<i>Background:</i> The value of measurement of glycosylated hemoglobin (HgbA_1C) in determining the degree of glycemic control in patients with chronic kidney disease (CKD) is unclear. <i>Methods:</i> A single-center, prospective cohort study was conducted in 128 veterans with diabetes mellitus and CKD. HgbA_1C was measured as clinically indicated and its relationship with random blood glucose (RBG) measurement evaluated prospectively over up to 10 years in three groups (end-stage renal disease (ESRD), CKD and controls who had diabetes but no CKD). <i>Results:</i> Between 1995 and 2011, in the control group, glycemic control as assessed by HgbA_1C was stable but improved when assessed by RBG. However, both the CKD and ESRD groups experienced declines in RBG and HgbA_1C. Declining HgbA_1C and RBG were noted prior to onset of dialysis. A fall in HgbA_1C remained after adjustment for RBG. A strong inverse relationship was seen between CKD stage and HgbA_1C even after adjusting for RBG such that the relationship between RBG levels and HgbA_1C was modified by CKD. <i>Conclusions:</i> In diabetic patients with late-stage CKD, glycemic control shows an improvement. However, HgbA_1C <7% may overestimate the degree of glycemic control. Therefore, reliance on HgbA_1C without home blood glucose monitoring may result in poor diabetes control.