Page 29 - The diagnostic work-up of women with postmenopausal bleeding

Data extraction, analysis and interpretation
From each systematic review or meta-analysis we extracted (if available) figures
for the sensitivity, specificity, likelihood ratios (LRs), the pre-test probability of
endometrial cancer and/or hyperplasia and the post-test probability. If an article 2 described both post- and premenopausal women we extracted the data relating
to the postmenopausal women. We used only data from studies evaluated as high
quality by the reviewers.
The LR indicates by how much a specific test raises or lowers the probability of having endometrial pathology.An LR of 1 indicates that the test has no predictive value for the outcome of interest. The higher an LR is above 1, the larger is the probability of pathology. An LR of less than 1 indicates that a negative test result is more likely to be true.In the present study,for a rating of‘high’diagnostic accuracy,the LR had to be over 10 for a positive test result or less than 0.1 for a negative result15. In this article, we use the LR of a negative test result (LR−) for reviews on the use of TVS, because we are interested in its accuracy in excluding endometrial cancer. For endometrial sampling and hysteroscopy we are interested in diagnosing cancer, so we used the LR of a positive test (LR+) and calculated the post-test probability of a positive test.When an article did not report post-test probability, we used Bayes’ theorem to calculate the post-test probability, using the following formula:
post-test probability LR x (pretest probability / (1-pre-test probability)) (LR x (pretest probability / (1-pre-test probability))) + 1
To compare study results we used pre-test probabilities extracted from the literature: a probability of 10% for endometrial cancer and a probability of 40% for focal benign or (pre)malign endometrial disease.16,17
Results
Study selection
A total of nine systematic reviews assessed (part of) the diagnostic pathway for women with PMB and met the criteria for inclusion (Figure 1). Of the selected systematic reviews, four articles assessed the use of TVS, 16,18-20 one described the use of SIS21, two assessed the use of outpatient endometrial sampling22,23 and two assessed the use of hysteroscopy in patients with abnormal uterine bleeding.24,25Table 1 shows further details of these studies.
Diagnostic evaluation
27