Comparison of visual assessment and image analysis in the evaluation of Ki-67 expression and their prognostic significance in immunohistochemically defined luminal breast carcinoma

Abstract

To compare the Ki-67 labeling index value obtained through immunohistochemistry analysis by human examiners to that obtained from computer-assisted image analysis, and to establish a cut-off value for Ki-67 labeling index for each method in luminal B breast carcinoma. Immunohistochemistry analysis for Ki-67 was performed on the formalin-fixed, paraffin-embedded tissue samples from 403 patients with primary luminal breast cancers. Whole slide images were obtained using the NanoZoomer (Hamamatsu Photonics, Hamamatsu, Japan) and thoroughly analyzed using the Definiens Tissue Studio version 1.1 (Definiens AG, Munich, Germany) to detect the percentage of positively-stained nuclei of carcinoma cells. Although a significant correlation was found between the Ki-67 labeling index obtained by manual assessment and computer-assisted image analysis (Spearman rank correlation coefficient, P<0.01), the Ki-67 labeling index value obtained by manual assessment was significantly higher than that obtained by computer-assisted image analysis (Wilcoxon signed rank test, P<0.0001). Disease-free survival was significantly lower in 403 patients with tumors having high Ki-67 labeling index values determined by automated analysis (cut-off value: 11.5%; P<0.00001) and visual counting (cut-off value: 28.5%; P<0.00001). Disease-free survival was also significantly lower in 288 patients who received adjuvant endocrine therapy alone having high Ki-67labeling index values determined by automated analysis (cut-off value: 11.5%; P<0.0001) and visual counting (cut-off value: 19.7%, P<0. 0001). The Ki-67 labeling index values determined by automated analysis and visual counting could equally predict disease-free survival in patients with luminal B breast carcinoma, including those who received endocrine therapy.