Lessons From Continuous Glucose Monitoring in Youth With Pre-Type 1 Diabetes, Obesity, and Cystic Fibrosis.

Glucose abnormalities exist in the prediabetic state well before a diagnosis of diabetes is made, and continuous glucose monitoring (CGM) is a sensitive method to detect these early abnormalities. Although insufficient data exist to support CGM for diagnosing diabetes, glucose patterns on CGM can provide insight into disease pathophysiology. Our group has used CGM to study early dysglycemia in several populations of youth at risk for diabetes—specifically, type 1 diabetes (T1D), type 2 diabetes (T2D), and cystic fibrosis–related diabetes (CFRD). We hypothesized that, in youth at risk for different types of diabetes matched by HbA1c, average sensor glucose would be no different, but specific CGM measures might differ among groups.

For this analysis, we combined data from three groups of youth: 1) antibody-positive (Ab+) children from DAISY (Diabetes Autoimmunity Study in the Young) (1), 2) youth with cystic fibrosis (CF) from the Glycemic Monitoring in Cystic Fibrosis Study (NCT02211235, ClinicalTrials.gov), and 3) overweight/obese youth with BMI ≥85th percentile at risk for T2D (2). Inclusion criteria were ages 10–18 years and HbA1c <6.5% (48 mmol/mol) with concurrent CGM data. Oral glucose tolerance testing (OGTT) was not performed in all youth and, therefore, not included in this analysis. The pre-T1D youth had ≥2 positive autoantibodies; CF youth had a confirmed diagnosis of CF based on newborn screen, sweat chloride testing, and/or genetic testing; and overweight/obese youth had a BMI ≥85th percentile. Exclusion criteria included known diabetes or use of medications affecting glucose metabolism including insulin. As there were fewer DAISY individuals, this group was matched up to 1:2 to CF and overweight/obese participants by HbA1c category, determined by dividing the participants’ HbA1c range into tertiles: 1) 4.9–5.3% (30–34 mmol/mol), 2) 5.4–5.8% (36–40 mmol/mol), and 3) 5.9–6.3% (41–45 mmol/mol).

A total of 108 participants were included (Table 1). As hypothesized, average sensor glucose levels on CGM were no different among groups. However, multiple measures of glycemic variability, including maximum glucose, standard deviation, and mean amplitude of glycemic excursions were significantly higher in youth with CF and pre-T1D.

Table 1

Demographics and clinical characteristics by CF, Pre-T1D, and obese/Pre-T2D group

	CF (N = 42)	Pre-T1D (N = 24)	Obese/Pre-T2D (N = 42)	P value
Demographic variable
Age (years)	13.3 (0.5)	15.6 (0.9)	14.6 (0.4)	0.02a
Male, n (%)	21 (50)	11 (46)	13 (31)	0.25
Race, n (%)				<0.001b
Non-Hispanic white	39 (93)	19 (79)	15 (36)
African American	0 (0)	0 (0)	4 (10)
Hispanic	2 (5)	5 (21)	22 (52)
Other	1 (2)	0 (0)	1 (2)
BMI z-score	−0.08 (0.1)	0.06 (0.2)	3.1 (0.2)	<0.001b
Weight (kg)	45.5 (2.1)	54.9 (3.2)	92.4 (3.8)	<0.001b
Height (cm)	153 (2.2)	162 (3.3)	163 (1.5)	0.001c
HbA1c (%)†	5.6 (0.04)	5.5 (0.08)	5.6 (0.05)	0.89
CGM variable*
Duration of CGM wear (days)	5.2 (0.1)	3.8 (0.1)	2.0 (0.1)	<0.0001a,b
Average glucose (mg/dL)	111 (2)	113 (3)	114 (2)	0.60
Day average glucose (mg/dL)	112 (2)	112 (3)	115 (2)	0.58
Night average glucose (mg/dL)	108 (3)	114 (4)	111 (3)	0.54
Maximum glucose (mg/dL)	208 (7)	202 (9)	162 (7)	0.0001b
Day maximum glucose (mg/dL)	206 (7)	200 (9)	160 (7)	0.0001b
Night maximum glucose (mg/dL)	164 (6)	174 (7)	145 (6)	0.01b
Minimum glucose (mg/dL)	64 (2)	60 (3)	77 (2)	<0.0001b
Day minimum glucose (mg/dL)	66 (2)	61 (3)	81 (2)	<0.0001b
Night minimum glucose (mg/dL)	73 (2)	74 (3)	86 (3)	0.004b
Standard deviation (mg/dL)	22 (1)	24 (3)	16 (1)	0.002b
Coefficient of variation	0.2 (0.01)	0.2 (0.01)	0.14 (0.01)	<0.0001b
Mean amplitude of glycemic excursions	41 (3)	45 (4)	27 (3)	0.006b
Mean of daily differences	22 (2)	26 (2)	17 (2)	0.02b
% Time >140 mg/dL	10 (3)	15 (3)	13 (3)	0.58
% Time >180 mg/dL	1.9 (0.8)	4.1 (1.0)	1.3 (0.7)	0.15
% Time >200 mg/dL	0.9 (0.5)	2.5 (0.7)	0.4 (0.5)	0.12
% Time >250 mg/dL	0.15 (0.3)	1.2 (0.4)	0.01 (0.3)	0.09
% Time <60 mg/dL	1.2 (0.4)	0.8 (0.5)	0.2 (0.4)	0.26
% Time <70 mg/dL	2.6 (0.7)	3.3 (1.0)	0.9 (0.8)	0.12

Data are mean (SE) unless otherwise indicated. Groups were compared using linear mixed models for continuous variables, conditional logistic regression for sex, and the χ2 test for race. Pairwise comparisons were adjusted using the Tukey method. Boldface type indicates statistical significance.

CGM P values adjusted for multiple comparisons to control the false discovery rate at 5%.

Minimum and maximum HbA1c were 5.2–6.2% (CF), 5.0–6.3% (pre-T1D), and 5.0–6.3% (obese/pre-T2D).

Pre-T1D different from CF.

Obese/pre-T2D different from CF and pre-T1D.

Pre-T1D and obese/pre-T2D different from CF.

This is the first report to compare glycemic profiles captured by CGM in three distinct populations of youth at risk for diabetes. These results highlight interesting similarities and differences in pathophysiology among the three disease states. Progressive β-cell failure and insulin deficiency are the primary drivers of disease in T1D and CFRD, and our findings suggest that early insulin insufficiency, as captured by CGM, manifests as glucose excursions and increased glycemic variability in the prediabetic state despite normal mean glucose levels. In T2D, progressive insulin resistance is followed by development of β-cell failure. In early stages of disease, insulin resistance, including hepatic insulin resistance, is the predominant metabolic defect and may be associated with increases in fasting and average glucose relative to normal-weight individuals (2), but with less glycemic variability than seen in youth at high risk for T1D and CFRD. We have previously shown that HbA1c does not underestimate average glucose in youth with CF as previously assumed (3). Here, we make the point that HbA1c may be normal, despite hyperglycemia and increased glycemic variability, not only in individuals with CF but also in Ab+ individuals pre-T1D. Therefore, the tendency for HbA1c to “underestimate” hyperglycemia is not unique to CF but, in fact, merely another example of the “fallacy of average” described by Bergenstal and colleagues (4).

These findings have implications for how we screen for diabetes in different populations. Given the high prevalence of diabetes in individuals with CF and low sensitivity of HbA1c relative to OGTT for diagnosing CFRD, screening with HbA1c is not recommended in this population. Given the relatively lower prevalence of T1D in the general community, routine screening for T1D is not current clinical practice. With autoantibody screening programs (e.g., https://www.askhealth.org/), however, individuals at high risk for progression to T1D are being identified well before symptoms manifest and, as seen in patients with CF, HbA1c in the Ab+ pre-T1D population may be normal despite the presence of hyperglycemia. In contrast, the prevalence of youth-onset T2D is still low. Arguably, given the ease of collecting HbA1c over OGTT, the often-rapid progression to diabetes, and low likelihood of undiagnosed cases of youth-onset T2D, HbA1c remains an adequate screening tool in the obese pediatric population.

In conclusion, CGM informs our understanding of the early stages of diabetes progression in youth with different forms of diabetes. As CGM becomes increasingly accurate, in certain groups with high prevalence of diabetes, it may eventually be the new standard for diagnosing diabetes (5). Longitudinal studies within our research group to determine which CGM variables predict increased risk for disease-specific complications are ongoing.