OBJECTIVE To assess the expression of the androgen receptor (AR) and oestrogen receptors (ERs) in bladder tumours because recent studies have shown conflicting results and the prognostic significance of their expression remains unclear. PATIENTS AND METHODS We investigated the expression of AR, ERα and ERβ in 188 bladder tumour specimens, as well as matched 141 non-neoplastic bladder and 14 lymph node metastasis tissues, by immunohistochemistry. We then evaluated the relationships between their expression and the clinicopathological features available for the present patient cohort. RESULTS AR/ERα/ERβ was positive in 80%/50%/89% of benign urothelium, 50%/67%/41% of benign stroma, 42%/27%/49% of primary tumours and 71%/64%/71% of metastatic tumours. Significantly lower expression of AR/ERα was found in high-grade tumours (36%/23%) and tumours invading muscularis propria (33%/19%) compared to low-grade tumours (55%; P= 0.0232/38%; P= 0.0483) and tumours not invading muscularis propria (51%; P= 0.0181/35%; P= 0.0139), respectively. Significantly higher expression of ERβ was found in high-grade tumours (58%) and tumours invading muscularis propria (67%) compared to low-grade tumours (29%; P= 0.0002) and tumours not invading muscularis propria (34%; P < 0.0001), respectively. Kaplan-Meier and log-rank tests further showed that positivity of ERβ (but not AR or ERα) was associated with the recurrence of low-grade tumours (P= 0.0072); the progression of low-grade tumours (P= 0.0005), high-grade tumours not invading muscularis propria (P= 0.0020) and tumours invading muscularis propria (P= 0.0010); or disease-specific mortality in patients with tumours invading muscularis propria (P= 0.0073). CONCLUSIONS Compared to benign bladders, a significant decrease in the expression of AR, ERα or ERβ in bladder cancer was seen. Loss of AR or ERα was strongly associated with higher grade/more invasive tumours, whereas ERβ expression was increased in high-grade/invasive tumours and predicted a worse prognosis. © 2011 BJU International.