Urine and vaginal cytology for endometrial cancer detection in women with postmenopausal bleeding in northwest England (DETECT): diagnostic accuracy study

Background Postmenopausal bleeding prompts urgent investigation with sequential invasive and costly tests that can be painful and distressing for women. A simple non-invasive test to identify cancer and safely reassure women with postmenopausal bleeding who do not have cancer would transform patient care. We aimed to prospectively validate the diagnostic test accuracy of urine and vaginal cytology for endometrial cancer detection in women with postmenopausal bleeding. Methods This multicentre diagnostic accuracy study was conducted at seven hospital sites in northwest England. We included consecutive women who presented with postmenopausal bleeding to a gynaecology clinic at one of the participating hospitals. Postmenopausal was defined as being at least 12 months after menstruation had stopped. All women provided a self-collected voided urine sample, and a Delphi screener-collected vaginal fluid sample was also obtained, before routine clinical investigations were done for suspected endometrial cancer. Samples were independently assessed by two of three cytopathologists unaware of participant cancer status, and discrepancies were settled by consensus review. Endometrial histopathology was the reference standard when performed. Disease-negative status was also defined by discharge from the cancer exclusion pathway and no cancer diagnosis within 3 months. From June 16, 2020, we also included women at higher risk of endometrial cancer (with endometrial thickness on transvaginal ultrasound >4 mm) and women with proven endometrial cancer before hysterectomy (the enriched cohort). Findings Between Sept 3, 2018, and March 31, 2021, 1890 participants were recruited, of whom 26 were withdrawn and 1864 were enrolled; 37 women comprised the enriched cohort. Of 1864 participants, 115 (6·2%) had either endometrial cancer (n=99) or other pelvic cancers (n=16). The other pelvic cancers comprised cervical cancer (n=7), ovarian cancer (n=3), leiomyosarcoma (n=2), colorectal cancer (n=2), bladder cancer (n=1), and metastatic pancreatic cancer (n=1). The sensitivity of combined urine and vaginal cytology for endometrial cancer was 80·8% (95% CI 71·7–88·0) and specificity was 92·6% (91·2–93·8). For any pelvic cancer, detection sensitivity was 80·0% (71·5–86·9) and specificity was 92·6% (91·2–93·8). The negative predictive value was 98·8% (98·2–99·3) for endometrial cancer detection and 98·6% (97·9–99·1) for any pelvic cancer detection. Vaginal cytology performed slightly better than urine cytology for endometrial cancer detection, with a sensitivity and specificity of 79·8% (70·5–87·2) and 94·4% (93·2–95·5) versus 73·5% (63·6–81·9) and 94·7% (93·5–95·7). Of 19 endometrial cancers originally missed by urogenital cytology, two (10·5%) had high-grade histology and one (5·3%) was stage 2 or worse, meaning that cytology detected 95·8% of aggressive histologies and 96·4% of locally advanced or metastatic diseases. Interpretation A novel diagnostic tool comprising urine and vaginal cytology showed good sensitivity and specificity for endometrial cancer detection. This method holds great promise for triaging women with postmenopausal bleeding before invasive tests. Studies assessing the clinical use of this diagnostic tool in practice are now required. Funding The Jon Moulton Charity Trust.