AIMS: With the availability of effective but expensive treatment in the form of imatinib, accurate diagnosis of gastrointestinal stromal tumour (GIST) is extremely important. The aims of this study were: to describe the clinicopathological, immunohistochemical and molecular features of cases referred to a cancer centre with a possible diagnosis of GIST; to identify pitfalls in the performance and interpretation of KIT immunohistochemistry; to define the role of KIT mutation testing in making a diagnosis of GIST. METHODS AND RESULTS: Morphological review, KIT immunohistochemistry and mutation testing were performed on all cases referred with a diagnosis of GIST or where the diagnosis was under serious consideration on the basis of KIT immunopositivity with a view to treating with imatinib. Thirty-seven cases met the inclusion criteria. Of these, 26 were classified as GIST and 11 as non-GIST. Most GISTs showed strong diffuse membranous, cytoplasmic or paranuclear KIT immunopositivity. Some non-GISTs demonstrated patchy cytoplasmic KIT immunopositivity related to the immunohistochemical protocol used in the external laboratory, which led to erroneous diagnoses of GIST in nine (24%) cases. KIT mutations involving exons 11 or 9 were identified in 22 (88%) GISTs tested and none of the non-GISTs. CONCLUSIONS: An accurate diagnosis of GIST can be made on clinicopathological and immunohistochemical criteria without the need for mutational analysis in most cases, provided proper attention is paid to the immunohistochemical protocol used and, most importantly, control material. False-positive diagnoses of GIST potentially leading to inappropriate treatment with imatinib are more common than missed diagnoses.