Article Text

Download PDFPDF
Accuracy of visual assessments of proliferation indices in gastroenteropancreatic neuroendocrine tumours
  1. Helen T M Young1,
  2. Norman J Carr1,
  3. Bryan Green1,
  4. Charles Tilley1,
  5. Vidhi Bhargava1,
  6. Neil Pearce2
  1. 1Cellular Pathology, University Hospital Southampton, Southampton, UK
  2. 2Hepatobiliary and Pancreatic Surgery, University Hospital Southampton, Southampton, UK
  1. Correspondence to Helen T M Young, Department of Cellular Pathology, University Hospital Southampton, Tremona Rd, Southampton, Hampshire, SO16 6YD, UK; htmy{at}


Aims To compare the accuracy of eyeball estimates of the Ki-67 proliferation index (PI) with formal counting of 2000 cells as recommend by the Royal College of Pathologists.

Methods Sections from gastroenteropancreatic neuroendocrine tumours were immunostained for Ki-67. PI was calculated using three methods: (1) a manual tally count of 2000 cells from the area of highest nuclear labelling using a microscope eyepiece graticule; (2) eyeball estimates made by four pathologists within the same area of highest nuclear labelling; and (3) image analysis of microscope photographs taken from this area using the ImageJ ‘cell counter’ tool. ImageJ analysis was considered the gold standard for comparison.

Results Levels of agreement between methods were evaluated using Bland–Altman plots. Agreement between the manual tally and ImageJ assessments was very high at low PIs. Agreement between eyeball assessments and ImageJ analysis varied between pathologists. Where data for low PIs alone were analysed, there was a moderate level of agreement between pathologists’ estimates and the gold standard, but when all data were included, agreement was poor.

Conclusions Manual tally counts of 2000 cells exhibited similar levels of accuracy to the gold standard, especially at low PIs. Eyeball estimates were significantly less accurate than the gold standard. This suggests that tumour grades may be misclassified by eyeballing and that formal tally counting of positive cells produces more reliable results. Further studies are needed to identify accurate clinically appropriate ways of calculating.

  • KI 67

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.