Comparison of Endometrial Cytology and Transvaginal Ultrasonography to Find Endometrial Malignancies

Abstract & Commentary

Synopsis: Transvaginal ultrasonography may be useful for the identification of endometrial abnormalities.

Source: Tsuda H, et al. Gynecol Oncol 1997;65: 383-386.

Six hundred postmenopausal women comprise the study cohort for this project that compares endometrial cytology to transvaginal ultrasonography (TVS) for the diagnosis of endometrial abnormalities. Office biopsies of the endometrium were used as the "gold standard" to which the other two modalities were compared. Three hundred seventy-five of the study women were asymptomatic, and 166 had postmenopausal bleeding. Fifty-nine women were dropped from the study because endometrial cytology and histology could not be obtained due to cervical stenosis.

Tsuda and associates subdivided the study women into those less than five years since menopause and those five years or more since menopause. They used a 4 mm cutoff value for women less than five years and a 3 mm cutoff value for those five years or more since menopause as "optimum for detection of endometrial malignancy by TVS."

Endometrial biopsy identified 16 cancers and 22 hyperplasias. Fifteen of the 16 cancers were among the women with postmenopausal bleeding as were 12 of the 22 hyperplasias. Fifteen of the 16 patients with endometrial cancer had a thickened endometrium, and all 22 cases of hyperplasia had thickened endometrium. Cytology diagnosed all 16 cancers but only 14 of the 22 endometrial hyperplasias.


This is not a good paper. However, I chose to review it in this issue of OB/GYN Clinical Alert because it serves as a good companion piece to the other article I reviewed (Ultrasonography-based Triage for Perimenopausal Patients with Abnormal Uterine Bleeding). However, before I comment on the combination of articles, let me point out the flaws in this study.

First, Tsuda et al state that they are investigating techniques for "endometrial cancer screening," yet 31% of their study cohort is comprised of women with postmenopausal bleeding. As I have repeatedly pointed out in other comments, screening techniques are only appropriate for patients without symptoms. Thus, these authors have studied a group of women that is comprised both of those who absolutely needed further evaluation and those for whom a screening technique might be appropriate.

Second, their two tables that examine the relationship between histologic findings and endometrial thickness make no sense whatsoever. The thickness categories are "0-3, 3-4, 4-6, etc." The problem with such division is that it is impossible to know into which category to place a patient. For example, in which category do you place a woman with an endometrial thickness of exactly 4 mm? This is especially important because their definition of an abnormal thickness is either 3 mm or 4 mm (depending on years since menopause), and both numbers are found in two categories.

Third, I have no idea how they obtained their endometrial biopsies. In the description of the study, they mention "office endometrial biopsies." In the following sentence, they state that they are performing "curettage" and occasionally used "minimal dilation." I don’t know whether they used a self-contained piston instrument such as a Pipelle, whether they used an office vacuum such as the Vabra, or whether they actually performed sharp curettage perhaps with some type of sedation. They also used the results of these biopsies as the definition of disease vs. non-disease state. As we know, office biopsies are far from 100% accurate. Indeed, for a prospective comparative study such as this was supposed to be, it is hard to imagine that anything short of hysterectomy could be used for diagnosis. It certainly is not acceptable to calculate sensitivity and specificity from their data (as did the authors).

Fourth, Tsuda et al once again raise the possibility that endometrial cytologic sampling might be useful, despite the many articles showing that it is not sensitive enough to be used for routine screening. This fact is not surprising since the diagnosis of early endometrial adenocarcinoma and the various degrees of hyperplasia is based largely on architecture, not cellular detail. Week after week, I am amazed at how benign the individual cells in a Grade I endometrial adenocarcinoma appear. It is interesting that 23 women in this study had positive cytology with a negative histology. Apparently, Tsuda et al believed that these were cytologic false-positives and did nothing further. Because office endometrial biopsies are far short of 100% sensitive for the detection of endometrial abnormalities, I wonder how many of these 23 women actually had endometrial pathology?

But, there is a bigger issue raised by this study and the other one that I abstracted: What is the best and most cost-effective method to evaluate postmenopausal bleeding? Little by little, TVS is becoming a more appealing technique. We already have largely discarded dilatation and curettage under anesthesia as our method of choice, and we have changed to office endometrial sampling. Unfortunately, office sampling (and, for that matter, D&C as well) is not 100% accurate. Endometrial cytology was discarded years ago. Eventually, TVS might emerge as the winner. Of course, it also will never be 100% accurate, and many women with thickened endometrial stripes will be found to have no disease. Finally, the accuracy of TVS depends to a great degree on the training and experience of the person performing the procedure. It is not always easy to obtain a view from which accurate measurements can be made. The large charges associated with the use of TVS represent another problem with widespread implementation. In addition, patient discomfort is a bigger factor than most authors who write about the technique will admit. In my experience, few women who have had a transvaginal ultrasound have found the examination to be comfortable. If saline infusion is added (as was done in the first article), the examination becomes even more costly, time-consuming, and uncomfortable. Nonetheless, eventually we may use TVS as our first method of evaluation in women with symptoms such as postmenopausal bleeding.