The effect of using a 30% cutoff for the proportion of HER2+ cells on the interobserver variability in the interpretation of HER2 immunohistochemical results was evaluated. Immunostained sections from 96 cases of breast carcinoma were reviewed by 10 pathologists and scored as positive (3+) when uniform strong membranous staining was identified in at least 10% of tumor cells; the actual percentage of cells with such staining was also estimated. The agreement rates and the kappa values using a 30% cutoff were compared with those using a 10% cutoff. These proved to be higher in 62% and 66% of measurements, respectively, with average interobserver rates and kappa values of 72% and 0.54 using the 30% cutoff and 70% and 0.49 using the 10% cutoff (P=.001 for all comparisons). Using a 30% cutoff for the percentage of HER2+ cells by immunohistochemical analysis modestly decreased interobserver variability in the interpretation of HER2 immunohistochemical results.