Predicting the clinical behaviour of gastrointestinal stromal tumours (GISTs) is difficult and criteria delineating benign from malignant cases are not firmly established. The aims of this study were to define the clinicopathological and molecular features of 66 GISTs, and to determine whether any specific parameters were associated with patient outcome. Archival cases of GIST from two major teaching hospitals in Western Australia were studied. Inclusion criteria for the study were: (1) appropriate morphology, (2) CD117 positivity, (3) adequacy of pathological material for study, and (4) exclusion of other tumour types on the basis of immunophenotypic and/or ultrastructural features. Expression of CD117, CD34, S100 protein, keratin (using broad spectrum MNF116), alpha-smooth muscle actin (SMA) was determined by immunohistochemistry. PCR and single strand conformation polymorphism (PCR-SSCP) analysis were used to screen for mutations in exons 11 and 9 of c-kit. There were equal numbers of males and females with a mean age at diagnosis of 60 years, followed up for a mean of 54 months. Thirteen patients (21%) had died of GIST by the end of the study. Tumours were mostly located in the stomach (67%) and small intestine (SI; 25%). The cell types were pure spindle (68%), pure epithelioid (12%) and mixed epithelioid/spindle (20%). c-kit mutations were found in 69% of GISTs, with the large majority (91%) occurring in exon 11. Size > or = 10 cm, tumour necrosis and pure epithelioid cell morphology each were the only factors significantly associated with adverse survival (p=0.038, and p=0.047 and p=0.028, respectively). Mitotic activity > or = 5/50 HPF showed a definite trend association with adverse survival, but unlike some other studies, did not achieve statistical significance (p=0.067). c-kit mutations were more frequent in small intestinal GISTs (p=0.05) and in those with pure spindle cell morphology (p=0.023) but were not associated with patient outcome. In this study, size > or = 10cm, necrosis and/or pure epithelioid cell morphology correlated significantly with adverse survival. Mitotic activity showed a strong association with survival but this did not reach statistical significance. c-kit mutations occurred mainly in GISTs of the SI, and in purely spindle cell tumours. While the mutation status did not associate with patient outcome in this series, this remains a controversial issue, and further studies are needed to assess whether the type of mutation affects response to tyrosine kinase inhibitor therapy in metastatic GISTs. CD117 staining of any mesenchymal lesion of the gastrointestinal tract should be mandatory for accurate classification. PCR-SSCP analysis is a fast, sensitive and relatively inexpensive method of analysing c-kit mutations, which may be important prognostically and also of therapeutic relevance in the assessment of new tyrosine kinase inhibitor therapies.