Aim a standardized postoperative score, the dulk (dutch leakage) score, has been demonstrated to be a useful clinical tool in the diagnosis of anastomotic leakage. It is complicated, however, and a simplification (the modified dulk score) based on fewer parameters derived from multiple logistic regression analyses has been developed. These include clinical condition, abdominal pain not localized at the wound, c-reactive protein level and respiratory rate. The accuracy of each was compared. Method data of all patients from five dutch centres operated on between 16 october 2007 and 1 november 2009 with an anastomosis in the colon or rectum were entered into a prospectively maintained database. Results in total, 782 patients were included of whom 81 (10.4%) had a clinically relevant anastomotic leakage. The dulk score gave an overall sensitivity of 97% for anastomotic leakage, overall specificity of 53%, a positive predictive value (ppv) of 16% and a negative predictive value (npv) of 99%. The modified dulk score used clinical condition, abdominal pain not localized at the wound, c-reactive protein level and respiratory rate. With at least one parameter present, overall sensitivity was 97%, overall specificity 57%, ppv 17% and npv 99.5%. With at least two points ppv was 41% and with three points 57%. Conclusion both the original and modified dulk scores are useful for the early diagnosis of clinically relevant anastomotic leakage. The modified dulk score offers the benefit of fewer parameters and so can easily be used in a clinical environment to estimate the likelihood of anastomotic leakage. However, the early diagnosis of anastomotic leakage remains difficult.