Objectives: Vascular complications are common poor prognosis in Takayasu arteritis (TAK). We aimed to develop machine learning (ML) models for prediction of vascular complications in TAK based on prospective data from the East China Takayasu Arteritis (ECTA) cohort. Methods: Data were collected from the ECTA cohort in which patients were enrolled from January 2009 to August 2020 and followed till February 2021 (n = 517). Predictor variables included 53 baseline features and outcome of interest was incident vascular complications. Data were randomly split into a training (85%) and test (15%) set. Logistic regression (LR), support vector machine, random forest (RF), k-nearest neighbors, XGBoost (XGB), and light gradient boosting machine models were trained using five-fold cross validation, and evaluated on the test set for recall, specificity, precision and area under ROC (AU-ROC) and precision-recall curves (AU-PRC). Permutation score was applied to assess feature importance to the outcome. Results: Over a median follow-up of 30 (15–44) months, incident vascular complications were observed in 29.0% (150/517) patients. The RF model demonstrated the best overall predictive performance (AU-ROC = 0.84, AU-PRC = 0.63). Both the RF and LR models had the highest specificity (0.98), and the XGB model had the highest recall (0.87). Progressive clinical course was an important feature significantly associated with the outcome for all models. Conclusions: It demonstrated the feasibility of developing ML models for prediction of vascular complications in TAK. The XGB model could help for early identification of high-risk patients, and RF and LR models could further confirm. Disclosures: All the authors declared no conflicts of interest.