To compare the effects of insulin sensitivity and β-cell function over time on HbA1c and durability of glycemic control in response to dual therapy.
GRADE participants were randomized to glimepiride (n = 1,254), liraglutide (n = 1,262), or sitagliptin (n = 1,268) added to baseline metformin and followed for mean ± SD 5.0 ± 1.3 years, with HbA1c assessed quarterly and oral glucose tolerance tests at baseline, 1, 3, and 5 years. We related time-varying insulin sensitivity (HOMA 2 of insulin sensitivity [HOMA2-%S]) and early (0–30 min) and total (0–120 min) C-peptide (CP) responses to changes in HbA1c and glycemic failure (primary outcome HbA1c ≥7% [53 mmol/mol] and secondary outcome HbA1c >7.5% [58 mmol/mol]) and examined differential treatment responses.
Higher HOMA2-%S was associated with greater initial HbA1c lowering (3 months) but not subsequent HbA1c rise. Greater CP responses were associated with a greater initial treatment response and slower subsequent HbA1c rise. Higher HOMA2-%S and CP responses were each associated with lower risk of primary and secondary outcomes. These associations differed by treatment. In the sitagliptin group, HOMA2-%S and CP responses had greater impact on initial HbA1c reduction (test of heterogeneity, P = 0.009 HOMA2-%S, P = 0.018 early CP, P = 0.001 total CP) and risk of primary outcome (P = 0.005 HOMA2-%S, P = 0.11 early CP, P = 0.025 total CP) but lesser impact on HbA1c rise (P = 0.175 HOMA2-%S, P = 0.006 early CP, P < 0.001 total CP) in comparisons with the glimepiride and liraglutide groups. There were no differential treatment effects on secondary outcome.
Insulin sensitivity and β-cell function affected treatment outcomes irrespective of drug assignment, with greater impact in the sitagliptin group on initial (short-term) HbA1c response in comparison with the glimepiride and liraglutide groups.
Introduction
Type 2 diabetes mellitus (T2DM) is a progressive disorder with worsening glycemic control over time (1,2). To date, metformin is considered the initial medication of choice for glycemic control in T2DM (3). However, in the setting of an inadequate response to metformin, the optimal therapy to be added is uncertain (4). Glycemia Reduction Approaches in Diabetes: A Comparative Effectiveness Study (GRADE) was designed to directly compare four glucose-lowering medications for determination of which would be most effective and have the longest durability of response when added to metformin. GRADE compared the effects of a long-acting insulin (glargine U-100), a sulfonylurea (glimepiride), a glucagon-like peptide 1 (GLP-1) receptor agonist (liraglutide) and a dipeptidyl peptidase 4 (DPP-4) inhibitor (sitagliptin) added to baseline metformin on glycemic control over time. The results confirmed the progressive loss of glycemic control over time with 71% of participants reaching the primary outcome (HbA1c ≥7% [53 mmol/mol]) and 48% reaching the secondary outcome (HbA1c >7.5% [58 mmol/mol]) despite treatment with two glucose-lowering medications over an average of 5 years (5). The incidence of primary and secondary outcomes differed by treatment groups, with the incidence of the primary outcome being the lowest in the insulin glargine and liraglutide groups and highest in the sitagliptin group and that of the secondary outcome lowest in the insulin glargine group and highest in the sitagliptin group (5).
Inadequate insulin secretion in the setting of insulin resistance is a key pathophysiologic factor contributing to the progression of T2DM (6–8). The study design of GRADE allows comparison of differential treatment responses to three commonly prescribed glucose-lowering medications in the context of insulin sensitivity and β-cell responses. In this report, we analyze data from repeated oral glucose tolerance tests (OGTTs) to determine whether measures of insulin sensitivity and β-cell function over time are associated with the risk and rate of glycemic deterioration and whether these associations differ by treatment with liraglutide, glimepiride, or sitagliptin added to metformin. We hypothesized that higher insulin sensitivity and greater β-cell responses would augment the initial glucose-lowering treatment effect and protect against a rise in HbA1c and loss of glycemic control over time. Given the differential treatment effects noted in the primary outcome results of GRADE, we also hypothesized that these relationships would differ by randomized treatment group.
Research Design and Methods
Study Design and Population
GRADE was a randomized, open-label, parallel-group, prospective multicenter clinical trial conducted from 2013 to 2021 in the U.S. comparing the effectiveness of four secondary glucose-lowering medications, when added to baseline metformin, on glycemic control over time. The treatment groups consisted of basal insulin (glargine U100), a sulfonylurea (glimepiride), a GLP-1 receptor agonist (liraglutide), and a DPP-4 inhibitor (sitagliptin). Randomization was conducted with use of a centralized Web-based system, with stratification by trial site. All clinical centers in GRADE obtained local institutional review board approval, and all participants gave written informed consent. The study was overseen by an independent data and safety monitoring board and registered on ClinicalTrials.gov (clinical trial reg. no. NCT01794143).
This analysis focuses on the three noninsulin treatment groups in GRADE. The insulin glargine treatment group was excluded for the following reasons. Exogenous insulin lowers fasting glucose and suppresses endogenous insulin secretion. As participants in the insulin glargine treatment group administered their dose of insulin glargine the night before undergoing the OGTT, based on the long half-life of glargine they would be expected to have lower fasting glucose and C-peptide (CP) concentrations. Additionally, glargine insulin and its metabolites are only partially recognized by the insulin assay used in GRADE (9), making it impossible to measure all active circulating insulin and thus accurately assess insulin sensitivity. Insulin sensitivity is also critical to take into account in assessing CP responses as measures of β-cell function.
Details of the study have previously been published (5,10), and the full protocol can be found on the GRADE website (https://grade.bsc.gwu.edu/). Briefly, eligibility criteria for participants included age ≥30 years (≥20 years if Native American) at time of diagnosis, duration of T2DM <10 years, glucose-lowering treatment with metformin alone, and an HbA1c level at the end of run-in between 6.8% and 8.5% inclusive (51 and 69 mmol/mol) on at least 1,000 mg and up to 2,000 mg/day metformin. Participants at baseline evaluation were on metformin as the only glucose-lowering medication and then randomized to one of the four treatment groups. Glimepiride was started at 1–2 mg daily and titrated up to a maximum dose of 8 mg daily based on HbA1c, fasting glucose, and hypoglycemia. Liraglutide was started at 0.6 mg daily and titrated up to a maximum dose of 1.8 mg daily as tolerated. Sitagliptin was given as 100 mg daily with adjustment for renal impairment.
Participants were assessed at quarterly visits for drug safety, compliance, and concurrent medications, at which time blood samples were collected for HbA1c measurements. At baseline and 1, 3, and 5 years, participants underwent an OGTT for assessment of insulin sensitivity and β-cell function. The primary outcome was defined as time to HbA1c ≥7% (53 mmol/mol) confirmed at the next visit and secondary outcome as time to HbA1c >7.5% (58 mmol/mol) confirmed at the next visit at which time glargine was added. Tertiary outcome was defined as HbA1c >7.5% (58 mmol/mol) despite addition of glargine but is not considered in these analyses as relatively few participants (29%) reached this outcome and because glargine had been added as rescue therapy.
OGTT Procedure
A 75-g OGTT was performed after an overnight fast, with metformin and study drug held on the morning of the OGTT (11). Samples were collected at 0, 30, 60, 90, and 120 min, processed, frozen, and then shipped to the central biochemistry laboratory (University of Minnesota, Minneapolis, MN), where they were assayed for glucose and CP concentrations.
Assays
Whole blood HbA1c and plasma glucose and CP were measured as previously described (11).
Calculations
Insulin Sensitivity
β-Cell Responses
As CP is cosecreted with insulin in a 1:1 molar ratio and is not affected by hepatic clearance, we used it to evaluate β-cell responses. The CP index (CPI), reflecting the early β-cell response to glucose, was calculated as the change in CP divided by the change in glucose from 0 to 30 min. The mean incremental areas under the curve (incAUCs) for glucose and CP were calculated for each OGTT as (AUC/120 min) − fasting, where AUC is area under the curve. The total CP response to the glucose stimulus was calculated as incAUC CP response divided by incAUC glucose (incAUC-CP/incAUCglu). Statistical models involving these responses as a measure of β-cell function were adjusted for insulin sensitivity as a covariate.
Statistical Analysis
Analyses presented here were prespecified as secondary analyses for GRADE, which included a total of 3,784 randomized participants after exclusion of the glargine group (see Consolidated Standards of Reporting Trials [CONSORT] diagram [Supplementary Fig. 1]). For all statistical models the time-varying metabolic measure included as the independent variable was calculated based on the most recent OGTT data. Since OGTTs were performed at baseline and again at 1, 3, and 5 years, analyses up to year 1 were based on baseline measures of insulin sensitivity and β-cell function (HOMA2-%S, CPI, total CP response). Baseline demographic, clinical, and fasting and OGTT-derived metabolic variables are reported by treatment group in Table 1, along with differences among the three treatment groups, tested with an ANOVA global F test (continuous) or a χ2 test (categorical).
Selected OGTT metabolic, demographic, and clinical variables by treatment group at baseline
. | Glimepiride . | Liraglutide . | Sitagliptin . | P (ANOVA or χ2)* . |
---|---|---|---|---|
Baseline (n) | 1,254 | 1,262 | 1,268 | |
Age (years) | 57.1 ± 10.1 | 57.4 ± 9.9 | 57.2 ± 10.1 | 0.704 |
Female sex | 476 (38.0) | 439 (34.8) | 470 (37.1) | 0.234 |
Race | 0.395 | |||
American Indian/Alaska Native | 30 (2.4) | 40 (3.2) | 34 (2.7) | |
Asian/Hawaiian/Pacific Islander | 54 (4.3) | 55 (4.4) | 62 (4.9) | |
Black or African American | 271 (21.6) | 251 (19.9) | 225 (17.7) | |
Other/unknown | 91 (7.3) | 101 (8.0) | 93 (7.3) | |
White | 808 (64.4) | 815 (64.6) | 854 (67.4) | |
Hispanic ethnicity | 234 (18.9) | 234 (18.6) | 241 (19.2) | 0.945 |
Diabetes duration (years) | 4.3 ± 2.8 | 4.2 ± 2.7 | 4.2 ± 2.7 | 0.422 |
BMI (kg/m2) | 34.3 ± 6.9 | 34.3 ± 6.7 | 34.1 ± 6.8 | 0.694 |
HbA1c, % (mmol/mol) | 7.5 ± 0.5 (58.2 ± 5.2) | 7.5 ± 0.5 (58.4 ± 5.3) | 7.5 ± 0.5 (58.4 ± 5.3) | 0.417 |
Fasting glucose (mg/dL) | 151.2 ± 31.2 (n = 1,251) | 150.9 ± 30.6 (n = 1,253) | 151.1 ± 29.6 (n = 1,258) | 0.968 |
Fasting CP (nmol/L) | 1.3 ± 0.6 (n = 1,250) | 1.3 ± 0.5 (n = 1,253) | 1.3 ± 0.6 (n = 1,257) | 0.708 |
HOMA2-%S based on CP | 34.4 ± 15.9 (n = 1,250) | 34.8 ± 16.3 (n = 1,252) | 34.3 ± 15.5 (n = 1,257) | 0.714 |
CPI (nmol/g) | 0.7 ± 0.5 (n = 1,149) | 0.8 ± 0.5 (n = 1,190) | 0.8 ± 0.5 (n = 1,194) | 0.525 |
Total CP response (nmol/mg) | 1.1 ± 0.6 (n = 1,109) | 1.0 ± 0.6 (n = 1,162) | 1.1 ± 0.6 (n = 1,157) | 0.462 |
. | Glimepiride . | Liraglutide . | Sitagliptin . | P (ANOVA or χ2)* . |
---|---|---|---|---|
Baseline (n) | 1,254 | 1,262 | 1,268 | |
Age (years) | 57.1 ± 10.1 | 57.4 ± 9.9 | 57.2 ± 10.1 | 0.704 |
Female sex | 476 (38.0) | 439 (34.8) | 470 (37.1) | 0.234 |
Race | 0.395 | |||
American Indian/Alaska Native | 30 (2.4) | 40 (3.2) | 34 (2.7) | |
Asian/Hawaiian/Pacific Islander | 54 (4.3) | 55 (4.4) | 62 (4.9) | |
Black or African American | 271 (21.6) | 251 (19.9) | 225 (17.7) | |
Other/unknown | 91 (7.3) | 101 (8.0) | 93 (7.3) | |
White | 808 (64.4) | 815 (64.6) | 854 (67.4) | |
Hispanic ethnicity | 234 (18.9) | 234 (18.6) | 241 (19.2) | 0.945 |
Diabetes duration (years) | 4.3 ± 2.8 | 4.2 ± 2.7 | 4.2 ± 2.7 | 0.422 |
BMI (kg/m2) | 34.3 ± 6.9 | 34.3 ± 6.7 | 34.1 ± 6.8 | 0.694 |
HbA1c, % (mmol/mol) | 7.5 ± 0.5 (58.2 ± 5.2) | 7.5 ± 0.5 (58.4 ± 5.3) | 7.5 ± 0.5 (58.4 ± 5.3) | 0.417 |
Fasting glucose (mg/dL) | 151.2 ± 31.2 (n = 1,251) | 150.9 ± 30.6 (n = 1,253) | 151.1 ± 29.6 (n = 1,258) | 0.968 |
Fasting CP (nmol/L) | 1.3 ± 0.6 (n = 1,250) | 1.3 ± 0.5 (n = 1,253) | 1.3 ± 0.6 (n = 1,257) | 0.708 |
HOMA2-%S based on CP | 34.4 ± 15.9 (n = 1,250) | 34.8 ± 16.3 (n = 1,252) | 34.3 ± 15.5 (n = 1,257) | 0.714 |
CPI (nmol/g) | 0.7 ± 0.5 (n = 1,149) | 0.8 ± 0.5 (n = 1,190) | 0.8 ± 0.5 (n = 1,194) | 0.525 |
Total CP response (nmol/mg) | 1.1 ± 0.6 (n = 1,109) | 1.0 ± 0.6 (n = 1,162) | 1.1 ± 0.6 (n = 1,157) | 0.462 |
Data are means ± SD for continuous variables or n (%) for categorical variables unless otherwise indicated. For OGTT measures (fasting glucose, fasting CP, HOMA2-%S, CPI, total CP response), available n for each baseline measure is provided in parentheses.
*ANOVA global F test for differences in means (continuous variables) or χ2 test for differences in proportions (categorical variables) among the four treatment groups.
For assessment of the relationship between each metabolic measure and HbA1c change, a linear repeated-measures regression model was fit for the change in HbA1c from baseline. Follow-up time was rescaled to zero at month 3 such that the mean change in HbA1c from baseline to 3 months became the intercept, and the slope reflected the rate of increase in HbA1c over time beyond 3 months. These models included as predictors the time-varying metabolic measure, treatment group, time from 3 months postbaseline, a three-way interaction term, and all corresponding two-way interactions. All models were adjusted for baseline HbA1c as a covariate, and models for CP responses were adjusted for insulin sensitivity (HOMA2-%S) as a covariate due to the known inverse relationship between CP responses and insulin sensitivity. As other factors may affect the relationship between CP responses and insulin sensitivity, including direct effects of GRADE treatments on the β-cell, analyses were repeated without adjustment for insulin sensitivity. Wald tests were conducted to test whether the association of the metabolic measure with the intercept (two-way interaction term between the metabolic measure and treatment group) or with the slope (three-way interaction term among the metabolic measure, treatment group, and time) differed by treatment group (test of heterogeneity). If the overall test of heterogeneity was significant for the intercept or the slope, then all pairwise comparisons of the association between treatment groups were tested. Otherwise, if the test of heterogeneity was not significant, then a test of the overall association (i.e., with all treatment groups together) was conducted.
For assessment of the association between each metabolic measure and primary metabolic failure, a Cox regression model was fit for the primary outcome, including as predictors the metabolic measure, treatment group, and two-way interaction between the metabolic measure and treatment group. Models for CP responses were run with adjustment for insulin sensitivity as a covariate to provide a measure of β-cell function that accounted for the secretory demand imposed by differences in insulin sensitivity. Models for CP responses were also run unadjusted. For testing of whether the association of the metabolic measure with the primary outcome differed by treatment group, a Wald test was conducted for the two-way interaction term between the metabolic measure and treatment group. Similar to the analysis of HbA1c change, if the overall test of heterogeneity was significant, then all pairwise comparisons of the association between treatment groups were tested. Otherwise, if the test of heterogeneity was not significant, then a test of the overall association was conducted. This analysis was repeated for the secondary outcome.
Subgroup analyses were conducted for assessment of whether the association of metabolic variables with the primary outcome by treatment group varied by the following baseline subgroup factors: age (<45, 45–59, ≥60 years), sex, BMI (<25, 25–<30, ≥30 kg/m2), race (Black, White, all others), Hispanic ethnicity, and diabetes duration (<3, 3–6, >6 years). For these analyses, separate Cox regression models were fit for the primary outcome for each subgroup factor, including as predictors the metabolic measure, treatment group, subgroup factor, the three-way interaction, and all corresponding two-way interactions. Similar to other analyses, models for CP responses were adjusted for insulin sensitivity as a covariate. A Wald test was conducted for the three-way interaction term between the metabolic measure, treatment group, and subgroup factor for assessment of whether the within-treatment associations of the metabolic measure with the primary outcome varied across different levels of the subgroup factor. The subgroup analyses were repeated for the secondary outcome.
Data and Resource Availability
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.
Results
Association of Insulin Sensitivity and CP Responses With Change in HbA1c by Treatment Group
Figure 1 illustrates the mean changes in HbA1c from baseline to 3 months (short-term) and the subsequent rise in HbA1c over time after 3 months (long-term). These changes are presented separately for each metabolic measure and treatment group, with the cohort stratified into tertiles based on these measures.
Short-term (0–3 months) and long-term (slope of rise in HbA1c after 3 months) mean HbA1c change by baseline tertiles of insulin sensitivity (HOMA2-%S [A and D]) and β-cell function as reflected by CP responses (CPI [B and E] and total CP response [C and F]) with models adjusted for HOMA2-%S as a covariate. For each panel, each treatment group is depicted separately with a test for overall treatment heterogeneity noted (i.e., a test of the interaction between treatment group and the continuous insulin sensitivity or CP response measure). For those with overall significance, any significant pairwise comparisons (P < 0.05) of the association of the metabolic measure and short- or long-term HbA1c change between treatment groups are denoted in the triangle with a bold black line between treatment groups: G, glimepiride (orange); L, liraglutide (blue); S, sitagliptin (red). Note, for D, there were no treatment group differences (all P > 0.05) and thus no pairwise comparisons were performed.
Short-term (0–3 months) and long-term (slope of rise in HbA1c after 3 months) mean HbA1c change by baseline tertiles of insulin sensitivity (HOMA2-%S [A and D]) and β-cell function as reflected by CP responses (CPI [B and E] and total CP response [C and F]) with models adjusted for HOMA2-%S as a covariate. For each panel, each treatment group is depicted separately with a test for overall treatment heterogeneity noted (i.e., a test of the interaction between treatment group and the continuous insulin sensitivity or CP response measure). For those with overall significance, any significant pairwise comparisons (P < 0.05) of the association of the metabolic measure and short- or long-term HbA1c change between treatment groups are denoted in the triangle with a bold black line between treatment groups: G, glimepiride (orange); L, liraglutide (blue); S, sitagliptin (red). Note, for D, there were no treatment group differences (all P > 0.05) and thus no pairwise comparisons were performed.
In all treatment groups, HbA1c decreased at 3 months and then started to rise. Notably, the highest tertile of baseline HOMA2-%S, representing higher insulin sensitivity, exhibited the most significant reduction in HbA1c at 3 months, while the lowest tertile had the smallest decrease (Fig. 1A). This relationship showed a significant difference across the treatment groups (as tested for heterogeneity). Specifically, the sitagliptin group had the smallest overall change in HbA1c at 3 months (short-term) but also demonstrated the greatest impact of insulin sensitivity on short-term HbA1c reduction (Fig. 1A and Table 2). However, long-term, the subsequent rise in HbA1c after the initial treatment response did not appear to be related to insulin sensitivity over time, and there was no effect of treatment group on this relationship (Fig. 1D and Table 2).
Relationship of measures of insulin sensitivity and β-cell function with short-term and long-term change in HbA1c (%)
. | Overall† . | Glimepiride* . | Liraglutide* . | Sitagliptin* . | Heterogeneity P* . | Overall P† . | Pairwise P‡ . |
---|---|---|---|---|---|---|---|
Difference in short-term HbA1c change per 1-unit increase in OGTT measure (95% CI) | |||||||
Log HOMA2-%S | −0.178 (−0.266, −0.090) | −0.224 (−0.310, −0.139) | −0.358 (−0.440, −0.275) | 0.009 | N/A | G-L 0.739, G-S 0.010, L-S 0.071 | |
CPI | −0.010 (−0.070, 0.050) | −0.060 (−0.108, −0.011) | −0.123 (−0.175, −0.071) | 0.018 | N/A | G-L 0.421, G-S 0.014, L-S 0.185 | |
Total CP response | −0.080 (−0.134, −0.025) | −0.045 (−0.076, −0.014) | −0.139 (−0.178, −0.099) | 0.001 | N/A | G-L 0.526, G-S 0.192, L-S 0.001 | |
Difference in long-term HbA1c change (slope) per 1-unit increase in OGTT measure (95% CI) | |||||||
Log HOMA2-%S | 0.011 (−0.004, 0.026) | 0.175 | 0.143 | N/A | |||
CPI | −0.061 (−0.082, −0.039) | −0.042 (−0.060, −0.025) | −0.016 (−0.034, 0.002) | 0.006 | N/A | G-L 0.388, G-S 0.005, L-S 0.101 | |
Total CP response | −0.064 (−0.083, −0.046) | −0.045 (−0.057, −0.034) | −0.013 (−0.026, 0.000) | <0.001 | N/A | G-L 0.199, G-S <0.001, L-S 0.001 |
. | Overall† . | Glimepiride* . | Liraglutide* . | Sitagliptin* . | Heterogeneity P* . | Overall P† . | Pairwise P‡ . |
---|---|---|---|---|---|---|---|
Difference in short-term HbA1c change per 1-unit increase in OGTT measure (95% CI) | |||||||
Log HOMA2-%S | −0.178 (−0.266, −0.090) | −0.224 (−0.310, −0.139) | −0.358 (−0.440, −0.275) | 0.009 | N/A | G-L 0.739, G-S 0.010, L-S 0.071 | |
CPI | −0.010 (−0.070, 0.050) | −0.060 (−0.108, −0.011) | −0.123 (−0.175, −0.071) | 0.018 | N/A | G-L 0.421, G-S 0.014, L-S 0.185 | |
Total CP response | −0.080 (−0.134, −0.025) | −0.045 (−0.076, −0.014) | −0.139 (−0.178, −0.099) | 0.001 | N/A | G-L 0.526, G-S 0.192, L-S 0.001 | |
Difference in long-term HbA1c change (slope) per 1-unit increase in OGTT measure (95% CI) | |||||||
Log HOMA2-%S | 0.011 (−0.004, 0.026) | 0.175 | 0.143 | N/A | |||
CPI | −0.061 (−0.082, −0.039) | −0.042 (−0.060, −0.025) | −0.016 (−0.034, 0.002) | 0.006 | N/A | G-L 0.388, G-S 0.005, L-S 0.101 | |
Total CP response | −0.064 (−0.083, −0.046) | −0.045 (−0.057, −0.034) | −0.013 (−0.026, 0.000) | <0.001 | N/A | G-L 0.199, G-S <0.001, L-S 0.001 |
For short-term change in HbA1c, data are presented as mean change in HbA1c from baseline to 3 months, and for long-term change in HbA1c, data are presented as the average rate of change per year after 3 months postrandomization through the end of follow-up. Results in this table are based on linear repeated-measures regression models including all available HbA1c measurements with the time-varying metabolic measure as a predictor in the model. All models include adjustment for baseline HbA1c as a covariate. All models for β-cell function (CPI, total CP response) also include adjustment for insulin sensitivity (HOMA2-%S) as a covariate. G, glimepiride; L, liraglutide; N/A, not applicable; S, sitagliptin.
*The heterogeneity P value is from the test of treatment group differences in the association of insulin sensitivity or β-cell function with HbA1c change. If there are significant treatment group differences, the treatment-specific columns present the difference in short-term or long-term change in HbA1c (95% CI) per unit increase in the insulin sensitivity or β-cell function index within that treatment group.
†If there are not significant treatment group differences, the overall P value is presented from the test of an overall association (i.e., across all treatment groups) of insulin sensitivity or β-cell function with HbA1c change, and differences in HbA1c change per unit increase in the metabolic variable are presented in the “overall” column. If there were significant treatment group differences, the test of an overall association was not conducted and “N/A” is presented for the overall P value.
‡If there are significant treatment group differences, then pairwise comparisons of the treatment groups were tested. “G-L” indicates the P value testing the comparison of glimepiride vs. liraglutide. “G-S” indicates the P value testing the comparison of glimepiride vs. sitagliptin. “L-S” indicates the P value testing the comparison of liraglutide vs. sitagliptin. If the test of heterogeneity across treatment groups is not significant, then tests of pairwise comparisons were not conducted, and “N/A” is presented in the pairwise P values column.
Higher CP responses adjusted for a given level of insulin sensitivity (i.e., from models with adjustment for insulin sensitivity as a covariate), indicating better β-cell function, showed an association with both a larger decrease in HbA1c at 3 months and a slower rise in HbA1c after 3 months (Fig. 1B, C, E, and F and Table 2) that differed significantly across treatment groups. Specifically, there was a greater impact of baseline CP responses on short-term change in HbA1c (Fig. 1B and C and Table 2) in the sitagliptin group as compared with the other two groups. However, adjusted CP responses over time were not associated with long-term change in HbA1c within the sitagliptin group but remained significant for glimepiride and liraglutide (Fig. 1E and F and Table 2). Also of note, analyses without adjustment for HOMA2-%S produced similar findings (data not shown).
Association of Insulin Sensitivity and CP Responses With Risk of Glycemic Failure by Treatment Group
Higher HOMA2-%S (i.e., time-varying) and greater recent CP responses (for a given level of HOMA2-%S) over time were associated with a reduced risk of reaching both the primary and secondary outcomes in all treatment groups, with hazard ratios <1 (Supplementary Table 1).
In examining the association between HOMA2-%S and outcomes, we observed significant treatment group differences for the primary outcome but not for the secondary outcome. Lower HOMA2-%S in the sitagliptin group was linked to a significantly higher risk of reaching the primary outcome in comparisons with the glimepiride and liraglutide groups (Fig. 2 and Supplementary Table 1). There was no significant treatment difference for HOMA2-%S and secondary outcome (Supplementary Fig. 2 and Supplementary Table 1).
The association of baseline tertiles of insulin sensitivity (HOMA2-%S) and CP responses (CPI, total CP response) with time to reach primary outcome is depicted with Kaplan-Meier plots for each treatment group. For those variables with significant treatment group differences in this association, identified with a global Wald test from Cox regression analysis (models for CP indices adjusted for HOMA2-%S as a covariate), significant pairwise comparisons (P < 0.05) within each panel are denoted in the triangle with a bold black line between treatment groups: G, glimepiride (orange); L, liraglutide (blue); S, sitagliptin (red).
The association of baseline tertiles of insulin sensitivity (HOMA2-%S) and CP responses (CPI, total CP response) with time to reach primary outcome is depicted with Kaplan-Meier plots for each treatment group. For those variables with significant treatment group differences in this association, identified with a global Wald test from Cox regression analysis (models for CP indices adjusted for HOMA2-%S as a covariate), significant pairwise comparisons (P < 0.05) within each panel are denoted in the triangle with a bold black line between treatment groups: G, glimepiride (orange); L, liraglutide (blue); S, sitagliptin (red).
The test for treatment group differences for total CP responses and the primary outcome was significant (P = 0.025). Higher total CP responses within the sitagliptin treatment group were associated with a relatively lower risk of the primary outcome when compared with the liraglutide group by pairwise comparison (Fig. 2 and Supplementary Table 1). For the CPI and primary outcome, as well as for CP responses and the secondary outcome, no significant treatment differences were observed (Supplementary Fig. 2 and Supplementary Table 1). Analyses of CP responses without adjustment for HOMA2-%S yielded similar findings (data not shown).
Subgroup analyses revealed that age, sex, duration of diabetes, and race and ethnicity did not have a significant impact on the association between any of the metabolic measures and risk of primary or secondary outcome by treatment group (Supplementary Table 2). While BMI was marginally significant (overall P = 0.028) in its association with the risk of the primary outcome for the total CP response by treatment group (Supplementary Table 2), it was no longer significant after adjustment for multiple comparisons (P = 0.084).
Conclusions
In this study, we sought to understand differential treatment effects of dual therapy on glycemic outcomes in the context of physiologic measures of insulin sensitivity and β-cell function over time. Both insulin sensitivity and β-cell function influenced the initial treatment response and subsequent risk of glycemic failure, with these effects being most pronounced in the sitagliptin treatment group. These associations were particularly noticeable in individuals in the sitagliptin group who were the most insulin resistant. These participants had the smallest change in HbA1c at 3 months and the highest risk of meeting the primary outcome. Similarly, those with the lowest β-cell function in the sitagliptin treatment group had the smallest initial treatment response, but β-cell function over time had less impact on the rate of rise in HbA1c in the sitagliptin treatment group compared with glimepiride and liraglutide. These data highlight the relative importance of insulin sensitivity and β-cell function and treatment interactions in glycemic control and progression of T2DM when patients are on dual therapy, adding to the main outcome findings from GRADE.
GRADE represents the first large-scale randomized clinical trial to explore whether insulin sensitivity and β-cell function over time have varying impacts on glycemic outcomes in response to dual therapy, including treatment with DPP-4 inhibitors and GLP-1 receptor agonists. To provide context, previous major clinical trials like UK Prospective Diabetes Study (UKPDS) and A Diabetes Outcome Progression Trial (ADOPT) compared the effects of monotherapy with insulin, sulfonylureas, metformin, and rosiglitazone. In these trials investigators observed a decline in glycemic control over time despite treatment, and this deterioration was associated with loss of β-cell function (1,7,14).
In ADOPT (7) and the Treatment Options for Type 2 Diabetes in Adolescents and Youth (TODAY) study (15), beneficial effects were observed with rosiglitazone in improving insulin sensitivity and stabilizing β-cell function. However, these studies did not report analyses of differential treatment responses relative to the impact of insulin sensitivity and β-cell function on glycemic outcomes. In a separate analysis of the ADOPT data investigators did find treatment-related differences by sex and BMI. They observed that nonobese men had a greater reduction in HbA1c on sulfonylureas in comparison with thiazolidinediones, while obese women had a greater reduction in HbA1c with thiazolidinediones compared with sulfonylureas (16). This suggests differential effects of insulin sensitivity on glycemic treatment response with these two medications.
GRADE significantly advances our knowledge through investigation of newer diabetes treatments within the context of dual therapy and conducting in-depth analyses focused on the interactions of insulin sensitivity and β-cell function with treatment approaches and how they impact glycemic outcomes.
Regarding treatment heterogeneity, our data highlight the differential treatment responses concerning both insulin sensitivity and β-cell function. This effect was particularly pronounced in assessment of the impact of insulin sensitivity on the initial treatment response and the risk of reaching the primary outcome within the sitagliptin group. The smaller initial treatment response observed in more insulin-resistant participants treated with sitagliptin likely contributed to the differential treatment effect of insulin sensitivity on the primary outcome. Such participants, with a smaller initial decrease in HbA1c, would be less likely to achieve and maintain an HbA1c <7%. This may simply reflect the lower potency of this class of drugs for glucose reduction; in the setting of insulin resistance, a larger proportional increase in insulin release would be required for the same glucose-lowering effect.
Also of interest is the observation of differences between short- and long-term effects on HbA1c. While we did not find an association between insulin sensitivity over time and the subsequent rise in HbA1c in any treatment group, higher insulin sensitivity still remained protective against primary and secondary outcomes for all treatment groups. Our results complement and expand on clinical data from the Predicting Response to Incretin Based Agents (PRIBA) study and the U.K. Clinical Practice Research Datalink (CPRD). These studies demonstrated an association between baseline markers of insulin resistance and reduced glycemic response and lower durability with DPP-4 inhibitors, but no such association was found with GLP-1 receptor agonist therapy (16). Together the findings from GRADE and the studies mentioned above support the hypothesis that there is less glycemic benefit with DPP-4 inhibitors in patients who are more insulin resistant.
Participants in the sitagliptin treatment group exhibited a distinctive impact of β-cell function on glycemic outcomes. Lower CP responses at baseline were associated with a smaller reduction in HbA1c at 3 months in all groups, but this effect was more pronounced with sitagliptin. CP responses over time protected against the progressive rise in HbA1c with glimepiride and liraglutide treatment but not with sitagliptin. One possible explanation for these observations is that the DPP-4 inhibitor (sitagliptin) is less potent in its impact on a dysfunctional β-cell compared with a GLP-1 receptor agonist or sulfonylurea. Our results contrast with those of a previous analysis based on clinical data that did not show an association between baseline markers of β-cell function (measured with HOMA2 of β-cell function) and glycemic response or treatment durability with DPP-4 inhibitors or GLP-1 receptor agonists (16).
A major strength of GRADE is the randomized controlled study design and the inclusion of a large, diverse cohort of adult men and women across a broad age range, making the results more applicable to the general population. GRADE is the first large randomized comparative trial to investigate how insulin sensitivity and β-cell function, as determined by CP responses to oral glucose, influence distinct treatment effects on glycemic outcomes. The results from GRADE may be applicable to a large segment of the population with T2DM. Based on National Health and Nutrition Examination Survey data, a little over 50% of the population with T2DM has historically not met HbA1c treatment targets of <7% (17) and ∼75% of patients with T2DM take metformin (18).
The confirmatory nature of the critical role of insulin sensitivity and β-cell function in responses to therapy could be considered a study limitation. However, the comparative analysis within the context of a large, carefully conducted randomized clinical trial and the differential treatment effects for the DPP-4 inhibitor treatment arm are novel aspects. The exclusion of the insulin glargine arm due to unreliable interpretation of insulin sensitivity and β-cell function unfortunately prevents us from fully leveraging the insights gained from GRADE results. Nonetheless, results from the three noninsulin treatment arms remain valid. Furthermore, these three medication classes comprise a significant portion of secondary treatment options, as initiation of insulin as secondary treatment is typically reserved for individuals with exceptionally high HbA1c levels. Another limitation is that we used a surrogate measure of insulin sensitivity that relies on fasting glucose and CP rather than a gold standard measure such as a hyperinsulinemic-euglycemic clamp. This is simply not practical for a large-scale study such as GRADE. All three of the medications in this analysis stimulate insulin secretion and thus can increase CP concentrations, but the HOMA2 computer model also accounts for the lower glucose levels achieved with treatment. Surrogate measures of insulin sensitivity have been shown to correlate well with measures derived from clamps even in those treated with sulfonylureas (r = 0.73) (19). Of note, glimepiride, liraglutide, and sitagliptin medications all act directly on the β-cell to enhance β-cell responses. Thus, changes in CP responses while patients are on treatment likely reflect both direct medication effects and intrinsic changes in β-cell function that occur over time.
The differential treatment effects observed concerning insulin sensitivity and β-cell function in individuals randomized to sitagliptin may offer some insight. But how this would translate into decision-making in clinical practice would need to be tested in a prospective trial. The hypothesis that insulin sensitivity in particular could guide treatment decisions regarding DPP-4 inhibitors like sitagliptin is intriguing, as this class of glucose-lowering medications is well tolerated and very unlikely to cause hypoglycemia. If our results are confirmed, DPP-4 inhibitors, although less potent than liraglutide or glimepiride, may provide a simple, safe, and effective oral treatment for patients who on clinical grounds appear more insulin sensitive, while those with clinical indicators of insulin resistance may not experience as robust of a response. Of course, detailed testing of insulin sensitivity is not feasible in clinical practice. However, clinicians are quite capable of judging insulin sensitivity based on easily observable parameters such as body mass and fat distribution.
Overall, the results of GRADE underscore the progressive nature of T2DM, even with treatment involving dual therapy. This highlights the importance of ongoing research to develop new treatments that can improve both insulin sensitivity and β-cell function to manage the condition more effectively. These findings have the potential to guide clinical decision-making and aid in tailoring treatments to individual patient needs for better outcomes in T2DM management.
Clinical trial reg. no. NCT01794143, clinicaltrials.gov
This article contains supplementary material online at https://doi.org/10.2337/figshare.24324946.
A complete list of members of the GRADE Research Group can be found in the supplementary material online.
This article is featured in podcasts available at diabetesjournals.org/care/pages/diabetes_care_on_air.
Article Information
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 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 NIH 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. J.B. reports ownership in stocks of Eli Lilly and Merck. R.M.B. has received research support from, has acted as a consultant for, or has been on the scientific advisory board for Abbott Diabetes Care, ARKRAY, Ascensia Diabetes Care Holdings, Bigfoot Biomedical, CeQur, Dexcom, Eli Lilly, embecta, Hygieia, Insulet, MannKind, Medtronic, Novo Nordisk, Onduo, Roche Diabetes Care, Tandem Diabetes Care, Sanofi, United Healthcare, Vertex Pharmaceuticals, and Zealand Pharma. R.A.D. reports grants/contracts from Boehringer Ingelheim, AstraZeneca, and Merck; payments or honoraria from AstraZeneca; and participation in a data safety monitoring or advisory board for AstraZeneca, Novo Nordisk, and Boehringer Ingelheim. C.D. reports consulting fees from Novo Nordisk and board membership for Nebraska Educational Biomedical Research Association (an Omaha VA Medical Center nonprofit). 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. 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 W.I.S. reports grant support from Iowa Fraternal Order of the Eagles and grant support from the National Institutes of Health outside the submitted work. K.M.U. reports personal fees from Nevro and research support from Lilly, and research support from Avid 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. K.M.U., J.B., C.D., T.E., S.E.K., N.R., W.M.V., and W.I.S. contributed to the conception and design of the research. K.M.U., N.Y., R.M.B., J.B., C.D., J.K., N.R., W.M.V., and W.I.S. contributed to the acquisition of data. N.Y. and N.M.B. contributed to the statistical analysis. K.M.U., N.M.B., A.B., R.M.B., J.B., C.D., R.A.D., J.K., S.E.K., N.R., W.M.V., and W.I.S. contributed to the interpretation of data and results. S.E.K. and W.I.S. contributed to the acquisition of funding. K.M.U., C.D., S.E.K., N.R., and W.I.S. contributed to the supervision and management of research. K.M.U., N.Y., N.M.B., J.B., S.E.K., and W.I.S. contributed to the drafting of the manuscript. All authors critically reviewed the manuscript. K.M.U. 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.