Using a higher cutoff for the percentage of HER2+ cells decreases interobserver variability in the interpretation of HER2 immunohistochemical analysis.