Guideline 61 - Supporting Material

Investigation of Post-Menopausal Bleeding

• Test performance (i.e. sensitivity, specificity and likelihood ratios)
• The probability that disease is present when the patient presents for possible investigation (i.e. pre-test probability)
• The cut-off value used to determine whether a result is normal or abnormal

These factors together determine the probability that disease is present following an investigation (i.e. post-test probability).

Test performance. Although transvaginal ultrasonography obtains an actual measurement of endometrial thickness, test results are typically reported simply as ‘positive’ or ‘negative’ depending on whether the thickness is above or below a specified threshold. For diagnostic tests providing a positive or negative result, performance across the study group is usually summarised in terms of sensitivity and specificity. Sensitivity is the probability of testing positive if the disease is truly present. Specificity is the probability of testing negative if the disease is truly absent. Based upon findings from a recent meta-analysis, transvaginal ultrasonography has a pooled sensitivity of 98% and pooled specificity of 53% for the detection of endometrial carcinoma, if a threshold of over 3 mm is used to define abnormal endometrial thickening.

Derived from data on sensitivity and specificity, the likelihood ratio (LR) indicates by how much a given test result will lower or raise the odds of having disease. Likelihood ratios are independent of the underlying prevalence of disease in the population being investigated. For a cut-off value of 3 mm, the likelihood ratio for a negative transvaginal ultrasonography result (LR -ve) is 0.04, meaning that negative result has reduced the odds of disease by 96% of what they were pretest, i.e. to 4%. In general, investigations with an LR -ve of less than 0.1 tend to be more helpful at excluding or ‘ruling out’ the target disease. The likelihood ratio for a positive transvaginal ultrasonography result (LR +ve) is 2.1. Generally investigations with an LR of 10 or greater tend to be more helpful at ‘ruling in’ the target disease.

The advantage of LRs over sensitivity and specificity is that LRs can be readily combined with relevant information about an individual’s pre-test risk of disease to provide a more useful estimate of the post-test probability of disease.

Table 1 summarises data from a meta-analysis of TVUS¹. This shows the 95% confidence intervals around the point estimates of likelihood ratios for different thresholds of endometrial thickness. Data presented in this guideline are derived from the point estimates only.

The pre-test probability (or prevalence) of the target disorder (endometrial cancer) in the group of patients being tested is just as important as performance in selecting and interpreting test results. A ‘negative’ TVUS result in a woman with PMB at lower risk of cancer (e.g. younger age group) is less likely to miss cancer than a ‘negative’ result in a woman at higher risk (e.g. older age group).

The pre-test probabilities are only an approximate guide because of uncertainty around estimates of the risk of cancer for different groups of women. The risk of individual women being assessed will not be identical to the risks provided for women in broad groups (e.g. age bands). Clinicians may wish to revise their own estimated baseline risks upwards or downwards according to other factors that influence their suspicion of the likelihood of cancer being present.

The cut-off value used for TVUS is often set locally. A ‘positive’ or ‘negative’ TVUS result is then determined according to whether the actual measurement of endometrial thickness is above or below that locally set value. In general, the greater the endometrial thickness, the higher the probability that cancer is present. A ‘negative’ TVUS result for a local cut-off point of 3 mm is therefore less likely to miss cancer (i.e. have a greater sensitivity) than cut-offs of 4 or 5 mm. There is a trade-off to be made, as lower cut-off points also result in a greater proportion of ‘false positives’ requiring further investigation. Adopting more than one cut off value may allow the interpretation of the test to be tailored to the patient’s pre-test probability (i.e. the patient risk group) and can help maximise detection of endometrial cancers.

The post-test probabilities can be derived from either the use of a formula or a normogram on the basis of the pre-test probability and likelihood ratios for TVUS. The post-test probability for a ‘positive’ result is also known as the positive predictive value. The post-test probability for a ‘negative’ result is related to the negative predictive value.

TVUS is most useful in determining which patients do not require further investigation. Therefore, the post-test probabilities for a negative result tend to be the most clinically useful. These probabilities were used in formulating the recommendations for the use of TVUS in this guideline. Further detail of how the recommendations were formulated given the available clinical data are presented in an option appraisal.

Table 1

Data on performance of transvaginal ultrasound in the detection of endometrial cancer.

1. Chien PFW, Voit D, Clark TJ, Khan KS, Gupta JKl. Ultrasonographic endometrial thickness for diagnosing endometrial pathology in women with postmenopausal bleeding: A meta-analysis. Acta Obstet Gynecol Scand. In press 2002.