Software Automated Counting of Ki-67 Proliferation Index Correlates with Pathologic Grade and Disease Progression of Follicular Lymphomas
Mark A Samols, Nathan E Smith, Milena Vuica-Ross, Christopher D Gocke, Kathleen H Burns, Michael J Borowitz, Toby C Cornish, Amy S Duffield. Johns Hopkins Medical Institutions, Baltimore, MD
Background: The WHO classification of follicular lymphoma (FL) divides the disease into low grade and high grade neoplasms; FL grade 1-2 has 0-15 centroblasts/high power field (HPF) within the atypical follicles, while grade 3 has >15 centroblasts/HPF. We examined the accuracy of software assisted measurement of Ki-67 index, and its correlation to the grade and clinical progression of FL.
Design: 32 cases of FL with an associated Ki-67 immunostain were selected from the archives. 18 cases were WHO grade 1-2 and 14 were WHO grade 3A or 3B. The H&E and Ki-67 slides were scanned using Scanscope CS (Aperio Inc., San Diego, CA). Fields were manually selected to cover a representative sample of atypical follicles, and 10 HPF-equivalent fields (550 x 550 micrometers) were extracted from the whole slide images. Aperio's Nuclear Quantitation algorithm was then used to identify and quantify positive and negative Ki-67 nuclei. For validation of automated counts, all 10 fields from three cases representing low, medium and high Ki-67 indices were also manually counted using an Image-J based macro.
Results: Aperio software automated quantitation of Ki-67 proliferation indices showed close positive correlation to the manual Ki-67% (Pearson r=0.99, DF=28, p<2.2e-16). The software consistently counted more positive and negative nuclei with an average of 10.9% more positive nuclei and 4.8% more negative nuclei per field. Ki-67 index was significantly higher in WHO grade 3 FL using a Wilcoxon signed-rank test: FL grade 1-2 showed a median Ki-67 index of 21.9% and FL grade 3 showed 40.3% (p=0.011). Clinical data was available for 26 of 32 cases, with disease progression defined by subsequent treatment with radiation or cytotoxic chemotherapy. The Ki-67 proliferation index was also significantly higher in cases that showed evidence of disease progression (21 patients, median 34.5%), versus those that did not (5 patients, median 10.0%, p=0.017). Higher WHO grades of FL were not associated with disease progression in this data set (p=0.21, Fisher exact test).
Conclusions: Software automated Ki-67% proliferation indices show high accuracy and close correlation to manual counts. A higher Ki-67 proliferation index was associated with both higher grade in FLs and clinical progression of disease, whereas higher grade was not associated with disease progression. This technique may be useful as an objective measure of pathologic grade and prognosis for FL.
Monday, March 19, 2012 1:00 PM
Poster Session II # 216, Monday Afternoon