Flexible Assessment of Skewed Exposure in Case-Control Studies with Case-Specific and Random Pooling

• Activity Number: 190
• Type: Contributed
• Date/Time: Monday, August 1, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Neil J. Perkins*+ and Brian Whitcomb and Robert H. Lyles and Enrique F. Schisterman
• Companies: Eunice Kennedy Shriver National Institute of Child Health and Human Development and University of Massachusetts and Emory University and Eunice Kennedy Shriver National Institute of Child Health and Human Development
• Keywords: gamma distribution ; set-based regression ; pooling ; biomarkers ; efficient design

Abstract

Pooled and hybrid, pooled-unpooled, designs have been proposed for epidemiologic study of biomarkers to minimize cost, while maintaining efficiency as well as several other physical and statistical benefits. Set-based logistic regression has been proposed for use with pooled data; however, analysis has been limited by assumptions regarding exposure distribution and logit-linearity of risk (i.e., constant odds ratio). We have developed a more flexible model for analysis of pooled or hybrid data using characteristics of the gamma distribution. A modified logistic regression is used to accommodate non-linearity corresponding to removal of the restriction of equal shape parameters in the gamma distributed exposure for cases and controls. Full maximum likelihood estimation is compared to a more standard logistic regression approach via simulation to assess consistency and efficiency of risk effect estimates given random and disease specific pooled, hybrid data. Our methods are demonstrated through effect estimates of pooled cytokines data on perinatal outcomes.