OBJECTIVE To determine the specificity, sensitivity and predictive value of hysteroscopic impression versus histological diagnosis of endometrial curettings in evaluating patients with abnormal uterine bleeding. In addition, to determine whether office hysteroscopy can eliminate hospital diagnostic dilatation and curettage for patients with abnormal uterine bleeding. METHODS A retrospective study of 556 patients who underwent hysteroscopy and dilatation and curettage for abnormal uterine bleeding between January 1995 and December 1998 at the Salmaniya Medical Complex in Bahrain. A comparison was made between hysteroscopic impression and histological examination. RESULTS Out of 556 patients who were included in the study, 53 were diagnosed to have endometrial polyps hysteroscopically, however only 13 patients (24.5%) were confirmed to have polyps histologically. Hysteroscopy had revealed submucous leiomyoma in 33 women but none of these were diagnosed histologically. Hysteroscopy was highly specific for diagnosis of both endometrial hyperplasia and endometrial carcinoma (specificity was 85% for endometrial hyperplasia and 99.5% for endometrial carcinoma), however the sensitivity of hysteroscopy for diagnosing endometrial cancer was 40% and 30% for endometrial hyperplasia. CONCLUSIONS Hysteroscopy was more sensitive than curettage in detecting endometrial polyps and submucous fibroids, but less sensitive than curettage in detecting endometrial hyperplasia and endometrial carcinoma. Hysteroscopy should be carried out in conjunction with curettage for evaluating women with abnormal uterine bleeding. Office hysteroscopy with directed biopsies could be carried out, to reduce hospital diagnostic dilatation and curettage.