1503 Background: The US Food and Drug Administration and other regulatory agencies recently warned prescribers that use of pioglitazone, a thiazolidinedione (TZD), may increase the risk of bladder cancer. France and Germany removed the drug from their markets, although the rest of Europe did not. However, no information was available about the risk from alternative TZDs. We aimed to compare bladder cancer risk over time with use of TZDs relative to sulfonylureas (SUs), and between pioglitazone and rosiglitazone. Methods: We conducted a cohort study of patients with type 2 diabetes who initiated treatment with a TZD or SU using The Health Improvement Network (2000-2010), a UK general practitioner medical record database. Incident cancers within the database were identified. We computed hazard ratios (HRs) of bladder cancer for TZDs in comparison to the reference group of SU users. Results: There were 60 incident diagnoses of bladder cancer in the TZD cohort (n=18,459) and 137 in the SU cohort (n=41,396). There was no significant difference in bladder cancer risk between the cohorts (HR 0·93, [95% CI 0·68-1·29]), but most use was short-term use. In contrast, the risk of bladder cancer increased with increasing time since initiation of TZD versus SU therapy (HRs 1·15, 1·40, and 1·72 for 3-4, 4-5, and ≥ 5 years, respectively; P-trend=0·033). Bladder cancer risk also increased with increasing time since initiation of pioglitazone (P-trend<0·001) and rosiglitazone (P-trend=0·006). Direct comparison of pioglitazone to rosiglitazone did not demonstrate significant differences in cancer risk with increasing time since initiation (P-trend=0·12) or duration of therapy (P-trend=0·75). Conclusions: Long-term TZD therapy is associated with an increased risk of bladder cancer, which appears to be a class effect.