Association Between Diabetes Technology Use and Glycemic Outcomes in Adults With Type 1 Diabetes Over a Decade Free

A minority of adults with type 1 diabetes in the U.S. meet the HbA_1c target of <7.0%, a goal recommended by the American Diabetes Association for most individuals without significant comorbidities (1–3). Only 23% of adults in the U.S. T1D Exchange Clinic Registry are able to meet the HbA_1c goal of <7.0% (2,3).

There have been remarkable advancements in diabetes technologies, with newer generation continuous glucose monitoring (CGM) systems that are smaller (4,5), do not require finger-stick glucose for calibration or insulin dosing (6), and are user friendly and more accurate than older generation systems (7,8). Numerous randomized trials have proven glycemia-lowering efficacy of CGMs in the management of type 1 diabetes (9–11). Similarly, clinical trials of automated insulin delivery (AID) devices in adults with type 1 diabetes have reported significant HbA_1c reduction compared with baseline or comparator treatment (12–15). Thus, the use of AID devices has taken the improvement in glycemia a step further since the approval of the first AID system by the U.S. Food and Drug Administration in 2016. With improving diabetes technology, we aimed to evaluate the uptake of diabetes technology and its association with change in mean clinic HbA_1c over time.

We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline (16). The Colorado Multiple Institutional Review Board approved this study under the exempt category.

In this single-center study, adults with type 1 diabetes (age ≥18 years) who visited the Barbara Davis Center for Diabetes Adult Clinic between 1 January 2014 and 31 December 2021 were included. The Barbara Davis Center for Diabetes is an academic center of excellence in clinical care and research in the field of type 1 diabetes, and the Adult Clinic caters to patients mostly from Colorado, serving both urban and rural areas. Only one visit (last visit of the year) per patient per year was included. Sex, race and ethnicity, age, diabetes duration, type of insurance, pregnancy status, HbA_1c measurements within 14 days of the encounter, BMI (kg/m²), CGM use (yes or no), and AID use (yes or no) were collected from the electronic medical record. If an HbA_1c value within 14 days of the encounter was missing, the next most recent HbA_1c value of that year was used. Per clinic standards, most individuals had point-of-care HbA_1c (DCA Vantage Analyzer [Siemens Health Care Diagnostics] from 2014 to 2021 and Afinion 2 Analyzer [Abbott Laboratories] from 2021 onward) at each clinic visit. In a sensitivity analysis, we excluded women who had a pregnancy during the study period (n = 329) because of variations in glycemic control associated with pregnancy.

Diabetes technology use was defined as CGM use (CGM alone with multiple daily injections or insulin pump without AID) or an AID system use. Low-glucose suspend systems were not considered to be AID, and users of these systems were included in the CGM use group. The Medtronic 670G and 770G and Tandem t:slim X2 with Control-IQ technology were the commercially available AID systems during the study period.

The primary outcome was mean clinic HbA_1c over the years with changing diabetes technology uptake. The secondary outcomes were the difference in HbA_1c between diabetes technology users and nonusers by year and the change in the percentage of patients who met the HbA_1c goal of <7.0% by diabetes technology use for each year.

Continuous data are presented as means and SDs, and categorical data are presented as absolute numbers and percentages. For each year, the number of unique patients and the mean (SD) age, diabetes duration, BMI, and HbA_1c values were calculated. Multivariable linear mixed models were used to examine HbA_1c levels by year and by diabetes technology use, adjusted for age and sex. A similar model was also used to investigate HbA_1c differences by diabetes technology use by time points, adjusted for age, sex, and visit year. In this model, we included an interaction between visit year and diabetes technology use. Another mixed model examined HbA_1c levels by race and diabetes technology use, adjusted for age, sex, race, insurance category, and visit year. An interaction term for race by diabetes technology use was included in this linear mixed model. For linear mixed models, least-squares mean HbA_1c values and SEs were reported. The χ² test was used for categorical comparison. Statistical significance was determined as a two-sided P < 0.05.

The data sets generated during and/or analyzed in the current study are available from the corresponding author upon reasonable request.

There were 4,174 unique patients (mean 1,988 patients/year) included in this analysis, with 15,903 clinic visits examined over 8 years. Characteristics of adults with type 1 diabetes by years are shown in Table 1. There was a significant increase in diabetes technology use over 8 years from 26.9% in 2014 to 82.7% in 2021. There was a modest decline in mean clinic HbA_1c from 7.7% in 2014 to 7.5% in 2021 (Fig. 1). In the linear mixed model adjusted for age, sex, and visit year, HbA_1c decreased over time (P < 0.001), while diabetes technology use was associated with a 0.53% reduction in HbA_1c (95% CI 0.49–0.58%, P < 0.001).

Diabetes technology users had significantly lower HbA_1c than diabetes technology nonusers in all 8 years (P < 0.001 for all time points) (Fig. 2A). The difference significantly increased over time (P < 0.001) from 0.36% (95% CI 0.26–0.47%, P < 0.001) in 2014 to 0.93% (95% CI 0.80–1.06%, P < 0.001) in 2021. HbA_1c among diabetes technology nonusers increased over time from 7.85% (SE 0.04%) in 2014 to 8.4% (SE 0.06%) in 2021.

The percentage of people who achieved the HbA_1c goal of <7.0% increased from 32.3% in 2014 to 41.7% in 2021 (P < 0.001); the overall change from 2014 to 2021 was significant (P < 0.0001), as was the test for trend (P < 0.0001). The percentage of diabetes technology users and diabetes technology nonusers who achieved the HbA_1c goal of <7.0% were 43.0 and 28.3% in 2014 (P < 0.0001), and 47.3 and 15.2% in 2021 (P < 0.0001), respectively (Fig. 2B). The percentage of diabetes technology users who achieved the HbA_1c goal was significantly higher than diabetes technology nonusers for all years (P < 0.001 for all years). Diabetes technology users were 2.6 times (95% CI 2.3–2.8, P < 0.001) more likely to achieve the HbA_1c goal than diabetes technology nonusers. The difference in odds of achieving the HbA_1c goal by diabetes technology use also increased significantly over time from an odds of 1.9 (95% CI 1.5–2.4) in 2014 to an odds of 5.7 (95% CI 3.6–8.9) in 2021 (P < 0.001 for year-by-technology use interaction).

In a sensitivity analysis, diabetes technology users were separated into CGM users and AID users from 2017 to 2021. AID users had significantly lower HbA_1c than diabetes technology nonusers (P < 0.001 for every year) from 2017 (7.4% [SE 0.09%] vs. 8.1% [SE 0.04%]) to 2021 (7.3% [SE 0.04%] vs. 8.4% [SE 0.06%]). CGM users also had significantly lower HbA_1c than diabetes technology nonusers at all time points (P < 0.001 for every year) from 2017 (7.6% [SE 0.05%] vs. 8.1% [SE 0.04%]) to 2021 (7.6% [SE 0.04%] vs. 8.4% [SE 0.06%]) (Fig. 3A). AID use was associated with a 0.20% (95% CI 0.13–0.27%) lower HbA_1c than CGM use and a 0.74% (95% CI 0.66–0.81%) lower HbA_1c than diabetes technology nonusers. CGM use was associated with a 0.54% (95% CI 0.48–0.60%) lower HbA_1c than diabetes technology nonusers. The percentage of people who achieved the HbA_1c goal of <7.0% differed significantly across CGM users, AID users, and diabetes technology nonusers for all years (P < 0.0001 for all years) (Fig. 3B).

The percentage of pregnant patients with type 1 diabetes remained the same throughout the 8 years (Table 1). Excluding pregnancy status did not change any of primary or secondary outcomes. The associations between HbA_1c and technology use (P < 0.0001) and visit year-by-technology use interaction (P < 0.0001) remained significant even after excluding pregnant patients. In a sensitivity analysis after excluding pregnant patients, diabetes technology use was associated with a 0.53% lower HbA_1c (95% CI 0.49–0.58) than technology nonuse. Similarly, in an analysis excluding pregnant patients, use of AID was associated with a 0.78% lower HbA_1c (95% CI 0.70–0.86%, P < 0.0001) than nonuse and a 0.22% lower HbA_1c (95% CI 0.15–0.29%, P < 0.0001) than CGM use. Use of CGM was associated with a 0.56% lower HbA_1c (95% CI 0.50–0.62%, P < 0.0001) than nonuse.

When the interaction of race and diabetes technology use was examined in the model, HbA_1c differed significantly by diabetes technology use and race (P for interaction = 0.04). Regardless of diabetes technology use, Non-Hispanic White patients had significantly lower HbA_1c than non-White patients (diabetes technology users: 7.5% [SE 0.03%] vs. 7.7% [SE 0.07%], P = 0.02; diabetes technology nonusers: 8.0% [SE 0.03%] vs. 8.3% [SE 0.06%], P < 0.001). Diabetes technology users had a lower HbA_1c than diabetes technology nonusers in both race groups (Non-Hispanic White patients: 7.5 vs. 8.0%; non-White patients: 7.7 vs. 8.3%; P < 0.001 for both).

To our knowledge, this study is the first to demonstrate changes in the diabetes technology use including AID systems by adults with type 1 diabetes and changes in clinic mean HbA_1c levels over a period of 8 years. Our study showed increasing diabetes technology adoption in the past decade, and with increasing use of CGM and AID, more people achieved the goal HbA_1c. Mean clinic HbA_1c was only modestly reduced (by 0.2%), specifically in the last 2 years, despite a significant reduction in HbA_1c among diabetes technology users. This is largely due to the observation that HbA_1c for people with type 1 diabetes not using diabetes technologies increased over the study period, resulting in a small reduction in the mean clinic HbA_1c.

The management landscape of type 1 diabetes is changing rapidly with the development and approval of newer diabetes technologies such as CGMs and AID systems. Randomized trials and U.S. national registry studies have demonstrated better glycemic control with the use of CGM compared with the self-monitoring of blood glucose (6) and with AID use compared with CGM and pump alone (also called sensor-augmented pump therapy) (17,18).

Newer generation CGMs, such as the Dexcom G6 and G7 and the Abbott Laboratory Libre 2 and 3, are smaller, last longer (10–14 days), and do not require finger-stick glucose for insulin dosing or calibration compared with CGM models that were available before 2020 (4,5). Improved human factors, such as size of the device, ease of use, and satisfaction with the use of the device, in addition to expansion of CGM coverage by private as well as government insurance (19) may have led to increasing adoption of CGM in our clinic. Our previous study demonstrated feasibility and improved outcomes with early initiation of CGM from type 1 diabetes diagnosis (20,21). This led to a change in the clinic practice of initiating CGM early from type 1 diabetes onset starting in 2018. Increasing use of CGM from type 1 diabetes onset may also have influenced the improved HbA_1c outcomes in our population in the last 3 years.

The U.S. Food and Drug Administration approved the first-generation AID system (Medtronic 670G) in 2016, which was commercially launched in March 2017 (22). We observed increasing use of AID systems in our clinic from 2017 onward. However, the mean clinic HbA_1c did not change significantly in 2017 and 2018. We believe that this is mainly due to the high rate of discontinuation of the first-generation AID system and low use of the automated mode because of human factor–related issues with the first-generation system (23). Studies have reported 30% discontinuation of first-generation AID systems in the U.S. within the first year of using this system (24). HbA_1c was still lower in those using an AID system compared with CGM alone or no use of diabetes technologies, suggesting that daily use of diabetes technology helps to improve glycemic outcomes.

We observed increasing HbA_1c among diabetes technology nonusers over time, and many factors may have contributed to this trend. We cannot rule out provider bias for not prescribing diabetes technology among those with higher HbA_1c or from disadvantaged socioeconomic backgrounds (25). Other plausible explanations could be that perceived lack of interest in caring for diabetes because of high HbA_1c may have led to not prescribing diabetes technologies or that socioeconomic disadvantages may have made it prohibitive to use these technologies, making diabetes control difficult with increasing HbA_1c over time. We previously reported higher rates of hypo- and hyperglycemia-related emergency visits among diabetes technology nonusers (21), and this may have led to hypoglycemia fear and higher HbA_1c.

Similar to a previous study (2), our study also reported higher HbA_1c among minorities compared with Non-Hispanic White individuals; however, HbA_1c reduction was significant in both racial groups, suggesting that diabetes technology use, especially AID use, may have led to significant improvement in HbA_1c regardless of race and ethnicity. Low recruitment of minorities in clinical trials of diabetes technologies (26) and language barriers with device use or device training (27) may be reasons for less diabetes technology use and suboptimal outcomes with the use of diabetes technologies in minorities with type 1 diabetes. More research on human factors to improve technology adoption and adherence is needed to improve glycemic outcomes and reduce disparities in diabetes care.

The findings of our study are in agreement with American Diabetes Association standards of care (28), recent international consensus (29) recommending CGM and AID for most people with type 1 diabetes, and early initiation of diabetes technology from the onset of type 1 diabetes. Increasing diabetes technology adoption may also reduce disparities in diabetes care and complications.

A large number of patients with records for each year and meticulous data extraction are major strengths of this study. The single-center and retrospective study design are major limitations. We did not have data on individual participants over time. Moreover, collection of data from electronic medical records is another limitation. Inadvertent inclusion of patients with type 2 diabetes or inappropriate assignment of participants to diabetes technology use versus nonuse because of medical record errors are possible. We did not collect information on socioeconomic or education status and the presence of comorbidities, which may affect the ability to achieve optimal glycemic control.

In conclusion, our study demonstrated significant increases in diabetes technology adoption among adults with type 1 diabetes along with decreased HbA_1c and more people able to achieve the HbA_1c goal of <7.0%. AID systems took the CGM system’s success a step further and provided a significant reduction in HbA_1c at the population level.

Duality of Interest. H.K.A. received, through the University of Colorado, research support from Dexcom, Medtronic, Tandem Diabetes, Eli Lilly, Senseonics, IM Therapeutics, REMD Biotherapeutics, and the International Association of Forensic Mental Health Services and honoraria from Mannkind and Senseonics for advisory board attendance. V.N.S. reported receiving, through the University of Colorado, research support from Novo Nordisk, Insulet, Tandem Diabetes, and Dexcom and honoraria from Medscape, Lifescan, Novo Nordisk, and DKSH Singapore for advisory board attendance and from Insulet and Dexcom for speaking engagements. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. K.E.K., H.K.A., and V.N.S. were involved in the conception, design, and conduct of the study and wrote the first draft of the manuscript. G.T.A. and J.K.S.-B. were involved in the analysis and interpretation of the results. All authors edited, reviewed, and approved the final version of the manuscript. V.N.S. 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.

Prior Presentation. Parts of this study were presented at the 16th International Conference on Advanced Technologies and Treatments for Diabetes, Berlin, Germany, 22–25 February 2023.