Objective To explore the correlation between the different glycemic variability (GV) indices calculated by self-monitoring of blood glucose (SMBG) and mean amplitude of glycemic excursion (MAGE) obtained from the continuous glucose monitor (CGM) system in patients with type 2 diabetes mellitus (T2DM). Methods This was a retrospective study. We analyzed the data of T2DM patients who received 48-72 h CGM and 7-point SMBG (pre and post-breakfast, lunch, dinner and prior to bedtime) simultaneously in Department of Endocrinology and Metabolic Disease, the Third Affiliated Hospital of Sun-Yat-sen University from January 2018 to October 2019. The GV indices calculated from the 7-point SMBG data included the standard deviation (SDBG) of the 7-point glucose profiles, the largest amplitude of glycemic excursions (LAGE) and the postprandial glucose excursion (PPGE), coefficient of variation of blood glucose (CV) and mean amplitude of glucose excursion (MAGE', calculated by SMBG profile). Spearman's correlation analysis, simple linear regression and multiple stepwise regression analysis were used to analyze the relationship between the different GV indices calculated from 7-point SMBG and MAGE obtained by CGM, and the receiver operator characteristic (ROC) curve was drawn to evaluate the ability of the former to predict the latter. Results Among 105 patients with T2DM, the SDBG, PPGE, LAGE and CV calculated by 7-point glucose profiles of SMBG were (2.02±0.77) mmol/L, (2.75±1.13) mmol/L, (5.62±2.13) mmol/L and (25.92±0.77)%, respectively, while the level of MAGE' and median MAGE were 4.11 (2.84, 5.92) mmol/L and 4.00 (2.65, 5.00) mmol/L, respectively. SDBG, PPGE, LAGE, CV and MAGE' were significantly correlative with MAGE (r=0.614, 0.499, 0.588, 0.533 and 0.473, respectively, all P<0.01). Multiple stepwise regression analysis was performed with MAGE as dependent variable and SDBG, PPGE, LAGE, CV and MAGE' as independent variables, and only SDBG entered the equation (P<0.01). The areas under ROC curve for SDBG [0.795, 95% confidence interval (CI) was 0.708-0.882], LAGE (0.782, 95%CI was 0.692-0.872) and CV (0.769, 95%CI was 0.677-0.846) were larger than those for PPGE (0.718, 95%CI was 0.620-0.816) and MAGE' (0.704, 95%CI was 0.607-0.789, all P<0.01) in reflecting the unqualified MAGE. Conclusions GV indices calculated from 7-point SMBG data including SDBG, LAGE, PPGE, CV and MAGE' are positively correlated with MAGE obtained from CGM. And SDBG has higher accuracy than the other four parameters. [ABSTRACT FROM AUTHOR]