Use of Twice-Daily Exenatide in Basal Insulin–Treated Patients With Type 2 Diabetes: A Randomized, Controlled Trial FREE

—The Editors

Glycemic control in type 2 diabetes mellitus becomes increasingly challenging with longer duration of disease. Progressive decline in β-cell function and insulin resistance, combined with increased hepatic glucose output due to glucagon dysregulation, lead to elevations in both fasting and prandial glucose levels. Pharmacologic treatment of diabetes usually involves the sequential addition of oral antihyperglycemic agents according to a target-driven strategy, usually followed by the addition of basal insulin and then prandial insulin (1). This progression, despite increasing numbers and doses of therapeutic agents, is generally associated with persistently elevated hemoglobin A_1c (HbA_1c) levels and decreased likelihood of achieving glycemic targets with longer duration of diabetes (1 - 5).

Several studies have demonstrated that basal insulin replacement can attain HbA_1c targets in 50% to 60% of patients when the insulin dose is systematically adjusted by following structured titration regimens (6 - 12). Prandial replacement is often required in patients with HbA_1c levels greater than 7.0% and long-standing type 2 diabetes (13 - 14). Glucagon-like peptide-1 receptor agonists are associated with improved glycemic control without increased risk for hypoglycemia, and weight loss also usually occurs (15 - 16). The rationale for combining twice-daily exenatide with basal insulin is thus based on complementary pharmacologic effects on prandial and fasting glycemia (17).

The use of exenatide and insulin in combination is not an approved regimen. Published supporting evidence of combining exenatide and insulin is limited to a short-term, small-scale, randomized, controlled clinical trial (18); a small, short-term, randomized trial comparing exenatide with sitagliptin added to optimized insulin glargine (19); 2 uncontrolled, nonrandomized, retrospective reports derived from clinical practice (20 - 21); a trial in which reduction in insulin dose and weight were the primary outcomes (22); and a recent summary report of small, observational studies and trials, which showed reductions in HbA_1c, body weight, and insulin dose (23).

We tested whether twice-daily exenatide injections result in greater reductions in HbA_1c level than placebo at 30 weeks in persons receiving insulin glargine.

Our 30-week, randomized, double-masked, parallel, placebo-controlled study was conducted in 59 centers in 5 countries (Greece, Israel, Mexico, United Kingdom, and United States) from 29 October 2008 to 4 January 2010. Participants, investigators, and other personnel involved in the conduct of the study were blinded to individual treatment assignments for the duration of the study. The primary objective was to determine whether exenatide injection, 10 µg twice daily, was superior to placebo, as measured by change in HbA_1c level, in participants with type 2 diabetes who were receiving insulin glargine with or without metformin or pioglitazone (or both agents). Secondary outcome measures included the percentage of participants with HbA_1c levels of 7.0% or less and 6.5% or less; 7-point self-monitored blood glucose (SMBG) profiles; change in body weight, waist circumference, and insulin dose from baseline; and safety (measured by self-reported hypoglycemic events and treatment-emergent adverse events). Exploratory measures included 1,5-anhydroglucitol level at baseline, week 18, and week 30.

The study was approved at each site by an institutional review board in accordance with principles described in the Declaration of Helsinki (24). All participants gave written informed consent before participating. No data safety monitoring board was involved, and no interim analyses were performed.

Participants were at least 18 years of age; had type 2 diabetes; had been receiving insulin glargine at a minimum of 20 U/d without any other insulin, alone or in combination with a stable dose of metformin or pioglitazone (or both agents) for at least 3 months; and had an HbA_1c level of 7.1% to 10.5%, body mass index of 45 kg/m² or less, and stable body weight (less than 5% change over 3 months). Participants were excluded if they had clinically significant hematologic, oncologic, renal, cardiac, hepatic, or gastrointestinal disease; had been in a weight-loss program in the 3 months before the study; received systemic glucocorticoid therapy in the 8 weeks before the study; had more than 1 episode of major hypoglycemia in the 6 months before the study; had an irregular sleep–wake cycle; or had a history of pancreatitis. Duration of diabetes was by self-report.

Participants continued their prestudy doses of oral antihyperglycemic agents. A computer-generated, random-sequence interactive voice-response system was used to randomly assign participants in blocks of 4, stratified by HbA_1c level (≤8.0% or >8.0%, as measured by a central laboratory; normal range, 4.3% to 6.1%), to receive exenatide (5 µg twice daily for 4 weeks and 10 µg twice daily thereafter) or placebo injections within 60 minutes before morning and evening meals. Placebo was indistinguishable from exenatide. At randomization, participants with HbA_1c levels greater than 8.0% continued to receive their current insulin glargine dose; those with HbA_1c levels of 8.0% or less decreased their dose by 20%. These doses were maintained for 5 weeks, after which participants began titration to achieve a fasting glucose level less than 5.6 mmol/L (<100 mg/dL), on the basis of the Treat-to-Target Trial algorithm (9). Investigators assessed participants' adherence to study medication, study diaries, and glycemic control at each visit. Participants recorded SMBG, and adjustments to insulin dose were made by the investigator on the basis of the algorithm at least weekly from week 5 to week 10, and every 2 weeks thereafter.

Sites were responsible for assessing medication distribution and use. Adherence to the insulin glargine algorithm was reviewed by study monitors at scheduled visits.

The primary outcome of the study was HbA_1c level at 30 weeks. Secondary and exploratory outcomes included fasting plasma glucose, lipid, and 1,5-anhydroglucitol levels measured at baseline and at 18 and 30 weeks; SMBG profiles were obtained during the week before these visits. Other secondary outcomes—vital signs, weight, waist circumference, and insulin doses—were obtained at baseline and each postrandomization visit.

Safety was monitored by site staff at telephone and office visits by diary review and participant interview. Hypoglycemic episodes were classified as minor (signs or symptoms associated with hypoglycemia and fingerstick blood glucose level <3 mmol/L [<54 mg/dL] that were either self-treated or resolved on their own) or major (blood glucose level <3 mmol/L [<54 mg/dL], resulting in loss of consciousness or seizure from which the participant promptly recovered in response to glucagon or glucose, or presumed hypoglycemia requiring the assistance of another person because of severe impairment of consciousness or behavior). All laboratory determinations were performed by a centralized laboratory (Quintiles Laboratories, Marietta, Georgia).

Additional exploratory outcomes that we assessed were optional 72-hour continuous glucose monitoring at study beginning and end, and changes in levels of adiponectin and high-sensitivity C-reactive protein. These outcomes are not reported in this article.

All statistical analyses were performed by using SAS software, version 9.1 (SAS Institute, Cary, North Carolina). Methods were documented in a prespecified analysis plan before unblinding. Post hoc analyses were conducted to support the prespecified analyses where appropriate. Analyses used 2-sided CIs where appropriate, and an α level of 0.05 was considered significant.

The analysis of efficacy and safety variables included data from all participants who received the study drug and had measurements at postbaseline visits. “Baseline” was defined as the last nonmissing value before randomization. The analysis of continuous variables used a mixed-model repeated-measures analysis with effects for treatment, visit, treatment-by-visit interaction, baseline HbA_1c stratum (≤8.0% or >8.0%) (except for HbA_1c analyses), baseline of the variable analyzed, and pooled investigative site, with an unstructured variance–covariance matrix to model the covariance structure among the repeated measurements by participant (25). All sites that enrolled 1 participant were pooled into a single group. Changes from baseline are summarized by least-squares means and 95% CIs from the mixed model.

The hypoglycemia rate was analyzed by using a generalized linear negative binomial model with effects for HbA_1c stratum and treatment, log of the number of days of follow-up as an offset, and a log link. The proportion of participants achieving HbA_1c targets was analyzed by using multiple imputation assuming missing at random, followed by a generalized estimating equation analysis (26). The variables included in the multiple imputation model were baseline body weight, baseline HbA_1c level, sex, and pre–30-week HbA_1c level, and data were imputed by using multiple regression methods under a monotone missing data pattern. Imputed data were analyzed by using a generalized estimating equation model with factors for treatment, visit, treatment-by-visit interaction, and baseline HbA_1c level.

With 260 randomized participants and a 20% dropout rate, 104 participants per treatment would provide 90% power, assuming a 0.5% difference in change from baseline HbA_1c value and an SD of 1.1%.

The Alliance of Eli Lilly and Company and Amylin Pharmaceuticals was the funding source and sponsor for the trial. The sponsor was involved in study design, data collection, data analysis, and manuscript preparation. All authors had full access to the data, participated in data analysis and manuscript development, and gave final approval of the manuscript. Dr. Buse was involved in the study design and made final decisions on manuscript content.

A total of 425 patients were screened, and 261 participants were randomly assigned to exenatide or placebo (1 person in each group withdrew before receiving the study drug and were excluded from analysis) (Figure 1). Of the 138 participants assigned to twice-daily exenatide treatment, 26 (19%) withdrew, mainly because of adverse events; 22 of 123 (18%) placebo recipients withdrew, mainly owing to participant decision. Baseline characteristics were similar between treatment groups, with the exception of sex and prestudy oral antihyperglycemic agent used (Table 1). Of the 55 active sites, 35 enrolled 4 or fewer participants, resulting in incomplete blocks, an imbalance in the number of participants assigned to each treatment, and possibly the difference in baseline HbA_1c level. Participants had a mean duration of diabetes of 12 years (SD, 7), and 14% had had diabetes for 20 or more years.

Reduction of HbA_1c level from baseline at 30 weeks with exenatide twice daily plus optimized insulin glargine was greater than with placebo plus optimized insulin glargine (change, −1.74% [95% CI, −1.91% to −1.56%] vs. −1.04% [CI, −1.22% to −0.86%]; between-group difference, −0.69% [CI, −0.93% to −0.46%]; P < 0.001) (Figure 2 and Appendix Figure). The difference in HbA_1c level between groups was not dependent on race, oral antihyperglycemic agent therapy, sex, or age. In both groups, participants with higher baseline HbA_1c values had greater HbA_1c reductions. At 30 weeks, the proportion of participants achieving the target HbA_1c value of 7.0% or less was 60% (CI, 51% to 69%) in the exenatide group and 35% (CI, 25% to 45%) in the placebo group (between-group difference, 25 percentage points [CI, 12 to 39 percentage points]); the proportion achieving the target HbA_1c value of 6.5% or less was 40% (CI, 30% to 49%) and 12% (CI, 6% to 17%), respectively (between-group difference, 28 percentage points [CI, 17 to 39 percentage points]) (Table 2).

The insulin dosage increased from baseline in both groups, but the increase was greater in the placebo group (20 U/d [CI, 16 to 24 U/d] vs. 13 U/d [CI, 9 to 17 U/d] in the exenatide group; between-group difference, −6.5 U/d [CI, −12.3 to −0.8 U/d]) (Table 2). Fasting plasma glucose level decreased similarly with twice-daily exenatide and placebo (Table 2).

At 30 weeks, SMBG was lower at all nonfasting time points in the exenatide group than in the placebo group (P < 0.001) (Figure 3, bottom), as were morning and evening postprandial glucose excursions (Table 2). Concentrations of 1,5-anhydroglucitol, a marker that is inversely proportional to average glycemia (27 - 30), were higher in the exenatide group than the placebo group (12.7 µg/mL [CI, 11.6 to 13.7 µg/mL] vs. 10.6 µg/mL [CI, 9.5 to 11.7 µg/mL]; between-group difference, 2.1 µg/mL [CI, 0.7 to 3.5 µg/mL]; P < 0.010).

Body weight decreased with exenatide (from 95.4 kg to 93.6 kg; change, −1.8 kg [CI, −2.5 to −1.1 kg]) but increased with placebo (from 93.8 to 96.3 kg; change, 1.0 kg [CI, 0.2 to 1.7 kg]) (between-group difference, −2.7 kg [CI, −3.7 to −1.7]; P < 0.001) (Figure 3, top). Waist circumference did not significantly differ between the groups (P = 0.23). There was no apparent relation between weight change and gastrointestinal adverse events in exenatide or placebo recipients (Appendix Table 1), but this was not formally tested because adverse events and weight change are postbaseline events and the data are further confounded by differences in baseline weight between participants with and without these events.

At 30 weeks, concentrations of triglycerides, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, and non–high-density lipoprotein cholesterol did not differ between the groups. Systolic and diastolic blood pressures decreased (P < 0.010 and P < 0.001, respectively) from baseline with exenatide (Table 2). Heart rate increased from baseline in the exenatide group (between-group difference, 3.0 beats/min [CI, 0.8 to 5.2 beats/min]) (Table 2).

The number of hypoglycemic events per participant per year did not differ significantly between the exenatide and placebo groups (1.4 [CI, 0.9 to 2.2] vs. 1.2 [CI, 0.8 to 2.0], respectively; estimate treatment ratio, 1.2 [CI, 0.6 to 2.2]; P = 0.49). Although different durations of follow-up due to early discontinuation confound interpretation of these values, the proportion of participants who had minor hypoglycemia seemed to be similar with exenatide and placebo (25% [34 of 137 participants] vs. 29% [35 of 122 participants], respectively) (Table 3). Only 1 participant (in the placebo group) experienced major hypoglycemia (2 nocturnal episodes, each of which required glucagon injection and 1 of which required hospitalization). Between-group differences in the following adverse events were greater in exenatide than in placebo recipients: nausea (32 percentage points [CI, 23 to 42 percentage points]), diarrhea (10 percentage points [CI, 2 to 18 percentage points]), vomiting (14 percentage points [CI, 7 to 21 percentage points]), headache (10 percentage points [CI, 3 to 16 percentage points]), and constipation (8 percentage points [CI, 3 to 14 percentage points]).

Table 3 shows all adverse events that occurred in at least 5% of exenatide recipients. More exenatide recipients than placebo recipients discontinued the study because of adverse events (13 [9%] vs. 1 [1%], respectively; P < 0.010). Appendix Table 2 shows discontinuation at each visit and cumulative values. Serious adverse events were evenly distributed between groups, and none was experienced by more than 2 participants in either group. One death (due to myocardial infarction) occurred in the placebo group. No treatment-emergent pancreatitis, acute renal failure, or cancer occurred.

The primary analysis of HbA_1c used a mixed model that is valid as long as missing outcomes are missing at random (25). To address the possible effect of informative dropout on our results, we conducted a sensitivity analysis in which we assumed that participants who dropped out from the exenatide and placebo groups had the same effect as placebo (that is, we used data only from placebo participants to impute 30-week results for both exenatide and placebo recipients even if pre–week-30 data were available for exenatide recipients) (Appendix Table 3). Under this conservative assumption, the change from baseline in HbA_1c values was still significantly greater in the exenatide group (−0.53% [CI, −0.76% to −0.31%; P < 0.001).

In our study, participants had diabetes for a mean of more than 10 years, and disease was suboptimally controlled with basal insulin glargine therapy. We hypothesized that the combination of twice-daily exenatide, an effective agent for managing postprandial glycemic excursions, with structured titration of basal insulin glargine therapy to achieve target fasting glucose levels would result in overall glucose control (as assessed by HbA_1c) and would be superior to basal insulin titration alone. Treatment with twice-daily exenatide plus optimized insulin glargine was associated with a 1.74% reduction in HbA_1c level, and more than one half of participants achieved an HbA_1c value of 7.0% or less; in contrast, treatment with optimized insulin glargine alone was associated with a clinically meaningful and statistically significant 1.04% reduction in HbA_1c, and only one third of participants achieved values of 7.0% or less. This greater glycemic control with exenatide was achieved with no difference in the proportion of participants reporting hypoglycemia and with modest weight loss, as opposed to weight gain seen with optimized insulin therapy alone.

Before this study, information on the use of insulin with exenatide was limited to data from small clinical trials and observational analyses (18 - 23). In a proof-of-concept study, Kolterman and colleagues (18) reported the effects of exenatide twice daily on prandial glycemic excursion in 24 participants, only 6 of whom were receiving insulin. Their data showed a reduction in prandial glycemic excursion.

More recently, Arnolds and associates' study (19) examined postprandial glucose control in which exenatide twice daily or sitagliptin was added to a regimen of insulin glargine and metformin that achieved fasting glucose levels less than 5.6 mmol/L (<100 mg/dL). Addition of either exenatide twice daily or sitagliptin was associated with a decrease in the 6-hour postprandial glucose level compared with treatment with glargine plus metformin alone. Reductions in mean HbA_1c level after 4 weeks of treatment in the exenatide, insulin glargine, and metformin group (−1.80%) and the insulin glargine plus metformin group (−1.23%) were similar to those in our study and were statistically different from one another (P < 0.050).

Yoon and colleagues (20) reviewed the records of 268 patients who had been treated with exenatide plus insulin and reported on 188 participants with sufficient data for analyses. The mean change in HbA_1c level over 6 to 24 months (time from addition of exenatide to insulin) ranged from −0.55% to −0.64%, similar to our study. Mean weight changes over these same periods ranged from −2.4 kg to −6.2 kg. Insulin doses decreased in these participants, but in contrast to our study, many participants in Yoon and colleagues' study were receiving basal plus bolus insulin regimens. The reduction in insulin dose may explain the greater weight reduction seen in their observational data compared with our study, in which insulin up-titration was part of the protocol.

In a similar retrospective analysis of 124 participants, Sheffield and coworkers (21) reported a reduction in HbA_1c level of 0.87% at 1 year, a weight reduction of 5.2 kg, and a reduction in insulin dose. Finally, a study of 160 participants receiving exenatide (to facilitate weight loss and reduction in insulin) reported no changes in HbA_1c values but reductions in weight (>10 kg) and insulin doses did occur (22). A recent summary report of small observational studies and trials consistently shows reductions in HbA_1c levels, weight, and insulin dose (23). Compared with our study, no previous study showed the same efforts to achieve specific fasting glucose targets through insulin up-titration, control for background therapy, or carefully assess hypoglycemia. Thus, our study adds to the limited body of evidence on use of twice-daily exenatide with insulin and specifically examines the potential of this combination to help patients reach defined glycemic targets.

With increasing duration of type 2 diabetes, glycemic control becomes progressively difficult to achieve. Titration of basal insulin to attain target fasting glucose levels is recommended as a relatively easy, effective technique to improve glycemic control in patients who do not achieve glycemic goals with oral antihyperglycemic agents, although this strategy is associated with weight gain and hypoglycemia (1). However, almost 50% of patients still do not achieve target values, and even fewer achieve target values if proper insulin titration to optimize basal insulin is not maintained (6 - 12,14,31 - 32). Average glycemia, as reflected by HbA_1c level in patients with insulin-treated diabetes, is substantially higher than in those treated with lifestyle therapy or oral antihyperglycemic agents, mainly reflecting differences in patient populations and stage of disease progression (1,3 - 5). The failure of insulin therapy to adequately control diabetes is related in part to inadequate titration of insulin in clinical settings, perhaps because of concern about weight gain and hypoglycemia.

Optimal glycemic control requires managing both fasting and prandial glucose values. We used a simple algorithm for titrating basal insulin glargine (9) and achieved mean fasting plasma glucose values similar to other studies by using this approach (6,8,10 - 11,31,33). The addition of twice-daily exenatide to therapy with insulin glargine produced a further 0.69% reduction of HbA_1c level despite lower insulin doses. The fasting plasma glucose values achieved in both study groups are probably the result of titration of insulin glargine. The additional postprandial effects of twice-daily exenatide are demonstrated in the SMBG patterns, in which all mean values were well within current treatment targets, and are confirmed by the differences between the exenatide and placebo groups in 1,5-anhydroglucitol level at 30 weeks. 1,5-Anhydroglucitol is a marker of average glycemia; is inversely proportional to HbA_1c; and is more sensitive to postprandial glycemic excursions, particularly at HbA_1c values less than 8.0% (27 - 30).

The SMBG results also demonstrate that glucose levels obtained at all time points (except the fasting value) were both statistically and clinically lower in the exenatide group than the placebo group (generally by >1 mmol/L [>18 mg/dL]). Furthermore, average postprandial glucose levels with twice-daily exenatide plus insulin glargine were well within current treatment targets, particularly after the morning and evening meals, when exenatide was administered.

Achieving an HbA_1c level of 7.0% or less in a majority of patients with poor glycemic control despite insulin therapy has been infrequently demonstrated. To our knowledge, insulin therapy has never been reported in this setting without weight gain (34) or increased risk for hypoglycemia. We reduced the dose of basal insulin glargine in participants with baseline HbA_1c levels of 8.0% or less during the first 5 weeks while twice-daily exenatide (or placebo) was being added to the therapeutic regimen; the relative safety of this approach is a validation of translational utility in clinical practice. The association of twice-daily exenatide therapy with relatively frequent, but usually not dose-limiting, gastrointestinal adverse events has been reported (35 - 40). The types or number of serious treatment-emergent adverse events did not differ between the exenatide and placebo groups, and no cases of renal failure, pancreatitis, or cancer were reported.

Our study has limitations. First, a small randomization imbalance resulted from block randomization by site, which caused a small difference in the distribution of baseline HbA_1c values and sex and an imbalance in concomitant glucose-lowering medications (especially thiazolidinediones). However, after we adjusted for these variables, none materially affected the primary outcomes.

Second, although our fasting glucose values would be acceptable for insulin titration in clinical practice and are similar to those in comparable studies (9 - 11,32 - 33), it is possible that lower fasting glucose values (and correspondingly lower HbA_1c values) can be achieved with even more aggressive insulin titration or in patients with diabetes who have better islet cell function earlier in their disease to further potentiate prandial glycemic control.

Third, the relative safety and efficacy of our approach to dose adjustment of basal insulin when exenatide twice daily was added may not apply to patients with baseline HbA_1c levels less than 7.0%, those with a recent history of major hypoglycemia, or those receiving long-acting glucagon-like peptide-1 receptor agonists.

Fourth, our study was only 6 months in duration. The durability and safety of such an approach with long-term therapy is not assured.

Fifth, some participants who dropped out in the first 18 weeks of the study did not have HbA_1c measurements done after randomization, and the methods for dealing with the sporadic HbA_1c values available before 18 weeks have limitations. However, imputations of missing data under conservative assumptions still confirm a clinically meaningful difference in change in HbA_1c greater than 0.53% favoring exenatide.

Finally, this trial compared only exenatide to placebo; no comparisons of other glucose-lowering agents that affect prandial glycemic excursion can be made. Furthermore, specific issues related to injection versus oral therapy or cost considerations cannot be addressed on the basis of the trial design or our results. We found a statistically significant increase in heart rate; however, a study of exenatide that included 24-hour blood pressure monitoring found a nonsignificant increase in heart rate in participants who received exenatide twice daily compared with those who received placebo (41). The clinical significance of the increased heart rate in exenatide recipients remains uncertain.

In conclusion, in a sample of patients with HbA_1c values greater than 7.0% despite treatment with insulin glargine, the addition of twice-daily exenatide therapy and consistent titration of basal insulin glargine provided improved control to a greater extent than did titration of insulin glargine alone. The improvement in HbA_1c level of −0.69% with the addition of twice-daily exenatide to optimized glargine was achieved with no increased risk for hypoglycemia and with modest weight loss. Nausea, diarrhea, vomiting, headache, and constipation were increased with exenatide compared with placebo. Long-term studies using regimens associated with weight loss, low risk for hypoglycemia, and improved glycemic control will be required to further examine the potential clinical benefits of the treatment strategies used in this study.