Endometrial Thickness as a Test for Endometrial Cancer in Women With Postmenopausal Vaginal Bleeding
Abstract & Commentary
Synopsis: Measurement of endometrial thickness by transvaginal ultrasonography is not sufficient as a single procedure for determining the presence or absence of endometrial cancer in symptomatic postmenopausal women.
Source: Tabor A, et al. Obstet Gynecol. 2002;99(4): 663-670.
Tabor and colleagues searched the english medical literature for the period January 1991 to September 30, 1997, for articles that discuss the use of vaginal ultrasonography for measurement of the thickness of the endometrium. They identified 48 potential studies, but eventually eliminated 39 of them for 1 or more reasons. The most common reason for exclusion was the inability of the researchers of the published articles to provide data on the median thickness of the endometrium in women in their study who were not affected by endometrial cancer.
The 9 studies that were selected for inclusion in this meta-analysis contained 3483 symptomatic women without endometrial cancer and 330 with cancer. The women were further subdivided into those who were premenopausal, postmenopausal not on HRT, and postmenopausal on HRT. In 6 of the studies, dilation and curettage (D&C) followed vaginal ultrasonography in all cases. In 2 studies, sonography was followed by endometrial biopsy, and 1 study included both a D&C and an endometrial biopsy.
The first (and perhaps most important) finding of the study was that the median thickness of the endometrium among women who did not have endometrial cancer varied significantly from 2.0 mm to 6.4 mm (P < .001). Two articles contained information about the median thickness of the endometrium among premenopausal women. In these women, the median thickness was 2-3 mm more than symptomatic postmenopausal women, not on HRT. Likewise, among those postmenopausal women who were on HRT (3 studies), the thickness was 2 mm more than postmenopausal women not on HRT.
Because of the differences in the medians, Tabor et al also transformed the data into multiples of the median (MoM). This adjustment did not smooth out the variability among the studies. The MoM with endometrial cancer ranged from 2.1 to 5.9. The average was 3.7 MoM.
Tabor et al then looked at false-positives rates for referral for histologic sampling of 50% and 10%. Overall, if a 10% false-positive rate was chosen, 63% of the women with cancer would have been sampled. At a 50% false-positive rate (no better than a coin-toss), 96% of cancers would have been identified, but 4% would have been missed.
Tabor et al have 2 major conclusions. First, it is critical that each center that performs transvaginal ultrasonography for determination of endometrial thickness develop their institution-specific median endometrial thickness for symptomatic women who are premenopause, postmenopause not on HRT, and postmenopause on HRT. Second, they conclude that vaginal ultrasonography alone is not sufficiently sensitive to rely on it as the only test for endometrial cancer in symptomatic women because of the 4% "false-negative rate," even with liberal guidelines for referral for D&C.
Comment by Kenneth L. Noller, MD
During the past 30 years, it has been interesting to watch the introduction of new technology into the practice of gynecology. Typically, there will be some skepticism regarding any new technique or test when it is first introduced. Indeed, more than half of all techniques that are introduced never "catch on" and are either disconnected or relegated to use in a few special situations. For those techniques that do happen to catch on, the initial skepticism is often followed by a wave of over-use and over-reliance. Eventually, after the publication of multiple articles, the new technique or test will eventually find its unique place in the day-to-day practice of gynecology.
For some years now, I have felt that we have been experiencing the over-use phase of vaginal ultrasonography for evaluation of postmenopausal women with vaginal bleeding. When it was first introduced, multiple small series applauded the technique as a major advance that nearly eliminated the need for the more painful procedure of endometrial biopsy. Virtually all of these early articles concluded that all one had to do was to perform a vaginal probe ultrasound examination, and if the endometrial lining was less than "x" in thickness, no further evaluation was necessary. There was, of course, an immediate indication that such a simple management scheme was certainly not appropriate since each article used a different "x." Nonetheless, I frequently see menopausal women who have experienced vaginal bleeding and have had no evaluation other than a vaginal ultrasound examination. An unusually large proportion of these women seems to be from specialists who do not usually perform endometrial biopsy and/or D&C.
This article by Tabor et al demonstrates clearly that it is impossible to detect every case of endometrial cancer by use of vaginal sonography. That should come as no surprise since virtually any test will fail occasionally. However, they found that unless at least half of all women who have endometrial thickness scanning are referred for D&C, the failure-to-detect rate is very high. In addition, they found (and I personally believe this is the most important information in the article) that there is no such thing as an "average" endometrial thickness above or below which it is possible to recommend referral for sampling. Rather, each institution must develop its own normal/abnormal thickness. Of course, almost no institution has done this, but rather relies on some isolated article that provided such information. That this is a mistake is clearly shown by Tabor et al. Any article an institution chooses is inappropriate for its population, as scans are read differently at the institution where the article was written and the institution that is using the printed material.
The bottom line is that it is not entirely safe to rely on endometrial thickness for evaluation of postmenopausal bleeding. Of course, we can all recite cases where it is absolutely critical to do so. Occasionally, a woman will be in such poor medical health that even an endometrial biopsy is inadvisable, or perhaps the cervix is entirely scarred and it is not possible to sample the endometrial lining. In such cases, if the lining is extremely thin the clinician can be somewhat reassured. Nonetheless, whenever it is possible to obtain a histologic sample from a postmenopausal woman who is bleeding, that is still the standard of care.
Dr. Noller is Professor and Chairman, Department of OB/GYN, Tufts University School of Medicine, Boston, Massachusetts.