Danping Liu

Main Content

Challenges and Methods in Biomarker Combination for Risk Prediction with Application to a Fetal Growth Study
10 December 2015 from 4:00 PM to 5:00 PM
201 Thomas Building
Contact Name
Contact Phone
Add event to calendar

Danping Liu

In disease prediction and diagnosis, the combination of multiple biomarkers often substantially improves the diagnostic accuracy over a single marker. Our work is motivated from the Scandinavian Fetal Growth Study, in which a cohort of pregnant women received multiple ultrasound examinations during pregnancy, but only a subset of the infants received further follow-up after birth. The first part of the talk will focus on using the longitudinal ultrasound measurements to predict the pregnancy outcome. We propose a pattern mixture model (PMM) framework, where the marker distribution given the disease status is estimated from a mixed effects model. A likelihood ratio statistic and the risk scores are computed as the combination rule, which is optimal in the sense of the maximum receiver operating characteristic (ROC) curve under the correctly specified mixed effects model. We further compared the efficiency and robustness of the PMM approach with existing methods. The second part of the talk is devoted to the scenario of predicting overweight infants at the one-year follow-up using the ultrasound measurements, where the disease status could be missing due to loss of follow-up. In estimating the ROC curve, it is well-known that the complete-case analysis often leads to biased estimator, known as “verification bias”. We investigate how verification bias impacts both the selection and combination of biomarkers. Several new approaches are proposed for biomarker combination that can handle missing disease status, based on reweighting and imputation techniques.


Filed under: ,

Navigation for this Section