A0595
Title: POI-SIMEX for conditionally Poisson distributed biomarkers from tissue microarrays
Authors: Aijun Yang - University of Victoria (Canada)
Finn Hamilton - BC Cancer (Canada)
Brad Nelson - BC Cancer (Canada)
Julian Lum - BC Cancer (Canada)
Mary Lesperance - University of Victoria (Canada)
Farouk Nathoo - University of Victoria (Canada) [presenting]
Abstract: In regression analysis, covariate measurement error is an important issue that has been studied extensively. The important case is considered, where covariates are derived from tissue microarrays. In such settings, biomarkers are obtained from tissue cores that are subsampled from a larger tissue area so that these biomarkers are only estimates of the true cell densities. The resulting measurement error is non-negligible but is rarely considered in cancer studies involving tissue microarrays. These discrete biomarkers are assumed to be conditionally Poisson distributed based on a Poisson process governing the spatial locations of marker-positive cells. SIMEX is extended to the conditional Poisson case (POI-SIMEX), where measurement errors are non-Gaussian with heteroscedastic variance. The resulting POI-SIMEX estimator is shown to be strongly consistent in a linear regression model under assumptions that include a conditional Poisson distribution for the biomarker. POI-SIMEX is applied to a study of high-grade serous ovarian cancer, examining the association between survival and the presence of Tregs CD3/CD8/FOXP3 in epithelial tissue.
