Introduction: Upper tract urothelial carcinoma (UTUC) accounts for 5% of all urothelial tumours. Due to its rarity, evidence regarding postoperative surveillance is lacking. The objective of this study was to develop a post-radical nephroureterectomy (RNU) surveillance protocol based on recurrence patterns in a large multi-institutional cohort of patients. Methods: Retrospective clinical and pathological data were collected from 1029 patients undergoing RNU over a 15-year period (1994-2009) at 10 Canadian academic institutions. A multivariable model was used to identify prognostic clinicopathological factors, which were then used to define risk categories. Risk-based surveillance guidelines were proposed based on actual recurrence patterns. Results: Overall, 555 (49.9%) patients developed recurrence, including 289 (25.9%) in the urothelium and 266 (23.9%) with loco-regional and distant recurrences. Based on multivariable analysis, three risk groups were identified: 1) low-risk patients with pTa-T1, pN0 disease and no adverse histological features (high tumour grade, lymphovascular invasion [LVI], tumour multifocality); 2) intermediate-risk patients with pTa-T1, pN0 disease with one or more of the adverse histological features; and 3) high-risk patients with a ≥pT2 tumour and/or nodal involvement. Low-, intermediate-, and high-risk patients were free of urothelial recurrence at three years in 72%, 66%, and 63%, respectively, and free of regional/distant recurrence in 93%, 87%, and 62%, respectively. The risks of loco-regional and distant recurrences (p<0.0001) and time to death (p<0.0001) were significantly different between the low-, intermediate-, and highrisk patients.