Differential diagnosis of a pelvic mass: improved algorithms and novel biomarkers

More than 200,000 women undergo exploratory surgery for a pelvic mass in the United States each year and 13%-21% of pelvic lesions are found to be malignant. Individual reports and meta-analysis indicate better outcomes when cancer surgery is performed by gynecologic oncologists. Despite the advantages provided by more thorough staging and cytoreductive surgery, only 30%-50% of women with ovarian cancer are referred to surgeons with specialized training in the United States. Imaging, menopausal status and biomarkers can aid in distinguishing malignant from benign pelvic masses to inform decisions regarding appropriate referral. The risk of malignancy index (RMI) uses ultrasound, menopausal status and CA125 and has been utilized in the United Kingdom for two decades, providing sensitivity that has ranged from 71%-88% and specificity it from 97%-74% for identifying patients with malignant disease. Criteria have been established by the Society of Gynecology Oncology and American College of Obstetrics and Gynecology for referral to a gynecologic oncologist, but these have lower sensitivity and specificity than the RMI. Recently, two new algorithms have been developed to identify women at sufficiently high risk to prompt referral to a specialized surgeon. The OVA1 multivariate index incorporates imaging, menopausal status, CA125 and four other proteomic biomarkers. Use of OVA1 provides 85%-96% sensitivity at 28%-40% specificity depending upon menopausal status. The negative predictive value for women judged to be at low risk is 94%-96%. The risk of malignancy algorithm (ROMA) includes CA125, human epididymal protein 4 and menopausal status, but not imaging results. The ROMA has yielded 93%-94% sensitivity at 75% specificity with a negative predictive value of 93%-98%. In a direct comparison, ROMA has achieved greater sensitivity (94%) than the RMI (75%) at 75% specificity. OVA1 has not been compared directly to ROMA, but is likely to be as sensitive, but substantially less specific. Both algorithms have high negative predictive values 94%-98%. Although a difference in specificity should not affect patient outcomes, it could affect distribution of medical resources.