Prediction of Overall Mortality in Patients Who Underwent Endovascular Treatment for Peripheral Artery Disease

초록

Background The K-VIS ELLA (Korean Vascular Intervention Society Endovascular Therapy in Lower Limb Artery Diseases) registry is a multicenter retrospective observational study in patients with lower extremity peripheral artery disease (PAD) undergoing endovascular treatment (EVT) in Korea. The aim of this study is to construct a prediction model to predict overall mortality in patients with PAD after EVT using the K-VIS ELLA registry. Methods We analyzed 2,831 patients from the registry. We divided this dataset into a training set (60%), which was used to construct a prediction model, and a testing set (40%), which was used to validate the established model in each gender. There was no difference between the training set and the testing set regarding input covariates. Cox regression was used to construct a model. We screened input covariates in the published research that have been demonstrated to be associated with mortality. Of those, we determined the final risk factors using backward stepwise logistic regression. The performance of the final model to calculate risk score, defined as the equation 0.0439 (age) + 0.0007(sex) − 0.0784(body mass index) − 0.0520(diabetes) − 0.1944(hypercholesterolemia) − 0.1672(current smoking) + 0.1538(congestive heart failure) + 0.9589(chronic kidney disease) − 0.0144(Rutherford classification) + 0.4665x(stroke) − 0.1221(hemoglobin), was checked by testing the dataset using the area under the curve. Results The mortality rates for men and women were 10.0% (233 of 2,339) and 12.0% (59 of 492), respectively. We reviewed previous studies and selected 11 covariates. After determining input features, we constructed a final model using a generalized linear model. The area under the curve of the established model was 0.736. On the basis of our model, the group with the top 10% of predicted probability had an HR of 31.53 compared with that with the lowest 10% of predicted probability. Conclusions We suggest a prediction model for overall mortality in patients with PAD who underwent EVT using baseline characteristics and risk factors.