Interobserver Variability in Assessing Ki-67 Proliferative Index in Gastrointestinal Well-Differentiated Neuroendocrine Neoplasms
D Dhall, DP Frishberg, G Galliano, L Chiles, F Chung, D Ines, HL Wang. Cedars-Sinai Medical Center, Los Angeles, CA
Background: Gastrointestinal well-differentiated neuroendocrine neoplasms (carcinoid tumors) with similar morphologic features may show variable biological behavior, which may be predicted by Ki-67 proliferative index (>1% Ki-67 associated with worse clinical outcome). Many practicing pathologists just eyeball immunostaining for reporting Ki-67 index, which may not be accurate or reproducible, particularly in tumors with a low proliferative index. Thus, using computer-assisted image analysis to assess Ki-67 index may be more desirable. This study examined interobserver variability in eyeballing Ki-67 proliferative index in carcinoid tumors and compared this method with computer-assisted image analysis.
Design: Sixty-eight Ki-67 immunostained slides from midgut carcinoid tumors were circulated among 5 pathologists for eyeballing. In cases where Ki-67 labeled cells were heterogeneously distributed, hot spots were circled to ensure that the same areas were evaluated. Ki-67 index was categorized as <1%, 1-2%, and numerical value for higher indices. Kappa statistics were calculated using 1% and 2% cutoffs. Computer-assisted image analysis was performed using Ariol system.
Results: With a cut off of 2%, an agreement was achieved by all 5 pathologists in 46 cases (68%) with a k-value of 0.66 (95% confidence interval: 0.59-0.74); indicating a substantial degree of agreement by eyeballing. With a cut off of 1%, the agreement was reduced to 35 cases (52%) and the k-value was in the moderate range (0.52). When individual pathologist performance was compared with computer-assisted image analysis, k-value was much lower (0.39 if the cut off was 2% and 0.36 if 1%). Computer-assisted image analysis generated a higher Ki-67 index in 21% of the cases comparing with individual pathologists, due to counting Ki-67 labeled lymphoid cells within the tumors.
Conclusions: There is a moderate to substantial agreement among pathologists in eyeballing Ki-67 index in carcinoid tumors, when the cut off for Ki-67 was kept as 1% or 2%. There is a stronger agreement among pathologists than between pathologists and computer. Eyeballing is thus a simple and relatively reliable method for assessing Ki-67 proliferative index in most carcinoid tumors. However, in borderline cases, computer-assisted image analysis may be helpful for more accurate counting if the areas examined contain no or only rare lymphoid cells.
Tuesday, March 10, 2009 1:00 PM
Poster Session IV # 93, Tuesday Afternoon