Assessing Inter-Observer Agreement for Multivariate IHC Scoring Data

Purdue University

Thursday, September 4, 2014 - 3:30pm

Immunohistochemical (IHC) staining is widely used in the diagnosis of cancer. Current methods for IHC scoring rely on a number of subjective evaluations.  A pathologist, for example, has to assess the overall level of staining of the tumor as negative, weak, intermediate, or strong, and then make a secondary subjective assessment of the percentage of cells that fall into each of these staining intensity categories.  Current inter- and intra-rater agreement methods do not account for the multivariate nature of the scoring and therefore can be misleading when it comes to the development of new IHC assays, training of pathologists, and calculating sample size and power for analytical validation studies designed to fulfill FDA requirements.  In this talk, I will discuss an approach to assess agreement based on these pairs of observed multivariate scores.  We estimate rater-specific cutpoints within the spectrum of staining intensities, accounting for intra-rater variation, and evaluate the shifts in these cutpoints as a measure of inter-rater agreement.

Bruce A Craig, Ningning Chen, Sameera Wijayawardana

Bruce A Craig is Professor of Statistics and Director of the Statistical Consulting Service (SCS) at Purdue University.  He received his M.S. and Ph.D. in statistics from the University of Wisconsin-Madison.  His research interests focus on the development of novel statistical methodology to address problems, primarily in the life sciences.  Areas of current interest are protein structure determination, diagnostic testing, and non-invasive capture/recapture modeling.

More information about Bruce Craig may be found at

This Colloquium is sponsored jointly by the University of Georgia Department of Statistics and the University of Georgia Department of Epidemiology and Biostatistics.