Personalizing Early Detection for Ovarian Cancer with Hierarchical Longitudinal Models of a Blood Biomarker: Experience in Multiple Prospective Screening Trials

• Activity Number: 216 - Statistical Opportunities in Disease Interception: Screening, Intervening, and Evaluating Benefit-Risk Trade-Offs
• Type: Topic Contributed
• Date/Time: Monday, July 31, 2017 : 2:00 PM to 3:50 PM
• Sponsor: Section on Medical Devices and Diagnostics
• Author(s): Steven Skates*
• Companies: MGH/Biostatistics Center
• Keywords: personalized diagnostics ; early detection ; ovarian cancer ; screening ; change-point models

Abstract

Early detection of ovarian cancer through regularly repeated screening tests offers the hope of mortality reduction. 3/4 of ovarian cancers are detected in late stage when prognosis is very poor, while detection in early stages when disease is limited to the ovaries has excellent prognosis. This contrast in prognosis coupled with the majority of cases detected in late stage makes ovarian cancer an ideal target for disease interception. CA125 is an ovarian cancer biomarker approved for clinical management of ovarian cancer, typically using a cutoff level. Over the past 15 years, a series of screening trials applied personalized CA125 testing for early detection of ovarian cancer, detecting cases with a rising CA125 above each woman's baseline level. The longitudinal model has an individual baseline, followed by a change-point, and then linearly increasing CA125 (log scale), with a hierarchy across cases and controls. Screening decisions such as referral to transvaginal ultrasound were made using the posterior probability of having a change-point. All trials increased the proportion detected in early stages, with some evidence of a mortality reduction in the largest trial.