The efficacy and safety of liraglutide added to metformin in patients with diabetes: a meta-analysis of randomized controlled trials

Liraglutide, a glucagon-like peptide (GLP-1) receptor agonist, has showed favorable effects in the glycaemic control and weight reduction in patients with type 2 diabetes mellitus (T2DM). The meta-analysis was to compare the efficacy and safety of liraglutide added to metformin with other treatments in patients with T2DM. A systematic literature search on PubMed, Embase, Web of Science and the Cochrane library databases were performed. Eligible studies were randomized controlled trials (RCTs) of patients with T2DM who received the combination treatment of liraglutide and metformin. Pooled estimates were performed using a fixed-effects model or random-effects model. A total of nine RCTs met the inclusion criteria. Compared with control (placebo, sitagliptin, glimepiride, dulaglutide, insulin glargine, and NPH), liraglutide in combination with metformin resulted in significant reductions in HbA1c, bodyweight, FPG, and PPG, and similar reductions in SBP, and DBP. Moreover, liraglutide combined with metformin did not increase the risk of hypoglycemia, but induced a higher incidence of gastrointestinal disorders. In conclusion, this meta-analysis confirmed the use of liraglutide as add-on to metformin appeared to be effective and safe for patients with T2DM. However, considering the potential limitations in this study, more large-scale, well-conducted RCTs are needed to identify our findings.

Inclusion criteria and study selection. All clinical trials assessing the efficacy and safety of liraglutide plus metformin in the treatment of T2DM were considered eligible for analysis. The predetermined study inclusion criteria were: (1) randomized controlled trials (RCTs); (2) adult patients had T2DM [HbA1c between either 6.5 or 7.0 and either 10.0 or 11.0%, depending on previous treatment]; (3) compared liraglutide in addition to metformin with another antidiabetic therapy or placebo; (4) reported the data on changes from baseline in HbA1c, bodyweight, fasting plasma glucose (FPG), postprandial plasma glucose(PPG), systolic blood pressure (SBP), and diastolic blood pressure (DBP); (5) treat patients more than 12 weeks after randomization.
Data extraction and quality assessment. Two independent investigators used a standardized tool to extract the following data from each study: first author's name, year of publication, country, number of study patients, baseline patient characteristics (age, sex, race, diabetes duration), mean changes from baseline in HbA1c, bodyweight, FPG, PPG, SBP, and DBP, and incidence of treatment-emergent adverse events (nausea, diarrhea, vomiting, dyspepsia, constipation, and nasopharyngitis).
We used the Jadad scale 22 to assess the methodological quality of the included studies. The Jadad scale consists of three items describing randomization (0-2 points), blinding (0-2 points), and dropouts and withdraws (0-1 point) to report the quality of a RCT 22 . A score of 1 point is given for each of the points described. A further point is obtained when the randomization and/or blinding is described and appropriate. The quality scale ranges from 0 to 5 points, and higher scale suggests better reporting. Any study with a score ≥ 3 is considered to be of high quality 22,23 . Statistical analysis. Data were analyzed using Stata version 12.0 (Stata Corporation, College Station, TX, USA). Before the data were synthesized, we first test the heterogeneity between the studies using Q chi-square test 24 , in which a P value < 0.10 was considered as significant heterogeneity. I 2 statistic was used to describe the percentage of the variability that attributed to heterogeneity across the studies rather than the chance. Studies with an I 2 statistic of < 25%, ~50%, ~75%, ~100% are considered to have no, low, moderate, and high degree of heterogeneity, respectively 25 . Pooled estimates were calculated using a fixed-effects model (Mantel-Haenszel method) 26 ; otherwise, a random-effects model (DerSimonian-Laird method) 27 was applied when significant heterogeneity among the included studies was found. If the heterogeneity was tested, subgroup analysis or sensitivity analysis was performed to explore the potential sources of heterogeneity.
Continuous variables, including mean changes from baseline in HbA1c, bodyweight, FPG, PPG, SBP, and DBP, were expressed as weight mean difference (WMD) with 95% confidence intervals (95%CIs); dichotomous variables, including the incidence of treatment-emerge adverse events, were expressed as relative risk (RR) with 95%CIs. The assessment of publication bias was evaluated by using Egger 28 and Begger 29 test. A P value less than 0.05 was judged as statistically significant, except where otherwise specified.

Results
Identification of eligible studies. The initial search yielded 916 relevant publications from PubMed, Web of Science, Embase, and the Cochrane library. Of these, 418 were excluded because of duplicate records, and 479 and 9 studies were removed after a review if title/abstract and full-text information, respectively (Fig. 1). Thus, 19 potential studies were identified for the final analysis. However, ten of them were excluded for the following reasons: six studies did not provide outcome of interest or available data [30][31][32][33][34][35] , two studies had liraglutide and metformin in both groups 36,37 , one study was a sing-arm design 38 , and one study 39 had overlap data with another trial 40 . Finally, nine RCTs 19,20,[40][41][42][43][44][45][46] met the inclusion criteria, and were included in this meta-analysis.
Characteristics of eligible studies and quality assessment. The main characteristics of the nine included RCTs are presented in Table 1. These studies were published between 2009 and 2015. The total number of included patients was 4,657, ranging from 63 to 1,091 patients per study. The clinical characteristics were well matched for age, sex distribution, duration of diabetes, BMI, waist circumference, HbA1c, FPG, SBP, and DBP in both groups at the beginning of each study. The study durations ranged from 12 to 52 weeks. All these studies Scientific RepoRts | 6:32714 | DOI: 10.1038/srep32714 provided information regarding the efficacy and safety of combination treatment of liraglutide and metformin in patients with T2DM.
There were 44 out of 1483 (2.97%) patients in metformin plus liraglutide group and 32 out of 397 (8.06%) patients in the control group (placebo, glimepiride, and NPH) that experienced hypoglycemia. Pooled results showed that, compared with control, metformin plus liraglutide did not increase the risk of hypoglycemia (RR = 0.33, 95%CI: 0.08, 1.44; P = 0.140). No cases of pancreatitis were reported in these included studies.

Publication bias. Assessment of publication bias was conducted by using Egger's and Begg test, and results
showed that no publication bias existed among the included studies (Egger's test: t = − 1.33, P = 0.201; Begg test: Z = 0.14, P = 0.887).

Discussion
The addition of GLP-1 receptor agonist is recommended by the ADA and the EASD as a therapeutic option for patients with T2DM whose HbA1c targets are not met or maintained with lifestyle modifications, with consideration of individual patient-related factors 47 . This meta-analysis investigated the efficacy and safety of liraglutide in combination with metformin, compared to other therapies for patients with T2DM. Overall, the results of our study suggest that compared with other therapies, liraglutide in combination with metformin showed greater reduction in terms of HbA1c levels, body weight, FPG, and PPG, and similar changes in SBP and DBP. In addition, when used as add-on therapy to metformin, liraglutide did not increase the risk of hypoglycemia, but induced a higher incidence of gastrointestinal disorders.
To the bests of our knowledge, this is the first comprehensive meta-analysis to compare the efficacy and safety of liraglutide add-on to metformin with other therapies in patients with T2DM. Our results suggest that, compared with control, combined therapy of liraglutide and metformin reduced the HbA1c significantly by − 0.36%. This result is consistent across the subgroup analysis based on dosage of liraglutide, in which 1. liraglutide decreased the HbA1c by − 0.47%, and − 0.35%, respectively. Furthermore, subgroup analysis based on comparators demonstrated that, liraglutide in combination with metformin was associated with a greater reduction of HbA1c than placebo, sitagliptin, insulin glargin, or dulaglutide, and similar change with glimepiride. The HbA1c reduction from baseline with the combination therapy of liraglutide and metformin in this meta-analysis was in line with a previously published study. In the trial conducted by Pratley RE, et al. 39 , 1.2 mg and 1.8 mg liraglutide decreased the HbA1c from baseline by − 1.24% and − 1.50%, respectively, whereas sitagliptin lowered the HbA1c by − 0.90%. The estimated treatment differences (ETD) for 1.2 mg, 1.8 mg liraglutide versus sitagliptin were − 0.34% (95%CI: − 0.51, − 0.61) and − 0.60% (95%CI: − 0.77, − 0.43), respectively 39 , which indicated that combination therapy of liraglutide and metformin resulted a more reduction of HbA1c than sitagliptin. The greater HbA1c reduction of combination therapy versus sitagliptin probably can be explained by the pharmacological concentrations of free liraglutide, whereas physiological concentrations of GLP-1 and glucose dependent insulinotropic polypeptide (GIP) are achieved with sitagliptin. Although the active GLP-1 concentrations are increased by two or three times with dipeptidyl peptidase-4 (DPP-4) inhibitors 48 , the stimulation of GLP-1 receptor activity by liraglutideis estimated to be several times higher than with DPP-4 inhibitors 14 .
Moreover, liraglutide has a long half-life (about 13 h), which may be another reason for the increased efficacy 49 . Despite sitagliptin has a similar pharmacokinetic half-life as liraglutide 50 , its effect on the increase of endogenous GLP-1 concentration occurs mainly after meals. Thus, the fasting concentrations of active GLP-1 remain considerably low overnight, so the FPG concentrations with sitagliptin reduced less than liraglutide 50 .
In this meta-analysis, liraglutide used as an add-on to metformin, reduced bodyweight more than other therapies did. This result can be attributed to the increased stimulation of GLP-1 receptor by liraglutide. T2DM increases morbidity and mortality mainly because of the cardiovascular and cerebrovascular disease. Obesity is one of the specific risk factors for the development of diabetes, and cardiovascular disease 51 . Thus, antidiabetic drugs should have favorable effects in the treatment of cardiovascular and cerebrovascular disease, as well as reduce the blood glucose. Among the controls (placebo, dulaglutide, insulin glargine, NPH, glimepiride, and sitagliptin) of the included studies, weight gain was observed in patients treated with insulin glargine (1.6 kg increase), and glimepiride (0.25 kg increase). However, when liraglutide is added to metformin, it significantly reduced the bodyweight (by 2.13 kg) compared with control. This result is in contrast to the data from a phase 3 trail, in which 1.8 mg once-daily liraglutide monotherapy was compared with 10 μ g twice-daily exenatide 16 . And liraglutide did not show a greater reduction in bodyweight than exenatide (liraglutide: − 3.24 kg VS exenatide: − 2.87 kg; ETD, − 0.38 kg, 95%: − 0.99, 0.23, P = 0.2235) 16 .
With regard to the FPG, our results showed that combined therapy of liraglutide and metformin was associated with a significant decrease on this clinical outcome compared with control (placebo, sitagliptin, glimepiride, dulaglutide, insulin glargine, and NPH). However, this significant FPG reduction was not observed in the subgroup analysis between liraglutide and glimepiride. Compared with glimepiride, 0.6 mg,1.2 mg, and 1.8 mg liraglutide reduced the FPG by 0.20 mmol/L, − 0.10 mmol/L, and − 0.22 mmol/L, respectively, though the difference was not significant. These results are in accordance with the findings from the LEAD-2 trial 19 . In that trial, liraglutide was found to have similar efficacy in FPG reduction compared to glimepiride. The decrease in FPG from baseline was 1.1 mmol/L, − 1.6 mmol/L, and − 1.7 mmol/L for the 0.6 mg, 1.2 mg, and 1.8 mgliraglutide, respectively, whereas the corresponding value for glimepiride was − 1.3 mmol/L 19 .
For the PPG, we found the pooled results remained confusing. Our finding showed that liraglutide in combination with metformin was associated with a significantly greater reduction in PPG than control (placebo, glimepiride, dulaglutide, insulin glargine, and NPH) (− 0.60 mmol/L). However, when it was compared with active comparators, no significant difference was found between them. Compared with glimepiride, 0.6 mg, 1.2 mg, and 1.8 mgliraglutide reduced PPG by 0.47 mmol/L, − 0.11 mmol/L, and − 0.49 mmol/L, respectively; however, the difference between them was not significant. These findings are in line with what has been observed in the LEAD trial 19 . In the LEAD trial, it has been found that the decreases from baseline in PPG were −1.7 mmol/L, − 2.3 mmol/L, and − 2.6 mmol/L for the 0.6 mg, 1.2 mg, and 1.8 mg liraglutide respectively, and − 2.5 mmol/L for glimepiride. The corresponding ETD for 0.6 mg, 1.2 mg, and 1.8 mg liraglutide versus glimepiride was 0.8 mmol/L, 0.2 mmol/L, and − 0.1 mmol/L, respectively. Similar results were also found in the comparison between liraglutide and dulaglutide 41 , insulin glargine 20 , or NPH 42 , which showed a comparable effect in PPG between them. We speculated that due to the strong response in the placebo group, liraglutide showed beneficial effects in PPG as compared with control.
Hypoglycemia is a challenge and obstacle in the treatment of T2DM. In this meta-analysis, combination treatment of liraglutide and metformin did not increase the risk of hypoglycemia (2.97% VS 8.06%) compared with other therapies (placebo, glimepiride, and NPH). Our results were consistent with the findings of a recently published meta-analysis by Zhang L, et al. 52 . In that study, another GLP-1 receptor agonist dulaglutide was compared with other antidiabetic drugs for T2DM. Their results indicated that, as a monotherapy, dulaglutide did not increase the risk of hypoglycemia compared with control (placebo, metformin and liraglutide) (7.8% VS 10.6, respectively); as in the combination with OAM and lispro, dulaglutide resulted in a similar incidence of hypoglycemia (24.5% VS 24.5%) compared with control (placebo, sitagliptin, exenatide and liraglutide) 52 .
In this meta-analysis, although the liraglutide in combination with metformin induced a higher incidence of gastrointestinal disorders than control, these adverse events were mainly mild to moderate. The incidence of diarrhea was higher in patients treated with liraglutide than in the control group, which is consistent with results from previous trials 19,20,39,46 . No cases of pancreatitis were reported in these included studies, thus we did not assess the potential association between incretin and pancreatitis. However, the recent epidemiological studies indicated that, incretin-based therapies did not increase the risk of pancreatitis as compared with other diabetes treatments, and general population with T2DM was not a risk factor for the development of pancreatitis 53 .
We admit that there are several potential limitations in this meta-analysis. First, our meta-analysis was conducted based on nine RCTs, and two of them had a relatively small sample size (less than 100). Although all of these included studies were of high-quality (Jadad ≥ 3), our findings might be overestimated by these small trials since trials with small sample size were more likely to overestimate the treatment effect compared with those larger trials. Second, considerable heterogeneity existed among these included studies. However, it should not be surprising when considering the inclusion criteria for patients, dosage of liraglutide, duration of study, and different comparators. To explore the potential sources of heterogeneity, we conducted subgroup analysis based on the dosage of liraglutide, and comparators. And the issue of heterogeneity was resolved when the data analysis was performed according to the subgroup analysis. Third, our exploration of the effect comparison between liraglutide with active comparators was insufficient because of sparse reporting among the included studies. Therefore, physicians should interpret our findings with caution when applying them into the clinical practice.
In conclusion, our meta-analysis indicated that, liraglutideas added-on to metformin showed greater reduction in HbA1c levels, body weight, FPG, and PPG, and similar change in SBP and DBP compared with other therapies. However, considering the potential limitations in this study, more large-scale, well-conducted RCTs are needed to identify our findings.