MEDLINE, EMBASE, and Cochrane Central Register of Controlled Trials were systematically searched to identify potentially eligible studies until July 31, 2021. We also searched ClinicalTrials.gov to identify unpublished studies. The search results were restricted to clinical trials and English publications. Cited references, reviews, and meta-analyses were checked to identify additional studies. Details of the study selection process are shown in Supplementary Table S1.

Studies were considered if they met the following inclusion criteria: 1) Randomized controlled trials (RCTs); 2) investigation of adult patients with type 2 diabetes; 3) ertugliflozin; 4) reporting the efficacy and safety endpoints; and 5) duration of intervention of at least 12 weeks. Exclusion criteria for studies were as follows: conference abstracts, reviews, letters, editorials, case reports, observation studies, long-term extension studies, and post hoc analyses. The primary efficacy outcomes comprised glycaemic control [HbA1c, the proportion of participants achieving an HbA1c level < 7%, and fasting plasma glucose (FPG)]; weight loss (body weight); and blood pressure control [systolic blood pressure (SBP) and diastolic blood pressure (DBP)]. Adverse events (AEs) with different degrees and prespecified AEs of SGLT2 inhibitors were selected to assess the safety and tolerability of ertugliflozin. The primary safety outcomes were any AEs, AEs related to study drug, serious AEs, deaths, discontinuations due to AEs, and predetermined AEs of interest for ertugliflozin [genital mycotic infection (GMI), urinary tract infection (UTI), symptomatic hypoglycaemia, and hypovolaemia]. AEs included in the analysis were mainly coded according to the Medical Dictionary for Regulatory Activities, as defined in the individual study (Supplementary Table S2). Considering the resource costs and social benefits of type 2 diabetes, we performed a cost-effectiveness analysis of ertugliflozin based on information provided by the included clinical literature. Detailed methods, results, and discussion were presented in Supplementary Material.

Using designed electronic forms, two authors (LL and F-HS) extracted the data for each article, including the first author’s name, publication time, NCT number, randomisation, intervention characteristics (type, dose, and duration of interventions), patient characteristics (background treatment, the proportion of men, mean age, duration of type 2 diabetes), and reported outcomes. Any dispute was resolved by consensus or by consultation with the corresponding authors (YW and Z-CG).

Two independent reviewers (LL and F-HS) performed the quality assessment using the Cochrane Collaboration’s tool (Higgins et al., 2011), and any disagreements were resolved by the corresponding authors (YW and Z-CG). Considering the risk of bias, we evaluated the following aspects: adequacy of random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, completeness of outcome data, selective outcome reporting, and other biases that could induce confounding effects.

All statistical analyses were performed using Stata version 12.0 (Stata Corporation, College Station, TX, United States). Meta-analysis estimates of the studies were derived and presented as forest plots. We applied a random-effects model to evaluate the overall estimated effects. Continuous variables, including least-squares, mean change versus control group for HbA1c, FPG, body weight, SBP, and DBP, were calculated as weighted mean difference (WMD) and associated 95% confidence intervals (CIs). A WMD less than 0 signified that the results favoured the use of ertugliflozin compared with other therapies. For dichotomous data, including patients with HbA1c < 7% and the rate of any AEs, we used risk ratios (RRs) and their 95% CIs. An RR less than 1 indicated a lower trend in ertugliflozin treatment than that in control. Subgroup analyses were performed for overall efficacy and safety outcomes by dosage (ertugliflozin 5 mg, ertugliflozin 15 mg), follow-up (≤ 26 weeks, > 26 weeks), and control (placebo or active agents). The interaction analysis (P for interaction) was performed to assess the comparability of efficacy and safety indexes in different dosages, follow-ups, and controls (Gu et al., 2020). A significant p indicated the difference among subgroups. Further leave-one-out and fixed-effect model sensitivity analyses were applied to detect the robustness of the results. The I ² test was used to measure the total variation between studies to test the heterogeneity among studies (significance for I ² > 50%) (Higgins and Thompson, 2002). Statistical significance was set at p < 0.05.

The initial search identified 236 records meeting the inclusion criteria, and nine studies (RCT trials) involving 5638 participants were finally included in the quantitative synthesis (Amin et al., 2015; Aronson et al., 2018; Dagogo-Jack et al., 2018; Grunberger et al., 2018; Miller et al., 2018; Pratley et al., 2018; Gallo et al., 2019; Hollander et al., 2019; Ji et al., 2019). The screening process is illustrated in Figure 1. The publication years of included trials were from 2015 to 2019. Background treatments included diet and exercise, metformin, and other antihyperglycaemic agents. Of the included trials, five trials were controlled with placebo (Amin et al., 2015; Dagogo-Jack et al., 2018; Grunberger et al., 2018; Miller et al., 2018; Ji et al., 2019), two trials were compared with glimepiride (Gallo et al., 2019; Hollander et al., 2019), and the remaining two were controlled with metformin, sitagliptin, respectively (Aronson et al., 2018; Pratley et al., 2018). Seven trials mainly focused on Caucasians (ranged from 64.8 to 90.4%), one trial focused on Asians (406 Chinese patients), and one trial did not report races. The duration of the trial follow-up ranged from 12 to 104 weeks. The mean age of participants was 57.6 years, and the mean duration of diabetes was 4.6–14.7 years. The mean HbA1c % ranged from 7.8 to 9.0%. Detailed demographic characteristics are presented in Table 1.

We evaluated the effects of ertugliflozin on glycaemic variables, body weight, and blood pressure, as shown in Figure 2. Meta-analysis results revealed that ertugliflozin significantly decreased HbA1c levels (%) (WMD −0.452%; 95% CI −0.774 to −0.129; I ² = 96.2%), HbA1c (mmol/mol) (WMD −7.722 mmol/mol; 95% CI −14.691 to −0.753; I ² = 98.3%), FPG (WMD −0.870 mmol/L; 95% CI −1.418 to −0.322; I ² = 96.6%), consequently increasing the rate of patients achieving target HbA1c (< 7%) (RR: 1.152; 95% CI: 1.073–1.951; I ² = 80.7%), compared with other hypoglycaemic agents or placebo. For body weight and blood pressure, the reduction in body weight from baseline was more considerable in ertugliflozin than non-ertugliflozin (WMD −1.774 kg; 95% CI −2.601 to −0.946; I ² = 97.6%). A greater reduction in blood pressure levels was also observed in patients receiving ertugliflozin compared with other treatments (SBP: WMD −2.572 mmHg; 95% CI −3.573 to −1.571; I ² = 93.8% and DBP: WMD −1.152 mmHg; 95% CI −2.002 to −0.303; I ² = 85.5%).

Table 2 shows the results of the subgroup analysis on the regulation of glycaemia, blood pressure, and body weight. Regarding glycaemic control, the reduction in HbA1c (%) and FPG based on different dosages were in line with the primary outcomes. For follow-up duration, patients administered with ertugliflozin decreased HbA1c mainly in short follow-up studies, rather than in long ones (≤ 26 weeks: WMD −0.788%; 95% CI −1.169 to −0.407; I ² = 88.6%; >26 weeks: WMD −0.287%; 95% CI −0.646 to 0.072; I ² = 95.6%; P _interaction of follow-ups = 0.061). For different controls, ertugliflozin presented a notable decrease in HbA1c compared with placebo (WMD −0.641%; 95% CI −0.984 to −0.298; I ² = 93.0%) and active agents (WMD −0.215%; 95% CI −0.629 to −0.199; I ² = 94.9%), but the interactive effect was significant (P _interaction of 0.039). Additionally, ertugliflozin had a significant reduction in FPG levels for less than 26 weeks compared with more than 26 weeks, with a P _interaction of 0.023. Similar results were also found in different controls (P _interaction of 0.006). As for reaching the glycaemia target, subgroup analyses showed that when used less than 26 weeks or compared with placebo, ertugliflozin contributed to a higher proportion of patients attaining HbA1c (< 7%), with the risk ratio of 4.045 (95% CI: 2.098–5.992; I ² = 41.0%), 3.834 (95% CI: 2.677–4.991; I ² = 11.6%), respectively. The interactive effects were of significant difference when follow-ups and controls were considered (P _interaction < 0.05). For different dosages, the results of subgroup analyses (5 mg RR: 1.196; 95% CI: 0.843–1.549; I ² = 64.9%; 15 mg RR: 1.326; 95% CI: 0.908–1.744; I ² = 69.2%) were not consistent with the primary results (RR 1.152; 95% CI: −1.073 to 1.951; I ² = 80.7%). Further, the results of leave-one-out sensitivity analysis were basically similar to the primary outcomes. Concerning the proportion of patients achieving HbA1c < 7%, two trials were found to influence the primary outcomes of individual trials (Dagogo-Jack et al., 2018; Ji et al., 2019) (Supplementary Table S3).

For body weight and SBP, the results of subgroup analyses across dosage, follow-up, or control were all in line with the primary outcomes. Regarding DBP, the overall subgroup analyses were not completely consistent with the primary analyses. For different dosages considered, only the use of 5 mg ertugliflozin was found to significantly decrease DBP levels (WMD −0.877 mmHg; 95% CI −1.695 to −0.059; I ² = 61.6%), although the use of 15 mg ertugliflozin was associated with a decreasing trend (WMD −0.586 mmHg; 95% CI −1.209 to 0.037; I ² = 38.3%), with a P _interaction of 0.579. For follow-up duration, a meaningful reduction of DBP was observed when ertugliflozin was used less than 26 weeks (WMD −2.042 mmHg; 95% CI −2.812 to 1.271; I ² = 69.7%). By contrast, ertugliflozin treatment more than 26 weeks failed to decrease DBP levels (WMD −0.566 mmHg; 95% CI −1.294 to 0.161; I ² = 69.7%) (P _interaction of 0.006). Similar results were observed in the subgroup analysis of different controls, and a P _interaction between placebo and active controls was 0.037. Additionally, further sensitivity analyses on body weight and blood pressure indicated that the outcomes remained invariant whatever either study was excluded. The results of the fixed-effect model remained consistent with that of the random-effect model (Supplementary Table S4).

The analysis assessed the tolerability of type 2 diabetes patients, as presented in Figure 2. The incidence of any AEs was 63.58% (2517/3959) in ertugliflozin group and 65.73% (1097/1669) in non-ertugliflozin group (Supplementary Table S5), indicating a similar risk between the two therapies (RR: 0.983; 95% CI 0.944–1.023; I ² = 0%). The results were consistent with those of all subgroups, as shown in Table 3. Also, no significant differences between ertugliflozin group and control group were observed in terms of serious AEs (RR: 1.132; 95% CI 0.909–1.409; I ² = 0%), death (RR: 1.174; 95% CI 0.054–2.739; I ² = 0%), and AEs leading to discontinuation (RR: 1.140; 95% CI 0.858–1.515; I ² = 0%). There were no notable differences between groups in the incidence of AEs related to study drug (1.158; 95% CI 0.963–1.391; I ² = 45.4%), although the incidence was higher when using ertugliflozin (19.22%) compared with other therapies (17.14%).

Regarding GMI, ertugliflozin use increased the 3-fold risk compared with other medications (RR: 4.004; 95% CI 2.504–6.402; I ² = 13.7%), with a higher incidence in ertugliflozin group (6.59%) versus in control group (1.44%). However, when the follow-up was less than 26 weeks, no correlation of higher GMI risk was observed with ertugliflozin treatment (RR: 1.977; 95% CI 0.756–5.165; I ² = 0%) (Table 3). AEs related to symptomatic hypoglycaemia were more common in non-ertugliflozin group (10.84%) than in ertugliflozin group (5.35%). However, there was no evidence that ertugliflozin use had a lower risk of symptomatic hypoglycaemia than other therapies (RR: 0.781; 95% CI 0.418–1.459; I ² = 83.7%). In the subgroup analysis by follow-up, a higher risk of symptomatic hypoglycaemia in ertugliflozin group was found in less than 26 weeks compared with in non-ertugliflozin group (RR: 3.896; 95% CI 1.058–14.337; I ² = 0%). We failed to find a higher risk of UTI in ertugliflozin treatment group compared with the control (RR: 0.906; 95% CI 0.737–1.113; I ² = 0%), with a lower incidence (6.92% in ertugliflozin group, 7.67% in control group, respectively). Also, there was no meaningful difference between the two groups in hypovolemia (RR: 1.296; 95% CI 0.577–2.910; I ² = 35.6%), with a comparable incidence (1.66% in ertugliflozin group, 1.18% in control group). Further, the results were similar in fixed-effect model sensitive analysis (Supplementary Table S6).

Seven trials used an interactive voice response system/integrated web response system when allocating concealment; two trials that did not report a similar system were evaluated as unclear. One trial that did not report the personnel assessing the outcomes was unclear. According to the bias tool item, nine trials were assessed to have a low risk of bias (Supplementary Table S7).

This study involved 5638 patients with type 2 diabetes from nine randomized clinical trials to evaluate the efficacy and safety of ertugliflozin comprehensively. Overall, the results revealed that ertugliflozin performed well in glycaemic control, blood pressure and weight management compared with other therapies (placebo, metformin, glimepiride, and sitagliptin). Ertugliflozin presented a better profile of glycaemic control and blood pressure regulation controlled with placebo rather than other active agents. For safety, ertugliflozin provided favourable tolerability in patients with type 2 diabetes but increased the risk of GMI, urging for personal hygiene education and close supervision in clinical treatment. The results of key subgroups were mainly consistent with the primary outcomes.

Ertugliflozin is a highly selective inhibitor with remarkable selectivity for SGLT2 over SGLT1 (> 2000-fold) (Cinti et al., 2017). Its oral bioavailability was nearly 100%, which was higher than that of dapagliflozin (78%) and canagliflozin (65%) (Kasichayanula et al., 2014; Devineni and Polidori, 2015; Fediuk et al., 2020). Therefore, ertugliflozin may have a preferable hypoglycaemic effect. We demonstrated that ertugliflozin lowered 0.45% HbA1c and raised 15.2% of the proportion of patients achieving target HbA1c (< 7%) than placebo and other glucose-lowering agents. Ertugliflozin resulted in a more significant reduction of HbA1c (%) in placebo-controlled groups than active agents-controlled ones, with a P _interaction of controls 0.039. It was reasonable and confirmed the credibility of our analysis. Meanwhile, subgroup analysis showed that 5 and 15 mg ertugliflozin reduced HbA1c by 0.43 and 0.47%, respectively. Generally, our results remained consistent with previous meta-analyses of ertugliflozin, in which ertugliflozin reduced HbA1c by 0.5–1.0% (Liu et al., 2020b; Cherney et al., 2020; Pratley et al., 2020). Analyses of three trials within 26 weeks showed ertugliflozin yielded the target HbA1c (< 7%) with a high RR of 4.045 (Amin et al., 2015; Miller et al., 2018; Ji et al., 2019). The results noted that ertugliflozin might exhibit better glycaemic control in the first few months. The mechanism is unclear, and the possible reason is that SGLT2 inhibitors seem to occur an early and rapid body water and fat loss up to approximately 8 weeks, then become a slower rate of sustained fat loss (Lee et al., 2018). Furthermore, a network meta-analysis performed by McNeill (McNeill et al., 2019) found that ertugliflozin was more effective in decreasing HbA1c than dapagliflozin and empagliflozin in patients with type 2 diabetes. When used alone, ertugliflozin 15 mg was more effective in reducing HbA1c than 10 mg dapagliflozin (0.36%) or 25 mg empagliflozin (0.31%). When combined with metformin, ertugliflozin was more effective than dapagliflozin (ertugliflozin 5 mg versus dapagliflozin 5 mg 0.22%; ertugliflozin 15 mg versus dapagliflozin 10 mg 0.26%). Furthermore, 15 mg ertugliflozin lowered HbA1c by 0.23% more than 25 mg empagliflozin. A model-based meta-analysis consolidated the results, which performed an indirect comparison of SGLT2 inhibitor efficacy, indicating that ertugliflozin was of comparable efficacy in magnitude to other SGLT2 inhibitors (Deryl Fediuk et al., 2019). Although there was no head-to-head study comparing the hypoglycaemic effect across all SGLT2 inhibitors, these results indicated that ertugliflozin had a relatively higher hypoglycaemic effect.

Obesity and hypertension are associated with cardiovascular morbidity and contribute to death in patients with type 2 diabetes (Low Wang et al., 2016). Improvement of these cardiovascular risk factors lowers the excess risk of death in adults with type 2 diabetes (Rawshani et al., 2018). SGLT2 inhibitors have a profile of weight loss and blood pressure control by a multifactorial mechanism (Chilton et al., 2015; Pfeifer et al., 2017). Weight loss might be directly related to glucose excretion in the kidneys by SGLT2 inhibitors, resulting in noticeable calorie loss (Pereira and Eriksson, 2019). The blood pressure drop caused by SGLT2 inhibitors is commonly explained by natriuresis and diuretic effects (Baker et al., 2017; Mazidi et al., 2017). SGLT2 inhibitors, by promoting urinary output, lead to increased urinary sodium excretion and some plasma volume contraction, resulting in blood pressure reduction (Mazidi et al., 2017).

Similarly, ertugliflozin might also induce weight loss and depressurisation as a class effect of SGLT2 inhibitors. Our analysis showed that ertugliflozin reduced body weight by 1.77 kg compared with other therapies. Moreover, ertugliflozin reduced blood pressure by 2.57 mmHg on SBP and 1.15 mmHg on DBP from baseline compared with control groups. In prior meta-analyses, ertugliflozin was also observed to reduce body weight by 0.9–3.5 kg (Liu et al., 2019b; Liu et al., 2020b; Heymsfield et al., 2020) and blood pressure (mainly SBP) by 1.8–7.2 mmHg (Liu et al., 2019a; Liu et al., 2020a; Pratley et al., 2020). These results are consistent with our findings, confirming the reliability of our results.

In this meta-analysis, ertugliflozin generally exhibited good tolerability in patients with type 2 diabetes, with an overall safety profile in monotherapy and combination therapy. There was no significant difference in the incidence of any AEs, AEs related to study drug, and serious AEs, and deaths and discontinuations due to AEs. Of the respecified AEs of interest for ertugliflozin, the incidence of UTI, symptomatic hypoglycaemia, and hypovolaemia failed to detect any significant difference. Subgroup analyses on symptomatic hypoglycaemia showed that ertugliflozin use within 26 weeks presented a greater relative risk with a wide 95% CI. The possible explanation was the favourable glycaemic control in the first few months. Our previous study, which included 78 relevant publications, found that SGLT2 inhibitors increased the risk of GMI (RR: 3.71) (Shi et al., 2019), of which six articles reported the risk of GMI (RR: 4.69) related to ertugliflozin. This issue was considered a class effect of SGLT2 inhibitors. The current study also confirmed that ertugliflozin was associated with a higher GMI risk (RR: 4.004). A greater risk of GMI was observed in more than 26 weeks (RR: 4.962) compared with less than 26 weeks (RR: 1.977). Results of subgroups by follow-up suggested that GMI risk of ertugliflozin might be related to a more prolonged therapeutic course, although the interactive effect was not found (P _interaction of 0.112). Generally, patients with type 2 diabetes are at an increased risk of infection, especially those with obesity or increased atherosclerotic plaque development (Peleg et al., 2007). Euglycaemia, BMI control, and decreased arteriosclerotic cardiovascular disease risk should be standard practices to reduce the risk of GMI. Thus, more attention should be paid before using ertugliflozin and close monitoring indicators clinically. Additionally, personal hygiene education is recommended for patients initiating SGLT2 inhibitors (Williams and Ahmed, 2019). If GMI occurred unavoidably with a treatment course, discontinuation of SGLT2 inhibitors was unnecessary for the usually mild infection, which could be resolved with oral antifungal or antifungal cream (Engelhardt et al., 2021).

This meta-analysis quantified the efficacy and tolerability of ertugliflozin on glycaemia regulation, body weight control, blood pressure reduction, and incidence of AEs for type 2 diabetes patients. Thus, it seemed that ertugliflozin might be a favourable alternative to other SGLT2 inhibitors for type 2 diabetes inadequate responders.

Our study comprehensively assessed ertugliflozin based on medication tolerability and clinical efficacy (with a relatively larger sample size), contributing to more effective clinical decision-making. The results of subgroups analysis and sensitivity analysis showed a basically consistent trend. Additionally, economic factors also matter when making clinical decisions for local medical providers. We performed a simple cost-effectiveness analysis of ertugliflozin to evaluate whether ertugliflozin had a favourable economic advantage (Supplementary Material). However, our study has some limitations. First, several outcomes showed moderate to high heterogeneity, and the baseline characteristics of included trials varied (different dosages, controls, or duration of follow-up), despite subgroup analyses and sensitivity analyses being conducted. Second, owing to the limited duration of follow-up, which ranged from 12 to 52 weeks mainly, a certain bias might exist undeniably, and the safety outcomes of ertugliflozin need a further long-term observation. Finally, a comparison between ertugliflozin and other SGLT2 inhibitors was not performed. Therefore, more comprehensive studies across all SGLT2 inhibitors would be meaningful and worth investigating.