Objectives Long‐term outcomes of endoscopic ultrasound‐guided choledochoduodenostomy (EUS‐CDS) performed with lumen apposing metal stents (LAMS) have been poorly evaluated in small or retrospective series, leading to an underestimation of LAMS dysfunction. Methods All consecutive EUS‐CDS performed in three academic referral centers were included in prospectively maintained databases. Technical/clinical success, adverse events (AEs), and dysfunction during follow‐up were retrospectively analyzed. Kaplan–Meier analysis was used to estimate dysfunction‐free survival (DFS), with Cox proportional hazard regression to evaluate independent predictors of dysfunction. Results Ninety‐three patients were included (male 56%; mean age, 70 years 95% confidence interval (CI) 68–72; pancreatic cancer 81%, metastatic disease 47%). In 67% of procedures, 6 mm LAMS were used. Technical and clinical success were achieved in 97.8% and 93.4% of patients, respectively, with AEs occurring in 9.7% (78% mild/moderate). Dysfunction occurred in 31.8% of patients after a mean of 166 days (95% CI 91–241), with an estimated 6 month and 12 month DFS of 75% and 52%, respectively; mean DFS of 394 (95% CI 307–482) days. Almost all dysfunctions (96%) were successfully managed by endoscopic reintervention. Duodenal invasion (hazard ratio 2.7 95% CI 1.1–6.8) was the only independent predictor of dysfunction. Conclusions Endoscopic ultrasound‐guided choledochoduodenostomy shows excellent initial efficacy and safety, although stent dysfunctions occurs frequently during long‐term follow‐up. Almost all stent dysfunctions can be managed successfully by endoscopic reinterventions. We propose a comprehensive classification of the different types of dysfunction that may be encountered and rescue procedures that may be employed under these circumstances. Duodenal invasion seems to increase the risk of developing EUS‐CDS dysfunction, potentially representing a relative contraindication for this technique.