QLS Seminar - Jaya Satagopan

Evaluation of interactions in clinical and molecular epidemiology studies

Jaya Satagopan, Rutgers University
Tuesday November 5, 12-1pm
McIntyre Medical Building, room 1034

Abstract: There is a long-standing interest in the evaluation of interactions between genetic and environmental factors in molecular epidemiology studies. Recent tumor genomics investigations have inspired evaluations of gene x treatment interactions in clinical studies. The interest in interactions is stimulated by the hope that significant interaction effects identified using statistical models can provide insights into optimal screening or preventive interventions, treatment options or patient selection into clinical trials. Interactions between risk factors are typically examined using a product term of the risk factors in statistical models. This talk presents the interpretation of such statistical interaction terms to highlight that certain interactions may be required in statistical models to obtain a good fit to the data because of the choice of the scale of the outcome. Therefore, a pragmatic approach to evaluate interactions would be to examine an interaction contrast on a clinically desired scale of the outcome. Some useful clinical scales of outcome will be discussed. Data from published cancer epidemiology studies of bladder cancer and advanced colorectal adenoma and a collaborative study of nevi in children will be used to illustrate interactions for binary and count outcomes. Since data from large clinical studies of treatment and biomarkers may not always be available, digitally extracted data from published phase III clinical trials of metastatic melanoma and metastatic colorectal cancer will be used for time-to-event outcomes.