[2137] Interobserver Variability in the Quantification of MIB-1 Labeling Index on Cytologic Samples from Well Differentiated Neuroendocrine Tumors (WDNETs) of the Pancreas (P) and Gastrointestinal Tract (GIT): A Comparative Analysis of Three Methods

Adele D Fung, Cynthia Cohen, Sravankumar Kavuri, Xin Gao, Michelle D Reid. Emory University School of Medicine, Atlanta, GA; Georgia Health Sciences University, Augusta, GA

Background: WDNETs of the P and GIT are classified in part by measuring MIB1 index. We determined the level of interobserver variability among 3 observers in the calculation of MIB1 labeling index by counting positive cells/400 tumor cells and doing eyeball estimation of MIB1 percentage (%) in 22 WDNETs to determine reliability/reproducibility of both tests compared to the automated cellular imaging system (ACIS, Dako).
Design: Twenty-two WDNETs of the P (n=13), GIT (n=4) and 5 liver metastases (2 colonic, 3 from P) were confirmed by cytology and immunohistochemistry. These were stained with MIB1 antibody (1;160 dilution, Dako, Carpinteria, CA). Number of MIB1-positive cells/400 tumor cells was calculated by 2 cytopathlogists (MR, SK) and 1 fellow (AF). In addition, the 3 reviewers did an eyeball estimate of MIB1% and scored cases into 3 categories; < 2%, 3-20% and > 20%. Statistical analysis using Spearman's correlation coefficient for all observers and methods was performed and compared to that obtained by ACIS.
Results: There was strong statistically significant correlation among observers when calculating MIB1-positive cells/400 tumor cells and when doing an eyeball estimate of MIB1%.

MIB1-Positive Cells/400 Tumor Cells
ObserverCC95% CIp-value
AF vs SK0.83(0.63, 1.0)<0.05
AF vs MR0.95(0.88, 1.0)<0.05
SK vs MR0.88(0.73, 1.0)<0.05
"Eyeball" Estimate of MIB-1 Index
ObserverCC95% CIp-value
AF vs SK0.75(0.48, 1.0)<0.05
AF vs MR0.86(0.69, 1.0)<0.05
SK vs MR0.79(0.55, 1.0)<0.05
CC, correlation coefficient; CI, confidence interval.

Additionally when both methods were compared there was also strong (statistically significant) correlation between them. However when ACIS was compared to MIB1-positive cells/400 tumor cells and to eyeball estimation there was a weaker, albeit statistically significant, correlation between the 3 methods.

Comparison Of Methods
MethodCC95% CIp-value
MIB1/400 vs Eyeballing0.78(0.53, 1.0)<0.05
MIB1/400 vs ACIS0.53(0.14, 1.0)<0.05
Eyeballing vs ACIS0.62(0.23, 1.0)<0.05
CC, correlation coefficient; CI, confidence interval..

Conclusions: The counting of MIB1-positive cells/400 tumors cells and eyeball estimation of MIB1 labeling index showed strong interobserver correlation. This would suggest that based on WHO recommendations both of these methods are more reliable and reproducible for grading of WDNETs of P and GIT on cytology samples, than ACIS.
Category: Techniques

Tuesday, March 20, 2012 2:15 PM

Platform Session: Section E, Tuesday Afternoon


Close Window