This protocol follows theguidelines of the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA-P) [15] (Additional file 1). This protocol will guide the review and any deviations while conducting the review will be reported including the reasons for the changes made in the method section of the final published manuscript. The review has been registered in the International Prospective Register of Systematic Reviews (PROSPERO) CRD42020149212.
Eligibility Criteria
The inclusion criteria of the review are: a) study design as having randomised controlled trials only; b) studies conducted in adults (aged ≥ 18 years old) diagnosed with T2D; c) have defined both use of real-time or non-real-time CGM as the intervention group and SMBG or routine glucose monitoring as the control group; and d) studies published in English. Studies published between February 2010 and March 2020 will be included as the field of CGMs among T2D has emerged over the last decade. Abstracts will be eligible for inclusion if sufficient information is provided to judge the quality and potential for bias of these trials. Non-RCT studies, non-type 2 diabetes, follow-up duration less than 6 weeks, and duplicate studies will be excluded.
Studies involving adults among T2D (aged ≥ 18 years old) have adopted CGMs as interventions will be included in this review. Inclusion criteria: a) Adults diagnosed with T2D (diagnostic criteria including the American Diabetes Association or World Health Organization or national guidelines); b) T2D of at least eight weeks duration; and 3) reporting HbA1c as one of the outcome measure. Exclusion criteria: a) Adolescents (under 18 years of age) and elderly people (over 70); b) Gestational diabetes mellitus; and c). Other types of diabetes (such as idiopathic diabetes or type 1 diabetes).
Information Sources and Search Strategy
The following databases will be searched: Cochrane Library, Science Direct, PubMed, EMBASE, CINAHL, PsycINFO, Scopus and grey literature for the identification of studies. The studies involving adults (aged ≥ 18 years old) will be included. We will include and summarize randomized clinical trials with respect to authors, publication type, year, status, and type of devices. Primary outcomes will be HbA1c, body weight, time spent with hypoglycaemia or hyperglycaemia, blood pressure, quality of life. Secondary outcome measures will be morbidity, all-cause mortality, user satisfaction, and barriers. Besides, the inclusion of grey literature (i.e. non-published, internal or non-reviewed articles, repositories, blogs) in the systematic review may help to overcome the publication bias that may arise due to the selective availability of data [16], thereby this review will include grey literature after reviewing the title and abstract accordingly.
Additionally, the reference list of identified systematic reviews and RCTs will also be updated to identify if references or bibliographies include relevant studies that might be included for the review (cross-referencing). Furthermore, indexed keywords in the Medical Subject Headings will be used to guarantee unified search terms. A comprehensive PubMed retrieval strategy (Additional file 2) will be developed in consultation with a medical librarian experienced in systematic database searching. Other databases will be searched and corresponding search strategies and logic grid will be adopted (Additional file 3).
Study Selection
Citation management system (Endnote X9) will be adopted to manage records exported from all the databases. First, all the studies will be screened by their titles using the Endnote X9. Then, the shortlisted studies will then be screened by their abstracts. Finally, the full-text of nominated studies will be screened according to the eligibility criteria. Study selection, data extraction, and risk of bias assessment will be conducted independently by at least two reviewers (MZ and YL). A pre-defined screening form will be designed and pilot testing will be conducted as per the eligibility criteria to ensure reliability of screening articles among the two reviewers (MZ and YL). Both reviewers will describe outcome measures after reviewing the studies to confirm the relevance of RCTs. A third author (WL) will determine and resolve discrepancies to make the final decision about whether the study meets the eligibility criteria for being including in the review in a consensus meeting. The process of study screening and selection will be reported via PRISMA flow diagram [17] (Additional file 4).
Data extraction and management
A data extraction form (Additional file 5) will be designed and edited by the review team after consultation. We will extract data by predesigned forms about characteristics of the studies to be included in the current study (including author, publication year, country, sample size, types of CGM devices, duration of diabetes, patient’s baseline, clinic history, basic treatment, and intervention/treatment duration). Continuous variables will illustrate as mean values, standard deviations, standard errors, or 95% CI, whereas binary variables will be expressed as frequencies and percentages (%). Studies comparing one SMBG group with two or more intervention groups will be treated as two or more studies sharing an SMBG group. Two reviewers (MZ and YL) will independently evaluate the quality of each study that meet the inclusion criteria of the systematic review. Another researcher (WL) will provide judgement when two authors have different reviews if necessary. We will evaluate duplicate publications, assess all available data simultaneously, maximizing the extraction of data for a bias assessment precisely. Authors will be contacted by emails to acquire missing or relevant material of their papers if necessary.
Quality Assessment of Included Studies
Cochrane’s risk of bias tool will be used for evaluating the quality of RCTs [18]. The following seven constructs will be evaluated as low, moderate, and high risk of bias or unclear risk of bias: random sequence generation, allocation concealment, blinding of personnel and participants, blinding of outcome assessors, incomplete data, selective reporting, and other potential risks (Additional file 6). Moreover, the quality of the evidence of the review will be assessed according to the Grading of Recommendations Assessment, Development and Evaluation Tool (GRADE) [19]. Two reviewers (MZ and YL) will independently evaluate the quality of each study. Any disagreement at this stage will be discussed and resolved by consultation and consensus within the review team if necessary. The summary of the quality of all included studies will be presented in a table.
Data synthesis and statistical analysis
This systematic review will synthesise a quantitative analysis known as meta-analysis. It seems that units may express the continuous variables differently in each included study; analysis will be performed using standardised mean difference (SMD) and 95% CIs. Binary variables will be analysed by risk ratio (RR) and 95% CIs. When only the standard error of p-value is reported, standard deviation (SD) will be considered according to the method recommended by Altman and Bland (1996). Moreover, if no description for SD is reported, we will calculate from 95% CIs, t-test values, or p-values. The sensitivity analysis will be used to test the robustness of the choices made, such as changing the cut-off for high - or low-quality included studies. Besides, the heterogeneity (by Cochrane chi-square χ² and the I² statistics) will be described via reporting differences in the study design and the characteristics of the study population [20]. A random-effect model will be used for analysis if appropriate; I ² will be used for evaluating statistical heterogeneity (I ² ≥ 50% is considered as heterogeneous), since it is the percentage of total variation provided between the studies (I ² values of 75%, 50% and 25% represent high, moderate and low heterogeneity, respectively) [21].
If sufficient RCTs are available and variability among those studies is low i.e. they are homogenous, a meta-analysis will be conducted as per the following subgroups if appropriate: a) real-time CGM vs SMBG; b) resorptive CGM vs SMBG; c) real-time CGM vs routine care, and d) resorptive CGM vs routine care. According to these categories, measures of associations such as relative risks, and odds ratios will be synthesized and reported. Beside this, when the number of included studies for this review is more than ten studies, a funnel plot will be plotted for evaluating publication bias. We will also use the Eggers' regression test to statistically evaluate the asymmetry of the funnel plot. Moreover, subgroup analysis via baseline HbA1c levels (< 6.5%, 6.6-7.9%, 8.0-11.0% and >11.0%) will be performed to assess the impact of baseline HbA1c on the effectiveness of CGM. Additionally, if heterogeneity is identified, a meta-regression analysis will be conducted on whether baseline information such as HbA1c, gender, age, and frequency and types of CGM sensor use has affected the impact of CGM on HbA1c levels.