Head-to-Head Comparison of the Hypoglycemic Efficacy and Safety Between Dipeptidyl Peptidase-4 Inhibitors and α-Glucosidase Inhibitors in Patients With Type 2 Diabetes Mellitus: A Meta-Analysis of Randomized Controlled Trials

Background: The α-glucosidase inhibitors (AGIs) are commonly prescribed in Asian patients with type 2 diabetes mellitus (T2DM), but with a high incidence of gastrointestinal side effects. This study was aimed to compare the efficacy and safety of dipeptidyl peptidase-4 (DPP4) inhibitors and AGIs in T2DM patients in a meta-analysis. Methods: Randomized controlled trials were identified via systematic search of PubMed, Embase, and Cochrane’s Library databases from inception to February, 2019. Meta-analyses were performed via a random or a fixed effect model according to the heterogeneity. Results: Eighteen studies with a total of 4,051 patients with T2DM were included. The DPP4 inhibitors were associated with lower reduction of glycosylated hemoglobin (HbA1c) as compared with AGIs [weighed mean difference (WMD): −0.37%, p < 0.001]. Subgroup analyses indicated that the benefit of DPP4 inhibitors as compared with AGIs on HbA1c were independent of study design, scale, baseline HbA1c, with or without concurrent medications, or follow-up durations. Moreover, compared to AGIs, DPP4 inhibitors was associated with lower reductions of fasting blood glucose (WMD: −0.53 mmol/L, P < 0.001) and postprandial glucose at 2h (WMD: −0.60 mmol/L, P = 0.04), moderately increased body weight (WMD: 0.34 kg, P = 0.02), and decreased risk of gastrointestinal adverse events [risk ratio (RR): 0.48, P < 0.001], but unaffected risk of symptomatic hypoglycemia (RR: 0.96, P = 0.90). Conclusions: The DPP4 inhibitors are superior to AGIs in T2DM patients for better glycemic control and lower risks of gastrointestinal side effects.


MATERIALS AND METHODS
This meta-analysis was designed, performed, and presented in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement (Moher et al., 2009) and Cochrane's Handbook guideline (Higgins and Green, 2011).

Database Search
We performed the initial electronic database search of the PubMed, Embase, and Cochrane's Library from inception to February, 2019, via the combination of the following terms: 1) "DDP4" OR "DDP-4" OR "DPP4 inhibitors" OR "sitagliptin" OR "vildagliptin" OR "linagliptin" OR "saxagliptin" OR "alogliptin" OR "dutogliptin, " 2) "alpha-GIs" OR "alpha-glucosidase inhibitors" OR "acarbose" OR "voglibose" OR "miglitol, " and 3) "random" OR "randomly" OR "randomized" OR "randomised. " The search was limited to studies in human, and no restriction was applied for the language of publication. The final literature search was performed on February 12th, 2019.

Inclusion and Exclusion Criteria
Studies were included if they met the following criteria: 1) designed as a parallel group RCT; 2) included adult patients with T2DM; 3) allocated to an intervention group of oral DPP4 inhibitor and a control group of AGI, with or without coadministration of other oral antidiabetic medications, such as metformin and/or sulfonylureas et al.; 4) with a follow-up duration of at least 8 weeks; and 5) reported at least one of the following outcomes: changes of glycosylated hemoglobin (HbA1c), fasting blood glucose (FBG), PPG at 2h, and body weight from baseline, and the incidences of symptomatic hypoglycemia and any gastrointestinal adverse events (GIAEs) in patients of both groups. Reviews, crossover trials, preclinical studies in animals, and repeated reports of already included RCTs were excluded.

Data Extraction and Quality Evaluation
We extracted the following study characteristics of each RCT: 1) location of the study; 2) design characteristics: single-blind, double-blind, or open label; 3) patients' characteristics: number, age, gender, baseline HbA1c, body mass index (BMI); 4) details of background OADs; 5) regimens of DPP4 inhibitors and AGIs; and 6) follow-up durations. The quality of the included RCT was evaluated by the Cochrane's risk of bias tool (Higgins and Green, 2011), which is based on the following seven domains: random sequence generation, allocation concealment, blinding in performance, blinding in outcome detection, incomplete outcome data, reporting bias, and the potential risk of other bias. The processes of database search, study identification, data extraction, and quality evaluation were independently performed by two authors. Discussion with a third author was indicated when discrepancies occurred.

Statistical Analyses
The statistical analyses were performed with RevMan software (version 5.1; Cochrane Collaboration, Oxford, UK) and Stata software (version 12.0; Stata Corporation, College Station, TX). Continuous variables were analyzed using weighed mean difference (WMD) and 95% confidence interval (CI), while categorized variables were analyzed using risk ratio (RR) and 95% CI. The heterogeneity among the included RCTs was evaluated by the Cochrane's Q test (Higgins and Green, 2011) and a P < 0.10 indicating significant heterogeneity. We also used I 2 statistic, which describes the percentage of total variation across studies that is due to heterogeneity rather than chance (Higgins et al., 2003), as an indicator of heterogeneity, and an I 2 > 50% indicates significant heterogeneity. A fixed effect model was used to pool the results of individual study during meta-analysis, if the heterogeneity was not significant according to the results of the Cochrane's Q test; otherwise, a random effect model was applied (Ma et al., 2018). The primary outcome of the study was the difference of changes of HbA1c between patients treated with DPP4 inhibitors and AGIs. Predefined subgroup analyses were performed to evaluate the potential influence of study characteristics including blinding, number of patients, baseline HbA1c, with or without background OADs, follow-up durations, and individual drugs of DPP4 inhibitor or AGI applied on the primary outcome. Medians of the continuous variables were used as cutoff value for stratification in subgroup analyses (Moher et al., 1998). The potential publication bias for the meta-analysis of each outcome was evaluated by the visual inspection of the symmetry of the funnel plots, as well as the Egger's regression test (Egger et al., 1997). A P value < 0.05 indicates statistical significance.

Database Search Result
The process of database search was summarized in Figure 1. Briefly, 450 studies were obtained via the initial database search, and 415 studies were excluded based on analyses of titles and abstracts, mostly because these studies were irrelevant to current study objective. Of the remaining 35 studies that underwent fulltext review, 17 studies were excluded because two of them did not include T2DM patients, two evaluated a combined effect of DPP4 inhibitors and AGIs, six were with follow-up durations <8 weeks, three were subgroup or extension studies of included RCTs, two were not with available outcome data, and another two were repeated reports of the included RCTs. Finally, 18 RCTs were included (Pan et al., 2008;Iwamoto et al., 2010a;Iwamoto et al., 2010b;Seino et al., 2011;Kawamori et al., 2012;Okada et al., 2013;Kobayashi et al., 2014;Mikada et al., 2014;Nakamura et al., 2014;Oe et al., 2015;Wang et al., 2015;Yokoh et al., 2015;Fujitani et al., 2016;Matsushima et al., 2016;Mori et al., 2016;Du et al., 2017;Koyama et al., 2018;Parthan et al., 2018).

Comparison Between DPP4 Inhibitors and AGIs on HbA1c in T2DM
All of the 22 comparisons reported the efficacy of DPP4 inhibitors versus AGIs on HbA1c changes in T2DM patients. Significant heterogeneity was detected among the included RCTs (P for Cochrane's Q test < 0.001, I 2 = 77%). Pooled results with a random effect model showed that treatment with DPP4 inhibitors was associated with a significant lower reduction of HbA1c as compared with AGIs (WMD: −0.37%, 95% CI: −0.48 to −0.26, p < 0.001; Figure 2A). Subsequent subgroup analyses indicated a significant lower reduction of HbA1c in patients treated with DPP4 inhibitors than AGIs, which was independent of the study design, patient number of the included RCTs, HbA1c level at baseline, with or without background OADs, and follow-up durations (P for efficacies all < 0.05; Table 3). Moreover, the results of stratified analyses indicated that the reduction of HbA1c by DPP4 inhibitors compared with AGIs was more remarkable in double-blinded studies than open-label trials (P for subgroup difference = 0.005), and in patients without background OADs than those with background OADs (P for subgroup difference = 0.01). In addition, a trend of lower reduction of HbA1c was observed in studies with longer follow-up duration (52 weeks) than those with shorter follow-up duration (P for subgroup difference = 0.07). Stratified analyses according to the specific medications of DPP4 inhibitors and AGIs used in the RCTs showed that the benefits of DPP4 inhibitors versus AGIs on HbA1c was significant in studies with alogliptin, linagliptin, and sitagliptin, but not significant in studies with saxagliptin or vildagliptin, while significant in studies with voglibose, but not significant in studies with acarbose or miglitol.

Publication Bias
Funnel plots for meta-analyses of the effects on changes of HbA1c, FBG, PPG-2h, and body weight, and the risks of symptomatic hypoglycemia and any GIAEs, were presented in Figure 4A-F, which were symmetric on visual inspection, indicating no significant publication biases. Results of Egger's regression tests also demonstrated no significance in publication biases (P for Egger's regression tests = 0.38, 0.52, 0.61, 0.58, 0.48, and 0.21 respectively).

DISCUSSION
In this updated meta-analysis of 18 RCTs with 4,051 T2DM patients, we found that DPP4 inhibitors confer better glycemic control than AGIs, as evidenced by significantly reduced HbA1c, FBG, and PPG-2h by DPP4 inhibitors compared to AGIs. Moreover, treatment with DPP4 inhibitors is associated with much less GIAEs and unaffected hypoglycemia, but a modest increased body weight as compared with AGIs. These results indicate that for T2DM patients, DPP4 inhibitors are superior to AGIs for better glycemic control but lower risks of GIAEs. A previous meta-analysis (Cai et al., 2015) including nine RCTs showed that treatment with DPP4 inhibitors was associated with a lower reduction of HbA1c and a less incidence of gastrointestinal discomfort as compared with AGIs, while the other meta-analysis (Gao et al., 2018) including 11 RCTs published up to 2016 showed that AGI treatment was associated with a significantly lower reduction in HbA1c than DPP4 inhibitors. Many related RCTs have been published since the previous meta-analyses (Oe et al., 2015;Fujitani et al., 2016;Matsushima et al., 2016;Mori et al., 2016;Du et al., 2017;Koyama et al., 2018;Parthan et al., 2018), which were included in this updated metaanalysis. Our meta-analysis has a few strengths as compared with the previous meta-analysis of the similar topic (Cai et al., 2015). Firstly, our updated meta-analysis included 18 RCTs of 4,051 T2DM patients, which is much larger than those of the  previous ones, which enables us to perform subgroup analyses to confirm the robustness of our findings regarding the superiority of DPP4 inhibitors on glycemic control over AGIs. Results of the subgroup analysis confirmed that treatment with the DPP4 inhibitors was associated with significantly reduced HbA1c as compared with AGIs. The superiority of glycemic control of DPP4 inhibitors over AGIs was independent of study design, patient number of the included RCTs, HbA1c level at baseline, with or without background OADs, and follow-up durations, indicating the stability of the findings. Secondly, results of our subgroup analysis also suggested some interesting findings. For example, we found that the reduction of HbA1c by DPP4 inhibitors compared with AGIs was more remarkable in doubleblinded studies than open-label trials. Since open-label trials are vulnerable to various bias, these results indicated that the actual hypoglycemic efficacy of DPP4 inhibitors compared with AGIs may be stronger than the overall results (−0.30% in previous meta-analysis and−0.37% in our study). In addition, we found a trend of lower reduction of HbA1c by DPP4 inhibitors compared with AGIs in studies with longer follow-up duration (52 weeks) than those with shorter follow-up duration. These results suggest the optimal durability of the hypoglycemic efficacy of DPP4 inhibitors. Thirdly, we found that despite of significant decreased FBG by DPP4 inhibitors compared with AGIs, treatment with DPP4 inhibitors also significantly decreased PPG as compared with AGIs in T2DM patients, which is not reported in previous meta-analysis (Cai et al., 2015). This is of clinical significance since AGIs are often considered as the OADs of choice to lower PPG because of its direct pharmacological action, which is to inhibit the absorption of carbohydrates during the meals (Liu and  Ma, 2017;Shimabukuro et al., 2017). Based on our results, DPP4 inhibitors should be considered in patients with high PPG levels. Moreover, meal intake-related glycemic excursion has been recognized as an important contributor to glycemic variability, which has been validated as another risk factor for morbidity and mortality in T2DM patients independent of glycemic control (Gorst et al., 2015;Nusca et al., 2018). Therefore, our finding indicated that DPP4 inhibitors may be an alternative category of OADs to stabilize glycemic variability, although studies are needed to confirm our findings. The potential mechanisms underlying the potential superiority of DPP4 inhibitors on glycemic control to AGIs may be primarily explained by the differences of the pharmacological mechanisms of the two categories of OADs. As mentioned previously, the AGIs mainly lower the PPG via inhibiting the absorption of carbohydrates during the meal (Liu and Ma, 2017). However, DPP4 inhibitors could exert both the fasting and postprandial hypoglycemic efficacies via reducing glucagon secretion during fasting and stimulating the glucose-dependent insulin secretion during the meal . Besides, DPP4 inhibitors may also be associated with better patient compliance, since most DPP4 inhibitors are to be taken once daily, while all of the AGIs are to be taken three times daily. This could be an alternative mechanism underlying the better glycemic control as observed in patients taking DPP4 inhibitors compared to those taking AGIs.
Our results confirmed the findings of the previous studies that treatment with DPP4 inhibitors in T2DM patients is associated with significantly decreased risk of GIAEs and unaffected hypoglycemia, but a moderate increased body weight as compared with AGIs (Cai et al., 2015;Wu et al., 2017). Because high incidence of GIAEs is a common reason for the discontinuation of AGIs in T2DM patients, based on our findings, DPP4 inhibitors should be recommended to patients who are unable to tolerate AGIs as an initial therapy for T2DM or in combination with other OADs. The mechanisms underlying the unaffected risk of hypoglycemia in patients taking DPP4 inhibitors as compared with those taking AGIs may rely on the glucose-dependent insulin stimulatory effect of DPP4 inhibitors (Flock et al., 2007;Vardarli et al., 2011). Moreover, the modest effect of body weight gain in DPP4 inhibitors as compared with AGIs could be explained by the increased GIP level after treatment of DPP4 inhibitors. The GIP has been recognized as a peptide which could enhance the uptake of lipids into the adipocytes in the overnutrition condition (Thondam et al., 2017). However, from our point of view, the gain of body weight after treatment with DPP4 inhibitors as compared with AGIs is moderate (0.34 kg), which does not outweigh the potential benefits of DPP4 inhibitors on glycemic control and GIAE risk.
Our study has limitations. Firstly, significant heterogeneities exist for meta-analysis comparing the efficacies of DPP4 inhibitors and AGIs on HbA1c, FBG, PPG-2h, body weight, and incidence of GIAEs. The differences of study characteristics, such as the baseline HbA1c, T2DM durations, comorbidities, and concurrent medications including OADs, may contribute to the heterogeneities. While the influences of these factors could not be determined since the stratified outcomes according to the above characteristics were not reported in the included RCTs. Moreover, we performed the meta-analysis on the study level instead of on the individual patient level because we did not have access to the individual patient data. This prevented us from performing intensified analyses of the heterogeneity. In addition, as compared with AGIs, the influences of DPP4 inhibitors on clinical outcomes, such as incidences of cardiovascular events and overall mortality, remained to be determined. Finally, the findings of the meta-analysis were almost all based on the RCTs in Asian patients. Since carbohydrates account for a larger proportion of diet structures of the Asians, AGIs are often prescribed in Asian T2DM patients (Zhu et al., 2013). Whether DPP4 inhibitors are superior to AGIs in glycemic control in other ethnic group should be investigated in the future.

CONCLUSIONS
In conclusion, our meta-analysis showed that DPP4 inhibitors are superior to AGIs in T2DM patients for better glycemic control but lower risks of GIAEs.

AUTHOR CONTRIBUTIONS
ZL and LZ carried out the acquisition and analysis of data, and drafting the manuscript. JY performed the drafting and revising of the manuscript. LY participated in the design and helped to revise the manuscript. All authors have read and approved the final version of the manuscript prior to submission.