[1839] Comparative Analysis of Different Counting Methodologies for Ki-67 in Pancreatic Neuroendocrine Tumors
Pelin Bagci, Nobuyuki Ohike, Nevra Dursun, Kee-Taek Jang, Takuma Tajiri, Olca Basturk, So Yeon Kong, Leslie Ducato, Michelle Reid, Volkan Adsay. Emory University, Atlanta; Showa University, Tokyo, Japan; MSKCC, New York
Background: Ki-67 labeling index is now one of the 2 parameters in the classification of pancreatic NETs in the WHO-2010 classification. However, in adapting this into daily practice, it becomes clear that there are serious challenges in the counting methodologies.
Design: Ki-67 immunohistochemical staining was performed on full sections from 21 NETs, and percentage of the tumor cells stained was counted by 4 different methods by 3 observers: 1) “Eye-balling”, 2) "Eye-counting" through the microscope, 3) Automated counting, 4)Manual counting on camera-captured/printed image of the hot spot.
Results: Considering the balance of accuracy, practicality, and reproducibility, the method that was found to be by far the most preferable (Table) was the manual counting on camera-captured/printed image.
| Impact on turnaround time | Average time it takes (minutes) | Practicality | Accuracy | Miscounting of non-target cells | Interobserver agreement (Pearson's correlation) | Additional Cost | |
| Eye-balling | None | 1 | Highest | Very low | Unlikely | R=23% | None |
| Eye-counting through microscope | None | 6 | Low | High | Unlikely | R=32% | None |
| Manual counting on camera captured/printed image of the hot spot | Minimal (depending on accessibility of a camera/printer setup) | 13 | Very high | Highest | Unlikely | R=43% | Printer + camera $5,400 |
| Automated | Highest (depends on the technician availability) | 5 | Low (accessibility issues) | Moderate | Very likely | N/A | Image analyzer $150,000 |