Effect of Early Glycemic Control in Youth-Onset Type 2 Diabetes on Longer-Term Glycemic Control and β-Cell Function: Results From the TODAY Study

Rates of type 2 diabetes have been increasing in youth over the last three decades, and previous studies suggested that youth-onset diabetes has a more aggressive trajectory than adult-onset type 2 diabetes (1–4), necessitating a deeper understanding of the pathophysiology of youth-onset type 2 diabetes and response to treatment. The efficacy and durability of successful treatment of type 2 diabetes are determined to a great extent by the ability of a specific intervention to ameliorate insulin resistance and/or prolong or restore effective β-cell function. Substantial information is available on the natural history of insulin sensitivity and secretion and the effects of various treatment regimens on these pathophysiological components of type 2 diabetes in adults, but less is currently known about these factors in youth-onset type 2 diabetes.

Current data show that insulin resistance is more severe and deterioration in β-cell function is more rapid in youth-onset than adult-onset type 2 diabetes (4–6), and we previously showed in the Treatment Options for Diabetes in Adolescents and Youth (TODAY) study that poor β-cell function is the best predictor of loss of glycemic control (HbA_1c >8%) (7–10). In addition, we previously reported that during the first 2–4 years of the TODAY study, those who lost glycemic control (HbA_1c ≥8%) were more likely to have higher baseline HbA_1c, fasting glucose concentration, and glucose variability than those who did not lose glycemic control (8,10). In addition, those who reached glycemic failure during the first 2–4 years of the TODAY study continued to have the highest HbA_1c at study year 10, with intermediate HbA_1c in those whose HbA_1c was rising but was <8% and lowest HbA_1c in those who had stable HbA_1c <8% (10). However, the effect of early glycemic control and particularly the effect of early tight glycemic control (HbA_1c <5.7% vs. 5.7 to <6.4% vs. 6.4 to <7.0% vs. 7–8%) on long-term glycemic control and β-cell failure has not been examined in youth-onset type 2 diabetes. Therefore, the overarching goal of this analysis was to determine the independent effect of glycemic control during the first 6 months of the study on long-term β-cell function and glycemic control in youth-onset type 2 diabetes. We hypothesized that tight early glycemic control, independent of initial insulin sensitivity and secretion, would predict better long-term β-cell function and glycemic control. The aims of this report are to 1) examine the effect of early glycemic control on long-term β-cell function and glycemic control when accounting for initial insulin sensitivity and secretion and 2) determine the impact of sex, race/ethnicity, and BMI on these relationships.

The design of the TODAY study has been previously described (ClinicalTrials.gov indentifier NCT00081328) (11,12). In brief, 699 participants with type 2 diabetes (American Diabetes Association [ADA] 2002 criteria) diagnosed before the age of 18 years, with duration of diabetes <2 years, BMI >85th percentile for age and sex, negative islet cell antibodies, and C-peptide >0.6 ng/mL, were enrolled at 15 participating diabetes centers in the U.S.

After screening to determine eligibility, participants completed a 2- to 6-month prerandomization run-in period in which they demonstrated mastery of standard diabetes education, were weaned from nonstudy diabetes medications, demonstrated tolerance of metformin 500–1,000 mg twice daily, maintained HbA_1c <8% (<64 mmol/mol) monthly for at least 2 months on metformin alone, and demonstrated adherence to study medication and visit attendance (13). To be eligible, participants had to take ≥80% of the prescribed study drug (metformin) during the 8-week run-in period.

Following the run-in phase, between 2004 and 2011, participants were randomized to receive metformin alone, metformin plus rosiglitazone, or metformin plus an intensive lifestyle intervention program. Investigators, study personnel, and participants were all masked to treatment group assignment. The primary outcome of the TODAY study was to evaluate the effects of the three treatment arms on time to treatment failure, defined as loss of glycemic control (HbA_1c ≥8% [≥64 mmol/mol] for six consecutive months or failure to wean from temporary insulin after acute metabolic decompensation). Insulin was initiated when participants met the primary outcome. After an average of 3.9 years (range 2–6 years) of follow-up, 46% (n = 319) of participants reached the primary outcome of loss of glycemic control, including 51.7%, 46.6%, and 38.6% of participants randomized to metformin alone, metformin plus lifestyle intervention, and metformin plus rosiglitazone, respectively (14).

In 2011, 572 (82%) of the TODAY participants enrolled in the TODAY2 postintervention follow-up study. Between 2011 and 2014, participants no longer received randomized treatment but continued to receive diabetes care from the TODAY staff at 3-month intervals and were treated with metformin and/or insulin as indicated. From 2014 to 2020, 518 (74% of the original cohort) TODAY participants transitioned to community diabetes care and continued to be followed by the TODAY study group at annual observational visits. We have previously reported that the characteristics of the cohort were nearly identical across all study phases (1).

TODAY and TODAY2 were approved by institutional review boards at all participating institutions, and all participants and guardians provided written informed assent and/or consent as appropriate for age and local guidelines.

Demographics, medical history, and medication use (including prescribed insulin) were collected as previously described (11,12). Measurements of height, weight, and calculated BMI (kg/m²) were obtained every 2 months during the first year, quarterly up to 2014, and annually between 2014 and 2020. Measurements of fasting HbA_1c, insulin, C-peptide, and glucose were performed centrally at the TODAY Central Biochemistry Laboratory (Northwest Lipid Metabolism and Diabetes Research Laboratories, University of Washington, Seattle, WA) according to standardized procedures (11,12). Blood was not collected during pregnancy, during lactation, or immediately postpartum. HbA_1c was measured at screening, at randomization, and at every study visit thereafter with a dedicated high-performance liquid chromatography method (TOSOH Biosciences Inc., South San Francisco, CA).

Oral glucose tolerance tests (OGTTs) with a 2-h sample frequency were performed after a 10- to 14-h overnight fast as previously described (9,12). OGTTs were obtained from every participant at randomization, at month 6, and at years 2, 3, 4, 5, 6, and 9 postrandomization. OGTTs were rescheduled if participants took study medications or were not fasting on the morning of the OGTT visit. Analyses included all available data collected at OGTT-scheduled study visits during TODAY and TODAY2 through study year 9. Results were used to derive surrogate markers of the following parameters:

• An insulin sensitivity measure (1/fasting insulin, or 1/I_F) was shown to correlate more strongly (r = 0.82) with the hyperinsulinemic-euglycemic clamp than OGTT-based or other fasting estimates of insulin sensitivity

• Insulin secretion:

  • C-peptide index was calculated as the ratio of the incremental C-peptide and glucose responses over the first 30 min of the OGTT (△C_0–30/△G_0–30); C-peptide index is a measure of C-peptide secretion in response to an acute increase in glucose concentration

  • Oral disposition index (C-oDI) is the product of insulin sensitivity multiplied by the C-peptide index ([1/I_F] × △C_0–30/△G_0–30); C-oDI is a measure of insulin secretion relative to the degree of insulin demand (sensitivity)

  • Molar ratio of C-peptide to glucose concentrations for the respective area under the curve (AUC) during the first 30 min of the OGTT (C_{AUC 0–30}/G_{AUC 0–30}) (7,9,12).

Early glycemia was defined as mean HbA_1c during the first 6 months postrandomization (3.8 ± 0.5 [mean ± SD] evaluations, range 1–4) and was categorized into five groups: <5.7%, 5.7 to <6.4%, 6.4 to <7.0%, 7.0 to <8.0%, and ≥8.0%. Categories were determined based on a balance of clinically meaningful cutoffs and groups of adequate sample size. Long-term outcomes of β-cell function and glycemic control were defined as the period between study years 2 and 9.

We summarized data for participant characteristics using means and SD for continuous variables and counts and percentages for categorical variables stratified by the five groups of early glycemia. We used the χ² test to evaluate group differences across categorical variables and the Mann-Whitney U test for continuous variables. Separate linear mixed models were used to evaluate the effect of early glycemia (categorized into five groups) on the rates of change (slopes) in the quantitative outcomes of interest (long-term glycemia, insulin sensitivity, and secretion) over repeated time points. Regression coefficients are provided to describe the directionality of the slopes (e.g., increasing, decreasing, or flat). Insulin sensitivity and β-cell function outcomes were log transformed prior to testing. Unadjusted and adjusted models were considered before and after adjustment for early insulin sensitivity and secretion (mean during the first 6 months postrandomization) as well as other relevant covariates (i.e., duration of diabetes, randomized treatment group, BMI, age, sex, and race/ethnicity). Tests for covariate effects and their interactions with early-glycemia groups were performed. Nonsignificant interactions suggest parallel slopes across the five groups of early glycemia, whereas significant interactions suggest differential rates of change across the five groups of early glycemia. Regression lines at specific time points (study years 2 and 9) within each of the five groups of early glycemia were plotted to illustrate interaction effects. Analyses were performed using SAS (version 9.4 for Windows; SAS Institute, Cary, NC) and considered exploratory, with statistical significance defined as P < 0.05.

Data collected for the TODAY and TODAY2 studies are available to the public through the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) Repository at repository.niddk.nih.gov/studies/today/.

Of the 699 TODAY participants, 656 (94%) were included in this analysis, after excluding 22 participants found to have monogenic diabetes mutations after randomization and 21 participants with <6 months of follow-up. Mean (± SD) length of follow-up was 6.4 ± 3.2 years. Table 1 shows the demographic, screening, and baseline (following a minimum of 2 months of run-in of metformin treatment) characteristics of the cohort, stratified by the five groups of early glycemia. There were no sex, age, or screening BMI differences, or differences in glycemic change during run-in, between the early-glycemia groups. BMI at baseline was significantly different across the five groups (P = 0.0145), with the lowest BMI being among participants with an early mean HbA_1c <5.7%. However, baseline BMI did not show a consistent rise across early-glycemia categories, and the difference between groups was no longer significant after adjustment for race/ethnicity (P = 0.0847).

Figure 1 shows the rates of change (slopes) for long-term outcomes of interest (HbA_1c, insulin sensitivity, and the three surrogate measures of insulin secretion, C-peptide index, C-oDI, and AUC_0–30 C-peptide/glucose) across the five groups with early glycemia.

At study year 2, HbA_1c was lowest in the <5.7% early-glycemia group (mean 6.0%) and highest in the ≥8.0% early-glycemia group (mean 9.7%). HbA_1c significantly increased in all five groups during the 7-year period (Fig. 1A). By study year 9, mean HbA_1c was 1) <5.7% group, 8.5%; 2) 5.7–6.3% group, 9.4%; 3) 6.4–6.9% group, 10.4%; 4) 7.0–7.9% group, 11.2%; and 5) ≥8.0% group, 10.6% (P < 0.0001 for difference in HbA_1c at study year 9, with the value for the <5.7% early-glycemia group being significantly lower than those of all other groups). Participants with a mean HbA_1c <5.7% in the first 6 months in the study had a steeper rate of increase (+0.40% per year) in long-term HbA_1c compared with the four other groups (all P < 0.05). Adjustment for duration of diabetes at baseline, randomized treatment group, age, sex, race/ethnicity, early (first 6 months following randomization) mean BMI, or early mean insulin sensitivity as covariates in separate models or as covariates included together in a multivariable model did not impact the rates of change in HbA_1c. In contrast, when added to the model, early C-oDI (surrogate measure of insulin secretion relative to demand) was a significant covariate (P < 0.0001) and was negatively associated with long-term HbA_1c (i.e., higher levels of long-term HbA_1c correlated with lower levels of early mean C-oDI [regression β estimate ± SE = −0.39 ± 0.10]). However, early mean C-oDI affected long-term HbA_1c similarly across the five early-glycemia groups (i.e., no interaction). Rates of change in HbA_1c across the five groups of early glycemia were similar when evaluated among groups stratified by randomized treatment group (Supplementary Fig. 1).

At study baseline, insulin sensitivity was higher in the <5.7% early-glycemia group than in all other early-glycemia groups (all P < 0.05); however, these differences lessened with adjustment for race/ethnicity in the model (P = 0.0501) (Table 1). At study year 2, similar but slightly attenuated results were observed, and by study year 9, insulin sensitivity in the <5.7% group was no longer significantly different from that of any other group (Fig. 1). There was no change in insulin sensitivity during the 7-year study period in any of the early-glycemia groups (Fig. 1B). Adjustment in the models for duration of diabetes at baseline, randomized treatment group, age, and race/ethnicity had no impact on the rates of change. On average, during the long-term period, female participants had lower insulin sensitivity than male participants, irrespective of early-glycemia group (P = 0.003); however, the slopes for male and female participants did not differ within each early-glycemia group (i.e., no interaction). Early mean BMI was a significant covariate in the model for long-term insulin sensitivity (P = 0.0009), and a significant interaction between early mean BMI and the early-glycemia groups was noted (P = 0.045). At study year 2 (Supplementary Fig. 2A), negative associations between insulin sensitivity and early mean BMI were observed among participants in the <6.4% early-glycemia group, and no association was observed among those with higher levels of early glycemia. At study year 9, a significant negative association between insulin sensitivity and early mean BMI remained among participants in the <5.7% group; however, the association was attenuated and no longer significantly different from that of any other early-glycemia group (Supplementary Fig. 2B). As anticipated because of the limited change of BMI in the participants over time, these results were unaffected when BMI change was added as an additional covariate in the regression models.

The surrogate measures of insulin secretion (C-peptide index) and insulin secretion relative to demand (C-oDI) at baseline were the highest in the early-glycemia group with lowest HbA_1c and progressively lower across groups with higher HbA_1c (all P < 0.05), with C-peptide index and C-oDI in the 7.0–7.9% and ≥8.0% groups being similarly low (Table 1). These differences were maintained at study year 2. At study year 6, C-peptide index remained significantly higher in the <5.7% early-glycemia group than in all other groups (all P < 0.05) and higher in the 5.7–6.3% group than in the two groups with the highest HbA_1c (all P < 0.05). By study year 9, all differences in C-peptide index between early-glycemia groups disappeared, except it was lowest in the two early-glycemia groups with the highest HbA_1c (P = 0.01 and P = 0.009, respectively). These results were similar for C-oDI and AUC_0–30 C-peptide/glucose, the molar ratio of C-peptide to glucose concentrations during the first 30 min of the OGTT.

C-peptide index and C-oDI significantly decreased in participants with early glycemia <7.0% (all P < 0.05) during the 7-year period (Fig. 1C and D), with the steepest rate of decline in the <5.7% early glycemia group (−0.013 ng/mL per mg/dL per year and −0.014 mL/µU × ng/mL per mg/dL per year, respectively). AUC_0–30 C-peptide/glucose decreased during the 7-year period in all early-glycemia groups except for the ≥8.0% early-glycemia group (all P < 0.05) (Fig. 1E). When added to the models, early mean BMI was significantly associated with long-term C-peptide index (P = 0.006), and a borderline significant interaction between early mean BMI and early-glycemia groups was noted (P = 0.045). At study year 2, participants with early glycemia <6.4% and lower early mean BMI had lower C-peptide index than those with higher early mean BMI, whereas participants with early glycemia ≥6.4% had similar C-peptide index, irrespective of early mean BMI (flat slopes). Specifically, at study year 2, the slope within the 5.7–6.3% early-glycemia group significantly differed from that of the 6.4–6.9% early-glycemia group (Supplementary Fig. 3A). Similar results were observed at study year 6 but did not reach statistical significance (Supplementary Fig. 3B).

Similar results were obtained when early mean insulin sensitivity was added to the model for long-term C-peptide index, except the results were reversed (i.e., lower early mean insulin sensitivity correlated with higher C-peptide index, P = 0.02) (Supplementary Fig. 3C–D); however, unlike with early mean BMI, no interaction between early mean insulin sensitivity and early-glycemia groups was noted, suggesting parallel slopes across the five early-glycemia groups.

Results for AUC_0–30 C-peptide/glucose were comparable with those observed with C-peptide index (Supplementary Fig. 4). Early mean BMI was not associated with long-term C-oDI, and the interaction between early mean BMI and early-glycemia groups was not significant (data not shown).

The overall goal of the current analysis was to address the hypothesis that tight early glycemic control (first 6 months after randomization), independent of initial insulin sensitivity and secretion, would predict better long-term (2–9 years after randomization) β-cell function and glycemic control in participants with youth-onset type 2 diabetes from the TODAY study. Early glycemia was unrelated to sex, screening age, BMI, baseline BMI (following 2 months of metformin during the run-in phase and after adjustment for race/ethnicity), glycemic change during run-in, or randomized treatment group, which confirms our previous findings (7,13). In contrast, early glycemia was higher in youth with longer diabetes duration at screening, lower baseline insulin sensitivity and β-cell function indices, and less BMI loss during run-in, and it varied by race/ethnicity. We now demonstrate the new finding that HbA_1c significantly increased in all five early-glycemia groups over time, even in the groups with the tightest initial glycemic control.

Thus, even in the three categories of youth where ADA guidelines for glycemic control (HbA_1c <7.0%) (15) were initially met or exceeded, β-cell failure still progressed. HbA_1c increased most quickly in the group with the lowest level of glycemia (HbA_1c <5.7%), demonstrating that even very tight early glycemic control alone does not improve the trajectory of diabetes control in the long run. Moreover, we also now demonstrate that despite the ability to produce goal glycemic control (HbA_1c <7%) in ∼80% of youth and very tight glycemic control (HbA_1c <5.7%) in ∼40% of youth, the treatments used in the TODAY study failed to slow β-cell failure in youth-onset type 2 diabetes. The only early (first 6 months after randomization) variable impacting the rate of change in HbA_1c over time was C-oDI, a measure of insulin secretion relative to demand, where lower early β-cell function correlated with higher long-term HbA_1c, irrespective of early glycemia. These results confirm the importance of underlying β-cell function in determining the course of youth-onset type 2 diabetes and support our previous findings that initial β-cell function is strongly predictive of subsequent short-term (2–4 years) and long-term (study year 10) glycemia (8,10). Of interest, a recent publication reported that tight glycemic control achieved in youth with recent-onset type 1 diabetes using continuous glucose monitoring to determine insulin infusion rate did not help preserve β-cell function. Although the pathophysiologies of type 1 and type 2 diabetes differ, these results both highlight the impact of underlying β-cell dysfunction at diagnosis and raise questions about the assumption that rapid attainment of tight glucose control is beneficial in slowing loss of β-cell function (16).

Insulin sensitivity is very low in youth with type 2 diabetes (5,6). We previously showed that insulin resistance was not predictive of glycemic failure (7), and our current analysis supports this finding by now showing no change in insulin sensitivity in any early-glycemia group over time despite using interventions hypothesized to improve insulin sensitivity (7,12). We also now show that among participants with lower early glycemia (<6.4%), higher BMI was correlated with lower insulin sensitivity, as expected. However, among higher levels of early glycemia (>6.4%), i.e., when β-cell function was already significantly compromised, insulin sensitivity was no longer significantly related to BMI, likely due to the association of persistent hyperglycemia and weight loss, disturbing the relationship between BMI and insulin sensitivity.

As reported in the Restoring Insulin Secretion (RISE) study (17), β-cell function in youth with type 2 diabetes did not improve with metformin in the TODAY study, and it did not improve in the TODAY study with rosiglitazone or intensive lifestyle intervention. Furthermore, in our current study, β-cell function significantly decreased over time in those with lower early glycemia, i.e., those who still had remaining β-cell function to lose, whereas those with early HbA_1c ≥7% and ≥8% had no further decline, implying that participants with early glycemia in the ≥7–8% range already had advanced loss of β-cell function. Similarly, in our current analysis, among participants with lower early glycemia (<6.4%), C-peptide index was higher in those with higher early BMI or lower early insulin sensitivity, likely because higher BMI and lower insulin sensitivity increase the demand for insulin secretion. The C-peptide index decreased over time, however, reflecting progressive β-cell failure. In contrast, in those participants with higher early glycemia (≥6.4%), there was no significant increase in C-peptide index with higher early BMI or lower early insulin sensitivity, suggesting that C-peptide secretion is already constrained once HbA_1c ≥6.4–7%.

Strengths of this study include the large sample size of youth with type 2 diabetes who were deeply phenotyped longitudinally and had a wide range of initial glycemic control. The study had a diverse sample regarding race/ethnicity and geographic location, standardized protocols, a robust definition of type 2 diabetes, including exclusions of participants with monogenic diabetes, measurements of HbA_1c every 3 months, and serial OGTTs performed up to 7 times postrandomization. Another strength of the study is it provides us the ability to assess the impact of screening BMI and HbA_1c and the impact of multiple variables after treatment with 2 months of metformin (the current first-line treatment for youth-onset type 2 diabetes). Limitations include inability to assess the impact of any medications used prior to enrollment in the TODAY study. While a potential limitation in longitudinal studies is loss of participants over time, 94% of the youth enrolled in the TODAY study (n = 656) had longitudinal data available, and we have previously shown that the characteristics of the long-term cohort align very well with the initial full TODAY cohort (1). In addition, while more robust than fasting measures alone, due to budget limitations imposed by a large sample size, insulin sensitivity and secretion were assessed by 2-h OGTT rather than hyperinsulinemic-euglycemic or hyperglycemic clamps, respectively.

Results from the current analyses suggest that even youth with type 2 diabetes who are in what has traditionally been considered tight initial glycemic control still experience significant deterioration of glycemic control over time. The ADA’s treatment targets are currently set at 7%, yet HbA_1c 6.4% or higher indicates severe β-cell dysfunction on multiple measures in the TODAY study. In alignment with this finding, we previously reported that HbA_1c 6.3% after initiation of metformin distinguishes between those who will have prolonged control on oral medication and those who will have deterioration more quickly (8,18). Taken together, these results suggest that intensification of therapy in youth-onset type 2 diabetes should be considered in individuals with HbA_1c >6.3%. Targeting lower HbA_1c with newer medications that do not cause hypoglycemia may result in improved long-term glycemia and prevent diabetes complications. However, future studies are needed to better understand the heterogeneity of type 2 diabetes and identify therapies that halt the deterioration of β-cell function.

Metformin, rosiglitazone, and intensive lifestyle interventions did not improve insulin sensitivity in youth with type 2 diabetes in the TODAY study (19). Metformin or insulin followed by metformin in youth with prediabetes or type 2 diabetes in the RISE study (2), and metformin treatment of normoglycemic youth with obesity in the Health Influences of Puberty (HIP) study, did not improve insulin sensitivity (20). Moreover, β-cell dysfunction is a critical determinant of the disease course in youth-onset type 2 diabetes, and the fact that β-cell dysfunction responded poorly to standard treatments used in the TODAY, RISE, and HIP studies highlights the need for better therapies aimed at preserving or improving β-cell function. Of note, the TODAY study was completed prior to the availability to youth of newer pathophysiology-focused agents, such as glucagon-like peptide 1 (GLP-1) agonists or sodium–glucose cotransporter 2 (SGLT-2) inhibitors. Recently published studies have now demonstrated the glycemia-lowering effects of multiple GLP-1 agonists and one SGLT-2 inhibitor, but these have been short-term studies, and research is needed to understand the long-term effects of these agents on disease course and β-cell function in youth. Finally, a better understanding of the natural history of insulin sensitivity and secretion during puberty and a clearer delineation of the factors that drive poor β-cell function in some youth are needed to design better prevention and treatment strategies.

TODAY Study Group Writing Committee. The writing committee for this work included Kristen J. Nadeau, University of Colorado Anschutz Medical Campus, Aurora, CO, Laure El ghormli, The Biostatistics Center, George Washington University, Rockville, MD, Silva Arslanian, University of Pittsburgh, UPMC Children’s Hospital of Pittsburgh, Pittsburgh, PA, Fida Bacha, Children’s Nutrition Research Center, Baylor College of Medicine, Houston, TX, Sonia Caprio, Yale University, New Haven, CT, Christine Chan, University of Colorado Anschutz Medical Campus, Aurora, CO, Lily C. Chao, Children’s Hospital Los Angeles, Los Angeles, CA, Elvira Isganaitis, Joslin Diabetes Center, Boston, MA, Maria Rayas, UT Health San Antonio, San Antonio, TX, Maggie K. Siska, St. Louis University Health Sciences Center, St. Louis, MO, and Philip Zeitler, University of Colorado Anschutz Medical Campus, Aurora, CO.

Acknowledgments. A complete list of individuals in the TODAY Study Group is presented in the Supplementary Material. The authors gratefully acknowledge the participation and guidance of the American Indian partners associated with the clinical center located at the University of Oklahoma Health Sciences Center, including members of the Absentee Shawnee Tribe, Cherokee Nation, Chickasaw Nation, Choctaw Nation of Oklahoma, and Oklahoma City Area Indian Health Service.

The opinions expressed in this article are those of the authors and do not necessarily reflect the views of the respective tribes and the Indian Health Service.

Funding and Duality of Interest. The following companies provided donations in support of the TODAY study’s efforts: Becton, Dickinson and Company, Bristol-Myers Squibb, Eli Lilly and Company, GlaxoSmithKline, LifeScan, Inc., Pfizer, and Sanofi. This work was completed with funding from the NIDDK and the National Institutes of Health Office of the Director through grants U01-DK61212, U01-DK61230, U01-DK61239, U01-DK61242, and U01-DK61254. No other potential conflicts of interest relevant to this article were reported.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The NIDDK project office was involved in all aspects of the study, including design and conduct, collection, management, analysis, and interpretation of the data, review and approval of the manuscript, and decision to submit the manuscript for publication.

Author Contributions. K.J.N. and P.Z. proposed the analyses, interpreted data, and wrote and edited the manuscript. L.E.g. conducted the statistical analyses, wrote parts of the manuscript, and edited the manuscript. S.A., F.B., S.C., C.C., L.C.C., E.I., M.R., and M.K.S. interpreted data and reviewed and edited the manuscript. K.J.N. and L.E.g. are the guarantors of this work and, as such, had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.