Glycaemic Metrics Improve Significantly Within 7 Days of Continuous Glucose Monitoring Initiation in Non‐Insulin–Using People With Type 2 Diabetes

1. Background

Type 2 diabetes (T2D) management relies heavily on daily self‐care behaviours, such as making healthy food choices and being physically active, yet maintaining healthy behaviours is challenging [1]. Continuous glucose monitoring (CGM) can offer people with T2D a comprehensive view of how food and lifestyle choices influence their glucose values, and CGM is now a recommended consideration for anyone with diabetes if it can help with their diabetes management [2]. Growing research suggests biofeedback from the device may help promote behaviour changes that lead to improved glycemia [3]. However, the timeframe and persistence of glycaemic changes after CGM initiation, specifically in non‐insulin–using people with T2D who use the device to guide lifestyle choices, are unknown. Understanding the timing and extent of glycaemic improvements after CGM initiation in this population could enhance shared‐decision making between people with diabetes and their care providers.

The UNITE (Using Nutrition to Improve Time in rangE) study was a randomized controlled trial that evaluated the glycaemic impact of two different CGM initiation approaches—a nutrition‐focused approach (NFA) and a self‐directed approach (SDA) in people with T2D who were not using insulin [4]. Before randomization, blinded CGM metrics were assessed. Throughout the intervention, participants in both arms continuously wore a CGM sensor and were instructed to use their glucose data to guide behaviour changes while keeping their diabetes medications stable.

The purpose of this post hoc analysis using UNITE study data is to describe the timeframe and persistence of glycaemic changes after CGM initiation. Specifically, this analysis describes the week‐over‐week changes in CGM‐derived glycaemic metrics compared to the blinded Baseline period in non‐insulin–using people with T2D who used real‐time CGM to help guide behavior change.

2. Methods

2.1. Study Overview

The UNITE study was a 2‐month intervention with a 4‐month observational follow‐up period conducted between July 2023 and December 2024 in non‐insulin–using adults with T2D who had no CGM use for at least 90 days prior to consent. Full details of the UNITE study methodology [4], the NFA design [5], the main trial outcomes [4], and the impact of CGM discontinuation during follow‐up [6], were previously published. In brief, after a 7‐ to 10‐day blinded CGM wear period with a physically blinded Dexcom G7 sensor (Dexcom Inc., San Diego, CA), participants were randomized to initiate real‐time CGM with an unblinded Dexcom G7 sensor using either the NFA or SDA. Education on how to use the CGM system was provided to participants by a registered dietitian or registered nurse at the time of randomization. All participants were encouraged to follow their glucose regularly, and all received guidance on glycaemic targets using internationally recognized standards for people with T2D (e.g., aim for glucose 70–180 mg/dL and > 70% time in range 70–180 mg/dL [TIR70‐180]) [7].

2.2. Study Outcomes and Statistical Analysis

The primary outcome for this analysis was the week‐over‐week changes in CGM‐derived glycaemic metrics compared to the blinded Baseline period regardless of the type of CGM initiation approach (as there was no difference in the differences of glycaemic changes from Baseline between the NFA and the SDA in the main trial) [4]. The secondary outcome was to evaluate whether the weekly changes in CGM‐derived glycaemic metrics compared to blinded Baseline differed by type of CGM initiation approach (NFA vs. SDA).

CGM‐based outcome metrics were calculated for non‐overlapping 7‐day periods from Baseline through the intervention period. Baseline data (Week 0) were derived from the first 7 days of blinded sensor wear. Intervention data were derived from the continuously worn unblinded sensors, which yielded approximately 7 weeks of intervention data per person. CGM metrics included: percent TIR70‐180, percent time above range 180 mg/dL (TAR180), percent time above range 250 mg/dL (TAR250), percent time below range 70 mg/dL (TBR70), mean glucose, and % coefficient of variation (%CV). Change from Baseline for CGM‐derived metrics at each follow‐up week was modelled using mixed models with repeated measures with an integer variable for Week (0–7), unstructured means, and repeated within‐person measurements. For the secondary outcome, an interaction term between Week and randomization assignment was included. Data were assumed to be missing at random; these models utilized all available data at all eight time periods and used maximum likelihood‐based ignorable methods, which yield valid inference for outcomes missing at random. No adjustments were made for multiple comparisons. Statistical significance was defined as p < 0.05 and analyses were performed in SAS v9.4 and R v4.4.1.

3. Results

The analytic cohort for the main intervention outcomes included N = 124 participants (NFA: N = 64 and SDA: N = 60), and all N = 124 are included in this analysis. Participants were a mean (SD) of 65 (10.3) years of age, had a T2D diagnosis for 10.7 (6.5) years, and a baseline A1c of 7.9% (0.7%). Additional participant characteristics are found in Table S1.

Statistically significant and large, clinically relevant improvements were detected for all CGM metrics except %CV from Baseline to Week 1 (Table 1); these improvements were maintained relative to Baseline through Week 7 (Figure 1). For the secondary outcomes, there were minimal differences in the differences of weekly changes from Baseline based on originally assigned CGM‐initiation approach (Table S2). The TAR250 decreased more (improved) for the NFA arm compared to the SDA arm beginning at Week 4, and there was significantly less TBR70 in the SDA arm compared to the NFA arm at almost all timepoints, but this was not clinically relevant as TBR70 was less than 0.5% at all timepoints.

TABLE 1.

Week‐over‐week changes in CGM‐derived glycaemic measures.

	Baseline	Week 1	Week 2	Week 3	Week 4	Week 5	Week 6	Week 7
% time in range 70–180 mg/dL
Weekly estimate	51 (47, 56)	63 (58, 67)	67 (63, 72)	70 (65, 74)	69 (65, 74)	71 (66, 75)	71 (66, 75)	71 (66, 76)
Change from Baseline	—	+12 (5, 18) p = 0.0005	+16 (10, 23) p < 0.0001	+18 (12, 25) p < 0.0001	+18 (12, 25) p < 0.0001	+20 (13, 26) p < 0.0001	+20 (13, 26) p < 0.0001	+20 (14, 26) p < 0.0001
Change from prior week	—	+12 (5, 18) p = 0.0005	+5 (−2, 11) p = 0.15	+2 (−4, 9) p = 0.51	0 (−7, 6) p = 0.93	+1 (−5, 8) p = 0.68	0 (−6, 7) p = 0.97	0 (−6, 7) p = 0.91
% time above range > 180 mg/dL
Weekly estimate	49 (44, 53)	37 (33, 42)	33 (28, 37)	31 (26, 35)	31 (26, 35)	29 (25, 34)	29 (25, 34)	29 (24, 34)
Change from Baseline		−11 (−18, −5) p = 0.0006	−16 (−23, −10) p < 0.0001	−18 (−25, −12) p < 0.0001	−18 (−25, −12) p < 0.0001	−19 (−26, −13) p < 0.0001	−20 (−26, −13) p < 0.0001	−20 (−26, −13) p < 0.0001
Change from prior week		−11 (−18, −5) p = 0.0006	−5 (−11, 2) p = 0.15	−2 (−9, 4) p = 0.52	0 (−6, 7) p = 0.94	−1 (−8, 5) p = 0.69	0 (−7, 6) p = 0.98	0 (−7, 6) p = 0.92
% time above range > 250 mg/dL
Weekly estimate	14 (12, 17)	9 (7, 11)	8 (5, 10)	6 (4, 8)	7 (4, 9)	5 (3, 8)	5 (3, 7)	5 (3, 7)
Change from Baseline		−5 (−9, −2) p = 0.001	−7 (−10, −4) p < 0.0001	−8 (−12, −5) p < 0.0001	−8 (−11, −5) p < 0.0001	−9 (−13, −6) p < 0.0001	−9 (−13, −6) p < 0.0001	−10 (−13, −6) p < 0.0001
Change from prior week		−5 (−9, −2) p = 0.001	−2 (−5, 2) p = 0.37	−1 (−5, 2) p = 0.39	1 (−3, 4) p = 0.76	−2 (−5, 2) p = 0.38	0 (−3, 3) p = 0.96	0 (−3, 3) p = 0.89
% time below range < 70 mg/dL a
Weekly estimate	0.21 (0.13, 0.28)	0.07 (0, 0.15)	0.13 (0.05, 0.15)	0.08 (0, 0.15)	0.12 (0.05, 0.2)	0.11 (0.04, 0.19)	0.10 (0.02, 0.18)	0.10 (0.02, 0.17)
Change from Baseline		−0.13 (−0.24, −0.02) p = 0.02	−0.08 (−0.19, 0.03) p = 0.16	−0.13 (−0.24, −0.02) p = 0.02	−0.08 (−0.19, 0.03) p = 0.14	−0.09 (−0.20, 0.02) p = 0.10	−0.11 (−0.21, 0) p = 0.06	−0.11 (−0.22, 0) p = 0.05
Change from prior week		−0.13 (−0.24, −0.02) p = 0.02	0.05 (−0.05, 0.16) p = 0.33	−0.05 (−0.16, 0.06) p = 0.34	0.05 (−0.06, 0.16) p = 0.38	0 (−0.12, 0.10) p = 0.86	0 (−0.12, 0.09) p = 0.80	0 (−0.11, 0.11) p = 0.97
Mean glucose (mg/dL)
Weekly estimate	189 (183, 195)	174 (169, 180)	169 (163, 175)	166 (160, 172)	166 (161, 172)	164 (158, 170)	163 (158, 169)	163 (158, 169)
Change from Baseline		−15 (−23, −6) p = 0.0006	−20 (−29, −12) p < 0.0001	−23 (−31, −15) p < 0.0001	−23 (−31, −14) p < 0.0001	−25 (−33, −17) p < 0.0001	−26 (−34, −17) p < 0.0001	−26 (−34, −17) p < 0.0001
Change from prior week		−15 (−23, −6) p = 0.0006	−6 (−14, 3) p = 0.18	−3 (−11, 5) p = 0.51	0 (−8, 9) p = 0.92	−2 (−11, 6) p = 0.56	−1 (−9, 8) p = 0.91	0 (−8, 8) p = 0.99
Coefficient of variation (%)
Weekly estimate	22 (21, 23)	21 (20, 22)	21 (20, 22)	21 (20, 21)	21 (20, 22)	21 (20, 22)	21 (20, 22)	21 (20, 22)
Change from Baseline		−1 (−2, 0) p = 0.10	−1.2 (−2.3, 0) p = 0.05	−1.5 (−2.7, −0.4) p = 0.01	−1.2 (−2.3, 0) p = 0.05	−1 (−2.2, 0.2) p = 0.09	−1 (−2.2, 0.1) p = 0.08	−1.2 (−2.4, 0) p = 0.04
Change from prior week		−1 (−2, 0) p = 0.10	−0.2 (−1.3, 1) p = 0.77	−0.4 (−1.5, 0.8) p = 0.53	0.4 (−0.8, 1.5) p = 0.53	0.2 (−1, 1.3) p = 0.79	0 (−1.2, 1.1) p = 0.95	−0.2 (−1.3, 1) p = 0.77

Note: Model‐based means, 95% CIs from mixed models with repeated measures unless otherwise indicated. Table denominators are N = 124 at Baseline and Weeks 1–4 and 6–7, and N = 123 at Week 5.

Abbreviation: CGM, continuous glucose monitoring.

^a Skewed distribution and clinically insignificant values. Interpret with caution.

FIGURE 1.

Week‐by‐week changes in CGM‐derived metrics. (A) Estimated weekly change from the blinded Baseline CGM period with estimates and 95% CI taken from mixed models with repeated measures, and (B) the observed overall distributions for CGM metrics by week; the data in (B) may not total 100% due to rounding. CGM, continuous glucose monitoring.

Several moderators were assessed (gender, educational attainment, baseline frequency of fingerstick blood glucose testing, or years with diabetes); however, none of the moderators significantly impacted the results or changed the interpretation of the main effect of time.

4. Conclusions

CGM can lead to improved glycemia in people with diabetes [2], and it is increasingly being recognized as an intervention, not just a glucose monitoring tool [8]. CGM metrics are reported to improve rapidly after CGM initiation in people with both type 1 diabetes and T2D who use insulin [9], including those with T2D who use basal insulin only [10] (improvements noted about 1–2 weeks after real‐time CGM initiation).

However, until now, little was known about the timeline and persistence for glycemic impact in non‐insulin–using people with T2D who use CGM to guide behaviour changes and not medication adjustment or titration. Anecdotal evidence and case reports suggest CGM metrics could improve rapidly in this population [11], but it is unclear if these stories are anomalies or apply broadly.

These data from the UNITE study analyses demonstrate that large, clinically relevant and significant improvements in glycaemic status appear almost immediately after CGM initiation, and without diabetes medication changes. In this non‐insulin–using population, within 1 week of CGM initiation, TIR70‐180 increased by 12 percentage points, and by Weeks 5–7, TIR70‐180 was 20 percentage points higher than Baseline; this is four times the amount of TIR improvement noted for clinical benefit [7]. The persistence and stability of the glycaemic impact in this population are noteworthy and are potentially related to the continued feedback from the device, as previous research does not show as much glycaemic benefit when the device is used intermittently [12].

Strengths of this analysis include use of high‐quality CGM data from a well‐defined intervention where diabetes medications were well‐monitored and stable. Limitations include the lack of a comparison arm where participants only received support from a registered dietitian or registered nurse without the CGM system. Additionally, UNITE participants were predominantly white, male, and they were also highly educated, which could imply differences in the acceptance or use of technology compared to those with less education, thus potentially limiting the generalizability of the results.

At a time when global estimates show that only about 21% of people treated for diabetes have optimal glycaemic status (based on A1c) [13], it is clear that treatment options that can move the glycaemic needle are a high priority. The UNITE data highlight what is possible when CGM is positioned specifically as a tool to guide lifestyle changes, and these data add to the growing evidence showing the value of CGM as part of therapy for non‐insulin–using people with diabetes. Further research is required to understand how long these glycaemic improvements can be sustained and when additional support may be most effective; this is especially true in light of research indicating that continued CGM use (> 75% of the time) is associated with better glycaemic management and suggesting that 6 months post‐CGM initiation could be a critical time to focus on counselling [14]. Nonetheless, preliminary evidence from the UNITE study suggests CGM can lead to large, clinically meaningful glycaemic improvements in a nearly immediate timeframe.

Funding

This work was supported by the Park Nicollet Foundation, the Dexcom Inc. and the American Diabetes Association, Grant #7‐22‐JDFN‐27.

Conflicts of Interest

H.J.W. consulted with and received educational grant funds from Abbott Diabetes Care; she consulted with Lingo and with Dexcom, and she also received research grant funding and study supplies from Abbott Diabetes Care and Dexcom during the conduct of this study. H.J.W. received no personal payment; all payments went to the non‐profit, HealthPartners Institute. S.K.G. has no conflicts of interest to report. M.M.J. has no conflicts of interest to report.

Supporting information

Table S1: Participant characteristics at baseline.

Table S2: Weekly glycaemic outcomes by treatment approach (NFA or SDA).

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.