Types of studies

Randomised controlled trials comparing any type of continuous glucose monitoring (CGM) system with conventional self‐monitoring of blood glucose levels or with another type of CGM system in patients with type 1 diabetes mellitus (DM).

Types of participants

Participants are males and females of any age who are classified as having type 1 DM using accepted criteria. To be consistent with changes in classification and diagnostic criteria of diabetes mellitus through the years, the diagnosis should have been established using the standard criteria valid at the time of the beginning of the trial (ADA 1999; WHO 1980; WHO 1985; WHO 1998). Ideally, diagnostic criteria should have been described. If necessary, authors' definition of diabetes mellitus will be used. Diagnostic criteria will be eventually subjected to a sensitivity analysis.

Types of interventions

Types of outcome measures

Electronic searches

We will use the following sources for the identification of trials:

• The Cochrane Library (latest issue);

• Specialised database of the Cochrane Metabolic and Endocrine Disorders Group;

• MEDLINE (2003 until recent);

• EMBASE (2003 until recent);

• CINAHL (until recent).

We will also search prospective trial registers to find ongoing trials:

• Dutch Trial Register (NTR);

• Australian New Zealand Clinical Trials Registry (ANZCTR);

• ISRCTN register (ISRCTN.org);

• ClinicalTrials.gov;

• Chinese Clinical Trial Register (ChiCTR);

• Clinical Trials Registry ‐ India (CTRI);

• Sri Lanka Clinical Trials Registry (SLCTR).

For detailed search strategies please see under Appendix 1. Additional key words of relevance may be detected during any of the electronic or other searches. If this is the case, electronic search strategies will be modified to incorporate these terms. Studies published in any language will be included.

Searching other resources

We will try to identify additional studies by searching the reference lists of included trials and (systematic) reviews, meta‐analyses and health technology assessment reports noticed. Furthermore, to find relevant but unpublished trials, experts on diabetes research will be contacted and abstract books of European and American diabetes conferences will be checked.

Selection of studies

To determine the studies to be assessed further, two authors will independently scan the title, abstract or both sections of every record retrieved. All potentially relevant articles will be investigated as full text. Interrater agreement for study selection will be measured using the kappa statistic (Cohen 1960). When there is only an abstract available, we will try to find the final report of the trial. Studies without a final report will be considered separately. In the case of duplicate publications and companion papers of a primary study, we will try to maximise yield of information by simultaneous evaluation of all available data. In cases of doubt, the original publication (usually the oldest version) will obtain priority.

The full text articles will be examined for compliance with eligibility criteria. Studies will be included in the review if:

• they are based on RCTs;

• they include patients with type 1 DM;

• the intervention includes a CGM system.

 Studies will be excluded if:

• the CGM system is not compared with conventional self‐monitoring of blood glucose levels or with another type of CGM system;

• none of the above mentioned outcomes are reported;

• the results on type 1 DM are not presented separately.

Study selection will be performed by two researchers independently. Where differences in opinion exist, they will by resolved by a third party. If resolving disagreement is not possible, the article will be added to those 'awaiting assessment' and authors will be contacted for clarification. An adapted PRISMA (Preferred Reporting Items for Systematic Reviews and Meta‐Analyses) flow‐chart of study selection will be attached (Liberati 2009).

Data extraction and management

For trials that fulfil the inclusion criteria, two authors will independently abstract relevant population and intervention characteristics using standard data extraction templates (for details see 'Characteristics of included studies' and Table 1, Appendix 2) with any disagreements to be resolved by discussion, or if required by a third party. Any relevant missing information on the trial will be sought from the original author(s) of the article, if required. 

The following data will be extracted:

(1) general information: title, authors, reference/source, year of publication and language of publication, reviewer, date

(2) in‐ and exclusion criteria (confirmation of eligibility, reason for exclusion)

(3) study characteristics:

• study design (RCT, parallel or cross‐over; single or multicenter, country, trial start year, duration of intervention and duration of follow up);

• patients: number, gender and age distribution, ethnic group distribution, setting, diagnostic criteria for type 1 diabetes mellitus, average duration of disease, baseline HbA1c, body mass index, insulin use (pump or injections), co‐morbidity, co‐medication, treatment before study, percentage pregnant women, percentage children (age less than 18 years), percentage patients with poorly controlled diabetes (HbA1c greater than 8.0%) and percentage patients with hypoglycaemia unawareness;

• interventions: type of CGM system, intermittent or continuous use, duration of CGM system use and type of self‐monitoring (times per day);

• outcomes: definition, timing and unit of measurement (for scales: upper and lower limits and whether a high or low score is favourable).

(4) results (for each outcome):

• dichotomous: number of patients with outcome and total number of patients in the intervention group and in the control group;

• continuous: number of patients, mean effect, standard deviation (SD) in the intervention group and in the control group;

• number of drop outs in the intervention group and in the control group.

(5) funding source.

Assessment of risk of bias in included studies

Two authors will assess each trial independently. Possible disagreement will be resolved by consensus, or with consultation of a third party in case of disagreement. Interrater agreement for key bias indicators (e.g. allocation concealment, incomplete outcome data) will be calculated using the kappa statistic (Cohen 1960).

Risk of bias will be assessed using the Cochrane Collaboration’s tool for assessing risk of bias (Higgins 2008). The following criteria will be used:

• was the allocation sequence adequately generated?

• was the allocation adequately concealed?

• was knowledge of the allocated interventions adequately prevented during the study (blinding)?

• were incomplete outcome data adequately addressed?

• are reports of the study free of suggestion of selective outcome reporting?

• is inappropriate influence of funders suspected?

• are the reports of the study free of conflicts of interest of the authors?

• was the study apparently free of other problems that could put it at risk of bias (baseline imbalance, early stopping)?

A judgement of ‘Yes’ indicates low risk of bias, ‘No’ indicates high risk of bias, and ‘Unclear’ indicates unclear or unknown risk of bias. For each criterion we will use the definitions of ‘Yes’, ‘No’ and ‘Unclear’ as described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2008).

Except for blinding and incomplete outcome data, the criteria will be addressed per trial. For blinding, we will assess the risk of bias for the subjective and objective outcomes separately. The item incomplete outcome data will be addressed for short‐term (three months and less) and long‐term (from three months onwards) endpoints.

Measures of treatment effect

Dichotomous outcome data (e.g. diabetic complications) will be expressed as relative risk (RR) with 95% confidence intervals (CI). In the case of rare events (incidence less than 1%) a Peto odds‐ratio will be calculated for each study (Bradburn 2007).

Continuous outcomes will be summarized as mean differences with 95% CI and an overall mean difference will be calculated in the meta‐analysis. For studies addressing the same outcome but using different outcome measures, for example different scales measuring quality of life, standardised mean differences (SMD) will be used.

Unit of analysis issues

Special issues in the analysis of studies with non‐standard designs, such as cluster‐randomized or cross‐over trials, will be described. For cluster RCTs and cross‐over studies we will extract the point estimates of the results and their standard error. These must be the result of a correct analysis (multilevel analysis and analysis of the paired differences, respectively). We will use the generic inverse variance method for combining those study results. If the results in cross‐over studies were presented as if the trial had been a parallel group trial with standard deviations for each intervention separately, we will estimate the standard error of the mean difference using these intervention‐specific standard deviations and impute a correlation coefficient of 0 (Higgins 2008). If we cannot obtain the results of a paired analysis in cross‐over studies, we will use the results of the first period.

Dealing with missing data

Relevant missing data will be obtained from authors, if feasible. Otherwise we will use, if possible, the imputation methods mentioned in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2008). Evaluation of important numerical data such as screened, randomised patients as well as intention‐to‐treat (ITT) and per‐protocol (PP) population will be carefully performed. Attrition rates, for example drop‐outs, losses to follow‐up and withdrawals will be investigated.

Assessment of heterogeneity

A priori the authors will evaluate clinical diversity of the included studies. In case of excessive clinical heterogeneity, that is the trials are not considered sufficiently homogeneous in terms of participants, interventions and outcomes, the results will not be pooled in a meta‐analysis.

We will assess statistical heterogeneity by visual inspection of the forest plots, by using a standard Chi² test and a significance level of α = 0.10, in view of the low power of such tests. We will quantify heterogeneity by the use of the I² statistic. I² values of 50% and more indicate a substantial level of heterogeneity (Higgins 2003). When heterogeneity is found, we will attempt to determine potential reasons for it by examining individual study and subgroup characteristics.

In case of considerable statistical heterogeneity (insufficient overlap of the 95% confidence intervals and an I² statistic higher than 75%) and availability of at least 10 trials a meta‐regression analysis will be performed to identify factors than may explain the heterogeneity. The following study characteristics will be considered:

• country (USA versus Europe versus other countries; because of differences in diabetes care and cultural factors);

• baseline HbA1c (disease severity; improvement in HbA1c is not to be expected in people with already low HbA1c values but suffering from high frequencies of hypoglycaemia);

• insulin use (pump versus injection; the benefits of CGM may be more readily discerned in those using the most optimal tool for insulin delivery).

Assessment of reporting biases

Funnel plots will be used to assess the potential existence of small study bias. When there are at least ten trials available, tests for funnel plot asymmetry will be performed. Possible sources of asymmetry in funnel plots are publication bias, poor methodological quality of smaller trials and true heterogeneity in effect associated with study size (Higgins 2008; Lau 2006; Sterne 2001).

Data synthesis

The following comparisons will be included in the analyses:

• CGM system versus conventional self‐monitoring;

• CGM system versus another type of CGM system.

For each comparison, separate analyses will be performed for four different patient groups:

1. adult (non‐pregnant) patients;

2. children (0 to 14 years);

3. adolescents (15 to 23 years);

4. pregnant women.

If possible, we will make a distinction between continuous and intermittent use of CGM.

We will carry out statistical analysis using the Review Manager software (RevMan 2008). Data will be summarised statistically if they are available, sufficiently similar and of sufficient quality. Statistical analysis will be performed according to the statistical guidelines referenced in the newest version of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2008).

Subgroup analysis and investigation of heterogeneity

For each patient group, the following subgroup analyses are planned:

• patients with hypoglycaemia unawareness; in these patients HbA1c is often already relatively low, therefore the number of episodes with severe hypoglycaemia will be used as primary outcome;

• patients with poorly controlled diabetes (defined as HbA1c greater than 8.0%).

Sensitivity analysis

We will perform sensitivity analyses by repeating the meta‐analyses excluding trials with:

• inadequate allocation concealment, inadequate blinding of the outcome assessors, incomplete follow up;

• suspected reporting bias;

• funding by a interested party (e.g. CGM system manufacturer) or possible conflicts of interest of the authors.