Algorithm Identifies Women at Risk of Ovarian Cancer

August 1, 1996

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.

PHILADELPHIA--New computer software is using an investigationalalgorithm to translate serial CA 125 values and other risk factorsinto a single number showing a postmenopausal women's risk ofdeveloping ovarian cancer, Steven J. Skates, PhD, assistant professorof medicine and biostatistics, Harvard Medical School, said athis 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 22,000postmenopausal women in an ovarian cancer screening study conductedby 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. Thesefigures combine to give a positive predictive value above 10%,which exceeds the minimum requirement for an ovarian cancer screeningtest 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 that12 to 16 of every 100 women identified by the program as highrisk will, in fact, have ovarian cancer. That represents at leasta sixfold improvement over the 2 in 100 women who will have thedisease 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 assayvariability. The algorithm triages women into one of three riskcategories: 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 testin 1 to 6 months; and high-risk women would be referred for furtherdiagnostic evaluation such as ultrasound.

"This study gives us sufficient evidence to conduct a randomizedclinical trial based on the software algorithm," Dr. Skatessaid. "Dr. Jacobs is initiating such a trial in the UnitedKingdom to determine if this type of multimodal, sequential screeningwill reduce the mortality rates of ovarian cancer."

The large UK trial, scheduled to last 7 years and include 120,000healthy postmenopausal women, will employ CA 125 II (Centocor,Malvern, Penn), an improved version of the CA 125 assay.

The CA 125 II assay has substantially less variability than theoriginal CA 125 test, thereby offering potentially higher specificity.The large UK study will also evaluate the usefulness of addinga second tumor marker, OVX-1, to the algorithm.