presents the baseline characteristics of the study sample by group. The groups were similar with respect to baseline weight, body mass index, BP, A1C, and diabetes therapy. The participants in the RT-CGM group were slightly younger on average, and there were more men in the RT-CGM group than in the SMBG group.
Characteristics of the Study Participants by Treatment Group
Changes-from-baseline in A1C are shown in . Mean (± SD) A1C decreased by 1.0% (±1.1%) for the RT-CGM group and 0.5% (±0.8%) for the SMBG group (p = .006). The median change-from-baseline in the RT-CGM and SMBG groups were -0.75% and -0.40%, respectively (z = 2.51 and p = .01). Note that three participants in the RT-CGM group opted not to use the technology following randomization but still are included in the RT-CGM group in these analyses. In multiple regression analyses, in which change-from-baseline for each outcome was regressed on age, gender, therapy, baseline value for the outcome, and treatment group, the RT-CGM group was a significant predictor of change in A1C. Specifically, the RT-CGM group had a statistically adjusted decline in A1C of 0.60% greater than the SMBG group (p = .002). Baseline A1C status was also a significant predictor (p < .0001).
Figure 2 Changes in A1C from baseline to 12 weeks, (A) by treatment group and (B) by RT-CGM usage group. The figure shows boxplots. The boxes themselves contain the 25th, 50th, and 75th percentiles. The whiskers of the boxes show the minimum and maximum values, (more ...)
The analyses of usage found that 16/50 used the RT-CGM <48 days and 34/50 used it ≥48 days. Mean (± SD) A1C decreased by 0.6% (±1.1%) for the participants in the RT-CGM <48 days group and 1.2% (±1.1%) for participants who used the technology per protocol. shows that the median A1C change-from-baseline in those using RT-CGM <48 days versus ≥48 days was -0.45% and -0.95%, respectively (chi-square 11.33 with two degrees of freedom and p = .005). The difference among those in the SMBG group, the RT-CGM <48 days group, and the RT-CGM ≥48 days group was significant (p = .002) in the ANOVA and persisted in the multiple regression models, which found that the RT-CGM group had an average, statistically adjusted decline in A1C of -0.48% greater than the SMBG group (p = .006) net of baseline A1C status, which was also significant (p < .0001).
The extent of the effect of treatment group and usage group on A1C differed by baseline A1C () so that a higher A1C predicted a better response. Further, the difference in A1C decline between the SMBG group and the group that used RT-CGM ≥48 days was narrowest among participants with a low baseline A1C and widest among participants with a high baseline A1C ().
Figure 3 Interaction of baseline A1C with group in the prediction of 12-week change in A1C. Figures were derived from multiple regression analyses. P values for interaction effects are shown in parentheses in the graphs and refer to the difference from the SMBG (more ...)
shows the mean glucose in the RT-CGM group taken both from their RT-CGM readings and their accompanying SMBG as well as from the SMBG group. The RT-CGM group tested 2.9 times per day while the SMBG group tested 2.4 times per day. Although data from the RT-CGM and SMBG readings are not directly comparable, we present this information because it is typically reported in RT-CGM studies and provides a context for interpreting the A1C results.
Summary Statistics of Glucose Readings, Baseline to 12 Weeks
We calculated the net change in hypoglycemic medica-tion prescriptions by subtracting the number of discontinu-ations from the number of initiations. We similarly calculated the number of dose changes of existing medications by subtracting the number of decreases from increases. The medications used during the study by the RT-CGM and SMBG groups, respectively, were metformin (37 and 43), glipizide (26 and 29), glyburide (2 and 4), glimepiride (2 and 3), pioglitazone (15 and 11), exenatide (10 and 1), liraglutide (0 and 1), sitagliptin (13 and 9), acarbose (1 and 0), and basal insulin (either glargine or detemir) (21 and 25). The RT-CGM group had about the same net change in new diabetes medications (11 versus 12) and similar increases in doses of existing medications (20 versus 23). However, only 6% (3/50) of patients in the RT-CGM group, compared with 16% (8 of 50) in the SMBG group, were begun on basal insulin.
The mean SUS score at baseline was 72.4 ± 13.5, suggesting moderately good usability. It was unchanged at 12 weeks (74.8 ± 14.6). There was a weak correlation of -0.27 (p = .07) between the baseline SUS score and change in A1C.
Changes-from-baseline of weight, BP, and PAID scores did not differ by treatment group or usage.