Triage method for endometrial biopsy in postmenopausal women: a multicenter retrospective cohort study.

This retrospective cohort study enrolled women who underwent endometrial biopsy with or without hysteroscopy in multicenter obstetrics and gynecology departments between March 2021 and March 2022. Data on the following characteristics were collected in an electronic database: age, body mass index (BMI) (categorized as underweight or normal weight [<24 kg/m 2 ], overweight [24-29.9 kg/m 2 ], and obese [≥30 kg/m 2 ] 11 ), menopause time, hypertension status, and diabetes status. 7 , 12 Eligible women included 40- to 80-year-old postmenopausal women with PMB and/or TVU abnormalities. Menopause was characterized by the permanent cessation of menstruation and is clinically diagnosed after 12 months of complete amenorrhea. 13 PMB was defined as any vaginal bleeding occurring at least 12 months after the last menstrual period in women who were not on hormonal therapy (HT) or had unscheduled bleeding while receiving HT. 14 TVU abnormalities included endometrial thickness ≥5 mm or abnormal imaging features (cystic endometrium, fluid in the cavity, suspected polyps, or other suspicious features). The exclusion criteria included unavailable data on age at menopause, incomplete TVU report, absence of PMB and/or TVU abnormalities, removal of the intrauterine contraceptive device, incomplete hysteroscopy or endometrial biopsy pathology report, failure of hysteroscopy, and prior history of malignancy. To further verify the results, information on women with TVU abnormalities who underwent endometrial biopsy in one of the four hospitals between May 2022 and May 2023 was collected. Ethical approval, with a waiver of informed consent, was obtained from the ethics committee of our hospital (date of approval: August 3, 2022, ethics No. 202208179). All hysteroscopists and ultrasonographers involved in the study had significant experience and appropriate skill sets. Women underwent hysteroscopy based on a local policy if they had PMB symptoms and/or TVU abnormalities. TVU was performed according to the International Endometrial Tumor Analysis Group guidelines. 15 The uterus was examined using two scanning planes: the sagittal plane, spanning from one cornu to the other, and the (oblique) transverse plane, extending from the cervix to the fundus. Endometrial thickness was measured as the maximum value in the sagittal plane, encompassing both endometrial layers. If the endometrium could not be clearly visualized, it was reported as “non-measurable.” The endometrial morphology, including endometrial echogenicity and midline, was also examined. The ultrasound images were independently reviewed by a second ultrasonographer who was completely blinded to the initial TVU results. All images obtained during the initial TVU for women with a thickened endometrium were retrieved, and the endometrial thickness (ET) was remeasured by the second reviewer. For women who were subsequently diagnosed with EC, magnetic resonance imaging was performed before surgery, and the ET was re-measured. The final ET used for further analysis was calculated as the mean of the measurements at the TVU, secondary TVU review, and magnetic resonance imaging, when available. Pathological diagnosis was considered the final diagnosis. When the tissue was unavailable for pathological diagnosis, direct vision under hysteroscopy was considered the reference for the final diagnosis. The endpoint was the final diagnosis, malignancy/premalignancy according to the pathology results, or nonmalignancy (normal and benign) according to the hysteroscopy report or biopsy pathology results. Benign included being diagnosed with conditions such as endometrial polyps, endometrial hyperplasia without atypia, or other benign histological findings (e.g., fibroids and adenomyoma). The study sample size was calculated considering a clinically acceptable level of precision, hypothesized values of sensitivity and specificity, and the estimated prevalence of disease in the study participants. 16 , 17 For the biopsy referral flow, our primary goal was to identify as many malignancies as possible for further invasive diagnostic flow; therefore, we only calculated the sample size based on high sensitivity. We selected anticipated sensitivity (S N ) = 0.90, the critical region (α) = 0.05, precision (L) = 0.10 with a 95% confidence interval (CI), and prevalence = 0.10, which suggested that the study sample size should be 346. Continuous variables are presented as means and standard deviations (SD), whereas qualitative variables are presented as absolute frequencies and percentages. The baseline characteristics of women with and without malignant/premalignant lesions were compared using the χ 2 test and analysis of variance (ANOVA) test when appropriate. Non-parametric Mann-Whitney and Kruskal-Wallis tests were used for asymmetric continuous variables. Three triage tests were employed: test A (women with PMB were referred for endometrial biopsy), test B (women with TVU abnormality were referred for endometrial biopsy), and test C (women with both PMB and TVU abnormalities were referred for endometrial biopsy). The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated using the three triage methods. Receiver operating characteristic (ROC) curve analysis, sensitivity, specificity, and NPV were used to calculate the most suitable cutoff value for endometrial thickness (ET) associated with malignancy/premalignancy in women with available ET data. A univariate logistic regression model was used to identify the potential risk variables for EC (age, menopausal age, hypertension, diabetes, and ET). Factors with a P value <0.25 in univariate analysis were included in the multivariate logistic analysis, which has been described in previous studies. 18 Based on the results of multivariate logistic regression analysis, a nomogram integrating valuable independent clinical variables was constructed to predict endometrial malignancy/premalignancy. The corresponding 95% confidence interval (CI) was calculated, and the statistical significance level was set at P < 0.05. Statistical analyses were performed using SPSS version 26 (Statistics Package for Social Sciences, IBM, Armonk, NY). The R software package (Version 4.2.2; R Foundation for Statistical Computing, Vienna, Austria) was used to construct the nomogram and internal and external validations of the nomogram. The optimal cutoff value was determined by maximizing the sensitivity through ROC curve analysis.

Overall, 470 women underwent endometrial biopsy with or without hysteroscopy in the gynecologic departments of four hospitals between March 2020 and March 2021. Sixty-five women were excluded for the following reasons: (1) incomplete clinical records ( n = 10), (2) uncertain final diagnoses ( n = 6), (3) absence of PMB and TVU abnormalities ( n = 18), and (4) history of malignancy ( n = 30). A total of 405 women were enrolled in this study. The pathological results revealed that 46 women had malignant diseases (including endometrial cancer, carcinosarcoma, and adenosarcoma), eight had premalignant disorders, 218 had benign disorders, and 133 had normal endometrium. Differences in hypertension, diabetes, and ET were statistically significant among women with normal, benign, and malignant/premalignant conditions (Table 1 ). However, no significant differences were found in the clinical characteristics between women assigned to each of the three triage methods, indicating that the population was comparable across different triage methods (Supplemental Digital Content 1, http://links.lww.com/MENO/B171 , demonstrating the baseline characteristics of participants in the three triage tests). Baseline characteristics of women enrolled The bold values mean that there is a statistical significance between two groups. BMI, body mass index; ET, endometrial thickness; PMB, postmenopausal bleeding. a Women without endometrial thickness data were not included. We compared the sensitivity, specificity, PPV, NPV, and number of biopsies per malignancy/premalignancy detected using the three triage methods. For tests A, B, and C, the sensitivity and specificity were 79.6% and 44.2%, 100% and 19.7%, and 79.6% and 63.8%, respectively. Among the tests, test B proved to be the most sensitive, not missing any malignancy, but it was also not specific. The PPV and NPV for tests A, B, and C were 18% and 93.4%, 100% and 6.2%, and 25.3% and 95.3%, respectively, while the numbers of biopsies per malignancy/premalignancy detected (women underwent biopsy/hysteroscopy to discover one malignancy/premalignancy) were 5.6, 6.2, and 4, respectively (Table 2 ). Of the 96 women with ultrasound findings without endometrial thickness, 25% (24/96) had malignant/premalignant disease, which was higher than the 9.7% (30/309) of women with measurable endometrial thickness. This finding may have led to bias and interference in our ET cutoff values. Therefore, in the subsequent analysis, we excluded women without ET data, and only included women with endometrial thickness data ( n = 309). To ensure the discovery of more malignant/premalignant disease, we suggested the optimal ET cutoff value for biopsy referral as 4 mm, which shows 93.3% sensitivity, 24.7% specificity, and 97.2% NPV (Supplemental Digital Content 2, http://links.lww.com/MENO/B172 , shows the ROC curve of ET for the detection of endometrial (pre)malignancy, area under curve [AUC], 0.718 (95% CI: 0.616 to 0.821]; and Supplemental Digital Content 3, http://links.lww.com/MENO/B173 , Supplemental Digital Content 4, http://links.lww.com/MENO/B174 , Supplemental Digital Content 5, http://links.lww.com/MENO/B175 , demonstrates number of malignancy/premalignancy, sensitivity, specificity, and NPV at different ET cutoff values in PMB, asymptomatic, and all enrolled women). Together, we found that it was better to refer all postmenopausal women with TVU abnormity (ET ≥ 4 mm or other abnormal imaging features) to endometrial biopsy. Sensitivity, specificity, PPV, NPV, and number of biopsies per malignancy/premalignancy detected of three triage tests Test A: women with PMB would be referred to endometrial biopsy service. Test B: women with TVU abnormity would be referred to endometrial biopsy service. Test C: women with both PMB and TVU abnormity would be referred to endometrial biopsy service. NPV, negative predictive value; PMB, postmenopausal bleeding; PPV, positive predictive value; TVU, transvaginal ultrasonography. To improve the applicability of the triage method further, we established a nomogram prediction model for women with TVU abnormalities. According to the univariate and multivariate logistic analyses, PMB (odds ratio [OR], 3.241; 95% CI, 1.073-9.789), diabetes (OR, 10.915; 95% CI, 3.389-35.156), and ET (OR, 1.277; 95% CI, 1.156-1.409) were independent predictive factors for endometrial (pre)malignancy and were used to construct the nomogram prediction model (Table 3 and Fig. 1 ). For each participant, an upward vertical line is drawn to the “Points” bar from “PMB,” “Diabetes,” and “ET” bars below to calculate the total points corresponding to the women's clinical characteristics. The sum calculated is used in “Total Points” axis to draw a downward line to the “Risk” axis, yielding the estimated percentage risk of (pre)malignancy for that specific woman. Univariate and multivariate logistic regressions for risk factors of malignancy/premalignancy in women with TVU abnormity ( n = 282) The bold values mean that there is a statistical significance between two groups. ET, endometrial thickness; PMB, postmenopausal bleeding; TVU, transvaginal ultrasonography. The use of nomogram for predicting the risk of malignancy/premalignancy in women with TVU abnormity ( n = 282). The nomogram consists of bars listing each independent risk factor involved, with a corresponding “Points” axis for score. The length of the bars for “Postmenopausal bleeding” and “Diabetes” demonstrate the score assigned to each risk factor. The ‘Endometrial thickness’ bar shows the points per millimeter of thickness. For each participant, vertical lines is drawn upward to “Points” to sum the total points according to the individual's clinical characteristics. The sum calculated is used in “Total points” axis to draw a downward line to the “Risk” axis, yielding the estimated percentage risk of (pre)malignancy for that specific women. ET, endometrial thickness; PMB, postmenopausal bleeding; TVU, transvaginal ultrasonography. In internal validation, the model achieved an AUC of 0.833 (95% CI, 0.740-0.927). To avoid missing one (pre)malignancy, a cutoff value of 22.5 was used; at this cutoff value, the nomogram had a sensitivity of 100.0% and specificity of 15.7% (Fig. 2 ). Accuracy of the prediction score of the nomogram for estimating the probability of malignancy/premalignancy ( n = 282). The area under the curve was 0.833 (95% CI, 0.740-0.927). We determined the ideal cutoff value as 22.5 (sensitivity 100.0% and specificity 15.7%) to identify all (pre) malignancies. A total of 112 women with TVU abnormalities were enrolled in the validation set (Supplemental Digital Content 6, http://links.lww.com/MENO/B176 ). Most clinical characteristics were not significantly different between the training and validation sets. In the external validation of the nomogram, the AUC was 0.798 (95% CI, 0.685-0.911). At a cutoff value of 22.5, the sensitivity was 100.0%, the specificity was 21.6%, the PPV was 11.1%, and the NPV was 100.0% (Supplemental Digital Content 7, http://links.lww.com/MENO/B177 ; ROC curves of the external verification set showed an AUC of 0.798 [95% CI, 0.685-0.911]). The additional inclusion of a nomogram could not only triage 100% of women with (pre)malignancy to undergo endometrial biopsy but could also prevent 25% (28/112) of women from invasive biopsy, thereby increasing the specificity of the triage flow compared to simply referring all postmenopausal women with TVU abnormalities to endometrial biopsy.

Referring women with TVU abnormalities for endometrial biopsy could identify 100% of the malignant/premalignant lesions despite a relatively low specificity (19.7%). Among women with measurable ET, we found that an ET cutoff value for biopsy referral of ≥4 mm ensured maximum identification of malignant/premalignant disease (sensitivity: 93.3% and specificity: 24.7%). Independent predictive factors for endometrial (pre)malignancy were PMB (OR, 3.241; 95% CI, 1.073-9.789), diabetes (OR, 10.915; 95% CI, 3.389-35.156), and endometrial thickness (OR, 1.277; 95% CI, 1.156-1.409). A nomogram prediction model was constructed to further improve the applicability of the triage method (AUC, 0.802; 95% CI, 0.715-0.889). The ideal cutoff point was 22.5, with a sensitivity of 100.0% and a specificity of 15.7%. The external validation achieved an AUC of 0.798 (95% CI, 0.685-0.911). In the existing literature, there is an urgent need to propose an ideal biopsy referral flow to identify as many endometrial malignancy/premalignancies as possible, while also referring as few women as possible to undergo invasive endometrial biopsy. After comparing the three triage methods, our results showed that referring all patients with TVU abnormalities to the endometrial biopsy service is optimal. The high sensitivity of TVU in the diagnosis of endometrial malignancy has been confirmed in various studies despite its low specificity and potentially unreliable results due to axial uterus, adenomyosis, and coexisting myomas. 19 - 20 However, the unequal sample size for the three triage methods is a limitation of this study and an important factor to consider when interpreting the results. Moreover, a significant difference in sample size can introduce statistical variability and affect the accuracy of the study's findings. In addition, we found that women whose ET was unmeasurable under TVU had a higher rate of malignant/premalignant disease, which has also been mentioned in a previous report. 21 One possible reason for this may be that endometrial malignancy myoinvasion makes the ET difficult for TVU physicians to depict. After excluding women with unknown ET (which may lead to an overestimation of the sensitivity of TVU abnormality), we found that the optimal ET cutoff value for biopsy referral was 4 mm. According to ACOG Committee Opinion, 4 TVU is sufficient for an initial evaluation of PMB when ET ≤ 4 mm, given that the NPV for EC will be greater than 99%. Our research suggests that women with PMB are referred to biopsy when the ET is ≥4 mm because 95.7% (22/23) of (pre)malignant lesions and 72.6% (61/84) of benign lesions could be identified at this cutoff. A stricter ET cutoff value will help us to identify the most benign lesions that should be treated under hysteroscopy at an early stage and prevent potential discomfort, as well as the possibility of malignant changes in the future. For asymptomatic women, when the ET cutoff value was expanded to 6 mm, the NPV for malignancy still reached 97.9%; however, 25.0% (22/88) of women with benign lesions could have been missed with this eased ET cutoff value. The need for the early treatment of benign lesions increases anxiety and the burden of repeated visits. As a result, we propose the use of a 4 mm threshold. Referring all women with TVU abnormity (ET ≥ 4 mm or other abnormal imaging features) for endometrial biopsy could provide a safer and more reasonable reference for postmenopausal women in triage to invasive endometrial biopsy, which deserves more studies in the future to test its clinical applications. Endometrial malignancy has been proven to be related to obesity, diabetes, metabolic syndrome, and age at menopause. 22 - 24 Here, we observed that PMB, diabetes, and ET were independent predictive factors for endometrial (pre)malignancy. The PMB, which has the potential to screen for up to 90% of malignant endometrial diseases, is an important clinical feature of endometrial (pre)malignancy. 25 Various studies have reported the effects of metabolic syndromes, especially diabetes, on the development and prognosis of EC. A diabetes risk-reducing diet may reduce the risk of endometrial cancer. 26 Elevated blood glucose levels and insulin resistance, commonly observed in individuals with diabetes, may contribute to the development and poor survival outcomes of EC, which may also affect the outcomes of conservative treatment for endometrial hyperplasia and early EC. 27 , 28 However, hypertension was not an independent risk factor for EC in our study, possibly due to the small sample size and selection bias. A nomogram for estimating the probability of endometrial malignancy was constructed based on these three risk factors. A nomogram score greater than the cutoff value of 22.5 identified a chance of malignant lesions; therefore, such cases must be rapidly referred for endometrial biopsy. For women with total scores less than 22.5, regular follow-ups may be optimal. In the external validation, the additional inclusion of the nomogram triaged 100% of women with (pre)malignancy to undergo an endometrial biopsy, as well as prevented 25% of women from undergoing an invasive biopsy, demonstrating its potential for clinical application. Notably, we also found that 53.8% of the women had benign lesions (uterine polys, endometrial hyperplasia without atypia) that could be treated using hysteroscopy. As these benign disorders can also cause symptoms of endometrial malignancy, these women still require medical intervention to rule out malignancy. 29 - 31 Biopsy under hysteroscopy is safe and feasible for the accurate diagnosis of intrauterine lesions, especially those with focal growth, which are otherwise missed by blind endometrial sampling. 32 In women with a thin endometrium whose tissue biopsy cannot be obtained by blind sampling, hysteroscopy can also be used for the initial evaluation. The detection of benign lesions in our study supports the necessity of biopsy referral under hysteroscopy, which not only allows for direct visual localization of suspicious lesions for biopsy but can also help with the treatment of some benign lesions. 33 , 34 This multicenter study was conducted among more than 400 postmenopausal women in the test set, and the results were further verified in another 100 women in the validation set; therefore, the results are representative and reliable as preliminary data, which can be used to help design future prospective studies. Our study has some limitations. First, the data were collected retrospectively from four hospitals, which likely introduced selection and information biases. Indeed, variations among ultrasonographers in different hospitals can introduce bias in the measurement of ET. Moreover, as this was not a randomized trial, the treatment strategies of different physicians may vary, which could influence our results. As we do not have follow-up information on all women, the pathology results may be false negatives, considering the heterogeneity of the biopsy sample and the future progression of the disease. Furthermore, it is important to note that the results obtained from a limited number of hospitals in one province may not be fully generalizable. Therefore, our findings should be regarded as preliminary evidence that supports the need for future studies involving a more extensive and diverse range of postmenopausal women. Expanding the scope of this study to include a broader population will enhance the reliability and applicability of our conclusions. Finally, it is crucial to acknowledge that the small sample size in our study resulted in statistical instability, as evidenced by the wider confidence intervals for independent risk factors. This limitation increases the possibility of overlooking certain predictive variables that may have significant implications.

To conclude, we found that postmenopausal women with TVU abnormalities (ET ≥ 4 mm or other abnormal imaging features) could be referred for endometrial biopsy. A nomogram model for estimating the probability of endometrial malignancy was constructed among women with TVU abnormalities and a score below the cutoff value of 22.5 suggested regular follow-up should be conducted instead of invasive endometrial biopsy.