1. ¹Department of Pathology, McGill University, Montreal, QC, Canada
2. ²Department of Mathematics and Statistics, McGill University, Montreal, QC, Canada
3. ³Department of Pathology, McGill University Health Centre, Montreal, QC, Canada
4. ⁴National Cancer Institute, Office of Biorepositories and Biospecimen Research, Bethesda, MD, USA

Correspondence: I Zlobec, Department of Pathology, McGill University, 3775 University Street, Room B22, Montreal, QC, Canada H3A 2B4. E-mail: inti.zlobec@mail.mcgill.ca

Received 31 January 2006; Revised 3 May 2006; Accepted 4 May 2006; Published online 2 June 2006.

Abstract

Molecular tumor markers are often studied in colorectal cancer using immunohistochemistry to determine their prognostic or predictive value. Protein expression is typically assigned a 'positive' score based on a predetermined cutoff. A semiquantitative scoring method that evaluates the percentage of positive tumor cells (0–100%) may provide a better understanding of the prognostic or predictive significance of these markers. The aim of this study was to assess and compare the interobserver agreement of immunohistochemistry scores using a percentage scoring method and three categorical scoring systems. Immunohistochemistry for p53, Bcl-2, vascular endothelial growth factor (VEGF) and apoptotic protease activating factor-1 (APAF-1) was performed on 87 tumor biopsies from patients with rectal carcinoma and scored independently by four pathologists as the percentage of positive tumor cells. Interobserver agreement was assessed by the intraclass correlation coefficient. The intraclass correlation coefficients for p53 and VEGF (>0.6) indicate substantial agreement between observers. The distribution of Bcl-2 and APAF-1 scores in addition to weaker interobserver agreement by percentage scoring suggest that this approach may not be appropriate for these proteins. In conclusion, p53 and VEGF protein expression assessed by immunohistochemistry in colorectal cancer and scored as a percentage of positive tumor cells may be a viable alternative scoring method.