Screening Based on the Risk of Cancer Calculation From Bayesian Hierarchical Changepoint and Mixture Models of Longitudinal Markers

• Published in: Journal of the American Statistical Association Vol. 96; no. 454; pp. 429 - 439
• Main Authors: Skates, Steven J, Pauler, Donna K, Jacobs, Ian J
• Format: Journal Article
• Language: English
• Published: Alexandria, VA Taylor & Francis 01.06.2001; American Statistical Association; Taylor & Francis Ltd
• Subjects: Applications; Applications and Case Studies; Bayesian analysis; Biological markers; Cancer; Cancer screening; Disease; Disease models; Disease risk; Exact sciences and technology; Longitudinal CA125; Markov chain Monte Carlo; Markovian processes; Mathematical models; Mathematics; Medical sciences; Medical screening; Medical treatment; Mixtures; Multivariate analysis; Ovarian cancer; Ovarian diseases; Parametric inference; Probability and statistics; Risk; Sciences and techniques of general use; Screening; Screening tests; Statistical models; Statistics; Tumors; Ultrasonography; Biometrics; Parameter estimation; Risk; Multivariate analysis; Markov chain; Posterior distribution; Bayes estimation; Mixed distribution; Monte Carlo method; Data analysis; Factor analysis; Prior distribution; Decision making; Statistical estimation; Marker; Markov model; Malignant tumor; Mean estimation; Cancer screening; Statistical method; Screening; Ovarian diseases; Sensitivity; Posterior probability; CA 125 antigen; Hierarchical model; Hierarchical system; Application; Principal component analysis
• ISSN: 0162-1459; 1537-274X
• DOI: 10.1198/016214501753168145

The standard approach to early detection of disease with a quantitative marker is to set a population-based fixed reference level for making further individual screening or referral decisions. For many types of disease, including prostate and ovarian cancer, additional information is contained in the subject-specific temporal behavior of the marker, which exhibits a characteristic alteration early in the course of the disease. In this article we derive a Bayesian approach to screening based on calculation of the posterior probability of disease given longitudinal marker levels. The method is motivated by a randomized ovarian cancer screening trial in the United Kingdom comprising 22,000 women screened over 4 years with an additional 5 years of follow-up on average. Levels of the antigen CA125 were recorded annually in the screened arm. CA125 profiles of cases and controls from the U.K. trial are modeled using hierarchical changepoint and mixture models, posterior distributions are calculated using Markov chain Monte Carlo methods, and the model is used to calculate the Bayesian posterior risk of having ovarian cancer given a new subject's single or multiple longitudinal CA125 levels. A screening strategy based on the risk calculation is then evaluated using data from an independent screening trial of 5,550 women performed in Sweden. A longitudinal CA125 screening strategy based on calculation of the risk of ovarian cancer is proposed. Simulations of a prospective trial using a strategy based on the risk calculated from longitudinal CA125 values indicate potentially large increases in sensitivity for a given specificity compared to the standard approach based on a fixed CA125 reference level for all subjects.