Introduction
Type 2 diabetes mellitus (T2DM) is characterized by increased insulin insensitivity coupled with a progressive failure of pancreatic β-cells, resulting in a gradual loss of glycemic control and hyperglycemia [
1]. If uncontrolled, hyperglycemia can lead to diabetic complications, including microvascular (e.g., retinopathy) and macrovascular (e.g., myocardial infarction) complications [
2]. Globally, the prevalence of diabetes is increasing and it is estimated that 642 million people aged 20–79 years will have diabetes by 2040; of these people, 87–91% will have T2DM [
3]. In Japan, a national survey has shown that the prevalence of diabetes has increased from 8.9 million in 2007 to 10 million in 2016 [
4]. Specifically, a significant increase in the prevalence of T2DM has also been recorded in a cohort study [
5]; this study attributed the increase in T2DM to higher rates of obesity and reduced physical activity levels [
5].
Treatment for T2DM is focused on the management of hyperglycemia and reducing the levels of glycated hemoglobin (HbA
1c) [
6]; the reduction in HbA
1c is associated with a reduction in the risk of diabetic complications [
2,
7]. In Japan, the main HbA
1c target set by the Japanese Diabetes Society is < 7% (53.0 mmol/mol), which was established from the perspective of preventing complications [
8]. However, despite the wide range of treatments available, a recent survey has shown that 47% of patients with T2DM in Japan do not achieve the recommended HbA
1c goal [
9]. In addition, patients with a higher body mass index (BMI) were less likely to achieve the HbA
1c target than patients with a lower BMI [
9]. Weight gain is also a known side effect of many oral antidiabetic drugs (OADs) and insulins used to treat T2DM, and may be associated with an increased risk of cardiovascular disease [
10,
11].
Glucagon-like peptide-1 receptor agonists (GLP-1 RA) are a class of drugs that act by increasing glucose-dependent insulin secretion, preserving β-cell function, suppressing glucagon levels, and slowing gastric emptying [
12,
13]. Unlike other therapies in T2DM, treatment with GLP-1 RA is typically associated with weight loss [
14‐
17] and may also reduce cardiovascular risk [
18]. In Japan, GLP-1 RAs are increasingly prescribed for the treatment of T2DM and several studies have demonstrated the efficacy and safety of GLP-1 RAs in the Japanese population [
19‐
23]. Studies have also shown that GLP-1 RAs are more effective in Japanese/Asian populations when compared with European populations [
24,
25], and are typically administered at a lower dose in the Japanese population [
26]. This is because T2DM in Asian patients is primarily characterized by higher β-cell dysfunction, rather than insulin resistance [
27], and GLP-1 RAs have a proven ability to improve β-cell function [
28]. Therefore, GLP-1 RAs are a favorable choice of therapy for this population. Japanese patients with T2DM are also generally less obese than their European counterparts and the exposure of a fixed dose GLP-1 RA will be greater in lighter patients.
Semaglutide once-weekly (QW) is a novel GLP-1 analogue administered at a 0.5 or 1.0 mg dose. The efficacy and safety of semaglutide QW has been assessed across the Semaglutide Unabated Sustainability in Treatment of Type 2 Diabetes (SUSTAIN) clinical trial program, which included two multicenter, head-to-head trials comparing semaglutide QW with exenatide QW or dulaglutide QW. In SUSTAIN 3, treatment with semaglutide QW provided significant reductions in HbA
1c and weight when compared with exenatide QW, both as an add-on to 1–2 OADs [metformin and/or sulfonylurea (SU) and thiazolidinedione (TZD)] [
29]. In SUSTAIN 7, semaglutide QW also provided significant reductions in HbA
1c and weight when compared with dulaglutide QW, both as an add-on to metformin [
30]; however, no Japanese patients were included in this trial.
The efficacy and safety of semaglutide QW in a Japanese population with T2DM were demonstrated in two recent trials. One trial compared semaglutide QW (0.5 and 1.0 mg) as monotherapy with sitagliptin 100 mg once-daily (QD) [
31]. After 30 weeks, both doses of semaglutide QW demonstrated significant reductions in HbA
1c and weight compared with sitagliptin 100 mg QD; the safety profile was also comparable to other GLP-1 RAs. Semaglutide QW (0.5 and 1.0 mg) was also assessed in a 56-week trial of patients who had poor glycemic control on diet and exercise or one OAD [
32]. The addition of semaglutide QW was associated with significant improvements in HbA
1c and weight when compared with the addition of another OAD.
Following the recent introduction of health technology appraisal (HTA) in Japan and the increasing amount of treatment options available to patients with T2DM, decision-makers need to assess the relative clinical benefits and risks of each treatment to make recommendations on their use within a limited budget. The availability of such comparative clinical data can also be used for cost-effectiveness analyses in Japan in the context of HTA. In the absence of head-to-head trials between semaglutide QW and other GLP-1 RAs in a Japanese population, a network meta-analysis (NMA) was performed. The objective was to assess the relative efficacy and safety of semaglutide QW vs GLP-1 RAs in Japanese patients, with a specific focus on the comparison between semaglutide 0.5 mg QW and dulaglutide 0.75 mg QW; dulaglutide is the latest QW GLP-1 RA in Japan (only available as a 0.75 mg QW dose) and semaglutide 0.5 mg QW is the expected maintenance dose for the Japanese population.
Methods
Systematic Review
A systematic review (SR) was performed in accordance with PRISMA guidelines [
33] to identify trials of GLP-1 RAs in Japanese patients with T2DM. Searches of MEDLINE
®, Embase, and the Cochrane Library were performed via Ovid on April 5, 2016, with updates occurring on October 3, 2016 and August 16, 2017. Searches of conference proceedings were also carried out for the European Association for the Study of Diabetes (EASD; 2014–2016), the International Society for Pharmacoeconomics and Outcomes Research (ISPOR; 2014–2017), the International Diabetes Federation (IDF; 2013 and 2015), and the American Diabetes Association (ADA) Scientific Sessions (2014–2017). The search results were then screened against the SR eligibility criteria to generate a list of potential studies to include in the NMA (Table S1). As the eligibility criteria of the SR were restricted to studies published in English, a supplementary search of four Japanese language databases (Japan Pharmaceutical Information Center, Japan Science and Technology Information Aggregator, J-GLOBAL, and Igaku Chuo Zasshi) was conducted to ensure all potentially relevant studies had been identified.
Citations of interest were identified by one reviewer and verified by a second independent reviewer on the basis of title and abstract. Full publications were obtained for all citations of interest and were assessed by one reviewer and verified by a second independent reviewer. Any uncertainties were resolved through discussion between reviewers. Data were then extracted into an Excel spreadsheet by one reviewer and checked by a second reviewer. All included references were assessed for risk of bias using a seven-criteria checklist as approved by the National Institute of Health and Care Excellence (NICE) [
34].
NMA Methodology
An NMA was performed to compare the efficacy and safety of GLP-1 RAs in Japanese patients, where the primary intervention of interest was semaglutide 0.5 mg QW and the comparators of interest were all other licensed doses of GLP-1 RAs in Japan—liraglutide QD, dulaglutide QW, exenatide twice-daily (BID), exenatide QW, and lixisenatide QD. In order to reduce variability between populations in different trials, the population of interest was aligned to the two Japanese SUSTAIN trials of semaglutide QW [NN9535-4091 trial (NCT02207374; now available as a full-text publication, Kaku et al. [
32]) and NN9535-4092 (NCT02254291; now available as a full-text publication, Seino et al. [
31])]. Therefore, while trials investigating a broader population were eligible for inclusion in the SR (Table S1), only trials investigating semaglutide QW or other licensed doses of GLP-1 RAs in Japan and in Japanese populations who have previously received 0–1 OADs were considered for further analysis.
The feasibility of performing an NMA at two time points (30 and 56 weeks), based on the study durations of the two Japanese SUSTAIN trials, was examined. All trials identified in the SR were examined for data on at least one outcome of interest at approximately 30 weeks (duration of the NN9535-4092 trial [
31]) and 56 weeks (duration of the NN9535-4091 trial [
32]), and the ability to form a best-case connected network was assessed. The feasibility of generating evidence networks for each of the 20 outcomes of interest (Table S1) was next examined; the outcomes of interest included HbA
1c outcomes [e.g., change from baseline, proportion of patients achieving a level < 7% (53 mmol/mol)], weight, BMI, systolic blood pressure (SBP), fasting plasma glucose (FPG), postprandial plasma glucose ,and safety outcomes (including the incidence of overall hypoglycemia).
The analysis of continuous outcomes (e.g., change from baseline in HbA1c) was performed using a normal likelihood, identity link, single parameter model (based on arm-level data), or a shared parameter model, which allows for a single coherent synthesis when outcome data is reported at both the arm level and trial level. For dichotomous outcomes, a binomial likelihood, logit link model was used for efficacy outcomes [e.g., proportion of patients achieving a HbA1c level < 7% (53 mmol/mol)], while a binomial likelihood, cloglog link model was used for safety outcomes (e.g. ,incidence of overall hypoglycemia). All analyses were performed using a fixed effects (FE) model; the FE model provided a better model fit compared with the random effects (RE) model in terms of deviance information criterion and residual deviance.
The NMA models were implemented using WinBUGS software (MRC Biostatistics Unit, Cambridge, UK) [
35] and employed a Bayesian framework with the inclusion of vague prior distributions. Three Markov Monte Carlo chains were used, starting from different initial values of selected unknown parameters. Convergence for all models were assessed by analyzing history and density plots, and Brooks–Gelman–Rubin diagnostic plots [
36]. In addition, autocorrelation plots were assessed to detect the presence of autocorrelation in the chains. Following this, model convergence inferences were made from data obtained by sampling for a further 20,000 iterations.
For continuous outcomes, a mean treatment effect with an associated 95% credible interval (CrI) is estimated and the treatment difference (95% CrI) for semaglutide 0.5 mg QW vs comparator is presented. For dichotomous outcomes, an odds ratio (OR) with an associated 95% CrI is calculated for semaglutide 0.5 mg QW vs comparator. A difference between semaglutide 0.5 mg QW and a comparator is assumed to only exist when the 95% CrI does not include the null value for treatment differences, or one for ORs.
Finally, this article does not contain any new studies with human or animal subjects performed by any of the authors.
Discussion
The objective of this study was to assess the relative efficacy and safety of semaglutide 0.5 mg QW vs GLP-1 RAs in Japanese patients, with a specific focus on the comparison between semaglutide 0.5 mg QW and dulaglutide 0.75 mg QW. Overall, semaglutide 0.5 mg QW was associated with significant reductions from baseline in HbA1c, weight, SBP, and FPG, compared with dulaglutide 0.75 mg QW. In addition, semaglutide 0.5 mg QW achieved significant reductions from baseline in HbA1c and FPG vs both liraglutide 0.6 and 0.9 mg QW, and a significant reduction in weight vs liraglutide 0.9 mg QW (liraglutide 0.6 mg QD was not available for comparison). Furthermore, semaglutide 0.5 mg QW was associated with significantly higher odds of achieving a HbA1c level < 7% (53.0 mmol/mol) compared with liraglutide 0.6 mg and 0.9 mg QD.
To our knowledge, this is the first NMA comparing semaglutide QW with other GLP-1 RAs in a Japanese population. However, the results of this NMA can be compared with the recent head-to-head, multicenter (Asia, Europe, and the USA), clinical trial SUSTAIN 7, which assessed semaglutide QW and dulaglutide QW, both as an add-on to metformin [
30]. The direction of effect for HbA
1c and weight demonstrated between semaglutide 0.5 mg QW and dulaglutide 0.75 mg QW in our analysis is broadly consistent with the data reported in SUSTAIN 7. In SUSTAIN 7 (which did not include any Japanese patients), semaglutide 0.5 mg QW demonstrated a statistically significant reduction in HbA
1c (− 1.5% vs − 1.1%) and weight (− 4.6 vs − 2.3 kg) compared with dulaglutide 0.75 mg at 40 weeks. This equates to a treatment difference of − 0.4% for HbA
1c and − 2.3 kg for weight (both in favor of semaglutide 0.5 mg QW), which are similar to the treatment differences estimated in our NMA at 52–56 weeks [− 0.61% (95% CrI − 0.92, − 0.30) and − 1.45 kg (95% CrI − 2.65, − 0.25), respectively]. SUSTAIN 7 also demonstrated that more patients achieved a HbA
1c level ≤ 7% (53.0 mmol/mol) with semaglutide 0.5 mg QW compared with dulaglutide 0.75 mg QW (69% vs 52%). In our analysis, semaglutide 0.5 mg QW was associated with an OR of 2.01 vs dulaglutide 0.75 mg QW for the proportion of patients achieving a HbA
1c level < 7% (53.0 mmol/mol), thereby corroborating what was observed in SUSTAIN 7; however, as the 95% CrI included 1, statistical significance could not be concluded.
Overall, our analysis provides a robust assessment of the efficacy of semaglutide QW across five outcomes compared with dulaglutide QW and liraglutide QD. In contrast, it was only possible to provide estimates for one safety outcome (incidence of overall hypoglycemia), which limits the ability to make any definitive conclusions from our study on the relative safety of semaglutide QW compared with other GLP-1 RAs. However, the two recent Japanese trials of semaglutide QW have demonstrated that its safety profile is comparable to other GLP-1 RAs [
31,
32]. Our analysis of the incidence of overall hypoglycemia is in agreement with these findings, as no significant differences for semaglutide 0.5 mg QW vs dulaglutide 0.75 mg QW or liraglutide 0.9 mg QD were detected. The NMAs were also subject to additional limitations. Although an SR and supplementary Japanese-specific searches were conducted to identify relevant trials, only four trials were available for the analysis at 52–56 weeks. Consequently, the direct comparisons within the evidence network were supported by data from only one trial each. In addition, while the overall risk of bias across the trials included in the analyses was low, the highest risk was associated with study blinding due to the presence of two open-label trials. The absence of double-blinding may be considered a limitation as this can potentially introduce performance bias into the NMA [
39].
Analysis of a few secondary outcomes assessed may have also been limited by poor reporting or low event rates. The overall incidence of hypoglycemia was low across the trials, which may have contributed to greater uncertainty in the analysis. Hypoglycemic episodes are often underreported by patients, which may be due to reasons such as fear of being judged, losing their job or driving license, and unawareness of nocturnal events [
40,
41]. In addition, studies often report the number of patients with hypoglycemic events, rather than the actual frequency of events [
41]. Furthermore, non-severe events are not always reported in trials. Patients may not consider such mild events as important enough to mention to their doctor [
41,
42]. Greater uncertainty was also observed in the analysis of SBP, when compared to the analysis of HbA
1c, FPG, and weight. The measurement of SBP is known to be variable and values can differ for the same patient throughout the day. In addition, there are often patient-, device-, and procedure-related inaccuracies in the way that blood pressure is monitored [
43]. Together, these may have contributed to the uncertainty present in our analysis of SBP. Despite this, a significant difference in the reduction of SBP from baseline with semaglutide 0.5 mg QW compared with dulaglutide 0.75 mg QW was still observed.