Abstract: Objective To investigate the accuracy of real-time continuous glucose monitoring（RT-CGM）system in the type 2 diabetes mellitus（T2DM）. Methods A total of 318 subjects hospitalized between 2013 May to 2014 August were recruited. Each subject received a RT-CGM system. In order to calibrate the interstitial glucose level, finger-stick blood glucose were measured at least four times every day. Patients were divided into three groups according to their fingerstick blood glucose level: ≤6.99 mmol/L, 7.0-10.0 mmol/L and ≥10.01 mmol/L. Pearson correlation was used to analyze the relationship between the finger-stick glucose and interstitial glucose gained from RT-CGM. Clarke error grid analysis was used to analyze the accuracy of RT-CGM. Results （1）A total of 2, 815 glucose meter values from finger-stick were com⁃ paired with glucose from RT-CGM. A positive relationship（r=0.847）were found between CGM and finger-stick blood glu⁃ cose. The correlation was closer with higher finger-stick blood glucose levels（r=0.457, 0.648 and 0.852 respectively in ≤ 6.99 mmol/L group, 7.0-10.0 mmol/L group and ≥10.01 mmol/L group. P＜0.001）.（2）The Clarke error grid analysis re⁃ vealed that 76.69% of the readings from RT-CGM were clinical perfect (zone A), 20.28% were acceptable (zone B), 3.03% were unacceptable (zone C-E). Clinical perfect rate was lower when the finger-stick glucose ≤6.99 mmol/L , compared with that when it is between 7.0-10.0 or ≥10.01 mmol/L group（73.12% vs 78.63% & 79.28%，P＜0.05）, there was no significant difference between group of 7.0-10.0 mmol/L with group ≥10.01 mmol/L (P＞0.05). Conclusion RT-CGM provides accu⁃ rate blood glucose values to guide the treatment of diabetes, and the accuracy is more reliable in high glucose environment.

Key words: diabetes mellitus, type 2, real-time continuous glucose monitoring, accuracy, error grid analysis