Evaluation of interobserver agreement in scoring immunohistochemical results of HER-2/neu (c-erbB-2) expression detected by HercepTest, Nichirei polyclonal antibody, CB11 and TAB250 in breast carcinoma.

To examine interobserver reproducibility in the judgments of HER-2/neu (c-erbB-2) oncoprotein expression, HER-2/neu expression was scored immunohistochemically. A total of 106 breast carcinomas were independently scored by five or six pathologists, using four commercially available anti-bodies (HercepTest, Nichirei polyclonal antibody, CB11 and TAB250). Two systems were used for the evaluation. In the Dako system, the intensity of membrane staining was primarily evaluated whereas in the Tohoku system, the area of the membrane-staining-positive cells was primarily evaluated. All four antibodies gave excellent concordance rates among the pathologists' judgments of the scores. Using the HercepTest and Nichirei polyclonal antibody, dividing the cases into two groups designated score 3+ and score 0/1+/2+ achieved a higher percentage of concordant cases than division into groups designated score 2+/3+ and score 0/1+. Using the HercepTest, the Dako system tended to achieve higher concordance than the Tohoku system. Conversely, CB11 and TAB250 showed similar concordance rates under both the Dako and the Tohoku systems, regardless of how the cases were divided. High interobserver agreement levels were achieved only when HER-2/neu overexpression was defined as a score of 3+, using any of the antibodies; these were similar between the Dako system and the Tohoku system. Positivity by TAB250 was lower than that by CB11. Using the HercepTest and Nichirei polyclonal antibody, cases with a score of 3+ almost always achieved complete interobserver agreement and showed HER-2/neu gene amplification. Conversely, cases with a score of 2+ achieved a low percentage of interobserver agreement and showed no gene amplification. Because judgment of a case as score 2+ was not very reproducible, and a score of 3+ was of biological significance and had high interobserver reproducibility, we suggest that only a score of 3+ be considered as overexpression. When cases of score 2+ are entered into a protocol, evaluation should be given by multiple observers and/or by using other tests such as fluorescence in situ hybridization.