Evidence review

This evidence summary updates and replaces an earlier evidence summary on lixisenatide, which was published in January 2013 (ESNM10). That summary was based on the 2 randomised controlled trials (RCTs) of lixisenatide that had been published when lixisenatide was marketed: GetGoal-Mono (Fonseca et al. 2012) and GetGoal-L-Asia (Seino et al. 2012). However, these 2 RCTs have not been prioritised for inclusion in this evidence summary. GetGoal-Mono investigated lixisenatide as monotherapy, which does not reflect the UK licensed indication. GetGoal-L-Asia investigated lixisenatide in combination with basal insulin with or without a sulfonylurea, but was conducted in an Asian population, which does not reflect the majority of the UK population. The GetGoal-L-Asia authors noted that incretin-based therapies, such as glucagon-like peptide-1 (GLP-1) mimetics, are particularly effective in people of Asian or Japanese family origin because of the underlying pathophysiology of diabetes in these groups of people.

This evidence summary is based on 3 RCTs of lixisenatide in people with type 2 diabetes, which have been published in full since January 2013:

  • GetGoal-P (Pinget et al. 2013) compared lixisenatide with placebo in people whose diabetes was poorly controlled on pioglitazone, with or without metformin.

  • GetGoal-X (Rosenstock et al. 2013) compared lixisenatide with exenatide in people on metformin monotherapy.

  • GetGoal-L (Riddle et al. 2013a) compared lixisenatide with placebo in people whose diabetes was poorly controlled on basal insulin, with or without metformin.

In addition, 2 other RCTs are briefly discussed: GetGoal-M (Ahrén et al. 2013), which compared lixisenatide with placebo in people whose diabetes was poorly controlled on metformin monotherapy; and GetGoal-Duo1 (Riddle et al. 2013b), which compared lixisenatide with placebo in people whose diabetes was poorly controlled on recently initiated insulin glargine, plus oral agents.

Two other placebo-controlled RCTs in the GetGoal programme have been completed but have not yet been published in full: GetGoal-F1 (lixisenatide in combination with metformin) and GetGoal-S (lixisenatide in combination with a sulfonylurea, with or without metformin). In accordance with the integrated process statement for evidence summaries: new medicines, they have not been included in this evidence summary.

In GetGoal-P, GetGoal-X and GetGoal-L, HbA1c results were reported only in percentages and these have not been converted to mmol/mol in this evidence summary. The Diabetes UK website indicates that a 0.5 percentage point difference in HbA1c is equivalent to a difference of about 5.5 mmol/mol, and a 1 percentage point difference is equivalent to a difference of about 11 mmol/mol. Note that these are rounded equivalents.

GetGoal-P (Pinget et al. 2013)

  • Design: a double-blind RCT conducted in 150 centres in 13 countries in Europe, India and North and South America. The study included a 24-week main study and a variable double-blind extension period that continued until the last patient had completed a total of 76 weeks' treatment. Allocation was concealed.

  • Population: 484 adults, mean age 56 years, with type 2 diabetes of at least 1 year's duration (mean 8.1 years), HbA1c 7.0% to 10.0% (mean 8.1%) and fasting plasma glucose (FPG) 13.9 mmol/l or less (mean 9.1 mmol/l), who had been treated with a stable dose of pioglitazone 30 mg/day or more (median dose 30 mg) for at least 3 months (mean 1.8 years). In addition, 81% of patients were treated with metformin 1.5 g/day or more (median dose 2 g/day); the median duration of metformin treatment was 3.4 years. At trial entry, 68% of patients had a body mass index (BMI) of 30 kg/m2 or greater (median 34 kg/m2).

  • Intervention and comparison: patients were randomised 2:1 to lixisenatide or placebo, administered no longer than 1 hour before breakfast. Patients randomised to lixisenatide received 10 micrograms once daily for 1 week, then 15 micrograms once daily for 1 week, then 20 micrograms once daily, if tolerated. If the target dose of lixisenatide could not be tolerated, the dose could be reduced. Patients continued on their established doses of pioglitazone and metformin.

  • Outcome: the primary efficacy outcome was the absolute change in HbA1c from baseline to week 24 for the modified intention-to-treat population (all randomised patients who received at least 1 dose of study treatment and had a baseline assessment and at least 1 post-baseline assessment). This was analysed using analysis of covariance (ANCOVA), stratified by screening level HbA1c and whether or not the person was taking metformin. The last observation carried forward (LOCF) method was used to handle missing data by taking the last available HbA1c as the HbA1c value at week 24: 35 patients randomised to lixisenatide (10.8%) and 24 patients randomised to placebo (14.9%) discontinued treatment prematurely. In addition, 74% of patients randomised to lixisenatide and 68% of those randomised to placebo completed the extension period. Secondary efficacy outcomes included the percentage of patients with an HbA1c of less than 7.0% and changes in FPG and body weight from baseline.

Table 1 Summary of GetGoal-P (Pinget et al. 2013)

Placebo

Lixisenatide

Analysis (lixisenatide versus placebo)

Efficacy a

n=148

n=308

Primary outcome: LS mean change in HbA1c from baseline to week 24b

−0.34% points from baseline of 8.1%

−0.9% points from baseline of 8.1%

LS mean change difference −0.56% points, 95% CI −0.73 to −0.39, p<0.0001

Selected secondary outcomes:

Patients with HbA1c of less than 7.0% at week 24b

26.4%

52.3%

p<0.0001

LS mean change in FPG from baseline to week 24b

−0.32 mmol/l from baseline of 9.1 mmol/l

−1.16 mmol/l from baseline of 9.1 mmol/l

LS mean change difference −0.84 mmol/l, 95% CI −1.21 to −0.47, p<0.0001

LS mean change in body weight from baseline to week 24b

+0.2 kg from baseline of 96.7 kg

−0.2 kg from baseline of 92.9 kg

LS mean change difference −0.41 kg, 95% CI −1.03 to 0.20, p=0.19 (not statistically significant)

Safety c

n=161

n=323

Serious adverse events

1.9% (3/161)

2.5% (8/323)

p value not stated

Death

0.6% (1/161)

0

p value not stated

Adverse events leading to discontinuation

5.0% (8/161)

6.5% (21/323)

p value not stated

Gastrointestinal disorders

28.6% (46/161)

36.5% (118/323)

p value not stated

Symptomatic hypoglycaemiad

1.2% (2/161)

3.4% (11/323)

p value not stated

Severe symptomatic hypoglycaemiae

0

0

Abbreviations: CI, confidence interval; FPG, fasting plasma glucose; LS, least square.

a Modified intention-to-treat population: all patients who received at least 1 dose of study treatment and had baseline assessment and at least 1 post-baseline assessment of any primary or secondary efficacy variable.

b ANCOVA analysis with last observation carried forward.

c All patients who received at least 1 dose of study treatment: results to week 24.

d Symptoms of hypoglycaemia with blood glucose less than 3.3 mmol/l and/or prompt recovery with carbohydrate.

e Symptoms of hypoglycaemia requiring the assistance of another person, with blood glucose less than 2.0 mmol/l or prompt recovery with carbohydrate.

GetGoal-X (Rosenstock et al. 2013)

  • Design: a 24-week, open-label, non-inferiority study conducted in 122 centres in 18 countries in Europe and North and South America. Allocation was concealed.

  • Population: 639 adults, aged 21 to 84 years (mean 57 years), with type 2 diabetes (mean duration 6.8 years), HbA1c 7.0% to 10.0% (mean 8.02%) and FPG 13.9 mmol/l or less (mean 9.7 mmol/l), who had been receiving metformin 1.5 g per day or more (mean daily dose 2 g). Mean BMI was 33.6 kg/m2. Patients had received no glucose-lowering agents other than metformin in the preceding 3 months.

  • Intervention and comparison: patients were randomised to receive lixisenatide (10 micrograms once daily for 1 week, then 15 micrograms once daily for 1 week, then 20 micrograms once daily) or exenatide (5 micrograms twice daily for 4 weeks and then 10 micrograms twice daily). All treatments were administered no longer than 1 hour before breakfast (lixisenatide and exenatide) or the evening meal (exenatide).

  • Outcome: the primary efficacy outcome was the change in HbA1c from baseline to week 24 for the modified intention-to-treat population (all randomised patients who received at least 1 dose of study treatment and had a baseline assessment and at least 1 post-baseline assessment). The pre-specified non-inferiority criterion was 0.4% or less, as recommended by regulatory guidelines when the study was designed. This was analysed using ANCOVA, with the treatment group, screening strata for HbA1c and BMI, and country as fixed effects, and with baseline HbA1c as a covariate. The LOCF method was used to handle missing data: 41 patients randomised to lixisenatide (12.9%) and 45 patients randomised to exenatide (14.2%) discontinued treatment prematurely. Secondary efficacy outcomes included the percentage of patients with an HbA1c less than 7.0% and changes in FPG and body weight from baseline to week 24.

Table 2 Summary of GetGoal-X (Rosenstock et al. 2013)

Exenatide

Lixisenatide

Analysis (lixisenatide versus exenatide)

Efficacy a

n=315

n=315

Primary outcome: LS mean change in HbA1c from baseline to week 24b

−0.96% points from baseline of 7.96%

−0.79% points from baseline of 7.97%

LS mean change difference 0.17% points, 95% CI 0.033 to 0.297

Selected secondary outcomes:

Patients with HbA1c of less than 7.0% at week 24b

49.8%

48.5%

p value not stated

LS mean change in FPG from baseline to week 24b

−1.45 mmol/l from baseline of 9.7 mmol/l

−1.22 mmol/l from baseline of 9.7 mmol/l

LS mean change difference 0.23 mmol/l, 95% CI −0.052 to 0.522

LS mean change in body weight from baseline to week 24b

−3.98 kg from baseline of 96.7 kg

−2.96 kg from baseline of 94.5 kg

LS mean change difference 1.02 kg, 95% CI 0.456 to 1.581

Safety c

n=316

n=318

Serious adverse events

2.2% (7/316)

2.8% (9/318)

p value not stated

Death

0.3% (1/316)

0.3% (1/318)

p value not stated

Adverse events leading to discontinuation

13.0% (41/316)

10.4% (33/318)

p value not stated

Gastrointestinal disorders

50.6% (160/316)

43.1% (137/318)

p value not stated

Symptomatic hypoglycaemiad

7.9% (25/316, 48 events)

2.5% (8/318, 8 events)

p<0.05

Severe symptomatic hypoglycaemiae

0

0

Abbreviations: CI, confidence interval; FPG, fasting plasma glucose; LS, least square.

a Modified intention-to-treat population: all patients who received at least 1 dose of study treatment and had baseline assessment and at least 1 post-baseline assessment of any primary or secondary efficacy variable.

b ANCOVA analysis with last observation carried forward.

c All patients who received at least 1 dose of study treatment.

d Symptoms of hypoglycaemia with blood glucose less than 3.3 mmol/l and/or prompt recovery with carbohydrate.

e Symptoms of hypoglycaemia requiring the assistance of another person, with blood glucose less than 2.0 mmol/l or prompt recovery with carbohydrate.

GetGoal-L (Riddle et al. 2013a)

  • Design: a 24-week, double-blind RCT conducted in 111 centres in 15 countries in Europe, North and South America, Egypt, India and South Korea. Allocation was concealed.

  • Population: 495 adults, aged 29 to 81 years (mean 57 years), with type 2 diabetes of at least 1 year's duration (mean 12.5 years), HbA1c 7.0% to 10.0% (mean 8.4%) and FPG 13.9 mmol/l or less (mean 8.1 mmol/l), who had been treated with a stable dose of basal insulin (at least 30 units per day, median 55 units) for at least 2 months (mean 3.1 years): 50% of patients were receiving insulin glargine, 40% were receiving NPH insulin and the remainder were receiving insulin detemir. At trial entry, 79% of patients were treated with metformin 1.5 g/day or more (mean dose 2 g/day). At trial entry, 60% of patients had a BMI of 30 kg/m2 or greater (mean 32 kg/m2).

  • Intervention and comparison: patients were randomised 2:1 to lixisenatide or placebo, administered no longer than 1 hour before breakfast. Patients randomised to lixisenatide received 10 micrograms once daily for 1 week, then 15 micrograms once daily for 1 week, then 20 micrograms once daily, if tolerated. Patients continued on their established dose of metformin, if taking it at study entry. The basal insulin dose was to remain stable (±20%) throughout the study, but if the patient's HbA1c was 7.5% or less at pre-entry screening, the dose was reduced by 20% and then progressively increased back to the previous dose between weeks 4 and 12. After week 12, no dose adjustments were permitted except reductions in response to hypoglycaemia.

  • Outcome: the primary efficacy outcome was the change in HbA1c from baseline to week 24 for the modified intention-to-treat population (all randomised patients who received at least 1 dose of study treatment and had a baseline assessment and at least 1 post-baseline assessment). This was analysed using ANCOVA, with the treatment group, screening strata for HbA1c and metformin use, and country as fixed effects, and with baseline HbA1c as a covariate. The LOCF method was used to handle missing data: 53 patients randomised to lixisenatide (16.1%) and 20 patients randomised to placebo (12.0%) discontinued treatment prematurely. Secondary efficacy outcomes included the percentage of patients with an HbA1c of less than 7.0% and changes in FPG and body weight from baseline to week 24.

Table 3 Summary of GetGoal-L (Riddle et al. 2013a)

Placebo

Lixisenatide

Analysis (lixisenatide versus placebo)

Efficacy a

n=166

n=327

Primary outcome: LS mean change in HbA1c from baseline to week 24b

−0.4% points from baseline of 8.4%

−0.7% points from baseline of 8.4%

LS mean change difference −0.4% points, 95% CI −0.6 to −0.2, p<0.0002

Selected secondary outcomes:

Patients with HbA1c of less than 7.0% at week 24b

12.0%

28.3%

p<0.0001

LS mean change in FPG from baseline to week 24b

−0.6 mmol/l from baseline of 8.0 mmol/l

−0.6 mmol/l from baseline of 8.1 mmol/l

LS mean change difference −0.1 mmol/l, 95% CI −0.6 to 0.4, p=0.76 (not statistically significant)

LS mean change in PPG from baseline to week 24b

−1.7 mmol/l from baseline of 15.9 mmol/l

−5.5 mmol/l from baseline of 16.4 mmol/l

LS mean change difference −3.8 mmol/l, 95% CI −4.7 to −2.9, p<0.0001

LS mean change in body weight from baseline to week 24b

−0.5 kg from baseline of 89 kg

−1.8 kg from baseline of 87 kg

LS mean change difference −1.3 kg, 95% CI −1.8 to −0.7, p<0.0001

Safety d

n=167

n=328

Serious adverse events

4.2% (7/167)

3.7% (12/328)

p value not stated

Death

0

0.3% (1/328)

p value not stated

Adverse events leading to discontinuation

4.8% (8/167)

7.6% (25/328)

p value not stated

Gastrointestinal disorders

20.4% (34/167)

40.2% (132/328)

p value not stated

Symptomatic hypoglycaemiae

21.6% (36/167, 132 events)

27.7% (91/328, 309 events)

p value not stated

Severe symptomatic hypoglycaemiaf

0

1.2% (4/328, 4 events)

p value not stated

Abbreviations: CI, confidence interval; FPG, fasting plasma glucose; LS, least square; PPG, post-prandial plasma glucose.

a Modified intention-to-treat population: all patients who received at least 1 dose of study treatment and had baseline assessment and at least 1 post-baseline assessment of any primary or secondary efficacy variable.

b ANCOVA analysis with last observation carried forward.

c 2-hour PPG levels after a standardised liquid breakfast meal.

d All patients who received at least 1 dose of study treatment.

e Symptoms of hypoglycaemia with blood glucose less than 3.3 mmol/l or prompt recovery with carbohydrate.

f Symptoms of hypoglycaemia requiring the assistance of another person, with blood glucose less than 2.0 mmol/l or prompt recovery with carbohydrate.

Clinical effectiveness

Lixisenatide in combination with oral therapy

GetGoal-P found that lixisenatide was statistically significantly superior to placebo after 24 weeks of treatment in terms of HbA1c reduction from baseline in patients treated with pioglitazone. Most but not all patients were also taking metformin, and the ANCOVA method of analysis would have ensured that the results were adjusted for this variable. However, the use of the LOCF method for handling missing data is a potential source of bias (see Evidence strengths and limitations for more information).

Lixisenatide was also compared with placebo in GetGoal-M (Ahrén et al. 2013). This had a very similar design and method of analysis to GetGoal-P, except that it also compared morning and evening administration of lixisenatide. It included 680 people with HbA1c 7% to 10% who were taking metformin monotherapy (mean daily dose, 2 g). GetGoal-M found a similar least squares mean reduction in HbA1c from baseline after 24 weeks in the lixisenatide group as in GetGoal-P. Administration of lixisenatide in the morning produced a reduction of 0.9 percentage points (9.8 mmol/mol), a 0.5 percentage point (5.5 mmol/mol) greater reduction than placebo (p<0.0001). Administration of lixisenatide in the evening produced a reduction of 0.8 percentage points (8.7 mmol/mol), a 0.4 percentage point (4.4 mmol/mol) greater reduction than placebo (p<0.0001). GetGoal-M and GetGoal-P both found a statistically significant reduction in fasting plasma glucose with lixisenatide compared with placebo. GetGoal-M found a reduction from baseline in 2-hour post-prandial plasma glucose in the lixisenatide morning group (least squares mean reduction 5.9 mmol/l, a 4.5 mmol/l greater reduction than with placebo, p<0.0001); this end point was not reported in GetGoal-P. There was no statistically significant effect from lixisenatide on body weight compared with placebo in either study.

GetGoal-X found that lixisenatide was non-inferior to exenatide for the primary outcome of reduction in HbA1c from baseline. However, in its public assessment report for lixisenatide (Lyxumia), the European Medicines Agency (EMA) concluded that non-inferiority to exenatide had not been shown robustly (see Evidence strengths and limitations for more information). Exenatide produced a statistically significantly greater mean reduction in bodyweight.

Lixisenatide in combination with insulin therapy

GetGoal-L found that lixisenatide was statistically significantly superior to placebo after 24 weeks of treatment in terms of HbA1c reduction from baseline in patients established on treatment with basal insulin. Most but not all patients were also taking metformin, and the ANCOVA method of analysis would have ensured that the results were adjusted for this variable. However, the use of the LOCF method for handling missing data is a potential source of bias (see Evidence strengths and limitations for more information).

Lixisenatide was also compared with placebo in conjunction with insulin in GetGoal-Duo1 (Riddle et al. 2013b). This double-blind, multi-national, multicentre RCT recruited 898 adults with type 2 diabetes and HbA1c 7% to 10% despite treatment with metformin alone or in combination with a sulfonylurea, glinide or glitazone. Patients were started on insulin glargine, the dose of which was titrated upwards over 12 weeks (patients continued on metformin and, if used previously, a glitazone). After this period, the 446 patients whose HbA1c remained in the range of 7% to 9% and whose fasting plasma glucose was 7.8 mmol/l or less were randomised to lixisenatide or placebo for 24 weeks, while adjustment of the insulin glargine dose continued. Lixisenatide was titrated up to 20 micrograms daily in a 2-step process, the same as in GetGoal-L.

Introduction of insulin glargine reduced mean HbA1c levels from 8.6% (70 mmol/mol) to 7.6% (60 mmol/mol) at 12 weeks. After 24 weeks of treatment with lixisenatide or placebo (or LOCF), the least squares mean reduction in HbA1c was 0.7 percentage points in the lixisenatide group and 0.4 percentage points in the placebo group (least squares mean difference −0.3 percentage points, 95% confidence interval −0.5 to −0.2, p<0.0001). More patients receiving oral therapy, insulin glargine and lixisenatide (56%) had an HbA1c less than 7.0% (53 mmol/mol) compared with those receiving oral therapy, insulin glargine and placebo (39%, p=0.0001).

Neither GetGoal-L nor GetGoal-Duo1 found a statistically significant reduction in fasting plasma glucose with lixisenatide compared with placebo, but both studies found a reduction in 2-hour post-prandial plasma glucose compared with placebo (in GetGoal-Duo 1 lixisenatide produced a 3.2 mmol/l greater reduction in least squares mean reduction from baseline compared with placebo, p<0.0001). In GetGoal-L, body weight reduced in both the lixisenatide and placebo groups, but to a statistically significantly greater extent in the lixisenatide group (difference −1.3 kg, p<0.0001). In GetGoal-Duo1, body weight increased in both study groups, although to a statistically significantly lesser extent in the lixisenatide group (+0.3 kg compared with +1.2 kg with placebo, p=0.0012).

Safety

Statistical analyses of most of the safety data from the studies included in this evidence summary were not reported, which limits the conclusions that can be drawn from them. In GetGoal-X, statistically significantly fewer patients using lixisenatide reported nausea compared with those using exenatide (24.5% compared with 35.1%, p<0.05). However, there was no statistically significant difference in the mean total patient assessment of upper gastrointestinal disorders – quality of life score. Statistically significantly fewer patients using lixisenatide reported symptomatic hypoglycaemia compared with those using exenatide (2.5% compared with 7.9%, p<0.05).

Gastrointestinal adverse effects are often reported with GLP-1 mimetics. The summary of product characteristics (SPC) states that the most frequently reported adverse reactions during clinical studies, in which more than 2600 people received lixisenatide, were nausea, vomiting and diarrhoea (very common: frequency 1 or more in 10). These reactions were mostly mild and transient. In addition, headache was also very common, as was hypoglycaemia when lixisenatide was used in combination with a sulfonylurea and/or a basal insulin. The SPC notes that GLP-1 mimetics have been associated with a risk of developing acute pancreatitis.

Evidence strengths and limitations

As with the other GLP-1 mimetics, there are no data from RCTs relating to patient-oriented outcomes, such as rates of macrovascular or microvascular events, or long-term safety data. The evidence of efficacy relates solely to surrogate end points, chiefly reductions in HbA1c. The clinical significance of these needs to be judged in the context of the wider evidence base relating to the management of type 2 diabetes, as discussed in the Introduction. In addition, in the studies discussed in this evidence summary the response to placebo on the primary outcome of reductions in HbA1c from baseline was large, which somewhat hampers assessment of lixisenatide's effects.

In all the clinical trials discussed in this evidence summary, allocation was concealed, therefore avoiding an important potential source of bias. In addition, the use of the ANCOVA method of analysis would have ensured that the results were adjusted for variables including concomitant treatment with specified oral hypoglycaemic drugs, baseline HbA1c and country of treatment. Most of the studies were double-blind, but GetGoal-X was open-label, which is a potential source of bias.

As the EMA notes in its guideline on missing data in confirmatory clinical trials, it is unrealistic to expect that all patients in any clinical trial will receive treatment with full compliance to the treatment schedule and with a complete follow-up as per protocol: some patients will drop out of the trial before the scheduled conclusion and among those who stay in, some will have data not recorded for some reason. The guideline states that there is no universally applicable method that adjusts the analysis to take these missing values into account, and different approaches may lead to different conclusions. All the studies discussed in this evidence summary were 24-week studies that used the last observation carried forward (LOCF) approach to take account of missing data. In this approach, regardless of when a patient left the trial (for example, after week 1, week 6 or week 23), the last available result for that patient was carried forward and analysed as though it were the result at the study end.

In the clinical trials discussed in this evidence summary, drop-out rates ranged from 8.6% to 16.1% among patients randomised to lixisenatide, and 5% to 14.9% among those randomised to placebo. The EMA's guideline on missing data in confirmatory clinical trials notes that people who do not complete a clinical trial may be more likely to have extreme values than those who do. Therefore, the loss of these 'non-completers' could lead to an underestimate of variability and therefore artificially narrow the confidence interval for the treatment effect. Because the choice of primary analysis will be based on assumptions that cannot be verified, the EMA guideline advises that it will almost always be necessary to investigate the robustness of trial results through appropriate sensitivity analyses that make different assumptions. Sensitivity analyses were not reported in any of the clinical trials discussed in this evidence summary.

The published analysis of the GetGoal-X non-inferiority study requires additional discussion. The analysis of the modified intention-to-treat (ITT) data set found that lixisenatide was statistically significantly less effective than, but non-inferior to, exenatide for the primary outcome of change in HbA1c from baseline; that is, the upper 95% confidence interval for the difference (0.297 percentage points) did not exceed the pre-specified non-inferiority criterion of 0.4 percentage points. The difference was also less than the criterion of 0.3 percentage points more recently recommended by the EMA in its guideline on clinical investigation of medicinal products in the treatment or prevention of diabetes mellitus. However, the EMA guidance on the points to consider on switching between superiority and non-inferiority advises that, although in a superiority study the ITT analysis is the analysis of choice, in a non-inferiority study the ITT and per-protocol analyses have equal value and their use should lead to similar conclusions for a robust interpretation of the results. The per-protocol analysis of GetGoal-X was not reported in the published account of GetGoal-X, but the EMA noted in its public assessment report for lixisenatide that that upper 95% confidence interval was 0.315 percentage points in the 'completer' population. This breached the EMA's currently required criterion. The EMA concluded that non-inferiority to exenatide had not been shown robustly and that applying the recommended non-inferiority margin of 0.3 percentage points indicated that the effect of lixisenatide may be inferior to exenatide.