Home monitoring of HbA1c in diabetes mellitus: A protocol for systematic review and narrative synthesis on reliability, accuracy, and patient acceptability

Population

Studies of adults aged 18 years or older who are being screened for, monitored, or previously diagnosed with diabetes mellitus. Studies which exclusively look at HbA1c measurement in people with haemoglobinopathies, anaemias, high erythrocyte turnover or other conditions known to affect HbA1c interpretation may be discussed but not included in analysis of validity. Animal studies will not be included.

Inclusion criteria and Index test

Articles which examine methods of capillary blood self-collection and storage for measurement of HbA1c, which look at test reliability, validity, or patient acceptability as detailed above, will be included. We will not include studies of novel methods still in experimental stages, where validation studies with patient blood specimens have not been carried out. We will include studies using venous blood to test novel methods for capillary blood collection (e.g. the use of venous blood on a dried blood spot). These studies will be differentiated from capillary studies in our analysis. We will only include studies of methods which would be practical to post to people with diabetes, thus excluding most point of care methods or those requiring high intensity storage such as refrigeration. Studies which look at the effect of time, temperature, and other storage variables on capillary HbA1c results will be included.

Target condition

Any form of diabetes mellitus where HbA1c would be used for diagnosis or monitoring. Routine use of HbA1c is not recommended in women in the second or third trimester of pregnancy and therefore studies in this cohort will not be included. [24]

Reference standard

For studies of validity or reliability, the reference standard must be a venous blood test analysed by a certified method for HbA1c testing as listed by the International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) or the National Glycohemoglobin Standardization Program (NGSP). [25] [26]We will exclude studies where the time interval between index test and reference standard is more than two weeks for >10% of participants.

Search strategy

The key search terms are ‘Diabetes’ AND ‘HbA1c’ AND ‘Capillary sampling’ which have been expanded in our full search strategy (Appendix 1 or available through PROSPERO (registration number CRD42021225606)). Free text, MESH terms, varied spellings and acronyms have been included to maximise capture of all relevant studies. We will search the Embase, MEDLINE, CINAHL and Web of Science Core Collection databases from database inception to September 2021. We will additionally search the grey literature through OpenGrey as well as using Google Scholar limiting retrieval to the first 200 relevant references as per guidance on optimal database combinations. [29] Ongoing studies and other grey literature will be sought from relevant conference abstracts from 2020 and 2021. Reference lists from the most relevant papers will be hand-searched using an ancestry approach. Results will be restricted to those available in the English language only.

Study Selection and Data Management

Titles and abstracts from search results will be downloaded to Endnote X9, where they will be de-duplicated. They will then be uploaded to Rayyan.ai for review. Two reviewers (JC and JB) will independently screen titles and abstracts against inclusion and exclusion criteria whilst blinded to the other’s decisions. Selected studies will then undergo full text review by both reviewers. Any discrepancies that arise during this process will be discussed between the two reviewers until a resolution is reached. Where a resolution cannot be found a third reviewer will be consulted. A PRISMA flow diagram will be used to demonstrate study selection and exclusion. Studies which undergo full text review will be documented, and if excluded at this stage we will keep record of the reasons for exclusion and make this information available.

We have created a data extraction tool using Microsoft Excel (Version 2102) which was piloted (Appendix 2). Two reviewers will carry out data extraction (JC and JB) and disagreements resolved through discussion.

The data being collected will include information regarding:

• Study Characteristics (Such as country, year, setting, objectives of the study, study inclusion and exclusion criteria)

• Participant Characteristics (Such as population description, age, sex, total number, diabetic status, and dropout rates)

• The capillary blood method/methods being investigated (including details about method used, person collecting the sample, location, storage and shipping conditions, laboratory methods including analysis device, laboratory blinding, resource requirements and timing between test and reference)

• The reference standard (including analysis device used, location of collection, storage and shipping conditions, laboratory blinding and number of available results)

• The primary outcome (including statistical methods used and outcomes relevant to that method, diagnostic thresholds if used, measures of error and bias and any other reported results)

• Additional outcomes (Patient acceptability, and where evidence exists to do so we will collect information on facilitators and barriers to patient uptake)

Quality assessment

We will be using ‘COnsensus-based Standards for the selection of health Measurement INstruments (COSMIN) risk of bias tool to assess the quality of studies on reliability and measurement’ to assess the quality of selected studies. The COSMIN tool will be piloted by both reviewers (JC and JB) on the first five selected studies. Following this if there is good agreement between both reviewers then the quality assessment for the remaining studies will be carried out by one reviewer only (JC). If there is poor agreement during the pilot this will be discussed between the two reviewers, the tool will be revised or replaced, repiloted, and this process will continue until good agreement is reached. As we are performing narrative synthesis, there will be no limit on what level of quality leads to inclusion. Instead, we will discuss quality of studies as part of the synthesis of their results.

Data Synthesis

From preliminary searches we have identified methodological heterogeneity in the studies identified. There is variability in how data has been presented, studies have either presented outcomes in terms of sensitivity and specificity of HbA1c as a diagnostic test, or at pre-defined clinically significant levels, or as measures of error, with correlation co-efficients, intercepts, Bland-Altman plots, and error grid analyses. There is also heterogeneity in the collection, storage and transportation methods and type of blood used for the index test. Full exploration of heterogeneity will take place following completion of searches, by visual examination of tables ordered by likely modifiers. Identified modifiers will be used to determine subgroups for synthesis, which will in turn be used to determine how these variables can impact validity.

Given the predicted level of heterogeneity, we anticipate the most appropriate method of synthesis will be narrative. We will be following the Synthesis Without Meta-Analysis (SWiM) reporting guidelines adapted for a non-interventional review. [30] The SWIM guidance is tailored towards reviews of intervention, and therefore it may need adapting as we undertake our synthesis. Where significant heterogeneity exists, meta-analysis can often fail to produce meaningful results by under-representing the overall body of evidence. Narrative synthesis will enable us to incorporate all available evidence despite heterogeneity.

We will discuss measurement outcomes and the appropriateness of each statistical method used to the study design. We will summarise outcomes by their direction of effect (which we define as whether the index test is or isn’t valid or reliable for measurement of HbA1c compared to reference standard), and use vote counting to inform our conclusions. We hope to not only present the current evidence, but also identify key gaps in the literature to guide future research.

For the secondary outcome of patient acceptability, the method of synthesis will depend largely on the amount of data available. We would ideally want to perform qualitative synthesis through the use of thematic analysis to further understand facilitators and barriers to patient uptake of home HbA1c measurement. If evidence for patient acceptability is limited, we will instead look at uptake and dropout rates, using this information to infer probable popularity of a proposed at home collection option.

Grading the strength of recommendations

We will be using the Grading of Recommendations, Assessment, Development and Evaluations framework (GRADE) in making our recommendations. [31]

Ethics and Dissemination

As a secondary analysis of existing published data with no primary identifiable personal data being collected no ethics review was required.

This review will be submitted for publication in a peer-reviewed open-access journal. Where invited to do so we will present our results at both national and international relevant conferences. These methods will allow dissemination to patient groups, clinicians, and guideline developers.

Patient and Public Involvement

Patients or the public were not involved in the design of this review.