Interobserver reproducibility of immunohistochemical HER-2/neu evaluation in human breast cancer: the real-world experience.

In spite of the large number of studies on technical problems affecting the interlaboratory reproducibility of IHC HER-2/neu determination, only little is known about factors limiting the intra- and interobserver reproducibility in the actual practice of HER-2/neu expression analysis. The aim of the present INQAT study was to evaluate the intra- and inter-observer reproducibility of IHC HER-2 analysis among pathologists routinely working in Italian laboratories. Twenty immunostained slides were distributed to 12 pathologists who had to report, for each slide, the semiquantitative analysis of the percentage of immunopositive cells and the qualitative evaluation of the intensity of membrane staining. The intra- and interobserver reproducibility as well as the reproducibility between each laboratory and the reference values were quantified adopting an approach based on computation of the weighted kappa statistic (Kw). Additionally, in order to evaluate the contribution of each category to the overall agreement, the kappa category-specific statistics (Kcs) were estimated for both classification criteria by jointly considering all the participating laboratories. The intraobserver analyses showed a satisfactory level of reproducibility for both the percentage of positive cells (median Kw, 0.94; range: 0.80-0.96) and membrane staining (median Kw, 0.86; range: 0.78-0.96). Similarly, a fairly good level of reproducibility for the percentage of cells (median Kw, 0.89; range, 0.73-0.96) and the intensity of membrane staining (median Kw, 0.84; range, 0.72-0.92) were observed from comparisons with reference values. When all possible pairwise comparisons were performed, a satisfactory level of interobserver reproducibility was found for most laboratories. Kw values varied between 0.51 and 0.98 (median Kw, 0.80) and between 0.61 and 0.94 (median Kw, 0.81) for semiquantitative and qualitative measurements, respectively. Analysis of the contribution of the extreme categories to the overall agreement showed a substantial or almost perfect agreement for both classification criteria. Conversely, the contribution of intermediate categories appeared to be scarce or slight for the percentage of immunostained cells and slight or fair for the intensity of membrane staining. We conclude that, overall, the interobserver reproducibility in IHC analysis of HER-2/neu expression is satisfactory, although classification of the intermediate categories is problematic, both with regard to the percentage of immunostained cells and the intensity of membrane staining.