Introduction

The burden of diabetes continues to increase globally, especially in developing countries [1]. Sodium-glucose cotransporter 2 (SGLT2) inhibitors are a new type of anti-hyperglycemic drug and have been recommended as first-line or second-line drugs for the treatment of diabetes by the European Diabetes Association and the American Diabetes Association. Recent studies have shown that SGLT2 inhibitors are beneficial for cardiovascular and renal outcomes in type 2 diabetes mellitus (T2DM) [2]. However, a systematic analysis of their efficacy in East Asian populations is lacking. Studies have shown that East Asian patients with T2DM have a lower body mass index (BMI), reduced insulin secretion, and higher insulin sensitivity than Caucasian patients with T2DM [3, 4]. It has also been reported that Chinese American and Japanese American premenopausal or early perimenopausal women without diabetes have a lower homeostatic model assessment (HOMA) of steady state beta cell function (%-β) than their non-Hispanic White counterparts [5]. These differences between East Asians and Western subjects in T2DM pathophysiology lead to discrepancies in therapeutic approaches. For example, studies suggest that dipeptidyl peptidase-4 (DPP-4) inhibitors and glucagon-like peptide-1 receptor agonists may be the more suitable therapeutic options for East Asian patients with T2DM [6]. Therefore, the aim of our study was to perform a systematic review and meta-analysis to evaluate the efficacy and safety profiles of SGLT2 inhibitors for the treatment East Asian patients with T2DM.

Methods

Search Strategy

This systematic review and meta-analysis were developed according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement [7]  (Electronic Supplemental Material [ESM] Table S1). We performed a systematic search of the PubMed, Embase, and the Cochrane Library databases to identify eligible studies published before October 2018 without the restrictions of country, language, or race. The following search terms were used: “East Asian” OR “East Asia” OR “China” OR “Japan” OR “Mongolia” OR “North Korea” OR “South Korea” AND “sodium glucose co-transporter 2 inhibitors” OR “dapagliflozin” OR “canagliflozin” OR “empagliflozin” OR “ipragliflozin” OR “tofogliflozin” OR “luseogliflozin” OR “sergliflozin” OR “remogliflozin.” Search conditions were adjusted to comply with the provisions of each database.

This article is based on previously conducted studies and does not contain any studies with human participants or animals performed by any of the authors.

Outcome Measures of Efficacy and Safety

The primary outcome of efficacy was the change in glycated hemoglobin (HbA1c) from baseline. The secondary results covered the proportion of patients with HbA1c < 7.0% (53 mmol/mol) and mean change in fasting plasma glucose (FPG), postprandial glucose (PPG) and body weight. The indicators of safety and tolerability events were hypoglycemia, urinary tract infections (UTIs), genital tract infections (GTIs), cardiovascular safety, acute renal failure, hypotension, and bone fractures.

Inclusion Criteria and Exclusion Criteria

Inclusion criteria were: (1) randomized clinical trials (RCTs) that compared SGLT2 inhibitors as either monotherapy or add-on therapy with other hyperglycemic drugs or placebo in T2DM subjects; (2) patients aged ≥ 18 years who were diagnosed with T2DM mellitus according to the 1999 World Health Organization diagnostic criteria or the 1997 American Diabetes Association [8, 9]; (3) treatment duration of at least 12 weeks; and (4) a percentage of East Asian patients in the study of > 50%.

The exclusion criteria were: (1) subgroup analysis, retrospective study, meta-analysis, uncontrolled trials, reviews, research in which HbA1c information cannot be extracted, duplicate publications, and case reports; and (2) patients with severe cardiac, hepatic, and renal insufficiency, as well as pregnant or lactating women.

Data Extraction

Two researchers independently screened and extracted the data according to the eligibility and exclusion criteria; any disagreements were discussed and a consensus reached. Following screening of the literature, two researchers assessed whether the trials would eventually be incorporated into the meta-analysis. Data on baseline demographics, interventions, efficacy, and safety outcomes were extracted.

Risk of Bias Assessment

The methodological quality of the included RCTs was assessed using the Cochrane Collaboration risk of bias tool (RevMan software version 5.3.5; Cochrane, London, UK). This tool assesses selection bias (random sequence generation; allocation concealment); performance bias (blinding of participants and personnel); detection bias (blinding of outcome assessors); attrition bias (incomplete outcome data); reporting bias (selective outcome reporting); and other bias parameters. Three levels were used to judge the risk bias of each study: “high risk,” “low risk,” and “unclear risk.” Two of the researchers conducted the quality assessment and consulted with a third reviewer if disagreements arose. Publication bias was assessed using funnel plots.

Statistical Analysis

RevMan (version 5.3.5; Cochrane Collaboration) was used for the analysis. For dichotomous data and continuous outcomes, risk ratios (RRs) and weighted mean differences (WMDs) were used. The Chi-square test (χ2) and I2 statistics were used to assess heterogeneity, and the random effect model was adopted regardless of I2. If the primary outcome data (such as the standard deviation and variance measures) were missing or incomplete, then we sent emails to the corresponding authors or sponsors of the study. If necessary, the standard deviation was calculated according to the confidence interval (CI) or standard error as described in the Cochrane Handbook. In addition, we conducted stratified analyses of the characteristic profiles based on HbA1c improvement and the risk of GTIs, respectively. The unpaired t test was selected in the GraphPad Prism 7.00 (Graphpad Software Inc., San Diego, CA, USA) to complete the analysis.

Results

Literature Search and Characteristics of the Included Studies

The selection steps and results are outlined in Fig. 1. Using the described search strategies, we screened 1450 items. After reviewing the articles according to titles, abstracts, and full texts in detail, we selected 33 trials for inclusion in the meta-analysis. Compared to placebo or other active antidiabetic agents (metformin, teneligliptin, sitagliptin, etc.), SGLT2 inhibitors were studied as monotherapy in 13 RCTs [10–22] and as add-on therapy in 20 RCTs [23–42]. Specifically, in these trials a total of 8496 patients were randomized, of which 6711 were assigned to an SGLT2 inhibitor: dapagliflozin (10 RCTs), canagliflozin (5 RCTs), empagliflozin (3 RCTs), ipragliflozin (9 RCTs), tofogliflozin (2 RCTs), and luseogliflozin (4 RCTs). The characteristics of the included studies are shown in ESM Table S2. Participants were aged > 18 years with HbA1c between 7% (53 mmol/mol) and 12% (107 mmol/mol). The BMI of most patients was > 24 kg/m2.

Fig. 1
figure 1

Flow diagram of the study selection process

Full size image

Risk of Bias Assessment

The risk of bias was estimated according to the Cochrane Collaboration risk of bias tool; the data are shown in ESM Fig. S1. Random sequence generation was evident in all but two of the RCTs [39, 42]; similarly, allocation concealment was obtained in 31 RCTs but was not explicitly described in these same two RCTs [39, 42]. In addition, four studies [39–42] were open label trials, and the blind law principle was not strictly enforced. The overall risk of bias in the included studies was low. Funnel charts showing publication bias are shown in ESM Fig. S2.

Efficacy of SGLT2 inhibitors

All 33 RCTs, involving 8469 patients, reported the change in HbA1c from baseline. The heterogeneity, as assessed by I2, was 92% (P < 0.00001). The meta-analysis (Fig. 2) using the random-effects model showed that the SGLT2 inhibitors improved HbA1c compared with the control group (WMD − 0.73%; 95% CI − 0.84, − 0.61; P < 0.00001). A subgroup analysis was also conducted to estimate the impact of different therapeutic regimens compared with placebo or active control. The results indicated that SGLT2 inhibitors as both a monotherapy and add-on therapy could significantly reduce HbA1c (monotherapy vs. control: WMD − 0.78%; 95% CI − 0.90, − 0.65; add-on therapy vs. control: WMD − 0.68%; 95% CI − 0.87, − 0.50) (Fig. 2).

Fig. 2
figure 2

The weighted mean difference in change from baseline in glycated hemoglobin (HbA1c) (%): sodium-glucose cotransporter 2 (SGLT2) inhibitors vs. control. SD Standard deviation, CI confidence interval, IV inverse variance

Full size image

Twelve trials (1275 patients) reported the percentage of patients with HbA1c  < 7% at the end of the   study The heterogeneity, as assessed by I2, was 80% (P < 0.00001). The meta-analysis showed that the percentage of patients with HbA1c  < 7% was 2.33-fold higher in the patients receiving SGLT2 inhibitors than in the control group [RR 2.33; 95% CI 1.74, 3.12). Compared with the control group, SGLT2 inhibitors significantly lowered the percentage of patients with HbA1c  < 7% , whether as monotherapy (RR 4.53;  95% CI 2.18, 9.40) or as add-on therapy (RR 1.81; 1.40, 2.33) (Fig. 3). The change in FPG and PPG was reported in 29 trials (7764 patients) and ten trials (2500 patients), respectively; the respective heterogeneities, as assessed by I2, were 90% (P  < 0.00001) and 95% (P < 0.00001). Our meta-analysis (ESM Figs. S3, S4) showed that both FPG (WMD − 28.47 mg/dl; 95% CI − 32.86, − 24.08)] and PPG (WMD − 52.32 mg/dl; 95% CI − 67.67, − 39.96) were improved in patients treated with SGLT2 inhibitors compared with the control group. Subgroup analysis revealed that SGLT2 inhibitors significantly lowered FPG compared to the control as both monotherapy (WMD − 29.99 mg/dl; 95% CI − 34.17, − 25.81) and as add-on therapy (WMD − 27.00 mg/dl; 95% CI − 34.71, − 19.30) (ESM Fig. S3). Similar results were found for PPG, with SGLT2 inhibitors clearly decreasing PPG as monotherapy (WMD − 58.14 mg/dl; 95% CI − 67.56, − 48.73) and as add-on therapy (WMD − 43.08 mg/dl; 95% CI − 68.82, − 17.35) (ESM Fig. S4). In the subgroup analysis, both monotherapy and add-on therapy distinctly improved FPG (monotherapy vs. control: WMD − 29.99 mg/dl, 95% CI − 34.17, − 25.81; add-on therapy vs. control: WMD − 27.00 mg/dl, 95% − 34.71, − 19.30) and PPG (monotherapy vs. control: WMD − 58.14 mg/dl, 95% CI − 67.56, − 48.73; add-on therapy vs. control: WMD − 43.08 mg/dl, 95% CI − 68.82, − 17.35).

Fig. 3
figure 3

Risk ratio in change from baseline in percentage of patients with HbA1c  <  7%: SGLT2 inhibitors vs.  control. M-H  Mantel-Haenszel

Full size image

The change in body weight was evaluated in 29 trials (7818 patients). The heterogeneity, as assessed by I2, was 98% (P < 0.00001). The meta-analysis showed (ESM Fig. S5) a significant reduction in body weight of − 1.73 kg (95% CI − 2.28, − 1.17) in patients receiving SGLT2 inhibitors as compared with those in the control group. In addition, compared with the control group, SGLT2 inhibitors significantly reduced body weight, whether as monotherapy or as add-on therapy (monotherapy vs. control: WMD − 1.73 kg, 95% CI − 1.56, − 1.90; add-on therapy vs. control: WMD − 1.70 kg, 95% CI − 2.53, − 0.87).

Safety of SGLT2 Inhibitors

The risk of all adverse events was reported in 30 trials (8163 patients). The heterogeneity, as assessed by I2, was 49% (P = 0.001). The meta-analysis (Fig. 4) using the random effects model showed that patients who received a therapeutic intervention with a SGLT2 inhibitor had a higher risk of adverse events than did the control group (RR 1.09; 95% CI 1.02, 1.12). Hypoglycemia was defined as a plasma glucose level of ≤ 3.9 mmol/L (70 mg/dl) with or without symptoms [43]. Thirty-one trials (8405 patients) reported hypoglycemia. The heterogeneity, as assessed by I2, was 50% (P = 0.001). The meta-analysis (ESM Fig. S6) showed that SGLT2 inhibitors did not increase the risk of hypoglycemia compared with the control group [RR 1.27; 95% CI 0.89, 1.82).

Fig. 4
figure 4

Risk of any adverse events: SGLT2 inhibitors vs. control

Full size image

UTI events were reported in 23 trials (7284 patients). The heterogeneity, as assessed by I2, was 0% (P = 0.97). The meta-analysis showed no higher risk of UTIs in patients taking SGLT2 inhibitors than in patients taking placebo or active controls (RR 0.93; 95% CI 0.68, 1.27) (ESM Fig. S7). Twenty-six trials (7687 patients) reported the GTI events. The heterogeneity as assessed by I2 was 0% (P = 0.89). The meta-analysis using the random-effects model showed that patients taking SGLT2 inhibitors were 1.73-fold more likely to develop GTIs than patients taking placebo or active controls (RR 1.73; 95% CI 1.02, 2.96) (Fig. 5).

Fig. 5
figure 5

Risk of genital tract infections: SGLT2 inhibitors vs. control group

Full size image

Five trials (1115 patients) using SGLT2 inhibitors for add-on therapy reported fractures. The heterogeneity, as assessed by I2, was 0% (P = 0.83). The meta-analysis showed that the risk of fracture in the experimental group did not increase compared with that of the control group (RR 1.60; 95% CI 0.48, 5.29) (ESM Fig. S8). Three trials [10, 24, 27] (1502 patients) focused on the indicator of hypotension (ESM Fig. S9), of which only one trial [24] reported a case of hypotension in the SGLT2 inhibitor (canagliflozin 100 mg) group. Ten trials (3333 patients) focused on the indicator of mortality (ESM Fig. S10); of these only one trial [16] reported a case of mortality in the placebo group.

The study duration of the RCTs included in this meta-analysis was mostly ≤ 24 weeks and, consequently, any long-term safety evaluation, such as cardiovascular mortality or other outcomes of special interest, could not be performed. Cardiovascular-related adverse events were reported in one article [25], with one event in the experimental group (teneligliptin + canagliflozin, n = 70) and two events in the control group (teneligliptin + placebo, n = 68), but no deaths were reported.

In addition, the total change in HbA1c was − 0.73%. Using this cutoff value, we performed a stratified analysis and compared the characteristics profile (age, duration of T2DM, HbA1c, FPG, body weight) of patients with a change in HbA1c levels of < 0.73% (Group 1) and those of patients with a change in HbA1c levels of > 0.73% (Group 2). The results showed that baseline HbA1c levels were higher in Group 2 patients (8.16% vs. 7.93%; P < 0.05), while the differences in other indicators were not statistically significant (ESM Table S3). At the same time, we used the stratified analysis to compare the characteristics of baseline data between people with no increased risk of GTIs (Group 3) and those with increased risk of GTIs (Group 4). The results of this analysis showed that the population in Group 4 had a higher mean body weight (69.95 vs. 67.69 kg; P < 0.05) and a longer mean duration of T2DM (8.18 vs. 6.48 years; P < 0.05) at the start of the respective study (ESM Table S4).

Discussion

In our systematic review, we evaluated the safety and efficacy of SGLT2 inhibitors (as monotherapy and add-on therapy) in East Asians with diabetes mellitus.

SGLT2 inhibitors are relatively new antihyperglycemic drugs and are not cheap. However, for patients who need to add other oral antihyperglycemic agents to metformin, SGLT2 inhibitors appear to be superior to other second-line drugs (such as DPP-4 inhibitors, sulfonylureas, etc.) [44]. Studies have shown that SGLT2 inhibitors reduce plasma glucose levels and body weight by increasing urine glucose excretion and thereby caloric loss, which is a different mode of action from several other hypoglycemic agents [45, 46]. This mechanism also plays a role in East Asian patients with T2DM. Our meta-analysis shows that compared with the control group, SGLT2 inhibitors improved both glycated hemoglobin (HbA1c) in patients (WMD − 0.73%; 95% CI − 0.84, − 0.61) and the percentage of patients with HbA1c  < 7%  (RR 2.33; 95% CI 1.74, 3.12); lowered both FPG (WMD − 28.47 mg/dl; 95% CI − 32.86, − 24.08) and PPG (WMD − 52.32 mg/dl; 95% CI − 67.67, − 39.96); reduced body weight (WMD − 1.73 kg; 95% CI − 2.28, − 1.17); and did not increase the risk of hypoglycemia, UTIs, hypotension, fractures, and mortality. However, SGLT2 inhibitors did increased the risk of GTIs (RR 1.73; 95% CI 1.02, 2.96). The meta-analysis based on the overall population showed that SGLT2 inhibitors significantly improved HbA1c levels (WMD − 0.66%; 95% CI − 0.73, − 0.58) compared with placebo [47, 48]. A meta-analysis aimed at evaluating the effects of SGLT2 inhibitors on UTIs and GTIs in subjects with T2DM reported that SGLT2 inhibitors did not increase the risk of UTIs, but did increase risk of GTIs [49]. These results were similar to those of our study.

The antihyperglycemic mechanism of SGLT2 inhibitors underlies the increased risk of polyuria, dehydration, and genitourinary infections due to increased levels of sugar in the urine [50]. Western population-based studies have shown that the risk of hypoglycemia with SGLT2 inhibitors is similar to that with other drugs, but that the risk of genitourinary infections increases [47, 51, 52]. Only the increased risk of urinary and genital infections has been consistently reported in clinical trials and observational studies [53]. Our meta-analysis indicates a slight increase in the risk of adverse reactions to SGLT2 inhibitors in East Asian T2DM patients compared to controls. In particular, the risk of developing GTIs was 1.73-fold higher in patients receiving SGLT2 inhibitors than in the control group. This result is consistent with the conclusion drawn by the authors of the Asian population-based study [54]. A meta-analysis involving a large population (n = 50,880) and no ethnic restrictions [55] found that SGLT-2 inhibitors did not appear to increase the risk of UTIs in patients with T2DM—with the exception of high-dose dapagliflozin (10 mg daily). In addition to race, sample size, drug type, and dose, study duration may also be a factor influencing the results of any analysis. To summarize, our results lead us to suggest that the treating physician should focus on GTIs in East Asian patients who use SGLT2 inhibitors and be alert to UTIs. The cardiovascular and renal safety of SGLT2 inhibitors in East Asian patients with T2DM could not be adequately assessed in our meta-analysis due to the limitation in study duration of the trials included in the meta-analysis. Therefore, further research is needed.

The results of the stratified analysis suggest that for East Asian patients with T2DM who use SGLT2, the level of HbA1c was higher at baseline and the improvement might be more pronounced at the end of the trial, compared with the control group. Our analysis also suggests that patients with higher body weight or a longer history of T2DM have an increased risk of developing GTIs after treatment with SGLT2 inhibitors. The specific mechanisms still need to be researched.

The efficacy data had a heterogeneity of > 90%, whereas the heterogeneity of the safety data was close to 0%. An important for this difference may be that the indicators involved in the effectiveness analysis (such as HbA1c) were reported in all 33 RCTs included in our meta-analysis, but the indicators involved in the safety analysis (such as UTIs) were only reported in 26 studies. The second explanation may be the data types. HbA1c is the continuous variable and GTIs are the dichotomous outcomes. Taking into account the obvious heterogeneity associated with HbA1c, body weight, FPG, PPG, and other indicators, we grouped the studies according to study protocol and performed a subgroup analysis based on drug type to explore the sources of heterogeneity. We divided the included studies into four groups (monotherapy vs. placebo; add-on therapy vs. placebo; monotherapy vs. active control; add-on therapy vs. active control) and established six subgroups (dapagliflozin, canagliflozin, empagliflozin, ipragliflozin, tofogliflozin, and luseogliflozin). The results are shown in ESM Figs. S12–20). Our final conclusion is that the heterogeneity may be due to the type of drug; however, we could not rule out the effect of drug dose. In general, this did not affect our assessment of the overall efficacy of SGLT2 inhibitors.

Our research had several potential limitations. First, considering the clinical heterogeneity of different randomized clinical trials, such as study design methods, drug doses, and medication regimens, we used a random effects model for all of the meta-analyses. Moreover, we excluded meeting abstracts, posters, and articles in which we did not have access to raw data.

Conclusion

This meta-analysis evaluated the efficacy and safety of SGLT2 inhibitors in the treatment of T2DM in East Asian patients. First, SGLT2 inhibitors exhibited beneficial effects in terms of the control of HbA1c, either as monotherapy or as an add-on to oral antihyperglycemic agents, compared to the control group. At the same time, SGLT2 inhibitor therapies achieved both significant reductions in FPG and PPG and a decrease in body weight. Our safety analysis showed a statistically significant increase in the risk of GTIs, while the risk of UTIs, hypoglycemia, hypotension, fractures, and mortality did not appear to increase. The stratified analysis showed that patients with higher HbA1c levels at baseline might achieve a greater improvement in HbA1c after taking SGLT2 inhibitors; those with higher body weight or a longer history of diabetes might have an increased risk of developing GTIs. In conclusion, further long-term studies are needed to fully assess the benefit/risk profiles of SGLT2 inhibitors, which in turn will help to determine the strengths and weaknesses of the use of SGLT2 inhibitors to treat East Asian patients with T2DM.