Quantification of Variation in Reading Immunohistochemical Assays

Abstract

Semiquantitative immunohistochemical assays have been used with increasing frequency. This study was designed to investigate the reproducibility of such measurements by observing how the measured level varied du.e to (a) the choice of tissue sample from a single or multiple tumors, (b) the immunohistochemical procedure and the influence of time on staining and (c) the subjective variability between readers (interobserver) and by the same reader (intra-observer). The study was meant to judge the reproducibility of the method, not its accuracy. The choice of the monoclonal antibody therefore did not influence the results. A total of 128 sets of sljdes from 8 colonic adenocarcinomas were analyzed by three pathologists using a randomized, symmetric, prospective, doubleblind study. There was surprisingly poor agreement between readings of the same case by the three pathologists (37%) and by the same pathologist over time (58%). Based on the component of variation analysis, 11% of the total variation was due to differences in the immunohistochemical procedure, 5% to variation of expression in different tumors, 5% to interobserver and 79% to intra-observer variability. Readings of semiquantitative immunohistochemical assays is limited by subjective intrinsic variability.