Glycemic Measures in Childhood as Predictors of Future Diabetes-Related Microvascular Complications in an Indigenous American Population

The prevalence of obesity among children and adolescents age 2–19 years was 19.7% in the U.S., according to the latest prepandemic nationally representative estimates (2017 to March 2020) from the National Health and Nutrition Examination Survey (1). With the onset of the COVID-19 pandemic and the severe social disruptions in its aftermath, the rate of BMI increases among U.S. children almost doubled compared with that in the prepandemic period (2). Those with pre-existing overweight or obesity and younger children exhibited the greatest increase in BMI (2).

There are significant racial/ethnic disparities in pediatric obesity prevalence and in early-life risk factors in minority populations that are likely to influence adiposity (3–5). Increasing severity of childhood obesity is associated with higher cardiometabolic risks, such as poor glycemic control, dyslipidemia, hypertension, and premature mortality (6–9). Recent studies have shown a rise in the incidence and prevalence of youth-onset diabetes, primarily type 2 diabetes, with the increase in childhood obesity (10,11). Youth with type 2 diabetes are at a higher risk of microvascular and other complications compared with their peers with type 1 diabetes (12). These complications are often solitary or appear in clusters, progress over time, and manifest in most patients by early adulthood (13). There is a high prevalence of diabetes-related complications and comorbidities observed in minority youth (13).

The American Diabetes Association (ADA) recommends risk-based screening for prediabetes and/or diabetes after pubertal onset or after age 10 years (whichever occurs earlier) in asymptomatic children and adolescents with overweight or obesity (age- and sex-specific BMI ≥85th percentile) by pediatric care providers. Risk factors include history of maternal gestational diabetes mellitus, first- or second-degree relatives with type 2 diabetes, belonging to specific race/ethnic groups (i.e., American Indian, African American, Latino, Asian American, and Pacific Islander), and exhibiting clinical signs of or conditions associated with insulin resistance (14). A recent study examining the usefulness of the risk-based ADA screening guidelines for prediabetes and diabetes in asymptomatic children found that despite many children being eligible for screening, few were diagnosed with these conditions (15). The U.S. Preventive Services Task Force (USPSTF) in its recently published guidelines concluded that there is currently insufficient evidence to assess the balance of the benefits and harms of screening asymptomatic children and adolescents for type 2 diabetes (16). The recommendation was based on a scientific review of the effects of screening and early diagnosis on long-term outcomes of type 2 diabetes, such as microvascular disease, mortality, and quality of life. The review found that most studies were of short duration and focused on intermediate health outcomes, and the benefits and harms of screening were not adequately evaluated (17).

The objective of this study was to examine the role of glycated hemoglobin (HbA_1c) and additional glycemic tests (i.e., 2-h plasma glucose [PG] and fasting PG [FPG]), performed during childhood, in predicting future type 2 diabetes-related microvascular complications, such as nephropathy and retinopathy, in a high-risk indigenous American population.

Data were obtained from a longitudinal study of diabetes and its associated conditions conducted over a 43-year period (1965–2007) in an indigenous American community in the southwestern U.S. with a high prevalence of type 2 diabetes (18). Residents of the community, almost all of whom reported American Indian heritage, who were age ≥5 years were invited to participate in biennial research examinations regardless of health status. In this community, almost all cases of diabetes, both in children and adults, are type 2 diabetes (19–21). In addition, a prior genetic analysis showed that none of the known maturity-onset diabetes of the young genes play an important role in the pathogenesis of early-onset diabetes in this population (22).

Data collected included anthropometric measures (height, weight, and waist circumference), medical history, glycemic measures, and spot urine specimens. The glycemic measures included FPG and 2-h PG tests after a 75-g oral glucose load. HbA_1c measurements were included beginning in 1989. Initial HbA_1c measurements were performed using high-performance liquid chromatography. In 2001, the laboratory changed the assay to the A1C 2.2 Plus Glycohemoglobin Analyzer (Tosoh Diagnostics, San Francisco, CA), which continued until the end of the study in 2007. To address differences between the two assays, results from the older assay (1989–2000) were converted to the newer assay by a linear regression equation (HbA_1c new method = −0.1916 + [0.9829 × HbA_1c old method]).

The albumin-to-creatinine ratio (ACR) in a spot urine specimen was used as a marker of nephropathy. Urine albumin was measured by nephelometric immunoassay (23), whereas urine creatinine was measured by a modified Jaffé reaction (24). Albuminuria was categorized using ACR as follows: <30 mg/g, normal; 30 to <300 mg/g, albuminuria; and ≥300 mg/g, severe albuminuria (25).

Retinopathy was assessed by direct ophthalmoscopy performed by a research physician who was trained in fundoscopic examination. Diabetic retinopathy was defined by the presence of at least one microaneurysm or hemorrhage or proliferative retinopathy. A prior study in this population comparing fundoscopy with retinal photography showed fair to good agreement between the two diagnostic methods (26). Funduscopic data were used for the present analyses, because retinal photography was only available in a subset of the population.

We used data collected since 1989 to compare the ability of the different glycemic measures (HbA_1c, 2-h PG, and FPG) to predict albuminuria, severe albuminuria, and retinopathy. Unlike HbA_1c, 2-h PG measurements were collected from study inception in 1965, resulting in more observations. To obtain more robust estimates of the associations of glycemia with future diabetes complications, we also analyzed 2-h PG data collected since 1965.

Our current study included children and youth age ≥5 to <20 years with no diagnosis of type 2 diabetes before the baseline screening. Type 2 diabetes was defined using ADA criteria as HbA_1c ≥6.5% (≥48 mmol/mol), FPG ≥126 mg/dL (≥7.0 mmol/L), or 2-h PG ≥200 mg/dL (≥11.1 mmol/L) (14). Because our objective was to assess the utility of screening for prediabetes and type 2 diabetes in high-risk asymptomatic children and adolescents, the inclusion criteria were designed to reflect a usual clinical scenario where the need for screening is contemplated. Thus, children with a pre-existing diabetes diagnosis (as indicated by their medical history) were excluded from the current analysis, but those diagnosed with diabetes at the baseline research examination were included. The current study included only children who had at least one follow-up research examination with data on diabetes and at least one complication.

Three different comparative data sets were compiled to compare the ability of HbA_1c, FPG, and 2-h PG to predict future adverse outcomes, which we named according to the complication analyzed as the outcome: 1) comparative albuminuria, 2) comparative severe albuminuria, and 3) comparative retinopathy. For each complication, the baseline examination was defined as the first examination at age 5 to <20 years, when all FPG, 2-h PG, and HbA_1c levels were measured, and the complication was assessed. For each complication, the follow-up examination was defined as the first examination at which the complication developed or, if the complication did not occur, the last examination at which the complication was assessed. Participants with the complication at the baseline examination were excluded as follows: in the comparative albuminuria data set, those with baseline ACR ≥30 mg/g; in the comparative severe albuminuria data set, those with baseline ACR ≥300 mg/g; and in the comparative retinopathy data set, those with retinopathy at baseline.

Another three 2-hr PG data sets were compiled to study the associations of glycemia based on 2-h PG, for which more data were available than the other glycemic measures, with the risk of future complications: 1) 2-h PG albuminuria, 2) 2-h PG severe albuminuria, and 3) 2-h PG retinopathy. In these data sets, the baseline examination was defined as the first examination where 2-h PG was measured, and the complication was assessed. For each complication, the follow-up examination was also defined as the first examination at which the complication was diagnosed or, if the complication did not develop, the last examination in which the complication was assessed. In these data sets, those with the complication at baseline were excluded in an analogous fashion from the analyses comparing all three glycemic measures. For all six data sets, all follow-up research visits where nephropathy and retinopathy assessments were performed were included, regardless of whether they occurred in childhood or adulthood. The study was approved by the Institutional Review Board of the National Institute of Diabetes and Digestive and Kidney Diseases.

Statistical analyses were conducted using SAS software (version 9.4; SAS Institute, Cary, NC). The incidence of complications per 1,000 person-years (PY) was calculated using the number of incident cases and PY of follow-up. Follow-up years were calculated from baseline to the first occurrence of the event or the last follow-up measurement, if event was not experienced. Incidence rates were calculated for each of the baseline glycemic measures based on three categories. HbA_1c was categorized as normal (≤5.6%; ≤38 mmol/mol), prediabetes (5.7–6.4%; 39–47 mmol/mol), or diabetes (≥6.5%; ≥48 mmol/mol). Categories based on 2-h PG were normal glucose tolerance (NGT; <7.8 mmol/L; <140 mg/dL), impaired glucose tolerance (IGT; 7.8–11.0 mmol/L; 140–199 mg/dL), and diabetes (≥11.1 mmol/L; ≥200 mg/dL). FPG was categorized as normal (<5.6 mmol/L; <100 mg/dL), impaired (5.6–7.0 mmol/L; 100–125 mg/dL), or diabetes (≥7.0 mmol/L; ≥126 mg/dL) (14). Incidence rate differences (RD) between subgroups based on glycemic measures were also calculated.

The associations of the three glycemic measures with the risk of complications were analyzed using Cox proportional hazards, with age and sex as covariates. To assess the influence of diagnosed diabetes on the results, three separate analyses were conducted: with all participants, with participants without diabetes at baseline, and with participants with diabetes at baseline. Areas under the receiver operating characteristic curve (AUCs) were used to compare the ability of HbA_1c, 2-h PG, and FPG to predict future complications, with the glycemic measures included as continuous and categorical predictors (27). The methods described by Pencina and D’Agostino (28) were used to calculate AUCs. Pairwise comparisons of AUCs were conducted using the Delong et al. (29) method. Receiver operating characteristic curves for each complication were plotted at the median follow-up time in years for that complication. All analyses described above were also conducted in subsets of participants with overweight or obesity (BMI ≥85th percentile) at the baseline examination.

The data sets analyzed during the current study are not publicly available because of privacy concerns.

Table 1 lists baseline characteristics for all participants included in HbA_1c and 2-h PG data sets, and Supplementary Table 1 lists them for subsets of participants with overweight/obesity. Most children included were female, and most had overweight or obesity (>57%); the median baseline age ranged from 11.4 to 15.2 years, and the median age at onset of the complication was lowest for albuminuria and highest for retinopathy. In children without diabetes at baseline included in the comparative data sets, diabetes incidence ranged from 12.2 to 13.0 cases per 1,000 PY in those with normal HbA_1c levels and from 66.9 to 70.3 cases per 1,000 PY in those with prediabetes HbA_1c levels. In children without diabetes at baseline included in the 2-hr PG data sets, diabetes incidence ranged from 14.0 to 14.8 cases per 1,000 PY in those with NGT at baseline and from 45.9 to 57.1 cases per 1,000 PY in those with IGT at baseline. The incidence of diabetes was higher in the subsamples of children with overweight/obesity than in the whole samples.

Table 2 lists incidence rates of microvascular complications stratified by data set and subgroup based on HbA_1c and 2-h PG levels. Table 3 lists associations of glycemic measures (HbA_1c/2-h PG) with risk of complications, assessed by Cox proportional hazards models, stratified by data set.

In children without diabetes at baseline, those with prediabetes HbA_1c levels had a higher incidence of albuminuria than those with normal HbA_1c levels (29.7 vs. 23.8 cases per 1,000 PY; RD 5.9; 95% CI −3.4 to 15.3; P = 0.21), although the difference was not statistically significant. Children with diabetes at baseline had the highest incidence of albuminuria (186.1 cases per 1,000 PY). Higher baseline HbA_1c levels, analyzed as a continuous variable, were associated with a significantly increased risk of albuminuria (hazard ratio [HR] 1.44 per 1% increase; 95% CI 1.17–1.77). This association remained significant in participants with diabetes at baseline (HR 1.59; 95% CI 1.06–2.39).

In participants with overweight/obesity and without diabetes at baseline, those with prediabetes HbA_1c levels also had a higher incidence of albuminuria compared with those with normal HbA_1c levels (34.8 vs. 24.9 cases per 1,000 PY; RD 9.9; 95% CI −0.7 to 20.5; P = 0.07). Higher HbA_1c levels, analyzed as a continuous variable, were also associated significantly with increased risk of albuminuria (HR 1.58 per %; 95% CI 1.27–1.95).

In children without diabetes at baseline, those with prediabetes HbA_1c levels had a higher incidence of severe albuminuria than those with normal levels (3.8 vs. 2.4 cases per 1,000 PY; RD 1.4; 95% CI −1.5 to 4.1; P = 0.36). The incidence was highest among children with diabetes HbA_1c levels at baseline (13.3 cases per 1,000 PY). Higher HbA_1c levels at baseline were associated with a significantly increased risk of severe albuminuria (HR 1.45; 95% CI 1.02–2.05).

In participants with overweight/obesity and without diabetes at baseline, those with prediabetes HbA_1c levels had a higher incidence of severe albuminuria than those with normal levels (4.3 vs. 2.8 cases per 1,000 PY; RD 1.5; 95% CI −1.8 to 4.8; P = 0.38), although the difference was not statistically significant. In those with overweight/obesity, higher HbA_1c levels, analyzed continuously, were associated with a significantly increased risk of severe albuminuria (HR 1.46; 95% CI 1.03–2.06).

In children without diabetes at baseline, those with prediabetes HbA_1c levels had a significantly higher incidence of retinopathy than those with normal levels (7.1 vs. 1.7 cases per 1,000 PY; RD 5.4; 95% CI 2.6–8.2; P = 0.0001). Children with diabetes HbA_1c levels at baseline had the highest retinopathy incidence (15.9 cases per 1,000 PY). Higher baseline HbA_1c values were associated significantly with an increased retinopathy risk, when analyzed continuously, (HR 1.65 per %; 95% CI 1.23–2.20). This association remained significant in participants without diabetes at baseline (HR 3.09; 95% CI 1.17–8.22) and was stronger than that in the whole sample.

In children with overweight/obesity and without diabetes, those with prediabetes HbA_1c levels had a significantly higher incidence of retinopathy than those with normal levels (7.9 vs. 2.5 cases per 1,000 PY; RD 5.4; 95% CI 1.9–9.0; P = 0.002). Higher HbA_1c levels, analyzed continuously, also increased the retinopathy risk significantly (HR 1.53 per %; 95% CI 1.11–2.11).

In children without diabetes, those with IGT had a significantly higher incidence of albuminuria than those with NGT (36.3 vs. 21.5 cases per 1,000 PY; RD 14.8; 95% CI 7.3–22.4; P = 0.0001). Children with diabetes at baseline had the highest albuminuria incidence (94.6 cases per 1,000 PY). Increases in 2-h PG, analyzed continuously, significantly increased the risk of albuminuria (HR 1.15 per 1-mmol/L increase; 95% CI 1.12–1.19). This association was still significant in children without diabetes at baseline (HR 1.13; 95% CI 1.06–1.47).

In children with overweight/obesity and without diabetes, those with IGT also had a significantly higher incidence of albuminuria than those with NGT (40.8 vs. 23.5 cases per 1,000 PY; RD 17.3; 95% CI 8.5–26.2; P = 0.0001). Higher 2-h PG levels significantly increased the risk of albuminuria in all participants, when analyzed continuously (HR 1.15 per mmol/L; 95% CI 1.11–1.19), and in those without diabetes at baseline (HR 1.13; 95% CI 1.05–1.22).

In children without diabetes at baseline, those with IGT had a significantly higher incidence of severe albuminuria than children with NGT (5.7 vs. 2.4 cases per 1,000 PY; RD 3.3; 95% CI 0.9–5.6; P = 0.006). Children with diabetes at baseline had the highest incidence of severe albuminuria (25.6 cases per 1,000 PY). Increases in 2-h PG, analyzed continuously, were associated significantly with an increased risk of severe albuminuria (HR 1.21 per mmol/L; 95% CI 1.16–1.27). This association remained significant in participants without diabetes at baseline (HR 1.25; 95% CI 1.06–1.47).

In those with overweight/obesity and without diabetes at baseline, children with IGT had a significantly higher incidence of severe albuminuria than those with NGT (6.7 vs. 3.2 cases per 1,000 PY; RD 3.5; 95% CI 0.5–6.5; P = 0.02). Higher 2-h PG levels, analyzed continuously, significantly increased the risk of severe albuminuria (HR 1.19 per mmol/L; 95% CI 1.14–1.25).

In children without diabetes at baseline, those with IGT had a significantly higher retinopathy incidence than those with NGT (8.3 vs. 2.3 cases per 1,000 PY; RD 6.0; 95% CI 4.0–8.0; P < 0.0001). Children with diabetes at baseline had the highest incidence of retinopathy (29.3 cases per 1,000 PY). Increases in 2-h PG were associated significantly with an increased risk of retinopathy, when analyzed continuously (HR 1.22 per mmol/L; 95% CI 1.17–1.26). This association remained significant in participants without diabetes at baseline and was stronger than in the whole sample (HR 1.48; 95% CI 1.31–1.67).

In children with overweight/obesity and without diabetes at baseline, those with IGT had a significantly higher retinopathy incidence than those with NGT (10.1 vs. 3.2 cases per 1,000 PY; RD 6.8; 95% CI 4.1–9.5; P < 0.0001). When analyzed continuously, increases in 2-h PG significantly increased the retinopathy risk (HR 1.19 per mmol/L; 95% CI 1.15–1.24) in those with overweight/obesity.

Associations of glycemic variables with the incidence of complications did not differ substantially by sex. Incidence rates of complications stratified by sex for both data sets are listed in Supplementary Table 2. For the comparative data sets, incidence rates and Cox hazards models for complications stratified by 2-h PG and FPG are listed in Supplementary Tables 3 and 4, respectively. Because measurement of HbA_1c began later in the longitudinal study, the HbA_1c data set had a smaller sample size. However, associations of 2-h PG with complications were comparable in this data set to those observed in the larger data set.

Supplementary Table 5 shows the AUC comparisons between the three glycemic measures as continuous and categorical variables when predicting the three diabetes complications using the comparative data sets. There were no significant differences between AUCs for models with HbA_1c, 2-h PG, and FPG as continuous variables when predicting albuminuria, severe albuminuria, or retinopathy (Supplementary Fig. 1). When predicting albuminuria and severe albuminuria, the AUCs for models with HbA_1c, 2-h PG, and FPG as categorical predictors were not significantly different (Fig. 1). Some AUCs for the three glycemic tests analyzed as categorical predictors showed small increases compared with AUCs for the tests analyzed as continuous. When predicting retinopathy, categories of 2-h PG had a marginally significantly higher AUC than categories of HbA_1c (0.73 vs. 0.69, respectively; P = 0.046) (Supplementary Table 5).

Similar results were obtained when comparing AUCs for models in subsamples of participants with overweight/obesity at baseline. AUCs ranged from 0.60 to 0.75. AUCs among models with HbA_1c, FPG, and 2-h PG as predictors of the three microvascular complications did not differ significantly in those with overweight/obesity, when analyzed as continuous or categorical predictors.

The recent USPSTF recommendations for screening of asymptomatic children for prediabetes and type 2 diabetes identify a lack of long-term outcome data on diabetes-related morbidity and mortality that would justify screening for these conditions in childhood (16,17). We leveraged data from a 43-year longitudinal study in a high-risk American Indian cohort to determine if glycemic tests performed during childhood predicted diabetes-related microvascular complications. We found that higher levels of glycemia in children who did not have previously diagnosed diabetes (and therefore would have been eligible for screening) were associated with an increased risk of nephropathy and retinopathy and that HbA_1c, 2-h PG, and FPG demonstrated comparable ability in predicting both complications.

In this study, children with diabetes at baseline had the highest incidence of albuminuria, severe albuminuria, and retinopathy compared with children with prediabetes and normoglycemia. Most importantly, we found that in children without diabetes at baseline, the incidence of complications in those in the prediabetes groups was consistently higher than in those with normoglycemia. Also, in children without diabetes at baseline, increasing glycemic levels increased the risk of complications. This might be due to the higher incidence of diabetes in those with prediabetes during the follow-up period. In children without diabetes at baseline, including diabetes as a time-dependent covariate in the Cox models always attenuated the HRs for the risks of complications; for example, the HR for retinopathy per 1% HbA_1c increase was 3.09, but this HR changed to 2.04 (95% CI 0.78–5.38) when including diabetes as a time-dependent covariate.

Analyses of data combining children with and without diabetes at baseline showed that higher levels of HbA_1c, FPG, and 2-h PG were associated with a higher risk of both future nephropathy and retinopathy. In children without diabetes at baseline, 2-h PG was associated significantly with the risk of future albuminuria and severe albuminuria, and both 2-h PG and HbA_1c were associated with the risk of future retinopathy. When analyzing participants with and without diabetes separately, differences in the significance of the associations of HbA_1c and 2-h PG with the risk of outcomes may be due in part to the larger data sets obtained when only considering 2-h PG as the baseline glycemic measure. In data sets in which HbA_1c was the baseline measure of interest, associations of 2-h PG with risk of outcomes were like those in the larger data set with this measure (Supplementary Table 4).

We previously showed that HbA_1c measured during childhood is a useful predictor of future diabetes (7,30). Also, HbA_1c, recommended as a screening tool for prediabetes/diabetes by the ADA in asymptomatic high-risk children, is a convenient glycemic measure because it does not require fasting, reflects chronic glycemia, and can be used to monitor glycemia over a time period (30). A cross-sectional study in U.S. youth from the National Health and Nutrition Examination Survey found a stronger association of cardiometabolic risk with HbA_1c-defined hyperglycemia (≥5.7%) versus with FPG-defined hyperglycemia (≥100 mg/dL) (15). The above and the lack of differences in predictive performance for complications among glycemic measures in the comparative data sets, where power is equalized, favor HbA_1c as the preferred measure for screening, because it is more conveniently obtained in nonfasting children. The replication of most results in children with overweight/obesity supports the rationale for risk-based screening for prediabetes and/or diabetes in this subset of the population.

Ours is one of the few studies that have examined the association between glycemic tests performed in asymptomatic children with future diabetes-related microvascular complications in a high-risk group. A previous study in this population, examining the ability of 2-h PG and FPG to predict complications, suggested that diabetes diagnosis thresholds for adults were applicable in children because they identify a population with a higher risk of future microvascular complications (31). Another study showed that childhood obesity is a predictor of premature mortality (death at age <55 years) resulting from endogenous causes, and this association was partially mediated by the development of glucose intolerance and hypertension during childhood (8).

A limitation is that the study population has a high prevalence of type 2 diabetes, and findings may not be generalizable. However, prior scientific observations related to diabetes and its complications in this population have been widely replicated in other groups. Also, we do not have information about lifestyle measures that may have been undertaken in this cohort. Another limitation is that although 2-h PG was measured from study initiation in 1965, measurements of HbA_1c started in 1989. This did not allow us to compare the predictive ability of the three glycemic measures since the beginning of the study. However, findings in the comparative data sets generally replicated findings in the 2-hr PG data sets with longer follow-up.

The USPSTF recommendation is based on a meta-analysis (17). However, the authors did not find any study that compared health outcomes among asymptomatic children screened for prediabetes or diabetes with sufficient follow-up time to assess these outcomes (17). In our study, biennial examinations during a time period of 43 years allowed for a significant number of longitudinal evaluations to occur and reasonable follow-up time for diabetes-related complications to manifest. In the just-published recommendations for the evaluation and treatment of children and adolescents with obesity by the American Academy of Pediatrics, the evidence supporting testing for abnormal glucose metabolism in children and adolescents age ≥10 years with obesity was strong; in those with overweight, the evidence was only moderate (32).

Our current study shows a positive association between higher glycemia and future microvascular disease, and the findings also hold in children with overweight or obesity. Results in participants without type 2 diabetes at baseline suggest that detection of early hyperglycemia (lower than that defining diabetes) may be useful. These observations have important public health implications. They provide evidence that justifies screening of high-risk children, including those with overweight/obesity, to prevent long-term adverse health outcomes from diabetes.

Acknowledgments. The authors thank Dr. William C. Knowler, National Institute of Diabetes and Digestive and Kidney Diseases, for advice and discussions during study development. Dr. Knowler received no financial support for participation. The authors thank the participants and their parents for participation in the study.

Funding. This research was supported by the Intramural Research Program of the National Institute of Diabetes and Digestive and Kidney Diseases.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. L.V. obtained study data, carried out the analyses, assisted in drafting the manuscript, and approved the final manuscript as submitted. E.V.A. assisted in study design, obtained study data, carried out the data analyses, critically reviewed the manuscript, and approved the final manuscript as submitted. R.L.H. assisted in study design, data analysis, and critical review of the manuscript and approved the final manuscript as submitted. M.S. conceptualized the study, assisted in study design and data analysis, drafted the initial manuscript, and approved the final manuscript as submitted. M.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. Results were presented in part at the 83rd Annual Scientific Meeting of the American Diabetes Association, San Diego, CA, 23–26 June 2023.