KI-67 PROLIFERATION INDEX IN NEUROENDOCRINE TUMORS OF GI SYSTEM ; A COMPARISON OF TWO COUNTING METHODOLOGIES.

Background: The Ki67 index is important for grading neuroendocrine neoplasms. However, different counting methods exists:(1) 'eye-ball' estimation (2) automated counting by image analyzer (3) manual eye-counting (eye under microscope without a grid (4) manual count of digital camera captured/ colour printed image. The objective of the current study was to compare traditionally followed; Manual counting of Ki-67 positive cells under microscope and recently advocated; Manual counting of Ki-67 positive cells in digital camera captured and colour printed image. Methods: Ki67 immunostaining was performed on 62 resected neuroendocrine tumnors of GI system. Two methods of quantication were employed : i. Manual counting of Ki-67 positive cells in 500 tumor cells by the reporting histopathologist under microscope, usingx20 magnication in tumor 'hotspots'. Only strong nuclear staining was considered to be positive for Ki-67. Light brown or pale staining were ignored (M1). ii. Manual counting of Ki-67 positive cells in 500 tumor cells by the reporting histopathologist using x20 magnication in digital camera captured and colour printed image in tumor'hotspots'(M2). Comparison was based on multiple parameters viz cost effectiveness, reproducibility, practical ease of counting, time required for counting, limitations of individual methods and areas of discordance without compromising in therapeutic and prognostic efcacy. Results: When M1 and M2 grades were compared, 95.16% cases showed agreement and 4.84%cases showed Disagreement. Discordance in Ki67 index was seen near the cut-off values of G1-G2 tumors. Mean time taken for Ki-67 indexing by M1 (36.49 sec) was signicantly less compared to mean time taken by M2 (7.11 min). The τ value obtained by M1 and M2 was 0.811; P <0.0001 and 0.628; P<0.0001 respectively. Near perfect interobserver reliability was obtained in both the methods with ICC of 0.9811 by M1 and ICC of 0.9860 by M2. The κ values for M1 shows moderate level of interobserver agreement (κ = 0.416, 95% CI=0.272-0.559) and κ values for M2 shows substantial agreement(κ = 0.697, 95% CI=0.565-0.828). The CCC for both the methodologies showed nearperfect correlation between the two histopathologists. M2 with CCC of 0.9719 (95%CI=0.9540 to 0.9829) had slightly better correlation than M1 with CCC of 0.9624 (95% CI=0.9409 to 0.9761).The concordance of Ki-67 proliferation index between M1 and M2 was excellent in all cases analysed with ICC of 0.9617 (95% CI: 0.9366 to 0.9769) and a CCC of 0.9252 (95% CI: 0.8789 to 0.9542). The cohen κ value for interobserver agreement between M1 and M2 was 0.891 (95% CI: 0.770 to 1.000) which shows near perfect agreement in grading by both the methodologies. Conclusion: It is important to assess Ki-67 indexing in the “hot-spot” area with the highest proliferative activity with a minimum count of 500 tumor cells. Majority of cases showed agreement in grading by both the methodologies. Disagreement was seen in G1-G2 tumors with Ki-67 values near the grade cut-offs. Ki67 indexing by manual counting of camera captured/color printed image may require additional cost and time but is highly reproducible, reliable and accurate. Ki-67 indexing by manual count under a microscope seems very practical in theory but is not easy to employ as cell distribution and overlapping of cells make the test difcult to perform resulting in interobserver variability.