Longitudinal Effects of Glucose-Lowering Medications on β-Cell Responses and Insulin Sensitivity in Type 2 Diabetes: The GRADE Randomized Clinical Trial

Both impaired glucose tolerance and type 2 diabetes are characterized by insufficient insulin secretion to compensate for the metabolic demand resulting from insulin resistance (1). Further, β-cell function progressively declines over time in type 2 diabetes, resulting in deterioration of glycemic control and the need for more glucose-lowering medication(s). The progressive loss of β-cell function (5–10% per year) (2) thus represents one of the most important challenges of maintaining long-term glycemic control in people with type 2 diabetes (3). Therefore, identifying effective interventions to delay or prevent this deterioration would be of great value (4).

Three glucose-lowering medications (rosiglitazone, glyburide, and metformin) when used long-term as initial monotherapy in recently diagnosed type 2 diabetes exhibited time-dependent differences in glycemic control that were related to differences in their mechanisms of action and ultimate impact on β-cell function and insulin sensitivity (5,6). However, studies comparing the long-term effects on glucose metabolism of combinations of glucose-lowering medications, including incretin-based therapies and basal insulin, in large and diverse populations are lacking.

Glycemia Reduction Approaches in Diabetes: A Comparative Effectiveness Study (GRADE) compared the glycemic control durability of agents from four distinct classes of glucose-lowering medications—long-acting insulin (insulin glargine U-100), sulfonylurea (glimepiride), glucagon-like peptide 1 receptor agonist (liraglutide), and dipeptidyl peptidase 4 inhibitor (sitagliptin)—when added to metformin. They were added to metformin in >1,250 study participants per treatment group, who were then followed for an average of 5 years (7). As previously reported, glargine and liraglutide were significantly, albeit modestly, more effective in achieving and maintaining the primary target glycated hemoglobin (HbA_1c) level (<7% [53 mmol/mol]) in the entire study cohort (7).

As the four medications exert different effects by providing exogenous insulin (glargine), stimulating insulin secretion (glimepiride, liraglutide), enhancing incretin action (sitagliptin, liraglutide), and modifying appetite and weight (liraglutide), we hypothesized that the differential long-term glucose-lowering effects could be due to differences in their effects on the physiology of glucose metabolism. In designing the study, we included serial oral glucose tolerance tests (OGTTs), allowing us to undertake the current secondary analysis to examine the effect over time of these four medications on insulin sensitivity, β-cell responses, and β-cell function. As the goal of this analysis was to describe changes in measures of insulin sensitivity and β-cell function while participants were only on their originally assigned dual therapy, we did not perform an intention-to-treat analysis. Our findings highlight that 1) the medications’ effects on the character of the β-cell responses differ, 2) the greatest effect to enhance β-cell responses occurred after a year of intervention, and 3) after the first year, β-cell responses declined in all four medication groups.

GRADE was a randomized, open-label, parallel-arm, prospective multicenter clinical trial comparing the effectiveness of four glucose-lowering medications (glargine, glimepiride, liraglutide, and sitagliptin) when added to baseline metformin in glycemic control over time. All participants provided written informed consent, and the study was approved by each center’s institutional review board. Details of the study design have previously been published (7,8). Briefly, participants were eligible to participate if they had been diagnosed with type 2 diabetes for <10 years at the time of screening and were ≥30 years of age (≥20 years if American Indian) at the time of diagnosis, with HbA_1c 6.8%–8.5% (51–69 mmol/mol), and taking at least 1,000 mg metformin daily at the end of the run-in period with a goal of 2,000 mg. After randomization, HbA_1c was measured quarterly. The primary metabolic outcome was HbA_1c ≥7% (53.0 mmol/mol). The secondary metabolic outcome was HbA_1c >7.5% (58.5 mmol/mol). Both outcomes required confirmation. At the time of the secondary metabolic outcome, glargine therapy was added in the three noninsulin treatment groups and intensification of insulin therapy (adding short-acting insulin) was initiated in the treatment group randomized to glargine treatment (7,8).

Per protocol, an OGTT was performed at baseline and serially at years 1, 3, and 5 (9). After a minimum of 8 h fasting overnight, the participant consumed a 75-g glucose drink within 5 min, and blood samples were drawn at 0, 30, 60, 90, and 120 min relative to the start of glucose ingestion. Participants were instructed to hold their morning dose of glucose-lowering medication(s) on the OGTT day. Participants in the glargine group took their usual dose of medication the night before the OGTT. All specimens were immediately placed on ice, then separated in a refrigerated centrifuge, and then frozen at −80°C. They were subsequently shipped to the study’s central laboratory at the University of Minnesota, Minneapolis, MN. Before assay, specimens were thawed at room temperature. The full protocol is available from https://grade.bsc.gwu.edu.

HbA_1c was measured in EDTA whole blood on the Automated Glycohemoglobin Analyzer HLC-723G8 (Tosoh Medics, San Francisco, CA) with an automated high-performance liquid chromatography method. Calibration of this method was evaluated with use of standard values derived by the NGSP. Glucose was measured in EDTA plasma by a hexokinase method on a cobas c501 chemistry analyzer (Roche Diagnostics, Indianapolis, IN). C-peptide was measured in EDTA plasma by immunoassay per the manufacturer’s instructions on a Roche cobas 6000 system (Roche Diagnostics).

HOMA of insulin sensitivity (HOMA2-%S) and β-cell function (HOMA2-%B) was calculated from C-peptide data with the HOMA2 Calculator, version 2.2.3 (Diabetes Trials Unit, University of Oxford, Oxford, U.K.) (10,11).

The early and total C-peptide responses to glucose during the OGTT were calculated as β-cell response measures. The C-peptide index (CPI) was computed as the increment in C-peptide above fasting (0 min) over the first 30 min of the OGTT divided by the increment in glucose over the same period: 100(C₃₀ − C₀)/(G₃₀ − G₀) (12). The C-peptide and glucose mean areas under the curve (AUCs) were calculated as C-peptide AUC/120 and glucose AUC/120, respectively. The incremental mean AUC (incAUC) for C-peptide and glucose was calculated as mean AUC minus fasting values, respectively. The total C-peptide response was calculated as the ratio of the incAUC of C-peptide to incAUC of glucose: 100 * (C-peptide incAUC)/(glucose incAUC).

The full GRADE cohort at randomization included 5,047 participants. Participants who did not have a baseline OGTT (n = 30), or those who did not attend at least one follow-up visit postrandomization and/or did not take at least one dose of their assigned medication (n = 216), were excluded, resulting in a total of 4,801 participants included in this analysis (Supplementary Fig. 1). The prespecified primary analysis included all 4,801 participants but not all of their visits: once a participant permanently discontinued metformin or the assigned second glucose-lowering medication (glargine, glimepiride, liraglutide, or sitagliptin), or added any other glucose-lowering medication, all the data from their subsequent visits were dropped from the analysis.

For further investigation of the effects of the study medications on β-cell responses, three sensitivity analyses were performed for those participants who remained on their assigned original study medication through year 1, year 3, and year 5, respectively. Of note, these analyses excluded participants who, at any time up to year 1, year 3, or year 5, had permanently discontinued either metformin or the assigned medication or added any other glucose-lowering medication. This is in contrast to the primary analysis estimated the curves using all participants, but only visits up to any deviation from the original study medication assignment.

For each measure (fasting C-peptide, fasting glucose, HOMA2-%S, total C-peptide response, CPI, and HOMA2-%B), we fit a general estimating equation (GEE) model with (AR)1 correlation structure, where AR is autoregressive (13,14), at years 1, 3, and 5 adjusting for the baseline value, with appropriate terms for visit number, treatment group, and a treatment-by-visit interaction. In addition, early and total β-cell responses (CPI and total C-peptide response), were adjusted for insulin sensitivity with use of HOMA2-%S to provide a measure of β-cell function. We assessed treatment group differences by testing the treatment-by-visit interaction term in each GEE model. If the interaction term was significant at the 0.05 level, the means of the measure in four treatment groups were compared with each other at every time point (years 1, 3, and 5) using six pairwise tests based on contrasts within the GEE model. The resulting 18 P values were jointly adjusted for multiple comparisons with use of the Benjamini-Hochberg false discovery rate adjustment.

OGTT profiles of glucose and C-peptide were also generated at baseline and years 1, 3, and 5 with a GEE model with terms for treatment group and OGTT time point (fasting, 30, 60, 90, and 120 min). The curves show the least squares means of glucose and C-peptide at each OGTT time point along with their pointwise 95% CIs from the GEE model.

This article is based on follow-up data and outcome assessments from the 5,047 participants enrolled into GRADE. This database will be available in the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) Central Repository in 2024.

Baseline characteristics of the 4,801 participants are summarized in Table 1. Mean ± SD age for the cohort was 57.1 ± 10 years and BMI 34.3 ± 6.8 kg/m², and 36.2% were women. Participants self-identified as 66.2% White, 19.4% Black, and 18.3% Hispanic. Self-reported duration of diabetes was 4.2 ± 2.7 years. None of these were significantly different from the baseline characteristics of the entire cohort.

At baseline, mean ± SD HbA_1c levels were 7.5% ± 0.5% (58.4 ± 5.3 mmol/mol) and levels did not differ between treatment groups (Table 1). In keeping with what was observed in the whole cohort (7), the liraglutide group had the lowest HbA_1c at year 1 while the sitagliptin group had the highest (Fig. 1A and Supplementary Table 1). At year 3, HbA_1c was highest with glimepiride, while it was similar among the other three groups. At year 5, there were no differences in HbA_1c among the four treatment groups (Fig. 1A and Supplementary Table 1).

Mean baseline fasting glucose and C-peptide levels were similar among the four groups, with clear treatment differences during the study (P < 0.001 for all) (Fig. 1B and C and Supplementary Tables 2 and 3). Overall, fasting glucose reached a nadir in all treatment groups at year 1 and subsequently increased gradually (Fig. 1B). Glargine had the lowest fasting glucose at years 1, 3, and 5. Among noninsulin groups, the liraglutide group had lower fasting glucose than the sitagliptin and glimepiride groups at all follow-up time points except year 5, when the difference between the liraglutide and glimepiride groups was not significant (Fig. 1B).

Fasting C-peptide decreased by 29% from baseline to year 1 with glargine (Fig. 1C and Supplementary Table 3). After year 1, fasting C-peptide gradually declined in all four groups (Fig. 1C). The glimepiride group had the highest fasting C-peptide levels at years 1, 3, and 5 (Fig. 1C). Fasting C-peptide was similar between the liraglutide and sitagliptin groups at all follow-up time points (except at year 1, when the liraglutide group had a statistically higher fasting C-peptide) (Fig. 1C).

We also examined the glucose and C-peptide profiles during the OGTT at the follow-up time points. At baseline, glucose and C-peptide levels after the glucose load were similar among the four treatment groups (Figs. 2A and B and Supplementary Tables 2 and 3). However, postload glucose and C-peptide levels differed among the four groups at all on-treatment time points (Fig. 2C–H). Postload glucose was higher in the glimepiride group and lowest in the liraglutide group at follow-up years (Fig. 2C, E, and G). The difference among groups was more prominent during 60–120 min of the OGTT. As with fasting C-peptide, the glargine group had the lowest stimulated C-peptide levels at all follow-up time points, but the profile of C-peptide during the OGTT with glargine paralleled that of glimepiride and sitagliptin (Fig. 2D, F, and H). Incretin-based treatments (liraglutide and sitagliptin) had distinct patterns for stimulated C-peptide curves (Fig. 2D, F, and H). Like the fasting C-peptide, the 30-min C-peptide levels were not different between liraglutide and sitagliptin, but the curves diverged after 30 min and the stimulated C-peptide levels at 60, 90, and 120 min were progressively higher with liraglutide compared with sitagliptin at the follow-up time points (Fig. 2D, F, and H).

The baseline HOMA2-%S in the entire cohort was 34.1 ± 15.5 units and did not differ between groups (Table 1). Except for an increase in HOMA2-%S from baseline to year 1 (from mean 34.39 [95% CI 34.34, 34.44] to 57.71 [56.06, 59.37]) with glargine, insulin sensitivity remained stable during the follow-up period. This resulted in treatment differences in this measure (P < 0.001); HOMA2-%S was significantly higher in the glargine group compared with the other three treatment groups at year 1 and remained so for the duration of follow-up (Fig. 3A and Supplementary Table 4). Among the noninsulin groups, the sitagliptin and liraglutide groups had similar HOMA2-%S during the study, which was modestly higher than that of the glimepiride group (Fig. 3A).

Baseline HOMA2-%B was similar among groups (Table 1), but there were treatment differences at the follow-up time points (P < 0.001) (Fig. 3B and Supplementary Table 5). HOMA2-%B was significantly higher at all follow-up time points in the liraglutide and glimepiride groups in comparison with the sitagliptin and glargine groups. Furthermore, HOMA2-%B was similar between liraglutide and glimepiride groups except at year 1, when it was significantly higher with liraglutide. HOMA2-%B was similar between sitagliptin and glargine at the different follow-up time points, except at year 3, when sitagliptin was higher than glargine.

The baseline measures of early and total C-peptide responses were similar among the four treatment groups (Table 1). All four treatments significantly increased the OGTT β-cell response relative to that of glucose over the first year whether it was the early (30 min) or the total (120 min) response to glucose load; however, the response declined gradually over time (Fig. 3C and D).

At all follow-up time points (years 1, 3, and 5), the early C-peptide response was mostly higher with the incretin-based treatments, followed by glargine and glimepiride (Fig. 3C and Supplementary Table 6). Liraglutide and sitagliptin had similar CPI in all on-treatment time points (except CPI was higher in sitagliptin at year 3). With sitagliptin, CPI was greater than glargine and glimepiride at all three time points. Further, liraglutide had a repeatedly higher CPI compared with glimepiride at all follow-up time points and similarly higher CPI compared with glargine at years 1 and 3, but the difference was not significant at year 5. CPI was higher with glargine compared with glimepiride at all follow-up time points.

The total C-peptide response was greatest with liraglutide, followed in descending order by sitagliptin, glargine, and glimepiride (Fig. 3D and Supplementary Table 7). While over time the absolute values of these measures decreased, the rank order was maintained throughout the study, with liraglutide producing the highest and glimepiride the lowest response to the glucose load.

As stimulated β-cell responses are modulated by insulin sensitivity, we also adjusted the CPI and total C-peptide response for HOMA2-%S at each time point to provide a measure of β-cell function (Fig. 3E and F). Doing so demonstrated differences between treatment groups, with the patterns being very similar to those observed in examining the β-cell responses except that the differences in CPI between glargine and glimepiride at years 1 and 3 became nonsignificant.

When we performed a sensitivity analysis examining the response of the β-cell (HOMA2-%B, CPI, total C-peptide response) among treatment groups in only the participants who remained on their original study medication (Supplementary Figs. 2–4), the findings were similar to those when we examined the full cohort.

In the GRADE cohort, C-peptide responses showed variable degrees of improvement at year 1, but gradually declined over time, independent of the randomized treatment. In contrast, insulin sensitivity remained essentially stable after the first year. Thus, β-cell function declined over time as glycemia deteriorated. These findings are in keeping with those from two other major studies (UK Prospective Diabetes Study [UKPDS] and A Diabetes Outcome Progression Trial [ADOPT]) in showing that loss of β-cell function is the more important driver for glycemic progression despite intervention (2,6). In GRADE, we have uniquely extended these observations by demonstrating that the three non-insulin-based therapies (sitagliptin, liraglutide, and glimepiride), while having different effects on the β-cell, were all unable to halt the progressive loss of β-cell function and glycemic deterioration over time. Further, their differential effect on glycemic control would appear to be driven by the magnitude of change in the measures of β-cell function that they achieved during the first year of intervention. Notably, we found that even treatment with the basal insulin glargine could not slow the progressive loss of β-cell function despite its ability to “rest” the β-cell.

Even though we did observe that C-peptide responses deteriorated over time, there were differences in the response of the β-cell to each of the four interventions. Liraglutide resulted in a greater total C-peptide response to oral glucose at all follow-up time points, followed by sitagliptin, glargine, and glimepiride, with the difference being most prominent at year 1. The robust enhancement of the β-cell response after 1 year of treatment with liraglutide is consistent with results of other studies (15,16). However, the sharper decline in β-cell function with liraglutide (compared with other treatment groups) after year 1 suggests that the beneficial effects of liraglutide are not durable and the results of short-term studies cannot be extrapolated to a long-term effect. Despite the gradual decline, the total C-peptide response to glucose with incretin-based treatments at the end of the study was still greater than at baseline. With glimepiride, it was lower than at baseline, demonstrating that sulfonylureas might have adverse effects on β-cell function over time. In the case of glargine, the lower C-peptide may represent in part the effect of insulin replacement to obviate the need for the β-cell to release insulin (17). These observations were largely similar in considering the early β-cell response with the exception that the two incretin-based therapies had similar effects on the early response.

The two incretin-based therapies (sitagliptin and liraglutide) exerted different effects on early and total C-peptide responses. In the first 30 min of the OGTT, the C-peptide and glucose increments were similar with sitagliptin and liraglutide. However, they differed during the 30- to 120-min period, with liraglutide having a more robust C-peptide response that was associated with a lower glucose profile, particularly toward the end of the test. These findings have two implications. First, by virtue of liraglutide’s prolonged action, pharmacological dosing, and direct action on the β-cells, liraglutide’s effect was greater and explains the superior glucose profile compared with that observed with dipeptidyl peptidase 4 inhibition. Second, the lower glycemic responses observed during the OGTT with liraglutide likely were mirrored during everyday life and thus contributed to our observation that liraglutide was superior to sitagliptin in long-term glycemic control in GRADE (7).

Glimepiride increased fasting C-peptide with very little effect on the postglucose responses, suggesting that this medication predominantly increases the basal release of insulin from the β-cell. This contrasts with liraglutide, which increased the fasting as well as the early and total C-peptide responses, while sitagliptin had no effect on fasting C-peptide and predominantly affected the early C-peptide response. It was of interest that the increase above baseline in HOMA2-%B with glimepiride was maintained over time, while the postglucose load measures increased slightly at year 1 and then declined to be subsequently lower than the baseline, which was different from what was observed with the incretin-based medications.

Our observations with glargine need to be considered in view of its relation to endogenous insulin. As mentioned previously, the glargine group took their usual dose of insulin the night before OGTT, while other participants held their morning dose of medications on the OGTT day. Due to the long half-life of glargine, the lower fasting C-peptide could reflect in part the physiologic effects of glargine on endogenous insulin secretion, as exogenous insulin infusion inhibits endogenous insulin release by 50–70% (18,19), as well as the decrease in fasting glucose. While glargine has the ability to reduce insulin secretory demand, the OGTT C-peptide responses highlight that the β-cell is still responsive. These responses also decrease over time, suggesting a loss of function. It should be noted that we used C-peptide for these measures as glargine is metabolized to active metabolites (predominately M1) that are not fully recognized in the insulin assay (20). Importantly, this inability of the insulin assay to recognize all glargine metabolites means that the calculation of HOMA2-%S using insulin is not feasible. And, given fasting C-peptide will be lower because of the metabolic activity of the glargine metabolites, the ∼60% increase in HOMA2-%S in the glargine group at year 1 is an artifact and likely overrepresents any improvement in insulin sensitivity in these participants. As HOMA2-%S remained unchanged in the glargine group, as for other participants, after year 1, this would mean that the β-cell responses over time likely reflect changes in β-cell function.

It is important to emphasize that the HOMA2-%B results were not concordant with the OGTT-driven measures of β-cell function, indicating that one needs to be careful in interpreting different measures of β-cell function in response to glucose-lowering agents with different effects on the physiology of glucose metabolism. HOMA2-%B is a static measure, and extrapolation of results to a more dynamic measure of β-cell function (i.e., OGTT) should be made with caution. The observation that liraglutide’s effect on this measure was similar to that of glimepiride and greater than that of sitagliptin provides further insight into the greater glucose-lowering effect of the glucagon-like peptide 1 receptor agonist. Collectively, these long-term data highlight the differential effects of these medications on fasting and postprandial β-cell function, a novel observation obtained through comparing them in GRADE.

By examining representative medications from four different classes of glucose-lowering agents, we have extended the findings from ADOPT. In ADOPT, glyburide, metformin, and rosiglitazone were compared for a median of 4 years and demonstrated more durable glycemic control with rosiglitazone and metformin (5,6). The response to glyburide in ADOPT was similar to the response to glimepiride in GRADE in that β-cell function improved initially but thereafter steadily declined (5). In contrast, both metformin and rosiglitazone improved insulin sensitivity and β-cell function declined less rapidly. Thus, perhaps reducing secretory demand on the β-cell by improving insulin sensitivity may contribute to the preservation of β-cell function. While liraglutide produced a mean weight loss of 3.5 kg and improved insulin sensitivity, it did not slow β-cell function loss. Whether the newer incretin-based therapies that produce greater weight loss (21,22) may have a more beneficial effect on β-cell function over time needs to be determined. Further, it would be informative to know whether the sodium–glucose cotransporter 2 inhibitors that reduce glycemia by inducing glycosuria are able to produce long-term benefits for the β-cell (23).

GRADE itself and the current analyses have both strengths and limitations. GRADE was a long-term study with a large sample size. The large number of participants followed for an average of 5 years precluded us from using more sophisticated measures of β-cell function and insulin sensitivity (such as the clamp and frequently sampled intravenous glucose tolerance test), but it did enable us to measure responses and calculate a surrogate measure of insulin sensitivity from the serial OGTTs to examine the effects of four different glucose-lowering agents on β-cell responses, β-cell function, and insulin sensitivity and the durability of these effects. We did not use the disposition index (the product of insulin response and insulin sensitivity) as a measure of β-cell function as this requires demonstration of a rectangular hyperbola describing the relationship between these two measures. As this relationship does not exist for the C-peptide responses and HOMA2-%S, we statistically adjusted the C-peptide responses for insulin sensitivity. In the current analysis we have not examined how different phenotypic characteristics determine the responses, as recently done for ADOPT (24,25); this is an area that should be pursued in the future. Lastly, as mentioned, GRADE did not include other medication classes. Therefore, a long-term study with serial measurements of insulin sensitivity and β-cell responses is needed to determine whether other classes or different combinations of glucose-lowering medications may have more a durable effect on β-cell function or insulin sensitivity.

In summary, GRADE has provided new and valuable insights from the long-term data on β-cell responses and insulin sensitivity during treatment with four different glucose-lowering medications. Overall, despite these interventions having different mechanisms of action, progressive deterioration of β-cell function and loss of glycemic control were observed in the case of all four medications when they were added to metformin.

Acknowledgments. The GRADE Research Group is deeply grateful for the participants, whose loyal dedication made GRADE possible.

Funding. GRADE was supported by a grant from the NIDDK of the National Institutes of Health (NIH) under award no. U01DK098246. The planning of GRADE was supported by a U34 planning grant from the NIDDK (U34-DK-088043). The American Diabetes Association supported the initial planning meeting for the U34 proposal. The National Heart, Lung, and Blood Institute and the Centers for Disease Control and Prevention also provided funding support. The Department of Veterans Affairs provided resources and facilities. Additional support was provided by grants P30 DK017047, P30 DK020541-44, P30 DK020572, P30 DK072476, P30 DK079626, P30 DK092926, U54 GM104940, UL1 TR000439, UL1 TR000445, UL1 TR001108, UL1 TR001409, 2UL1TR001425, UL1 TR001449, UL1 TR002243, UL1 TR002345, UL1 TR002378, UL1 TR002489, UL1 TR002529, UL1 TR002535, UL1 TR002537, 2UL1 TR001425, and UL1 TR002548. Educational materials have been provided by the National Diabetes Education Program. Material support in the form of donated medications and supplies has been provided by Becton, Dickinson and Company, Bristol-Myers Squibb, Merck & Co., Novo Nordisk, Roche Diagnostics, and Sanofi.

The content of this manuscript is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Duality of Interest. N.R. reports grants, personal fees, and nonfinancial support from Novo Nordisk; grants, personal fees, and nonfinancial support from Eli Lilly; grants and personal fees from Sanofi; grants from Allergan; and grants from Boehringer Ingelheim outside the submitted work. R.A.D. reports other support from AstraZeneca (advisory board, research support, speakers bureau), Novo Nordisk (advisory board), Boehringer Ingelheim (advisory board, research support), Intarcia (advisory board), and Merck (research support) outside the submitted work. J.B.M. reports grants from Medtronic, grants from Novo Nordisk, personal fees from Bayer, personal fees from Boehringer Ingelheim, personal fees from MannKind, personal fees from Novo Nordisk, grants from NIH, and grants from JDRF outside the submitted work, personal fees from Salix Pharmaceuticals, personal fees from Provention Bio, and personal fees from Thermo Fisher Scientific outside the submitted work. R.M.C. reports stock ownership from Bristol-Myers Squibb and stock ownership from Pfizer outside the submitted work. S.E.K. reports grants from Seattle Institute for Biomedical and Clinical Research, during the conduct of the study, personal fees from Bayer, personal fees from Boehringer Ingelheim, personal fees from Casma Therapeutics, personal fees from Eli Lilly, personal fees from Intarcia, personal fees from Merck, personal fees from Novo Nordisk, personal fees from Pfizer, and personal fees from Third Rock Ventures outside the submitted work. W.I.S. reports grant support from Iowa Fraternal Order of the Eagles and grant support from NIH outside the submitted work. K.M.U. reports research support from Avid, personal fees from Nevro, and research support from Lilly outside the submitted work. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. All authors affirmed that authorship is merited based on the International Committee of Medical Journal Editors (ICMJE) authorship criteria. N.R., A.G., R.M.C., K.M.U., and S.E.K. contributed to the conception and design of the study. N.R., A.G., J.A., R.M.C., J.B.M., W.I.S., W.V.T., K.M.U., and S.E.K. contributed to acquisition of data. N.R., N.Y., A.G., and S.E.K. contributed to statistical analysis. All authors contributed to interpretation of data. N.R. and S.E.K. contributed to acquisition of funding. N.R., J.A., and K.M.U. contributed to the supervision and management of the research. N.R., N.Y., A.G., J.A., R.A.D., L.S.K., and S.E.K. drafted the manuscript. All authors contributed to the critical review of the manuscript. N.R. and N.Y. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

K.M.U. and S.E.K. are editors of Diabetes Care but were not involved in any of the decisions regarding review of the manuscript or its acceptance.

Prior Presentation. Parts of this study were presented in abstract form at the 82nd Scientific Sessions of the American Diabetes Association, New Orleans, LA, 3–7 June 2022.