Interpathologists discrepancies in Ki67 assessment in the PACS01 trial: An independent prognosis factor

• Published in: Journal of clinical oncology Vol. 30; no. 15_suppl; p. 543
• Main Authors: Penault-Llorca, Frederique Madeleine, Goubar, Aicha, Raoelfils, Ines, Sagan, Christine, Lacroix-Triki, Magali, Denoux, Yves, Verrielle, Veronique, Jacquemier, Jocelyne, Martin, Anne-Laure, Roche, Henri, Andre, Fabrice
• Format: Journal Article
• Language: English
• Published: 20.05.2012
• ISSN: 0732-183X; 1527-7755
• DOI: 10.1200/jco.2012.30.15_suppl.543

Abstract only 543 Background: Several reports suggest an inter-observer variability in Ki67 assessment. Nevertheless, there is no large study that evaluates the rate of discrepancies, together with their impact. Methods: Ki67 expression was assessed on 663 samples from patients with ER-positive breast cancers included in the PACS01 trial (Roche, J Clin Oncol, 2006). Ki67 staining was done using MiB1 antibody (Dako, Copenhagen, Denmark, Dilution 1:250). Prognostic and predictive values have been reported previously (Penault-Llorca, J Clin Oncol, 2009). A second central review was done by a senior breast pathologist from a French Cancer center. A discrepancy was defined as either a false positive or false negative result. Cut-off for positivity was defined at 15% according to data from Cheang et al (JNCI, 2009). Results: The rate of discrepancy was correlated with the percentage of stained tumor cells. A 10% discrepancy rate between the 2 pathologists was observed when the first pathologist reported <10% tumor cells stained. Same low rate of discrepancy (10%) was observed if more than 30% cancer cells were stained according to the first assessment. At the opposite, discrepancy rates were 47%, 45%, 22%, 34% when the first pathologist reported 10-15%, 15-20%, 20-25% and 25-30% tumor cells stained. Overall, 36% of the patients presented a grade II tumor together with Ki67 <10% or >30%. We then evaluated the impact of discrepancy in terms of prognosis. Patients presenting a concordant result between the two pathologists showed a better outcome as compared to patients presenting a discrepancy, independently to the percentage of tumor stained (p=0.05, patients with concordant Ki67 ranged between 10-30% versus those with one reader <10% and the other one >10%-<30%). Conclusions: Discrepancy rates between pathologists are acceptable when Ki67 is either <10% or >30%. A Ki67 ranged between 10 and 30% could define a grey zone in which Ki67 should be reported with caution or be double checked by another pathologist. Survival analysis suggested that inter-observer discrepancies could act a prognostic factor. Such finding could reflect underlying intra-tumor heterogeneity.