INTRODUCTION This study investigated the accuracy of general practitioner (GP) referrals under the two-week wait pathway for soft tissue sarcomas and whether the current National Institute for Health and Care Excellence criteria should be refined. METHODS All patients referred under the two-week wait system to one centre over the course of one year were reviewed. Comparison was made between the criteria identified by the GP and those confirmed by the centre to assess the accuracy of the referrals, and to identify what criteria predicted malignancy. RESULTS Overall, 135 patients were referred to our unit with a mean age of 56.4 years. Of these, 45 (33%) were found to have a malignant tumour. Factors identified by the GP were accurate in 74% of cases. The best predictor of malignancy was 'size >5cm' (76% sensitivity) while 'pain' was the least useful (27% sensitivity). Lowering the threshold for concern to a size of >4cm increased sensitivity to 89%. Although 106 patients had undergone some form of imaging prior to referral, this did not increase the likelihood of malignancy being detected. The combination of factors most likely to predict malignancy was a size of >5cm, increase in size, deep location and no pain (10 out of 13 referrals, 77% accuracy). CONCLUSIONS Based on the results of this study, we recommend an adaption of the existing features for concern. The new feature for concern should be 'size >4cm' and the factor 'pain' should be removed from the urgent referral form.