The Impact of Baseline User Characteristics on the Benefits of Real-Time Versus Intermittently Scanned Continuous Glucose Monitoring in Adults With Type 1 Diabetes: Moderator Analyses of the ALERTT1 Trial

Abstract

Background: ALERTT1 showed that switching from intermittently scanned continuous glucose monitoring (isCGM) without alerts to real-time CGM (rtCGM) with alert functionality improved time in range (TIR; 70-180 mg/dL), glycated hemoglobin (HbA1c), time <54 mg/dL, and Hypoglycemia Fear Survey version II worry subscale (HFS-worry) score after six months in adults with type 1 diabetes (T1D). Moderator analyses aimed to identify certain subgroups that would benefit more from switching to rtCGM than others.

Methods: Post hoc analyses of ALERTT1 evaluated the impact of 14 baseline characteristics on the difference (delta) in mean TIR, HbA1c, time <54 mg/dL, and HFS-worry score at six months between rtCGM and isCGM. Therefore, the delta was allowed to depend on each of these variables by including interactions in the moderator analysis model. Analyses were performed separately for each variable; variables with P < .10 in the univariable analysis were combined into a single model.

Results: Univariable analyses showed no dependency of delta TIR, HbA1c, or time <54 mg/dL on variables other than CGM type. Only delta HFS-worry score depended on baseline HbA1c (P = .0059), indicating less worries with rtCGM in people with baseline HbA1c <6.5% or ≥8%. Given P < .10 for dependency of delta TIR on insulin therapy type (favoring multiple daily injections), baseline HbA1c, and baseline TIR, these variables were combined into a multivariable analysis; interactions were not statistically significant.

Conclusions: Except for HFS-worry score, no interactions between 14 baseline characteristics and the six-month intervention effect of rtCGM on TIR, HbA1c, or time <54 mg/dL were observed, supporting the conclusion of ALERTT1 that switching from isCGM without alerts to rtCGM with alert functionality is beneficial for a wide range of people with T1D.

Keywords: continuous glucose monitoring; intermittently scanned continuous glucose monitoring; moderator analysis; real-time continuous glucose monitoring; type 1 diabetes.

Figure 1.

Schematic overview of moderator analyses of the ALERTT1 trial. This post hoc analysis of ALERTT1 evaluated the impact of different baseline user characteristics (so-called moderators, as indicated in the box) on the difference (delta [Δ]) in mean time in range, HbA1c, time <54 mg/dL, and hypoglycemia worry at six months between rtCGM and isCGM. For this evaluation, the delta was allowed to depend on each of these variables by including interactions in a model, the moderator analysis. Variables were binary or continuous and selected on the basis of previously published literature. Abbreviations: rtCGM, real-time continuous glucose monitoring; isCGM, intermittently scanned continuous glucose monitoring; HbA1c, glycated hemoglobin.

Figure 2.

Univariable moderator analyses with continuous moderators—time in range. Results from univariable moderator analyses, evaluating whether the difference in means of rtCGM versus isCGM at six months (delta) for time in range (sensor-glucose 70-180 mg/dL; plotted on the y-axis), depend on the level of different baseline variables (plotted on the x-axis): age (panel a); diabetes duration (panel c); HbA1c (panel e); isCGM scan frequency (panel g); HFS-worry score (panel i); and different time in ranges (panels b, d, f, h, j). Dark blue lines represent the predicted mean delta, blue shaded areas represent the 95% confidence interval. A flatter line indicates less interaction, while a more curved line indicates more interaction. For each considered moderator, the P value for interaction is reported in the graph. Note that only continuous moderators are reported here. To convert glucose ranges from mg/dL to mmol/L, divide by 18. To convert HbA1c from % to mmol/mol: (10.93 × [HbA1c %]) − 23.5 mmol/mol. Abbreviations: rtCGM, real-time continuous glucose monitoring; isCGM, intermittently scanned continuous glucose monitoring; HFS-worry, Hypoglycemia Fear Survey version II worry subscale; HbA1c, glycated hemoglobin.

Figure 3.

Univariable moderator analysis—HFS-worry score and baseline HbA1c. Result from a univariable moderator analysis, evaluating whether the difference in means of rtCGM versus isCGM at six months (delta) for HFS-worry score (plotted on the y-axis), depends on the level of baseline HbA1c (plotted on the x-axis). Dark blue lines represent the predicted mean delta, blue shaded areas represent the 95% confidence interval. A flatter line indicates less interaction, while a more curved line indicates more interaction. The P value for interaction is reported in the graph. Results show that there was dependency of delta HFS-worry score on HbA1c, indicating less worries with the use of rtCGM compared with isCGM in people with low (<6.5% [48 mmol/mol]) or high (≥8% [64 mmol/mol]) baseline HbA1c. To convert HbA1c from % to mmol/mol: (10.93 × [HbA1c %]) − 23.5 mmol/mol. Abbreviations: HFS-worry, Hypoglycemia Fear Survey version II worry subscale; rtCGM, real-time continuous glucose monitoring; isCGM, intermittently scanned continuous glucose monitoring; HbA1c, glycated hemoglobin.