Article Text

Download PDFPDF

Effect of adjunctive glucose-lowering drugs on body weight in people with type 1 diabetes: a systematic review and network meta-analysis protocol
  1. Christian Laugesen1,
  2. Ajenthen G. Ranjan1,
  3. Signe Schmidt1,
  4. Lauge Neimann Rasmussen2,
  5. Ole Nørgaard2,
  6. Robin Christensen3,4,
  7. Kirsten Nørgaard1
  1. 1Clinical Research, Steno Diabetes Center Copenhagen, Gentofte, Denmark
  2. 2Education, Diabetes Knowledge Center, Steno Diabetes Center Copenhagen, Gentofte, Denmark
  3. 3Musculoskeletal Statistics Unit, the Parker Institute, Frederiksberg and Bispebjerg Hospital, Copenhagen, Denmark
  4. 4Research Unit of Rheumatology, Department of Clinical Research, University of Southern Denmark, Odense University Hospital, Odense, Denmark
  1. Correspondence to Dr Christian Laugesen; christian.laugesen{at}


Introduction Obesity increases the risk of comorbidities and diabetes-related complications and, consequently, efforts to prevent and reduce excess weight in people with type 1 diabetes are essential. The aim of this systematic review and network meta-analysis is to assess the effect of adjunctive glucose-lowering drugs on body weight and other important health outcomes in people with type 1 diabetes.

Methods and analysis This systematic review and network meta-analysis will include randomised controlled trials (RCTs) evaluating the use of adjunctive glucose-lowering drugs for treatment of people with type 1 diabetes. MEDLINE, EMBASE, Cochrane Central Register of Controlled Trials, Cochrane Database of Systematic Reviews, and WHO International Clinical Trials Registry Platform will be searched from inception to present. Key eligibility criteria include: RCT study design; adult participants with type 1 diabetes; treatment with a glucose-lowering drug for ≥24 weeks; and comparison of the intervention to placebo, usual care or another glucose-lowering drug. The primary outcome is change in body weight. Other major outcomes include change in HbA1c and total daily insulin dose and risk of hypoglycaemia and other adverse events. Dual study selection, data extraction and risk of bias assessment will be performed. Results from the meta-analysis will be presented as weighted mean differences for continuous outcomes and risk ratios for dichotomous outcomes. Sources of heterogeneity will be explored by subgroup and sensitivity analysis. A network meta-analysis for the primary outcome will be performed using an arm-based random-effects model based on the Bayesian framework while assessing for transitivity across studies and consistency between direct and indirect estimates. The overall quality of the evidence will be assessed using the Grading of Recommendations, Assessment, Development and Evaluation approach for each outcome.

Ethics and dissemination No ethical assessment is required. The results of this review will be disseminated through peer-reviewed publication and conference presentation.

PROSPERO registration number CRD42020158676

  • diabetes & endocrinology
  • statistics & research methods
  • therapeutics
  • clinical trials

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See:

View Full Text

Statistics from

Strengths and limitations of this study

  • This systematic review and network meta-analysis will comprehensively evaluate the effect on body weight, glycaemic control and adverse events of all adjunctive glucose-lowering drug classes for treatment of type 1 diabetes.

  • The thorough and transparent methodological approach undertaken will minimise the risk of possible biases. Quality of evidence will be assessed to provide confidence in the effect estimates.

  • To be able to assess meaningful changes in body weight, eligible studies are limited to those of interventions with a duration of ≥24 weeks.

  • Common to any meta-analysis, some study heterogeneity across and within drug classes may exist.



Several studies have demonstrated a concerning trend in the prevalence of obesity in people with type 1 diabetes.1–3 The Type 1 Diabetes Exchange registry and the Pittsburgh Epidemiology of Diabetes Complications Study found that around 50%–60% of the adult population with type 1 diabetes was either overweight or obese,1 4 and the prevalence of obesity among people with type 1 diabetes has been shown to increase at a faster rate compared with the general population.5 Weight management is an important health-related topic of interest as excess body weight increases the risk of several adverse health outcomes; there is solid evidence that obesity is associated with increased mortality, increased risk of cardiovascular disease and a lower health-related quality of life in the general population.6 7 Although less is known about the health consequences in people with type 1 diabetes, several studies indicate that increasing body mass index (BMI) is associated with increased mortality and risk of cardiovascular disease, heart failure and microvascular complications in this population.8–11 Consequently, efforts to prevent and reduce excess weight in people with type 1 diabetes are essential.

Several approaches to reduce weight have been investigated. Lifestyle interventions such as diets and exercise intensification are key therapeutic avenues to reduce body weight, but these can often be difficult to implement for the same reasons as in people without diabetes. Further, the risk of hypoglycaemia also constitutes a significant barrier for many people with type 1 diabetes with respect to implementing the lifestyle changes. As insulin induces weight gain and pump therapy is associated with lower total daily insulin dose (TDD),12 this would theoretically cause less weight gain. However, studies have not been able to demonstrate any significant differences in body weight between people using pump and injection therapy.13 14

Despite aforementioned strategies, weight management remains difficult. Therefore, additional approaches to managing body weight can be beneficial. In people with type 2 diabetes, some glucose-lowering agents have shown to reduce weight.15 Several studies have investigated the effects of various adjunctive therapies on glycaemic control and body weight in people with type 1 diabetes, and numerous systematic reviews have evaluated the outcomes of specific drug classes. However, a complete and comparative overview of the weight-reducing and glycaemic effects of agents across all adjunctive glucose-lowering drug classes is warranted.


The primary objective of this systematic review and meta-analysis is to assess the effect of adjunctive (non-insulin) glucose-lowering drugs on body weight in adults with type 1 diabetes. To assess the comparative effectiveness of each drug class with respect to the primary outcome, a network meta-analysis (NMA) using both direct and indirect trial evidence will be performed (if applicable). The secondary objectives are to assess the effect of adjunctive glucose-lowering drugs on glycaemic (HbA1c), non-glycaemic (change in TDD) and safety outcomes (risk of hypoglycaemia, diabetic ketoacidosis and serious adverse events (SAEs)).


The protocol was developed and reported according to the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA-P)16 and registered with the International Prospective Register of Systematic Reviews (PROSPERO). In the event of protocol amendments, these will be submitted to PROSPERO accompanied by a description of the change and rationale.

Patient and public involvement

No patients or public entities were involved in the conception of this systematic review and meta-analysis protocol.

Eligibility criteria

Study design

Only randomised controlled trials (RCTs) are eligible for inclusion, excluding cluster RCTs and controlled (non-randomised) clinical trials. Results from the first period of crossover studies will be included if they are reported explicitly and are otherwise in accordance with the inclusion criteria.

Report characteristics

Both abstracts and full-text articles will be considered for eligibility. No limitations regarding year of publication will be applied.

Records in languages other than English (LOEs) will be screened on title and abstract level. When abstracts are not available in English, Danish, Swedish, Norwegian or German (languages spoken by the review team), LOE abstracts will be translated using Google Translate ( LOE abstracts which are deemed relevant based on this assessment will be reported in the supplemental material of the review specifying the languages identified by Google Translate, but are excluded from the analysis unless the full text-publication is published in a language spoken by the review team (ie, Danish, Swedish, Norwegian, German and English). Although this limitation might introduce bias, our review team has limited language skills, does not have funding to hire professional translation facilities (see support statement above), and is not a part of an institution, for example, university, with informal access to people with the additional language skills potentially needed. The risk of bias due to this limitation will be discussed in the review.17 18


Studies examining adult (≥18 years) men and non-pregnant women with type 1 diabetes will be eligible for inclusion.


Studies examining treatment with any adjunctive (non-insulin) glucose-lowering drug (defined as pharmacological therapy primarily intended to lower blood glucose) in combination with insulin therapy for a duration of ≥24 weeks will be eligible for inclusion. Studies examining adjunctive glucose-lowering drugs with a simultaneous co-intervention (eg, diet or exercise intervention) will only be eligible for inclusion if the co-intervention is applied in all study arms.

Drug classes of a priori interest are based on the sources listed in the Search strategy section and include:

  • α-Glucosidase inhibitors.

  • Amylin mimetics.

  • Biguanides.

  • Bile acid sequestrants.

  • Dopamine-2 agonists.

  • Dipeptidyl peptidase-4 inhibitors.

  • Gastric inhibitory peptide analogues.

  • Glucagon-like peptide-1 receptor agonists.

  • Meglitinides.

  • Sodium-glucose co-transporter−1 and 2 inhibitors.

  • Somatostatin analogues.

  • Sulfonylureas.

  • Thiazolidinediones.


Studies comparing the experimental intervention with placebo, usual care (without placebo) or another glucose-lowering drug will be eligible for inclusion.


Eligibility of trials will not be restricted by outcome criteria. Thus, studies meeting the remaining eligibility criteria will be included irrespective of the outcomes reported; this will enable assessment of the risk of selective outcome reporting. The predefined outcome measurements of the meta-analysis are listed in the Outcomes section.


There will be no restrictions on type of setting.

Information sources

The following databases will be searched from inception to present:

  • MEDLINE (Ovid).

  • EMBASE (Ovid).

  • Cochrane Central Register of Controlled Trials.

  • Cochrane Database of Systematic Reviews, which covers WHO International Clinical Trials Registry Platform database and

In addition, reference lists of included studies will be used as sources to identify eligible studies. Study authors or organisations will be contacted for information about unpublished studies if required.

Search strategy

The search strategy was developed by two health information specialists with input from the remaining project team. The MEDLINE search strategy (table 1) will be adapted to the syntax and subject headings of the other databases.

Table 1

MEDLINE search (Ovid interface)

The glucose-lowering drugs (drug classes as well as individual drugs) used in the search strategy are based on (1) American Diabetes Association’s (ADA) ‘Pharmacological Approaches to Glycaemic Treatment—Standards of Medical Care in Diabetes’ (2020),19 (2) drugs listed as diabetes medicine in the European Medicines Agency’s public assessment reports20 and on,21 22 and (3) authors’ subject matter expertise. The intervention criteria used for selection of studies are based on the definitions listed in the Eligibility criteria section. Thus, eligible interventions are not limited to the drugs included in the search strategy.

Study records

Data management

Literature search results will be uploaded to EPPI-Reviewer V.4.023, which will be used for screening. The health information specialists will provide members of the review team with adequate training in the software and support during the review process.

Selection process

The search strategy described will be used to obtain study titles and abstracts for potentially eligible trials. Two reviewers will independently screen titles and abstracts to identify studies for full-text assessment and subsequently determine whether each study meets the eligibility criteria by assessing full-text articles. Reasons for excluding studies after full-text review will be documented. Disagreement will be resolved first by discussion and then by consulting a third review author for arbitration. In studies where a subset of the included participants meets the inclusion criteria, data deriving separately from the eligible subgroup will be included. If subgroup data are not reported, study authors will be contacted with enquires to share subgroup data if available. In case it is not possible to gain access to the relevant subgroup data, the study is excluded. A PRISMA flow diagram24 will be produced to document and transparently illustrate the selection process.

Data collection process

Baseline characteristics and outcome data will be collected independently by two reviewers using predefined and piloted extraction forms. Appropriate software (eg, Microsoft Excel or EPPI-Reviewer) will be used to collect the data. If more than one publication reports on a specific trial, reports will be collected and grouped, and relevant data from each report will be used in the analyses. Any discrepancy between published versions will be highlighted. If outcomes are not reported as defined in this protocol, conversion to the relevant metric is not statistically possible, and it is not possible to obtain these data after contact to the study authors, the study will still be included in the systematic review (ie, qualitative synthesis).

Data items

The following data items will be collected: publication data (title, first author, year and source of publication, identifier (NCT), source of funding), trial design, baseline characteristics of the study population (sample size, sex, age, duration of type 1 diabetes, body weight, BMI, HbA1c, total daily insulin dose, treatment modality), details of the intervention(s) (generic drug name, dose, frequency, treatment duration), the comparator used and the outcome measures (listed below). Data items needed for assessment of the risk of bias will also be collected. For crossover studies, data will be extracted for the first period only because of possible carry-over effects.


Primary outcome:

  • Change in body weight (or BMI) from baseline to end of the intervention.

Other major outcomes:

  • Change in HbA1c from baseline to end of the intervention.

  • Risk of mild hypoglycaemia.

  • Risk of severe hypoglycaemia.

  • Change in TDD (or TDD/kg body weight) from baseline to end of the intervention.

  • Risk of diabetic ketoacidosis (DKA).

  • Risk of SAEs.

  • Drop-out rate.

  • Withdrawals due to adverse events.

The criteria used for the classification of mild/severe hypoglycaemia, DKA and SAEs will follow the definitions applied by the respective study authors. In cases where outcomes are reported on various time points during the study (eg, at 26 and 52 weeks), the outcome closest to the end of the active intervention will be used (ie, only data respecting the original randomised design will be included; excluding early rescues and open label extension phases).

Outcome measures were chosen based on clinical relevancy and use of outcome reporting from relevant literature. In the section ‘Non-insulin Treatments for Type 1 Diabetes’ of ADA’s ‘Standards of Medical Care 2020’, body weight, HbA1c and DKA are used to describe the agents.19 Risk of hypoglycaemia and SAEs are essential outcome measurements when evaluating glucose-lowering drugs and TDD is a valuable outcome measure with respect to interpreting the primary outcome (body weight). Lastly, drop-out rate and withdrawals due to adverse events are widely used components in the assessment of adverse effects.25

Geometry of the network

A network graph will be used to present the evidence base for the primary outcome following our systematic review: treatments will be represented by nodes and head-to-head studies between treatments are represented by edges.26 The sizes of edges and nodes are proportional to the available numbers of studies comparing the different interventions and the numbers of patients studied with each treatment. The description of the network of interventions will include numerical summary statistics to describe the current evidence available for the competing interventions and to identify gaps and potential bias. At the level of drug classes, it will be examined whether head-to-head comparisons are between agents in the same class or between agents in different classes. In the networks of drug class comparisons, each drug class will be drawn by a node and randomised comparisons between drug classes shown by links between the nodes.

Assessment of risk of bias

Two reviewers will independently assess the risk of bias for each study using the Cochrane Risk of Bias tool.27 Each study will be rated as having a low, high or unclear risk of bias for each of the following aspects: sequence generation, allocation concealment, blinding of participants, blinding of personnel, blinding of outcome assessors, incomplete outcome data, selective outcome reporting and other sources of bias. Disagreement will be resolved first by discussion and then by consulting a third review author for arbitration.

Data synthesis

Measures of treatment effect: contrast-based comparisons

When each study uses the same outcome measure and the units of that measure are intuitively interpretable (eg, body weight in kilograms and HbA1c in percentages), presenting differences in means is usually desirable. Thus, the meta-analyses for the continuous outcomes measuring on the same scale will be conducted by calculating the weighted mean difference with 95% CIs.28 For binary (dichotomous) outcomes, risk ratios will be the preferred measure of relative effect and, where applicable, control group risks will be used to generate absolute risks.29

Handling of missing data

RCT level: an intention-to-treat (ITT) analysis estimates the effect of treatment assignment in a particular trial, not the effect of the treatment itself. ITT effects are agnostic about post-randomisation decisions, including treatment discontinuation and the use of concomitant therapies prohibited by the study protocol. The true effect of selecting a new management strategy is a combination of biological effects, variations in compliance or adherence and other patient characteristics that influence efficacy. For all the meta-analyses, estimates apparently derived from the ITT population will be used; that is, if these analyses did not explicitly (and appropriately) handle the missing data in a specific trial, this will be noted as a risk of attrition bias.

Meta-analysis level: for missing standard deviations (SDs) of mean changes and where the p value is provided for a comparison between the treatment and control groups, the SD will be calculated by converting the p value into a t value with appropriate degrees of freedom (df). If neither the SDs nor the p values are available, a SD will be imputed based on studies with comparable measurement methods, duration and measurement error. Missing outcome data that were probably measured will be retrieved by contacting the corresponding authors of the RCTs via email. Where this is unsuccessful, missing data will be calculated from the raw numbers given in tables and/or estimated from bar charts if possible. The overall robustness of the findings will be assessed by manually imputing a ‘no change’ into the analysis. Thus, consistence between the sensitivity analyses and the primary analyses will strengthen the robustness of the results.

Assessment of heterogeneity

Statistical heterogeneity will be assessed using the inconsistency index (I2), where an I2 value of more than 50% indicates significant heterogeneity. Sources of heterogeneity will be explored by subgroup and sensitivity analysis. Random-effects models will be applied per default, but the 95% CIs will be compared with the point estimate from the corresponding fixed-effects meta-analysis. Agreement between the models will indicate robustness against small-study bias.

Data synthesis (meta-analysis)

Each outcome will be combined using appropriate statistical software according to the statistical guidelines referenced in the current version of the Cochrane Handbook for Systematic Reviews of Interventions. Data will be combined into nodes for each drug class using a random-effects model as a substantial variability in the trial methodology and subsequent treatment effects across studies may be expected. If heterogeneity is substantial, confidence in the estimates will be rated down in the Summary of Findings table as recommended by the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) Working Group. For the primary outcome (change in body weight), a random-effects NMA will also be conducted to assess the comparative effectiveness of the adjunctive glucose-lowering drugs. Local and global methods for evaluation of inconsistency will be employed. Quality of evidence contributing to network estimates of the primary outcomes will also be assessed using GRADE.

When two drugs are compared with a common standard, the difference in effect between these two drugs with respect to the common standard forms the basis of indirect comparisons. Indirect treatment comparisons in a meta-analysis can be analysed by various methods according to the different networks applied, including the star, ladder, closed and partially closed-loop designs. We will perform mixed-effects models using an arm-based, random-effects model within an empirical Bayes framework.30 The linear mixed model incorporates a vector of random-effects and a design matrix for the random-effects. Allowance will be made for differences in heterogeneity of effects between different drugs by specifying that the linear predictor varies at the level of study and the drug across study. Heterogeneity for the indirect comparison analyses will be evaluated using estimated tau-squared, which measures the statistical heterogeneity across the population of studies. If the collected data do not allow for an NMA of the primary outcome, this will be stated in the manuscript.

Reporting the NMA

An NMA will be performed within the ‘frequentist framework’ to synthesise the available evidence from the entire network of trials (reporting on change in body weight) by integrating direct and indirect estimates for each comparison into a single summary treatment effect. A frequentist random-effects model will be used (ie, empirical Bayes based on mixed-effects models) applying the methodology of multivariate meta-analysis to assess the comparative effectiveness of eligible interventions.

To check that the model fits the data well, hypothesis tests based on deviance statistics will be used. Issues of incoherence (direct and indirect effect estimates are not similar) will be identified by comparing direct evidence (ie, estimates from pairwise comparisons) with indirect evidence (ie, estimates from NMA).31 In this approach, incoherence will be assessed locally by statistical evaluation of the difference between direct and indirect estimates for a specific comparison in the loop. A common heterogeneity estimate across the network will be assumed. In case of incoherence in a closed loop of evidence, the certainty of evidence of each estimate will enable us to decide which estimate to believe.

Treatment rankings

An objective assessment of the strength of information in the network and the magnitude of absolute benefits should accompany rankings to minimise potential biases. Following the NMA, information about the hierarchy of competing interventions (drug classes) in terms of their mutual mean difference and their 95% CIs (and credibility intervals) will be provided. As recommended by the GRADE Working Group,32 mutual rankings can be judged along with corresponding estimates of pairwise comparisons between interventions. Rankings (eg, the surface under the cumulative ranking (SUCRA) curve) might unfortunately be misinterpreted since these may exaggerate small differences in relative effects, especially if they are based on limited information. For these reasons, standard forest plots will be used to summarise findings for all contrast-based meta-analyses. The large number of treatment comparisons coming from arm-based NMAs will be presented using a league table (ie, a tabular approach used to succinctly present all possible pairwise comparisons between treatments) as suggested by PRISMA Extension Statement for Reporting Network Meta-analyses.26 The primary attention will be on reporting differences between means of a specific drug class of interest versus ‘management as usual’/placebo.

Using the GRADE approach instead of the SUCRA curve, the adjunctive glucose-lowering drug classes will be evaluated first based on their effectiveness versus placebo, second versus other competing interventions and finally according to GRADE certainty of evidence ratings. Based on this, drug classes will be sorted into three groups: among the most effective (superior to both placebo and to at least one intervention superior to placebo or no treatment); superior to placebo, but not superior to any other intervention; or no more effective than placebo.

Subgroup and sensitivity analysis

Outcomes will be reported separately for each drug class. Subgroup analysis will be used to explore possible sources of heterogeneity based on the duration of intervention and dose administered. A meta-regression analysis will be performed to evaluate the potential influence of glycaemic control at baseline, duration of diabetes and baseline body weight. Selection of a statistical model for NMAs where comparisons between treatments are largely based on single studies can represent a challenge. Sensitivity analysis will be performed in order to explore the sources of heterogeneity based on internal validity components (eg, full-text publications vs abstract) and risk of bias (high vs low risk of bias). Performance of additional analyses retains an important role in establishing the robustness of our findings. Various ways to structure the treatment network (such as lumping and splitting in relation to name of active drug and potentially dose levels, method of administration or exclusion of certain doses) will be reconsidered, accounting for the effect of covariates on summary effect measures (using meta-regression analysis) and use of different statistical models (including a Bayesian approach, where different prior distributions will be chosen). Findings from such analyses will be reported so that readers have all available information for judging robustness of primary findings.


For outcomes reported in ≥10 studies, comparison-adjusted funnel plots will be drawn. If funnel plot asymmetry is observed, reasons for its prevalence (eg, selective reporting, publication bias, heterogeneity and inconsistency) will be examined.

Confidence in cumulative evidence

The certainty associated with the comparisons will be assessed using the GRADE approach.33 For both direct and indirect comparisons, the starting point for certainty in the estimates will be high (further research is very unlikely to change the authors’ confidence in the estimate of effect), but can be rated down to moderate (further research is likely to have an important impact on the authors’ confidence in the estimate of effect and may change the estimate), low (further research is very likely to have an important impact on the authors’ confidence in the estimate of effect and is likely to change the estimate) or very low (any estimate of effect is very uncertain). The certainty of the evidence will be assessed across the domains of risk of bias, consistency, directness, precision and publication bias. Additional domains may be considered where appropriate. Interventions will be categorised from most to least effective interventions: superior to both placebo and alternatives; superior to placebo, but inferior to alternatives; and no better than placebo. The GRADE assessment will be conducted by three reviewers.

Ethics and dissemination

As no primary data collection will be undertaken, no ethical assessment is required. Data will be processed according to protocol, merged into at least one scientific article and published in an international peer-reviewed scientific journal. The analysis will be reported according to PRISMA guidelines.


View Abstract


  • Contributors CL is the guarantor. CL, AGR, SS and KN drafted the protocol manuscript. LNR and ON critically reviewed the methodological content and developed the search strategy in collaboration with CL, AGR, SS and KN. RC provided statistical and methodological expertise. All authors read, provided feedback and approved the final protocol.

  • Funding CL, AGR, SS, LNR, ON and KN are employed by Steno Diabetes Center Copenhagen (SDCC), who is the sponsor of the research initiative. SDCC is owned by the Capital Region of Denmark and supported by a donation from the Novo Nordisk Foundation (NNF16SA0024138). The Musculoskeletal Statistics Unit at the Parker Institute, Bispebjerg and Frederiksberg Hospital (RC), is supported by a core grant from the Oak Foundation (OCAY-18-774-OFIL).

  • Disclaimer The funders have not been involved in the conception or any other part of the systematic review, and no external funding has been received for this research initiative.

  • Competing interests CL, AGR, SS, LNR, ON and RC have nothing to disclose. KN owns shares in Novo Nordisk and has received research grants and fees for lecturing from Novo Nordisk and Zealand Pharma. No other potential conflicts of interest relevant to this article were reported.

  • Patient consent for publication Not required.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.