Diagnostic methods for fast-track identification of endometrial cancer in women with postmenopausal bleeding and endometrial thickness greater than 5 mm

OBJECTIVE: This study aims to evaluate the diagnostic efficiency of pattern recognition by transvaginal ultrasonography (TVS) and gel infusion sonography (GIS) for identifying endometrial pathology and to compare this setup with a standard setup of endometrial sampling (ES), hysteroscopy with pattern evaluation (HY(pattern)), or magnetic resonance imaging (MRI). METHODS: This study used a prospective cohort of 174 women with postmenopausal bleeding and endometrial thickness of 5 mm or greater. Resectoscopic biopsy (hysteroscopy with biopsy) samples or hysterectomy served as reference standard. Malignant and benign endometrial patterns were evaluated with TVS, GIS and HY(pattern) were then added. The efficiency of each diagnostic strategy, including ES and MRI findings (n = 83), was compared and evaluated against the reference standard. RESULTS: ES, TVS, GIS, and HY(pattern) had high diagnostic efficiency (area under the curve) for malignancy diagnosis (ES, 0.90; TVS, 0.88; GIS, 0.92; HY(pattern), 0.91). When insufficient samples were incorporated, ES was less efficient than the other techniques. ES was not more efficient in the subgroup of women without localized lesions than in the subgroup of women with localized lesions. MRI and HY(pattern) added limited efficiency, whereas hysteroscopy with biopsy was most efficient. CONCLUSIONS: As a first-line technique, pattern recognition on TVS, GIS, and HY(pattern) correctly identifies 9 of 10 women with malignancy and is superior to pattern recognition on ES when insufficient samples are included. Endometrial pattern evaluated with TVS and GIS is a fast and efficient first-line diagnostic tool that outperforms ES in women with or without localized lesions. Malignant patterns on TVS/GIS should warrant fast-track evaluation, whereas women with benign patterns may be selected for office or operative hysteroscopy. A fast-track diagnostic setup based on pattern recognition is presented.