Accuracy of hysteroscopy in the diagnosis of endometrial cancer and hyperplasia: a systematic quantitative review.

CONTEXT Hysteroscopy (direct endoscopic visualization of the endometrial cavity) is used extensively in the evaluation of common gynecologic problems, such as menorrhagia and postmenopausal bleeding. However, there is a continuing debate about the value of this technology in the diagnosis of serious endometrial disease. OBJECTIVE To determine the accuracy of hysteroscopy in diagnosing endometrial cancer and hyperplasia in women with abnormal uterine bleeding. DATA SOURCES Relevant articles were identified through searches of the Cochrane Library, MEDLINE, and EMBASE (1984-2001), manual searches of bibliographies of known primary and review articles, and contact with manufacturers. STUDY SELECTION Studies were selected blindly, independently, and in duplicate if accuracy of hysteroscopy was estimated in women with abnormal uterine bleeding, using histopathologic findings as a reference standard. Our search identified 3486 articles; 208 of these were deemed to be potentially eligible and were retrieved for detailed data extraction. Sixty-five primary studies were analyzed, including 26 346 women. DATA EXTRACTION Data were abstracted on characteristics and quality from each study. Results for diagnostic accuracy were extracted to form 2 x 2 contingency tables separately for endometrial cancer and endometrial disease (cancer, hyperplasia, or both). Pooled likelihood ratios (LRs) were used as summary accuracy measures. DATA SYNTHESIS The pretest probability of endometrial cancer was 3.9% (95% confidence interval [CI], 3.7%-4.2%). A positive hysteroscopy result (pooled LR, 60.9; 95% CI, 51.2-72.5) increased the probability of cancer to 71.8% (95% CI, 67.0%-76.6%), whereas a negative hysteroscopy result (pooled LR, 0.15; 95% CI, 0.13-0.18) reduced the probability of cancer to 0.6% (95% CI, 0.5%-0.8%). There was statistical heterogeneity in pooling of LRs, but an explanation for this could not be found in spectrum composition and study quality. The overall accuracy for the diagnosis of endometrial disease was modest compared with that of cancer, and the results were heterogeneous. The accuracy tended to be higher among postmenopausal women and in the outpatient setting. CONCLUSION The diagnostic accuracy of hysteroscopy is high for endometrial cancer, but only moderate for endometrial disease (cancer or hyperplasia).