The systematic review will be conducted and reported in accordance to the ‘Preferred Reporting Items for Systematic Reviews and Meta-Analysis Protocols (PRISMA-P) 2015 checklist’ (20). View Additional file 1 for the completed checklist relevant to this review protocol. The protocol has been registered with the International Prospective Register of Systematic Reviews (PROSPERO) database (ID number: 168032).
Both published and unpublished studies including a weight maintenance intervention will be analysed. Studies will be included irrespective of the country where they have been carried out and the language of the report. Studies may be both randomised and non-randomised, and blinding is not required given that it is not always possible in behavioural interventions. The use of the findings from this review to develop a novel intervention and design a randomised controlled trial justify the inclusion of non-randomised studies (21). Studies meeting the following criteria will be deemed eligible for analysis (PICO structure):
Types of Participants
Participants will be adults (age ≥ 18 years), with a clinical diagnosis of T2DM or pre-diabetes according to the criteria of the World Health Organisation (22) or equivalent international standards (e.g. 23). When it is unclear whether a diagnosis was conducted, the study authors will be contacted to obtain the information. Further, studies where participants display impaired fasting glucose or non-diabetic hyperglycaemia will also be included. Participants will additionally have had to achieve weight loss within 24 months prior to the start of the weight maintenance intervention or have been instructed to maintain their weight as a preventative measure. No restriction will be placed on the amount of weight loss achieved during a weight loss programme given differential targets and success rates across studies (24). Participants may have been recruited from the general community, hospital, clinical care centre or other health services. Exclusion criteria include individuals with overweight or obesity who do not have a diagnosis of pre-diabetes or T2DM, individuals with type 1 diabetes mellitus, and individuals with conditions requiring antipsychotic drugs.
Types of Intervention
Interventions applied can focus on behavioural/lifestyle change, pharmacological (drugs must be approved as a weight loss drug by the European Medicines Agency), can be delivered online or in-person, and can entail food replacement(s)/supplement(s). Both stand-alone and combined interventions will be included. Papers will be excluded if the intervention focuses solely on weight loss or the weight loss component is not clearly distinguished from the weight maintenance phase. Single-cohort studies will be accepted, as well as studies where the comparator is no intervention, standard or minimal care, waitlist, or the use of a placebo. When the details of the information are unclear, authors will be contacted to provide further information.
Types of outcomes
The primary outcomes will be weight, glycaemic control (including HbA1c, fasting plasma glucose, insulin sensitivity/resistance), and adverse effects (recovery of weight lost, physical and/or psychological side effects such as disordered eating behaviour). Secondary outcomes will include cardiovascular risk factors (total cholesterol, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, triacylglycerol, diastolic blood pressure, systolic blood pressure), psychological wellbeing (including health related quality of life), change to glucose medication and waist circumference. Quantitative (both continuous and categorical) outcomes will be considered.
Should amendments be necessary, these shall be documented, reported, and fully justified.
A systematic search of the literature available from 1940 to present will be carried out in Web of Science, Medline, PsychINFO (via OVID), Embase, international trial registers (Cochrane Central Register of Controlled Trials, WHO International Clinical Trials Registry portal (ICTRP) and ClinicalStudyResults.org). We will additionally conduct hand searches of reference lists, and we will review the grey literature (e.g. conference papers/conference proceedings, theses, dissertations, studies published in non-indexed journals) via OpenGrey. Personal letters and e-mails will be sent to the corresponding authors of papers in the field of weight management in T2DM. The authors will be asked for information regarding unpublished or ongoing studies.
Search terms identified for each of the following relevant categories: population (“type 2 diabetes” or “diabetes mellitus” or “diabetes mellitus type 2” or “pre-diabetes” or “prediabetes” or “hyperglycaemia”), intervention (“weight maintenance” or “maintained weight” or “weight maintenence” AND “behav*” or “lifestyle*” or “online” or “computer” or “web” or “pharma*” or “food repla*” or “*supple*”), and outcome (“weight” or “body mass index” or “BMI” or “glycaemic* control” or “cardiovascular*” or “psych*”). Boolean searches will be carried out with terms combining categories and variations of the terms (via truncation). Medical Subject Headings (MeSH) keywords and Emtree keywords will be used. See Additional file 2 for the draft search strategy for MEDLINE.
The search will be managed using the Rayyan (http://rayyan.qcri.org) software app. Results will be imported into Rayyan, which automatically detects duplicates. Duplicates will also be detected through the manual data screening process.
Two authors of the team will screen all titles and abstracts identified by the search strategy based on the inclusion criteria. Articles that appear to meet the inclusion criteria will be extracted and reviewed in full by two authors. Inconsistencies will be discussed, and if unresolved another member of the team will be brought in to make the final decision. Studies deemed irrelevant will be excluded, and reason will be provided. Authors of unpublished studies will be contacted.
Search results will additionally be uploaded to AbstrackR, a citation screening programme functioning on an algorithmic framework that predicts the likelihood citations are relevant for further inspection (25). Screening will be initiated by two authors until the programme indicates it is ready to make predictions. Citations extracted through this approach will be compared and cross-checked against the output of the manual screening and selection process. Precision, false negative rate, number of relevant citations missed, and researcher time saved will be examined to determine the usefulness of AbstrackR.
Data extraction and quality assessment
Two authors will independently extract the data of the included studies using a standardised data extraction form (Additional file 3) and the GRADE pro software (https://gradepro.org/). The authors will extract information needed to assess the methodological quality of the studies, as well as sample characteristics (size, inclusion/exclusion criteria, sex, age), intervention details, and outcome measures. Inconsistencies in data extraction will be discussed and recorded. If needed, a third author will be included in the discussion to reach a final decision. When data is missing or inadequately described, three attempts will be made to request the information from the corresponding author.
The Cochrane Risk of Bias Version 3 tool (26) will be used for randomised controlled trials, while the ROBINS-I tool (27) for observational studies. The strength of the overall body of evidence for each outcome will be assessed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) system (28). Any discrepancies will be solved by a third author.
A narrative synthesis of eligible studies will be conducted. A qualitative summary of all extracted data will be provided, and quantitative data will be reported in a table. One author from the team will synthesise all the data, and another will review the process. Inconsistencies will be resolved by a third author. Overall effect size on weight will be reported as weighed mean differences with 95% confidence intervals. If there is homogeneity in the outcome variables, a meta-analysis of the studies will be carried out with the Comprehensive Meta-Analysis software (29). Clinical heterogeneity between studies will be evaluated by identifying variability in participants, baseline data, interventions, and outcomes. The I2 statistic will be calculated to quantify and interpret statistical heterogeneity (30). Funnel plots will be carried out to assess publication bias (if over 10 studies are identified). Pooled risk ratios (RRs) and 95% confidence intervals (CIs) will be calculated for dichotomous outcomes. Pooled mean differences and 95% Cis or standardised mean differences and 95% CI will be calculated for continuous outcomes when results are reported on the same scale (or can be converted to the same scale), or if results are reported on different scales, respectively.
A random effect model will be generated, where study weighing is based on in-study and between-study variances. When available data does not enable statistical pooling, results will be reported in a narrative format.
If data is not available in the format required, we will first contact study authors or we will back-calculate the data if possible (e.g., standard deviation from standard errors or 95% CIs, mean and standard deviation from median and range, etc.). Reasons for missing data will be recorded (e.g. drop-outs, losses to follow-up and withdrawals).
A sensitivity analysis will be conducted including only trials classified as low risk of bias.