CONTEXT: In trials, thiazolidinediones (TZDs) increase fracture risk in women, but the effects of discontinuation are unknown. OBJECTIVE: The objective was to investigate the effects of TZD use and discontinuation on fractures in women and men. DESIGN: This was a longitudinal observational cohort study using data from the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial bone ancillary study. Duration of TZD use and discontinuation during ACCORD, assessed every 2-4 months at clinic visits, were modeled as time-varying covariates in proportional hazards models for occurrence of first non-spine fracture. PARTICIPANTS: We studied a total of 6865 participants in ACCORD BONE. MAIN OUTCOME MEASURES: Main outcome measures were centrally adjudicated non-spine fracture. RESULTS: Average age was 62.4 (SD, 6.6) years; average duration of diabetes was 11.1 (SD, 7.8) years. Rosiglitazone was used by 74% and pioglitazone by 13% of participants. During a mean follow-up of 4.8 (SD, 1.5) years, 262 men and 287 women experienced at least one non-spine fracture. The fracture rate was higher in women with 1-2 years of TZD use (hazard ratio [HR] = 2.32; 95% confidence interval [CI], 1.49, 3.62) or >2 years of TZD use (HR = 2.01; 95% CI, 1.35, 2.98), compared with no use. The fracture rate was reduced in women who had discontinued TZD use for 1-2 years (HR = 0.57; 95% CI, 0.35, 0.92) or > 2 years (HR = 0.42; 95% CI, 0.24, 0.74) compared with current users. TZD use and discontinuation were not associated with non-spine fractures in men. CONCLUSIONS: TZD use was associated with increased non-spine fractures in women, but not men, with type 2 diabetes. When women discontinued TZD use, the fracture effects were attenuated.