Objective: After radical nephroureterectomy, substantial numbers of patients with upper tract urothelial carcinoma are ineligible for adjuvant chemotherapy owing to diminished renal function. Accurate pre-operative prediction of worse pathological findings in radical nephroureterectomy specimens can guide appropriate patient selection for neoadjuvant chemotherapy. Herein, we evaluated pre-operative voided urine cytology and the additional efficacy of selective ureteral cytology for predicting pathological features in upper tract urothelial carcinoma patients. Methods: This retrospective cohort study comprised 722 patients with upper tract urothelial carcinoma who underwent radical nephroureterectomy. Patients with concomitant bladder cancer and those who received neoadjuvant therapy were excluded. Finally, 437 patients with urinary cytology data were enrolled in the study. We assessed the positive voided urine and selective ureteral cytology for predicting higher pathological T stage (≥pT3), higher tumor grade (3) and positive lymphovascular invasion. Results: Previous bladder cancer, tumor location, clinical T stage and voided urine cytology (P = 0.029) were independently associated with ≥pT3, whereas selective ureteral cytology was not. Gender, clinical N category and voided urine cytology (P = 0.017) were independently associated with tumor Grade 3, whereas selective ureteral cytology was not. Hydronephrosis, clinical T stage, clinical N category and voided urine cytology (P = 0.0021) were independently associated with lymphovascular invasion, whereas selective ureteral cytology was not. Conclusions: Pre-operative positive voided urine cytology was an independent predictor for worse pathological findings in radical nephroureterectomy specimens, while selective ureteral cytology had no additional efficacy. However, further studies with larger numbers of patients and complete data sets are needed to select patients for more aggressive treatments including neoadjuvant chemotherapy.