Practical issues concerning the implementation of Ki-67 proliferative index measurement in breast cancer reporting

Commercial molecular tests which rely heavily on proliferation markers to stratify breast cancer are in increasing demand, but are expensive and not widely available. There is heightened interest in the use of Ki-67 immunohistochemistry as a marker of proliferation. This study sought to examine practical issues in the incorporation of Ki-67 measurement into breast cancer reporting. We conducted a prospective study of Ki-67 proliferative activity in 85 breast carcinomas in 70 patients. We considered whether dual staining with cyto-keratin and Ki-67 was necessary to exclude background cells in automated digital image analysis (DIA) and how well a semi-quantitative assessment (SQA) method of Ki-67 proliferation and formal manual counting by two pathologists correlated with DIA. Our study showed good correlation between single and dual stained specimens by DIA (Spearman correlation coefficient 0.8), with a kappa statistic of 0.51 (moderate agreement) but with significantly fewer positive cells identified in dual stained sections. There was fair correlation between SQA and DIA by two pathologists (Spearman correlation coefficient 0.7 and 0.7). Using a ≥10% cut-off to define cases with a 'low' and 'high' proliferative index gave a kappa statistic of 0.25 and 0.32 (fair agreement). There was fair correlation between formal manual counts between two pathologists (Spearman correlation coefficient 0.7; kappa 0.32). Repeat DIA on all cases showed excellent correlation (Spearman coefficient 0.98; kappa 1.0). Automated digital analysis of Ki-67 PI is likely to be more accurate and consistent than semi-quantitative assessment and more practicable than formal manual counting. There remain challenges in standardisation of technique within and across laboratories, interpretation of results and in evaluating clinical relevance.