Integration of survival data from multiple studies
Steffen Ventz, PhD
Assistant Professor of Biostatistics | Division of Biostatistics
School of Public Health | University of Minnesota
Where: Hybrid Event | 2001 McGill College, Room 1140; Zoom
Biomedical technologies enable the use of omics information for prognostic purposes, to quantify the risk of diseases or to predict response to treatments. Risk stratification in oncology often utilizes a set of biomarkers to predict cancer progression or death within a time period. The number of covariates can often exceed the sample size, which makes the identification of relevant genomic features for risk prediction and the development of accurate models challenging. In this talk I introduce a statistical procedure that integrates datasets from multiple biomedical studies to predict patients’ survival, based on individual clinical and genomic profiles. The procedure accounts for potential differences in the relation between predictors and outcomes across studies, due to distinct patient populations, treatments, and technologies to measure outcomes and biomarkers. These differences are modeled explicitly with study-specific parameters. We use hierarchical regularization to shrink study-specific parameters towards each other and to borrow information across studies. The estimation of the study-specific parameters utilizes a similarity matrix, which summarizes differences and similarities of the relations between covariates and outcomes across studies. We illustrate the method in simulation studies and using a collection of gene expression datasets in ovarian cancer. We show that the proposed model increases the accuracy of survival predictions compared to alternative meta-analytic methods.