PRISMA-P (Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols) will be used to guide the protocol for this systematic review and NMA . A completed PRISMA-P checklist was utilized to guarantee the quality of the protocol, as indicated in Additional file1. In addition, the review will be published using the PRISMA extension statement for network meta-analysis. Our protocol has been registered with PROSPERO; any amendments and their reasons will be added to the platform.
Types of participants
We will include studies with patients diagnosed with type 2 diabetes using the criteria formulated by WHO and who are 18 years old and taking DPP-4is or other active antidiabetic medications, regardless of gender, race, or nationality.
Types of interventions and comparators
We will take into account the research that evaluated the following therapies involving DPP-4is (mainly including Sitagliptin, Saxagliptin, Vildagliptin, Linagliptin, and Alogliptin) compared with placebo or other hypoglycemic drugs (like Biguanides, Sulfonylureas, Thiazolidinediones, α-glycosidase inhibitors, SGLT-2 inhibitor, and DLP-1 receptor agonist) in patients with T2DM.
Types of studies
Only randomized controlled trials (RCTs) will be enrolled in this review, which compared DPP-4i therapies with other active hypoglycemic medications or placebo in T2DM patients with a minimum intervention length of 4 weeks. Either blinding or open-label studies will be considered. Furthermore, the literature will only be available in English. Other types of studies, such as in vivo or in vitro studies, case reports, reviews, and non-RCTs, will be eliminated.
The primary outcome of our study is pancreatic safety events, such as pancreatitis and pancreatic cancer. The secondary outcome of interest will include the change of pancreatic enzyme from baseline.
Medline, Embase, Pubmed, Web of Science, the Cochrane Central Register of Controlled Trials (CENTRAL) were searched from their inception dates to August 10th, 2021. Clinical trial registries (such as www.ClinicalTrials.gov) were also searched for unpublished trials. Moreover, we will check the systematic reviews and meta-analyses to search for further relevant studies and contact the authors to obtain incomplete data. Two researchers will perform the search, and all discrepancies will be resolved by discussion with a third reviewer. To find qualifying studies in PubMed, we will use the following search strategy, and it will be tweaked for each database. Table 1 shows the search strategies and retrieval type which will be employed in this investigation.
Table1 Search strategies for PubMed
Diabetes Mellitus, Type 2 [Mesh]
Diabetes Mellitus, Type 2 [Title/Abstract] OR Diabetes Mellitus, Noninsulin-Dependent [Title/Abstract] OR Diabetes Mellitus, Ketosis-Resistant [Title/Abstract] OR Diabetes Mellitus, Ketosis Resistant [Title/Abstract] OR Ketosis-Resistant Diabetes Mellitus [Title/Abstract] OR Diabetes Mellitus, Non Insulin Dependent [Title/Abstract] OR Diabetes Mellitus, Non-Insulin-Dependent [Title/Abstract] OR Non-Insulin-Dependent Diabetes Mellitus [Title/Abstract] OR Diabetes Mellitus, Stable [Title/Abstract] OR Stable Diabetes Mellitus [Title/Abstract] OR Diabetes Mellitus, Type II [Title/Abstract] OR NIDDM [Title/Abstract] OR Diabetes Mellitus, Noninsulin Dependent [Title/Abstract] OR Diabetes Mellitus, Maturity-Onset [Title/Abstract] OR Diabetes Mellitus, Maturity Onset [Title/Abstract] OR Maturity-Onset Diabetes Mellitus [Title/Abstract] OR Maturity Onset Diabetes Mellitus [Title/Abstract] OR MODY [Title/Abstract] OR Diabetes Mellitus, Slow-Onset [Title/Abstract] OR Diabetes Mellitus, Slow Onset [Title/Abstract] OR Slow-Onset Diabetes Mellitus [Title/Abstract] OR Type 2 Diabetes Mellitus [Title/Abstract] OR Noninsulin-Dependent Diabetes Mellitus [Title/Abstract] OR Noninsulin Dependent Diabetes Mellitus [Title/Abstract] OR Maturity-Onset Diabetes [Title/Abstract] OR Diabetes, Maturity-Onset [Title/Abstract] OR Maturity Onset Diabetes [Title/Abstract] OR Type 2 Diabetes [Title/Abstract] OR Diabetes, Type 2 [Title/Abstract] OR Diabetes Mellitus, Adult-Onset [Title/Abstract] OR Adult-Onset Diabetes Mellitus [Title/Abstract] OR Diabetes Mellitus, Adult Onset [Title/Abstract]
#1 OR #2
Dipeptidyl-Peptidase IV Inhibitors [Mesh]
Dipeptidyl-Peptidase IV Inhibitors [Title/Abstract] OR Dipeptidyl Peptidase IV Inhibitors [Title/Abstract] OR DPP-4 Inhibitor [Title/Abstract] OR DPP 4 Inhibitor [Title/Abstract] OR Inhibitor, DPP-4 [Title/Abstract] OR DPP-IV Inhibitor [Title/Abstract] OR DPP IV Inhibitor [Title/Abstract] OR Inhibitor, DPP-IV [Title/Abstract] OR DPP-4 Inhibitors [Title/Abstract] OR DPP 4 Inhibitors [Title/Abstract] OR DPP-IV Inhibitors [Title/Abstract] OR DPP IV Inhibitors [Title/Abstract] OR Gliptin [Title/Abstract] OR Dipeptidyl Peptidase 4 Inhibitor [Title/Abstract] OR Dipeptidyl-Peptidase IV Inhibitor [Title/Abstract] OR Dipeptidyl Peptidase IV Inhibitor [Title/Abstract] OR Inhibitor, Dipeptidyl-Peptidase IV [Title/Abstract] OR Dipeptidyl-Peptidase 4 Inhibitor [Title/Abstract] OR Inhibitor, Dipeptidyl-Peptidase 4 [Title/Abstract] OR Dipeptidyl-Peptidase 4 Inhibitors [Title/Abstract] OR Dipeptidyl Peptidase 4 Inhibitors [Title/Abstract] OR Gliptins [Title/Abstract] OR DPP4 Inhibitor [Title/Abstract] OR Inhibitor, DPP4 [Title/Abstract] OR DPP4 Inhibitors [Title/Abstract] OR Sitagliptin [Title/Abstract] OR Saxagliptin [Title/Abstract] OR Vildagliptin [Title/Abstract] OR Linagliptin [Title/Abstract] OR Alogliptin [Title/Abstract]
#4 OR #5
Randomized Controlled Trial [Publication Type] OR RCT [Title/Abstract] OR randomized [Title/Abstract] OR controlled [Title/Abstract] OR placebo [Title/Abstract]
pancrea* [Title/Abstract] OR safety [Title/Abstract]
#3 AND #6 AND #7 AND #8
Study selection process
Two reviewers will conduct literature screening independently. They will screen titles and abstracts of all the retrieved records to find potentially eligible studies and then examine the entire text to select studies that match the inclusion criteria. After rechecking the source papers, debate among the two reviewers, and adjudicating by the third, disagreements will be resolved with an entire consensus before inclusion. A PRISMA flow diagram will outline the research selection procedure and reasons for exclusions , and it was indicated in Additional file2. If a study was published in duplicate, we would choose the version with the most detailed content and data.
We will ensure a standardized data extraction template in advance, and two investigators will independently extract the data from all eligible studies in duplicate. Any discrepancies will be resolved by consensus or arbitrated by the third. The following trial information will be extracted: study author, publication year, study design, sample size, types of intervention and control, background therapy, and funding source. Moreover, population characteristics containing mean age, gender proportions, racial, duration of disease, length of the trial, loss of follow-up, and baseline level of HbA1c will also be collected. What is more, the following outcome measures will be gathered: Relative Risk of pancreatic safety event (involving pancreatitis and pancreatic cancer) and post-intervention values or changes from baseline with corresponding standard deviations for the pancreatic enzyme. Additionally, if the supplied data is incomplete, we will contact the authors for more information.
Risk of bias assessment
Two assessors will assess the risk of bias and quality of all included studies, with the third reviewer participating in the discussion as needed. They will pilot numerous samples before the formal evaluation to get to an agreement on assessment standards. Cochrane risk of bias tool (ROB tool) will be used to assess the risk of bias for trials concerning seven points: the judgment of the random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other sources of bias, which are graded as low, high or unclear risk of bias. Furthermore, we will try to gather all the information related to this review to control publication bias, and if more than 10 trials are included, we will assess the reporting bias through a funnel plot.
The collected evidence will be interpreted by the GRADE Working Group approach for rating the quality of treatment effect estimates from NMA. According to GRADE, the assessment requires estimates from direct, indirect, and combined evidence from direct and indirect sources, as well as quality ratings for the direct and indirect comparisons, which are primarily used to assess the quality of each piece of evidence in five areas: limitation, imprecision, inconsistency, indirectness, and publication bias.
Data Synthesis and Analysis
A descriptive overview of available data will be compiled and reported, namely trial and population characteristics, interventions, results, and risk of bias evaluations. We will plot a network regarding the evidence structure of direct comparisons, in which the size of nodes will be proportional to the sample size of each intervention, and the thickness of the lines will be proportional to the cumulative number of RCTs for each pairwise comparison. We will also use the contribution plot to present the influence of each direct piece of evidence. Both traditional pairwise meta-analysis and the network meta-analyse will be conducted. All quantitative analyses will be carried out using Stata 16 and WinBUGS.
A random effects model will be employed for traditional pairwise meta-analyses if the included studies are of heterogeneity. The I2 value will measure the degree of heterogeneity. Relative Risk (RR) for dichotomous data will be calculated as effect measures with 95% confidence intervals, and continuous data will be presented as weighted mean difference (WMD) with 95% confidence intervals. If different scales are involved in studies, we will use standardized mean differences (SMD) to present continuous data to eliminate its effect on results. Besides, sensitivity analysis will be performed to validate the stability of the results or exclude studies with a high risk of bias.
NMA is an upgrade of classical meta-analysis, which can simultaneously combine direct and indirect analysis, estimate the efficacy of interventions through common comparators even if they have not been investigated head to head in randomized clinical trials, and thus analyze the effects of multiple interventions compared with each other.
Before the analysis, examine the assumptions of consistency, heterogeneity, and similarity of the included studies within and across connections in the entire network of interventions to determine whether direct and indirect evidence is reasonable. An approach of loop inconsistency is used to evaluate the presence of inconsistency in each closed loop in which an intervention effect measured using an indirect comparison is not equivalent to the intervention effect measured using direct comparison. The I2 statistic will quantify the global heterogeneity, and the Cochran Q test and its P-value will also be used to evaluate the heterogeneity. The predictive interval graph and confidence interval will be plotted to present the influence of heterogeneity on each pairwise comparison. Furthermore, the similarity will be evaluated by comparing the critical clinical and methodological characteristics that can influence the effects of studies between two sets.
Suppose the above assumptions of consistency, heterogeneity, and similarity are reasonable. In that case, NMA will be performed in a random effects model, by a Bayesian framework using Markov Chains Monte Carlo (MCMC) methods with WinBUGS 1.4.3 and Stata 16 software to synthesize all the available evidence. The results of all pairwise comparisons will be reported as appropriate effect values with 95% confidence intervals. The network geometry of interventions will be plotted to present their concise characteristics, while forest plots and contribution figures will be used to show the combined effect value. Besides, we will display values of the Surface Under the Cumulative Ranking (SUCRA) curve for each intervention as well as rankings of effects. And convergence will be assessed by the Gelman Rubin statistic method and inspection of Monte Carlo errors.
Subgroup analyses and meta-regression
We will conduct subgroup analysis and meta-regression to explore the source of heterogeneity if significant heterogeneity and sufficient data exist. Subgroup analyses will combine effect sizes for each subgroup from the perspectives of clinical or methodological heterogeneity, including age, duration of disease, length of the trial, loss of follow-up, the baseline level of HbA1c, and quality of study. And in meta-regression, regression analysis will be used to explore the influence of some covariables on the merger effect in meta-analysis.