Objective: Assessing diabetes therapy performance in Real World Data (RWD) is difficult as the data quantity of SMBG measurements shows strong variability. In a previous study, we proposed a new metric, Gkn/N ⇔ n out of N days :|SM BG| ≥ k, which ensures data quality so patients with skewed SMBG values or highly motivated users are not overrepresented. The classification separates users by logging habits, e.g., G1 ≅ 1 log a day . The aim of this study is to validate our metric with CGM data and determine if more logs a day increase metrics accuracy.

Method: We analyzed 299 users of a data set with CGM and SMBG logs . The inclusion criteria was 70% CGM logs for 1 month with associated SMBG logs. Each user was classified into a G-group. The distribution was biased to the G4-G5 groups. Therefore, users were randomly selected and had logs randomly removed to create lower G-group users, providing an equal distribution. The percent errors of the users’ mean BG, standard deviation (SD), and tests-in-range (TIR) between the SMBG and CGM logs were analyzed. With random sampling the analysis was repeated and mean results used.

Results: Results had G1 users with the largest error in all metrics tested (Mean Error:BG =8%, SD =15.4%, TIR =17.5%). The error did not substantially decrease as the logs increased (correlating to a higher G-groups). Because differences were not apparent in the mean BG, SD, or TIR;t-tests were performed for each group to determine statistical differences from the other groups based on BG or SD error. The only G-group to be statistically different (p<0.05) was G1 from all higher groups.

Conclusion: Results allow separating users into two groups. The error between G2 and all higher logging groups was not significant, implying metric accuracy does not increase with more logs. The validation allows the use of metrics based on mean BG or SD to a larger portion of users while limiting the bias on users’ logging habits. Further analyses are necessary to compare values from each G-class with clinically reported metrics.


R.P. Biven: None. J. Wrede: None. R.P. Bankosegger: Employee; Self; mySugr. J. Kober: Employee; Self; mySugr. B. Petersen: Employee; Self; Roche Diabetes Care. C. Ringemann: Employee; Self; Roche Diabetes Care. T. Huschto: Employee; Self; Roche Diabetes Care.

Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at http://www.diabetesjournals.org/content/license.