Review question

This review of studies published in the past 30 years, from 1 January 1986 to 31 August 2016, should answer the following question:

What is the global prevalence of DM among patients with TB?

Inclusion criteria

We will include cross-sectional, case–control or cohort studies conducted in patients suffering from pulmonary drug-sensitive or resistant TB and reporting on the prevalence of DM or providing enough data to compute this estimate. We will consider extrapulmonary TB diagnosed by culture of Mycobacterium tuberculosis and also those which will have been treated as such despite the absence of culture of M. tuberculosis. However, this second group will not be considered for the meta-analysis but will be used for the narrative part review. The diagnosis of diabetes will have to have been made by a physician or defined based on measured fasting plasma glucose, oral glucose tolerance test or self-report, according to WHO criteria.12 TB cases must have been diagnosed based on WHO criteria.13

Exclusion criteria

We will exclude:

1. Commentaries, editorials and cases series with less than 50 patients.

2. Duplicates: for studies published in more than one paper, the most comprehensive one reporting the largest sample size will be considered.

3. Studies whose key data will not be accessible even after request from the authors.

Bibliographic database searches

1. We will search PubMed, Excerpta Medica Database (Embase), Index Medicus and African online journals to identify relevant abstracts on the prevalence of DM among patients with TB published from inception to 31 December 2016. Key search terms will include: ‘tuberculosis’, ‘TB’, ‘mycobacterium’, ‘diabetes’, ‘diabetic patients’ and ‘hyperglycemia’. The PubMed search strategy is shown in table 1, and will be adapted for other databases.

2. Abstracts of all eligible papers will be reviewed and their full articles in the second time. To supplement the bibliographic database searches and identify potential additional data sources. Where relevant data will not be available, we will directly contact the authors to request supplementary information or the full texts of their articles.

Selection of studies deemed relevant for inclusion in the review

Assessment of eligible papers will be independently run by two authors using an assessment guide to ensure that the selection criteria are reliably applied by them all (ATT and JJB). Two investigators (ATT and JJB) will independently screen titles and abstracts obtained from the searches and retrieve all full texts of potentially eligible papers. These full texts will be screened using a standardised form for final inclusion in the review. Disagreements between assessors will be resolved by consensus, with consultation of a third investigator (JJN) for arbitration if needed. Level of agreement between review authors will be measured using the Cohen’s Kappa statistic.14

Assessment of methodological quality and reporting of data

The methodological quality of included studies will be assessed using the Newcastle-Ottawa Scale.15

Data extraction and management

A data extraction sheet will be used to collect information relating to the country, the region, year of publication, type of study, period of the study, study design, study setting, number of participants, mean/median age or age range of the population, diagnostic criteria for each condition, the presence of another important comorbidity like HIV and the prevalence of DM. Authors will be contacted to request lacking or additional data. We will conduct a subgroup analysis using comorbidities, different diagnosis criteria and period of the study. In case of multinational studies, we will disaggregate the results to present the prevalence of DM among patients with TB in each country. If it is not possible to separate the data by country, the study will be presented as one and the countries in which the studies was conducted will be shown.

Statistical analysis

A meta-analysis will be conducted for data from studies in which DM will have been diagnosed using the same diagnosis criteria. We will first determine SEs for the study-specific estimates from the point estimate and the appropriate denominators, assuming a binominal distribution. After stabilising the variance of individual studies with the Freeman-Tukey double arc-sine transformation,16 we will then pool the study-specific estimates through a random-effects meta-analysis model to obtain an overall summary estimate of the prevalence across studies. Assessment of publication bias will be performed by using funnel plots and the Egger’s test. Heterogeneity will be evaluated by the χ² test on Cochrane’s Q statistic and quantified by the I². I² values of 25%, 50% and 75% will be considered as representing low, medium and high heterogeneity, respectively.17 In the case of substantial heterogeneity, we will perform a subgroup analysis to detect possible sources of heterogeneity using the following grouping variables: age group, the period of diagnosis (beginning or ending of treatment), positivity of sputum culture at microscopy, relapse or recurrence, association to others comorbid conditions such as HIV, continent or geographical area and study quality. Statistical analyses will be performed using Stata  V.14 software.

Results reporting and presentation

The proposed systematic review will be reported following the MOOSE Guidelines for Meta-Analyses and Systematic Reviews of Observational Studies.17 The study selection process will be summarised using a Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols flow diagram including reasons for studies’ exclusion. Tables and forest plots will serve to summarise quantitative data where appropriate. Prevalence data will be presented by continent, time period of diagnosis, presence of others comorbid conditions and classification of TB infection depending on available data. The methodological quality of included studies will be presented in a table and narratively summarised.