Controlled analysis of factors associated with insufficient sample on outpatient endometrial biopsy
Correspondence: Dr K. S. Khan, Academic Department of Obstetrics and Gynaecology, Birmingham Women's Health Care NHS Trust, Birmingham B15 2TG, UK.
We examined the relative significance of hysteroscopic and ultrasonographic evidence of endometrial atrophy in relation to insufficient sample on outpatient endometrial biopsy in women with abnormal uterine bleeding. Multivariate logistic regression modelling was used to evaluate the independent effects of age, menopausal status, hysteroscopic findings and sonographic endometrial thickness on outpatient endometrial sampling (sufficient or insufficient) used as the binary dependent variable. Insufficient sample on endometrial biopsy was associated with hysteroscopic finding of endometrial atrophy (OR 4.79, 95% CI 1.05–21.91, P= 0.04) and sonographic endometrial thickness below 5 mm (OR 0.19, 95% CI 0.07–0.53, P= 0.001). There was no association with patient's age and menopausal status. In conclusion, when reassuring women with insufficient sample on outpatient endometrial biopsy, one can be confident about absence of pathology provided the hysteroscopic and sonographic endometrial assessment is consistent with endometrial atrophy.
In ambulatory investigation of abnormal uterine bleeding, insufficient sample is obtained from outpatient endometrial biopsy in 25–33% cases1,2. Generally, an insufficient sample is not believed to be a cause for concern, as failure to obtain an endometrial specimen from a device correctly positioned within the uterine cavity is considered to be an assurance that no significant intrauterine pathology is present3. Do we have the confidence that insufficient sample is benign in women with abnormal uterine bleeding? One study supporting this view4 showed that in 42% of cases with hysteroscopically atrophic endometrium no tissue was obtained on traditional dilatation and curettage for histological assessment. Such data provides a plausible, but not scientific, evidence of association because of lack of a comparison with a hysteroscopically normal control group. Moreover, such data also suffer from lack of an adjustment for confounding variables in the analysis like age, menopausal status and ultrasonographic endometrial thickness. Apart from this methodological debate, information obtained from inpatient dilatation and curettage can hardly be considered generalisable to outpatient biopsy devices.
To have the confidence that insufficient sample on outpatient biopsy is benign in women with abnormal uterine bleeding, we decided to evaluate the significance of hysteroscopic and ultrasonographic evidence of endometrial atrophy in relation to outpatient endometrial sampling. We conducted a controlled statistical analysis to delineate the relative significance of factors associated with insufficient sample on outpatient endometrial biopsy.
We used a database of 248 consecutive women with abnormal uterine bleeding investigated in our one-stop clinic from November 1996 to December 19975. The clinic was held in a large teaching hospital and women were referred to the service from gynaecology outpatient clinics or directly from general practice for investigation of abnormal uterine bleeding (e.g. menorrhagia, irregular menstrual loss, postmenopausal bleeding and unscheduled bleeding on tamoxifen or hormone replacement therapy). The characteristics of the women attending the one-stop clinic are shown in Table 1.
Table 1. Characteristics of women attending outpatient one stop hysteroscopy clinic. Values are given as mean (95%CI) or n (%) [95% CI]. HRT = hormone replacement therapy
|Age (years)||50.7 (0.26–0.38)||50.6 (0.61–0.73)|
|Parity||2.4 (.−26–0.38)||2.4 (0.61–0.73)|
|Post menopause||56 (32.2) [25.3–39.7]||56 (75.6) [64.3–84.9]|
|HRT||44 (25.2) [19.0–32.4]||29 (39.2) [28.0–51.2]|
|Cancer||8 (4.6) [2.0–8.9]||0 (0) [0–4.8]|
In the clinic, a combination of pelvic ultrasound scan, outpatient hysteroscopy, and pipelle endometrial sampling were performed. Pelvic transvaginal ultrasonography was performed using a portable scanner (Hitachi Sumi, Japan) with an endovaginal 6.5 MHz transducer. Endometrial thickness was measured in millimetre including both layers of the endometrium. Outpatient hysteroscopy was performed using a 1.2 mm microhysteroscope with a 2.5 mm outer rigid sheath (Karl Storz, Tuttlingen, Germany). The depth of mucosa was assessed by simple pressure of the top of the endoscope. Hysteroscopic appearances of thin smooth nonvascular endometrium were considered to be atrophic. Outpatient endometrial biopsies were performed using a pipelle sampling device (Laboratoire CCD, Paris, France). Of the 248 women, 74 (29.8%) had insufficient endometrial sample on pathology report. We followed all 74 women using the hospital clinical computer system for any further admission episodes related to abnormal uterine bleeding over a period of two years. We performed inpatient assessment and sampling on those with recurrent symptoms. None of these women had endometrial cancer or hyperplasia.
To delineate the relative significance of various factors (independent variables) on sufficiency of outpatient endometrial biopsy (binary dependent variable), we built a multivariate logistic regression model6. The primary null hypothesis we wanted to test was, ‘the value of the β-coefficient for the various risk factors in the logistic model is zero or its exp(β) is 1.0. To obtain a valid estimate of the association between endometrial atrophy (on hysteroscopy and on ultrasonography) and insufficient endometrial specimen, the analysis was adjusted for the confounding effects of age, menopausal status, parity and use of hormone therapy by including them as independent variables in the multivariate logistic regression analysis. In the model building process, the significance of the independent variables was tested individually and in various combinations. Variables were tested for interaction and they were analysed using both nominal and continuous scales of measurement. The assumption of linearity of association was checked for the different independent variables. Cases with missing data were excluded from the multivariable analysis. The model that best predicted the odds of insufficient sample on endometrial biopsy with the lowest deviance and highest goodness of fit was selected. The Hosmer-Lemeshow test using the ‘deciles of risk’ method was used to assess the goodness of fit.
Univariate regression (Table 2) revealed that atrophic endometrium on hysteroscopy increased the odds of having insufficient sample on endometrial biopsy (OR 5.53, P= 0.001) while endometrial thickness above 5 mm on ultrasonography decreased the odds (OR 0.15, P= 0.001). Increasing patient's age (OR 1.04, P= 0.001) and postmenopausal status (OR 4.49, P= 0.001) were also associated with insufficiency of outpatient sampling in unadjusted analysis. When these variables were combined in a multivariate regression analysis to control for confounding, only endometrial atrophy on hysteroscopy (OR 4.79, P= 0.04) and endometrial thickness < 5 mm on ultrasonography (OR 0.19, P= 0.001) remained significant in predicting an insufficient sample on outpatient endometrial biopsy. In a similar but separate multivariate model using endometrial thickness as a continuous variable, we found that the odds of insufficient sample decreased with each 1 mm increase in endometrial thickness (OR 0.88, 95% CI 0.78–0.99, P= 0.03).
Table 2. Logistic regression analysis to determine the relationship between various factors associated with insufficient specimen on outpatient endometrial biopsy. Values are given as OR (95% CI), unless otherwise indicated.
|Age (years)||1.04 (1.01–1.06)||0.001||1.02 (0.98–1.06)||0.24|
|Menopausal status (pre vs post)||4.49 (2.49–8.09)||0.001||1.92 (0.66–5.61)||0.23|
|Ultrasonographic endometrial thickness (5 mm cut off)||0.15 (0.06–0.37)||0.001||0.19 (0.07–0.53)||0.001|
|Hysteroscopic endometrial thickness (atrophic vs nonatrophic)||5.53 (2.47–11.60)||0.001||4.79 (1.05–21.91)||0.04|
To our knowledge, this type of controlled analysis to delineate the significance of factors associated with insufficient specimen on outpatient endometrial biopsy has not been reported so far. Insufficient sample on endometrial biopsy was only associated with hysteroscopic finding of endometrial atrophy and ultrasonographically thin endometrium. There was no association with patient's age, menopausal status, parity and use of hormone replacement therapy in multivariable analysis (Table 2). Hysteroscopic finding of atrophy is somewhat subjective, whereas measurement of endometrial thickness is objective. However, both of these tests independently influenced the odds of insufficient sample on outpatient biopsy.
There is controversy about the need for additional testing if outpatient endometrial biopsy is insufficient. Some gynaecologists prefer to perform inpatient curet-tage under anaesthesia, while others suggest that further tests may not be necessary3. It would appear from our analysis that we can be confident in reassuring women with insufficient sample on outpatient endometrial biopsy, provided the hysteroscopic and sonographic endometrial assessment is consistent with endometrial atrophy. This means that inpatient dilatation and curet-tage can possibly be avoided in such women. However, the usual precautions of reinvestigation on recurrence of symptoms would still be warranted.