Our aim was to determine the diagnostic accuracy of endometrial thickness measurement by pelvic ultrasonography for predicting endometrial carcinoma and disease (hyperplasia and/or carcinoma) during an investigation of postmenopausal bleeding. We performed a systematic quantitative review of the available published literature, which consisted of online searching the MEDLINE and EMBASE databases (1966-2000) coupled with scanning of bibliography of known primary and review articles. The selection of studies, assessment of study quality, and extraction of data were performed in duplicate under masked conditions. Included in the analyses were 57 studies with 9031 patients. Accuracy data were summarized using likelihood ratios for various cut-off levels of abnormal endometrial thickness. The commonest cut-offs were 4 mm (9 studies) and 5 mm (21 studies), measuring both endometrial layers. None of the nine studies using the < or = 4 mm cut-off level were of good quality. Only four studies (out of the 21) used the < or = 5 mm cut-off level, which employed the best-quality criteria. Using the pooled estimates from these four studies only, a positive test result raised the probability of carcinoma from 14.0% (95% CI 13.3-14.7) to 31.3% (95% CI 26.1-36.3), while a negative test reduced it to 2.5% (95% CI 0.9-6.4). In conclusion, ultrasound measurement of endometrial thickness alone, using the best-quality studies cannot be used to accurately rule. However, a negative result at < or = 5 mm cut-off level measuring both endometrial layers in the presence of endometrial pathology rules out endometrial pathology with good certainty.