Undernutrition as a risk factor for tuberculosis disease

Objectives

This is a protocol for a Cochrane Review (prognosis). The objectives are as follows:

To determine the prognostic value of undernutrition in the general population of adults and children for predicting tuberculosis disease.

Population	Adults and children from the general population
Index prognostic factors	Undernutrition (wasting, stunting, underweight)
Outcomes	Incidence of tuberculosis (adjusted risk)
Timing	Short‐ and long‐term follow‐up
Setting	General population/outpatients
Study design	Cohort studies

Background

Description of the health condition and context

Tuberculosis (TB) is one of the leading causes of death worldwide and the second‐leading cause of death for a single infectious agent after COVID‐19 (WHO 2022). The burden of tuberculosis has a widely inequitable distribution across the globe. More than two‐thirds of episodes are detected in World Health Organization (WHO) South East Asia and Africa Regions, with lower incidence rates found in the Americas and Europe. The estimated number of deaths from TB increased between 2019 and 2021, in contrast with the previous trend of continuing decline (WHO 2022). These fluctuations have negatively affected the milestones of the international endeavours to end the TB epidemic by 2035 as part of one of the UN Sustainable Development Goals (WHO 2022).

An estimated 2000 million people, equal to 25% of the global population, have been infected with TB bacteria (WHO 2022). Without treatment, approximately 5% to 10% of immunocompetent infected individuals will develop TB disease at some point in their lives, but some people – especially those under immunosuppressive conditions such as uncontrolled HIV infection – are at a higher risk of falling ill (CDC 2021). Data gathered by the WHO showed that in 2021, about 10.6 million people fell ill with TB, and 1.6 million (including 187,000 HIV‐positive people) died from the disease (WHO 2022).

Efforts to prevent the spread of TB include the detection and effective treatment of episodes and the provision of preventative therapy to close contacts of people with TB disease, infection prevention and control measures (including adequate ventilation and masking), bacillus Calmette‐Guérin (BCG) vaccination, and the detection and control of the underlying risk factors that increase the risk of TB, primarily HIV infection (Brett 2020). On a larger scale, preventive efforts also involve tackling the social determinants of health that lead to higher transmission and death rates amongst more vulnerable populations (Hargreaves 2011), including, for example, include people living in prisons, migrants, other socially marginalised people, people living with HIV, diabetes, or other immunosuppressive disorders, and children (WHO 2017).

Description of the prognostic factor

The WHO Global Tuberculosis Programme includes estimates of the risk for TB to estimate the number of TB episodes attributable to the following risk factors: alcohol use disorders, diabetes, HIV, tobacco smoking, and undernutrition. These estimates are part of the annual Global Tuberculosis Report (see Table 1).

Table 1. Global estimates of tuberculosis episodes in 2021 attributable to selected risk factors.

Risk factor	Risk ratio (uncertainty interval)		Number of people with the risk factor (millions)	Attributable TB episodes (millions, uncertainty interval)
Alcohol use disorders	3.3	(2.1 to 5.2)	292	0.74	(0.50 to 1.0)
Diabetes	1.5	(1.3 to 1.8)	498	0.37	(0.27 to 0.49)
HIV infection	16	(14 to 18)	38	0.86	(0.70 to 1.0)
Smoking	1.6	(1.2 to 2.1)	981	0.69	(0.49 to 0.92)
Undernourishment	3.2	(3.1 to 3.3)	711	2.2	(2.0 to 2.3)

Undernutrition is a condition that refers to the deficient intake of essential nutrients resulting from either a lack of food or an imbalanced diet that does not provide adequate quantities of the necessary nutrients (Global Nutrition Report 2020). Undernutrition affects every country in the world, and particularly poses children to be more vulnerable to disease and death. The WHO estimated that in 2022, one in five (148 million) children across the globe and younger than five years were too short for their age (stunted); 45 million were too thin for their height (wasted); and more than 37 million were overweight or obese (UNICEF‐WHO‐World Bank 2023). For adults, the last available age‐standardised estimates stated that in 2016, about 9.0% of people aged 20 years and older were underweight, having a body mass index (BMI) below 18.5 (NCD‐RisC 2017).

Health outcomes

Tuberculosis can be classified as tuberculosis infection (TBI) or tuberculosis disease (TB disease), depending on the ability of the host to control the infection. Individuals with TBI are infected with Mycobacterium tuberculosis but do not have active TB disease and cannot spread the infection to others. In some people, M tuberculosis overcomes the immune system and multiplies, progressing from TB infection to overt TB disease (CDC 2021). Half of the progressions to TB disease occur during the first year, whilst the incidence of progression decreases significantly beyond five years (Menzies 2021).

TB disease can occur in pulmonary (lungs) and extrapulmonary sites (elsewhere in the body). Pulmonary TB disease is the most common form of the disease, and usually presents with a cough and an abnormal chest radiograph. Extrapulmonary TB can involve any other organ, most commonly affecting the larynx, lymph nodes, pleura, brain, kidneys, bones, and joints; however, extrapulmonary TB is usually not infectious (CDC 2021).

The diagnosis of TB commonly involves a clinical evaluation, imaging studies, and laboratory microbiology tests (Lewinsohn 2017). The clinical evaluation involves the assessment of disease symptoms and medical history. Imaging diagnostics, including a chest X‐ray or computed tomography (CT) scan, can indicate characteristic patterns associated with pulmonary TB. Diagnostic microbiological tests include sputum smear microscopy, a culture‐based technique which aids in detecting the presence of M tuberculosis. Other molecular tests, such as polymerase chain reaction assays, detect the genetic material of the bacteria. The diagnostic approaches used in different studies may vary in availability and feasibility depending on local resources. In resource‐limited settings or in those where access to laboratories is limited, clinical diagnosis and radiological findings may be relied upon for prompt treatment initiation (Lewinsohn 2017).

TB can also occur in individuals following a successfully completed treatment for TB, also known as recurrent TB, which encompasses both relapse and reinfection (Vega 2021). Relapse refers to the reactivation of the same strain of M tuberculosis that caused the initial infection, whereas reinfection involves acquiring a new strain of M tuberculosis following a previous TB infection. Recurrent TB disease encompasses cases where it is challenging to differentiate between relapse and reinfection due to similar clinical presentations, although most cases of recurrent TB might be related to relapse (Vega 2021).

Why it is important to do this review

Previous systematic reviews have provided estimates for the multivariable‐adjusted risk of TB attributable to alcohol (2017), diabetes (2018), tobacco (2010), HIV (data from UNAIDS), and undernutrition (2010) (Hayashi 2018; Imtiaz 2017; Lönnroth 2010a; Lönnroth 2010b). The WHO has therefore commissioned updated reviews to assess the current prognostic value of these risk factors. Two of these risk factors (diabetes and undernutrition) were prioritised to inform the WHO TB Report, and this is the protocol for one of these two reviews.

Objectives

To determine the prognostic value of undernutrition in the general population of adults and children for predicting tuberculosis disease.

Population	Adults and children from the general population
Index prognostic factors	Undernutrition (wasting, stunting, underweight)
Outcomes	Incidence of tuberculosis (adjusted risk)
Timing	Short‐ and long‐term follow‐up
Setting	General population/outpatients
Study design	Cohort studies

Methods

Criteria for considering studies for this review

Types of studies

We will include retrospective and prospective cohort studies. We will include studies regardless of their publication status or language.

Targeted population

We will include adults and children from the general population. These may include individuals from different settings (outpatient and inpatient cohorts) with variable distribution of comorbidities and risk exposure to tuberculosis infection. We will include studies across all regions of the world.

Type of prognostic factor

We will define undernutrition as wasting (low weight‐for‐height), stunting (low height‐for‐age), and underweight (low weight‐for‐age) (WHO 2023). Common measures for this risk factor include the following.

• Children and adolescents

  • BMI < −2 standard deviations (SD) below the median

  • Underweight prevalence amongst children under five years of age (% weight‐for‐age < −2 SD)

• Adults and adolescents: low BMI scores (e.g. BMI < 18.5 age‐standardised estimates)

The moment of prognostication will be upon enrolment/baseline.

Types of outcomes to be predicted

Primary outcome

Incidence of active tuberculosis (TB) or TB disease, including pulmonary (lungs) and extrapulmonary sites (elsewhere in the body). This diversity of diagnostic approaches, including clinical evaluation, imaging studies, and study microbiology, will inherently contribute to clinical heterogeneity across the included studies.

Secondary outcomes

Incidence of recurrent TB disease, including due to relapse, reinfection, or non‐specified.

Time points for outcome assessments

We will consider group outcomes at short‐ (≤ 1 year), medium‐ (1 to 5 years), and long‐term (≥ 5 years) follow‐ups.

Search methods for identification of studies

Electronic searches

We will conduct a unified search for two risk factors, diabetes and undernutrition, that will inform two Cochrane Reviews of each risk factor independently due to the expected overlap in studies assessing multiple risk factors simultaneously. To identify relevant studies, we have developed an empirically derived search strategy for MEDLINE. First, we sourced relevant cohort studies from five known systematic reviews on 'tuberculosis and diabetes' (= 4) (Al‐Rifai 2017; Hayashi 2018; Jeon 2008; Obels 2022) and 'tuberculosis and undernutrition' (= 1) (Lönnroth 2010a). This yielded a set of 25 relevant references as a study pool for diabetes and five relevant studies for undernutrition. We complemented the undernutrition topic by carrying out exploratory searches, resulting in an expanded study pool for undernutrition of 20 studies. After the removal of duplicates, the final study pool for both topics combined resulted in 43 relevant references of cohort studies (year range = 1957 to 2022), which were all indexed in MEDLINE.

We conducted a text analysis of this study pool using the tools PubReMiner (hgserver2.amc.nl/cgi-bin/miner/miner2.cgi) and Yale MeSH Analyzer (mesh.med.yale.edu/) (Adam 2023). Our final search strategy was able to retrieve 39 of 43 relevant references from MEDLINE. Two older references (from 1957 and 1971) could not be retrieved because they did not have an abstract, and the title was not specific enough. Two further references could not be retrieved because they focused on post‐transplant patients, and the abstracts did not mention 'body mass index' or 'underweight' as a risk factor.

Using this strategy (see Appendix 1), we will conduct a comprehensive search with no restrictions on language of publication, publication status, or study filter in the following sources:

• MEDLINE (via PubMed);

• WHO Global Index Medicus, which comprises five information sources:

  • AIM (African Index Medicus);

  • LILACS (Latin American and Caribbean Health Science Information database);

  • IMEMR (Index Medicus for the Eastern Mediterranean Region);

  • IMSEAR (Index Medicus for South‐East Asia Region);

  • WPRIM (Western Pacific Region Index Medicus).

• World Health Organization International Clinical Trials Registry Platform (WHO ICTRP) (trialsearch.who.int/Default.aspx).

Searching other resources

We will screen the reference lists of the 43 relevant studies identified by extraction of studies from known systematic reviews and exploratory searches described in the previous section. We will also screen the reference lists of other potentially eligible studies or ancillary publications identified by our search. Furthermore, we will contact the study authors of included studies to identify any articles that we may have missed.

Data collection and analysis

Selection of studies

We will use Covidence software to identify and remove potentially duplicate records (Covidence). Two review authors (BB, JVAF) will independently scan abstracts and titles to determine which studies should be assessed further following an initial pilot test of the inclusion criteria. Two review authors will categorise all potentially relevant records as full‐text or map records to studies and classify them as included studies, excluded studies, studies awaiting classification, or ongoing studies, according to the criteria in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2022). Any disagreements between the two review authors will be resolved through consensus or by recourse to a third review author (BR). If a resolution is not possible, we will designate the corresponding study as 'awaiting classification'. We will contact the study authors as needed for clarification to determine the health status or diagnostic criteria of included participants. If we receive no response, clinical experts in our review group will classify the study, or we will list studies as 'awaiting classification'. We will document the reasons for the exclusion of excluded studies in the 'Characteristics of excluded studies' table. We will present a PRISMA flow diagram showing the study selection process (Page 2021).

Data extraction and management

We will develop a dedicated data extraction form based on the Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modelling Studies (CHARMS) checklist adapted for prognostic factor studies (Moons 2014; Riley 2019). We will pilot‐test the data extraction form ahead of time. Two review authors (BB and JVAF) will independently abstract the following information from included studies, which we will present in the 'Characteristics of included studies' table.

• Study design (e.g. prospective or retrospective cohort) and sources (surveys, laboratory data, clinical health records, claims data, etc.).

• Study dates, settings and country, inclusion/exclusion criteria.

• Participants: eligibility and recruitment method, characteristics at baseline, including data on other potential risk factors, including some of the following: diabetes, undernutrition, HIV infection, recent TB infection, history of untreated or inadequately treated TB disease, immunosuppressive therapy, cigarette smokers, drug or alcohol use disorders, and socioeconomic status, amongst others.

• Sample size: calculation, number of participants, outcomes used to define sample size, and number of events.

• Prognostic factors: definition and method of measurement of undernutrition, including timing and handling of prognostic factors in the statistical modelling.

• Outcome to be predicted: definition (TB type (pulmonary or extrapulmonary)) and method for measurement of outcome(s), type of outcome(s), time of outcome(s), occurrence, or summary of duration of follow‐up.

• Missing data: number of participants with any missing value, details of attrition (loss to follow‐up); for time‐to‐event outcomes: number of censored observations, and handling of missing data.

• Analysis: modelling method, assumptions, unadjusted and adjusted prognostic effect estimates, adjustment factors used.

• Results: interpretation of presented results, comparison with other studies, discussion of generalisability, strengths, and limitations.

• Study funding sources.

• Declarations of interest by primary investigators.

We will resolve any disagreements by discussion or, if required, by consultation with a third review author (BR).

We will provide information, including study identifiers, about potentially relevant ongoing studies in the 'Characteristics of ongoing studies' table.

We will contact the authors of included studies to obtain key missing data as needed.

Dealing with duplicate and companion publications

In the case of duplicate publications, companion documents, or multiple reports relating to a primary study, we will maximise the yield of information by mapping all publications to unique studies and collating all available data. We will use the most complete data set aggregated across all known publications. In case of doubt, we will give priority to publications reporting the longest follow‐up times associated with our primary or secondary outcomes.

Assessment of risk of bias in included studies

We will use the Quality In Prognosis Studies (QUIPS) tool to assess the risk of bias of included studies (Hayden 2013), including the following specifications on the use of this tool for our review question.

• Study participation

  • Source population (e.g. general population, hospitalised people, or chosen by a diagnosis or location)

  • Sampling frame, recruitment methods, and study participation

  • Period and place of recruitment

  • Inclusion and exclusion criteria (e.g. age cut‐offs or explicit diagnostic criteria)

  • Baseline characteristics: diabetes, undernutrition, sex and age, history of TB, immunosuppressive therapy or status including HIV infection, alcohol use disorder, tobacco use, and socioeconomic status. In studies focused on children, some of these might not be relevant (diabetes, tobacco, and alcohol use).

• Study attrition

  • Proportion available for analysis

  • Attempts to collect data on the characteristics of participants who dropped out, including reasons, outcome, and prognostic factor information and differences from those with complete data

• Prognostic factor measurement

  • Clear, valid, and reliable definition and measurement of undernutrition (weight and height) for all participants

  • Adequate categorisation of the continuous variables for BMI and weight‐for‐age for undernutrition

  • Proportion and handling of missing data on prognostic factors

• Outcome measurement

  • Clear, valid, and reliable definition and measurement of TB disease (microbiological, clinical, radiological, or obtained through pharmacy or clinical health records) for all participants

• Adjustment for other prognostic factors

  • Clear, valid, and reliable definitional measurement of relevant other prognostic factors for all participants

  • Methods used to handle missing data on other prognostic factors

  • Important other prognostic factors are accounted for in the study design or analysis (matching, stratification, restriction or adjustment)

    • For adults, important other prognostic factors include sex, age, diabetes, immunosuppression, alcohol, and tobacco.

    • For children, important other prognostic factors include at least two of the following: sex, age, immunosuppression, history of TB, or socioeconomic status.

• Statistical analysis and reporting: presentation of the analytical strategy, model development, and reporting of results

Two review authors (BB and JVAF) will judge each domain at low, moderate, high, or uncertain risk of bias according to the criteria set out in the tool, considering specificities of rating for observational studies, following an initial pilot test of the rating criteria. We will add an unclear risk option for judging the domains given the expected inconsistent reporting of the prognostic studies. We will also reach an overall judgement of low risk of bias for studies rated as low risk of bias in the domains of adjustment and statistical analysis and reporting. We will also explore this definition of low risk of bias in a sensitivity analysis.

We will resolve any disagreements by discussion or, if required, by consultation with a third review author (BR).

Measures of association to be extracted

If we are unable to retrieve the necessary information to analyse time‐to‐event outcomes, we will assess for dichotomous outcomes the risk ratios and odds ratios in order to maximise the use of data, including 95% confidence intervals (CIs) and P values. We will focus on association measures that were adjusted for other prognostic factors (diabetes, sex and age, history of TB, immunosuppressive therapy or status, alcohol use disorder, tobacco use, and socioeconomic status), but we will also extract and present unadjusted estimates. The type of effect measure will be considered as a source of heterogeneity.

Dealing with missing data

We will contact study authors for additional or updated information that is not available from the published articles. If measures of the association have not been reported, or details on precision are missing, we will estimate these from other information, including sample size, number of events, log‐rank P values and CIs, and Kaplan‐Meier curves (Altman 2003; Parmar 1998; Tierney 2007).

Assessment of heterogeneity

Between‐study heterogeneity concerning the role of undernutrition as a risk factor for tuberculosis can be attributed to three crucial aspects, as follows.

• Clinical heterogeneity: this encompasses differences in study populations (e.g. age and comorbidities), variations in the definition of the exposure (undernutrition), and co‐interventions (e.g. use of antiretrovirals) within study cohorts. Additionally, differences in outcome measurement, based on culture, clinical assessment and imaging, or indirect assessments (e.g. codes in electronic health records or prescriptions of antituberculosis medication).

• Methodological heterogeneity: this refers to the diversity of methods employed and the extent to which studies were robustly conducted, considering the risk of bias. Variations in the approach to analysis also contribute to methodological heterogeneity. Sample size and potential covariates in regression models can affect the reliability and generalisability of the findings.

• Statistical heterogeneity: we will quantify this using Tau². We will also assess between‐study heterogeneity using the I² statistic following the guidance in the Cochrane Handbook (Higgins 2022), considering:

  • 0% to 40%: might not be important;

  • 30% to 60%: may represent moderate heterogeneity;

  • 50% to 90%: may represent substantial heterogeneity;

  • 75% to 100%: considerable heterogeneity.

Assessment of reporting bias

It is likely that reporting bias is an important problem in prognostic factor research. In addition, the prospective registration of prognostic factor studies is not common, making it impossible to check whether there were reporting deficiencies for all included studies (Peat 2014).

We will explore the presence of small‐study effects by creating a funnel plot with the measure of association and its standard error. We will perform a test for funnel plot asymmetry using the R package 'metamisc' and produce contour‐enhanced funnel plots (Debray 2019). We will interpret the results with caution, as these tests often have limited power to detect asymmetry, and many tests yield inadequate type‐I error rates (Debray 2018).

Data synthesis and meta‐analysis approach

If possible, we will perform a meta‐analysis for the adjusted hazard ratios, risk ratio, or odds ratio for the incidence of active tuberculosis and the incidence of any type of tuberculosis disease. We will present the results of different effect measures separately and will stratify our analyses by time points. We will only conduct meta‐analyses when studies are sufficiently similar and other covariates in the models are similar. If there are sufficient studies at low risk of bias (e.g. 10 or more), we will restrict our main meta‐analyses to this subgroup. We will conduct our main meta‐analyses on studies that adjust the analysis for the following risk factors (CDC 2021).

• For adults: sex, age, diabetes, immunosuppression, alcohol, and tobacco

• For children: at least two of the following: sex, age, immunosuppression, history of TB, or socioeconomic status

We will use a random‐effects approach for meta‐analyses, employing the restricted maximum likelihood estimation in the R package 'metafor'. In addition to calculating a CI around the pooled estimate, we will calculate a 95% (approximate) prediction interval (Riley 2011).

Where meta‐analysis is not possible, we will conduct alternative forms of synthesis, including the summary of effect estimates (McKenzie 2022).

Subgroup analysis and investigation of heterogeneity

If there is considerable heterogeneity as judged by the prediction interval, we will consider performing a meta‐regression analysis to explore possible causes. We will consider parameters related to clinical and methodological heterogeneity as described above. Whilst this method can identify effect modifiers, with the possibility of exploring simultaneously more variables, including continuous variables, it requires sufficient data to yield informative results. We will therefore also explore heterogeneity through subgroup analyses considering the following factors.

• Age and sex

• HIV status or immunosuppression status

• Socioeconomic status

• Local/geographical TB incidence

• Definition and degree of undernutrition (mild, moderate, and severe undernutrition, e.g. marasmus, kwashiorkor, or BMI < 16 kg/m²)

• Type of outcome data: time‐to‐event versus dichotomous outcome data

We will interpret the results of subgroups with caution, considering formal statistical testing as indicated in Section 10.11.3.1 of the Cochrane Handbook (Higgins 2022). If none of these approaches is possible, we will perform a narrative discussion of sources of clinical heterogeneity.

Sensitivity analysis

We will estimate the robustness of our analysis by performing the following sensitivity analyses.

• Risk of bias: restricting our analysis to studies at low risk of bias for the domains of adjustment and statistical analysis and reporting or all domains.

• Complete information: restricting the analysis to studies that reported complete information on risk estimates and precision versus those for which we had to calculate this, based on other reported information as described above in the section 'Dealing with missing data'.

• Adjustment for risk factors: including studies with fewer risk factors in their analysis.

Summary of findings table

We will present absolute risks in summary of findings tables by each defined outcome (incidence of active tuberculosis and incidence of all forms of tuberculosis). We will use the GRADE approach for prognostic reviews to assess the quality of evidence of the listed outcomes (Foroutan 2020; Iorio 2015). The GRADE system classifies the certainty of evidence into one of four grades, as follows.

• High: we are very confident that the variation in risk associated with the prognostic factor (probability of future events in those with/without the prognostic factor) lies close to that of the estimate.

• Moderate: we are moderately confident that the variation in risk associated with the prognostic factor (probability of future events in those with/without the prognostic factor) is likely to be close to the estimate, but there is a possibility that it is substantially different.

• Low: our certainty in the estimate is limited: the variation in risk associated with the prognostic factor (probability of future events in those with/without the prognostic factor) may be substantially different from the estimate.

• Very low: we have very little certainty in the estimate: the variation in risk associated with the prognostic factor (probability of future events in those with/without the prognostic factor) is likely to be substantially different from the estimate.

The certainty of evidence can be downgraded by one (serious concern) or two levels (very serious concern) for the following reasons: risk of bias, inconsistency, indirectness, imprecision, and publication bias. Conversely, the certainty of the evidence can also be upgraded by one level due to a large summary effect.

Notes

The methods of this protocol are based on a previous protocol co‐authored by the lead author of this protocol (Garegnani 2023) and a recently published review (Perais 2023).

Appendices

Appendix 1. Search strategy

MEDLINE (PubMed)

#1 tubercul*[tw]

#2 diabet*[tw] OR BMI[tiab] OR body mass index[tiab] OR underweight[tiab] OR malnutri*[tw] OR undernutri*[tiab] OR wasting[tw] OR undernourish*[tiab] OR malnourish*[tiab] OR Kwashiorkor*[tw] OR nutritional deficien*[tiab]

#3 risk[tw] OR prevalence[tw] OR incidence[tw] OR cohort[tw]

#4 #1 AND #2 AND #3
(= 3,811)

WHO Global Index Medicus

Title, abstract subject: tubercul* AND (diabet* OR BMI OR "body mass index" OR underweight OR malnutri* OR undernutri* OR wasting OR undernourish* OR malnourish* OR Kwashiorkor* OR "nutritional deficien*")
(= 1,438)

WHO ICTRP (Standard search)

tubercul* AND (diabet* OR BMI OR "body mass index" OR underweight OR malnutri* OR undernutri* OR wasting OR undernourish* OR malnourish* OR Kwashiorkor* OR "nutritional deficien*")

(= 75)

PubMed IDs of 43 relevant articles used to develop the search strategy

3792471[PMID] OR 4992917[PMID] OR 8593374[PMID] OR 11532111[PMID] OR 13470303[PMID] OR 15220232[PMID] OR 16913973[PMID] OR 17278679[PMID] OR 17592104[PMID] OR 18400769[PMID] OR 18790239[PMID] OR 20505496[PMID] OR 20819255[PMID] OR 21109542[PMID] OR 22238171[PMID] OR 22331390[PMID] OR 22571347[PMID] OR 22791739[PMID] OR 23343909[PMID] OR 23510593[PMID] OR 23554875[PMID] OR 23631563[PMID] OR 24236069[PMID] OR 24278293[PMID] OR 24365557[PMID] OR 25310745[PMID] OR 25478954[PMID] OR 25884596[PMID] OR 26048371[PMID] OR 26459528[PMID] OR 27986673[PMID] OR 28169875[PMID] OR 29437750[PMID] OR 31319659[PMID] OR 31557282[PMID] OR 31931911[PMID] OR 32476692[PMID] OR 34926543[PMID] OR 35387594[PMID] OR 35709199[PMID] OR 35921384[PMID] OR 36171396[PMID] OR 36269203[PMID]

Contributions of authors

Juan Victor Ariel Franco, Agostina Risso, Maria‐Inti Metzendorf, and Brenda Bongaerts wrote the first draft of the protocol.

Johanna AAG Damen and Bernd Richter provided input on the review methods.

Annabel Baddeley, Jennifer Manne‐Goehler, Mathieu Bastard, Anna Carlqvist, Bianca Hemmingsen, Farai Mavhunga, Maria Nieves Garcia‐Casal, Kerri Viney, Melanie Boeckmann, and Sabrina Schlesinger provided content and methodological expert input for the development of this protocol.

All authors have approved this version of the protocol. Juan Victor Ariel Franco is the guarantor of this protocol.

Sources of support

Internal sources

• Institute of General Practice, Medical Faculty of the Heinrich‐Heine‐University Düsseldorf, Düsseldorf, Germany

  In‐kind support for JVAF, MIM, BB, and BR

External sources

• World Health Organization, Switzerland

  Funding for the team at the Heinrich Heine University Düsseldorf to support the conduct of this review

Declarations of interest

Juan Victor Ariel Franco: none known.

Maria‐Inti Metzendorf: none known.

Agostina Risso: none known.

Melanie Boeckmann: none known.

Sabrina Schlesinger: Novo Nordisk (Independent Contractor), Alpro Foundation (Grant/Contract).

Johanna AAG Damen: none known.

Bernd Richter: none known.

Annabel Baddeley: employee of the World Health Organization.

Anna Carlqvist: consultant for the World Health Organization.

Mathieu Bastard: employee of the World Health Organization.

Maria Nieves Garcia‐Casal: employee of the World Health Organization.

Bianca Hemmingsen: employee of the World Health Organization.

Farai Mavhunga: employee of the World Health Organization.

Jennifer Manne‐Goehler: none known.

Kerri Viney: employee of the World Health Organization.

Brenda Bongaerts: none known.

Juan Victor Ariel Franco, Maria‐Inti Metzendorf, Brenda Bongaerts, and Bernd Richter are editors for the Cochrane Metabolic and Endocrine Disorders Group, but they were excluded from the editorial processing of this protocol.

New