PDM: Methods (2023)

  • Increase Font Size
  • Decrease Font Size
  • View as PDF

PDM: Methods (2023)

Protocol and Registration

This systematic review adhered to the Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) checklist1, and the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) method2, as well as methods from the Academy of Nutrition and Dietetics3, The protocol for this systematic review, including the research question, search strategy, inclusion/exclusion criteria and outcomes of interest, were specified a priori and registered at the International Prospective Register of Systematic Reviews (https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=351421). 

Eligibility Criteria

Articles were eligible for inclusion if they met the pre-specified Population, Intervention, Control and Outcome (PICO) criteria as described in the search plan linked to each research question. Participants in eligible studies were required to be free-living adults ≥ 18 years of age diagnosed with impaired glucose tolerance as confirmed by biochemical testing of one or more of the following parameters as defined by the American Diabetes Association: a Hemoglobin A1c (HbA1c) of 5.7-6.4%, fasting blood glucose (FBG) of 100-125mg/dl, and/or an oral glucose tolerance test (OGTT) blood glucose of 140-199 mg/dl 4. Studies were excluded if any or all participants were on glucose-altering medication prior to starting or during the study or were diagnosed with a chronic disease including but not limited to organ disease or transplantation, cancer, or any form of diagnosed diabetes mellitus. Interventions included MNT with at least two dietitian contacts over the course of at least one month compared to a control group receiving standard care. The intervention could be MNT alone or as part of a multidisciplinary intervention. In this review, most studies implementing the DPP did not utilize a dietitian, specify the credentials of the lifestyle coach delivering the intervention, or provide individualized nutrition interventions, and were therefore excluded. Articles were required to be randomized controlled trials (RCTs) published in English in peer-reviewed publications from 1995 to reflect contemporary practice. This review was limited to RCTs because they minimize the influence of confounding factors and are the most reliable form of evidence to test the effectiveness of an intervention. Conference proceedings where only the abstract was published were excluded. The primary outcome measures were incidence of T2DM, HbA1c and FBG levels. Secondary outcomes included lipid profiles, anthropometric and blood pressure measures, cardiovascular events or cardiovascular disease risk, quality of life, cost-effectiveness, adverse events, hospitalizations, and mortality.

Search Strategy

The search was conducted on August 5th, 2022, using the electronic databases MEDLINE (Ebsco), CINAHL (Ebsco) and Cochrane Central (Ebsco).  Relevant systematic reviews were hand searched to identify additional studies missed during the database search. 

Study Selection

Titles and abstracts identified in the database search were uploaded into Rayyan for title and abstract screening5. Each title and abstract was reviewed independently by two reviewers. Conflicts were discussed among reviewers to reach consensus. Following the title and abstract screening, full-texts of articles identified for potential inclusion were screened by two independent reviewers. Discrepancies were either discussed among reviewers to reach consensus or were examined by a third reviewer. When consensus could not be reached or eligibility was unclear, articles were evaluated by content experts. 

Data Extraction

One analyst extracted data from each included article into a standardized template. Accuracy of the data was confirmed through inspection from a second reviewer and consensus between the two reviewers when applicable. Data extracted included bibliographic information, population, and demographic data (e.g., age, sex, study location), source of funding, intervention characteristics (e.g., intervention duration, number of contacts with the dietitian, other health professionals involved), comparator description, outcomes of interest reported and quantitative data for each group including sample size, mean change and variance for continuous variables and number of events for categorical variables. The corresponding author was contacted if additional data or clarification of methods was needed.

Risk of Bias

The risk of bias was assessed using version 2 of the Cochrane risk-of-bias tool for randomized trials (RoB2), which assesses potential bias resulting from the randomization process, deviations from the intended intervention, missing outcome data, measurement of the outcome, and/or selection of the reported results6. Risk of bias is rated high, some concerns or low based on the five aforementioned domains. Articles were assessed for risk of bias independently by two reviewers. Discrepancies were discussed among reviewers to reach a consensus. Risk of bias was reported using the Robvis data visualization tool7.

Data Analysis

A meta-analysis was conducted in OpenMeta[Analyst] when quantitative data from multiple studies was available for a specific outcome measure8. Studies that did not provide the quantitative data required for meta-analysis were described in narrative synthesis only. When outcome measures were reported at multiple time points, only data reported closest to the end of the intervention was used in the meta-analysis. Sensitivity analysis was conducted by comparing effect sizes based on study quality. Subgroup analyses were conducted for anthropometric outcomes and glycemic outcomes based on whether >80% of the population had overweight or obesity. Subgroup analyses were also conducted for glycemic outcomes based on the format of MNT (individual, blended or group) and interventions provided (MNT only or MNT and additional non-dietetic interventions). The meta-analyses utilized random effects models and reported outcomes as mean difference (MD) and 95% confidence intervals (CI) for continuous variables. The sole categorical variable of interest, incidence of T2DM, was narratively synthesized due to lack of quantitative data available for conducting meta-analysis. The meta-analyses were reported using forest plots, funnel plots were generated to assess the presence of publication bias, and I2 measures were used to assess the degree of heterogeneity across studies8. 

Certainty Assessment

Certainty of evidence was assessed for each outcome using the GRADE method9 and a summary of findings table was generated using GRADEpro GDT2. Certainty of evidence was graded as high, moderate, low, or very low based on total sample size, study design, indirectness, inconsistency, risk of bias, imprecision, and other factors. Reasons for reducing certainty of evidence are documented in the summary of findings table. 

References:

  1. Systematic Reviews [Internet]. BioMed Central. [cited 2022 Dec 20]. Available from: https://systematicreviewsjournal.biomedcentral.com/
  2. GRADEpro GDT: GRADEpro Guideline Development Tool [Software]. McMaster University and Evidence Prime, 2022. Available from gradepro.org. 
  3. Handu D, Moloney L, Wolfram T, Ziegler P, Acosta A, Steiber A. Academy of Nutrition and Dietetics Methodology for Conducting Systematic Reviews for the Evidence Analysis Library. J Acad Nutr Diet. 2016 Feb;116(2):311–8. 
  4. Diagnosis | ADA [Internet]. [cited 2022 Dec 22]. Available from: https://diabetes.org/diabetes/a1c/diagnosis 
  5. Ouzzani M, Hammady H, Fedorowicz Z, Elmagarmid A. Rayyan—a web and mobile app for systematic reviews. Systematic Reviews. 2016 Dec 5;5(1):210. 
  6. Sterne JAC, Savovic J, Page MJ, Elbers RG, Blencowe NS, Boutron I, et al. RoB 2: a revised tool for assessing risk of bias in randomised trials. BMJ. 2019 Aug 28;366:l4898. 
  7. McGuinness LA, Higgins JPT. Risk-of-bias VISualization (robvis): An R package and Shiny web app for visualizing risk-of-bias assessments. Research Synthesis Methods. 2021;12(1):55–61. 
  8. Wallace BC, Dahabreh IJ, Trikalinos TA, Lau J, Trow P, Schmid CH. Closing the Gap between Methodologists and End-Users: R as a Computational Back-End. J Statistical Software. 2012 Jun 30;49:1–15. 
  9. Malmivaara A. Methodological considerations of the GRADE method. Ann Med. 2015 Feb;47(1):1–5.