Weight Gain Was Associated With Worsening Glycemia and Cardiovascular and Kidney Outcomes in Patients With Type 2 Diabetes Independent of Diabetes Medication in the GRADE Randomized Controlled Trial Free

Weight management is a key component of effective type 2 diabetes (T2D) management (1,2). Glucose-lowering medications have different effects on weight. Given the demonstrated benefit of glycemic control on microvascular outcomes even in association with weight gain on insulinotropic therapies (3,4), weight gain is generally tolerated in pursuit of glycemic management. Nonetheless, there are concerns that benefits of glycemic management may be undermined by treatment-associated weight gain.

The Glycemia Reduction Approaches in Diabetes: A Comparative Effectiveness Study (GRADE) offers a unique cohort for assessing the effect of commonly used glucose-lowering medications on weight and of treatment-associated weight change on subsequent glycemic, cardiovascular, microvascular, and patient-reported outcomes. GRADE randomly assigned participants with metformin-treated T2D to insulin glargine, the sulfonylurea glimepiride, the glucagon-like peptide 1 receptor agonist liraglutide, or the dipeptidyl peptidase 4 inhibitor sitagliptin. Participants were monitored over a mean of 5 years for the primary glycemic outcome of HbA_1c ≥7%, confirmed, and were continued on randomly assigned therapy through the secondary glycemic outcome of HbA_1c >7.5%, confirmed, at which point glargine was added, followed by aspart, if needed. A high rate of worsening glycemia in all treatment groups has previously been reported, with differences in reaching the secondary glycemic outcome: sitagliptin (55%), glimepiride (50%), liraglutide (46%), and glargine (39%) (P < 0.001) (5). Baseline BMI was not associated with differences in worsening glycemia and did not explain differential effectiveness of medications on glycemic outcomes (5,6). It has previously been reported that over 4 years, average weight loss in GRADE was 3.5 kg in liraglutide, 2 kg in sitagliptin, 0.73 kg in glimepiride, and 0.61 kg in glargine groups. In addition, there were group differences in the rate of ≥10% weight gain, with 13% of glargine, 12% of glimepiride, 9% of sitagliptin, and 6% of the liraglutide groups experiencing this outcome (5). Detailed weight outcomes have not been reported.

In this report, we performed a secondary analysis of the GRADE cohort to gain a greater understanding of weight outcomes associated with the four medication classes added to metformin and to identify baseline factors that predict weight change. In addition, we assessed whether weight gain while on treatment was associated with glycemic, cardiovascular, microvascular, and treatment satisfaction outcomes.

GRADE and its major outcomes have been previously described (5–9). The GRADE Research Group and protocol, including the inclusion and exclusion criteria, are presented in the Supplementary Appendix and are available online at https://clinicaltrials.gov/study/NCT01794143. The trial was approved by the Institutional Review Boards of The George Washington University and all clinical centers.

GRADE was a parallel-group comparative effectiveness randomized clinical trial conducted in 36 clinical centers across the U.S. funded by the National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK). Randomization was performed via a central web-based system and stratified by site. Treatment assignment was unmasked to the participants and clinic staff only.

Participants had T2D diagnosed at age ≥30 years (≥20 in self-reported American Indian/Alaska Native), diabetes duration <10 years, treatment with metformin, and HbA_1c 6.8–8.5% (50.8–69.4 mmol/mol) at randomization. There were no inclusion or exclusion criteria based on weight. Potential participants with history of a major cardiovascular event within the prior year or New York Heart Association Functional Classification II–IV heart failure were excluded. Participants were recruited between July 2013 and August 2017 and followed for a mean of 5.0 years (range 0–7.6) through April 2021 (5,9).

Participants were randomly assigned to start insulin glargine 100 units/mL, glimepiride, liraglutide, or sitagliptin in addition to metformin. All medications were dosed according to their labeling.

Participants had in-person study visits with quarterly measurement of HbA_1c and weight. Weight was measured in kilograms on calibrated electronic or balance scales while participants wore light clothing without shoes. There was 92% overall adherence to study visits (89% during the coronavirus disease 2019 pandemic), with no differences across treatment groups. Missing weight data were not imputed. Waist-to-hip ratio (WHR) was calculated at baseline and at follow-up visits at 2, 4, and 6 years. The Matsuda Index, a measure of whole-body insulin sensitivity, was calculated using plasma glucose and insulin concentrations while fasting and during the baseline 75-g oral glucose tolerance test (10); lower values indicate greater insulin resistance, with values <3 indicating insulin resistance (10,11).

According to the GRADE protocol, participants continued on the randomly assigned medication plus metformin through the primary glycemic outcome of HbA_1c ≥7% until the protocol-defined secondary glycemic outcome of HbA_1c >7.5%, confirmed 3 months later. At that point, the protocol specified the addition of glargine insulin in the non-glargine groups and aspart in the glargine group. For this report, we selected the secondary glycemic outcome (HbA_1c >7.5%, confirmed) as the main outcome of interest to maximize assigned medication exposure time for evaluation of weight outcomes.

We report weight change at 1 year and over the trial. Clinically significant weight change was defined as ≥5% and ≥10% gain or loss. Cardiovascular disease (CVD) was defined as a broad composite outcome of cardiovascular death, myocardial infarction, stroke, unstable angina requiring hospitalization or revascularization, heart failure, coronary artery bypass grafting, interventional cardiology procedure, or other vascular/peripheral vascular interventions adjudicated by a committee masked to treatment assignment (12). Serum creatinine was collected annually, and a spot urinary albumin-to-creatinine ratio (ACR) was collected every 6 months (13). Kidney disease progression was defined as a composite of ACR ≥30 mg/g if the baseline ACR was <30 mg/g, ACR ≥300 mg/g if the baseline ACR was ≥30 but <300 mg/g, or progression to estimated glomerular filtration rate <60 mL/min/1.73 m², dialysis, kidney transplant, or death. New-onset diabetic peripheral neuropathy (DPN) was defined as a score ≥7 on the Michigan Neuropathy Screening Instrument (MNSI) or ≥2.5 on the MNSI clinical examination, performed annually (14). The patient-reported outcome measure of diabetes medication treatment satisfaction, measured by the Diabetes Treatment Satisfaction Questionnaire (DTSQ) (15,16), was collected at baseline and 1 year. The DTSQ is scored from 0 to 6, with higher values indicating greater treatment satisfaction.

Analyses of treatment homogeneity were conducted both by intention-to-treat (ITT) and on randomly assigned therapy (abbreviated “AT” for this report). ITT analyses include participants who had a baseline and at minimum one follow-up weight measurement whether or not they were adherent to medications in GRADE. AT analyses include participants who took at least one dose of metformin and of the second randomized medication (glargine, glimepiride, liraglutide, or sitagliptin), but exclude any information (weight, outcomes, etc.) that occurred after the participant deviated from the assigned treatment, including the addition of glargine or aspart per protocol, any nonprotocol glucose-lowering medication, or the permanent discontinuation of metformin or the randomly assigned second medication.

Treatment homogeneity was tested as follows: for weight change in the first year, we used an F test from a linear model with treatment as the only covariate. For weight change after year 1, we tested the interaction term in a mixed-effects model with weight at each visit as the response, treatment group, time, and a treatment-by-time interaction as fixed effects, and random slopes and intercepts. For time-to-event outcomes (weight gain and loss of ≥5% and ≥10%), we used Wald tests from Cox proportional hazards models with treatment as the only covariate, robust estimates for the SEs, and the Efron method for tie handling. In all models, if homogeneity was rejected, pairwise treatment comparisons were based on contrasts from the model, and type I error was controlled using a Tukey adjustment in the linear model and a Holm adjustment in the remaining models.

We assessed baseline predictors of weight change using Classification and Regression Tree (CART) models, including treatment group, age, sex, self-reported race (3 categories), self-reported ethnicity, BMI, HbA_1c, diabetes duration, and the Matsuda Index as candidate predictors. The final tree was selected by pruning at the minimum error estimate obtained from 10-fold cross-validation.

Finally, in AT analyses only, we evaluated the association, reported as hazard ratio (HR) per kilogram of weight change, between weight change during the first 6 and 12 months of GRADE and subsequent outcomes (HbA_1c >7.5%, CVD, kidney disease progression, new-onset neuropathy, and diabetes treatment satisfaction), including only outcomes occurring after the time point of exposure using proportional hazard models with age, sex, self-reported race and ethnicity, baseline HbA_1c, weight change from baseline to 6 or 12 months, treatment group, and a weight change-by-treatment interaction. We used mixed effects models for the patient-reported outcome of diabetes treatment satisfaction (DTSQ). Tests of homogeneity and pairwise treatment comparisons were conducted as described above.

GRADE is funded by the NIDDK. This manuscript is based on data collected at baseline and follow-up and outcome assessments from the 5,047 participants enrolled into the study who met eligibility criteria for this report. This database is available in the NIDDK Central Repository at https://repository.niddk.nih.gov/study/151.

This report includes 4,980 GRADE participants with available baseline and follow-up weight data (98.6% of the original cohort) (Supplementary Table 1). Mean age was 57 ± 10 years, 36% were women, 44% were non-White, and baseline HbA_1c was 7.5 ± 0.5%. At baseline, mean weight was 100 ± 22.3 kg, WHR was 0.99 ± 0.08, and BMI was 34 ± 6.8 kg/m², with no differences across treatment groups. The mean baseline Matsuda Index was 2.1 ± 1.4. Ninety-five percent of participants remained on assigned treatment (metformin plus the assigned second medication) at the time of the secondary glycemic outcome (HbA_1c >7.5%, confirmed).

The greatest changes in weight occurred during the first treatment year. In ITT analyses from baseline to year 1, there was weight loss (mean [95% CI] kg) in the liraglutide (−3.5 [−3.8,−3.2]) and sitagliptin (−1.07 [−1.35,−0.78]) groups, and weight gain in the glimepiride (0.89 [0.60, 1.18]) and glargine groups (0.45 [0.16, 0.74]; P < 0.0001). Between years 1 and 5, in the ITT analysis, all groups except liraglutide experienced modest weight loss, while the liraglutide group maintained weight (−0.12 [−0.25, 0.01]). The AT analysis displayed similar results between baseline and year 1, with more pronounced weight loss overall from years 1 to 5, especially in the liraglutide and sitagliptin groups (Fig. 1A and B and Supplementary Table 2A and B).

There was substantial variability in weight change, with individuals in the glargine and glimepiride groups being the likeliest to gain ≥10% weight over time, and the liraglutide group being least likely (P < 0.001), as previously reported in ITT analysis (5). Approximately 44% of the glargine and 43% of the glimepiride groups gained ≥5% weight, followed by 31% of the sitagliptin and 21% of the liraglutide group in ITT analyses (P < 0.0001). Compared with ITT, the AT analyses showed lower rates of ≥5% and ≥10% weight gain in all treatment groups (Fig. 2A–D). ITT analyses showed >40% of participants in all treatment groups lost ≥5% weight. Weight loss was highest in the liraglutide group and lowest in the glimepiride and glargine groups. AT analyses showed a similar pattern of weight loss, but with lower rates of ≥5% weight loss (Fig. 2E–H). There were no significant differences between groups in WHR change from baseline to 2 years (data not shown). Fewer than 2% of GRADE participants had bariatric surgery, with no difference by treatment group.

In CART models (Fig. 3A and B), treatment group was the strongest predictor of weight change in both ITT and AT models. Treatment with liraglutide predicted more weight loss than the other three medications, and participants with a baseline BMI ≥39 kg/m² lost more weight on liraglutide than those with baseline BMI <39 kg/m². In the glimepiride and glargine treatment groups, participants with a very low baseline Matsuda Index (<0.69) lost weight compared with those with a slightly higher Matsuda Index (≥0.69) in the ITT analysis. In the AT analysis, BMI replaced the Matsuda Index as the major predictor of weight loss (baseline BMI ≥46 kg/m²) or gain (BMI <46 kg/m²) in the glimepiride- and glargine-treated participants.

Within the first 6 months of GRADE, <0.5% of the cohort met the secondary glycemic outcome of HbA_1c >7.5%, confirmed. Weight gain in the first 6 months of the trial was associated with subsequent increased risk of HbA_1c >7.5% over the course of follow-up (P < 0.0001). There were some differences in the HR of the secondary glycemic outcome across treatment groups: 1.08 (95% CI 1.06, 1.11) per kilogram of weight gain in glimepiride, 1.07 (1.04, 1.1) in liraglutide, 1.03 (1.003, 1.06) in glargine, and no effect in the sitagliptin group (pairwise P was significant for difference between sitagliptin and glimepiride [P < 0.0001] and between sitagliptin and liraglutide [P = 0.012]) (Table 1). Weight change from baseline to year 1 was associated with a HR of 1.05 [1.04, 1.07] per kilogram weight gain for HbA_1c >7.5% at any time thereafter, with no differences across treatment groups.

Nearly 11% of cardiovascular outcomes occurred within the first 6 months. Excluding these outcomes, weight gain between baseline and 6 months was associated with an increased risk of subsequent composite cardiovascular outcomes (HR per kilogram weight gain 1.03 [1.01, 1.06], P = 0.018), with no differences by treatment group (Supplementary Table 3). Kidney disease progression occurred in 16.3% of participants by 12 months. Excluding these outcomes, weight change from baseline to 12 months was associated with subsequent kidney disease progression (HR per kilogram weight gain 1.03 [1.01, 1.06], P = 0.0020), with no difference across treatment groups (Supplementary Table 4). Baseline weight (P < 0.0001), but not weight change, was associated with new-onset neuropathy, with no difference across treatment groups (Supplementary Table 5).

Weight gain from baseline to 6 months was weakly associated with reduced DTSQ score (−0.006 [95% CI −0.012, −0.006], P = 0.022), with similar associations with 1-year weight gain, and no differences across treatment groups, indicating that weight gain was associated with slightly lower levels of diabetes treatment satisfaction (Supplementary Table 6).

Treatment with liraglutide and sitagliptin led to weight loss, while glimepiride and glargine led to slight weight gain during the first year, followed by stable or progressive weight loss in all treatment groups between years 1 and 5 of GRADE trial, providing greater detail than the original GRADE report (5) on patterns of weight change over time in people with T2D initiating a second glucose-lowering medication. After the first year of treatment, while the liraglutide and sitagliptin groups sustained lower weight, the difference in average weight among the four treatment groups diminished over time. CART analysis identified treatment group as the strongest predictor of weight change, with liraglutide treatment associated with greater weight loss than the other three medications. Higher baseline BMI was linked to greater absolute weight loss in participants on liraglutide, glargine, and glimepiride. Among those treated with glimepiride and glargine, extreme insulin resistance (Matsuda Index <0.69), a correlate of high BMI, was associated with more weight loss than less extreme insulin resistance (Matsuda Index ≥0.69).

The analyses also suggest that treatment-associated weight gain is linked to adverse outcomes. Weight gain on the assigned treatment of metformin plus a second glucose-lowering medication was associated with worsening glycemia regardless of medication, with modest differences by treatment group, corresponding to a 5% increased risk of HbA_1c >7.5% per kilogram of weight gain and slightly lower satisfaction with treatment. Five kilograms of weight gain would be predicted to be associated with 25% risk of worsening glycemia to this threshold. In addition, greater degrees of weight gain increased HRs both for subsequent CVD and for kidney disease outcomes by 3% per kilogram gained, or 15% for 5 kg weight gain, with no differences by treatment group.

Following the weight changes observed over the first year of the study, average weight stabilized or decreased in all four treatment groups between years 1 and 5, even in the glargine and glimepiride groups that experienced initial weight increase. This is consistent with the natural history of weight observed both in many long-term T2D trials and in usual care. Participants in the standard diabetes and education arm of Look Action for Health in Diabetes (AHEAD) lost 0.88% body weight at 4 years and 3.5% by trial end (13 years) (17,18). Other long-term diabetes medication trials have also reported stable weight. In the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial, patients in the standard arm had stable weight (0.3 ± 6.3 kg) (19). In the Action in Diabetes and Vascular Disease: Preterax and Diamicron MR Controlled Evaluation (ADVANCE) trial, weight was largely stable over time (standard arm −0.70 [95% CI 0.53, 0.83] kg; intensive arm 0.16 [95% CI −0.02, 0.34] kg) (20). Participants who join trials are likely motivated to improve diabetes management and weight and derive benefit from participation (possibly from reinforcement of diabetes self-management behaviors). However, any focus on diabetes, such as at the time of diagnosis, may promote weight loss: usual care T2D observational studies demonstrate weight loss for up to 5 years after diagnosis (21,22). Additionally, weight loss over time might reflect the recognized effects of aging on weight. The results from this report, which dissect weight change over time rather than reporting an overall average, show that after initial treatment-related weight changes in the first year, weight tended to decrease or stabilize, consistent with results from these other cohorts.

As in other trials, treatment was a strong predictor of weight gain, with more weight gain in the insulin and sulfonylurea groups of GRADE. In the UK Prospective Diabetes Study (UKPDS), participants in the glibenclamide and insulin arms gained 4.8 kg at 3 years compared with 1.7 kg in the dietary therapy arm (in which one-quarter crossed over to medication) (23). The weight gain in the intensive arm of ACCORD (3.0 ± 7.0 kg) was associated with higher baseline HbA_1c and treatment with insulin or thiazolidinediones (19). In the ADVANCE trial, despite stable overall weight, treatment combinations, including insulin and thiazolidinediones, were associated with the greatest weight gain, with sulfonylureas associated with comparatively less weight gain (20). In GRADE, the protocol specified that participants with confirmed HbA_1c >7.5% add basal insulin, and, if needed, prandial aspart. This report shows that there were higher rates of ≥5% and ≥10% weight gain in the ITT versus the AT analyses (which excluded those using rescue insulin), particularly in the liraglutide and sitagliptin groups, suggesting that rescue insulin and discontinuation of liraglutide or sitagliptin at the time of glycemic failure may have contributed to weight gain.

We evaluated the relationship between weight change within the first year of GRADE and subsequent outcomes over the course of the trial only in participants on assigned therapy (i.e., adherent to the randomly assigned second medication and metformin) to restrict the analysis to the impact of weight change on the specified treatment rather than weight change that might have been due to another treatment. In these analyses, adjusted for age, sex, self-reported race and ethnicity, baseline HbA_1c, weight change from baseline to 6 or 12 months, treatment group, and a weight change-by-treatment interaction, higher baseline weight and weight gain within the first year of treatment were associated with worsening glycemia. While the medications had different effects on average weight, the effect of weight gain on outcomes did not differ by treatment group. For example, weight gain on glimepiride and liraglutide was associated with HR per kilogram of 1.07 and 1.06 increased risk of HbA_1c >7.5%, respectively, compared with 1.04 HR per kilogram risk with glargine and sitagliptin. Thus, although people treated with liraglutide were likelier to lose weight, weight gain on liraglutide was associated with a relatively high rate of worsening glycemia, possibly because this unexpected outcome on liraglutide suggests nonadherence or nonresponse to the medication. Similar analyses evaluating the effect of weight gain on treatment with medical outcomes showed that weight gain on treatment was associated with a 1.03 HR per kilogram increased risk of subsequent cardiovascular and kidney outcomes, suggesting weight gain of <1 kg, as observed with glargine and glimepiride in the first year of GRADE, is unlikely to be clinically significant. More substantial weight gain would be expected to yield correspondingly higher risk in this model, as is biologically plausible given the increased risk of vascular wall inflammation and glomerulopathy with increasing degrees of obesity. Finally, while baseline weight was associated with new-onset neuropathy, weight gain was not. This may be related to the high baseline rate of diabetic peripheral neuropathy measured by the MNSI in this cohort (41.9%), with 55% of the cohort progressing over the course of the trial (9), suggesting that overall risk is high and not specifically related to further weight gain.

GRADE has previously reported minimal differences in overall health-related quality of life in association with treatment group and weight (24); however, in this report, we further explored the effect of weight gain on diabetes treatment satisfaction. Treatment satisfaction in GRADE was high, with very little variability. Nonetheless, participants who experienced weight gain did report a small, but statistically significant, lower satisfaction with diabetes treatment. In general, satisfaction with diabetes treatment is related to perceived efficacy of therapy, which in GRADE was centered on glycemic outcomes (25).

While the national multicenter participation, longitudinal experience, high retention rate, and data quality in this report are strengths, the findings should be interpreted within the context of limitations that are inherent to clinical trials. Participants were likely to be more motivated and had closer follow-up than those seen in usual care. There was no specific lifestyle intervention as part of GRADE, and data on diet and exercise were not collected. However, participants received diet and exercise advice at quarterly follow-up visits as part of the diabetes care provided in the study. This advice, plus regular follow-up, may have supported participants in achieving weight outcomes superior to those that would be obtained in usual care. Thus, while the observational analyses of risk of outcomes with weight change are internally valid, the overall estimates of weight change on these medications may not be fully generalizable. Moreover, as in any observational analysis, association does not imply causality. In the ITT analyses, greater weight change may be an indicator of failure to control glucose (which may drive negative outcomes), leading to glargine and aspart rescue and weight gain. Similarly, while lower adherence to liraglutide might lead to higher weight, this may confound a possible association between a reduced putative direct effect of liraglutide on cardiac and kidney outcomes rather than any effect mediated through weight.

Nonetheless, overall, these findings support the conclusion that differences in weight effects among diabetes medications are less pronounced over time than would be suggested by shorter-term trials. Weight changes with diabetes medications mostly occur in the first year of treatment, with a tendency for modest weight loss thereafter. Weight gain on treatment is associated with worsening glycemia, slightly decreased satisfaction with treatment, and cardiovascular outcomes and kidney disease progression. While modest weight gain is associated with very small increased risk, risk accumulates in association with greater degrees of weight gain. The comparative effectiveness aspect of this trial, in which average glycemic management was excellent in all treatment groups, is important to consider. A recently published long-term follow-up of the UKPDS cohort showed mortality, microvascular, and cardiovascular benefits of treatment with insulin and sulfonylurea compared with conventional management (i.e., worse glycemic management), notwithstanding initial weight gain on insulin and sulfonylurea (26). Despite the overall benefit conferred by glycemic management on preventing complications in diabetes, the current findings suggest that substantial weight gain on treatment may be a barrier to effective diabetes care and that glycemic targets should be achieved without substantial weight gain whenever possible.

Funding. The GRADE Study was supported by a grant from the NIDDK of the National Institutes of Health under Award 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, P30 DK020572, P30 DK072476, P30 DK079626, and P30 DK092926 from NIDDK; U54 GM104940 from the National Institute of General Medical Sciences; and UL1 TR000170, UL1 TR000439, UL1 TR000445, UL1 TR001102, UL1 TR001108, UL1 TR001409, 2UL1 TR001425, UL1 TR001449, UL1 TR002243, UL1 TR002345, UL1 TR002378, UL1 TR002489, UL1 TR002529, UL1 TR002535, UL1 TR002537, UL1 TR002541, and UL1 TR002548 from the National Center for Advancing Translational Sciences. Educational materials were provided by the National Diabetes Education Program. Material support in the form of donated medications and supplies were provided by Becton, Dickinson and Company, Bristol-Myers Squibb, Merck, 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. D.J.W. reports participation on data monitoring committees for Novo Nordisk A Research Study to See How Semaglutide Works Compared to Placebo in People With Type 2 Diabetes and Chronic Kidney Disease (FLOW) and Semaglutide Cardiovascular Outcomes Trial (SOUL) trials outside the submitted work. W.T.G. reports receiving consulting fees for membership on advisory boards for Boehringer Ingelheim, Eli Lilly, Novo Nordisk, Pfizer, Fractyl Health, Alnylam Pharmaceuticals, Inogen, Zealand, Allurion, Carmot/Roche, and Merck; payments for participation on data safety monitoring board in phase 3 trials for Boehringer-Ingelheim and Eli Lilly; honoraria for invited lectureships at universities: Brown University, Auburn University, Wayne State, University of Illinois, Mt Sinai, Cleveland Clinic, and University of Miami; membership (unpaid) on Lancet Commission on Obesity; and publicly traded stock with Eli Lilly, Pfizer, Novartis, Merck, Ionis, and Bristol-Myers Squibb (each <$20,000 in value and managed independently by an investment firm), outside the submitted work. A.J.A. reports participation on a scientific advisory board for Medtronic Diabetes insulin pumps outside the submitted work. J.B.F. reports membership (unpaid) for the American Diabetes Association National Board outside the submitted work. S.P.F. reports research contracts from Pfizer paid to the institution for vaccine studies outside the submitted work. S.M. reports speakers bureau payment from AstraZeneca Pharmaceuticals outside the submitted work. All authors affirm that authorship is merited based on the ICMJE authorship criteria. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. D.J.W., W.T.G., A.G., E.J.K., H.K.-S., A.J.A., S.C., T.A.E., S.P.F., J.A.K., S.M., and N.Y. contributed to the interpretation of data and results. D.J.W., W.T.G., A.G., E.J.K., and N.Y. contributed to the drafting of the manuscript. D.J.W., W.T.G., H.K.-S., J.B.-F., S.P.F., S.M., M.T., and N.Y. contributed to the supervision and management of the research. D.J.W., W.T.G., A.J.A., T.A.E., S.P.F., J.A.K., and S.M. contributed to the acquisition of data. D.J.W., A.G., H.K.-S., A.L.C., and N.Y. contributed to the conception and/or design of the research. A.G., E.J.K., and N.Y. contributed to the statistical analysis of data. H.K.-S. contributed to the acquisition of funding. All authors contributed to the critical review and revision of the manuscript. D.J.W. 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.

Handling Editors. The journal editor responsible for overseeing the review of the manuscript was Matthew C. Riddle.