Page 140 - The diagnostic work-up of women with postmenopausal bleeding

Chapter 7
in seven studies.12-16,22,23 In three studies the diagnosis of endometrial cancer detected
by endometrial sampling was confirmed by hysterectomy and not by hysteroscopy or D&C.12,13,21
All 12 studies reported on the fraction of women in whom endometrial sampling failed,mostly due to cervical stenosis.The failure rates of endometrial sampling varied between 1% and 53%, with a weighted failure rate of 11 %. Eight studies reported on the fraction of women in whom insufficient material was found at histology, which varied between 7% and 76%, with a weighted insufficient rate of 31 %.14,16-21,23 In the article by Batool et al the rate of insufficient samples was much higher than in the other studies (42/55). In 37 of these women with an insufficient sample, material was also insufficient for diagnosis by D&C, which might explain the high insuffient-rate.20
The weighted percentage of women with endometrial (pre) cancer among those who had failed or insufficient sampling is 7% (range 0%-18% in seven studies). Goldberg et al described a percentage of 18% endometrial cancer in women with insufficient or failed samples.This ar ticle from 1982, lacked detail on the small number of women (n=12) included.19
Diagnosis of endometrial cancer
From all 12 articles we could extract data on the sensitivity and specificity regarding the diagnosis endometrial cancer (Table 3).The sensitivity of endometrial sampling was 100% in all five studies using blind D&C, but varied between 50-100% in the seven studies using hysteroscopy with histology as a reference standard, with a weighted sensitivity of 90%. Specificity was 99-100% regardless of the reference standard that was used. Figure 3A shows an ROC plot of the performance of the 12 studies that allowed the calculation of both sensitivity and specificity.
Diagnosis of (pre)cancer of the endometrium
With respect to the diagnosis of endometrial (pre-) cancer, i.e. atypical hyperplasia or endometrial cancer we could calculate sensitivity and specificity from the data in all five studies using D&C as a reference and in four studies using hysteroscopy as a reference (table 3). The weighted sensitivity in studies using D&C was 92% (range 71-100%), whereas the weighted sensitivity in studies using hysteroscopy as a reference standard was 82% (range 56-94%). Specificity was 99-100% in all studies. Figure 3B shows an ROC plot of the performance of the twelve studies that allowed the calculation of both sensitivity and specificity.
138