OBJECTIVES: to identify risk factors for infrainguinal bypass occlusion and quantify the predictive value of data available before and after surgery. DESIGN: prospective study of 2650 patients who participated in a randomised trial of oral anticoagulants or aspirin after infrainguinal bypass surgery. MATERIALS AND METHODS: risk factors were determined by univariate Cox regression analysis, and entered in multivariate analyses which distinguished two models: analysis of factors available from history and clinical examination, completed by radiological and surgical data in the second model. To compare the information content of the two models, receiver-operator characteristic (ROC) curves were computed. RESULTS: in all patients female gender, critical ischaemia, femorocrural bypass grafting and non-venous graft material were independent risk factors. In patients with femoropopliteal bypasses female gender, critical ischaemia, poor run-off and non-venous graft material, the latter even in patients with supragenicular bypasses, were independent risk factors. The only significant risk factor in patients with femorocrural bypass grafts was use of a non-venous graft. The information contained in the first model was poor, whereas the second model had a higher predictive value. CONCLUSIONS: the major risk factor, even in above-knee bypasses, is non-venous graft material. The venous bypass graft should be offered to patients whenever possible. Copyright 2000 Harcourt Publishers Ltd.
OBJECTIVES: to identify risk factors for infrainguinal bypass occlusion and quantify the predictive value of data available before and after surgery. DESIGN: prospective study of 2650 patients who participated in a randomised trial of oral anticoagulants or aspirin after infrainguinal bypass surgery. MATERIALS AND METHODS: risk factors were determined by univariate Cox regression analysis, and entered in multivariate analyses which distinguished two models: analysis of factors available from history and clinical examination, completed by radiological and surgical data in the second model. To compare the information content of the two models, receiver-operator characteristic (ROC) curves were computed. RESULTS: in all patients female gender, critical ischaemia, femorocrural bypass grafting and non-venous graft material were independent risk factors. In patients with femoropopliteal bypasses female gender, critical ischaemia, poor run-off and non-venous graft material, the latter even in patients with supragenicular bypasses, were independent risk factors. The only significant risk factor in patients with femorocrural bypass grafts was use of a non-venous graft. The information contained in the first model was poor, whereas the second model had a higher predictive value. CONCLUSIONS: the major risk factor, even in above-knee bypasses, is non-venous graft material. The venous bypass graft should be offered to patients whenever possible. Copyright 2000 Harcourt Publishers Ltd.