Hemoglobin A1c (HbA1c) has been used for the diagnosis of diabetes and glycemic monitoring. However, using HbA1c for glycemia estimation has some fallacies in anemic persons. Zidovudine (AZT) treatment is associated with anemia and/or increased mean corpuscular volume (MCV). This study aimed to compare the correlation between HbA1c and plasma glucose in HIV-infected individuals who were receiving AZT and non-AZT containing regimens. A cross-sectional study was conducted in 150 HIV-infected individuals. We evaluated the correlation of paired fasting plasma glucose (FPG), random plasma glucose (RPG), mean plasma glucose (MPG) and HbA1c values by using Pearson correlation. Multivariate linear regression was used to determine the associated factors of HbA1c. The mean age was 49.0 ± 10.5 years, and 60.0% were male. Thirteen patients (8.7%) had diabetes and 14 patients (9.3%) had anemia. There were significant correlations between HbA1c and plasma glucose (FPG, RPG, and MPG; p < 0.05, all). The correlation between HbA1c and MPG in patients receiving AZT [HbA1c = 3.18 + 0.02MPG; R2=0.44] and not receiving AZT [HbA1c = 3.76 + 0.02MPG; R2=0.43] indicated that HbA1c in patients receiving AZT was 0.58% underestimated. Multivariate linear regression analysis showed that hematocrit [β 0.192; 95% confidence interval (CI) 0.003, 0.690; p = 0.032] and MCV [β -0.195; 95% CI -0.326, -0.002; p = 0.047] were associated with HbA1c levels. HbA1c underestimates glycemia in HIV-infected individuals receiving AZT containing regimens. Factors associated with decreased HbA1c levels in HIV-infected individuals included decreased hematocrit and increased MCV. In HIV-infected individuals receiving AZT, using HbA1c for diabetes diagnosis or glycemia monitoring should be cautiously interpreted.