Overweight and obese children with optimal control in the T1D Exchange Registry: How are they different from lean children with optimal control?

Abstract

Aims:

Increased adiposity is a risk factor for suboptimal diabetes control and cardiovascular disease (CVD) complications.

Our goal was to identify modifiable behavioral characteristics of overweight and obese pediatric patients with type 1 diabetes mellitus (T1DM) who achieve optimal glycemic control and to evaluate their CVD risk compared to lean patients. Our hypothesis was that optimally controlled obese and overweight participants require more total daily insulin and are at higher CVD risk compared to optimally controlled lean participants.

Methods:

We analyzed a cohort of 9,263 participants with T1DM aged<21 years in the T1D Exchange Registry. Optimal diabetes control was defined as HbA1c ≤7.5% (58mmol/mol). We compared factors that influence glycemic control in lean, overweight and obese participants with optimal vs. suboptimal control, using logistic regression.

Results:

Age, race, overweight status, continuous subcutaneous insulin infusion (CSII) and continuous glucose monitoring (CGM) use were important variables influencing glycemic control. In the optimally controlled cohort, 27% of participants were overweight or obese versus 30% in the suboptimally controlled cohort (P<0.001). Overweight and obese participants with optimal control were not significantly different from lean participants in terms of CSII use, total daily insulin dosage per kg of bodyweight, glucose checks per day, boluses with bedtime snack, use of CGM, but had higher LDL cholesterol and triglycerides, and lower HDL cholesterol (P<0.05).

Conclusions:

There were no differences in modifiable behavioral characteristics between the obese, overweight and lean optimally controlled participants. However, predictors of cardiovascular disease were higher in the overweight and obese group.

1.1. Introduction

In the US, approximately 200,000 children live with type 1 diabetes mellitus (T1DM).¹ The American Diabetes Association (ADA) has recently updated its recommendation of a target HbA1c ≤7.5% (58mmol/mol) for all children with T1DM; however, the majority of them don’t achieve this goal.^2,3 An analysis of the T1D Exchange Registry data showed that less than 21% of the 13-20 year old group achieve optimal control ². Poor glycemic control is associated with increased risk of developing micro- and macrovascular complications and higher mortality rates. ^4,5

The alarming increase in global obesity also affects children with T1DM, with an estimated prevalence of 12.5% to 33% ^6–8. Obesity is a risk factor for cardiovascular disease by itself, and potentially makes optimal diabetes control more challenging.^9,10

Our goal was to identify modifiable behavioral characteristics of overweight and obese pediatric participants of the T1D Exchange Registry who achieve optimal glycemic control so that future intervention trials could target these behaviors in order to improve outcomes. We were also interested in the cardiovascular risk of the overweight and obese participants compared to lean individuals. Our hypothesis was that among optimally controlled participants with T1DM, those that are overweight or obese require more intensive insulin and glucose monitoring and are at higher CVD risk compared to those that are lean.

2.1. Subjects, Materials and Methods

Data were collected from the public dataset of the T1D Exchange Registry for the years 2010-2012.Clinical data are obtained from medical chart extraction and self-reported statements. Inclusion criteria were: i) diagnosis of T1DM( as defined in the registry’s criteria of eligibility)¹¹ and ii) age of less than 21 years at consent. Participants were excluded if they had no HbA1c measurement within three months of enrollment, had other missing information (see below) and/or had a body mass index (BMI) >40 kg/m² or a BMI below the 5^th percentile for age and sex¹², in order to avoid potential inclusion of participants with combined type 1 and type 2 diabetes, and to avoid potentially inaccurate values of BMI. Participants were also excluded if they were transgender in order to avoid misclassification of sex variable.

Data were extracted from a total of 17,047 participants (see figure 1). Participants with missing data for bolus and basal insulin or short-acting insulin and long-acting insulin respectively (n=4,009), those who had interchangeably used both multiple daily injections (MDIs) and continuous subcutaneous insulin infusion (CSII) (n=145), participants on fixed dosages of insulin (n= 494) were excluded. Participants with a % total daily bolus dose under 10% or over 90% of total daily insulin dose were also excluded to eliminate the effect of outliers (n=191). Lastly, participants were excluded if information on race was missing (n = 3) and if duration of their diabetes was less than 1 year (n = 886). The remaining 9,263 participants who met inclusion criteria were then separated into distinct groups according to their BMI. Participants with a BMI < 85^th percentile for their age/sex [as defined by the Centers for Disease Control and Prevention (CDC)] were classified as normal weight, 85-94.9^th percentile as overweight and above the 95^th percentile as obese. We also categorized participants into three age groups to capture the majority of prepubertal children (<12 years old), pubertal children (12-17 years old) and children who have transitioned to young adulthood (18-21 years old).

Figure 1: Flow Diagram of cohort selection.

BMI: Body Mass Index, MDI: Multiple Daily Injections

The mean HbA1c was calculated using the most recent available HbA1c measurement (from 6 months prior to 1 month after consent). The analysis was treated cross-sectionally and participants were considered as optimally controlled if the mean most recent HbA1c was less or equal to 7.5% (58mmol/mol). To identify associations between behavioral factors and glycemic control, we extracted data on the following parameters: timing of bolus dose administration, frequency of missing doses, frequency of bolus for day and bedtime snacks, coverage for snacks, frequency of blood glucose monitoring, as well as whether continuous glucose monitoring (CGM) was used. We also collected data on T1DM-related events in the previous 12 months, namely diabetic ketoacidosis (DKA) hospital admissions and frequency of severe hypoglycemic episodes resulting in seizure or loss of consciousness. These were chosen for their high clinical significance and as markers of possible poor compliance with the regimen as well as for their association with CVD risk. We also extracted demographic data on household income, insurance status; we classified households as low income if earned $50,000 and below, mid income between 50,000 to 100,000 and high income if they earned $100,000 and above. For the analysis on lipid measurements, all available lipids measurements were included.

For the descriptive analysis, means and standard deviations were calculated for continuous variables and for the frequency and proportion of categorical variables, respectively. The optimally controlled and suboptimally controlled cohorts were compared using two-sided t-tests and chi-square tests. Further analyses were performed to explore differences associated with daily insulin doses. The total daily dose of insulin was analyzed against method of delivery and glycemic control status within the same insulin delivery method group. One-way ANOVAs and logistic regressions were performed to assess differences after controlling for demographic factors including race, sex, BMI category, and age group. We also adjusted for multiple comparisons using the Holm-Bonferroni correction. Finally, we designed a regression model to identify factors that are important for optimal control in the whole group. P values <0.05 were considered significant. SAS version 9.4 (SAS Institute, Cary, NC) was used for the cohort construction and R 3.3 (. R Foundation for Statistical Computing, Vienna, Austria) for the data analysis.

3.1. Results

Demographic characteristics of all optimally and suboptimally controlled 9,263 participants are presented in the additional files. There were significantly fewer optimally controlled participants in the 12-17-year-old subgroup, among Black and Hispanic individuals, in the overweight and obese subgroups, in females and in MDI users (Table A.1). In the optimally controlled cohort, there were more pump users (58.7% vs 46.3%, P<0.001) and they used significantly less insulin (0.85±0.4 vs 0.97±0.46 units/kg/day), P<0.001), even after adjusting for age, sex, race, duration of diabetes and insurance status. Also, participants in the optimally controlled cohort checked their blood glucose more frequently (6.5± 2. 6 vs 5.6± 2.3 times per day, P<0.001) and a higher proportion of the them used CGM (5.5% vs 2.3%, P<0.001). The optimally controlled cohort had fewer severe hypoglycemic episodes (11.6% vs 17.7%, P<0.001) and hospitalizations for DKA (0.03% vs 0.15%, P<0.001) in the previous 12 months. These results also remained significant after adjusting for age, BMI, race, sex, T1DM duration, income, insurance and multiple comparisons, except for the severe hypoglycemia events (Additional File 2). Finally, the suboptimally controlled cohort had higher mean low density lipoprotein (LDL) cholesterol and total cholesterol and triglycerides levels, but no difference was found in the mean high density lipoprotein (HDL) cholesterol compared to the optimally controlled cohort (Table B.1).

The descriptive statistics of optimally controlled participants according to their BMI status is shown in Table 1. Adolescents (12-17 years old) were more frequently overweight or obese compared younger study participants (P<0.01). Also, obesity rates are higher among Black (23.3%) and Hispanic children (15%), compared to white children (8.1%), (P<0.001).

Table 1:

Descriptive characteristics of the optimally controlled participants according to their BMI

Variable	Normal n=1720	Overweight n=430	Obese n=217	P
HbA1c %	6.98 (0.44)	7.04 (0.4)	6.99 (0.5)	0.077
HbA1c (mmol/mol)	53 (0)	53 (0)	53 (0)	0.077
Age n (%)				0.010
< 12 years	719 (73.7)	152 (15.6)	104 (10.7)
12 – 17 years	733 (71.6)	200 (19.5)	91 (8.9)
18 – 20 years	268 (72.8)	78 (21.2)	22 (6)
Race n (%)				<0.001
1.White Non-Hispanic	1484 (73.8)	364 (18.1)	162 (8.1)
2.Black Non-Hispanic	24 (55.8)	9 (20.9)	10 (23.3)
3.Hispanic or Latino	138 (66.7)	38 (18.4)	31 (15)
4.Native Hawaiian/Other Pacific Islander	0 (0)	1 (33.3)	2 (66.7)
5.Asian	36 (78.3)	6 (13)	4 (8.7)
6. American Indian/Alaskan Native	1 (33.3)	0 (0)	2 (66.7)
7.More than one race	37 (67.3)	12 (21.8)	6 (10.9)
Sex n (%)				0.042
Female	789 (74.2)	195 (18.3)	80 (7.5)
Male	931 (71.5)	235 (18)	137 (10.5)
BMI (kg/m2)	19.34 (2.99)	23.71 (3.47)	27.33 (4.84)	<0.001

Values are represented as mean +/− SD. Values in parentheses represent percent % or the standard deviation SD. Adjusted p values based on logistic regression of overweight/obese controlling for age, race, sex, DM duration, income, insurance, mean Hba1C and multiple comparisons.

In terms of clinical management, we found frequent pump use in the optimally controlled participants across all BMI groups, with slightly lower rates of use among obese children [58.9% among participants with normal weight, 64% with overweight,47% with obesity (P<0.001)]. However, after adjustment for age, race, sex, T1DM duration, income, and insurance and multiple comparisons, this difference lost statistical significance. The total daily insulin dose expressed as units per bodyweight per day was similar across all BMI categories (P=0.096) within the optimally controlled cohort. The use of CGM was not very common among optimally controlled children [5.6% among normal weight children 6.5% of overweight and 2.8% of obese children (P=0.233)]. Hypoglycemic episodes were more frequent among overweight (17.4%) compared to normal weight (10.3%) and obese children (11.3%) (P=0.013), but this difference did not persist after adjustment for demographic and socio-economic factors and multiple comparisons (P=0.864). Despite their optimal control, overweight and obese children had higher concentrations of LDL cholesterol (P=0.021) and triglycerides (P<0.001) and lower levels of HDL cholesterol (P<0.001) compared to normal weight children (Table 2). This relationship persisted for HDL and triglycerides after adjustment for socioeconomic and demographic factors, but not for LDL. No differences were seen in the number of DKA admissions, the frequency of boluses for bedtime snack, consumption of a bedtime snack or the level of exercise. The behavioral and CVD risk characteristics of participants according to BMI are shown in Table 2.

Table 2:

Behavioral and CVD risk characteristics of the optimally controlled participants according to their BMI

Variable	Normal n=1720	Overweight n=430	Obese n=217	P	Adjusted P
Insulin Method (%)				<0.001	1.000
CSII	1013 (58.9)	275 (64)	102 (47)
MDI	707 (41.1)	155 (36)	115 (53)
Total daily insulin dose, (units/kg/day)	0.84 (0.4)	0.89 (0.37)	0.86 (0.43)	0.096	1.000
Number Meter Check	6.62 (2.57)	6.46 (2.4)	6.12 (2.77)	0.020	1.000
CGM Used (%)				0.233	1.000
1.Yes	97 (5.6)	28 (6.5)	6 (2.8)
2.No	1588 (92.3)	395 (91.9)	209 (96.3)
3.Unknown	35 (2)	7 (1.6)	2 (0.9)
Severe Hypoglycemic Events (%)				0.013	0.864
1.Yes	130 (10.3)	54 (17.4)	18 (11.3)
2.No	1131 (89.5)	256 (82.3)	141 (88.7)
3.Don’t know	3 (0.2)	1 (0.3)	0 (0)
Number of Hospitalizations for DKA	0.03 (0.29)	0.04 (0.2)	0.06 (0.48)	0.192	1.000
Number of Bolus per Day	6.65 (3.97)	6.56 (5.13)	6.48 (3.94)	0.817	1.000
Frequency of Missing Dose				0.265
1.Never	1064 (61.9)	245 (57)	146 (67.3)
2.Rarely	431 (25.1)	130 (30.2)	47 (21.7)		1.000
3.Sometimes	179 (10.4)	44 (10.2)	20 (9.2)		1.000
4.Often	36 (2.1)	8 (1.9)	3 (1.4)		1.000
5.Very often	6 (0.3)	0 (0)	0 (0)		1.000
6.At least once a day	3 (0.2)	3 (0.7)	1 (0.5)		1.000
Frequency of Bolus Bed Time Snack (%)				0.884
1.Never	119 (11.2)	30 (11.9)	19 (15.8)
2.Rarely	81 (7.7)	22 (8.7)	10 (8.3)		1.000
3.Sometimes	173 (16.4)	44 (17.4)	21 (17.5)		1.000
4.Most of the time	252 (23.8)	54 (21.3)	24 (20)		1.000
5.Always	433 (40.9)	103 (40.7)	46 (38.3)		1.000
Eat Bed Time Snack (%)				0.280	1.000
1.Yes	1059 (61.9)	252 (59.2)	120 (55.8)
2.No	638 (37.3)	172 (40.4)	92 (42.8)
3.Don’t know	13 (0.8)	2 (0.5)	3 (1.4)
LDL (mg/dl)	84.32 (23.94)	86.37 (23.13)	90.06 (28.42)	0.021	0.182
HDL (mg/dl)	58.23 (14.53)	53.99 (12.85)	51.4 (14.06)	<0.001	<0.001
Total Cholesterol (mg/dl)	157.64 (29.51)	157.26 (28.09)	162.04 (35.61)	0.355	1.000
Triglycerides (mg/dl)	80.27 (47.83)	89.62 (72.09)	97.42 (58.96)	<0.001	0.005
Days of Exercise per week	5.28 (1.68)	5.12 (1.64)	5 (1.86)	0.081	0.987

Values are represented as mean +/− SD. Values in parentheses represent percent % or the standard deviation SD. Adjusted p values based on logistic regression of overweight/obese controlling for age, race, sex, DM duration, income, insurance and multiple comparisons.

MDI: Multiple Daily Injections,, CGM: Continuous glucose monitoring , , CSII: continuous subcutaneous insulin infusion

Finally, in the logistic regression model we found that age, sex, race, overweight status, duration of diabetes, family income, insurance type, method of insulin delivery (MDI vs CSII), number of glucose checks, use of CGM, history of DKA, significantly affected glycemic control (See table 3 and figure 2).

Table 3:

Logistic regression model for factors that are important for optimal control

Variable	OR	CI	P
Sex			0.011
Male	1.14	[1.04,1.27]
Female	-ref
BMI
Normal (5-84.9th %)	-ref
Overweight (85th-94.9th %)	0.87	[0.77,0.99]	0.039
Obese (≥95th %)	0.87	[0.73,1.04]	0.134
Age	1.09	[1.07,1.11]	<0.001
Race
White	-ref
Black	0.46	[0.33,0.65]	<0.001
Hispanic	1.13	[0.94,1.35]	0.201
Asian	1.51	[1.07,2.36]	0.020
Other	0.68	[0.50,0.93]	0.011
Diabetes Duration (per 2 years)	0.84	[0.81,0.87]	<0.001
Annual Income
High	-ref
Low	0.61	[0.52,074]	<0.001
Mid	0.70	[0.63,0.80]	<0.001
Not Present	0.81	[0.70,0.94]	0.003
Insurance
Medicaid	-ref
Medicare	0.89	[0.63,1.24]	0.476
No Coverage	0.862	[0.34,1.85]	0.688
Other	1.28	[1.02,1.61]	0.036
Private	1.50	[1.25,1.79]	<0.001
Unknown	1.20	[0.93,1.55]	0.134
Total daily insulin dose (units/kg/day)	0.54	[0.47,0.62]	<0.001
Method
CSII	-ref
MDI	0.77	[0.68,0.86]	<0.001
Bed Time Snack
Yes	-ref
No	1.10	[0.99,1.23]	0.065
Number of Hospital DKA’s	0.51	[0.41,0.63]	<0.001
Number of Times Meter Checked	1.18	[1.16,1.21]	<0.001
CGM Not Used	0.51	[0.39,0.66]	<0.001

OR=Odds ratio, CI=Confidence Interval, Diabetes duration refers to participants that had diabetes for at least one year. Model regresses the odds of controlled Hba1C < 7.5%(58mmol/mol] against sex, BMI, age, race, income of the family, insurance, insulin method and insulin amount used, number of hospital DKA’s, number of meter checks, CGM use, and if bedtime snack was taken.

BMI: Body Mass Index, CGM: Continuous glucose monitoring, DKA: diabetic Ketoacidosis , CSII: continuous subcutaneous insulin infusion, MDI: Multiple Daily Injections

Figure 2: Forest plot of the multiple linear regression with controlled HgA1c<7.5% (58 mmol/mol) as dependent variable.

MDI: Multiple Daily Injections, CGM: Continuous glucose monitoring, DKA: diabetic ketoacidosis

4,1. Discussion

Poor glycemic control is associated with a higher risk of diabetes complications, such as retinopathy, cardiovascular disease and diabetic kidney disease¹³ Children with T1DM who are obese or overweight have an additional risk factor for diabetes complications ^9,10 and identifying ways to improve their glycemic control is important in order to minimize this risk. To our knowledge, the effects of demographic and behavioral factors on optimal glycemic control optimal glycemic control defined as HbA1c ≤7.5% (58mmol/mol) has not been previously studied. In this analysis, we confirmed that overweight status was associated with sub optimal control. Our study additionally highlights that within the optimally controlled cohort, whether of normal weight, overweight or obese, use of CSII, CGM and frequency of blood glucose checks were not different. Likewise, optimally controlled participants had similar daily insulin requirements per kg body weight across all BMI categories. In contrast, the optimally controlled obese youth with T1DM had lipid abnormalities that are predictive of higher cardiovascular risk compared to the lean participants.

4.1.1. BMI and glycemic control

The role of BMI in predicting glycemic control is controversial and overall BMI has been shown to have a negative or no effect in glycemic control ^14–20. In our sample, more overweight and obese participants (27%) were in the cohort with suboptimal blood glucose control than in the well controlled cohort (30 %). In addition, we found that overweight status is an important factor in the logistic regression model for control status. It is possible that the smaller sample size of obese youth compared to overweight youth was not enough to reach statistical significance in the logistic regression model; we suggest that this relationship is further explored by future large scale studies. The relationship between adiposity and higher insulin requirements as a result of insulin resistance may negatively impact diabetes management.²¹ In a Finnish cohort of 105 children and adolescents with T1DM, higher BMI correlated with worse glycemic control.²² No apparent link of BMI with HbA1c was found in the longitudinal and cross-sectional analyses of 340 children aged 9.0 to 14.9 years at four pediatric endocrinology clinics in the US.¹⁹ In contrast, evidence from the Search for Diabetes in Youth study for the years 2001-2005 showed worse glycemic control in underweight or normal youth with T1D.¹⁷ Authors speculated that poorer residual beta cell function in children with low BMI may play a role. Based on our own data, we recommend that physicians should promote healthy weight maintenance in light of the short and long-term complications associated with increased adiposity.³

4.1.2. BMI and insulin delivery method, use of CGM and blood glucose checks

When comparing the two insulin delivery methods across the BMI subgroups in the optimally controlled cohort we found high rates of use of CSII in all subgroups, although CSII was used less frequently among obese participants. This finding suggests that increased use of CSII, along with the use of CGM and frequent blood glucose checks, can be as effective in overweight as normal-weight youths. The logistic regression model highlighted that the method of insulin delivery was also an important factor for optimal control and the use of MDI was associated with poorer control compared to CSII. Several studies have previously reported lower HbA1c levels in CSII users compared to MDI users.^23–25 In 2008, a meta-analysis of randomized control trials compared the short-term metabolic control of children with T1DM managed with CSII and MDI and concluded that the use of CSII had a favorable effect on HbA1c. Additionally, in that meta-analysis insulin requirements were lower by 0.22 units/kg/day ²⁶. Similar results were reported when data from three large pediatric diabetes registries were compared.²⁴ Among 54,410 children with T1DM, HbA1c levels were on average 0.5% lower in children receiving insulin via insulin pumps²⁵. Data from the T1D Exchange Registry for children <6 years old demonstrated a statistically significant reduction of HbA1c by 0.2% after initiation of CSII regimen among previous MDI users that was independent of anthropometric and socioeconomic factors.²³

CGM use was not very popular among children in our cohort (roughly 6%) , which could be due to older versions of CGMs (used in 2010) that were perhaps not as user friendly as the newer versions of CGM that are available today. Overall, CGM use was similar among all BMI subgroups. Inconsistent results have been published about the use of CGM in improving glycemic control in pediatric patients²⁷. In a large randomized study by the Juvenile Diabetes Research Foundation (JDRF), the use of CGM was not associated with improved HbA1c among the pediatric population.²⁸ However, this could be partially explained by inconsistent daily use, which remains a challenge for many patients.^29,30

4.1.3. BMI and total daily dose of insulin

Our analysis revealed that optimally controlled participants across all BMI subgroups used the same amount of total daily insulin per kg even after adjusting for age, race, duration of diabetes, income, sex, insurance and multiple comparisons Although this constitutes a borderline statistical trend (p=0.096), it failed to reach statistical significance despite the large number of participants. This suggests that perhaps BMI is not the best marker to characterize the metabolic aspects of obesity and estimate insulin sensitivity in youth with T1DM. Another explanation could be similar insulin sensitivity between the three groups, despite their differences in BMI. Decreased insulin sensitivity has been documented among participants with T1DM, although the precise mechanisms behind it are not fully elucidated.³¹ Two recent studies showed the beneficial effects of metformin as insulin sensitizer in youth with T1DM. Both studies showed that metformin improved vascular health in children with T1DM and decreased their insulin requirements, indicating that addressing insulin sensitivity could be beneficial for cardiovascular health of youth with T1DM ^32,33. We were unable to further address the question of insulin sensitivity in this study, because neither anthropometric measures such as waist circumference nor serial testing (frequently samples oral or venous glucose tolerance tests or clamp studies) are available in the T1D Registry, and merits further studying by large scale cohorts.

4.1.4. BMI and dyslipidemia

We found that overweight and obese children with optimal blood glucose control had worse lipid profiles than those with normal weight. It is known that children with T1DM have a higher cardiovascular disease risk compared to healthy children and that poor glycemic control and obesity are important contributors to this risk.^10,34 Participants who followed an intensive insulin regimen during the DCCT trial gained more weight than the participants who had a less aggressive regimen. This could be attributed to increased lipogenesis with intensive therapy.³⁵ However, in our analysis, the similar insulin requirements among different BMI groups of optimally controlled participants indicate that other factors are important, rather than insulin itself. For example, poor diet, decreased exercise, napping, increased screen time, not having regular breakfast and dinner or genetic factors may play a role.⁸ Therefore, counselling children with T1DM to maintain a healthy lifestyle is clearly important to achieve optimal glycemic control and reduce cardiovascular risk.

4.1.5. Non-modifiable factors for optimal control

Finally, we found that age, sex and race were important factors in our logistic regression model that determine optimal control, which is in agreement with other studies. Adolescence has previously been reported as a challenging period for diabetes management ^3,36–38 which may be attributed to decreased insulin sensitivity due to the effect of growth hormone and other puberty related hormonal changes. Adolescent females have been identified as a high-risk group for poor metabolic control.^9,36,39–45 Regarding race and ethnicity, significantly higher mean HbA1c levels have been reported in black children with T1DM compared to white and non-black Hispanic children, even after adjustment for socioeconomic factors, diabetes duration, age and sex.^46–49 Limited access to healthcare, lower rates of enrollment to health insurance ⁵⁰ and lack of prescription coverage ⁵¹ may significantly limit successful diabetes management. Notable disparities in the insulin delivery methods have also been reported.^46,52 In a US multicenter study involving 2,743 participants <20 years old, 52.7% of the non-Hispanic white children were treated with CSII compared to 19.1% of black children and 27.8% of Native American children.³⁸

4.1.6. Stengths and limitations

Our study’s main strengths are the quality and the large size of available data provided by the T1D Exchange Registry. We analyzed a large cohort of 9,263 well-characterized children and adolescents with T1DM who received care in a wide range of settings throughout the US. Limitations include that black children are underrepresented in this registry (5.9%), as are children aged less than 6 years. Also, data collected in this database are self-reported and participants are self-enrolled, which might have introduced some errors. Moreover, the database provides limited information about the insulin sensitivity and C-peptide measurements of participants and further adjustments for these parameters were not possible. Since information about pubertal status was not available, age groups were created as an approximation to include most children in the prepubertal, pubertal and post-pubertal era. Further breakdown of age groups by sex to account for differences in age of puberty was beyond the scope of this paper. Finally, since 2014 the ADA recommends that a HbA1C target of <7.5% (58 mmol/mol) should be considered for all children and adolescents. Data in this cohort are from participants enrolled from 2010-2012 who were treated by their providers using a more relaxed HbA1c goal of less than 8.5% (69 mmol/mol) for children under 6 years of age, less than 8.0% (64mmol/mol) for children 6-12 years of age and less than 7.5% (58mmol/mol) for adolescents. It is therefore possible that nowadays providers recommend more intensive insulin regimens are used across all ages, and that some of our results might not be representative of the current pediatric practice. However, because we used the newest definition of HbA1c target to define optimal control, we believe our results can guide current management of overweight and obese children with T1DM. Furthermore, although an HbA1C target of <7.0% is recommended in non-pregnant adult populations, the designated target for children and adolescents was used for all participants included to ensure consistency. Finally, although fasting lipid measurements are traditionally used for screening, there were not available fasting measurements for all participants, so all measurements were included in the analysis. Although this is a limitation, there is a recent trend towards interchangeable use of fasting and non-fasting for metabolic screening for patients who are not receiving therapy for hyperlipidemia and who are not being investigated for inherited lipid conditions.⁵³ In a recent nationwide cross-section pediatric study, the comparison of fasting and non-fasting measurements did not yield clinically significant differences.⁵⁴

5.1. Conclusions

In summary, we found that in the group with HbA1c ≤ 7.5% (58mmol/mol) there were no significant differences in use of CSII, CGM and frequent blood glucose checks between overweight and obese participants as compared to the lean participants. However, predictors of cardiovascular disease were higher in the overweight and obese group. The use of CSII versus MDI, use of CGM and the multiple checks of blood glucose are important modifiable behaviors to achieve optimal glycemic control in lean overweight and obese children with T1DM. Furthermore, overweight and obese children with optimal glycemic control don’t have to use more insulin/kg/day to achieve their glycemic goal. Future studies can further investigate the role of CSII use, insulin sensitivity and increased residual C-peptide levels in maintaining excellent control in overweight/obese participants. Finally, counselling about lifestyle changes to maintain a healthy weight is important for the cardiovascular health of all overweight and obese children with optimal glycemic control.

Highlights.

We analyzed a cohort of 9,263 children and adolescents with T1DM in the US.

Overweight/obese children with optimal GC use the same insulin/kg/day as lean ones

There is increased CVD risk among the overweight/obese group compared to the lean one

The use of CSII, CGM and the multiple checks of BG are important modifiable behaviors

Abbreviations

BG

Blood glucose

CSII

continuous subcutaneous insulin infusion

CVD

cardiovascular disease

CGM

continuous glucose monitoring

GC

glycemic control

T1DM

Type I diabetes mellitus

Table A1:

Descriptive characteristics of the optimally and suboptimally controlled participants

Variable	Optimal control (n=2367)	Suboptimal control (n=6896)	P
HBA1c	6.99 (0.44)	8.92 (1.28)	<0.001
Age (%)			0.009
< 12	975 (41.2)	2763 (40.1)
12 - 17	1024 (43.3)	3208 (46.5)
18 - 20	368 (15.5]	925 (13.4]
Race (%)			<0.001
1.White Non-Hispanic	2010 (84.9]	5469 (79.3]
2.Black Non-Hispanic	43 (1.8]	386 (5.6]
3.Hispanic or Latino	207 (8.7]	661 (9.6]
4.Native Hawaiian/Other Pacific Islander	3 (0.1]	15 (0.2]
5.Asian	46 (1.9]	87(1.3]
6.American Indian/Alaskan Native	3 (0.1]	35 (0.5]
7.More than one race	55(2.3]	243 (3.5]
Sex (%)			0.009
Female	1064 (45]	3335 (48.4]
Male	1303 (55]	3561 (51.6]
BMI (%)	20.86 (4.22]	21.32 (4.47]	<0.001
Normal (5-84.9th %]	1720 (72.7]	4617 (67]
Overweight (85th to 94.9th %]	430 (18.2]	1464 (21.2]
Obese (≥95th %]	217 (9.2]	815 (11.8]

Values are represented as mean +/− SD. Values in parentheses represent percent % or the standard deviation (SD).

Table B1:

Behavioral and clinical characteristics of the optimally and suboptimally controlled participants

Variable	Optimal control (n=2367)	Suboptimal control (n=6896)	P	Adjusted P.
Insulin Method (%)			<0.001	<0.001
Pump	1390 (58.7)	3190 (46.3)
MDI	977 (41.3)	3706 (53.7)
Insulin Amount, units/kg/day	0.85 (0.4)	0.97 (0.46)	<0.001	<0.001
Number Meter Check	6.54 (2.56)	5.61 (2.29)	<0.001	<0.001
CGM Used (%)			<0.001	<0.001
1.Yes	131 (5.5)	161 (2.3)
2.No	2192 (92.6)	6659 (96.6)
3.Unknown	44 (1.9)	76 (1.1)
Severe Hypoglycemic Events (%)			<0.001	0.138
1.Yes	202 (11.6)	920 (17.7)
2.No	1528 (88.1)	4225 (81.5)
3.Don’t know	4(0.2)	40 (0.8)
Number of Hospitalizations for DKA	0.03 (0.3)	0.15 (0.65)	<0.001	<0.001
Number of Bolus per Day	6.41 (4.39)	6.61 (4.05)	0.143	<0.001
Missing Insulin Dose			<0.001
1.Never	3001 (43.5)	1455 (61.5)
2.Rarely	1693 (24.6)	608 (25.7)		<0.001
3.Sometimes	1299 (18.8)	243 (10.3)		<0.001
4.Often	631 (9.2)	47 (2)		<0.001
5.Very often	143 (2.1)	6 (0.3)		<0.001
6.At least once a day	119 (1.7)	7 (0.3)		<0.001
Frequency of Bolus Bed Time Snack (%)			<0.001
1.Never	168 (11.7)	564 (12.7)
2.Rarely	113 (7.9)	389 (8.7)		1.000
3.Sometimes	238 (16.6)	1013 (22.8)		0.463
4.Most of the time	330 (23.1)	1075 (24.2)		1.000
5.Always	582 (40.7)	1409 (31.7)		0.058
Bed Time Snack (%)			0.002	0.138
1.Yes	1431 (60.9)	4426 (65)
2.No	902 (38.4)	2345 (34.4)
3.Don’t know	18 (0.8)	43 (0.6)
LDL	85.24 (24.29)	91.57 (27.89)	<0.001	<0.001
HDL	56.79 (14.38)	57.41 (14.45)	0.141	0.463
Total Cholesterol	157.99 (29.88)	165.27 (32.12)	<0.001	<0.001
Triglycerides	83.63 (54.62)	97.53 (85.77)	<0.001	<0.001
Days of Exercise per week	5.22 (1.69)	5.2 (1.72)	0.645	1.000

Values are represented as mean +/− SD. Values in parentheses represent percent % or the standard deviation SD. Adjusted p values based on logistic regression controlling for age, BMI race, sex, DM duration, income, insurance and multiple comparisons.