16. Assessment of ki67 proliferative activity in breast carcinoma: a comparison of rapid manual assessment with computerised image analysis

Abstract

Background There has been increasing interest in recent years in molecular assessment of proliferative activity (PA) as a means of stratifying prognosis in some subsets of invasive breast carcinoma. In addition to commercial molecular risk stratification tools, which are strongly influenced by proliferative markers, recent consensus guidelines have suggested incorporation of Ki67 assessment of PA by immunohistochemistry into treatment selection decisions in some circumstances. Despite this, there are significant practical obstacles to measurement of PA outside a research setting. In particular, recommended manual methods involve counting a minimum of 1000 cells, which is likely to be too time consuming for ready incorporation into a busy clinical workload. Aims and Methods We conducted a prospective study of 70 cases in routine practice to compare two less time consuming methods of Ki67 assessment: a rapid semi-quantitative manual method and computerised image analysis. The latter has previously been shown to be strongly correlated with formal manual assessment. We also sought to determine whether dual staining with cytokera-tin and Ki67 was necessary to exclude background cells from automated proliferation assessment. Results We found that rapid semi-quantitative manual assessment showed only a fair correlation with automated counting (correlation coefficient 0.7). There was a strong correlation between single stained and dual stained specimens by automated assessment (correlation coefficient 0.9). Conclusions Our results suggest that computerised image analysis may be the best alternative to formal manual Ki67 counting when that method is impractical in busy clinical practice.