Proliferation Index in Breast Cancer: Is Digital Imaging Accurate?

Abstract

Recent clinical trials have shown that Ki-67, a marker of proliferation, has important prognostic and predictive significance in breast cancer. Exact proliferation indices provide critical cutoffs for treatment decisions. However, the measurement of Ki-67 has not yet been standardized. We sought to determine whether digital imaging would improve the accuracy of Ki-67 measurement in a cohort of patients from the VAST clinical trial. At our institution, the VAST clinical trial has enrolled 12 patients since 2010 to test valproic acid as a neoadjuvant antitumor agent, and one objective is to determine whether the drug decreases the Ki-67 index of treated tumors. Immunohistochemistry for Ki-67 antigen (MIB-1 clone, DAKO North America, Carpinteria, CA) was performed on unstained slides from formalin-fixed breast tissue. H&E and immunohistochemical slides were digitally scanned with Aperio XT Scanscope (Aperio Technologies, Vista, CA) and quantitatively analyzed with image analysis algorithms (software from Aperio Technologies and Definiens, Parsippany, NJ). By using light microscopy, one pathologist (K.N.) counted 1,000 malignant cells from the same region of tumor that was scanned. An “eyeballing” estimate of MIB-1 staining was performed by another pathologist (R.E.F.). A positive result was reported as a percentage of positively stained malignant cells out of the total examined. The result showed excellent correlation among manual counting, digital analysis, and the eyeballing estimate, with a Cronbach α value of .93. However, eyeballing systematically underestimated the number of MIB-1-positive cells. Digital analysis had lower sensitivity and specificity for detecting negative cells on immunohistochemistry slides compared with manual counting. The sensitivity and specificity improved when H&E slides were analyzed to obtain total cell count. Digital imaging seems to accurately analyze MIB-1 positivity. Future studies will include optimization of image detection parameters and use of various cell-specific markers to improve sensitivity and specificity.