OC12.06: Validation of objective measurements to predict myometrial or cervical stromal invasion and prediction models to predict high-risk endometrial cancer

To estimate and validate cut-offs for objective measurements to predict deep myometrial invasion (MI), cervical stromal invasion (CSI). To validate the subjective and objective 2-step model to identify high risk cancer. Prospective multicentre study in 1714 women with endometrial cancer undergoing standardised expert transvaginal ultrasound examination: MI/CSI were evaluated subjectively and compared to objective measurements (at optimal and published cut-offs) to predict MI: Tumour/Uterine AP ratio and CSI: distance from outer-cervical-os to lower margin of tumour (Dist-OCO). We also validated a 2-step strategy to predict high risk cancer. First step: biopsy grade 3/non-endometrioid cancer classified as high-risk cancer. Second step: mathematical model with biopsy grade 1-2 and subjective assessment of MI/CSI (subjective model) or grade 1-2 & minimal-tumour-free margin (objective model). Histological assessment from hysterectomy was used as gold standard. After exclusions, 1538 women were available for analysis. In this series, the optimal cut-off for MI for tumour/uterine AP ratio was 0.52 (accuracy 71%, AUC 0.77), for CSI for Dist-OCO was 23mm (accuracy 70%, AUC 0.77). When validated, the published cut-off for Tumour/Uterine AP ratio; 0.53 and Dist-OCO; 21 mm showed similar performance (accuracy 71%, AUC 0.77 and 70%, 0.72) as in the original paper on 144 patients (74%, AUC 0.80 and 65%, AUC 0.75). Subjective impression and grade had an accuracy of 80% to predict high risk cancer. The subjective model was slightly superior to objective model (AUC 0.76 versus 0.75), in line with the original series (accuracy of 80% (AUC 0.76) versus 74% (AUC of 0.75). Tumour/uterine AP ratio at the 0.52-0.53 cut-off and dist-OCO of 21-23mm had an accuracy of 70-74% and 65-70% to predict MI and CSI, respectively. High risk disease was best identified by subjective impression or by the subjective 2-step strategy at an accuracy of 80%.