P11.02: Performance of IOTA ADNEX model in differentiating histological subtypes of borderline ovarian tumours

Abstract

To analyse the performance of IOTA ADNEX model for the differentiation of histological subtypes of borderline ovarian tumours (BOT). Retrospective analysis of ultrasound (US) dataset of patients diagnosed with BOT on the final histology was done at the tertiary oncology centre in the period of 2009–2016. The IOTA ADNEX model (web application) was used to calculate the absolute risk (AR) and the relative risk (RR) for the mass being BOT. 57 (72.2%) of patients were diagnosed with serous BOTs (s_BOTs) and 22 (27.8%) with mucinous BOTs (m_BOTs). Without CA-125, according to AR, the ADNEX correctly classified 37 (64.9%) serous and 12 (54.5%) mucinous BOTs with no difference between the groups. According to RR, the test correctly differentiated 44 (77.2%) serous BOTs and 15 (68.2%) mucinous BOTs (p=0.409 between the groups). When tumours were classified according to RR, the performance of ADNEX increased by 35% and 30% in s_BOTs and m_BOTs groups respectively (p<0.05). When CA-125 was added, according to AR, 35 (68.6%) of s_BOTs and 11 (57.9%) m_BOTs were correctly classified. According to RR, 41 (80.4%) of s_BOTs and 14 (73.7%) of m_BOTs were classified correctly. The performance of ADNEX model increased by 37.5% in both groups when differentiating tumours according to RR vs. AR (p<0.05). The mean AR and RR was significantly higher in s_BOTs compared to m_BOTs group (30.8% vs. 18.1% and 4.89 vs. 2.88, respectively; p=0.002). However, when CA-125 was added, the difference of mean AR (29.4% vs. 21.4%) and RR (4.67 vs. 3.39) slightly decreased and was no longer significant (p=0.106). IOTA ADNEX model works well in differentiating BOTs of different histological subtypes. The performance is better when tumours are differentiated according to the RR and the values of risk are higher in serous BOTs compared to mucinous ones.