Abstract P3-05-09: Interlaboratory variability of Ki67 labeling index in breast cancer tissue microarrays

Abstract

Abstract Objective: Assessment of tumor proliferation using the Ki67 labeling index (Ki67-LI) is increasingly recommended for prognostication and adjuvant chemotherapy decisions in breast cancer. Our aim was to investigate interlaboratory variance of Ki67-LI results using TMA and centralised assessment to exclude preanalytic influences and postanalytic variance, respectively. Methods: Nine pathology laboratories (8 German and 1 Dutch) performed Ki67 staining of a TMA slide according to their routine in house protocol (including internal and external quality assurance). 40 samples per lab were centrally analyzed. The Ki67-LI was calculated after counting first all tumor cells and subsequently all Ki67 positive tumor cells of each sample regardless of staining intensity. For each tissue sample we evaluated the range of Ki67-LIs between different labs. Further 20 labs of the German Breast Screening Pathology Initiative will participate in 2013. Results: The range of Ki67-LIs between the labs was: 1-5% in 4 (10%), 6-10% in 6 (15%), 11-15% in 7 (17,5%), 16-20% in 2 (5%), 21-25% in 7 (17%), 26-30% in 4 (10%), 31-35% in 1 (1%), 36-40% in 4 (10%), 41-45% in 2 (5%), 46-50 in 0 (0%) and >50% in 3 samples (7%), respectively. Thus, in 35% of results of the 9 labs (26 of the 40 TMA specimens) the Ki67-LI differed by more than 25%. The analysis of variance (ANOVA) came out with F = 4.24 which is much larger than the critical F value of 1.97 for these study results (p = 7.74902E-05). This means that the observed interlaboratory variance of 4.24 is systematic and not due to sampling error. The respective standard deviation is 33.4%. Conclusion: In a setting strictly standardised in terms of preanalytic influences by using TMA and postanalytic variance by centralised quantification, Ki67-LI seems to be heavily influenced by laboratory-specific analytic variables. Taking this into consideration, there may be a risk of prognostic or predictive misclassification in some breast cancer patients in daily practice. Citation Information: Cancer Res 2013;73(24 Suppl): Abstract nr P3-05-09.