Online Program

Return to main conference page

Friday, January 12
Fri, Jan 12, 10:30 AM - 12:15 PM
Crystal Ballroom A
Statistical Methods for Cancer Screening Using Risk Prediction Models, Polygenic Risk Scores, and Simulations

Simulation modeling of prostate cancer screening - a powerful tool for policy development and medical decision making (304235)

*Ruth Etzioni, Fred Hutchinson Cancer Research Center 
Roman Gulati, Fred Hutchinson Cancer Research Center 
Jane Lange, Fred Hutchinson Cancer Research Center 
Lurdes Inoue, University of Washington 
Alex Tsodikov, University of Washington 
Eveline Heijnsdijk, Erasmus University 
Harry de Koning, Erasmus University 

Keywords: Simulation modeling, prostate cancer screening, model calibration, racial disparities, personalized medicine, randomized trials, cancer surveillance

PSA screening for prostate cancer has been a part of clinical practice for 30 years, yet many critical questions remain. Many questions cannot be addressed by randomized studies and evidence gaps persist. I will present some recent applications of a simulation model of prostate cancer screening that we have developed with the support of NCI’s Cancer Intervention and Surveillance Modeling Network. After reviewing the statistical basis for the model, I will show how we have used it to (a) reconcile apparently conflicting results of the two largest prostate cancer screening trials; (b) develop recommendations for how screening should be different in high-risk populations like black men and (c) project risks of overdiagnosis that are personalized according to age, grade, PSA, and comorbidity at diagnosis. I will show that the screening trials are not as different as they seem and combine to indicate a strongly beneficial screening test. Further, it makes sense to begin screening African-American men up to ten years earlier than the general population. Finally, the chance of being overdiagnosed can range from 3 to 80 percent depending on a patient’s characteristics at diagnosis.