Several studies1–3 suggested that tamoxifen produced estrogenic changes in the endometrium ranging from hyperplasia, polyps, or invasive carcinoma. Sonographic endometrial thickness measurement was proposed as the first line of investigation for those conditions.4 Two studies5,6 indicated that tamoxifen-treated women with endometrial thicknesses of at least 4 mm should have additional diagnostic procedures. Other investigators7,8 reported low accuracy of transvaginal ultrasound in asymptomatic women for identifying abnormal histologies, resulting in many unnecessary invasive procedures.
The purposes of this study were to assess the independent accuracy of transvaginal ultrasound in identifying women at risk for endometrial disorders, after adjusting for known confounding variables, and evaluate whether a cutoff greater than 4 mm better identifies women who need endometrial histologic assessments.
Materials and Methods
The study population consisted of postmenopausal women with breast cancer who were receiving tamoxifen (20 mg/day) for at least 6 months, admitted for gynecologic surveillance to the Obstetrics and Gynecology Department of the University of Insubria between November 1994 and October 1997. Menopause was defined as absence of menses (physiologic or treatment-induced) for at least 6 months. Inclusion criteria were endometrial thickness at ultrasound greater than 4 mm (asymptomatic women), or vaginal bleeding (symptomatic women). Exclusion criteria were previous medical or surgical treatments for endometrial hyperplasia or gynecologic malignant tumors before tamoxifen therapy, vaginal bleeding during the first 6 months of tamoxifen treatment, and endometrial thickness at ultrasound greater than 4 mm before tamoxifen therapy.
All women had gynecologic evaluations including pelvic examinations, Papanicolau smears, and endovaginal sonography of the uterus and ovaries before tamoxifen therapy, then every 12 months after. An additional endovaginal sonography was done if vaginal bleeding occurred. No women received postmenopausal estrogen replacement therapy (ERT).
Endometrial thickness was measured with electronic calipers on midline sagittal scans including anterior-to-posterior endometrium. All ultrasound examinations were done with a Hitachi 790A unit (Hitachi Medical Corp., Tokyo, Japan) with a 5-to-6.5-MHz endovaginal transducer. A hysteroscopy with selective endometrial biopsies was done on all women with a 30°, 4-mm, rigid hysteroscope (Karl Storz, Strombeek-Bever, Belgium) with videomonitoring. When feasible, hysteroscopy was done in an outpatient setting. In women for whom assessment was not possible because of cervical stenosis or discomfort, hysteroscopy was done under general anesthesia. When a polyp was found at diagnostic hysteroscopy, an operative hysteroscopy was scheduled within 1 week. If a woman had more than one evaluation, the last endometrial thickness measured, or that immediately preceding endometrial sampling, was considered for analysis.
Informed consent was obtained from all subjects. An abnormal histopathologic finding was defined as the presence of endometrial hyperplasia, endometrial polyps, adenocarcinoma in situ, or malignant endometrial carcinoma.9 This study was approved by the Ethical Research Committee of our hospital.
Statistical analysis was done with EpiStat 4.0 (EpiStat Services, Richardson, TX) and with SPSS 7.0 (Statistical Package for Social Sciences; SPSS Inc., Chicago, IL). Student t test and Mann-Whitney U test were used for comparison of continuous variables, whereas proportions were compared with χ2 or Fisher exact test. Receiver operating characteristic (ROC) curves were constructed to describe the relationship between sensitivity and false-positive rate for different values of endometrial thickness, and duration of tamoxifen treatment in detection of abnormal histopathologic findings at hysteroscopy. Logistic regression was used to investigate relationships between covariates and abnormal histologic findings. Vaginal bleeding, parity, age at menarche, age at menopause, body mass index (BMI), endometrial thickness, and duration of tamoxifen treatment were used as covariates. All variables were entered into the logistic regression model as dichotomous variables, using the cutoff of parity 0 or more, age at menarche less than 15 years or 15 years or more, age at menopause less than 52 years or 52 years or more, BMI less than 27 kg/m2 or 27 kg/m2 or more.9 Cutoffs for endometrial thickness and duration of tamoxifen therapy were derived from the ROC curves analysis. The regression relationship between log odds of an abnormal histologic finding and the explanatory variables was examined using overall logistic regression and a stepwise covariate selection procedure P < .05 was considered statistically significant.
During the study, 166 women met inclusion criteria. Three refused to participate, leaving 163 cases for analysis. Forty-six women (28.2%) were symptomatic. Patient characteristics according to presence or absence of vaginal bleeding are presented in Table 1. Hysteroscopy was done under general anesthesia in 31 women (15 asymptomatic and 16 symptomatic). The histopathologic findings according to presence or absence of vaginal bleeding are displayed in Table 2. The proportion with abnormal pathologic findings was higher among symptomatic than asymptomatic women (56.5% [26 of 46] versus 16.2% [19 of 117], P < .001). The median (range) endometrial thickness was greater in women with abnormal pathologic findings than in those with normal endometria (13.1 mm [3.6–26] versus 4.5 mm [2.6–16.3], P < .01). Among symptomatic women, endometrial carcinomas developed in two—one serous papillary carcinoma stage IV, grade 3, and one endometrial adenocarcinoma stage Ib, grade 1. The duration of tamoxifen therapy in those two women was 28 and 11 months, respectively.
Receiver operating characteristic curve analysis showed a significant relationship between endometrial thickness and abnormal histopathologic finding at hysteroscopy (area under the curve 0.79, P < .001) (Figure 1). That figure remained significant, considering only asymptomatic women (n = 117) (area under the curve 0.77, P < .001). There was a significant relationship between the duration of tamoxifen therapy and abnormal findings at hysteroscopy (area under the curve 0.67, P <.001) (Figure 2). This relationship remained statistically significant when analysis was restricted to asymptomatic women (n = 117) (area under the curve 0.65, P < 0.01). The cutoff values, derived from the ROC curves, used to calculate the diagnostic indices, were 9 mm for endometrial thickness and 27 months for duration of tamoxifen treatment, respectively. The proportion of women with abnormal histologic findings was higher among those with endometrial thicknesses greater than 9 mm, compared with those with endometrial thicknesses 9 mm or less (60% [39 of 65] versus 6.1% [six of 98]), P < .001]. Among women with abnormal endometrial findings, there was no difference between asymptomatic and symptomatic women in the proportion with endometrial thicknesses greater than 9 mm (84.2% [16 of 19] versus 88.4% [23 of 26], respectively) and in the proportion of women who received tamoxifen for at least 27 months (47.4% [nine of 19] versus 76.9% [20 of 26]). The proportion with abnormal histologic findings was higher among women who received tamoxifen for more than 27 months than those who received tamoxifen for less time (46% [29 of 63] versus 16% [16 of 100], P < .005). Diagnostic indices and positive and negative predictive values for endometrial thicknesses greater than 9 mm and for tamoxifen treatment longer than 27 months are presented in Table 3.
Logistic regression analysis found that when vaginal bleeding, endometrial thickness, parity, age at menarche, age at menopause, BMI, and duration of tamoxifen treatment were entered independently into the model, a relationship with the presence of an abnormal histopathologic finding was present only for vaginal bleeding (χ2 = 41.49, odds ratio [OR] = 11.08, 95% confidence interval [CI] 5.22, 23.51, P < .001), endometrial thickness (χ2 = 31.07, OR = 7.06, 95% CI 3.38, 14.73, P < .001), and duration of tamoxifen treatment (χ2 = 10.30, OR 2.97, 95% CI 1.50, 5.85, P < .005). Using a stepwise covariate selection procedure, logistic regression found that when endometrial thickness was entered last into the model, after vaginal bleeding and the duration of therapy, a significant increment in the χ2 was present (from χ2 = 46.45 to χ2 = 60.92, P < .001). A computed regression analysis found that when vaginal bleeding, endometrial thickness, and duration of tamoxifen treatment were entered in the model, an endometrial thickness greater than 9 mm was associated strongly with abnormal findings at hysteroscopy (Table 4).
A MEDLINE search was done to identify studies on endometrial abnormal findings in women with breast cancer who were receiving tamoxifen therapy, published between 1966 and August 1998. The search terms used, alone or in combination, were tamoxifen, endometrium, vaginal bleeding, endometrial hyperplasia, hysteroscopy, and endometrial cancer. Additional articles used were searched by cross-referencing. The present study is the largest in women with breast cancer receiving tamoxifen in which hysteroscopies with selective endometrial biopsies were done to screen for histologic endometrial abnormalities.
Our results agree with those of Hann et al,10 who reported 58% of histologic abnormalities among tamoxifen-treated women with sonographic endometrial thicknesses greater than 8 mm. Our findings differ from those of Kedar et al,11 who reported a significantly higher positive predictive value in asymptomatic woman than the present study (100% versus 60%, P < .01) for identifying atypical hyperplasia or polyps using endometrial thicknesses of at least 8 mm. This was possibly because of fewer subjects (n = 16) with endometrial thicknesses greater than 8 mm and a different method of endometrial biopsy from ours,11 using a sterile, disposable suction curette. Several studies3,12 showed the highest accuracy detecting endometrial abnormality was with hysteroscopy with selective biopsies; therefore, minimal lesions could be underestimated using blind methods of endometrial biopsy.
Our results agreed with Cheng et al,13 who reported that among premenopausal and postmenopausal women with breast cancer receiving tamoxifen, all in whom endometrial diseases developed were exposed for more than 14 months to tamoxifen, and that all cases with atypical hyperplasia or endometrial carcinoma received tamoxifen for more than 25 months. Correlations between duration of tamoxifen treatment and dysfunctional endometrial changes were reported by other investigators,14,15 as was the increase in severity of lesions with time of exposure to tamoxifen.1,4
Increasing the cutoff value from 4 to 9 mm at which to do additional procedures would result in a 70% reduction of hysteroscopy in asymptomatic women with minimally increased risk of missing an endometrial disease. Using 9 mm as cutoff, the risk of unnecessary invasive procedures with atrophy can be reduced. McGonigle et al16 reported a higher incidence of atrophy and cystic atrophy in tamoxifen-treated women when the sonographic endometrial thickness was between 5 and 8 mm than when it was greater than 8 mm (66.7% [four of six] versus 20.7% [six of 29], P < .05).
Although endometrial lesions in asymptomatic tamoxifen-treated women are benign or premalignant, controversies still exist regarding continuous effect of tamoxifen on endometrial disorders.17 It was proposed that tamoxifen might be involved in regulating endometrial and myometrial expression of mediators of estrogen-stimulated uterine proliferation, and that it can be involved in the regulation of carcinogenic factors. Elkas et al18 reported recently altered expression of insulinlike growth factor-1 and insulinlike growth factor-binding protein-1, proteins with autocrine or paracrine activity, in the endometria of tamoxifen-treated women. Those proteins seem to affect early carcinogenesis.
Alternative diagnostic procedures (ie, sonohysterography) less invasive than hysteroscopy were proposed19 in women with suspicious endometria at ultrasound. Timmerman et al20 recently proposed transvaginal sonography with sonohysterography was at least as accurate as office hysteroscopy for identifying endometrial polyps. The present study and that of other investigators21 showed that hysteroscopy with selective endometrial biopsies could be done in an outpatient setting, with more than 80% of women, with minimal discomfort.
Until the natural history of tamoxifen-induced endometrial lesions is clearer, women with breast cancer receiving tamoxifen who had normal endometria before therapy should have a hysteroscopy, regardless of vaginal bleeding, when sonographic endometrial thickness is greater than 9 mm.
1. Clarke M, Collins R, Davies C, Godwin J, Gray R, Peto R. Tamoxifen for early breast cancer: An overview of the randomized trials. Lancet 1998;351:1451–67.
2. Cohen JC. Tamoxifen and endometrial cancer: Tamoxifen effects on the human female genital tract. Semin Oncol 1997;24:55–64.
3. Ugwumadu AHN, Carmichaels PL, Neven P. Tamoxifen and the femal genital tract. Int J Gynecol Cancer 1998;8:6–15.
4. Berliere M, Charles A, Galant C, Donnez J. Uterine side effects of tamoxifen: A need for systematic pretreatment screening. Obstet Gynecol 1998;91:40–4.
5. Cohen I, Rosen D, Tepper R, Cordoba M, Shapira Y, Altares M, et al. Ultrasonographic evaluation of the endometrium and correlation with endometrial sampling in postmenopausal patients treated with tamoxifen. J Ultrasound Med 1993;5:275–80.
6. Osmer R, Volksen M, Schauer A. Vaginosonography for early detection of endometrial carcinoma. Lancet 1990;335:1569–71.
7. Hulka CA, Hall DA. Endometrial abnormalities associated with tamoxifen therapy for breast cancer. Sonographic and pathologic correlations. Am J Roentgenol 1993;160:809–12.
8. Goldstein SR. Unusual ultrasonographic appearance of the uterus in patients receiving tamoxifen. Am J Obstet Gynecol 1994;170:447–51.
9. DiSaia PJ, Creasman WT. Clinical gynecologic oncology. 5th ed. St. Louis, Missouri: Mosby, 1997.
10. Hann LE, Giess LE, Bach AM, Tao Y, Baum HJ, Barakat RR. Endometrial thickness in tamoxifen-treated patients. Correlation with clinical and pathologic findings. Am J Roentgenol 1997;168:657–81.
11. Kedar RP, Bourne TH, Powles TJ, Collins WP, Ashley SE, Cosgrove DO, et al. Effects of tamoxifen on uterus and ovaries of postmenopausal women in a randomized breast cancer preventional trial. Lancet 1994;343:1318–21.
12. Neven P, De Muylder X, Van Belle Y, Vanderick G, De Muylder E. Hysteroscopic follow-up during tamoxifen treatment. Eur J Obstet Gynecol Reprod Biol 1990;35:235–8.
13. Cheng WF, Lin HH, Torng PL, Huang SC. Comparison of endometrial changes among symptomatic tamoxifen-treated and non-treated premenopausal and postmenopausal breast cancer patients. Gynecol Oncol 1997;66:233–7.
14. Uziely B, Lewin A, Brufman A, Dorembus D, Mor-Yosef S. The effect of tamoxifen on the endometrium. Breast Cancer Res Treat 1993;26:101–5.
15. Cohen I, Altares MM, Shapira J, Tepper R, Rosen DJ, Cordoba M, et al. Time-dependent effect of tamoxifen therapy on endometrial pathology in asymptomatic postmenopausal breast cancer patients. Int J Gynecol Pathol 1996;15:152–7.
16. McGonigle KF, Shaw SL, Vasilev SA, Odom-Maryon T, Roy S, Simpson JF. Abnormalities detected on transvaginal ultrasonography in tamoxifen-treated postmenopausal cancer patients may represent endometrial cystic atrophy. Am J Obstet Gynecol 1998; 178:1145–50.
17. Carlson RW. Overview from a medical oncologist. Semin Oncol 1997:24(Suppl 1):151–7.
18. Elkas J, Gray K, Howard L, Petit N, Pohl J, Armstrong A. The effect of tamoxifen on endometrial insulin-like growth factor-1 expression. Obstet Gynecol 1998;91:45–50.
19. Bourne TH, Lawton F, Leather A, Granberg S, Campbell S, Collins WP. Use of intracavitary saline instillation and transvaginal ultrasonography to detect tamoxifen associated endometrial polyps. Ultrasound Obstet Gynecol 1994;4:73–5.
20. Timmerman D, Deprest J, Bourne T, Van den Berghe I, Collins WP, Vergote I. A randomized trial on the use of ultrasonography or office hysteroscopy for endometrial assessment in postmenopausal patients with breast cancer who were treated with tamoxifen. Am J Obstet Gynecol 1998;179:62–70.
21. Nagele F, O'Connor H, Davies A, Badawy A, Mohamed H, Magos A. 2500 outpatient diagnostic hysteroscopies. Obstet Gynecol 1996; 88:87–92.