OBJECTIVE

Mealtime insulin is commonly added to manage hyperglycemia in type 2 diabetes when basal insulin is insufficient. However, this complex regimen is associated with weight gain and hypoglycemia. This study compared the efficacy and safety of exenatide twice daily or mealtime insulin lispro in patients inadequately controlled by insulin glargine and metformin despite up-titration.

RESEARCH DESIGN AND METHODS

In this 30-week, open-label, multicenter, randomized, noninferiority trial with 12 weeks prior insulin optimization, 627 patients with insufficient postoptimization glycated hemoglobin A1c (HbA1c) were randomized to exenatide (10–20 µg/day) or thrice-daily mealtime lispro titrated to premeal glucose of 5.6–6.0 mmol/L, both added to insulin glargine (mean 61 units/day at randomization) and metformin (mean 2,000 mg/day).

RESULTS

Randomization HbA1c and fasting glucose (FG) were 8.3% (67 mmol/mol) and 7.1 mmol/L for exenatide and 8.2% (66 mmol/mol) and 7.1 mmol/L for lispro. At 30 weeks postrandomization, mean HbA1c changes were noninferior for exenatide compared with lispro (–1.13 and –1.10%, respectively); treatment differences were –0.04 (95% CI –0.18, 0.11) in per-protocol (n = 510) and –0.03 (95% CI –0.16, 0.11) in intent-to-treat (n = 627) populations. FG was lower with exenatide than lispro (6.5 vs. 7.2 mmol/L; P = 0.002). Weight decreased with exenatide and increased with lispro (−2.5 vs. +2.1 kg; P < 0.001). More patients reported treatment satisfaction and better quality of life with exenatide than lispro, although a larger proportion of patients with exenatide experienced treatment-emergent adverse events. Exenatide resulted in fewer nonnocturnal hypoglycemic episodes but more gastrointestinal adverse events than lispro.

CONCLUSIONS

Adding exenatide to titrated glargine with metformin resulted in similar glycemic control as adding lispro and was well tolerated. These findings support exenatide as a noninsulin addition for patients failing basal insulin.

Type 2 diabetes is a progressive disease requiring treatment intensification to maintain glycemic control (1,2). For patients unable to achieve individualized glycemic targets with oral glucose-lowering agents, basal insulin is usually initiated and up-titrated to a fasting glucose (FG) target, but postprandial glucose may remain elevated. Although mealtime insulin may control postprandial glucose (1), weight gain, increased hypoglycemia risk, and poor patient acceptance may result (1). Recent studies suggest that adding a short-acting glucagon-like peptide 1 (GLP-1) receptor agonist to target postprandial glucose may be a viable option, but, to date, no studies directly compared the efficacy and safety of a GLP-1 receptor agonist with mealtime insulin in patients receiving optimized basal insulin (3,4). Placebo-controlled studies have been criticized for providing insufficient data for translation into clinical practice (5).

We prospectively compared glycated hemoglobin A1c (HbA1c) and other outcomes over 30 weeks between the short-acting GLP-1 receptor agonist exenatide twice daily (exenatide) and titrated mealtime insulin lispro (lispro), both added to titrated insulin glargine (glargine) in metformin-treated patients with insufficient glycemic control despite 12 weeks of intensive protocol-based glargine titration.

Study Design and Patients

This open-label, randomized, controlled, noninferiority study (NCT00960661) was conducted at 108 centers in 17 countries between September 2009 and August 2012. The 44-week study included a 2-week screening period, 12-week basal insulin optimization (BIO) phase, and 30-week intervention phase. Enrolled patients were males and females 18 years and older with type 2 diabetes treated with insulin glargine and metformin ± sulfonylurea with HbA1c of 7.0% (53 mmol/mol) to 10.0% (86 mmol/mol) and BMI of 25.0 kg/m2 (23.0 for South Korean participants) to 45.0 kg/m2. Exclusion criteria included use of other glucose-lowering agents and/or clinical history, condition, or concomitant medication that could confound efficacy or safety (Supplementary Table 1). At entry, patients continued metformin and discontinued sulfonylurea. Bedtime glargine was titrated to FG of 5.6 mmol/L or lower without hypoglycemia (i.e., glucose <3.0 mmol/L) (6), based on self-monitored blood glucose (International Federation of Clinical Chemistry and Laboratory Medicine plasma-equivalent referent meters; Roche Diagnostics, Indianapolis, IN) and dosing aid (Supplementary Fig. 1). The BIO-phase identified patients requiring additional therapy by failure to reach HbA1c 7.0% (53 mmol/mol) or less on titrated basal insulin and metformin.

All patients provided prior written informed consent. The protocol was approved by ethics and regulatory committees and institutional review boards in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines.

Randomization

Patients were allocated 1:1 to treatment groups using block randomization with computer-generated random sequence stratified by country and prior sulfonylurea use.

Procedures

At randomization (defined as 0 weeks), daily glargine was reduced 10% or more in patients allocated to exenatide with HbA1c of 8.0% (64 mmol/mol) or less. In a prior study comparing addition of exenatide twice a day or placebo to insulin glargine, reduction of glargine by 20% in patients with HbA1c of 8.0% or less was associated with a low risk of major hypoglycemia; however, there was no clear evidence that a reduction of this magnitude was necessary to reduce hypoglycemia risk. A more conservative approach of a 10% reduction was taken in this study (3). Patients were instructed to inject exenatide twice daily before the two largest meals, with at least 6 h between dosing. Regimen was 5 μg twice daily per injection for the first 4 weeks and 10 μg per injection thereafter. Exenatide dose reduction was allowed based on tolerability.

Patients randomized to lispro reduced daily glargine by one-half or one-third, at the investigator’s discretion. In previous studies that added mealtime lispro to glargine, reductions by one-half were associated with a low incidence of severe hypoglycemia (7,8). The reduced amount was replaced with three doses of lispro injected before meals to maintain the same total insulin dose. Thereafter, glargine was titrated as in the BIO-phase (Supplementary Fig. 1) (6), and lispro was titrated based on self-monitored premeal glucose values (Supplementary Fig. 2) (7).

The recommended insulin titration algorithms were prospectively planned, aiming to assure consistent insulin titration across multiple sites and countries involved in the study. Patients were contacted once weekly in person or by telephone for the first 8 weeks after randomization and biweekly thereafter. Participants were asked to monitor FG levels daily (6,7), perform a four-point self-monitoring glucose profile twice weekly (before breakfast, lunch, dinner, and bedtime), and perform a seven-point self-monitoring glucose profile within 2 weeks of scheduled visits. Patient compliance was assessed by the investigators or their designee at each visit based on a review of glycemic control, adherence to visit schedule, completion of study diaries, and regular study drug use. Noncompliant patients received education and training as required; consistently noncompliant patients could be discontinued from the study. Nevertheless, no specific study data were collected for analysis of treatment compliance and therefore were not reported separately.

Anthropometric characteristics, blood pressure, and heart rate were measured while fasting at baseline, randomization, and follow-up visits. Patient-reported outcome assessments included Impact of Weight on Quality of Life (IWQOL) questionnaire (short form) (9) and Diabetes Treatment Satisfaction Questionnaire (DTSQ) status or change version (10,11) with measurements at baseline, randomization, week 8, and end point. Samples were collected at all visits for HbA1c and fasting serum glucose. Fasting samples were collected at weeks 0 and 30 for lipids and exploratory biomarkers (Quintiles Central Laboratories, Marietta, GA). Antiexenatide antibodies (12) were measured at weeks 0, 8, and 30 (Millipore Corp., St. Charles, MO). High antibody positivity was defined as a titer ≥1/625.

Primary outcome was change from randomization to 30 weeks in HbA1c. Secondary outcomes included weight, percentage of patients achieving HbA1c targets with weight gain of 1 kg or less, HbA1c of 6.5% (48 mmol/mol) or less or 7.0% (53 mmol/mol) or less, FG, seven-point self-monitored blood glucose profiles, insulin dose, blood pressure, and lipid measurements. Exploratory biomarkers included 1,5-anhydroglucitol, adiponectin, hs-CRP, and urine albumin/creatinine. Safety measures included hypoglycemic events (Supplementary Table 2) and adverse events.

Statistical Analyses

The primary objective of the study was to compare the difference in HbA1c change from randomization to 30 weeks between exenatide or lispro added to glargine in the per-protocol (PP) population. Noninferiority was assessed using an HbA1c margin of 0.4%. If this objective was met, a second noninferiority comparison would be conducted with an HbA1c margin of 0.3%. If noninferiority was established using this stricter margin, superiority would be tested in the intent-to-treat (ITT) population. Due to the ordered nature of objectives, no adjustment to significance level was required to maintain the study-wise significance level. Significance testing was conducted at the two-sided 5% level.

A PP population of 464 patients was calculated to provide 96% power to conclude noninferiority between the treatments in change from randomization HbA1c assuming that there was no difference and the SD was 1.15%. This sample size would provide 80% power to conclude noninferiority with a margin of 0.3%. Based on previous studies (3,13), it was assumed that 20% of randomized patients would be excluded from the PP analysis; therefore, 580 randomized patients were required.

The primary noninferiority objective was assessed using the PP population, as ITT analyses of noninferiority are prone to increased chance of incorrectly concluding noninferiority when one treatment is actually inferior. The PP population was defined as all randomized patients who completed the study and had no violation of the inclusion/exclusion/discontinuation criteria. As supportive evidence, the primary analysis was repeated using the ITT analysis set, i.e., all randomized subjects receiving at least one dose of study drug grouped according to randomized treatment, regardless of the study drug actually received. Secondary analyses were conducted in the PP population. Exploratory end points and post hoc subgroup analyses were conducted using the ITT population. Safety analyses were conducted on the as-treated analysis set, i.e., all randomized subjects taking at least one dose of study drug according to the treatment received. The ITT and as-treated populations were identical.

Mixed models for repeated measures (MMRM) were used for the primary analysis with baseline HbA1c, country, prior use of sulfonylureas, randomized treatment, visit, and treatment-by-visit interaction as fixed effects. Noninferiority was concluded if the upper bound of the 95% CI for the treatment contrast at week 30 excluded the corresponding noninferiority margin. ANCOVA analyses using last observation carried forward were also conducted on PP and ITT populations. The proportion of patients achieving glycemic targets were analyzed using logistic regression models with baseline HbA1c, prior use of sulfonylureas, and country as fixed effects. Variables measured twice (e.g., lipids) were analyzed with an ANCOVA model. The proportion of patients with one or more hypoglycemic episodes was compared using Fisher exact test. All treatment comparisons used a two-sided 5% significance level.

Study Population

Of 1,036 patients screened, 917 entered the 12-week BIO-phase (Fig. 1) during which mean HbA1c decreased from 8.5% (69 mmol/mol) to 8.2% (66 mmol/mol) and FG from 8.4 to 7.2 mmol/L (Table 1). At week 12, 92 patients achieved HbA1c of 7% (53 mmol/mol) or less, with HbA1c decreasing from 7.9% (63 mmol/mol) to 6.7% (50 mmol/mol) and FG from 7.1 to 5.9 mmol/L; 173 patients withdrew for reasons specified in Fig. 1. Despite up-titration of glargine, HbA1c decreased from 8.6% (70 mmol/mol) to 8.4% (68 mmol/mol) and FG from 8.6 to 7.3 mmol/L in 652 patients. Of these eligible patients, 637 were randomized to add exenatide or lispro to titrated glargine and 627 patients continued in the study; 36.2% had used sulfonylureas prior to enrollment in the BIO-phase. Characteristics at randomization were similar between arms (Table 2). Randomized patients had a mean age of 59.8 years, median diabetes duration of 12 years, and mean HbA1c of 8.2% (66 mmol/mol). The majority were Caucasian and overweight (Table 2). Discontinuation rates were comparable for exenatide (316 ITT; 247 PP) and lispro (321 ITT; 263 PP) (Fig. 1).

Figure 1

Enrollment and outcomes.

Figure 1

Enrollment and outcomes.

Close modal
Table 1

Baseline and end of BIO-phase characteristics of the study population

CharacteristicEnrolledRespondersFailures
(n = 917)(n = 92)(n = 652)
Demographic    
 Men 457 (50%) 56 (61%) 315 (48%) 
 Age (years), mean (SD)  59.8 (9.5) 57.4 (8.9) 59.7 (9.4) 
 Duration of diabetes (years), median (IQR) 12 (8–16) 10 (6–14) 12 (8–16) 
 Ethnic origin    
  White 809 (88%) 79 (86%) 568 (87%) 
  Asian 49 (5%) 5 (5%) 37 (6%) 
  African American 5 (<1%) 0 (0%) 3 (<1%) 
  Other* 53 (6%) 8 (9%) 43 (7%) 
Clinical    
 Use of glucose-lowering medication    
  Metformin daily dose (mg), mean (SD)  2,027 (655) 2,037 (703) 2,010 (648) 
  Insulin glargine (units/day), mean (SD)    
   Start of BIO-phase 40.7 (24.2) 34.5 (17.0) 40.6 (24.2) 
   End of BIO-phase 56.8 (30.5) 47.3 (22.2) 58.0 (31.0) 
  Insulin glargine (units/day), median (IQR)    
   Start of BIO-phase 34 (24–46) 30 (24–40) 34 (25–46) 
   End of BIO-phase 48 (38–70) 41 (31–56) 50 (38–70) 
 Glycemic measures    
  HbA1c (%) [mmol/mol], mean (SD)     
   Start of BIO-phase 8.5 (1.0) [69 (10.9)] 7.9 (0.7) [63 (7.7)] 8.6 (0.8) [70 (8.7)] 
   End of BIO-phase 8.2 (1.0) [66 (10.9)] 6.7 (0.4) [50 (4.4)] 8.4 (0.9) [68 (9.8)] 
  FG (mmol/L), mean (SD)     
   Start of BIO-phase 8.4 (2.8) 7.1 (2.2) 8.6 (2.8) 
   End of BIO-phase 7.2 (2.4) 5.9 (1.7) 7.3 (2.4) 
 Anthropometric measures    
  Body weight (kg), mean (SD)    
   Start of BIO-phase 88.9 (17.4) 89.4 (18.1) 88.9 (16.9) 
   End of BIO-phase 89.3 (17.1) 90.1 (17.5) 89.3 (17.0) 
  BMI (kg/m2), mean (SD)    
   Start of BIO-phase 32.1 (5.0) 31.8 (4.7) 32.2 (4.8) 
   End of BIO-phase 32.3 (4.8) 31.5 (4.5) 32.3 (4.8) 
  Blood pressure (mmHg), mean (SD)    
   Systolic    
    Start of BIO-phase 138 (16) 137 (18) 137 (15) 
    End of BIO-phase — — — 
   Diastolic    
    Start of BIO-phase 80 (10) 80 (10) 79 (10) 
    End of BIO-phase — — — 
  Pulse rate (beats per min), mean (SD)     
   Start of BIO-phase 76 (11) 77 (11) 75 (11) 
   End of BIO-phase — — — 
CharacteristicEnrolledRespondersFailures
(n = 917)(n = 92)(n = 652)
Demographic    
 Men 457 (50%) 56 (61%) 315 (48%) 
 Age (years), mean (SD)  59.8 (9.5) 57.4 (8.9) 59.7 (9.4) 
 Duration of diabetes (years), median (IQR) 12 (8–16) 10 (6–14) 12 (8–16) 
 Ethnic origin    
  White 809 (88%) 79 (86%) 568 (87%) 
  Asian 49 (5%) 5 (5%) 37 (6%) 
  African American 5 (<1%) 0 (0%) 3 (<1%) 
  Other* 53 (6%) 8 (9%) 43 (7%) 
Clinical    
 Use of glucose-lowering medication    
  Metformin daily dose (mg), mean (SD)  2,027 (655) 2,037 (703) 2,010 (648) 
  Insulin glargine (units/day), mean (SD)    
   Start of BIO-phase 40.7 (24.2) 34.5 (17.0) 40.6 (24.2) 
   End of BIO-phase 56.8 (30.5) 47.3 (22.2) 58.0 (31.0) 
  Insulin glargine (units/day), median (IQR)    
   Start of BIO-phase 34 (24–46) 30 (24–40) 34 (25–46) 
   End of BIO-phase 48 (38–70) 41 (31–56) 50 (38–70) 
 Glycemic measures    
  HbA1c (%) [mmol/mol], mean (SD)     
   Start of BIO-phase 8.5 (1.0) [69 (10.9)] 7.9 (0.7) [63 (7.7)] 8.6 (0.8) [70 (8.7)] 
   End of BIO-phase 8.2 (1.0) [66 (10.9)] 6.7 (0.4) [50 (4.4)] 8.4 (0.9) [68 (9.8)] 
  FG (mmol/L), mean (SD)     
   Start of BIO-phase 8.4 (2.8) 7.1 (2.2) 8.6 (2.8) 
   End of BIO-phase 7.2 (2.4) 5.9 (1.7) 7.3 (2.4) 
 Anthropometric measures    
  Body weight (kg), mean (SD)    
   Start of BIO-phase 88.9 (17.4) 89.4 (18.1) 88.9 (16.9) 
   End of BIO-phase 89.3 (17.1) 90.1 (17.5) 89.3 (17.0) 
  BMI (kg/m2), mean (SD)    
   Start of BIO-phase 32.1 (5.0) 31.8 (4.7) 32.2 (4.8) 
   End of BIO-phase 32.3 (4.8) 31.5 (4.5) 32.3 (4.8) 
  Blood pressure (mmHg), mean (SD)    
   Systolic    
    Start of BIO-phase 138 (16) 137 (18) 137 (15) 
    End of BIO-phase — — — 
   Diastolic    
    Start of BIO-phase 80 (10) 80 (10) 79 (10) 
    End of BIO-phase — — — 
  Pulse rate (beats per min), mean (SD)     
   Start of BIO-phase 76 (11) 77 (11) 75 (11) 
   End of BIO-phase — — — 

Enrolled, patients who enrolled in the BIO-phase; responders, patients at the end of the BIO-phase who had HbA1c of 7% or less after 12 weeks of titrated insulin glargine; failures, patients at the end of the BIO-phase who had HbA1c >7% and who were eligible to enroll in the intervention phase of the study.

*American Indian or Native Alaskan.

†As-treated population.

Table 2

Characteristics and results of patients at randomization and at study end point

Characteristic/variableExenatide (n = 247)Lispro (n = 263)Exenatide-lispro
Demographics    
 Men 128 (52%) 133 (51%)  
 Age (years), mean (SD)  59.5 (9.6) 59.4 (9.3)  
 Duration of diabetes (years), median (IQR)  12 (8–17) 11 (8–15)  
 Ethnic origin    
  White 222 (90%) 229 (87%)  
  Asian 11 (5%) 14 (5%)  
  African American 2 (<1%) 1 (<1%)  
  Other* 12 (5%) 18 (7%)  
Use of glucose-lowering agents    
 Metformin daily dose (mg), mean (SD)  2,038 (633) 1,998 (648)  
 Previous sulfonylurea use 85 (34%) 99 (38%)  
 Insulin glargine (units/day)    
  Randomization, mean (SD) 61.5 (30.9) 61.1 (35.2)  
  End point, mean (SD) 56.9 (29.4) 51.5 (31.4)  
  End point, LS mean (SE) (95% CI) 56.8 (1.2) 51.5 (1.2) 5.3 (2.1, 8.5) 
  Randomization, median (IQR) 54 (40–76) 52 (38–76)  
  End point, median (IQR) 50 (36–74) 46 (34–62)  
 Insulin lispro (units/day)    
  1 week, mean (SD)  — 24.6 (12.7)  
  End point, mean (SD)  — 42.1 (30.8)  
  1 week, median (IQR) — 22 (16–30)  
  End point, median (IQR)  — 34 (24–52)  
 Insulin glargine + insulin lispro (units/day)    
  1 week, mean (SD) — 64.0 (32.9)  
  End point, mean (SD)  — 93.7 (53.9)  
  1 week, median (IQR) — 56 (42–80)  
  End point, median (IQR)  — 82 (60–108)  
 Exenatide (μg/day)    
  1 week, mean (SD) 9.9 (0.8) —  
  End point, mean (SD) 18.6 (3.6) —  
  Randomization, median (IQR) 10 (10–10) —  
  End point, median (IQR) 20 (20–20) —  
Glycemic measures    
 HbA1c (%) [mmol/mol]    
  Randomization, mean (SD) 8.3 (1.0) [67 (11)] 8.2 (0.9) [66 (10)]  
  End point, mean (SD) 7.2 (1.0) [55 (11)] 7.2 (1.0) [55 (11)]  
  Change from randomization, LS mean (SE) (95% CI) −1.13 (0.05) [–12.4 (0.5)] −1.10 (0.05) [–12.0 (0.5)] −0.04 (–0.18, 0.11) 
 Patients with HbA1c ≤7% (% [95% CI]) 49.6 [42.6, 55.4] 49.0 [42.9, 55.3]  
 Patients with HbA1c ≤6.5% (% [95% CI]) 26.2 [20.6, 31.8] 25.5 [20.3, 31.2]  
 FG (mmol/L)    
  Randomization, mean (SD) 7.1 (2.3) 7.1 (2.5)  
  End point, mean (SD) 6.5 (2.0) 7.2 (2.8)  
  Change from randomization, LS mean (SE) (95% CI) −0.46 (0.16) 0.18 (0.15) −0.64 (–1.05, –0.24)§ 
 Self-monitored blood glucose (mmol/L)    
  Morning 2 h postmeal    
   Randomization, mean (SD)  9.97 (2.94) 10.21 (2.89)  
   End point, mean (SD)  7.42 (1.97) 7.74 (2.10)  
   Change from randomization, LS mean (SE) (95% CI) −2.57 (0.14) −2.30 (0.14) −0.27 (–0.63, 0.09) 
  Midday 2 h postmeal    
   Randomization, mean (SD) 10.4 (2.7) 10.9 (3.0)  
   End point, mean (SD) 8.5 (2.4) 7.6 (2.2)  
   Change from randomization, LS mean (SE) (95% CI) −2.18 (0.16) −3.11 (0.16) 0.93 (0.52, 1.34) 
  Evening 2 h postmeal    
   Randomization, mean (SD) 10.9 (3.2) 11.1 (2.9)  
   End point, mean (SD) 8.2 (2.6) 7.8 (2.4)  
   Change from randomization, LS mean (SE) (95% CI) −2.88 (0.17) −3.16 (0.17) 0.28 (–0.16, 0.72) 
 1,5 anhydroglucitol (μg/mL)**    
  Randomization, median (IQR) 5.7 (3.2–9.0) 5.1 (2.9–9.1)  
  End point, median (IQR)†† 9.8 (5.3–14.0) 10.0 (5.5–15.4)  
  Change from randomization, LS mean (SE) (95% CI)‡‡ 1.2 (3.0) 5.3 (3.0) −4.2 (–11.4, 3.1) 
 Fasting glucagon (pmol/L)    
  Randomization, mean (SD) 24.8 (10.1) 24.3 (9.2)  
  End point, mean (SD) 25.6 (12.0) 24.3 (10.1)  
  Change from randomization, LS mean (SE) (95% CI)‡‡ 1.3 (0.7) 0.3 (0.7) 1.1 (–0.6, 2.7) 
Weight and composite measures    
 BMI (kg/m2   
  Randomization, mean (SD) 32.7 (4.7) 32.3 (4.7)  
  End point, mean (SD) 31.8 (4.6) 33.1 (5.1)  
  Change from randomization, LS mean (SE) (95% CI) −0.9 (0.1) 0.8 (0.1) −1.7 (–1.9, –1.4) 
 Body weight (kg)    
  Randomization, mean (SD) 91.1 (16.6) 89.4 (17.0)  
  End point, mean (SD) 88.6 (16.5) 91.4 (17.9)  
  Change from randomization, LS mean (SE) (95% CI) −2.5 (0.3) 2.1 (0.2) −4.6 (–5.2, −3.9) 
 Waist circumference (cm)    
  Randomization, mean (SD) 108 (13) 108 (13)  
  End point, mean (SD) 106 (13) 108 (14)  
  Change from randomization, LS mean (SE) (95% CI) −2.0 (0.4) 0.7 (0.3) −2.7 (–3.6, –1.8) 
 Hip circumference (cm)    
  Randomization, mean (SD) 110 (12) 110 (11)  
  End point, mean (SD) 109 (11) 111 (12)  
  Change from randomization, LS mean (SE) (95% CI) −1.8 (0.4) 0.9 (0.4) −3.0 (–3.6, –1.9) 
 HbA1c ≤7% and ≤1 kg weight gain (% [95% CI]) 44.6 [37.4, 50.2] 22.9 [17.9, 28.4]ǁ  
 HbA1c ≤6.5% and ≤1 kg weight gain (% [95% CI]) 23.1 [17.6, 28.4] 14.5 [10.4, 19.3]§§  
Cardiovascular parameters    
 Systolic blood pressure (mmHg)    
  Randomization, mean (SD) 137 (16) 135 (15)  
  End point, mean (SD) 133 (14) 136 (16)  
  Change from randomization, LS mean (SE) (95% CI) −4.1 (1.0) 0.4 (0.9) −4.5 (–7.0, –2.0) 
 Diastolic blood pressure (mmHg)    
  Randomization, mean (SD) 79 (10) 78 (9)  
  End point, mean (SD) 79 (10) 79 (10)  
  Change from randomization, LS mean (SE) (95% CI) −0.6 (0.6) −0.1 (0.6) −0.5 (–2.1, 1.1) 
 Pulse rate (beats per min)    
  Randomization, mean (SD) 74 (11) 73 (10)  
  End point, mean (SD) 75 (11) 74 (11)  
  Change from randomization, LS mean (SE) (95% CI) 1.5 (0.6) 0.9 (0.6) 0.6 (–1.1, 2.2) 
 Cholesterol (mmol/L)    
  Total cholesterol    
   Randomization, mean (SD) 4.6 (1.0) 4.6 (1.0)  
   End point, mean (SD) 4.5 (1.0) 4.6 (1.1)  
   Change from randomization, LS mean (SE) (95% CI)‡‡ −0.1 (0.1) −0.0 (0.0) −0.1 (–0.2, 0.0) 
  LDL cholesterol    
   Randomization, mean (SD) 2.5 (0.8) 2.6 (0.9)  
   End point, mean (SD) 2.5 (0.8) 2.6 (0.9)  
   Change from randomization, LS mean (SE) (95% CI)‡‡ −0.1 (0.0) −0.0 (0.0) −0.1 (–0.2, 0.0) 
  HDL cholesterol    
   Randomization, mean (SD) 1.2 (0.3) 1.2 (0.3)  
   End point, mean (SD) 1.2 (0.3) 1.2 (0.3)  
   Change from randomization, LS mean (SE) (95% CI)‡‡ −0.04 (0.01) 0.03 (0.01) −0.07 (–0.10, –0.04) 
 Triglycerides (mmol/L)    
  Randomization, median (IQR) 1.6 (1.2–2.1) 1.5 (1.1–2.0)  
  End point, median (IQR) 1.6 (1.2–2.3) 1.4 (1.1–2.0)  
  Change from randomization, LS mean (SE) (95% CI)‡‡ 0.06 (0.06) −0.05 (0.06) 0.11 (–0.03, 0.24) 
 Adiponectin (ng/L)**    
  Randomization, median (IQR) 5,771 (3,957–7,737) 5,406 (3,869–7,778)  
  End point, median (IQR) 6,061 (4,103–8,581) 5,201 (3,541–7,985)  
  Change from randomization, LS mean (SE) (95% CI)‡‡ −491 (247) −651 (249) 160 (–430, 750) 
 hs-CRP (mg/L)**    
  Randomization, mean (SD) 4.9 (5.7) 4.5 (5.9)  
  End point, mean (SD)†† 4.3 (6.6) 4.9 (6.5)  
  Change from randomization, LS mean (SE) (95% CI)‡‡ −0.6 (0.4) 0.2 (0.4) −0.8 (–1.7, 0.1) 
  Randomization, median (IQR) 3.2 (1.5–6.2) 2.6 (1.3–5.4)  
  End point, median (IQR)  2.7 (1.1–5.3) 2.7 (1.1–5.9)  
 Urine albumin/creatinine (mg/mmol)**    
  Randomization, median (IQR) 1.6 (0.6–5.0) 1.4 (0.6–5.2)  
  End point, median (IQR)†† 1.1 (0.5–2.9) 1.2 (0.5–5.2)  
  Change from randomization, LS mean (SE) (95% CI)‡‡ −4.2 (2.6) −3.7 (2.6) −0.5 (–6.7, 5.7) 
Characteristic/variableExenatide (n = 247)Lispro (n = 263)Exenatide-lispro
Demographics    
 Men 128 (52%) 133 (51%)  
 Age (years), mean (SD)  59.5 (9.6) 59.4 (9.3)  
 Duration of diabetes (years), median (IQR)  12 (8–17) 11 (8–15)  
 Ethnic origin    
  White 222 (90%) 229 (87%)  
  Asian 11 (5%) 14 (5%)  
  African American 2 (<1%) 1 (<1%)  
  Other* 12 (5%) 18 (7%)  
Use of glucose-lowering agents    
 Metformin daily dose (mg), mean (SD)  2,038 (633) 1,998 (648)  
 Previous sulfonylurea use 85 (34%) 99 (38%)  
 Insulin glargine (units/day)    
  Randomization, mean (SD) 61.5 (30.9) 61.1 (35.2)  
  End point, mean (SD) 56.9 (29.4) 51.5 (31.4)  
  End point, LS mean (SE) (95% CI) 56.8 (1.2) 51.5 (1.2) 5.3 (2.1, 8.5) 
  Randomization, median (IQR) 54 (40–76) 52 (38–76)  
  End point, median (IQR) 50 (36–74) 46 (34–62)  
 Insulin lispro (units/day)    
  1 week, mean (SD)  — 24.6 (12.7)  
  End point, mean (SD)  — 42.1 (30.8)  
  1 week, median (IQR) — 22 (16–30)  
  End point, median (IQR)  — 34 (24–52)  
 Insulin glargine + insulin lispro (units/day)    
  1 week, mean (SD) — 64.0 (32.9)  
  End point, mean (SD)  — 93.7 (53.9)  
  1 week, median (IQR) — 56 (42–80)  
  End point, median (IQR)  — 82 (60–108)  
 Exenatide (μg/day)    
  1 week, mean (SD) 9.9 (0.8) —  
  End point, mean (SD) 18.6 (3.6) —  
  Randomization, median (IQR) 10 (10–10) —  
  End point, median (IQR) 20 (20–20) —  
Glycemic measures    
 HbA1c (%) [mmol/mol]    
  Randomization, mean (SD) 8.3 (1.0) [67 (11)] 8.2 (0.9) [66 (10)]  
  End point, mean (SD) 7.2 (1.0) [55 (11)] 7.2 (1.0) [55 (11)]  
  Change from randomization, LS mean (SE) (95% CI) −1.13 (0.05) [–12.4 (0.5)] −1.10 (0.05) [–12.0 (0.5)] −0.04 (–0.18, 0.11) 
 Patients with HbA1c ≤7% (% [95% CI]) 49.6 [42.6, 55.4] 49.0 [42.9, 55.3]  
 Patients with HbA1c ≤6.5% (% [95% CI]) 26.2 [20.6, 31.8] 25.5 [20.3, 31.2]  
 FG (mmol/L)    
  Randomization, mean (SD) 7.1 (2.3) 7.1 (2.5)  
  End point, mean (SD) 6.5 (2.0) 7.2 (2.8)  
  Change from randomization, LS mean (SE) (95% CI) −0.46 (0.16) 0.18 (0.15) −0.64 (–1.05, –0.24)§ 
 Self-monitored blood glucose (mmol/L)    
  Morning 2 h postmeal    
   Randomization, mean (SD)  9.97 (2.94) 10.21 (2.89)  
   End point, mean (SD)  7.42 (1.97) 7.74 (2.10)  
   Change from randomization, LS mean (SE) (95% CI) −2.57 (0.14) −2.30 (0.14) −0.27 (–0.63, 0.09) 
  Midday 2 h postmeal    
   Randomization, mean (SD) 10.4 (2.7) 10.9 (3.0)  
   End point, mean (SD) 8.5 (2.4) 7.6 (2.2)  
   Change from randomization, LS mean (SE) (95% CI) −2.18 (0.16) −3.11 (0.16) 0.93 (0.52, 1.34) 
  Evening 2 h postmeal    
   Randomization, mean (SD) 10.9 (3.2) 11.1 (2.9)  
   End point, mean (SD) 8.2 (2.6) 7.8 (2.4)  
   Change from randomization, LS mean (SE) (95% CI) −2.88 (0.17) −3.16 (0.17) 0.28 (–0.16, 0.72) 
 1,5 anhydroglucitol (μg/mL)**    
  Randomization, median (IQR) 5.7 (3.2–9.0) 5.1 (2.9–9.1)  
  End point, median (IQR)†† 9.8 (5.3–14.0) 10.0 (5.5–15.4)  
  Change from randomization, LS mean (SE) (95% CI)‡‡ 1.2 (3.0) 5.3 (3.0) −4.2 (–11.4, 3.1) 
 Fasting glucagon (pmol/L)    
  Randomization, mean (SD) 24.8 (10.1) 24.3 (9.2)  
  End point, mean (SD) 25.6 (12.0) 24.3 (10.1)  
  Change from randomization, LS mean (SE) (95% CI)‡‡ 1.3 (0.7) 0.3 (0.7) 1.1 (–0.6, 2.7) 
Weight and composite measures    
 BMI (kg/m2   
  Randomization, mean (SD) 32.7 (4.7) 32.3 (4.7)  
  End point, mean (SD) 31.8 (4.6) 33.1 (5.1)  
  Change from randomization, LS mean (SE) (95% CI) −0.9 (0.1) 0.8 (0.1) −1.7 (–1.9, –1.4) 
 Body weight (kg)    
  Randomization, mean (SD) 91.1 (16.6) 89.4 (17.0)  
  End point, mean (SD) 88.6 (16.5) 91.4 (17.9)  
  Change from randomization, LS mean (SE) (95% CI) −2.5 (0.3) 2.1 (0.2) −4.6 (–5.2, −3.9) 
 Waist circumference (cm)    
  Randomization, mean (SD) 108 (13) 108 (13)  
  End point, mean (SD) 106 (13) 108 (14)  
  Change from randomization, LS mean (SE) (95% CI) −2.0 (0.4) 0.7 (0.3) −2.7 (–3.6, –1.8) 
 Hip circumference (cm)    
  Randomization, mean (SD) 110 (12) 110 (11)  
  End point, mean (SD) 109 (11) 111 (12)  
  Change from randomization, LS mean (SE) (95% CI) −1.8 (0.4) 0.9 (0.4) −3.0 (–3.6, –1.9) 
 HbA1c ≤7% and ≤1 kg weight gain (% [95% CI]) 44.6 [37.4, 50.2] 22.9 [17.9, 28.4]ǁ  
 HbA1c ≤6.5% and ≤1 kg weight gain (% [95% CI]) 23.1 [17.6, 28.4] 14.5 [10.4, 19.3]§§  
Cardiovascular parameters    
 Systolic blood pressure (mmHg)    
  Randomization, mean (SD) 137 (16) 135 (15)  
  End point, mean (SD) 133 (14) 136 (16)  
  Change from randomization, LS mean (SE) (95% CI) −4.1 (1.0) 0.4 (0.9) −4.5 (–7.0, –2.0) 
 Diastolic blood pressure (mmHg)    
  Randomization, mean (SD) 79 (10) 78 (9)  
  End point, mean (SD) 79 (10) 79 (10)  
  Change from randomization, LS mean (SE) (95% CI) −0.6 (0.6) −0.1 (0.6) −0.5 (–2.1, 1.1) 
 Pulse rate (beats per min)    
  Randomization, mean (SD) 74 (11) 73 (10)  
  End point, mean (SD) 75 (11) 74 (11)  
  Change from randomization, LS mean (SE) (95% CI) 1.5 (0.6) 0.9 (0.6) 0.6 (–1.1, 2.2) 
 Cholesterol (mmol/L)    
  Total cholesterol    
   Randomization, mean (SD) 4.6 (1.0) 4.6 (1.0)  
   End point, mean (SD) 4.5 (1.0) 4.6 (1.1)  
   Change from randomization, LS mean (SE) (95% CI)‡‡ −0.1 (0.1) −0.0 (0.0) −0.1 (–0.2, 0.0) 
  LDL cholesterol    
   Randomization, mean (SD) 2.5 (0.8) 2.6 (0.9)  
   End point, mean (SD) 2.5 (0.8) 2.6 (0.9)  
   Change from randomization, LS mean (SE) (95% CI)‡‡ −0.1 (0.0) −0.0 (0.0) −0.1 (–0.2, 0.0) 
  HDL cholesterol    
   Randomization, mean (SD) 1.2 (0.3) 1.2 (0.3)  
   End point, mean (SD) 1.2 (0.3) 1.2 (0.3)  
   Change from randomization, LS mean (SE) (95% CI)‡‡ −0.04 (0.01) 0.03 (0.01) −0.07 (–0.10, –0.04) 
 Triglycerides (mmol/L)    
  Randomization, median (IQR) 1.6 (1.2–2.1) 1.5 (1.1–2.0)  
  End point, median (IQR) 1.6 (1.2–2.3) 1.4 (1.1–2.0)  
  Change from randomization, LS mean (SE) (95% CI)‡‡ 0.06 (0.06) −0.05 (0.06) 0.11 (–0.03, 0.24) 
 Adiponectin (ng/L)**    
  Randomization, median (IQR) 5,771 (3,957–7,737) 5,406 (3,869–7,778)  
  End point, median (IQR) 6,061 (4,103–8,581) 5,201 (3,541–7,985)  
  Change from randomization, LS mean (SE) (95% CI)‡‡ −491 (247) −651 (249) 160 (–430, 750) 
 hs-CRP (mg/L)**    
  Randomization, mean (SD) 4.9 (5.7) 4.5 (5.9)  
  End point, mean (SD)†† 4.3 (6.6) 4.9 (6.5)  
  Change from randomization, LS mean (SE) (95% CI)‡‡ −0.6 (0.4) 0.2 (0.4) −0.8 (–1.7, 0.1) 
  Randomization, median (IQR) 3.2 (1.5–6.2) 2.6 (1.3–5.4)  
  End point, median (IQR)  2.7 (1.1–5.3) 2.7 (1.1–5.9)  
 Urine albumin/creatinine (mg/mmol)**    
  Randomization, median (IQR) 1.6 (0.6–5.0) 1.4 (0.6–5.2)  
  End point, median (IQR)†† 1.1 (0.5–2.9) 1.2 (0.5–5.2)  
  Change from randomization, LS mean (SE) (95% CI)‡‡ −4.2 (2.6) −3.7 (2.6) −0.5 (–6.7, 5.7) 

All values are for the PP population unless otherwise noted. LS, least squares.

*American Indian or Native Alaskan.

†Calculated using MMRM.

P = 0.001.

§P = 0.002.

¶From self-monitored blood glucose profiles.

P < 0.001.

**ITT population.

††End point values are last observation carried forward.

‡‡Calculated using ANCOVA at end point for change from randomization.

§§P = 0.005.

Primary Outcome

The difference in HbA1c change from randomization to end point at 30 weeks between exenatide or lispro added to glargine was –0.04 (95% CI –0.18, 0.11) in the PP population. The upper limit of the 95% CI for the treatment difference was <0.4 and <0.3%, demonstrating noninferiority. Changes from randomization in HbA1c with the addition of exenatide or lispro to glargine were –1.13 (12.4 mmol/mol) (95% CI –1.24, –1.03) and –1.10 (12.0 mmol/mol) (95% CI –1.20, –1.00) (P = 0.627), respectively (Table 2 and Fig. 2A). In the ITT population, the difference in change in HbA1c from randomization to end point between exenatide or lispro added to glargine was –0.03 (95% CI –0.16, 0.11). Changes from randomization in HbA1c with added exenatide or lispro were –1.10 (95% CI –1.20, –1.00) and –1.07 (–1.17, –0.97) (P = 0.702), respectively (Fig. 2B and Supplementary Table 3).

Figure 2

Primary and secondary outcomes from randomization to 30 weeks. A: Least-squares (LS) mean change (SE) in HbA1c values from the MMRM model in the PP population. B: LS mean change (SE) in HbA1c values from the MMRM model in the ITT population. C: LS mean change (SE) for FG in the PP population (*P = 0.002). D: Self-monitored blood glucose mean values (SE) at randomization (open symbols, dashed lines) and at 30 weeks (closed symbols, solid line) (†P < 0.001, lispro vs. exenatide). E: LS mean change (SE) for weight in the PP population (‡P < 0.001). F: Insulin dose (units/day) mean (SD) for basal insulin glargine in the exenatide and lispro arms (open symbols) and the total insulin dose (glargine plus lispro) in the lispro arm (closed symbol).

Figure 2

Primary and secondary outcomes from randomization to 30 weeks. A: Least-squares (LS) mean change (SE) in HbA1c values from the MMRM model in the PP population. B: LS mean change (SE) in HbA1c values from the MMRM model in the ITT population. C: LS mean change (SE) for FG in the PP population (*P = 0.002). D: Self-monitored blood glucose mean values (SE) at randomization (open symbols, dashed lines) and at 30 weeks (closed symbols, solid line) (†P < 0.001, lispro vs. exenatide). E: LS mean change (SE) for weight in the PP population (‡P < 0.001). F: Insulin dose (units/day) mean (SD) for basal insulin glargine in the exenatide and lispro arms (open symbols) and the total insulin dose (glargine plus lispro) in the lispro arm (closed symbol).

Close modal

Secondary Outcomes

Proportions of patients with HbA1c ≤7.0% (53 mmol/mol) or 6.5% (48 mmol/mol) were similar (Table 2). Achievement of composite HbA1c and weight targets favored exenatide (P < 0.001) (Table 2). FG decreased with exenatide (P = 0.002) (Fig. 2C). Changes in postprandial glucose were similar except after lunch (P < 0.001) (Table 2 and Fig. 2D). After 30 weeks, average weight had decreased with exenatide and increased with lispro (P < 0.001) (Table 2 and Fig. 2F). No significant between-group differences were noted for fasting glucagon or total cholesterol, LDL cholesterol, or triglycerides; however, a between-group difference for HDL cholesterol was noted (Table 2).

At study end, mean glargine dose had decreased somewhat in both groups (Table 2 and Fig. 2E), while the mean glargine dose was 56.8 units in the exenatide group and 51.5 units in the lispro group (P < 0.001). In patients receiving exenatide, 73.3% administered 10 μg per injection and 11.1% administered 5 μg per injection. Insulin titration deviations for glargine dosing were 81.4% in the exenatide arm and 80.2% in the lispro arm.

Systolic blood pressure decreased with exenatide (between-group difference, P < 0.001), whereas no between-group differences were noted for diastolic blood pressure or pulse rate (Table 2). No between-group differences were noted for hs-CRP, adiponectin, 1,5-anhydroglucitol, or urine albumin/creatinine (Table 2).

Both exenatide and lispro recipients reported significant improvements in overall perceived treatment satisfaction (DTSQ) and perceived frequency of “high blood sugar” compared with baseline (start of BIO-phase); improvements were greater with exenatide (P = 0.003) (Supplementary Table 5). More “low blood sugar” was perceived for lispro than exenatide. On the DTSQ, average scores for both exenatide and lispro at end point indicated high levels of treatment satisfaction. In addition, exenatide recipients reported significant (P < 0.001) improvement in perceived impact of weight on quality of life (IWQOL total score) from randomization, whereas lispro recipients did not (P = 0.037 between groups).

Safety and Tolerability

Overall, 70 (22%) exenatide patients tested positive for anti-exenatide antibodies at study end; of these, 10 had high-titer antibodies (Supplementary Fig. 4) (12). In patient subpopulations determined by antibody status, antibody positivity was associated with similar HbA1c changes (Supplementary Table 5) and incidences of treatment-emergent adverse events (74.3 vs. 72.2%, respectively).

The proportion of patients experiencing treatment-emergent adverse events was greater for exenatide than lispro (Supplementary Table 4). Gastrointestinal-related adverse events, including nausea, vomiting, and diarrhea, were more common for exenatide (47 vs. 13% for lispro) (Supplementary Table 4). Incidence of hypoglycemia was greater with lispro for minor (41 vs. 30% for exenatide) and confirmed nonnocturnal hypoglycemia (34 vs. 15% for exenatide) (Supplementary Table 4). Two exenatide and seven lispro recipients had at least one major hypoglycemic episode. Nocturnal hypoglycemia was similar for exenatide and lispro (25 vs. 27%, respectively).

One patient died before randomization from multiple organ failure. In the exenatide arm, one patient died of hemorrhagic stroke, and one patient (treated with exenatide for 71 days) was diagnosed with pancreatic cancer, discontinued exenatide and the study, and died 1.5 years poststudy. No other cancers were observed in either arm. The percentage of patients with serious adverse events was similar between groups (5.7% for exenatide and 7.4% for lispro). No pancreatitis was observed. One exenatide-treated patient was hospitalized for acute renal failure and recovered.

For patients with advanced type 2 diabetes unable to achieve glycemic control despite titrated basal insulin titration over 12 weeks (6,14), adding the short-acting GLP-1 receptor agonist exenatide before the two largest meals improved glycemic control similarly to mealtime (thrice daily) lispro. Additionally, exenatide was associated with lower FG, less nonnocturnal hypoglycemia, weight loss, reduced systolic blood pressure, and better quality of life. More adverse events, mainly gastrointestinal, were observed with exenatide than lispro. To date, available studies combining a GLP-1 receptor agonist with basal insulin used placebo as a comparator and failed to first optimize FG and identify patients needing additional therapy targeting postprandial hyperglycemia (3,4,13,15). By directly comparing a short-acting GLP-1 receptor agonist added to titrated basal insulin with a standard basal-bolus insulin regimen (1), this study provides clinically relevant data on the benefits and risks of this strategy.

Greater FG reduction was observed for exenatide versus lispro despite similar HbA1c changes at end point. Although both groups had comparable FG and daily glargine doses after the initial 12-week glargine titration period, glargine dose was reduced more aggressively at randomization in the lispro group to allow initiation of thrice-daily mealtime insulin. Even though the glargine was titrated according to the protocol, the possibility still remains that the lower dosage of glargine in the lispro group at study end could have affected the primary outcome. This strategy may have resulted in the lower doses of glargine with lispro at study end, although in both groups a similar titration algorithm for insulin glargine was used, which also was used in the BIO-phase. Interestingly, the between-group difference in FG disappeared in the subset of patients with confirmed nocturnal hypoglycemia (data not shown), suggesting that glargine dose titration was guided by clinical symptoms rather than by FG values. Additionally, fasting glucagon does not explain the FG difference since glucagon concentrations tended to be higher with exenatide, predictive of increased hepatic glucose output (16,17). The intrinsic ability of short-acting GLP-1 receptor agonists to lower FG concentrations via effects on insulin secretion and/or glucagon suppression (1822), which differs from the short-acting insulins such as lispro, may provide the best explanation of the FG difference. Nevertheless, it remains unclear whether exenatide’s stronger effects on FG compared with lispro are due to clinical factors such as titration practices or to physiological differences in response to therapy.

Control of postmeal glycemic excursions was generally similar for both treatments, with the exception of the midday meal when exenatide was not administered. Mechanisms controlling postprandial glucose differ between these therapies: whereas both exenatide and rapid-acting insulin analogs suppress hepatic glucose output and stimulate peripheral glucose uptake (18,19), only short-acting GLP-1 receptor agonists decelerate gastric emptying (23,24), resulting in slower meal-derived carbohydrate entry into the circulation. The importance of decelerated gastric emptying is demonstrated by the reduction of postmeal insulin responses with exenatide (24) despite the insulinotropic potential of GLP-1 receptor stimulation (23). It is unknown whether similar efficacy can be obtained by combining long-acting GLP-1 receptor agonists with basal insulin because long-acting formulations have less impact on gastric emptying delay and thus reduce mealtime glucose less than short-acting formulations (25). Different mechanisms of action may explain differential risk of nonnocturnal hypoglycemic events, which largely occurred postmeal. The use of rapid-acting insulin analogs is associated with increased hypoglycemia risk (26,27), whereas GLP-1 receptor agonists are associated with low risk of hypoglycemia due to glucose-dependent action (4). Most hypoglycemic events were reported after 12 weeks postrandomization (Supplementary Fig. 3B), when HbA1c and glargine doses were stable, suggesting that hypoglycemia was associated with lispro. The similar incidence of nocturnal hypoglycemia was likely attributable to glargine, as both lispro and exenatide, having short half-lives, were administered during the day.

The observed difference in weight may be important over time as weight gain contributes to both insulin resistance and cardiovascular risk (1,28). Combined improvements in HbA1c, weight, and blood pressure may also favorably affect cardiovascular risk. Although HDL cholesterol was lowered with exenatide, total cholesterol and LDL cholesterol were also lowered by both treatments (the between-group difference did not reach statistical significance).

Patient adherence is an important factor in any therapeutic regimen and is associated with clinical and economic benefits (29). It is possible that some of the randomized patients failed basal insulin during BIO-phase due to unwillingness to titrate, so clinical data on patient adherence to these regimens are needed. A patient’s perception of how treatment affects his or her life is an important clinical consideration and may affect adherence (1). Improved treatment satisfaction was previously noted with GLP-1 receptor agonists (22,30). Here, increased treatment satisfaction was observed for both regimens, suggesting that patients value clinical improvements more than they oppose injections (22,30). Factors such as weight loss, less hypoglycemia, or some combination thereof, may provide greater improvement in treatment satisfaction with exenatide, despite more gastrointestinal adverse effects.

The insulin titration protocols chosen for this study may be considered a study limitation. In general, patient care and follow-up procedures were intended to reflect daily clinical practice, allow individualization of patient care according to clinical judgment, and minimize hypoglycemia risk. Dosing instructions were more straightforward for the exenatide group, for whom routine glargine titration was not reduced at randomization except for patients with an HbA1c of 8.0% or less at introduction of exenatide. To support patient safety, glargine in these patients was reduced by 10% since a previous study of this combination demonstrated acceptable hypoglycemia rates with a 20% initial reduction of glargine in patients with HbA1c of 8.0% or less (3). In contrast, the titration protocols for glargine with lispro were based on or similar to ongoing or recently completed trials of basal-bolus insulin (the IOOQ Study [7] and IOOV/DURABLE Study [8]). The division of daily insulin dose into 50% basal and 50% bolus insulin was common practice when this study was initiated (31), and the final total daily insulin dose was consistent with published studies (32). Nevertheless, some physicians now initiate bolus insulin with the more recent “basal-plus” method in which basal insulin is not reduced and low doses of bolus insulin are added to one or more meals a day (32). The results of comparing a “basal-plus” strategy with addition of exenatide to basal insulin are not known, but we believe the approach used for initiating insulin will not affect the overall results of the study, particularly differences in body weight and hypoglycemia incidence. Addition of exenatide to glargine was associated with improvement in multiple clinical and patient-reported outcome measures, and this approach was supported as viable for routine clinical practice. It remains possible that more efficacious insulin treatment regimens will be developed in the future, for instance those using the new long-acting analogs.

Discontinuation rates, while significant, were consistent with similar studies (3,13). In the exenatide arm, one patient was diagnosed with pancreatic cancer after treatment with exenatide for 71 days. The patient discontinued exenatide and the study and died 1.5 years poststudy. No other cancers were observed in either arm. Based on the time course, it is unlikely that there is a causal relationship with the use of exenatide. Type 2 diabetes itself has been associated with an increased risk of various types of cancer, including pancreatic cancer (33). In a recent review, Nauck and Friedrich (34) reported that there is no firm evidence that supports the hypothesis that GLP-1–based therapy can increase the risk for specific malignant diseases like pancreatic carcinoma and thyroid cancer, nor is there firm evidence to rule out any such increased risk based on results available at present. These conclusions have recently been affirmed by the U.S. Food and Drug Administration and the European Medicines Agency (35).

Other limitations of our study include no extensive mechanistic measurements and past use of diverse therapies. Furthermore, guidance for treating patients who fail despite exenatide addition, which may include rapid-acting insulin, is absent. Lastly, while these data demonstrate the safety and tolerability of exenatide compared with lispro when added to glargine over 30 weeks, additional studies are needed to provide data on the long-term use of these therapies.

Insulin and its analogs have long been the only injectable option for patients with advanced type 2 diabetes. Exenatide, like insulin, also requires injection, albeit with fixed dosing, fewer daily injections, and less need for glucose monitoring. The findings of similar HbA1c reduction but less nonnocturnal hypoglycemia, more weight loss, and improved treatment satisfaction for addition of exenatide instead of lispro to titrated basal insulin challenge current treatment recommendations (1) by suggesting that adding a short-acting GLP-1 receptor agonist to basal insulin may be preferred over mealtime and basal insulin for some patients with type 2 diabetes.

Clinical trial reg. no. NCT00960661, clinicaltrials.gov.

A slide set summarizing this article is available online.

Acknowledgments. The authors thank Birgit Schinzel (Principal Statistician, Accovion GmbH), Claudia Nicolay (Senior Research Scientist, Group Leader Statistics, Eli Lilly and Company), and Antje Hottgenroth (Comparative Data Scientist Data Sciences & Solutions, Eli Lilly and Company) for their contributions to the study and analysis of the data; Mary Beth DeYoung (Global Publications Lead), Pete Griffin (Senior R&D Associate Medical Research), and Julie Ellison (Medical Communications Scientist, Bristol-Myers Squibb Company) for their assistance with the preparation of figures, writing, and editing of the manuscript; and Sushma Soni (inScience Communications, Springer Healthcare) for assistance with copyediting. The authors also thank the participating research physicians, nurses, and patients who made this study possible. The coauthors wish to acknowledge the many years of outstanding dedication and commitment of Professor Michaela Diamant to diabetes care and research. She died on 9 April 2014. She was a committed scholar, a highly valued colleague, and a good friend to all of us.

Duality of Interest. This study was part of the Eli Lilly and Company/Amylin Pharmaceuticals Alliance and the Bristol-Myers Squibb/AstraZeneca Alliance. M.D. was a consultant for Abbott, Bristol-Myers Squibb/AstraZeneca, Boehringer Ingelheim, Eli Lilly and Company, Merck Sharp & Dohme (MSD), Novo Nordisk, Poxel Pharma, and Sanofi and a speaker for Bristol-Myers Squibb/AstraZeneca, Eli Lilly and Company, Novo Nordisk, and Sanofi. Through M.D., the VU University Medical Center received research grants from Abbott, Bristol-Myers Squibb/AstraZeneca, Boehringer Ingelheim, Eli Lilly and Company, MSD, Novo Nordisk, and Sanofi. M.D. reported receiving no personal payments in connection to the above-mentioned activities, but all payments were directly transferred to the Diabetes Centre nonprofit Research Foundation. M.A.N. has received research grants (to his institution, the Diabeteszentrum Bad Lauterberg, Bad Lauterberg, Germany) from Berlin-Chemie AG/Menarini (Berlin, Germany), Eli Lilly and Company (Indianapolis, IN), MSD (München, Germany), and Novartis Pharma AG (Basel, Switzerland) (mono- or oligocentric studies) and from AstraZeneca (Södertälje, Sweden), Boehringer Ingelheim (Ingelheim, Germany), GlaxoSmithKline (King of Prussia, PA), Lilly Deutschland GmbH (Bad Homburg, Germany), MetaCure Inc. (Orangeburg, NY), Roche Pharma AG (Grenzach-Wyhlen, Germany), Novo Nordisk Pharma GmbH (Mainz, Germany), and Tolerx Inc., a Delaware Corporation (Cambridge, MA), for participation in multicentric clinical trials. He has received consulting fees and/or honoraria for membership in advisory boards and/or honoraria for speaking from Amylin Pharmaceuticals, Inc. (San Diego, CA), AstraZeneca (Mölndal, Sweden), Berlin-Chemie AG/Menarini, Boehringer Ingelheim, Bristol-Myers Squibb EMEA (Rueil-Malmaison, France), Diartis Pharmaceuticals, Inc. (Redwood City, CA), Eli Lilly and Company, F. Hoffmann-La Roche Ltd. (Basel, Switzerland), GlaxoSmithKline, Intarcia Therapeutics, Inc. (Hayward, CA), Lilly Deutschland GmbH, MannKind Corp. (Danbury, CT), MSD GmbH (München, Germany), MSD Corp. (Whitehouse Station, NJ), Novartis Pharma AG, Novo Nordisk A/S (Bagsvaerd, Denmark), Novo Nordisk Pharma GmbH, Sanofi Pharma (Bad Soden/Taunus, Germany), Takeda (Deerfield, IL), Versartis (Sunnyvale, CA), and Wyeth Research (Collegeville, PA), including reimbursement for travel expenses in connection with the above-mentioned activities. He owns no stock and is employed by Diabeteszentrum Bad Lauterberg. R.S., M.R., and D.d.V. are employees of Eli Lilly and Company. J.K.M., S.C., and B.J.H. are employees and shareholders of Eli Lilly and Company. L.M. is an employee of Bristol-Myers Squibb. B.H.R.W. has received grant support for clinical studies and also consulting fees for serving on advisory boards and as a speaker for Eli Lilly and Company, GlaxoSmithKline, Novo Nordisk, and Pfizer. He has also received consulting fees from Eli Lilly and Company as a member of the 4B Study and the DURABLE Trial Data Monitoring Committee. No other potential conflicts of interest relevant to this article were reported.

The sponsor took part in study design, safety monitoring, data collection, data analysis, data interpretation, and writing of the report.

Author Contributions. M.D. and B.H.R.W. were principal investigators of the study and contributed to the design and conduct of the study, researching data, interpreting and discussing results, writing the manuscript, and critically revising all subsequent versions. M.A.N. contributed to researching data, interpreting and discussing results, drafting parts of the manuscript, and revising all subsequent versions. R.S. was the lead physician of the study, wrote the protocol, and contributed to the design and conduct of the study, researching data, interpreting and discussing results, writing the manuscript, and revising all subsequent versions. J.K.M., B.J.H., and L.M. contributed to the design and conduct of the study, researching data, interpreting and discussing results, writing the manuscript, and revising all subsequent versions. S.C. was involved in the study design, conduct, data analysis, data interpretation, manuscript writing, and revising all subsequent versions. M.R. was involved in the study design, conduct, data interpretation, manuscript writing, and revising all subsequent versions. D.d.V. was clinical trial manager for the study and was involved in the study design, conduct, data interpretation, results discussions and manuscript writing, and revising all subsequent versions. B.H.R.W. agreed to submit the final version of the manuscript for publication. All authors had full access to the data and responsibility for the content of the report. All authors had final responsibility for the decision to submit. B.H.R.W. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. This study has been published previously in abstract form in the following publications: 1) Diamant M, Nauck M, Shaginian R, et al. Exenatide BID vs insulin lispro TID added to titrated insulin glargine QD in metformin-treated T2DM patients resulted in similar glycemic control but weight loss and less hypoglycemia: the 4B study. Diabetes 2013;62(Suppl. 1):A17; 2) Wolffenbuttel B, Nauck M, Shaginian R, et al. Glycaemic control and hypoglycaemia in metformin-treated T2DM patients with exenatide BID vs insulin lispro TID added to titrated insulin glargine QD: the 4B trial. Diabetologia 2013;56(Suppl. 1):S7; 3) Diamant M, Nauck M, Shaginian R, et al. Impact on cardiovascular risk factors of exenatide BID vs insulin lispro TIDM added to titrated insulin glargine QD in metformin-treated T2DM patients: the 4B trial. Diabetologia 2013;56(Suppl. 1):S393.

1.
Inzucchi
SE
,
Bergenstal
RM
,
Buse
JB
, et al
American Diabetes Association (ADA)
European Association for the Study of Diabetes (EASD)
.
Management of hyperglycemia in type 2 diabetes: a patient-centered approach: position statement of the American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD)
.
Diabetes Care
2012
;
35
:
1364
1379
[PubMed]
2.
Turner
RC
,
Cull
CA
,
Frighi
V
,
Holman
RR
UK Prospective Diabetes Study (UKPDS) Group
.
Glycemic control with diet, sulfonylurea, metformin, or insulin in patients with type 2 diabetes mellitus: progressive requirement for multiple therapies (UKPDS 49)
.
JAMA
1999
;
281
:
2005
2012
[PubMed]
3.
Buse
JB
,
Bergenstal
RM
,
Glass
LC
, et al
.
Use of twice-daily exenatide in basal insulin-treated patients with type 2 diabetes: a randomized, controlled trial
.
Ann Intern Med
2011
;
154
:
103
112
[PubMed]
4.
Balena
R
,
Hensley
IE
,
Miller
S
,
Barnett
AH
.
Combination therapy with GLP-1 receptor agonists and basal insulin: a systematic review of the literature
.
Diabetes Obes Metab
2013
;
15
:
485
502
[PubMed]
5.
Nathan
DM
.
Time for clinically relevant comparative effectiveness studies in type 2 diabetes
.
Ann Intern Med
2011
;
154
:
131
132
[PubMed]
6.
Yki-Järvinen
H
,
Juurinen
L
,
Alvarsson
M
, et al
.
Initiate Insulin by Aggressive Titration and Education (INITIATE): a randomized study to compare initiation of insulin combination therapy in type 2 diabetic patients individually and in groups
.
Diabetes Care
2007
;
30
:
1364
1369
[PubMed]
7.
Rosenstock
J
,
Ahmann
AJ
,
Colon
G
,
Scism-Bacon
J
,
Jiang
H
,
Martin
S
.
Advancing insulin therapy in type 2 diabetes previously treated with glargine plus oral agents: prandial premixed (insulin lispro protamine suspension/lispro) versus basal/bolus (glargine/lispro) therapy
.
Diabetes Care
2008
;
31
:
20
25
[PubMed]
8.
Miser
WF
,
Arakaki
R
,
Jiang
H
,
Scism-Bacon
J
,
Anderson
PW
,
Fahrbach
JL
.
Randomized, open-label, parallel-group evaluations of basal-bolus therapy versus insulin lispro premixed therapy in patients with type 2 diabetes mellitus failing to achieve control with starter insulin treatment and continuing oral antihyperglycemic drugs: a noninferiority intensification substudy of the DURABLE trial
.
Clin Ther
2010
;
32
:
896
908
[PubMed]
9.
Kolotkin
RL
,
Crosby
RD
.
Psychometric evaluation of the Impact of Weight on Quality of Life-Lite questionnaire (IWQOL-Lite) in a community sample
.
Qual Life Res
2002
;
11
:
157
171
[PubMed]
10.
Bradley C. The Diabetes Treatment Satisfaction Questionnaire. In Handbook of Psychology and Diabetes: A Guide to Psychological Measurement in Diabetes Research and Practice. Bradley C, Ed. Chur, Switzerland, Harwood Academic Publishers, 1994, p. 111–132
11.
Bradley
C
.
Diabetes Treatment Satisfaction Questionnaire. Change version for use alongside status version provides appropriate solution where ceiling effects occur
.
Diabetes Care
1999
;
22
:
530
532
[PubMed]
12.
Fineman
MS
,
Mace
KF
,
Diamant
M
, et al
.
Clinical relevance of anti-exenatide antibodies: safety, efficacy and cross-reactivity with long-term treatment
.
Diabetes Obes Metab
2012
;
14
:
546
554
[PubMed]
13.
Seino
Y
,
Min
KW
,
Niemoeller
E
,
Takami
A
EFC10887 GETGOAL-L Asia Study Investigators
.
Randomized, double-blind, placebo-controlled trial of the once-daily GLP-1 receptor agonist lixisenatide in Asian patients with type 2 diabetes insufficiently controlled on basal insulin with or without a sulfonylurea (GetGoal-L-Asia)
.
Diabetes Obes Metab
2012
;
14
:
910
917
[PubMed]
14.
Riddle
MC
,
Rosenstock
J
,
Gerich
J
Insulin Glargine 4002 Study Investigators
.
The treat-to-target trial: randomized addition of glargine or human NPH insulin to oral therapy of type 2 diabetic patients
.
Diabetes Care
2003
;
26
:
3080
3086
[PubMed]
15.
Riddle
MC
,
Aronson
R
,
Home
P
, et al
.
Adding once-daily lixisenatide for type 2 diabetes inadequately controlled by established basal insulin: a 24-week, randomized, placebo-controlled comparison (GetGoal-L)
.
Diabetes Care
2013
;
36
:
2489
2496
[PubMed]
16.
Gerich J. Abnormal glucagon secretion in type 2 (noninsulin-dependent) diabetes mellitus: causes and consequences. In Diabetes Mellitus: Pathophysiology and Therapy. Creutzfeldt W, Lefebvre P, Eds. Berlin, Heidelberg, SpringerVerlag, 1989, p. 127–133
17.
Roden
M
,
Perseghin
G
,
Petersen
KF
, et al
.
The roles of insulin and glucagon in the regulation of hepatic glycogen synthesis and turnover in humans
.
J Clin Invest
1996
;
97
:
642
648
[PubMed]
18.
Paquot
N
,
Roulin
D
,
Schneiter
P
, et al
.
Effects of regular insulin or insulin LISPRO on glucose metabolism after an oral glucose load in patients with type 2 diabetes mellitus
.
Diabetes Metab
1998
;
24
:
523
528
[PubMed]
19.
Drucker
DJ
,
Nauck
MA
.
The incretin system: glucagon-like peptide-1 receptor agonists and dipeptidyl peptidase-4 inhibitors in type 2 diabetes
.
Lancet
2006
;
368
:
1696
1705
[PubMed]
20.
Drucker
DJ
,
Buse
JB
,
Taylor
K
, et al
DURATION-1 Study Group
.
Exenatide once weekly versus twice daily for the treatment of type 2 diabetes: a randomised, open-label, non-inferiority study
.
Lancet
2008
;
372
:
1240
1250
[PubMed]
21.
DeFronzo
RA
,
Ratner
RE
,
Han
J
,
Kim
DD
,
Fineman
MS
,
Baron
AD
.
Effects of exenatide (exendin-4) on glycemic control and weight over 30 weeks in metformin-treated patients with type 2 diabetes
.
Diabetes Care
2005
;
28
:
1092
1100
[PubMed]
22.
Pratley
RE
,
Nauck
M
,
Bailey
T
, et al
1860-LIRA-DPP-4 Study Group
.
Liraglutide versus sitagliptin for patients with type 2 diabetes who did not have adequate glycaemic control with metformin: a 26-week, randomised, parallel-group, open-label trial
.
Lancet
2010
;
375
:
1447
1456
[PubMed]
23.
Nauck
MA
,
Holst
JJ
,
Willms
B
,
Schmiegel
W
.
Glucagon-like peptide 1 (GLP-1) as a new therapeutic approach for type 2-diabetes
.
Exp Clin Endocrinol Diabetes
1997
;
105
:
187
195
[PubMed]
24.
Linnebjerg
H
,
Park
S
,
Kothare
PA
, et al
.
Effect of exenatide on gastric emptying and relationship to postprandial glycemia in type 2 diabetes
.
Regul Pept
2008
;
151
:
123
129
[PubMed]
25.
Fineman
MS
,
Cirincione
BB
,
Maggs
D
,
Diamant
M
.
GLP-1 based therapies: differential effects on fasting and postprandial glucose
.
Diabetes Obes Metab
2012
;
14
:
675
688
[PubMed]
26.
Holman
RR
,
Thorne
KI
,
Farmer
AJ
, et al
4-T Study Group
.
Addition of biphasic, prandial, or basal insulin to oral therapy in type 2 diabetes
.
N Engl J Med
2007
;
357
:
1716
1730
[PubMed]
27.
Holman
RR
,
Farmer
AJ
,
Davies
MJ
, et al
4-T Study Group
.
Three-year efficacy of complex insulin regimens in type 2 diabetes
.
N Engl J Med
2009
;
361
:
1736
1747
[PubMed]
28.
Rodbard
HW
,
Jellinger
PS
,
Davidson
JA
, et al
.
Statement by an American Association of Clinical Endocrinologists/American College of Endocrinology consensus panel on type 2 diabetes mellitus: an algorithm for glycemic control
.
Endocr Pract
2009
;
15
:
540
559
[PubMed]
29.
Davies
M
,
Speight
J
.
Patient-reported outcomes in trials of incretin-based therapies in patients with type 2 diabetes mellitus
.
Diabetes Obes Metab
2012
;
14
:
882
892
[PubMed]
30.
Lind
M
,
Jendle
J
,
Torffvit
O
,
Lager
I
.
Glucagon-like peptide 1 (GLP-1) analogue combined with insulin reduces HbA1c and weight with low risk of hypoglycemia and high treatment satisfaction
.
Prim Care Diabetes
2012
;
6
:
41
46
[PubMed]
31.
LaSalle
JR
 Jr
,
Berria
R
.
Insulin therapy in type 2 diabetes mellitus: a practical approach for primary care physicians and other health care professionals
.
J Am Osteopath Assoc
2013
;
113
:
152
162
[PubMed]
32.
Abrahamson
MJ
,
Peters
A
.
Intensification of insulin therapy in patients with type 2 diabetes mellitus: an algorithm for basal-bolus therapy
.
Ann Med
2012
;
44
:
836
846
[PubMed]
33.
Lee
MS
,
Hsu
CC
,
Wahlqvist
ML
,
Tsai
HN
,
Chang
YH
,
Huang
YC
.
Type 2 diabetes increases and metformin reduces total, colorectal, liver and pancreatic cancer incidences in Taiwanese: a representative population prospective cohort study of 800,000 individuals
.
BMC Cancer
2011
;
11
:
20
[PubMed]
34.
Nauck
MA
,
Friedrich
N
.
Do GLP-1-based therapies increase cancer risk
?
Diabetes Care
2013
;
36
(
Suppl. 2
):
S245
S252
[PubMed]
35.
Egan
AG
,
Blind
E
,
Dunder
K
, et al
.
Pancreatic safety of incretin-based drugs—FDA and EMA assessment
.
N Engl J Med
2014
;
370
:
794
797
[PubMed]

Supplementary data