OBJECTIVE: To examine longitudinal differences in glycemic control between non-Hispanic white and non-Hispanic black veterans with type 2 diabetes. DESIGN: Retrospective cohort study. SETTING: VA facility in the Southeastern United States. PARTICIPANTS: A 3-month person-period dataset was created for 8813 veterans with type 2 diabetes between June 1997 and May 2006. MAIN OUTCOME MEASURES: Primary outcome was mean change in hemoglobin A1c (HbA1c) over time. Secondary outcome was the odds of poor glycemic control over time (HbA1c >8%). For the primary outcome, a linear mixed model (LMM) approach was used to model the relationship of HbA1c levels and race/ethnicity over time. For the secondary outcome, generalized LMMs were used to assess whether glycemic control changed over time and whether change in glycemic control varied by racial/ethnic group. RESULTS: Mean age was 66.3 years, 36% were non-Hispanic black (NHB), 98% were male, 65% were married, and 50% were unemployed. Mean follow-up time was 4.4 years. Least square mean HbA1c levels from LMM adjusted for time and relevant confounders showed that NHBs had higher HbA1c values over time (mean difference of 0.54% [P < 0.001]). The final model with poor versus good glycemic control as the dependent variable, race/ethnicity as primary independent variable adjusted for time, and relevant confounders showed that NHBs were likely to have poor control compared with NHWs (OR: 1.8, 95% CI, 1.7; 2.0, P < 0.0001). CONCLUSIONS: NHB veterans were more likely to have higher mean HbA1c values and less likely to have good glycemic control over time compared with NHW veterans