at ultrasonography was available in 594 women.
Table 1 presents subject characteristics. There were 66 (11%) nulliparous women. Diabetes was present in 68 patients, of whom 10 were diet controlled, 30 were treated with oral drugs, and 28 were insulin dependent. Hypertension requiring medical treatment was present in 129 patients, whereas 34 patients were hypertensive but were not taking medication. The BMI was known in 510 women. Among the included patients, 29% had a BMI of 30 or greater (obese).
Among the 594 patients, 62 (10%) had endometrial cancer, and 6 (1%) had atypical hyperplasia. Among the 94 patients who were excluded from the study because of missing or not interpretable ultrasonography measurements, 11 women had carcinoma and 2 had atypical hyperplasia, resulting in a prevalence of 14%. In the remaining 594 patients in whom endometrial thickness as measured at ultrasonography was available, a receiver operating characteristic curve was constructed demonstrating the capacity of endometrial thickness to discriminate between patients with and without premalignant endometrium (Fig. 1). The area under the curve was 0.87 (standard error [SE] 0.03).
Figure 2A shows a scatter plot of endometrial thickness in diabetic and nondiabetic patients, stratified for endometrial cancer. Among the 68 patients with diabetes, 14 (21%) had a (pre)malignancy of the endometrium. Among the 526 without diabetes, a (pre)malignancy from the endometrium was present in 54 patients (10%; P = .002). In patients with a (pre)malignancy of the endometrium, the mean endometrial thickness was 15.2 (standard deviation [SD] 8.2) and 13.7 mm (SD 7.0) in diabetic and nondiabetic patients, respectively (P = .67). The mean endometrial thickness in diabetic and nondiabetic patients without a (pre)malignancy was 8.6 (SD 7.3) and 5.2 mm (SD 4.6), respectively (P = .01).
Figure 3A shows a scatter plot of endometrial thickness in hypertensive and normotensive patients, stratified for (pre)malignancy of the endometrium. Among the 163 patients with hypertension, 26 (16%) had a (pre)malignancy of the endometrium. Among the 431 without hypertension, a (pre)malignancy of the endometrium was present in 42 patients (9.7%). In patients with a (pre)malignancy of the endometrium, the mean endometrial thickness was 16.2 (SD 8.0) and 12.6 mm (SD 6.4) in hypertensive and normotensive patients, respectively (P = .24). The mean endometrial thickness in patients without a (pre)malignancy of the endometrium with and without hypertension was 6.3 (SD 5.1) and 5.3 mm (SD 5.0), respectively (P = .01).
Figure 4A shows a scatter plot of endometrial thickness in obese and nonobese patients stratified for the presence of (pre)malignancy. The BMI was known in 510 patients. Among the 148 patients with obesity, 26 (18%) had a (pre)malignancy of the endometrium. Among the 362 women without obesity, a (pre)malignancy of the endometrium was present in 31 patients (8.6%). In patients with (pre)malignancy of the endometrium, the mean endometrial thickness was 13.0 (SD 5.8) and 14.8 mm (SD 7.8) in obese and nonobese patients, respectively (P = .99). The mean endometrial thickness in patients without (pre)malignancy of the endometrium in obese and nonobese patients was 8.0 (SD 5.3) and 4.7 mm (SD 4.6), respectively (P = .001). There was a significant correlation between BMI and endometrial thickness in noncancer patients (0.29; P < .001), but this correlation was absent in patients with cancer (0.03, P = .80).
In view of the statistically significant differences in endometrial thickness in noncancer patients (between diabetic and nondiabetic patients, between hypertensive and nonhypertensive patients, and between obese and nonobese patients), we decided to construct receiver operating characteristic curves in these subcategories of patients. Figure 2B shows the receiver operating characteristic curves stratified for the presence of diabetes. The area under the curve was 0.75 (SE 0.07) for women with diabetes and 0.88 (SE 0.03) for women without diabetes. Figure 3B shows that the receiver operating characteristic curves were virtually similar in women with and without hypertension, with areas under the curve of 0.88 (SE 0.04) and 0.87 (SE 0.03), respectively. Figure 4B shows the receiver operating characteristic curves for obese and nonobese patients, with areas under the curve of 0.74 (SE 0.05) and 0.90 (SE 0.03), respectively.
The distribution of endometrial thickness in women with diabetes, obesity, or hypertension and in patients with none of these risk indicators is summarized in Table 2. In patients with diabetes and in obese patients, the likelihood ratio was only increased if the endometrial thickness exceeded 15 mm. In hypertensive patients, the likelihood ratio was increased if the endometrial thickness exceeded 10 mm, as was the accuracy in women without any risk indicators. However, in women with hypertension, the likelihood ratio of an endometrial thickness between 10 mm and 15 mm was 3.2, whereas the likelihood ratio for a similar test result in women without any risk indicators was 5.3.
Because diabetes and obesity are correlated to each other, the above data do not demonstrate whether diabetes and obesity are independent factors in the decrease of diagnostic accuracy. To evaluate this potential interaction between diabetes and obesity, we performed subgroup analysis. There were 27 patients with diabetes who had a BMI less than 30. Three of these 27 patients had cancer or atypical hyperplasia, resulting in a prevalence of 11%. There were 117 patients with a BMI greater than 30 mg/kg2 who did not have diabetes. Of these patients, 17 had cancer or atypical hyperplasia, resulting in a prevalence of 15%. Nine of 31 obese diabetics had cancer or atypical hyperplasia, resulting in a prevalence of 29%. The areas under the receiver operating characteristic curve were 0.88 (SE 0.14) for nonobese patients with diabetes, 0.80 (SE 0.06) for obese patients without diabetes, and 0.59 (SE 0.11) for obese diabetics.
This study shows that the accuracy of ultrasonographic transvaginal endometrial thickness measurement in the diagnosis of endometrial cancer in women with obesity or diabetes is decreased compared with nonobese and nondiabetic patients. In women with hypertension, the diagnostic performance of transvaginal endometrial thickness measurements was not affected.
The guideline of the Dutch Society of Obstetrics and Gynecology does, among many other guidelines, recommend that endometrium sampling is not indicated if transvaginal ultrasonography shows a double layer less than 5 mm.12 Therefore, histology was obtained when the endometrial thickness exceeded 4 mm. This might have led to verification bias, which occurs when verification of the diagnosis depends on the test under study. Information on the subsequent development of a malignancy in the women with reassuring results at first diagnoses was not obtained unless they had recurrent bleeding. Therefore, our study design may have underestimated the presence of endometrial cancer in women with an endometrial thickness of 4 mm or less. It is important to realize that further assessment of the endometrium was only dependent on the findings at ultrasonography and not on other risk indicators assessed in the present study, such as obesity, diabetes, or hypertension, thus limiting the impact of verification bias on other findings.
In the literature, the accuracy of a diagnostic test is commonly reported in terms of sensitivity, specificity, and likelihood ratios. When such parameters are used, the crucial underlying assumption is that these indices remain constant for patients with different clinical characteristics.15,16 A diagnostic test should decrease the posttest risk of the presence of endometrial cancer to a level of approximately 5%, not only when used in women with a low pretest change for atypical hyperplasia or a malignancy of the endometrium, but also in women with a high pretest change. In women with a negative test (eg, in those with endometrial thickness under a certain cutoff point), further invasive diagnostic procedures can be omitted. In the present study, receiver operating characteristic analysis showed that particular patient characteristics, that is, the presence of obesity and diabetes, decreased the accuracy of transvaginal endometrial thickness measurement in detecting endometrial cancer.
Two factors are important in understanding the decreased value of transvaginal ultrasonography in women with diabetes and obese women. First, our study confirms previous reports that the incidence of malignancy is higher in women with postmenopausal vaginal bleeding and obesity (18%) or diabetes (21%), compared with women without one of these risk factors (8.0%).17–19 In obese women with diabetes, the incidence was as high as 29%. Second, this study shows that in the absence of malignancy, symptomatic women with obesity and/or diabetes have thicker endometria than women without these risk factors. In women diagnosed with a malignancy, endometrial thickness did not differ between patients with or without risk factors. Thus, whereas the pretest probability for malignancy was higher, the potential of the test to reduce the posttest probabilities to less than 5% was very limited.
Previous reports on this topic are scarce. In a sample of 559 asymptomatic postmenopausal women with (33%) or without HRT, the current use of HRT was the most important factor associated with endometrial thickness.7 Others found increased endometrial thickness in asymptomatic obese postmenopausal women.8,10 Our findings are consistent with those of van der Bosch et al,19 who reported a significant positive correlation between both weight (0.24, P < .01) and BMI (0.26, P < .01) and the endometrial thickness in postmenopausal women with vaginal bleeding or endometrial cells on cervical cytology smear. From our study results, it is not clear that diabetes and obesity are independent factors that affect the diagnostic accuracy of transvaginal ultrasonography, and a synergistic effect cannot be excluded. We found a clear decrease in the accuracy of transvaginal ultrasonography in obese women with diabetes compared with obese women without diabetes, for a strong increase in the incidence of cancer, thus indicating an independent effect. However, because of the relatively small number of patients with combined risk factors in our cohort, definite conclusions on this topic cannot be drawn.
We found no relation between endometrial thickness and hypertension in both asymptomatic and symptomatic women. The relation between endometrial thickness and hypertension has been examined in asymptomatic women.9,10,20,21 After correction for weight, Serdar Serin et al,10 found no relation between hypertension and endometrial thickness. Pardo et al21 showed that, in women with an endometrial thickness exceeding 7 mm, endometrial atrophy was present in 84% of the patients on nifedipine, compared with 41% of women not on antihypertensive drugs. They stated that a drug effect on the endometrium caused a false-positive test in women on nifedipine comparable with the phenomenon described for tamoxifen.
In conclusion, it is debatable whether transvaginal measurement of endometrial thickness is of use in all postmenopausal women with vaginal bleeding. In obese women and in women with diabetes, it might be preferable to perform endometrial sampling irrespective of the findings at transvaginal ultrasonography.
1. Smith-Bindman R, Kerlikowske K, Feldstein VA, Subak L, Scheidler J, Segal M, et al. Endovaginal ultrasound to exclude endometrial cancer and other endometrial abnormalities. JAMA 1998;280:1510–7.
2. Runowicz CD. Can radiological procedures replace histologic examination in the evaluation of abnormal vaginal bleeding? Obstet Gynecol 2002;99:529–30.
3. Gull B, Karlsson B, Milsom I, Granberg S. Can ultrasound replace dilatation and curettage? A longitudinal evaluation of postmenopausal bleeding and transvaginal sonographic measurement of the endometrium as predictors of endometrial cancer. Am J Obstet Gynecol 2003;188:401–8.
4. Goldstein RB, Bree RL, Benson CB, Benacerraf BR, Bloss JD, Carlos R, et al. Evaluation of the woman with postmenopausal bleeding: Society of Radiologists in Ultrasound-Sponsored Consensus Conference statement. J Ultrasound Med 2001;20:1025–36.
5. Dijkhuizen FP, Mol BW, Brölmann HA, Heintz AP. Cost-effectiveness of the use of transvaginal ultrasonography in the evaluation of postmenopausal bleeding. Maturitas 2003;45:275–82.
6. Tabor A, Watt HC, Wald NJ. Endometrial thickness as a test for endometrial cancer in women with postmenopausal vaginal bleeding. Obstet Gynecol 2002;99:663–70.
7. Gull B, Karlsson B, Milsom I, Granberg S. Factors associated with endometrial thickness and uterine size in a random sample of postmenopausal women. Am J Obstet Gynecol 2001;185:386–91.
8. Andolf E, Dahlander K, Aspenberg P. Ultrasonic thickness of the endometrium correlated to body weight in asymptomatic postmenopausal women. Obstet Gynecol 1993;82:936–40.
9. Alcazar JL. Endometrial sonographic findings in asymptomatic, hypertensive postmenopausal women. J Clin Ultrasound 2000;28:175–8.
10. Serdar Serin I, Ozçelik B, Basbug M, Ozsahin O, Yilmazsoy A, Erez R. Effects of hypertension and obesity on endometrial thickness. Eur J Obstet Gynecol Reprod Biol 2003;109:72–5.
11. Weiderpass E, Persson I, Adami HO, Magnusson C, Lindgren A, Baron JA. Body size in different periods of life, diabetes mellitus, hypertension, and risk of postmenopausal endometrial cancer (Sweden). Cancer Causes Control 2000;11:185–92.
12. Diagnostiek bij abnormaal vaginaal bloedverlies in de postmenopauze. Available at: http://www.nvog.nl/files/rl04_abnormal_bloedverlies_postmenopauze.pdf
. Retrieved July 20, 2004.
13. Gibbons JD, Chakraborti S. Nonparametric statistical inference. 3rd ed. New York (NY): Marcel Dekker Inc; 1992.
14. Sackett DL, Haynes RB, Guyatt GH, Tugwell P, editors. Clinical epidemiology: a basic science for clinical medicine. Boston (MA): Little, Brown; 1991.
15. Feinstein AR. Clinical epidemiology: the architecture of clinical research. Philadelphia (PA): Saunders; 1985.
16. Kraemer HC. Evaluating medical tests: objective and quantitative guidelines. Newbury Park (CA): SAGE Publications; 1992.
17. Tornberg SA, Carstensen JM. Relationship between Quetelet's index and cancer of breast and female genital tract in 47,000 women followed for 25 years. Br J Cancer 1994;69:358–61.
18. La Vecchia C, Parazzini F, Negri E, Fasoli M, Gentile A, Franceschi S. Anthropometric indicators of endometrial cancer risk. Eur J Cancer 1991;27:487–90.
19. Van den Bosch T, Vandendael A, Van Schoubroeck D, Lombard CJ, Wranz PA. Age, weight, body mass index and endometrial thickness in postmenopausal women. Acta Obstet Gynecol Scand 1996;75:181–2.
20. Bornstein J, Auslender R, Goldstein S, Kohan R, Stolar Z, Abramovici H. Increased endometrial thickness in women with hypertension. Am J Obstet Gynecol 2000;183:583–7.
21. Pardo J, Aschkenazi S, Kaplan B, Orvieto R, Nitke S, Ben-Refael Z. Abnormal sonographic endometrial findings in asymptomatic postmenopausal women: possible role of antihypertensive drugs. Menopause 1998;5:223–5.
DUPOMEB (Dutch Study in Postmenopausal Bleeding)
Other DUPOMEB members are Maurice V. A. M. Kroeks, MD, PhD, Diakonessenhuis, Utrecht, The Netherlands; Peter H. M. van de Weijer, MD, PhD, Gelre Hospital, Apeldoorn, the Netherlands; Aad A. F. Planken, MD, PhD, Mesos Medical Center, Utrecht, The Netherlands; and M. Jitze Duk, MD, PhD, Meander Medical Center, Amersfoort, The Netherlands.