Ultrasound assessment of endometrial morphology and vascularity to predict endometrial malignancy in women with postmenopausal bleeding and sonographic endometrial thickness ≥ 4.5 mm

Abstract

To determine which endometrial morphology characteristics as assessed by gray-scale ultrasound and which endometrial vessel characteristics as assessed by power Doppler ultrasound are useful for discriminating between benign and malignant endometrium in women with postmenopausal bleeding (PMB) and sonographic endometrial thickness >or= 4.5 mm and to develop logistic regression models to calculate the individual risk of endometrial malignancy in women with PMB, endometrial thickness >or= 4.5 mm, good visibility of the endometrium and detectable Doppler signals in the endometrium. Of 223 consecutive patients with PMB and sonographic endometrial thickness >or= 4.5 mm, 120 fulfilled our inclusion criteria. They underwent transvaginal gray-scale and power Doppler ultrasound examination, which was videotaped for later analysis by two examiners with more than 15 years' experience in gynecological ultrasonography. They independently assessed endometrial morphology and vascularity using predetermined criteria. Their agreed-upon description was compared with the histological diagnosis. Univariate and multivariate logistic regression analyses were used. The best diagnostic test was defined as the one with the largest area under the receiver-operating characteristics curve (AUC). Thirty (25%) endometria were malignant. Inter-observer agreement for the description of endometrial morphology and vascularity was moderate to good (Kappa 0.49-0.78). The best ultrasound variables to predict malignancy were heterogeneous endometrial echogenicity (AUC 0.83), endometrial thickness (AUC 0.80), and irregular branching of endometrial blood vessels (AUC 0.77). A logistic regression model including endometrial thickness and heterogeneous endometrial echogenicity had an AUC of 0.91. Its mathematically best risk cut-off yielded a positive likelihood ratio of 4.4, and a negative likelihood ratio of 0.1. Adding Doppler information to the model improved diagnostic performance marginally (AUC 0.92). In selected high-risk women with PMB and an endometrial thickness of >or= 4.5 mm, calculation of the individual risk of endometrial malignancy using regression models including gray-scale and Doppler characteristics can be used to tailor management. These models would need to be tested prospectively before introduction into clinical practice.