Impact of race and delayed adoption of diabetes technology on glycemia and partial remission in type 1 diabetes

Introduction

Racial inequity impedes quality diabetes care for minority youth with type 1 diabetes (T1D) [1], 2], a population that is rapidly growing in the United States (US) [3]. T1D is a chronic, debilitating disease that affects 1.7 million adults of age >20 years and 304,000 children and adolescents of age <20 years [4]. Strategies to reduce the medical and economic burden of T1D led to the development of diabetes technology such as the insulin pump and continuous glucose monitors (CGM), which provide real-time or on-demand blood glucose profiles, enabling precise responses to glucose fluctuations [5]. CGM use decreases hemoglobin A1c (A1c), a marker of glycemia, regardless of insulin delivery modality [6], [7], [8]. CGM usage also improves the quality of life and is often preferred by parents over fingerstick blood glucose checks [9].

Despite these benefits, the impact of diabetes technology on the partial clinical remission (PR) phase, also known as the honeymoon phase, of T1D remains unclear [10], [11], [12]. The PR phase is characterized by residual endogenous insulin secretion from the surviving β-cells [13]. Prolonging the PR phase is associated with improved glycemic control and reduced prevalence of long-term T1D complications [11], 14]. Data from the Diabetes Control and Complications Trial and some recent studies [12] suggest that intensive glycemic control could preserve residual β-cell function (RBCF) and prolong PR [11], 15]. However, other studies [10], 16] and systematic reviews [17] have not supported this premise. A study by Enander et al. [18] showed that RBCF at two years was associated with initial A1c and C-peptide levels at diagnosis, but not with initial insulin regimens (subcutaneous vs. intravenous). Thus, further research is needed to clarify the impact of glycemia and diabetes technology on PR and RBCF.

Though diabetes technology improves both short- and long-term health outcomes in T1D, minority youth are 50 % less likely to use CGM than White youth [3], 19]. Minority youth face significant individual, cultural, and structural barriers [2] that hinder access to and engagement with diabetes technology, such as CGM [2], 3]. These disparities result in poorer glycemic control, earlier onset of T1D complications, and worse outcomes in adulthood, including renal failure, amputations, blindness, cardiovascular events, and shortened lifespan, compared to their Non-Hispanic White peers [20]. These racial/ethnic disparities in CGM and insulin pump use [3] call for equity-informed studies and interventions to generate data that can increase CGM and insulin pump use among minority youth with T1D. This is crucial because existing technology-based educational interventions do not adequately address the unique multilevel factors limiting technology adoption in these populations [21].

We designed this study to investigate the outcomes of diabetes technology use in minority and majority youth in the US, with a particular focus on remediable factors contributing to disparities in CGM use. We use race/ethnicity not as a biological construct but as a marker of inadequate access to medical care, potentially influenced by socioeconomic factors. Our aim was to determine any differential impact of diabetes technology use on glycemia and PR in majority vs. minority children with T1D and the factors responsible for such differences. We hypothesized that early introduction of diabetes technology would decrease glycemia and prolong PR duration in both groups.

Materials and methods

Subjects

In this longitudinal, retrospective, single-center study, we identified 154 subjects who were diagnosed with T1D between 2018 and 2024, who were started on either Dexcom G5 or G6, and who had no prior history of CGM use (Figure 1). We excluded subjects who had no data on the date of CGM initiation, those who discontinued their CGM in the first 12 months from initiation, and those not wearing their CGM consistently. We further excluded subjects with other types of diabetes (e.g., type 2 diabetes and cystic fibrosis-related diabetes), subjects with disordered eating, and those with chronic illnesses including sickle cell disease, liver failure, and heart failure. Of the 154 subjects, we excluded 9 subjects with no record of clinic visits in the first 6 months or only one visit in the first 12 months. A total of 145 subjects met inclusion criteria.

Figure 1:

Flow diagram of study subject inclusion and exclusion. The diagram illustrates the process of subject selection for the study, showing the number of subjects included and excluded at each stage based on specific criteria. An exaggerated partial remission was defined as not receiving long-acting insulin for >6 months (n=5) and no insulin for >2 months (n=1). A total of 4 subjects with unspecified race/ethnicity were excluded, including 2 females and 2 males. The mean (SD*) age among these 4 subjects was 115.25 (25.02) months. The mean (SD*) for A1c at 12 and 24 months for these patients was 7.7 % (0.2) and 8.3 % (0.6), respectively. *SD=standard deviation.

Results

Of the 145 subjects included, 139 (75 male and 64 female) were analyzed for glycemia (Table 1). Of the subjects classified as Other, 4 were Hispanic White, 20 were Black/African American, 18 were Asian, and 26 identified as other. There were 105 early adopters and 34 late adopters. Sixty early adopters and 19 late adopters had data at both baseline and 24 months. One hundred and forty-five subjects were included in the PR analyses: this number includes the 6 subjects who were excluded from the glycemic analyses for having an exaggerated PR. Summary statistics of the absolute change in A1c, IDAA1c, and BMI-SDS for the overall cohort are reported in Table 2.

Longitudinal analyses of glycemia and PR

Hemoglobin A1c

A1c decreased significantly over time after adjusting for insulin pump status, race and ethnicity, age at CGM initiation, BMI SDS, and early adopter status (p<0.0001). A1c decreased by 31.1 % (95 % CI: 28.5–33.5 %) 3 months after CGM initiation relative to baseline. However, A1c did not decrease significantly at 6, 9, 12, 15, 18, 21, or 24 months relative to the 3-month visit, p>0.05 for all (Figure 2). This suggests that the decreased glycemia in the first 3 months following CGM initiation was maintained for 24 months.

Figure 2:

Graph of the temporal trends in A1c between the non-Hispanic White (NHW) subjects and the Other during the 24 months of observation following the initiation of continuous glucose monitoring (CGM). The figure shows that the estimated mean A1c levels were significantly lower in the NHW subjects vs. the Other, p<0.0001.

The results show that for each 3-month increase in age at CGM start, hemoglobin A1c decreased by 0.3 % (95 % CI: 0.1–0.5 %, p=0.001). BMI SDS was also significantly associated with A1c (p<0.0001), with a 4.9 % decrease (95 % CI: 2.9–6.5 %) per unit increase in BMI SDS. There was an equally significant association between insulin pump status and hemoglobin A1c (p=0.03), such that subjects on insulin pumps had a 4.9 % decrease in A1c compared to those who did not use such technology (95 % CI: 0.5–9.1 %). In contrast, early adoption of CGM was not significantly associated with a change in A1c, (p=0.08). There was also no significant association between % change in A1c and race and ethnicity after adjusting for visits, pump status, age, BMI SDS, and early adopter status (p=0.10).

Insulin-dose-adjusted hemoglobin A1c (IDAA1c)

IDAA1c decreased significantly over time after adjusting for insulin pump status, race and ethnicity, age at CGM initiation, BMI SDS, and early adopter status, p<0.0001. IDAA1c decreased by 29.2 % (95 % CI: 26.6–31.7 %) relative to baseline in the first 3 months after CGM initiation. BMI SDS was significantly associated with changes in IDAA1c over time (p<0.0001), with a 3.7 % decrease in IDAA1c (95 % CI: 1.8–5.5 %) for each unit increase in BMI SDS. Early CGM adopter status was significantly associated with change in IDAA1c (p=0.02), with a 7.1 % greater decrease in IDAA1c in early compared to late adopters (95 % CI: 1.3–12.5 %). In contrast, there were no significant associations between IDAA1c and pump status, age at CGM initiation, and race and ethnicity (p=0.069, p=0.25, and p=0.12, respectively).

Total daily dose of insulin (TDDI)

Although there was a significant decrease in the TDDI in the first several months following CGM initiation, an increase in TDDI was noted over time. TDDI decreased significantly in the first 3–6 months of CGM initiation, (p<0.001). TDDI decreased by 0.099 ± 0.02 units/kg/day in the first 3 months and by 0.07 ± 0.02 units/kg/day at 6 months, relative to baseline. No significant changes in TDDI occurred during the 9–12 months of CGM use compared to baseline (p=0.5). However, TDDI rose significantly from 12 to 24 months of CGM use, compared to baseline, p<0.014 (0.055 ± 0.02 units/kg/day at 15 months, 0.096 ± 0.02 units/kg/day at 18 months, and 0.13 ± 0.02 units/kg/day at 21 and 24 months).

TDDI decreased significantly in the early compared to the late CGM adopters, 0.09 ± 0.04 units/kg/day, p=0.03. The changes in the TDDI were not significantly associated with BMI SDS, age, insulin pump status, or race and ethnicity (p=0.29, p=0.07, p=0.37, and p=0.33, respectively).

CGM metrics

Our unadjusted analysis of 67 subjects with 3-month and 24-month CGM data showed a significant increase in mean glucose values, p=0.0025. Specifically, the mean ± SD change in mean glucose from 3 to 24 months (calculated as mean glucose at 24 months minus mean glucose at 3 months) was 16.6 ± 43.1 mg/dL. During this period, there was also a significant increase in mean fasting plasma glucose (12.1 ± 48.6 mg/dL, p=0.047) and a significant decrease in mean TIR (−10.3 ± 26.3, p=0.005). As TIR decreased, TBR and TAR increased as follows: from 3 to 24 months, the mean ± SD change in TBR class 1 (mild hypoglycemia with blood glucose 69–54 mg/dL) was an increase of 0.25 ± 1.07; the median (Q1–Q3) was 0 (0–0.5). The mean change in percent TBR class 2 (severe hypoglycemia with blood glucose <54 mg/dL) ± SD from 3 to 24 months was an increase of 0.30 ± 0.57; the median (Q1–Q3) was 0 (0–1). There was a significant increase in mean TAR class 1 (hyperglycemia 181–250 mg/dL) (3.4 ± 13.5, p=0.045) and class 2 (hyperglycemia >250 mg/dL) (6.3 ± 17.7, p=0.005) from 3 to 24 months. The longitudinal analysis of TIR from 3 to 24 months showed that TIR did not differ significantly based on insulin pump status, race and ethnicity, or early adopter status (p=0.10, p=0.26, p=0.44, and p=0.20, respectively).

Analysis of the occurrence of partial clinical remission (PR) based on race and ethnicity, time of CGM adoption, use of insulin pump, and anthropometry

There was no significant difference in the occurrence of PR between the NHW and the Other at the specified time points of 6 and 12 months following the diagnosis of T1D. Out of 108 subjects with data at 6 months, 30/51 (58.8 %) NHW subjects compared to 25/57 (43.9 %) subjects in the Other category had an IDAA1c of ≤9 (unadjusted p=0.12, FDR-adjusted p=0.38). At 12 months, out of 99 subjects with data, IDAA1c was ≤9 in 46 % (23/50) of NHW subjects compared to 26.5 % (13/49) of subjects in the Other category (unadjusted p=0.04, adjusted p=0.33). An assessment of the proportion of all subjects who had IDAA1c of ≤9 within 12 months after T1D diagnosis showed that 72.5 % (50/69) of NHW subjects experienced PR compared to 51.4 % (37/72) of the Other (unadjusted p=0.01, adjusted p=0.12). A similar assessment over 24 months revealed that 73.9 % of NHW subjects and 52.8 % of the Other achieved PR (unadjusted p=0.0093, adjusted p=0.12).

Insulin pump status, age at CGM initiation, or BMI SDS had no significant impact on PR status at any specified time points. Though PR status did not differ significantly between the early vs. late adopter groups at 6, 12, and 18 months, our analysis at 24 months showed that PR occurred in 33.3 % (21/63) of early CGM adopters compared to 5 % (1/20) of late adopters (unadjusted p=0.012, adjusted p=0.12). Looking at the trends over time, the unadjusted analyses show that early CGM adopters had a greater predicted probability of achieving PR over time compared to late adopters (p<0.0001) (Figure 3), and the predicted probability of achieving PR was significantly higher over time among the NHW subjects (p<0.0001) (Figure 4).

Figure 3:

Longitudinal comparison of partial clinical remission (PR) status between the early adopters and later adopters. The graph shows the predicted probabilities of experiencing the partial clinical remission (PR) of type 1 diabetes between the early adopters of continuous glucose monitoring (CGM) technology vs. the late CGM adopters during the 24 months of observation following CGM initiation. The early CGM adopters had a higher probability of experiencing PR, defined as an insulin-dose-adjusted A1c (IDAA1c) level of ≤9), compared to the late adopters, p<0.0001.

Figure 4:

Longitudinal comparison of partial clinical remission status between the non-Hispanic White (NHW) subjects and the Other. A graph of the predicted probabilities of partial clinical remission (PR) between the non-Hispanic White (NHW) youth vs. the Other during the 24 months of observation following the initiation of continuous glucose monitoring (CGM). This graph suggests that the NHW youth had a higher probability of achieving PR than the minority youth, p<0.0001.

Discussion

This study examining the impact of race/ethnicity and the timing of the adoption of diabetes technology in youth with T1D has 5 key findings. First, diabetes technology use is associated with sustained reductions in both A1c and IDAA1c in both NHW and minority youth with newly diagnosed T1D. Additionally, the combination of CGM with an insulin pump is superior to the use of CGM alone for reductions in both glycemia and IDAA1c in these populations. Second, a disparity exists in technology adoption: 90 % of NHW children adopted technology early compared to only 63 % of minority children. This difference appears to influence glycemic control and PR.

Third, our model showed that the early adoption of diabetes technology was associated with decreased glycemia. Specifically, the adjusted model showed that the decreased glycemia in the first 2 years of diagnosis of T1D was associated with diabetes technology, increased age, and BMI, and not with race/ethnicity or the time of CGM adoption. However, the unadjusted analysis showed a significant relationship between early adoption of diabetes technology, in this case, CGM, and decreased hemoglobin A1c. This suggests that the superior glycemic control observed in NHW children likely stems from early and consistent technology use, rather than inherent racial/ethnic factors.

Fourth, diabetes technology, not race or ethnicity, positively impacts PR [8]. Early technology adoption, continued use, and the combination of CGM and insulin pumps were associated with a reduction in IDAA1c, a clinical marker of PR. While a greater proportion of NHW youth achieved PR at various time points, this did not reach statistical significance, likely due to the p-value adjustment for multiple testing and limited sample size. However, unadjusted GLMM analyses suggest a clinically meaningful relationship between early CGM adoption (in the NHW group) and PR attainment. Specifically, our data suggest a role for early CGM adoption in PR attainment as a greater proportion of early adopters tended to reach PR at each study time point than the late CGM adopters, though these differences did not reach statistical significance. In contrast, insulin pump status, age, and BMI SDS were not associated with PR status at any time point.

Fifth, CGM metrics showed no significant differences in hyperglycemia (TAR) or hypoglycemia (TBR) between NHW and minority children, demonstrating comparable safety profiles across groups and highlighting the competency of both groups in using the technology. This parity in safety profile is consistent with our previous studies where we found no differences in TAR and TBR between the racial and ethnic groups [22], 25].

These findings underscore that modifiable factors like access to technology, rather than non-modifiable factors such as race and ethnicity, drive optimal glycemic control and PR in youth with T1D. Specifically, lower temporal trends for hemoglobin A1c and IDAA1c likely stemmed from early adoption of CGM, the use of an insulin pump, and the combination of insulin pump and CGM. The uneven proportion of early adoption of diabetes technology in youth with T1D is concerning, as this could explain some of the disparities in the prevalence of complications of T1D in various racial and ethnic groups [11]. Several factors limit the adoption and retention of diabetes technology, according to studies examining barriers to its use, including patients’ concerns regarding wearing visible devices, socio-economic status, type of health insurance, health insurance eligibility requirements, access to endocrinologists, and skin conditions [2], 3], 26], 27]. Addressing the disparities in technology access, including those rooted in social determinants of health (SDOH) and potential implicit biases within healthcare [19], is crucial for equitable diabetes care including optimal adoption of diabetes technology to decrease glycemia and improve PR.

Limitations of the study include its retrospective design which precludes causality, and the relatively small sample size, which could have prevented the detection of clinically significant differences in some variables. We cannot exclude the possibility of a selection bias favoring children from families with higher socioeconomic status. Strengths include the comparable cohorts, prolonged follow-up, and inclusion of IDAA1c as a marker of PR, which allowed us to examine the long-term consequences of differential adoption of diabetes technology.

Conclusions

Early adoption of diabetes technology augments euglycemia and appears to promote PR in youth with T1D. However, early CGM adoption occurs in 90 % of NHW children vs. 63 % of minority children. This disparity in early CGM adoption could help to explain the gap in outcomes for different populations with T1D. A nationwide campaign for prompt and sustained access to diabetes technology for all children will help to close the gap in diabetes care and reduce the burden of complications.