Increased Technology Use Associated With Lower A1C in a Large Pediatric Clinical Population

Trends in pediatric glycemic target achievement deteriorated between 2010–2012 and 2016–2018 across age-groups. Newer-generation continuous glucose monitors (CGMs) and hybrid closed-loop (HCL) systems result in improved A1C in clinical trials (1–6). The first HCL system became available in the U.S. in spring 2017, followed by the first factory-calibrated CGM in 2018 and a second HCL system in 2020 (7–9).

To date, real-world evidence of change in A1C associated with these technologies in children has been limited by study size, lack of a nontechnology-user comparison group, or separate analysis of HCL users (10–13). We evaluated A1C trends in children and young adults with type 1 diabetes between two 6-month periods in 2016–2017 and 2020–2021 at a single, large U.S. center, assessing whether diabetes technology use was associated with lower A1C.

This study was approved by the Colorado Multiple Institutional Review Board (20-2686), which determined it met criteria for exemption from Institutional Review Board review (category 4, secondary research). All criteria were met for a full waiver of Health Insurance Portability and Accountability Act authorization.

Electronic medical records at the Barbara Davis Center for Diabetes were queried for patients aged <22 years, with type 1 diabetes, who attended the pediatric clinic between October 2016 through March 2017 or October 2020 through March 2021, had diabetes duration >3 months, and had available laboratory-measured A1C and diabetes technology data at the same clinical encounter. All patients are offered CGM and insulin pumps, and there are no requirements for failure of previous therapy for technology eligibility. Our clinic widely adopted HCL systems when commercially available, although we had participants in our clinic using all three commercial systems during their development and U.S. Food and Drug Administration (FDA) approval trials. All FDA-approved technologies were covered by insurers in our area.

Patient demographics, A1C, insulin regimen, and use of CGM, CGM time in range 70–180 mg/dL (3.9–10 mmol/L), insulin pump, and/or HCL use at the most recent clinical encounter within each 6-month period were extracted from the medical record. If CGM, pump, or HCL use were not documented in discrete fields, our study team reviewed the device download systems to confirm use. If technology use remained undetermined, we used the preceding encounter within the time period. Two patients who did not use multiple daily injections (MDIs) were excluded, as were 35 in each time period whose insulin regimen remained unclear after review. CGM users were defined as those with a nonzero time on CGM in the 14 days preceding the most recent visit; all others were classified as a blood glucose meter (BGM) user. HCL users were defined as those whose medical record documentation indicated HCL use. Thus, there were five groups: MDI/BGM, pump/BGM, MDI/CGM, pump/CGM (without HCL), and HCL.

A1C values were modeled with a linear mixed-effects model for each time segment (2016–2017, 2020–2021) and technology group with an interaction term of time and technology, adjusting for age, diabetes duration, sex, insurance, and race/ethnicity. A random intercept for subject was added to account for subjects with repeated measurements. Contrasts of estimated marginal means (least squares means) were used to determine difference within technology groups at each time point.

The primary analysis included all patients. Three subanalyses included 1) only patients who had data in both time periods, 2) only patients who remained in the same technology group, and 3) only patients who switched between groups.

Analyses were performed using R 4.1 software (R Core Team, Vienna, Austria).

A total of 4,103 patients met inclusion criteria, 2,827 in 2016–2017 and 2,731 in 2020–2021. Of these, 1,455 patients had data in both time periods (Supplemental Table 1). Group selection patterns are shown in Fig. 1 and patient characteristics by technology use group in Table 1.

From 2016–2017 to 2020–2021, there were increases in pump/CGM (26.9% vs. 28.3%), HCL (0.3% vs. 27.9%), and MDI/CGM groups (5.8% vs. 19.1%), while other groups decreased: pump/BGM (34.3% vs. 9.8%) and MDI/BGM (32.8% vs. 14.9%). The proportion of patients using CGM increased (32.9% to 75.3%), and the percentage using insulin pumps increased (61.4% to 66.0%) between time periods. In 2020–2021, among HCL users, mean and median HCL use was 52% and 69%, respectively, although four users’ data indicated HCL use without ascertainable use percentage.

Patients identifying as non-Hispanic, Black, or Hispanic or patients with Medicaid insurance, used CGM, pumps, and HCL less frequently than non-Hispanic White and privately insured patients in both time periods. Increases in CGM and pump use were similar between the groups, but HCL use was much lower in Hispanic and Medicaid insured patients than non-Hispanic White and privately insured patients in 2020–2021.

Mean A1C decreased over time (8.9% [73.8 mmol/mol] to 8.6% [70.5 mmol/mol], P < 0.0001) using linear mixed-effects model for all patients regardless of technology use group (Table 1 and Fig. 2). A1C increased in the MDI/BGM group (9.4% [79.2 mmol/mol] to 9.9% [84.7 mmol/mol], P < 0.0001) and in the pump/BGM group (9.2% [77.0 mmol/mol] to 9.8% [83.6 mmol/mol], P = 0.003). There was no significant within-group A1C change for the MDI/CGM, pump/CGM, or HCL groups.

When including only patients with data in both time periods (Fig. 3 and Supplemental Table 1), A1C was unchanged over time by linear mixed-effects model, while A1C for the BGM groups increased: pump/BGM (9.1% [76.0 mmol/mol] to 9.8% [83.6 mmol/mol], P = 0.01) and MDI/BGM (9.1% [76.0 mmol/mol] to 10.3% [89.1 mmol/mol] P < 0.0001). There was no significant difference between times for the MDI/CGM, pump/CGM, or HCL groups.

Of 493 patients who remained in the same technology group, mean A1C increased from 8.8% (72.7 mmol/mol) to 9.3% (78.1 mmol/mol; P < 0.0001) (Supplementary Fig. 1). When linear mixed-effects modeling was used, A1C increased from 9.2% (77.0 mmol/mol) to 10.2% (88.0 mmol/mol) in the MDI/BGM group (P < 0.0001), while the other groups showed no difference in A1C over time (Supplemental Table 1).

Of the patients with data in both time periods, 962 (66%) switched groups, with 20 possible combinations. No patients switched from HCL to MDI/BGM, MDI/CGM, or pump/BGM. In the linear mixed-effects model including patients who switched, A1C decreased for pump/BGM to HCL (8.9% [73.8 mmol/mol] to 8.0% [63.9 mmol/mol], P < 0.001) and from pump/CGM to HCL (8.0% [63.9 mmol/mol] to 7.6% [59.6 mmol/mol], P = 0.001), while A1C increased with switch from pump/CGM to pump/BGM (9.1% [76.0 mmol/mol] to 10.1% [86.9 mmol/mol], P = 0.021) (Supplementary Fig. 2). The remaining comparisons were not significantly different.

Between the 2016–2017 period and the 2020–2021 period, mean A1C across our pediatric clinic decreased by 0.4% (4.4 mmol/mol), associated with sharp increases in CGM and HCL uptake. Insulin pump use increased slightly.

In the cross-sectional analysis comparing the entire clinic cohorts in the two periods, A1C was higher in 2020–2021 in the groups not using CGM. As all patients are offered CGM, individuals who did not adopt the new technology may represent a higher-risk group. We hypothesize that the A1C increase in patients not on CGM may be because patients with fewer care barriers started CGM. Compared with non-Hispanic Whites, Hispanic and non-Hispanic Black people in the U.S. are at higher risk of poverty and poorer health outcomes (14). It was encouraging that minority and poor patients had similar increases in pump and CGM use, but more needs to be done to close gaps and improve HCL uptake for these groups.

Although CGM technology improved between the periods, patients using CGM had similar A1C across the time periods, a trend that remained when restricting analysis to those who had data in both periods. Thus, the overall decrease in A1C across the population is consistently associated with the relative increases in the size of CGM and HCL groups, in line with previous cross-sectional findings in our clinic (15). Although the first FDA-approved HCL did not become commercially available in the U.S. until early 2018, eight patients in the 2016–2017 cohort were in clinical research studies or were using do-it-yourself systems.

In the 2016–2018 T1D Exchange registry cohort, 6% and 15% of patients aged 6–17 years reached A1C <7% (53 mmol/mol) and <7.5% (58.5 mmol/mol), respectively, whereas our 2020–2021 cohort were 20% and 33%, respectively (16). Although our current trends remain higher than reports from European registries, they represent an improvement for a U.S. cohort. Higher A1C among our non-HCL groups and those in the T1D Exchange Quality Improvement registry than in all similar groups in the German/Austrian/Luxembourgian Diabetes-Patienten-Verlaufsdokumentation registry indicates that there are insufficiently addressed factors beyond technology use affecting glycemia in our clinic and in other U.S. pediatric diabetes clinics (13).

Contrary to the primary analysis using all patients, mean A1C significantly increased among patients who remained in their same technology group. This is likely attributable to the fact that the majority in this analysis were not using CGM and only two patients were using HCL.

Among patients who switched groups, significant A1C decreases only occurred in those who initiated HCL; significant increase was only among pump users who discontinued CGM. These data further support the associations of CGM and HCL use with lower A1C.

Notably, CGM-using groups were younger with shorter diabetes duration than those using BGM, which may reflect provider or family preferences. This represents a target of opportunity for increasing CGM uptake among older adolescents, the age-group previously shown to have the highest A1C (16).

Strengths of this analysis include the large number of patients and the temporal comparison of clinic-wide trends from before the introduction of nonadjunctive CGM use, factory-calibrated CGM, and HCL to a period afterward. Limitations include the retrospective design and data from a single center.

In conclusion, this real-world study demonstrates sharp increases in CGM and HCL use being strongly associated with lower A1C across a pediatric clinic population, a welcome finding after many years of apparent stagnation or deterioration in A1C trends seen in the T1D Exchange clinic registry (16). With a quarter of patients not on CGM, more than two-thirds not on HCL, and some demographic subgroups with even lower use rates, addressing barriers to uptake of technologies may further improve A1C trends, reduce disparities, and decrease risks of diabetes-related complications.

Acknowledgments. The authors thank Bing Wang (University of Colorado Anschutz Medical Campus) for database support.

Funding. Supported by the University of Colorado Diabetes Research Center Clinical Resources Core, National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases grants P30DK116073, 5T32DK063687-17, and K12DK094712. H.K.A. has received research support from the Institute for the Advancement of Food and Nutrition Sciences.

Duality of Interest. G.P.F. conducts research sponsored by Medtronic, Dexcom, Abbott, Tandem, Insulet, Eli Lilly, and Beta Bionics and has been a speaker/consultant/advisory board member for Medtronic, Dexcom, Abbott, Tandem, Insulet, Eli Lilly, and Beta Bionics. H.K.A. received research support from Tandem, Dexcom, Medtronic, Senseonics, Eli Lilly, IM Therapeutics, and REMD Pharmaceuticals and has been a consultant/advisory board member for MannKind, Eli Lilly, and Senseonics. B.I.F. has been an advisory board member for Provention Bio. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. G.T.A., T.M.T., H.K.A., M.E.P., M.S., G.P.F., and B.I.F. researched data and reviewed and edited the manuscript. G.T.A. and B.I.F. conceptualized the study. C.S. and L.P. assisted with study design, performed statistical analyses, and reviewed and edited the manuscript. G.T.A. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.