PHILADELPHIA--New computer software is using an investigational algorithm to translate serial CA 125 values and other risk factors into a single number showing a postmenopausal women's risk of developing ovarian cancer, Steven J. Skates, PhD, assistant professor of medicine and biostatistics, Harvard Medical School, said at his American Society of Clinical Oncology poster presentation.
22,000 Postmenopausal Women
Dr. Skates and his colleagues applied the Risk of Ovarian Cancer (ROC) algorithm to blood samples collected from more than 22,000 postmenopausal women in an ovarian cancer screening study conducted by Ian Jacobs, MD, of St. Bartholomew's Hospital, London.
The results showed a sensitivity of 86% and specificity of 99.7% for the ROC algorithm in identifying women at high risk. These figures combine to give a positive predictive value above 10%, which exceeds the minimum requirement for an ovarian cancer screening test as suggested in the medical literature, he said.
Dr. Skates, who developed the algorithm along with Robert C. Knapp, MD, professor emeritus, Harvard Medical School, estimates that 12 to 16 of every 100 women identified by the program as high risk will, in fact, have ovarian cancer. That represents at least a sixfold improvement over the 2 in 100 women who will have the disease when identified using only a single elevated CA 125 level.
To calculate risk, the algorithm uses serial CA 125 assay values, changes in those levels over time, the woman's age, and assay variability. The algorithm triages women into one of three risk categories: normal (ROC level of 0.05% and below); intermediate (0.05% to 4%); and elevated (above 4%).
Those in the normal category would continue annual CA 125 screening; intermediate scores would indicate the need for a repeat test in 1 to 6 months; and high-risk women would be referred for further diagnostic evaluation such as ultrasound.
