Existing risk prediction methods for elective abdominal aortic aneurysm repair do not predict short-term outcome following endovascular repair

Abstract

Improving the safety of elective abdominal aortic aneurysm (AAA) repair has become an imperative. Five well-described risk-scoring systems developed on open aneurysm repair (OR) were tested on a multicenter contemporary sample of patients undergoing endovascular repair of AAA (EVR) to determine if they predicted 30-day morbidity and mortality. The Glasgow score (GAS), combined prognostic index (CPI), and its modification (M-CPI), the Leiden score and the Vascular Biochemical and Haematological Outctome Model (VBHOM) score were studied using a retrospective database of 846 patients. Thirty-day mortality and serious morbidity were used as end-points. A receiver-operator characteristic curves was plotted and the area under this (known as the c-statistic) was calculated to determine discriminatory ability of each model. Incidence of postoperative mortality was 2.2% and serious morbidity was 12.3%. All scores were predictive of mortality except the Leiden score, which had a c-statistic of 0.603 (95% CI, 0.485-0.720; P = .123). The VBHOM score and the M-CPI had a c-statistic of 0.649 (95% CI, 0.514 -0.783; P = .026) and 0.653 (95% CI, 0.544-0.763; P = .026), respectively. The best performing scores were the GAS and CPI, which had a c-statistic of 0.677 (95% CI, 0.559-0.795; P = .008) and 0.679 (95% CI, 0.572-0.787; P = .007), respectively. No score effectively predicted morbidity. None of the available scores predicted the outcome of EVR with enough accuracy to be recommended for clinical use. To improve preoperative risk prediction in EVR validation of new systems is required, taking into account morphologic features of the aneurysm to predict medium-term morbidity and re-intervention.