OBJECTIVES: Using a large, US administrative pharmacy claims database, calculate proportion of days covered (PDC) using actual days supply, and compare estimated PDC (ePDC) using days supply derived from drug-specific Defined Daily Dose (DDD) criteria. METHODS: Continuously eligible patients filling non-insulin anti-diabetic medication were targeted from a large sample of pharmacy claims during 2008 and 2009.  Medications were grouped into ATC diabetes drug classes.  Proportion of days covered (PDC) was calculated as the number of days a patient had medication in their possession divided by the number of days in the period.  PDC was first calculated using actual days supply, then ePDC was calculated using an estimated days supply from DDD, strength, and pill quantity.  The percent of patients adherent to therapy was defined by a PDC > 0.80. The reliability of each method was assessed by Pearson correlation coefficients and agreement above chance was assessed using Kappa statistics. RESULTS: Adherence was calculated for 163,750 patients taking non-insulin anti-diabetics.  Overall, the PDC was 0.69 and ePDC 0.57. The two measures were also highly and significantly correlated (r=0.73; p<.0001).  The percent adherent was 48.8% (PDC) and 34.4% (ePDC), (Kappa=0.50; p<.0001).  At the medication class level, differences in PDC and ePDC ranged from 0.01 to 0.35, with correlation coefficients ranging from 0.40 to 0.93.  Differences in the percent adherent metric ranged from -2.3 to 23.2, and kappa values from 0.22 to 0.89. CONCLUSIONS: Applying DDD estimates for the purposes of diabetes adherence estimation when lacking days supply values may provide reasonable estimates of adherence based on results presented here.  At the medication class level there is greater variability in the reliability measures. Including claims from the U.S. only is a limitation of this analysis, as local treatment patterns may vary, and DDD values were not available for all U.S. medications.