Impact of Continuous Glucose Monitoring Versus Blood Glucose Monitoring to Support a Carbohydrate-Restricted Nutrition Intervention in People with Type 2 Diabetes

Abstract

Introduction:

Low- and very-low-carbohydrate eating patterns, including ketogenic eating, can reduce glycated hemoglobin (HbA1c) in people with type 2 diabetes (T2D). Continuous glucose monitoring (CGM) has also been shown to improve glycemic outcomes, such as time in range (TIR; % time with glucose 70–180 mg/dL), more than blood glucose monitoring (BGM). CGM-guided nutrition interventions are sparse. The primary objective of this study was to compare differences in change in TIR when people with T2D used either CGM or BGM to guide dietary intake and medication management during a medically supervised ketogenic diet program (MSKDP) delivered via continuous remote care.

Methods:

IGNITE (Impact of Glucose moNitoring and nutrItion on Time in rangE) study participants were randomized to use CGM (n = 81) or BGM (n = 82) as part of a MSKDP. Participants and their care team used CGM and BGM data to support dietary choices and medication management. Glycemia, medication use, ketones, dietary intake, and weight were assessed at baseline (Base), month 1 (M1), and month 3 (M3); differences between arms and timepoints were evaluated.

Results:

Adults (n = 163) with a mean (standard deviation) T2D duration of 9.7 (7.7) years and HbA1c of 8.1% (1.2%) participated. TIR improved from Base to M3, 61–89% for CGM and 63%–85% for BGM (P < 0.001), with no difference in change between arms (P = 0.26). Additional CGM metrics also improved by M1, and improvements were sustained through M3. HbA1c decreased by ≥1.5% from Base to M3 for both CGM and BGM arms (P < 0.001). Diabetes medications were de-intensified based on change in medication effect scores from Base to M3 (P < 0.001). Total energy and carbohydrate intake decreased (P < 0.001), and participants in both arms lost clinically significant weight (P < 0.001).

Conclusion:

Both the CGM and BGM arms saw similar and significant improvements in glycemia and other diabetes-related outcomes during this MSKDP. Additional CGM-guided nutrition intervention research is needed.

Introduction

Nutrition has long been recognized as a critical component of diabetes care and management.¹ Current standards of care in diabetes state there is no single eating pattern or ideal macronutrient distribution that works for everyone with diabetes.² However, for people with type 2 diabetes (T2D), low- and very-low-carbohydrate eating patterns have been shown to significantly improve short-term glycemic outcomes, such as glycated hemoglobin (HbA1c).² Glycemic outcome results from longer-term research on very-low-carbohydrate eating patterns, such as ketogenic diets, are mixed,³ but positive outcomes have been observed when ketogenic eating patterns are supported by remote, continuous care.^4,5

In people with diabetes who require insulin, use of continuous glucose monitoring (CGM) has also been shown to improve glycemic outcomes compared with usual care, which is often blood glucose monitoring (BGM).⁶ Research in people with T2D who use less-intensive insulin therapy is limited, but growing evidence suggests CGM may also be useful in this population when it is used to help establish and guide dietary or lifestyle changes.⁷

Reports of nutrition interventions using CGM to guide dietary intake are sparse, and the study designs and outcomes assessed in previous trials leave more to be desired. For example, Griauzde et al. conducted a pragmatic quality improvement study in people with T2D, which compared 12-month changes in HbA1c levels between people receiving usual care and people using intermittent CGM with low-carbohydrate nutrition counseling (<100 g/day).⁸ Choe et al. randomized people with T2D to receive either structured nutrition education plus CGM or standard care with BGM.⁹ In both of these examples, HbA1c improved more in the CGM arms; however, given that the nutrition guidance was not the same in the comparison groups, it is difficult to understand the true impact of the glucose monitoring device.

The purpose of this research was to assess differences in diabetes-related outcomes based on the type of glucose monitoring used to guide a medically supervised ketogenic diet program (MSKDP). The primary objective was to compare differences in change in CGM-derived time in range (TIR) from baseline to 3 months in participants randomized to use CGM or BGM as part of the MSKDP. It was hypothesized that the continuous feedback provided by the CGM would improve adherence to dietary guidance and improve medication management, therefore leading to improved glycemia. The secondary and exploratory objectives were to assess changes from baseline to 3 months in additional CGM metrics, HbA1c, medication effect scores (MES), ketones, dietary intake, body weight, and patient-reported outcomes such as diabetes distress.

Methods

Design

The IGNITE (Impact of Glucose moNitoring and nutrItion on Time in rangE) study was a randomized, two-arm, parallel-group, virtual-care study (ClinicalTrials.gov identifier NCT 05516797) conducted by the International Diabetes Center (Minneapolis, MN) and Virta Health (Denver, CO).

This study was conducted according to the ethical standards described in the Declaration of Helsinki and according to the study protocol approved by the Advarra Institutional Review Board. Participants provided electronic informed consent prior to study procedures.

Participants

All people with T2D who enrolled in the Virta Health MSKDP between September 2022 and May 2023 were screened for the IGNITE study; they were invited to participate if they met all inclusion and exclusion criteria (Supplementary Table S1). In brief, inclusion criteria were ≥18 years, T2D, HbA1c 7.0–11.5% within 180 days, stable diabetes medication regimen and lifestyle patterns within 30 days, use of at least one glucose-lowering medication, ability to download and use the FreeStyle Libre 2 app on a personal smartphone, willing to use study-provided CGM devices and to perform fingersticks to test glucose and ketones twice daily, and intention to participate in Virta’s MSKDP for at least 7 months. Exclusion criteria were type 1 diabetes; insulin pump use or >3 insulin injections per day; following a self-reported, very low-carbohydrate diet; current or planned personal CGM use; contraindicated medical conditions; history of ketoacidosis; planned pregnancy, pregnant, or lactating; medical grade adhesive allergy; participation in another interventional trial; and/or unsuitable based on investigator discretion.

Sample size

The analytic sample size was selected to provide 80% power (alpha = 0.05) to test a 3-month postintervention difference in TIR between CGM and BGM of 10% with an expected TIR standard deviation of 20%. To account for expected retention of 85% at 3 months, the analytic sample size goal of 63 per arm (N = 126 total) was inflated to at least 75 (N = 150 total) participants with baseline TIR data.

Intervention

All participants voluntarily enrolled in the MSKDP, which is a remote-care metabolic clinic designed to help people with T2D reverse their condition through nutritional ketosis.

The intervention under study was not the MSKDP but rather the type of glucose monitoring (CGM or BGM) used to help guide dietary and lifestyle choices and medication management decisions as part of the MSKDP. BGM is considered usual care in the MSKDP. This study was entirely virtual with no in-person contact.

CGM arm participants were encouraged to wear their CGM sensor and use the CGM data continually throughout the MSKDP. Digital education on how to interpret and use the CGM data was provided to all CGM arm participants. The CGM device required scanning at least every 8 h to obtain complete glucose data; however, participants were encouraged to scan upon waking, with meals and activities, and before bed. Participants were encouraged to use the real-time and retrospective glucose data to learn about the impact of their food choices, behaviors, and medications. Target glucose range and %TIR targets, along with guidance on how to use high/low glucose alerts, were provided by the care team on an individual basis. The care team used cloud-based CGM data to support participants throughout the MSKDP. Ambulatory Glucose Profile Reports were systematically reviewed with CGM-arm participants by a care provider (an endocrinologist) around days 30, 60, and 90.

BGM arm participants were encouraged to perform one to two fingerstick glucose tests daily as is standard in the MSKDP. The care team monitored fingerstick glucose testing regularly and made adjustments to the number of glucose tests requested on an as-needed basis. Fingerstick glucose data were recorded in the program’s smartphone-based application (app) and used by the care team to support participants throughout the MSKDP.

Overall, provision of care from the health coaches and care providers to the study participants was identical between the two arms; the only systematic difference was the type of glucose monitoring data available and the addition of a CGM data review by a care provider at approximately days 30, 60, and 90 for the CGM arm participants.

In addition, all participants in the MSKDP were encouraged to achieve and sustain blood beta-hydroxybutyrate (BHB) levels within the range of 0.5–3.0 mmol/L. Initially, all participants were advised to consume 30 g of carbohydrates per day or less as part of a well-formulated ketogenic diet that encouraged whole, unprocessed vegetables, meats, eggs, dairy, some fruits, nuts, and seeds. Dietary carbohydrate was gradually increased over time according to the individual’s carbohydrate tolerance and health goals. Suggestions for protein intake started at 1.5 g/kg for the individual’s medium-frame reference body weight, and this level was tailored over time based on glucose and ketone values. Dietary fat was consumed for satiety. Program participants engaged with their dedicated care team (health coach and medical provider) through the program app. Participants regularly entered biometric data including, BHB levels, and body weight into the app, and these data informed the need for dietary modifications and/or medication adjustments to achieve and sustain nutritional ketosis.

Figure 1 provides the study overview. A baseline (Base) assessment period spanned approximately 21 days. During this time, up to 14 days of blinded CGM data, demographic and survey information, HbA1c, medication usage, a 24-h dietary recall, body weight, and height were collected. The MSKDP and the randomly assigned glucose monitoring method were initiated on day 0 after all baseline assessments were completed. Outcomes were assessed at 1 month (M1), 3 months (M3), and a follow-up at 6 months (M6; data not described).

FIG. 1.

Study overview. BGM, blood glucose monitor; CGM, continuous glucose monitoring; MSKDP, medically supervised ketogenic diet program; T2D, type 2 diabetes.

Randomization

After confirming at least five days of blinded CGM data, participants were randomly assigned to either the CGM or the BGM arm. The randomization schedule used permuted blocks of two and four and was generated by the study statistician using the random number generator in SAS. Blinding was not possible because the study compared two physically different glucose monitoring methods.

Study outcomes

The primary outcome was the between-arm comparison of change in up to 14-day CGM-derived TIR (% time with glucose 70–180 mg/dL) from Base to M3. Secondary outcomes included comparison of (1) change in other CGM-derived core metrics from Base to M1 and Base to M3; (2) percent reaching CGM-derived consensus targets, including composite scores, at M3; (3) change in HbA1c from Base to M3; and (4) 90-day mean BHB levels. Exploratory outcomes included the comparison of changes in diabetes medication use, dietary intake, body weight, and patient-reported outcomes from Base to M1 and/or Base to M3.

Outcome measurements

CGM assessments were collected using data from either a remote-start Freestyle Libre Pro system (blinded data) or the Freestyle Libre 2 (Abbott Diabetes Care, Alameda, CA). To align with guidance on CGM use in clinical trials,¹⁰ the blinded Libre Pro readers were programmed to use the Libre 2 algorithm so that all CGM data were equivalent. A minimum of 5 days, and up to 14 days, of CGM data were required for all participants during the Base, M1, and M3 assessment windows. The Freestyle Libre Pro system was used by all participants for the Base assessments and for the BGM arm participants during the M1 and M3 assessments. Neither participants nor the MSKDP care team had access to the blinded CGM data at any point during the study. For the CGM arm, data from the Libre 2 sensors were retrieved from LibreView by the study team during the M1 and M3 assessment windows.

Fingerstick glucose and BHB levels were assessed using the Precision Xtra Blood Glucose and Ketone Monitoring System (Abbott Diabetes Care); fingerstick glucose testing was prescribed on an individual basis, and ketones were requested twice daily for all participants. HbA1c was assessed by any clinical laboratory improvement amendments (CLIA)-certified laboratory using an NGSP HbA1c method at Base and M3. Dietary intake was assessed using the National Cancer Institute’s (NCI) Automated Self-Administered 24-hr recall system (ASA24), and the Healthy Eating Index-2015 (HEI) score was calculated using an NCI-provided code.¹¹ One 24-h recall was collected for each participant at Base, M1, and M3. Body weight was obtained from cellular-connected scales (BodyTrace, New York, NY). Height was self-reported. Medication use was tracked continuously, and the MES to describe medication intensity was calculated using established methods¹² at Base and on study days 30 and 90. Diabetes distress was assessed at Base and M3 using the 17-item Diabetes Distress Scale (DDS-17); total scores and subscale scores were calculated using established methods.¹³

Statistical analysis

Participant demographics, baseline clinical measurements, and counts of glucose-lowering medications at Base, M1, and M3 are summarized with descriptive statistics. The analysis of CGM-derived TIR and other CGM metrics, dietary intake, weight, and body mass index (BMI) utilized general linear mixed models with unstructured means and a compound symmetric heterogenous covariance matrix of residuals to model repeated measurements over time. These models utilized all available data at all time periods and accounted for missing data using maximum likelihood-based ignorable methods to yield valid inference for outcomes missing at random. The primary prespecified contrast estimated from these models consisted of the difference in differences (DID) estimating differential change by study arm from Base to M3. The DID contrast contrasting Base to M1 was also estimated. These models report model-based means and confidence intervals (CIs) for each time point by study arm, for change within study arm, and for the prespecified DID contrasts. The analysis strategy for continuous variables measured only at Base and M3 (HbA1c, DDS total score, and subscale score) followed the same approach. The analysis of repeatedly measured binary endpoints (e.g., lost >5% body weight and % reaching time below range [TBR] <4%) utilized generalized linear mixed models (logit link and binary error distribution and random intercept for participant) and reported odds ratios, CIs, and ratios of odds ratios to quantify the DID contrasts of interest. HEI total and component scores are reported as means and CIs estimated with the population ratio method, which produces summaries at the study arm level. HEI component scores were rescaled as a percentage of total for radar plots. All statistical testing was done with alpha = 0.05, and no adjustments for multiple comparisons were made. Analysis was conducted with SAS v9.4 (SAS Institute, Cary, NC).

Data management

Data sources included data recorded in Research Electronic Data Capture (REDCap) by the study team (e.g., from participants’ verbal or electronic communication) or direct REDcap data entry by study participants (e.g., surveys); data from Virta’s electronic medical record and program app; CGM data obtained from LibreView; and data downloaded from the ASA24 server.

Adverse events and serious adverse events

Adverse events were recorded and reported if they were (1) severe CGM sensor-related reactions or infections requiring in-person medical care, (2) severe hypoglycemia defined as requiring assistance of another person due to altered consciousness and requiring another person to actively administer carbohydrate, glucagon, or other resuscitative actions, or (3) severe hyperglycemia if the event involved diabetic ketoacidosis hyperosmolar or hyperglycemic state, as defined in Kitabchi et al.¹⁴

Serious adverse events (SAEs) were recorded and reported if they were deemed serious, unexpected, and probably, possibly, or definitely related to the glucose monitoring device or research procedures.

Results

Participants

Adequate blinded CGM data at baseline were obtained for 163 participants; 81 were randomized to use CGM and 82 were randomized to use BGM (Fig. 2). Participants were 49% female, 74% white, with a mean (standard deviation) age of 52.8 (9.5) years and a mean T2D duration of 9.7 (7.7) years (Table 1). At baseline, mean BMI was 36.1 (7.6), HbA1c was 8.1% (1.2), and the mean number of glucose-lowering medications was 2.1 (0.9).

FIG. 2.

Study flow diagram.

Table 1.

Participant Characteristics at Baseline

	Total N = 163	Continuous glucose monitoring N = 81	Blood glucose monitoring N = 82
Female, N (%)	79 (48.5)	41 (50.6)	38 (46.3)
Age, mean (SD)	52.8 (9.5)	53.5 (9.7)	52.3 (9.3)
Race, N (%)
Asian	7 (4.3)	5 (6.2)	2 (2.4)
Black	22 (13.5)	9 (11.1)	13 (15.9)
Native Hawaiian/Pacific Islander	2 (1.2)	0 (0)	2 (2.4)
White	121 (74.2)	62 (76.5)	59 (72.0)
More than one race	5 (3.1)	2 (2.5)	3 (3.7)
Unknown	6 (3.7)	3 (3.7)	3 (3.7)
Hispanic, N (%)^a	14 (8.6)	7 (8.6)	7 (8.5)
Weight at baseline (kg), mean (SD)^b	106.3 (25.2)	103.2 (25.3)	109.5 (24.9)
BMI at baseline (kg/m²), mean (SD)^b	36.1 (7.6)	34.9 (7.6)	37.2 (7.5)
BMI at baseline (kg/m²), N (%)^b
<25	9 (5.7)	7 (8.9)	2 (2.5)
25 to <30	28 (17.7)	14 (17.7)	14 (17.7)
30 to <35	36 (22.8)	19 (24.1)	17 (21.5)
35 to <40	37 (23.4)	21 (26.6)	16 (20.3)
40+	48 (30.4)	18 (22.8)	30 (38.0)
HbA1c %, mean (SD)^c	8.1 (1.2)	8.1 (1.2)	8.1 (1.2)
Years since T2D diagnosis, mean (SD)	9.7 (7.7)	8.8 (7.1)	10.5 (8.1)
Number of glucose-lowering medications including insulin, mean (SD)^d	2.1 (0.9)	2.0 (0.9)	2.2 (0.9)
% Time in range 70–180 mg/dL, mean (SD)	61.9 (31.5)	61.2 (31.8)	62.7 (31.5)
Education, N (%)^e
High school or less	18 (11.3)	9 (11.4)	9 (11.1)
Some college	44 (27.5)	21 (26.6)	23 (28.4)
College graduate	57 (35.6)	30 (38.0)	27 (33.3)
Some graduate school	10 (6.3)	5 (6.3)	5 (6.2)
Completed graduate school	31 (19.4)	14 (17.7)	17 (21.0)
Income, N (%)^a
<$50K	23 (14.5)	9 (11.4)	14 (17.5)
$50K to <$100K	67 (42.1)	35 (44.3)	32 (40.0)
$100K to <$150K	37 (23.3)	18 (22.8)	19 (23.8)
$150K+	32 (20.1)	17 (21.5)	15 (18.8)

N = 163 unless otherwise indicated.

N = 159.

N = 158.

N = 154.

Medication counts include all glucose-lowering medications, including metformin and insulin.

N = 160.

BMI, body mass index; HbA1c, glycated hemoglobin; SD, standard deviation; T2D, type 2 diabetes.

Glycemic outcomes

For the primary outcome, there was no difference in change in TIR from Base to M3 between the CGM and BGM arms (P = 0.26); however, within each arm, TIR improved significantly from Base to M3, 28% (61%–89%) for CGM and 22% (63%–85%) for BGM; P < 0.001 (Table 2). TIR improved quickly after initiating the MSKDP, as indicated by the significant increase at M1 that was maintained without additional changes from M1 to M3 (Fig. 3).

Table 2.

Glycemic Outcomes

	Continuous glucose monitoring N = 81					Blood glucose monitoring N = 82					Difference in changes,base to M3		Additional difference in changes (P-values)
	Base	M1	M3	Change base to M1	Change base to M3	Base	M1	M3	Change base to M1	Change base to M3	DID estimate	DID P value	Base to M1	M1 to M3
Primary:% TIR70–180 mg/dL^a,c	61 (54 to 68)	87 (83 to 91)	89 (84 to 93)	26*** (19 to 32)	28*** (21 to 34)	63 (56 to 70)	84 (80 to 89)	85 (80 to 90)	22*** (15 to 28)	22*** (15 to 29)	5 (−4 to 15)	0.26	0.36	0.77
% TAR>180 mg/dL^a,c	39 (31 to 46)	13 (8 to 17)	11 (6 to 15)	−26*** (−32 to −20)	−28*** (−34 to −21)	37 (30 to 44)	13 (8 to 17)	13 (8 to 18)	−24*** (−31 to −17)	−23*** (−30 to −17)	−5 (−14 to 5)	0.33	0.68	0.46
% TBR <70 mg/dL^a,c,e	0.3 (−0.1 to 0.7)	0.4 (−0.7 to 1.5)	0.7 (0 to 1.5)	0.1 (−1.0 to 1.2)	0.4 (−0.4 to 1.2)	0.6 (0.2 to 1.1)	3.0 (1.8 to 4.2)	1.9 (1.1 to 2.6)	2.4*** (1.1 to 3.6)	1.2** (0.4 to 2.0)	−0.8 (−1.9 to 0.4)	0.18	0.008	0.11
% TBR <54 mg/dL^a,c,e	0.04 (−0.02 to 0.11)	0.02 (−0.01 to 0.05)	0.03 (−0.003 to 0.07)	−0.02 (−0.1 to 0.05)	−0.01 (−0.09 to 0.06)	0.06 (−0.01 to 0.13)	0.07 (0.03 to 0.10)	0.05 (0.01 to 0.09)	0.007 (−0.07 to 0.08)	0.04 (−0.09 to 0.06)	0 (−0.11 to 0.11)	0.99	0.55	0.39
Mean sensor glucose mg/dL^a,c	178 (166 to 190)	135 (127 to 142)	130 (122 to 139)	−43*** (−54 to −33)	−48*** (−58 to −37)	172 (160 to 183)	125 (118 to 133)	130 (121 to 139)	−46*** (−57 to −36)	−42*** (−53 to −31)	−5 (−21 to 10)	0.48	0.71	0.17
% Coefficient of variation^a,c	23 (21 to 24)	18 (16 to 19)	18 (17 to 20)	−5*** (−6 to −4)	−4*** (−6 to −3)	23 (22 to 24)	18 (16 to 19)	19 (18 to 21)	−5*** (−7 to −4)	−4*** (−5 to −2)	0 (−2 to 2)	0.64	0.86	0.53
% reaching TIR >70%^b,c	44 (31 to 57)	86 (76 to 93)	90 (81 to 95)	8.2*** (3.7 to 18.2)	12.1*** (4.9 to 29.9)	53 (40 to 66)	82 (70 to 91)	85 (73 to 93)	4.1*** (1.8 to 9.3)	5.1*** (1.8 to 11.9)	2.4 (0.7 to 8.2)	0.17	0.23	0.81
% reaching TBR <4% ^b,c	98 (91 to 99)	97 (90 to 99)	92 (83 to 96)	0.95 (0.1 to 7.0)	0.3 (0.1 to 1.4)	95 (88 to 98)	84 (73 to 92)	84 (72 to 91)	0.3* (0.1 to 0.9)	0.2* (0.1 to 0.8)	1.1 (0.1 to 8.7)	0.90	0.28	0.23
% reaching TIR >70% and TBR <4%^b,c	42 (31 to 54)	82 (72 to 90)	81 (70 to 89)	6.5*** (3.1 to 13.7)	5.9*** (2.8 to 12.5)	49 (37 to 60)	66 (52 to 77)	66 (53 to 78)	2.0* (1.0 to 4.0)	2.1* (1.0 to 4.2)	2.9 (1.0 to 7.9)	0.04	0.02	0.82
HbA1c %^a,d	8.1 (7.8 to 8.4)	NA	6.5 (6.3 to 6.7)	NA	−1.6*** (−1.8 to −1.3)	8.1 (7.8 to 8.3)	NA	6.6 (6.4 to 6.8)	NA	−1.5*** (−1.7 to −1.2)	−0.1 (−0.5 to 0.2)	0.48	NA	NA
% Time in tight range 70–140 mg/dL^a,c	34 (27 to 40)	64 (57 to 71)	69 (62 to 76)	30*** (24 to 37)	35** (28 to 42)	38 (31 to 44)	68 (61 to 76)	70 (63 to 78)	31** (23 to 38)	33** (26 to 40)	2 (−8 to 12)	0.66	0.97	0.66

Model-based means, 95% CIs from mixed model repeated measures regression models.

Model-based percentages, odds ratios, 95% CIs from generalized linear mixed models.

Denominators, mean days of glucose readings (range): Base: CGM N = 81, 13 days (5–14), BGM N = 82, 13 days (6–14); M1 CGM N = 77, 11 days (5–14), BGM N = 63, 12 days (5–14); M3 CGM N = 72, 11 days (5–14), BGM N = 64, 12 days (5–14).

Denominators: Base: CGM N = 75, BGM N = 79; M3: CGM N = 64, BGM N = 64.

Skewed distribution. Interpret with caution.

*P < 0.05; **P < 0.01; ***P < 0.001.

Base, baseline; BGM, blood glucose monitoring; CGM, continuous glucose monitoring; CI, confidence interval; DID, difference in differences; M1, month 1; M3, month 3; TAR, time above range; TIR, time in range; TBR, time below range.

FIG. 3.

Time in ranges by study arm. TAR, time above range; TBR, time below range; TIR, time in range.

Within each arm, several other glycemic endpoints also improved. Time above range (TAR; % time with glucose >180 mg/dL), mean sensor glucose, % coefficient of variation (%CV), and time in tight range (TITR; % time with glucose 70–140 mg/dL) all improved from Base to M1 (P < 0.01) and Base to M3 (P < 0.01). TBR (% time with glucose <70 mg/dL) was infrequent in both arms, which resulted in skewed data; however, TBR (but not % time with glucose <54 mg/dL) did increase significantly from Base to M1 and Base to M3 in the BGM arm (P < 0.05). The percent of people reaching both the CGM-derived consensus targets, TIR >70% and TBR <4%, favored the CGM arm at M3 (P = 0.04).

Within arms, from Base to M3, HbA1c improved by 1.6% (8.1%–6.5%) for CGM and 1.5% (8.1%–6.6%) for BGM (P < 0.001); this was not different between arms. Among 124 people with an HbA1c measured at Base and M3, 73.4% had an improvement in HbA1c of at least 1% from Base to M3.

Diabetes medications

The number of diabetes medications taken by participants and the medication class types are shown in Table 3. Overall, medications were de-intensified based on the change in MES from Base to M1 and M3 for both the CGM and BGM arms (P < 0.001); there were no differences in change between arms (P = 0.79). The MES dropped from 1.6 to 1.1 and from 1.8 to 1.3 from Base to M3 for CGM and BGM, respectively (Table 3). Consistent with the overall medication reduction based on the MES, the total daily dose (TDD) of insulin also decreased in both arms. Of the people taking insulin at Base and/or M3, TDD decreased by approximately 60% in the BGM arm (N = 12; from mean TDD at Base = 66 units to mean TDD at M3 = 27 units) and approximately 59% in the CGM arm (N = 12; from mean TDD at Base = 45 units to mean TDD at M3 = 18 units).

Table 3.

Medication Use by Count, Class, and Intensity

	Baseline			Month 1			Month 3
	Total N = 163	CGM N = 81	BGM N = 82	Total N = 159	CGM N = 79	BGM N = 80	Total N = 152	CGM N = 74	BGM N = 78
Medication count
Total number of glucose-lowering medications, mean (SD)^a	2.1 (0.9)	2.0 (0.9)	2.2 (0.9)	1.6 (0.7)	1.5 (0.7)	1.7 (0.7)	1.6 (0.8)	1.5 (0.9)	1.7 (0.8)
Number of glucose-lowering medications, N (%)^a
0	0	0	0	3 (1.9)	1 (1.3)	2 (2.5)	9 (5.9)	6 (8.1)	3 (3.9)
1	47 (28.8)	25 (30.9)	22 (26.8)	77 (48.4)	46 (58.2)	31 (38.8)	64 (42.1)	36 (48.7)	28 (35.9)
2	68 (41.7)	35 (43.2)	33 (40.2)	63 (39.6)	25 (31.7)	38 (47.5)	59 (38.8)	22 (29.7)	37 (47.4)
3	32 (19.6)	13 (16.1)	19 (23.2)	16 (10.1)	7 (8.9)	9 (11.3)	18 (11.9)	9 (12.2)	9 (11.5)
≥4	16 (9.8)	8 (9.9)	8 (9.8)	0	0	0	2 (1.3)	1 (1.4)	1 (1.3)
Metformin only, N (%)^b	34 (20.9)	18 (22.2)	16 (19.5)	55 (34.6)	30 (38.0)	25 (31.3)	48 (31.6)	25 (33.8)	23 (29.5)
Medication class
Metformin, N (%)	121 (74.2)	57 (70.4)	64 (78.1)	125 (78.6)	59 (74.7)	66 (82.5)	116 (76.3)	53 (71.6)	63 (80.8)
Insulin, N (%)	23 (14.1)	10 (12.4)	13 (15.9)	23 (14.5)	11 (13.9)	12 (15.0)	19 (12.5)	8 (10.8)	11 (14.1)
Sulfonylurea or glinide, N (%)	50 (30.7)	23 (28.4)	27 (32.9)	2 (1.3)	1 (1.3)	1 (1.3)	6 (4.0)	4 (5.4)	2 (2.6)
Thiazolidinedione, N (%)	10 (6.1)	3 (3.7)	7 (8.5)	9 (5.7)	3 (3.8)	6 (7.5)	9 (5.9)	3 (4.1)	6 (7.7)
DPP-4 inhibitor, N (%)	11 (6.8)	10 (12.4)	1 (1.2)	7 (4.4)	7 (8.9)	0	4 (2.6)	4 (5.4)	0
SGLT2i, N (%)	60 (36.8)	34 (42.0)	26 (31.7)	15 (9.4)	5 (6.3)	10 (12.5)	16 (10.5)	5 (6.8)	11 (14.0)
GLP-1 RA, N (%)	69 (42.3)	29 (35.8)	40 (48.8)	70 (44.0)	31 (39.2)	39 (48.8)	67 (44.1)	30 (40.5)	37 (47.4)
Combination, N (%)^c	15 (9.2)	7 (8.6)	8 (9.8)	3 (1.9)	2 (2.5)	1 (1.3)	2 (1.3)	1 (1.4)	1 (1.3)
Medication intensity
MES, mean (SD)^d		1.6 (1.4–1.8)	1.8 (1.6–2.0)		1.2*** (1.0–1.3)	1.4*** (1.2–1.5)		1.1*** (0.9–1.2)	1.3*** (1.2–1.5)

Descriptive data only unless otherwise indicated.

Medication counts include all glucose-lowering medications, including metformin and insulin.

Excludes people taking metformin as part of a combination medication.

Medication containing more than one glucose-lowering medication.

Data are model-based means and 95% CIs from mixed model repeated measures regression models. Asterisks indicate P values for change within arm from baseline to M1 and baseline to M3.

*P < 0.05; **P < 0.01; ***P < 0.001.

MES, medication effect score.

Ketones, dietary intake, and weight

The mean 90-day BHB levels did not differ between arms; they were 0.54 mmol/L for CGM and 0.48 for BGM (P = 0.24). Ketones can be considered a proxy for carbohydrate intake; therefore, these data suggest both groups adhered equally to the dietary guidance provided. Mean 90-day BHB levels were negatively correlated with carbohydrate intake at M3 (r = −0.27; P = 0.01; data not shown).

Dietary data are shown in Table 4. In both study arms, total energy and carbohydrate intake decreased significantly without regular food logging or calorie counting (P < 0.001), whereas total protein and fat intake stayed stable or decreased. All changes in dietary intake were evident by M1, and most of these changes were maintained at M3. Dietary fiber intake decreased significantly for both arms and was well below the daily recommended intake level of 14 g per 1000 kcals based on the Dietary Guidelines for Americans. Intake of other nutrients of public health concern was either relatively unchanged (calcium and vitamin D) or decreased (potassium) from Base to M1 and M3. Diet quality, as measured by the total HEI score, did not change from Base to M3; however, total HEI scores at Base and M3 were below the desired score of ≥80 on a 0–100 scale (52.1 for CGM and 51.5 for BGM at M3). HEI component scores and radar plots depicting food intake changes during the intervention are provided in Supplementary Table S2 and Supplementary Figure S1.

Table 4.

Dietary Intake Data Based on a 24-H Dietary Recall at Each Time Point

	Continuous glucose monitoring					Blood glucose monitoring					Difference in changesBase to M3		Additional difference in changes (P values)
Dietary intake^a,b	Base	M1	M3	Changebase to M1	Changebase to M3	Base	M1	M3	Changebase to M1	Changebase to M3	DID estimate	DID P value	Base to M1	M1 to M3
Kcals	2245 (2014 to 2477)	1584 (1410 to 1758)	1607 (1430 to 1783)	−662*** (−916 to – 408)	−638*** (−894 to −383)	2459 (2251 to 2667)	1571 (1384 to 1757)	1607 (1425 to 1789)	−888*** (−1129 to −647)	−851*** (−1092 to −611)	213 (−139 to 564)	0.23	0.20	0.93
Carbohydrate grams	217 (186 to 248)	57 (45 to 70)	80 (64 to 97)	−160*** (−190 to −130)	−137*** (−167 to −106)	234 (206 to 262)	65 (52 to 79)	68 (51 to 85)	−168*** (−195 to −141)	−165*** (−193 to −137)	29 (−13 to 70)	0.18	0.68	0.12
% calories carbohydrate	38 (34 to 41)	15 (12 to 18)	20 (17 to 23)	−22*** (−26 to −19)	−18*** (−22 to −14)	38 (35 to 41)	18 (15 to 20)	16 (13 to 19)	−20*** (−24 to −17)	−22*** (−25 to −18)	4 (−2 to 9)	0.19	0.38	0.03
Protein grams	99 (86 to 111)	101 (91 to 112)	99 (86 to 112)	3 (−12 to 17)	0 (−16 to 17)	107 (96 to 118)	99 (88 to 111)	105 (91 to 119)	−7 (−22 to 7)	−2 (−18 to 14)	2 (−21 to 25)	0.86	0.33	0.48
% calories protein	18 (16 to 20)	27 (25 to 29)	25 (23 to 27)	9*** (7 to 12)	7*** (5 to 10)	18 (17 to 19)	26 (24 to 28)	27 (25 to 29)	8*** (5 to 11)	9*** (7 to 12)	−2 (−5 to 2)	0.29	0.52	0.14
Total fat grams	111 (98 to 124)	103 (91 to 116)	98 (86 to 110)	−8 (−24 to 9)	−13 (−29 to 3)	120 (108 to 132)	97 (83 to 110)	100 (88 to 113)	−23** (−39 to −7)	−20* (−35 to −4)	7 (−16 to 29)	0.54	0.19	0.47
% calories total fat	45 (42 to 47)	58 (55 to 60)	55 (52 to 58)	13*** (10 to 16)	11*** (7 to 14)	44 (42 to 46)	56 (53 to 59)	57 (54 to 60)	12*** (9 to 16)	13*** (9 to 16)	−2 (−7 to 3)	0.40	0.72	0.28
Saturated fat grams	37 (32 to 41)	37 (32 to 42)	36 (31 to 41)	0 (−6 to 6)	0 (−7 to 6)	39 (34 to 43)	33 (28 to 38)	35 (29 to 40)	−6 (−12 to 0)	−4 (−10 to 2)	4 (−5 to 12)	0.40	0.17	0.61
% calories saturated fat	15 (14 to 16)	22 (20 to 23)	21 (19 to 24)	7*** (5 to 8)	6*** (4 to 8)	15 (14 to 16)	20 (18 to 22)	21 (19 to 23)	5*** (3 to 7)	6*** (4 to 8)	0 (−3 to 3)	0.93	0.22	0.39
Monounsaturated fat grams	39 (34 to 44)	37 (32 to 41)	35 (30 to 41)	−3 (−9 to 4)	−4 (−10 to 2)	43 (39 to 47)	35 (30 to 40)	36 (31 to 41)	−8* (−14 to −2)	−7* (−13 to −1)	3 (−6 to 11)	0.55	0.24	0.58
% calories monounsaturated fat	17 (15 to 18)	22 (20 to 23)	21 (19 to 22)	5*** (3 to 7)	4*** (2 to 6)	17 (16 to 18)	21 (20 to 23)	21 (20 to 23)	5*** (3 to 6)	5*** (3 to 6)	−1 (−3 to 1)	0.44	0.83	0.37
Polyunsaturated fat grams	25 (21 to 29)	21 (17 to 24)	18 (15 to 21)	−5 (−10 to 1)	−7** (−12 to −2)	28 (24 to 31)	20 (16 to 24)	19 (16 to 22)	−7** (−12 to −2)	−8*** (−13 to −4)	1 (−6 to 8)	0.75	0.48	0.65
% calories polyunsaturated fat	10 (9 to 12)	12 (11 to 14)	11 (10 to 12)	2 (0 to 3)	0 (−1 to 2)	10 (9 to 11)	13 (11 to 15)	11 (10 to 13)	3** (1 to 5)	1 (0 to 3)	−1 (−3 to 2)	0.51	0.45	0.87
Fiber grams	14 (12 to 16)	8 (6 to 9)	9 (7 to 11)	−6*** (−8 to −4)	−5*** (−7 to −2)	17 (16 to 19)	9 (7 to 11)	9 (7 to 11)	−9*** (−11 to −6)	−9*** (−11 to −6)	4 (0 to 8)	0.04	0.14	0.40
Calcium mg	974 (796 to 1151)	871 (727 to 1015)	940 (786 to 1094	−103 (−323 to 117)	−34 (−260 to 193)	1094 (937 to 1252)	848 (693 to 1003)	939 (779 to 1098)	−246* (−458 to −34)	−156 (−371 to 60)	122 (−191 to 435)	0.44	0.36	0.89
Potassium mg	2483 (2186 to 2780)	2079 (1857 to 2301)	2107 (1819 to 2395)	−404* (−754 to −53)	−376 (−766 to 14)	2925 (2660 to 3189)	2187 (1948 to 2425)	2226 (1928 to 2523)	−738*** (−1072 to −404)	−699*** (−1074 to −324)	323 (−218 to 864)	0.24	0.17	0.96
Vitamin D mcg	5 (3 to 7)	5 (4 to 6)	5 (4 to 6)	0 (−2 to 2)	−1 (−3 to 2)	6 (4 to 7)	4 (3 to 5)	4 (3 to 5)	−2 (−4 to 0)	−2 (−4 to 0)	1 (−2 to 4)	0.45	0.29	0.64
Total Healthy Eating Index-2015 score (0–100)^c,d,e	49.1 (45.1 to 53.1)	NA	52.1 (49.5 to 54.8)	NA	+3.0	54.5 (48.6 to 60.5)	NA	51.5 (48.3 to 54.0)	NA	−3.0	NA	NA	NA	NA

Dietary recalls were collected using the National Cancer Institute’s Automated Self-Administered 24-h recall system (ASA24).

Table denominators except where otherwise noted were Base: CGM N = 52, BGM N = 66; M1: CGM N = 52, BGM N = 45; M3: CGM N = 45, BGM N = 42.

Data are presented as model-based means, 95% CIs from mixed model repeated measures regression models, except where otherwise noted.

Healthy Eating Index-2015 (HEI) is a measure of diet quality independent of diet quantity, and it is based on the degree to which food intake aligns with the recommendations from the Dietary Guidelines for Americans (DGA). The total HEI score is calculated using ASA24 data and is comprised of “component scores” from 13 food categories, 9 adequacy components (e.g., points for intake of total vegetables, total proteins, whole grains, etc.) and 4 moderation components (e.g., points for minimizing intake of added sugars, sodium, etc.). A score of 100 indicates perfect alignment with the DGAs. Of note, the DGAs are not intended to represent a specific macronutrient intake level, and therefore, it is unclear what an ideal HEI score may be for groups or individuals choosing a low or very-low-carbohydrate diet.

Denominators: Base: CGM N = 35, BGM N = 39; M3: CGM N = 35, BGM N = 39.

Means and CIs from HEI population ratio method. No statistical testing is done.

*P < 0.05; **P < 0.01; ***P < 0.001.

Dietary changes contributed to clinically meaningful reductions in body weight from Base to M3, with participants in the CGM arm losing an average of 7.2 kg and those in the BGM arm losing 7.8 kg; this was about 7% of total body weight loss for both arms (Table 5). More than 64% of all participants lost at least 5% body weight from Base to M3, and the percentage of people who lost >5% body weight nearly doubled from M1 to M3.

Table 5.

Body Weight Outcomes

	Continuous glucose monitoring					Blood glucose monitoring					Difference in changesBase to M3		Additional differences in change (P values)
	Base	M1	M3	ChangeBase to M1	ChangeBase to M3	Base	M1	M3	ChangeBase to M1	ChangeBase to M3	DID estimate	P value	Base to M1	M1 to M3
Weight kg^a,c	103.5 (98.0 to 108.9)	99.5 (94.4 to 104.7)	96.3 (91.1 to 101.4)	−3.9*** (−4.9 to −3.0)	−7.2*** (−8.2 to −6.2)	110.0 (104.5 to 115.5)	105.2 (100.0 to 110.4)	102.2 (97.0 to 107.30	−4.8*** (−5.8 to −3.9)	−7.8*** (−8.8 to −6.9)	0.6 (−0.7 to 2.0)	0.35	0.19	0.71
% with >5% body weight loss^b,d	NA	34.6 (24.9 to 45.9)	68.6 (56.7 to 78.4)	NA	NA	NA	38.5 (28.3 to 49.7)	64.9 (53.3 to 74.9)	NA	NA	NA	0.64^e	0.62^f	NA
BMI (kg/m²)^a,c	35.0 (33.4 to 36.7)	33.7 (32.2 to 35.3)	32.6 (31.0 to 34.2)	−1.3*** (−1.6 to −1.0)	−2.4*** (−2.8 to −2.1)	37.3 (35.7 to 39.0)	35.7 (34.2 to 37.3)	34.7 (33.1 to 36.3)	−1.6*** (−1.9 to −1.3)	−2.6*** (−2.9 to −2.3)	0.2 (−0.3 to 0.6)	0.41	0.22	0.68

Model-based means, 95% CIs from mixed model repeated measures regression models.

Model-based percentages, 95% CIs from generalized linear mixed model.

Denominators: Base: CGM N = 79, BGM N = 79; M1: CGM N = 80, BGM N = 80; M3: CGM N = 72, BGM N = 76.

Denominators: Base to M1 change: CGM N = 78, BGM N = 78; Base to M3 change: CGM N = 70, BGM N = 74.

P value for study arm comparison of % with >5% body weight loss from baseline to M3.

P value for study arm comparison of % with >5% body weight loss from baseline to M1.

*P < 0.05; **P < 0.01; ***P < 0.001.

Patient-reported outcomes

Based on previous validation research using the 17-item DDS-17,¹⁵ participants in both arms reported moderate diabetes distress at baseline; total DDS-17 scores were 2.5 for CGM and 2.8 for BGM. The total DDS-17 score decreased significantly from Base to M3 in both arms (and exceeded the minimal clinically important difference of 0.25),¹⁶ but the changes were not different between arms (−0.7 and −0.7, respectively; P = 0.89). Supplementary Table S3 provides details on changes in distress scores including by subscales.

Participant confidence using glucose data to help manage diabetes was rated higher for participants using CGM compared with BGM at M3 (based on a 5-point scale with 1 = not at all confident and 5 = very confident and an independent samples t-test with Satterthwaite approximation); mean score was 4.4 for CGM and 3.7 for BGM (P = 0.009).

Adverse events and protocol deviations

There were no reportable adverse events or SAEs. There were no major protocol deviations. Minor protocol deviations were infrequent and included missed study procedures, assessments outside of study windows, not following instructions for CGM sensor use or placement, and unreturned study supplies.

Discussion

To the best of our knowledge, this is one of the first studies to compare differences in TIR when people with T2D were randomized to use CGM or BGM to support a nutrition intervention.

From baseline to 3 months, several glycemic outcomes—TIR, TAR, and HbA1c—improved similarly and significantly with the use of either CGM or BGM during a MSKDP but with no differences in these improvements between arms. These data fail to reject the null hypothesis; CGM use did not lead to greater improvement in TIR compared with BGM use in this MSKDP for people with T2D. It is likely there were no differences between the CGM and BGM arms because the carbohydrate-restricted diet intervention with continuous remote care overpowered the impact of the glucose monitoring method. At month 3, the mean glucose was 130 mg/dL, and the %CV was ≤19 for both arms, perhaps suggesting that continuous glucose data may not have had much to add for people consuming minimal carbohydrate. It is possible the results would be different if comparing the use of CGM versus BGM in people following a more moderate carbohydrate diet intervention where the CGM data would help users identify which carbohydrate types or amounts work best for glycemic goal attainment (i.e., as part of a meal, is pasta, rice, or bread more likely to lead to glucose 70–180 mg/dL?). The impact of CGM or BGM to guide a dietary intervention would likely also be different for people using intensive insulin therapy who have a risk for severe hypoglycemia.

Clinically meaningful glycemic improvements happened quickly in the IGNITE study, as evidenced by the CGM metrics at M1 and the sustained improvements at M3. TIR improved by a mean of 22–28% from baseline to M3 despite a significant reduction in diabetes medication use, which is noteworthy when considering that every 5% increase in TIR is considered clinically beneficial.¹⁷ It is also notable that at least 85% of all participants reached a TIR of >70%, as a higher TIR has been associated with reduced risk of albuminuria, retinopathy, cardiovascular disease mortality, and other health outcomes.¹⁸

Typically, the goal of glycemic management is to maximize TIR while minimizing TBR. The percent of IGNITE study participants with both TIR >70% and TBR <4% was higher in the CGM arm compared with the BGM arm. However, it is somewhat difficult to interpret the clinical significance of this between-arm difference because it appears driven mainly by the BGM arm’s increased time with glucose between 54 and 69 mg/dL, which is often referred to as level 1 or not clinically significant hypoglycemia. In addition, this level 1 hypoglycemia occurred in participants who were consuming limited carbohydrate and using little to no medications that carry hypoglycemia risk.

Some guidance suggests that glucose values between 54 and 69 mg/dL may not need to be reported in certain settings.¹⁹ Further research is needed to understand the importance of time with glucose <70 mg/dL, but ≥54 mg/dL, in people with low-risk for hypoglycemia—especially given the growing interest in CGM use by people with prediabetes and people with T2D who are not using insulin.²⁰

Overall, the glycemic improvements—including the ≥1.5% reduction in HbA1c and average HbA1c of <7% at M3—were impressive for a nutrition and lifestyle intervention that included diabetes medication deprescription. These results are similar or greater to those that have been reported in some pharmacologic interventions. However, despite the notion that CGM is recognized as a helpful tool for guiding medical nutrition therapy and other lifestyle behaviors,⁶ there are very few randomized clinical trials for comparison that have evaluated the impact of using CGM versus usual care to guide a defined diet or lifestyle intervention. Aronson et al. compared the impact of CGM versus BGM in noninsulin using people with T2D who received six sessions of diabetes self-management education (DSME) over 4 months. Their study did not include a specific dietary intervention per se, but the emphasis on nutrition and food choices as part of the DSME was similar for both arms. At the end of 4 months, TIR was significantly greater (∼10%) for the CGM plus DSME arm than it was for the DSME alone arm (TIR: 56.3%–76.3% for CGM plus DSME and 57.5%–67.6% for DSME alone).²¹ Similarly, Wada et al. randomized people with T2D to use either CGM or BGM as a tool for adjusting diet and lifestyle choices based on glucose levels.²² In this study, general dietary guidance was provided, but diet and lifestyle changes were not assessed. Based on comparisons of change in blinded CGM data from baseline and 12 weeks, TIR improved more in the CGM arm compared with the BGM arm (∼2.3 h more TIR). These studies suggest that under some circumstances, CGM can be used not only for glucose monitoring but also as an intervention to guide people with diabetes in making diet and lifestyle changes that improve glycemia.²³ In our study, the impact of the glucose monitoring method appeared to be minimized by the structured, carbohydrate-restricted eating plan with continuous, remote support provided by the care team.

Several reviews have described the impact of ketogenic eating programs on diabetes-related outcomes such as HbA1c and body weight.^3,24,25 Results from the IGNITE study aligned with the general consensus that ketogenic eating patterns can lead to reductions in HbA1c and body weight for people with T2D, especially in the first 3–6 months. Interestingly, very few of the studies included in these reviews assessed participants’ actual dietary intake during the intervention; instead, studies typically reported the intended dietary intake based on the protocol (e.g., <50 g carbohydrate/day, <10% kcals from carbohydrate), which may or may not be what was actually consumed.

The IGNITE study assessed participants’ actual dietary intake using a validated 24-h recall, ASA24.¹¹ Total energy and carbohydrate intake decreased significantly from Base through M3 for both study arms, with minimal differences in total protein and fat intake. These findings help explain the approximately 7% weight loss and indicate generally good adherence to the prescribed eating pattern. However, based on an average consumption of about 15%–20% of kcals from carbohydrate, this would likely be considered a low-carbohydrate diet and not a ketogenic diet, as described by a recent expert consensus report on nutrition and low-carbohydrate diets.²⁶ This is worth considering because it demonstrates participants were able to achieve improved glycemic results and weight loss while still consuming some carbohydrates. It is also interesting to note that CGM arm participants appeared to consume a slightly greater percentage of calories from carbohydrate compared with BGM arm participants without compromising glycemic outcomes. It is possible participants in the CGM arm were more comfortable consuming foods containing some amount of carbohydrate because they could see the full impact of their foods or meals compared with the limited information provided by BGM, although this requires further research. CGM arm participants did, however, report significantly greater confidence using glucose data to manage their diabetes compared with BGM arm participants.

It is important to note that while many different eating patterns can help manage diabetes, ketogenic eating is best done in partnership with a health care team—as some research suggests concern for inadequate micronutrient intake and/or the potential for severe adverse health outcomes in certain individuals.^27,28 Participants in the IGNITE study were supported by a care team and encouraged to choose a well-formulated ketogenic diet, which emphasized nutrient-dense whole foods. Current standards of care recommend all people with diabetes should be encouraged to work with a registered dietitian nutritionist (RDN)²; this is especially true for people choosing low- and very-low-carbohydrate eating patterns to help manage diabetes.²⁸ Working with an RDN increases the likelihood of achieving glycemic goals with the healthiest food choices available for any dietary pattern.

This study had several strengths. First, it was well-powered, the participants were geographically diverse, and they were racially similar to the U.S. population (however, Hispanic representation was about half the current U.S. estimate). Second, CGM use followed many best practice recommendations. For example, all participants were required to collect up to 14 days of blinded CGM data at baseline, and BGM arm participants were asked to do the same during specified assessment periods; this allowed for meaningful comparisons of all CGM metrics between arms. For the CGM arm, CGM data were continuously available to the participants and the care team, and the data were systematically reviewed with the participants by endocrinologists each month. It is also notable that participants were able to remote-start their CGM and connect to their care team with only minimal printed instructions from the study team. Third, this research was conducted in a well-defined MSKDP, and actual dietary intake was assessed using a validated method, which is often not included in ketogenic diet interventions, thus helping to put the CGM metrics into context.

The IGNITE study also recognizes limitations. CGM was compared with BGM, which is the current usual care in the MSKDP, but many people with T2D who do not require insulin do not regularly use glucose monitoring and this state was not evaluated. Another limitation is that up to 14 days of CGM data were desired, but only 5 days of CGM data were required for each assessment period. Since the writing of the IGNITE protocol, it is clear that 14 days of CGM data with at least 70% active time are optimal, while 5 days of data may be ok but are less desirable.²⁹ (Of note, during IGNITE, a mean of 12.8 [2.2] days of CGM data was collected at baseline.) Furthermore, this 3-month intervention period was short-term and does not provide insight into the durability of these results over time (however, 6-month follow-up data are forthcoming). Another aspect to consider is both a strength and a limitation. The IGNITE study assessed the impact of glucose monitoring in people who voluntarily chose a carbohydrate-restricted diet program with continuous, remote care for diabetes management. On one hand, this eliminated a common barrier in nutrition research, which is the assignment of a specific dietary intervention. IGNITE participants were motivated and willing to restrict carbohydrates, and they all chose the nutrition intervention before they were invited to participate in the study. On the other hand, these results may not be generalizable to people following different diets with or without continuous remote care.

Future research should explore the impact of CGM to guide high-quality eating patterns with a more moderate carbohydrate load in people with diabetes who do and do not use insulin. It is possible the CGM would be more effective for users with moderate or higher carbohydrate intake because of the potential to identify which carbohydrate types and amounts work best for achieving glycemic goals. Assessing the impact of CGM on dietary interventions with intermittent CGM use could also be of benefit. Additionally, future studies may need to consider whether alternate CGM targets are necessary for certain groups of people with T2D (e.g., low glucose thresholds for people with diabetes who do not use insulin).

Conclusions

Both the CGM and BGM arms saw similar and significant improvements in glycemia and other diabetes-related outcomes during this MSKDP delivered via continuous remote care. TIR improved by a mean of 28% for the CGM arm and 22% for the BGM arm from baseline to 3 months, and most participants achieved >70% TIR.

CGM has been shown to be both beneficial and cost-effective in people with type 1 diabetes and those with T2D using insulin. However, based on these results, it appears CGM may be less impactful for people with T2D who have a low-complexity diabetes medication regimen and are embarking on a well-supported, very-low-carbohydrate diet program; this appears due to the effectiveness of the diet alone on glycemic management. The impact of CGM in people who choose a more moderate carbohydrate diet plan or a very-low-carbohydrate diet plan under other circumstances—such as with complex diabetes medication regimens such as intensive insulin therapy, without a continuous care team, or later in a lifestyle intervention when adherence may wane—is unknown and should be considered in future research.

Footnotes

Acknowledgments

The authors would like to thank McKenzie Granger for the immense efforts that went into shipping all of the study supplies, managing the incoming blinded CGM data, and her all-around support of the study. They also thank the care providers who supported the participants during this MSKDP: Dana Artinyan, RD, CDCES; Kayla Cunningham, MD, MHs, FACE; Brandon Fell, MS, RD; Collin Foster, MS, RD; Lynn Matthusen, MS, RDN; Jennifer McGill, RN, CDCES; Greeshma Shetty, MD; Jay Short, MS, RD; and Detrick Snyder, MPH, RDN, and all of the study participants who gave generously of their time to participate in this research.

Authors’ Contributions

H.J.W.: Conceptualization, methodology, investigation, writing—original draft, project administration, and funding acquisition. S.E.A.: Conceptualization, methodology, formal analysis, data curation, and writing—review and editing. A.L.M.: Conceptualization, methodology, and writing—review and editing. R.N.A.: Investigation, data curation, and writing—review and editing. C.G.P.R.: Investigation and writing—review and editing. B.M.V.: Conceptualization, methodology, and writing—review and editing. S.K.: Investigation and writing—review and editing. S.J.A.: Data Curation and writing—review and editing. A.R.Z.: Investigation and writing—review and editing. R.M.B.: Conceptualization, methodology, investigation, writing—review and editing, project administration, and funding acquisition.

Author Disclosure Statement

H.J.W.: Received research support, consulted with, and/or has been on an advisory board for Abbott Diabetes Care and Dexcom. H.J.W.’s employer, the nonprofit HealthPartners Institute, contracts for her services and no personal income goes to H.J.W. S.E.A.: S.E.A.’s employer, the nonprofit HealthPartners Institute, contracts for his services and no personal income goes to S.E.A. A.L.M.: Former employee of Virta Health and has received stock options; current employee of Abbott’s Lingo Biowearables. R.N.A.: Employee of Virta Health and has received stock options. C.G.P.R.: Employee of Virta Health and has received stock options. B.M.V.: Former employee of Virta Health and has received stock options; current employee with the Ketogenic Foundation. S.K.: No disclosures. S.J.A.: Employee of Virta Health and has received stock options. A.R.Z.: Employee of Virta Health and has received stock options. R.M.B.: Received research support, consulted with, or has been on a scientific advisory board for Abbott Diabetes Care, Ascensia, Dexcom, Lilly, Hygieia, Insulet, Novo Nordisk, Onduo, Sanofi, Tandem, Medtronic, Roche, and Zealand. R.M.B.’s employer, nonprofit HealthPartners Institute, contracts for his services and no personal income goes to R.M.B.

Funding Information

Funding and all study-related supplies needed to conduct this study were provided by Abbott Diabetes Care.

Supplementary Material

Supplementary Table S1

Supplementary Table S2

Supplementary Table S3

Supplementary Figure S1

References

Evert

, Dennison

, Gardner

, et al. Nutrition therapy for adults with diabetes or prediabetes: A consensus report. Diabetes Care, 2019; 42(5):731–754; doi: 10.2337/dci19-0014

American Diabetes Association Professional Practice Committee. Facilitating positive health behaviors and well-being to improve health outcomes: Standards of Care in Diabetes—2024. Diabetes Care, 2024; 47(Suppl 1):S77–S110; doi: 10.2337/dc24-S005

Choy

KYC

, Louie

JCY

. The effects of the ketogenic diet for the management of type 2 diabetes mellitus: A systematic review and meta-analysis of recent studies. Diabetes Metab Syndr, 2023; 17(12):102905; doi: 10.1016/j.dsx.2023.102905

Athinarayanan

, Adams

, Hallberg

, et al. Long-term effects of a novel continuous remote care intervention including nutritional ketosis for the management of type 2 diabetes: A 2-Year Non-randomized Clinical Trial. Front Endocrinol (Lausanne), 2019; 10:348; doi: 10.3389/fendo.2019.00348

Athinarayanan

, Vantieghem

, McKenzie

, et al. 832-P: Five-year weight and glycemic outcomes following a very-low-carbohydrate intervention including nutritional ketosis in patients with type 2 diabetes. Diabetes, 2022; 71(Supplement_1); doi: 10.2337/db22-832-P

American Diabetes Association Professional Practice Committee. Diabetes technology: Standards of Care in Diabetes—2024. Diabetes Care, 2024; 47(Suppl 1):S126–S144; doi: 10.2337/dc24-S007

Ajjan

, Battelino

, Cos

, et al. Continuous glucose monitoring for the routine care of type 2 diabetes mellitus. Nat Rev Endocrinol, 2024; 20(7):426–440; doi: 10.1038/s41574-024-00973-1

Griauzde

, Ling

, Wray

, et al. Continuous glucose monitoring with low-carbohydrate nutritional coaching to improve type 2 diabetes control: Randomized Quality Improvement Program. J Med Internet Res, 2022; 24(2):e31184; doi: 10.2196/31184

Choe

, Rhee

, Won

, et al. Effects of patient-driven lifestyle modification using intermittently scanned continuous glucose monitoring in patients with type 2 diabetes: Results from the Randomized Open-label PDF Study. Diabetes Care, 2022; 45(10):2224–2230; doi: 10.2337/dc22-0764

10.

Battelino

, Alexander

, Amiel

, et al. Continuous glucose monitoring and metrics for clinical trials: An international consensus statement. Lancet Diabetes Endocrinol, 2023; 11(1):42–57; doi: 10.1016/S2213-8587(22)00319-9

11.

Automated Self-Administered 24-Hour (ASA24®) Dietary Assessment Tool. Available from: https://epi.grants.cancer.gov/asa24/ [Last accessed: July 29, 2024 ].

12.

Alexopoulos

, Yancy

, Edelman

, et al. Clinical associations of an updated medication effect score for measuring diabetes treatment intensity. Chronic Illn, 2021; 17(4):451–462; doi: 10.1177/1742395319884096

13.

Polonsky

, Fisher

, Earles

, et al. Assessing psychosocial distress in diabetes: Development of the diabetes distress scale. Diabetes Care, 2005; 28(3):626–631; doi: 10.2337/diacare.28.3.626

14.

Kitabchi

, Umpierrez

, Miles

, et al. Hyperglycemic crises in adult patients with diabetes. Diabetes Care, 2009; 32(7):1335–1343; doi: 10.2337/dc09-9032

15.

Fisher

, Hessler

, Polonsky

, et al. When is diabetes distress clinically meaningful? Establishing cut points for the Diabetes Distress Scale. Diabetes Care, 2012; 35(2):259–264; doi: 10.2337/dc11-1572

16.

Banks

, Amspoker

, Vaughan

, et al. Ascertainment of minimal clinically important differences in the diabetes distress scale–17: A secondary analysis of a randomized clinical trial. JAMA Netw Open, 2023; 6(11):e2342950-e2342950; doi: 10.1001/jamanetworkopen.2023.42950

17.

Battelino

, Danne

, Bergenstal

, et al. Clinical targets for continuous glucose monitoring data interpretation: Recommendations from the international consensus on time in range. Diabetes Care, 2019; 42(8):1593–1603; doi: 10.2337/dci19-0028

18.

Yapanis

, James

, Craig

, et al. Complications of diabetes and metrics of glycemic management derived from continuous glucose monitoring. J Clin Endocrinol Metab, 2022; 107(6):e2221–e2236; doi: 10.1210/clinem/dgac034

19.

Glucose concentrations of less than 3.0 mmol/L (54 mg/dL) should be reported in clinical trials: A joint position statement of the American Diabetes Association and the European Association for the Study of Diabetes. Diabetes Care, 2017; 40(1):155–157; doi: 10.2337/dc16-2215

20.

Bergenstal

. Roadmap to the effective use of continuous glucose monitoring: Innovation, investigation, and implementation. Diabetes Spectr, 2023; 36(4):327–336; doi: 10.2337/dsi23-0005

21.

Aronson

, Brown

, Chu

, et al. IMpact of flash glucose Monitoring in pEople with type 2 Diabetes Inadequately controlled with non-insulin Antihyperglycaemic ThErapy (IMMEDIATE): A randomized controlled trial. Diabetes Obes Metab, 2023; 25(4):1024–1031; doi: 10.1111/dom.14949

22.

Wada

, Onoue

, Kobayashi

, et al. Flash glucose monitoring helps achieve better glycemic control than conventional self-monitoring of blood glucose in non-insulin-treated type 2 diabetes: A randomized controlled trial. BMJ Open Diabetes Res Care, 2020; 8(1):e001115; doi: 10.1136/bmjdrc-2019-001115

23.

Wright

Jr , Subramanian

. Is continuous glucose monitoring a tool, an intervention, or both? Diabetes Technol Ther, 2023; 25(S3):S48–S55; doi: 10.1089/dia.2023.0025

24.

Rafiullah

, Musambil

, David

. Effect of a very low-carbohydrate ketogenic diet vs recommended diets in patients with type 2 diabetes: A meta-analysis. Nutr Rev, 2022; 80(3):488–502; doi: 10.1093/nutrit/nuab040

25.

Zhou

, Wang

, Liang

, et al. Ketogenic diet benefits to weight loss, glycemic control, and lipid profiles in overweight patients with type 2 diabetes mellitus: A meta-analysis of Randomized Controlled Trails. Int J Environ Res Public Health, 2022; 19(16); doi: 10.3390/ijerph191610429

26.

Volek

, Yancy

, Gower

, et al. Expert consensus on nutrition and lower-carbohydrate diets: An evidence- and equity-based approach to dietary guidance. Front Nutr, 2024; 11:1376098; doi: 10.3389/fnut.2024.1376098

27.

Churuangsuk

, Griffiths

, Lean

MEJ

, et al. Impacts of carbohydrate-restricted diets on micronutrient intakes and status: A systematic review. Obes Rev, 2019; 20(8):1132–1147; doi: 10.1111/obr.12857

28.

Kirkpatrick

, Bolick

, Kris-Etherton

, et al. Review of current evidence and clinical recommendations on the effects of low-carbohydrate and very-low-carbohydrate (including ketogenic) diets for the management of body weight and other cardiometabolic risk factors: A scientific statement from the National Lipid Association Nutrition and Lifestyle Task Force. J Clin Lipidol, 2019; 13(5):689–711.e1; doi: 10.1016/j.jacl.2019.08.003

29.

Bailey

, Calhoun

, Bergenstal

, et al. Assessment of the glucose management indicator using different sampling durations. Diabetes Technol Ther, 2023; 25(2):148–150; doi: 10.1089/dia.2022.0284

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.12 MB

0.18 MB

0.17 MB

0.25 MB