Challenges in quantifying the Ki67 proliferation index in well-differentiated gastroenteropancreatic neuroendocrine tumours: assessment of a standardised manual technique

Abstract

<h3>Aims</h3> Treatment decisions for gastroenteropancreatic neuroendo-crine tumours (GEP-NETs) are currently based on grading systems that incorporate the Ki67 proliferation index (PI). The reported Ki67 PI is often an ‘eyeball' estimation that is prone to inter-observer variability. We trialled the use of a standardised method of manual Ki67 quantification. <h3>Methods</h3> The Miller Square is an ocular graticule that is typically used in haematology to facilitate the rapid, accurate and reproducible counting of reticulocytes. We used this method to calculate the Ki67 PI in a retrospective review of thirty consecutive GEP-NETs. <h3>Results</h3> The Miller Square method produced high inter-observer agreement, with an intra-class correlation coefficient (ICC) of 0.99 (95%CI 0.98–0.99). When restricted to grade 1 and 2 tumours, there was only moderate inter-observer agreement, ICC 0.49 (95%CI 0.14–0.73). Using this method, five (18%) of 27 grade 1 and 2 tumours were classified differently by the two observers. Three (11%) of the tumours classified as grade 1 by eyeball estimation were upgraded by the Miller technique. <h3>Discussion</h3> Precise quantification of the Ki67 PI at low values to distinguish grade 1 and grade 2 tumours is challenging, both by eyeball estimation and when using a standardised counting technique. These results highlight the practical difficulties of current grading systems.