Reproducibility of the Rapid Bud Count Method for Assessment of Tumour Budding in Stage II Colorectal Cancer
BD Hayes, A Maguire, N Conlon, D Gibbons, LM Wang, K Sheahan. St Vincent's University Hospital and University College Dublin, Dublin, Ireland
Background: We recently described a rapid assessment method for the characterisation of tumour budding in colorectal cancer which was independently associated with a poor outcome (hazard ratio = 4.76). Although highly reproducible in the initial study (Wang et al, Am J Surg Pathol 2009;33:134), the reproducibility of the method between pathologists within an academic department with a training programme has not been evaluated.
Design: A test set of forty stage II colorectal cancer cases (twenty each of low and high budding) was used to assess reproducibility. Each of the four study participants underwent brief training in the rapid bud count method comprising review of a PowerPoint presentation, review of the initial study manuscript and review of sample cases, not included in the study, with one of the authors (KS). Five 200x fields at the invasive tumour front were examined per slide (3-9 slides per case). The total number of fields containing at least one tumour bud (defined as an isolated cluster of less than five cells) was noted. Cases with more than 50% of the fields positive were classed as high budding. Agreement was measured using percentages and free-marginal kappa statistics for multiple raters of dichotomous data.
Results: There was substantial agreement between the pathologists (kappa = 0.617) on the presence of low or high budding. There was total agreement (4:0) in 62.5% of cases (25/40), and total disagreement (2:2) in one case (2.5%). In the remaining 14 cases (35%) there was three-way agreement: ten cases with one pathologist underestimating high budding, and four with overestimation of low budding. The mean deviation in the number of positive fields by the disagreeing pathologist was low and was similar in cases of underestimation (4.1 or 28%) and overestimation (4.125 or 33%). In three of the ten cases of underestimation, the degree of deviation in the number of positive fields was marginal (less than 1, or 4%), such that a single additional positive microscopic field would have classed the case as high budding.
Conclusions: There is substantial reproducibility of the rapid bud count method for assessment of tumour budding in this cohort of pathologists, despite minimal training and experience of the technique. Borderline-low tumour budding cases may benefit from additional studies, either through cytokeratin staining or repeat analysis by a second pathologist.
Tuesday, March 23, 2010 9:30 AM
Poster Session III # 109, Tuesday Morning