Skip navigation
Link to Back issues listing | Back Issue Listing with content Index | Subject Index

Hysteroscopy for endometrial cancer and hyperplasia

Review
Results
Endometrial cancer
Endometrial disease
Comment

Diagnostic test evidence is different from that of therapeutics. It is more difficult because the quality of studies is often very low, so that they suffer from bias (Bandolier 26 & 70), and high quality studies are rare. It is also difficult because the way in which studies and meta-analyses are reported are hard to understand, and are neither intuitive nor, for most people, immediately useful (Bandolier 28). A systematic review of hysteroscopy [1] offers an opportunity to examine some of these issues again.

Review

The review focused on observational studies in which hysteroscopy was compared with results of the reference standard of endometrial histology. Verification of hysteroscopy diagnosis was followed either at the same time or after a short delay. The outcome was the accuracy of endometrial cancer and hyperplasia diagnosis.

Searching involved MEDLINE and EMBASE to the end of 2001, as well as the Cochrane Library, and without language restriction. Studies retrieved were assessed for methodological quality using a five-item hierarchy (Table 1). Studies in levels 1-3 were considered to be high quality, and 4 and 5 low quality. Technical failure in hysteroscopy so that no diagnosis was made were categorised as failed procedures. Information from the studies was collected on setting and pre or postmenopausal status.

Results

There were 65 primary studies involving 26,346 women. Of these studies, only 12 (18%) were of high quality (Table 1). Only one study was of the ideal quality, that is, an independent, blind comparison with reference standard among an appropriate population of consecutive patients.

Table 1: Definitions of quality criteria for study and number of studies with particular quality scores. Scores 1-3 were rated high quality

 

Quality
Definition
Number
1
An independent, blind comparison with reference standard among an appropriate population of consecutive patients
1
2
An independent, blind comparison with reference standard among an appropriate population of nonconsecutive patients or confined to a narrow population of study patients
1
3
An independent, nonblind comparison with reference standard among an appropriate population of consecutive patients
10
4
An independent, nonblind comparison with reference standard among an appropriate population of nonconsecutive patients or confined to a narrow population of study patients
42
5
An independent, blind comparison among an appropriate population of patients, but reference standard not applied to all patients
11

In 35 studies failure rates were reported, with an overall failure to make a diagnosis with hysteroscopy of 3.6%. Potentially severe uterine complications were reported in eight cases, but only 19 studies with 9,413 procedures explicitly stated an intention to report them. A worst case risk of a serious complication would then be 1 in 1177 cases.

Endometrial cancer

For endometrial cancer there were 56 unique studies with 61 sets of data, only 11 of which were deemed of high quality. The post-test probability of endometrial cancer with positive hysteroscopy and with a prevalence of 3.9% is shown in Figure 1. There was considerable variability according to high versus low quality, by different settings, and by menopausal status.

Figure 1: Post-test probability of endometrial cancer, by quality, setting, and menopausal status

In high quality studies the likelihood ratio for a positive test was reported as 35, giving a post-test probability of endometrial cancer of 59%. The likelihood ratio for a negative test was reported as 0.2, giving a post-test probability of endometrial cancer of 0.8%.

Endometrial disease

Endometrial disease was defined as endometrial cancer, hyperplasia, or both. For endometrial disease there were 41 unique studies with 71 sets of data, only 12 of which were deemed of high quality. The post-test probability of endometrial disease with positive hysteroscopy and with a prevalence of 10.6% is shown in Figure 2. There was considerable variability according to high versus low quality, by different settings, and by menopausal status.

Figure 2: Post-test probability of endometrial disease, by quality, setting, and menopausal status

In high quality studies the likelihood ratio for a positive test was reported as 5.5, giving a post-test probability of endometrial disease of 39%. The likelihood ratio for a negative test was reported as 0.3, giving a post-test probability of endometrial disease of 3.5%.

Comment

The definition of high quality studies in this review was probably justified. Certainly those in grades 4 and 5 are studies known to be associated with bias. There may still be a question mark over whether non-blind comparisons are subject to bias, and if they were, then 10 out of the 12 studies rated high quality would also be subject to possible bias. That would leave only two studies, not much grist for the mill of meta-analysis there.

Why then bother to perform an analysis using studies known to have potential faults? Perhaps because there's nothing else. Good studies of diagnostic tests are exceptional, as this review proves again.

So even if we include studies of poor design, what was the result? One way of looking at it for cohorts of 1,000 theoretical women is shown in Figure 3.

Figure 3: Natural frequencies for endometrial cancer and endometrial disease in cohorts of 1,000 women

But these results may just be the figments of the imagination of biased trials. To know how good hysteroscopy actually is, we have to wait for new studies. We also need reassurance that the gold standard, of histopathology, really is gold and not some shiny base metal. Agreement in histopathology is not always high, and the quality of histopathology will have a major impact on the results.

Once again, the ethics of involving patients in studies that are unlikely to produce a useful result, in this case the best part of 24,000 women, should be considered. Once again, the deficient design quality of diagnostic test studies has been highlighted.

Reference:

  1. TJ Clark et al. Accuracy of hysteroscopy in the diagnosis of endometrial cancer and hyperplasia. A systematic quantitative review. JAMA 2002 288: 1610-1621.

previous story