SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. References

Objectives  To compare transvaginal ultrasound measurements of endometrial thickness with direct anatomical measurements and consider the implications of these findings on clinical practice.

Design  Prospective observational study using two modalities blinded to each other's findings.

Setting  Singleton Hospital, Swansea, a medium-sized District General Hospital.

Sample  Forty-seven women admitted for hysterectomy.

Methods  All women underwent transvaginal ultrasound scan to measure the endometrial thickness within 16 hours of surgery. Anatomical measurement of the fresh specimen was carried out immediately after surgery.

Main outcome measures  Agreement between ultrasound and anatomical measurements of the endometrial thickness.

Results  No ultrasound measurement was possible in 15% of patients. When both values were obtained, transvaginal ultrasound measurements were >2 mm different from the ruler measurement in 13/40 (33%) with an obvious tendency of the ultrasound scan to over-estimate the endometrial thickness. The mean difference between the two measurements was −0.8 mm (limits of agreement −7.1 to +5.5 mm). The discrepancy was greater in women with endometrial thickness ≤5 mm (−1.6 mm, limits of agreement −5.7 to +2.6 mm) compared with that in women with endometrial thickness >5 mm (−0.2 mm, limits of agreement −7.6 to +7.2 mm). Kappa statistics showed good agreement between the two measurements in discriminating between thin and thick endometrium in 77% (κ= 0.55). Transvaginal ultrasound misdiagnosed a thick endometrium as thin in 3/40 (8%) and misdiagnosed a thin endometrium as thick in 6/40 (15%).

Conclusions  Transvaginal ultrasonography is of limited value as a screening test for abnormal endometrium in patients with postmenopausal bleeding if the only parameter of normality is an endometrial thickness of 5 mm or less.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. References

Transvaginal ultrasound has been widely used for many years for the measurement of the endometrial thickness in patients with postmenopausal bleeding, and in women undergoing ovulation induction or assisted reproduction treatment. Nevertheless, there have been only very few well-documented studies examining the reliability and validity of the transvaginal ultrasound measurement of endometrial thickness.1–3 Two studies have shown that endometrial thickness measured by ultrasound correlated well with values obtained by histopathological measurement,4,5 but both these studies used transabdominal ultrasound, which is rarely used in current gynaecological practice. Other studies examining the accuracy of transvaginal ultrasound measurement of the endometrium thickness have compared transvaginal ultrasound with the hysteroscopic assessment of the endometrium rather than with histopathological measurement.6

The purpose of this study was to compare pre-operative transvaginal ultrasound measurements of the endometrial thickness with the anatomical measurement of the endometrial thickness in hysterectomy specimens, and then to consider the implications of these findings on clinical practice.

METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. References

Forty-seven women undergoing abdominal or vaginal hysterectomy were prospectively included in the study during a 12-month period. The indications for hysterectomy were dysfunctional bleeding, cervical intraepithelial neoplasia or prolapse. Patients included in this study were aged between 33 and 70 years. Eight women were postmenopausal and 39 were premenopausal at variable stages of their cycles at the time of hysterectomy. Women on any form of hormonal treatment (e.g. hormone replacement therapy or contraceptive pill) and women with a clinically enlarged uterus or a known gynaecological malignancy were excluded.

The Local Research Ethics Committee approved this study and each participating woman gave informed consent.

Each woman underwent transvaginal ultrasonography no more than 16 hours before her hysterectomy. Three experienced ultrasonographers (JAB, CH, DW) carried out the transvaginal ultrasound scanning using a transvaginal transducer of variable frequency (7.5, 6.5 or 5 MHz) on either a Siemens Sonoline Versa Pro (manufactured by Matsushita Communications Industrial, Japan, for Siemens Medical Solutions Systems, Issaquah, Washington, USA), or Aloka 1700 (Aloka, Tokyo, Japan) ultrasound machine. Endometrial thickness was measured at the thickest part of the endometrium in the longitudinal plane (approximately 1 cm from the endometrial–myometrial interface at the fundus) and included both endometrial layers; from the base of the hyper-echoic (bright echo) layer of the posterior endometrium to the base of the hyper-echoic layer on the anterior endometrium as described by Granberg et al.7(Fig. 1). Only one operator scanned each patient but intra- and inter-operator testing is part of the quality assurance of the ultrasound department and no one sonographer produced measurements which consistently over- or under-estimated the endometrial thickness when compared with the anatomical measurements.

image

Figure 1. Endometrial thickness measured by transvaginal ultrasound.

Download figure to PowerPoint

Immediately after hysterectomy, the fresh, unfixed uterus was sent to the histopathologist (SW), who was blinded to the transvaginal ultrasound measurement of the endometrial thickness. The uterus was placed on a clean dissecting board. A metal probe was gently placed into the uterine cavity and a freshly sharpened 8-in. surgipath knife was used to slice open the uterus in the sagittal plane. Using a Luxo magnifying glass with strip lighting, the endometrial thickness was measured with a standard ruler. The combined thickness of the endometrium of the anterior and posterior walls was measured in millimetres to give the equivalent of the transvaginal ultrasound measurement.

Agreement between the two measurements of endometrial thickness was assessed using ‘limits of agreement’ method described by Bland and Altman.8 For limits of agreement, a plot of the difference between the histological and the ultrasound measurements of the endometrial thickness against the mean difference between the two measurements was used to provide a visual assessment of the agreement between the two sets of measurements. The mean difference between the histological and the ultrasound measurements is an estimate of the systematic error or bias introduced by the ultrasound measurement, and the 95% limits of agreement provides a measure of the random error.8 Both the mean difference and limits of agreement are plotted on the same graph.

The above analysis was applied separately for uteri with thin (≤5 mm) and thick (>5 mm) endometrium to determine the effect of the varying endometrial thickness on the amount of ‘error’ in transvaginal ultrasound measurement.

Using an endometrial thickness of 5 mm as the maximum value for the normal thickness, the values were defined categorically (thin and thick endometrium), allowing the use of kappa (κ) statistics to measure the level of agreement between the two measurements.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. References

Among the 47 women included in the study, ultrasound measurement of the endometrial thickness was possible in 40 women: 8 postmenopausal and 32 premenopausal. The results of the ultrasound and histological measurements of the endometrial thickness in these women are shown in Table 1. In the remaining seven patients, the endometrial thickness could not be measured satisfactorily by transvaginal ultrasound. Their details are shown in Table 2 and it can be seen that in five of the seven cases, the transvaginal ultrasound did identify the correct pathology even though the endometrial thickness could not be measured.

Table 1.  Ultrasound and anatomical measurements of endometrial thickness in 40 women undergoing hysterectomy.
S/NEndometrial thickness measurements (mm)
Transvaginal ultrasoundAnatomical
112.010.0
23.03.5
37.08.0
44.04.5
54.06.0
613.08.0
76.04.0
82.05.5
916.015.0
107.07.0
1116.09.0
127.07.0
136.02.0
141.73.0
159.07.0
163.03.0
172.01.0
181.31.0
192.14.0
203.02.0
217.74.0
223.48.0
2313.011.0
2412.013.0
2513.010.0
264.02.0
2714.013.0
283.02.0
297.02.5
306.016.0
315.02.0
325.75.0
332.51.0
3416.910.0
357.88.0
362.02.0
3711.04.0
387.06.0
3910.014.0
4012.013.0
Table 2.  Cases where the endometrial thickness could not be measured satisfactorily by transvaginal ultrasound scan. ET = endometrial thickness.
Transvaginal ultrasoundAnatomical
PolypPolyp (ET = 6 mm)
PolypPolyp (ET = 3.5 mm)
FibroidFibroid (ET = 4 mm)
Endometrium not identifiedET = 2 mm
PolypPolyp (ET = 2.3 mm)
Endometrium not identifiedET = 2 mm (difficult specimen)
FibroidFibroid (ET = 3.5 mm)

The results show that in 13 cases (33%), the transvaginal ultrasound measurements were >2 mm different from the ruler measurement. Transvaginal ultrasonography relatively under-estimated the measurement of the endometrial thickness in 33% and over-estimated it in 57% of cases.

Table 3 demonstrates the mean endometrial thickness as measured by transvaginal ultrasound, the mean discrepancy between the true histological and the transvaginal ultrasound measurement of endometrial thickness and the limits of agreement for the whole study group as well as for the thin (≤5 mm) and thick (>5 mm) endometrial thickness subgroups. The results show that transvaginal ultrasound tends to over-estimate the endometrial thickness by 0.8 mm in the whole study group. In women with endometrial thickness ≤5 mm, the estimate was 1.6 mm compared with 0.2 mm in women with endometrial thickness >5 mm.

Table 3.  The mean endometrial thickness (SD), mean difference between the histological and ultrasound measurements of the endometrial thickness (SD) and limits of agreement (2SD) for the whole study group, and the thin (≤5 mm) and thick (>5 mm) endometrial thickness subgroups. ET = endometrial thickness.
 Mean ET by transvaginal ultrasound mm (SD)Mean histological ET − ultrasound ET mm (SD)Limits of agreement (mm)
Whole study group (n= 40)7.2 (4.6)−0.78 (3.22)−7.1 to 5.5
ET ≤5 mm (n= 19)4.2 (2.5)−1.6 (2.3)−5.7 to 2.6
ET >5 mm (n= 21)9.9 (4.3)− 0.2 (3.8)−7.6 to 7.2

Figures 2–4 demonstrate the limits of agreement between the ultrasound and the histological measurement of the endometrial thickness in whole study group (Fig. 2), in women with endometrial thickness ≤5 mm (Fig. 3) and in women with endometrial thickness >5 mm (Fig. 4). The overall agreement between the two measurements is low as evidenced by the wide limits of agreement (−7.1 to +5.5 mm) of the whole study group. The agreement between the two measurements is higher in the subgroup of women with thin endometrium (limits of agreement −5.7 to +2.6 mm) compared with that of women with thick endometrium (limits of agreement −7.6 to +7.2 mm).

image

Figure 2. Limits of agreement plot of the difference between the histological thickness and the ultrasound thickness of the endometrium against the mean of histological thickness in the whole study group (n= 40).

Download figure to PowerPoint

image

Figure 3. Limits of agreement plot of the difference between the histological thickness and the ultrasound thickness of the endometrium against the mean of histological thickness in women with histological endometrial thickness ≤5 mm (n= 19).

Download figure to PowerPoint

image

Figure 4. Limits of agreement plot of the difference between the histological thickness and the ultrasound thickness of the endometrium against the mean of histological thickness in the whole study group (n= 40) in women with histological endometrial thickness >5 mm (n= 21).

Download figure to PowerPoint

Table 4 compares the numbers of ‘thin’ and ‘thick’ values estimated by each method using a cutoff at 5 mm. Analysis of the agreement between the two measurements in determining thin and thick endometrium, using kappa statistics, showed an agreement in 31/40 (77%) cases (κ= 0.55). In the remaining 9/40 (23%) cases, there were disagreements between the transvaginal ultrasound and the histological measurements: 3/40 (8%) misdiagnosed a thick endometrium as thin and 6/40 (15%) misdiagnosed a thin endometrium as thick. In other words, of the 16 women with ultrasonographic endometrial thickness of ≤5 mm, 3 (19%) were found to have a histological thickness >5 mm. On the other hand, 6 of the 24 (25%) women with ultrasonographic endometrial thickness >5 mm were found to have histological endometrial thickness of ≤5 mm.

Table 4.  Numbers of ‘thin’ and ‘thick’ endometrial measurements by each method, using a cutoff of 5 mm. Kappa statistics were used to test the agreement between the two measurements in determining the normal or thickened endometrium.
 Anatomical measurementκ
Thin (≤5 mm)Thick (>5 mm)Total
TVS measurement
Thin (≤5 mm)133160.55
Thick (≥5 mm)61824
Total192140

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. References

In this paper, we have sought to evaluate the clinical validity of transvaginal ultrasound measurement of the endometrial thickness by correlating the transvaginal ultrasound measurements with the values obtained by histopathological measurement in unfixed uteri immediately after hysterectomy. To the best of our knowledge, this is the first study to compare ultrasound and anatomical measurements made in this way.

Where both measurements could be made our study shows a good agreement (≤2 mm) between the transvaginal ultrasound and the anatomical measurements in about two-third of cases (26/40) and very good agreement (≤1 mm) in 18/40 (45%) of cases. However, in 35% of cases, a substantial disagreement (3–10 mm) was observed with a tendency of transvaginal ultrasound to over-estimate. The discrepancy between the two measurements was greater in the group of women with thin endometrium (1.6 mm) compared with that of women with thick endometrium (0.2 mm) (Table 3). However, the agreement between the two measurements is higher in the group of women with thin endometrium (limits of agreement −5.7 to 2.6) compared with that in women with thick endometrium (−7.6 to 7.2). In other words, although the error of the transvaginal ultrasound measurement seems to be greater when the endometrial thickness is thin, the agreement appears to be better in this group of women than those with thick endometrium.

Variations of the transvaginal ultrasound measurements are not surprising as the transvaginal ultrasound endpoints are often relatively soft and may not correspond exactly with the anatomical planes. This would not, however, explain the predominant over-estimation by transvaginal ultrasound. A thin layer of fluid in the uterus at the time of the transvaginal ultrasound might be the explanation, as might loss of turgor in the exsanguinated specimen. The over-estimation was more marked in cases with a thinner endometrium, perhaps due to the difficulty in obtaining an accurate measurement in the very thin endometrium with a relatively large cursor. There is a natural tendency to measure to the outer extreme and a slight movement of the cursor will make a significant difference to the measurement.

Another possible factor that could have contributed to the variation of the transvaginal ultrasound measurements of the endometrial thickness is the inter-observer variability or using two different US machines. This is unlikely to be valid, as previous studies have confirmed the inter-observer reliability of transvaginal ultrasound measurements of endometrial thickness. Furthermore, as mentioned above, the inter-observer variability would not explain the tendency of transvaginal ultrasound to over-estimate the endometrial thickness. Another possible explanation of the relative inaccuracy of the ultrasound measurement of the endometrial thickness is the variation of the angle of insonation of the endometrial echo. However, as stated above, this factor has been overcome as far as possible by taking the endometrial thickness measurements where the endometrial echo was perpendicular to the ultrasound beam.

Our results are in disagreement with the earlier report by Fleischer et al.4 who showed good correlation between the ultrasound and the histopathological measurements, reporting that the endometrial thickness was accurately assessed by sonography (within 1 mm of the histopathological measurement) in 87% (33/38) of the patients. In that study, the time interval between the pre-operative scan and the post-operative measurement was not clearly documented but this is unlikely to have improved the correlation. The results described by Saxton et al.5 are more like ours in that no ultrasound measurement could be made in 17% and in the others the mean difference between the two measurements was small but the range was large (mean 0.63 mm, 95% range −3.3 to +4.5 mm). Very few of the women described by Saxton et al.5 had an ultrasound estimate of endometrial thickness of 5 mm or less and the suggestion following that study was only that ultrasound measurements could usefully complement vaginal examination. Both of these two studies, however, used transabdominal ultrasound scans, which have relatively poor resolution compared with the transvaginal ultrasound from which one would expect a more accurate measurement. We cannot confidently explain the discrepancy between our results and those of Fleischer et al.4 but we are confident that our methodology was sound and reflects the clinical use of transvaginal ultrasound.

In our study, no measurement was possible in 15% of subjects; in the others, using a maximum value of 5 mm to discriminate between thin and thick endometrium, transvaginal ultrasound showed a reasonable agreement with histological measurement in 77% (κ= 0.55) with a tendency to over-estimate in 15% and to under-estimate in 8%(Table 4). In 19% of women with ultrasonographically thin endometrium, the endometrium was found to be histologically thick and in 25% of those with thick endometrium on transvaginal ultrasound the endometrium was found to be histologically thin. This means that the chance of missing a thick endometrium is about 8% of all the women scanned and 19% of women with ultrasonographically thin endometrium. On the other hand, over-estimating the endometrial thickness in 15% of all women undergoing transvaginal ultrasound and in 25% of those with ultrasonographically thick endometrium may result in unnecessary intervention (e.g. hysteroscopy), in this proportion of women.

Several previous studies have suggested that women with postmenopausal bleeding and endometrial thickness of <5 mm need no further investigations.7,9–18 In a large multicentre study, Ferrazzi et al.19 examined the endometrial thickness in 930 women with postmenopausal bleeding. They reported a negative predictive value of 99% in the detection of malignancy when an endometrial thickness of 4 mm was used as the upper limit of normal. A recent meta-analysis of 35 studies examining the transvaginal ultrasound measurements of endometrial thickness showed that the sensitivity of detecting endometrial abnormalities was 92% and for detecting cancer was 96% when an endometrial thickness of 5 mm was used as the upper limit of normal.20 However, several studies have reported that endometrial cancer occurred in women with endometrial thickness of <5 mm as measured by transvaginal ultrasound.21–23 This may be explained by the finding in our study that there is an 8% chance of missing a thick endometrium when transvaginal ultrasound is used to measure the endometrial thickness.

CONCLUSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. References

This study demonstrated some of the limitations of transvaginal ultrasound as a screening test for abnormal endometrium in patients with postmenopausal bleeding (if they were not postmenopausal then 5 mm would not be appropriate) if the only parameter used is maximum thickness of 5 mm. We therefore recommend that transvaginal ultrasound measurement should be interpreted with caution and patients with persistent postmenopausal bleeding and a thin endometrium on transvaginal ultrasound should undergo a hysteroscopic examination.

References

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. METHODS
  5. RESULTS
  6. DISCUSSION
  7. CONCLUSION
  8. References

Accepted 16 January 2004