Objective: To examine the performance of the Risk of Malignancy Index (RMI) and Risk of Ovarian Malignancy Algorithm (ROMA) by histologic subtype and stage of disease in a cohort of women with ovarian cancer.
Methods: All patients with confirmed ovarian cancer at the Princess Margaret Hospital between February 2011 and January 2013 were eligible for study inclusion. Preoperative cancer antigen 125, human epididymis protein 4, and ultrasound findings were reviewed, and the sensitivity and false-negative rates of the RMI and ROMA were determined by stage of disease and tumor histology.
Results: A total of 131 patients with ovarian cancer were identified. High-grade serous (HGS) histology was most frequently associated with stage III/IV disease (n = 46 [72% of stage III/IV]) vs stage I (n = 5 [11% of stage I]; P < 0.0001). Clear cell (CC) and endometrioid (EC) histology presented most commonly with stage I disease (n = 9 [20%] and n = 13 [29% of stage I cases], respectively). Median cancer antigen 125 and human epididymis protein 4 values were significantly higher for HGS than for EC or CC histology. Risk of Malignancy Index II demonstrated the highest sensitivity of the 3 RMI algorithms. All RMIs and ROMA were significantly more sensitive in predicting malignancy in patients with HGS than EC or CC histology. Risk of Malignancy Index II (n = 38) and ROMA (n = 35) exhibited sensitivities of 68% and 54% and false-negative rates of 32% and 46%, respectively, for patients with stage I disease vs sensitivities of 94% and 93% and false-negative rates of 6% and 7% for patients with stage III/IV disease.
Conclusion: Both RMI and ROMA performed well for the detection of advanced ovarian cancer and HGS histology. These triaging algorithms do not perform well in patients with stage I disease where EC and CC histologies predominate. Clinicians should be cautious using RMI or ROMA scoring tools to triage isolated adnexal masses because many patients with stage I malignancies would be missed.