The proposed systematic review was registered in the PROSPERP database(CRD42021253822).
Randomized controlled trials fulfilling the following requirements will be eligible in our network analysis:
Patients aged over 18 and with T2DM will be included, regardless of gender, or ethnic origins. The diagnosis of T2DM should have been established using standard criteria. Pregnant or breastfeeding female patients and patients with T2DM on Maintenance Hemodialysis will be excluded.
DPP-4 inhibitors, including sitagliptin, saxagliptin, vildagliptin, lindagliptin, alogliptin, omarigliptin and trelagliptin.
DPP-4 inhibitors compared with each other, traditional antidiabetic agents includes α-glucosidase inhibitors, sulphonylureas, non-sulfonylureas and thiazolidinediones, or placebo
Studies with clinically relevant outcomes will be included. Interested outcomes include changes in HbA1c concentration, weight and FPG from baseline to the end of treatment and proportions of patients in achieving HbA1c < 7.0%, patients with treatment-emergent adverse events leading to treatment discontinuation and patients with hypoglycaemia.
Parallel-group randomized controlled trials (RCTs) will be included, regardless of the blinding design. To investigate long-term efficacy and safety of different kinds of DPP-4 inhibitors, RCTs with duration of trial＜24 weeks will be excluded.
English language studies will be searched in the electronic databases(PubMed, the Cochrane Library and EMBASE), without time restriction. In addition, complementary sources will be used, including reference lists from relevant studies assessed for eligibility and researches registered in ClinicalTrials.gov.
The search strategy will be defined by (i) patient’s condition (i.e. T2DM), (ii) studied drugs (i.e. sitagliptin, saxagliptin, vildagliptin, lindagliptin, alogliptin, omarigliptin and trelagliptin) and (iii) study design. For the latter, the specific search strategy of RCTs defined in the Cochrane Handbook will be used. Literature reviews will be excluded. Each parameter will be defined by several MeSH terms and/or free terms in titles and abstracts.
All identified studies will be imported into the Endnote reference manager for study selection. Firstly, Automatic and manual de-duplication will be carried out. Secondly, titles and abstracts of identified references are screened by two reviewers independently. Thirdly, The eligible references will be reviewed on full texts.Conflicts will beresolved by discussion among the team. A PRISMA flowchart will be drawn to document the screening process.
Data extraction and Risk of bias assessment
Two investigators will extract data using pre-specified forms with epidata software and independently assess the accuracy of abstractions and resolve any discrepancies by consensus after discussion with the third investigator. The following characteristics will be extracted for included studies: first author, year of publication, study design, number of participants enrolled, study design, total study duration, sequence generation, allocation sequence concealment, participant characteristics (age, gender, BMI, total number, ethnicity), interventions and outcomes(change in HbA1c, weight and FPG from baseline to the end of treatment, proportions of patients in achieving HbA1c < 7.0%, treatment-emergent adverse events leading to treatment discontinuation and with hypoglycaemia.
Risk of bias assessment was conducted by two reviewers independently, according to the Cochrane Collaboration risk of bias tool across 5 domains (sequence generation, allocation concealment, blinding, detection bias, and attrition bias).
Data synthesis and statistical analysis
The main objective of this data synthesis is to compare the effectiveness and tolerability of different kinds of DPP-4 inhibitors. NMA is useful in synthesizing all available data, including direct and indirect comparison. Statistical heterogeneity was assessed by I² statistic. If the source of heterogeneity can't be located, single literature exclusion method is used to find out the literature that has great influence on the study of heterogeneity. When direct evidence is available from at least two studies, we will perform pairwise meta-analysis in a random-effects model by STATA 13.0 software for each outcome. A Bayesian network meta-analysis was conducted for indirect and mixed comparisons using WinBUGS 1.4.3 software. And effect estimates(RR, MD) for the outcome measures will be calculated.
Assessment of transitivity and inconsistency
To evaluate the assumption of transitivity, we will compare the similarity of the included populations and study settings in terms of age, gender, BMI and Basic hypoglycemic agents.
A design-by-treatment interaction model will be applied to assess potential design inconsistency and loop inconsistency.
Exploratory subgroup analyses will be performed, if a sufficient number of studies are identified. Analyses will focus on the difference in efficacy, safety and tolerability among different ethnic groups and between once-weekly and once-daily preparation.