Scolaris Content Display Scolaris Content Display

Cochrane Database of Systematic Reviews Protocol - Intervention

The effects of oral vitamin D supplementation on linear growth and non‐communicable diseases among infants and children younger than five years of age

This is not the most recent version

Collapse all Expand all

Abstract

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

To assess effects of oral vitamin D supplementation on preventing and treating stunting and non‐communicable diseases among infants and children younger than five years of age.

Background

Description of the condition

Improving health among children younger than five years of age

Suboptimal health among children younger than five years of age remains a major global challenge (UNICEF, WHO, World Bank, UN 2015; WHO 2016). Most of the 5.9 million deaths among children younger than five years of age in 2015 could be attributed to preventable causes with available treatment options, including malnutrition and other non‐communicable diseases (UNICEF, WHO, World Bank, UN 2015).

Stunting

Stunting is the most prevalent form of undernutrition globally among children younger than five years of age (UNICEF, WHO, World Bank 2017). Nearly 155 million children younger than five years of age had stunting in 2016 (UNICEF, WHO, World Bank 2017); furthermore, two of every three of these children live in low‐ and middle‐income countries (UNICEF, WHO, World Bank 2017).

Child stunting, defined as measurements more than two standard deviations (SDs) below the World Health Organization (WHO) reference standard (length‐ or height‐for‐age z‐score [LAZ or HAZ, respectively]; WHO 2006), reflects poor cumulative nutrition in utero and postnatally (Dewey 2011). With prolonged undernourishment, the human body conserves energy in various ways, including reducing the rate of linear growth during childhood (Dewey 2011). Critically, childhood stunting is associated with numerous indicators of poor health (shorter adult height, reduced lean body mass, lower cognitive function, decreased physical work capacity) and low socioeconomic status (lower educational attainment, decreased earnings and wages) during later life (Black 2008; Dewey 2011; Haas 1996; Victora 2008). In terms of intergenerational implications, short maternal stature is linked with stunted offspring and other adverse birth outcomes (Martorell 2012; Prendergast 2014).

Given the widely recognized burden of disease caused by childhood stunting in diverse populations (Black 2008; Black 2013; De Onis 2012; Prendergast 2014), many global research and policy efforts have sought to reduce growth faltering (Victora 2010; WHO 2014a). It has been estimated that improved understanding and scaling up of effective evidence‐informed safe and effective interventions can prevent stunting among 33.5 million children (Bhutta 2013; Huey 2016; WHO 2014a). In particular, investigators have explored vitamin D supplementation as an intervention to prevent and mitigate childhood stunting (Kumar 2011). Optimal vitamin D status, which is often assessed by measuring calcifediol (i.e. 25[OH]D), allows calcium absorption and growth to support active vitamin D (i.e. calcitriol [1,25{OH}2D3]) (Holick 2010). Prolonged inadequate status of vitamin D impairs transcriptional regulation of skeletal homeostasis and linear growth, which could result in stunting (Holick 2010).

Prior observational studies have provided evidence that stunting is associated with suboptimal vitamin D (status or dietary intake) among children (Walli 2017). Therefore, vitamin D supplementation as a potentially modifiable risk factor that can prevent or treat stunting requires further evaluation.

Non‐communicable diseases

Non‐communicable diseases are the leading causes of death globally (68%; WHO 2014b), affecting children as well as adults. These diseases include diabetes, cancers, autoimmune disorders, and musculoskeletal disorders. Previous studies have noted respective associations between 25(OH)D and these non‐communicable outcomes (Holick 2010).

Description of the intervention

Vitamin D status

One billion people have suboptimal vitamin D status, according to global estimates (Holick 2010). Even in countries with sun exposure all year round, low vitamin D status is a global problem among all age groups (Palacios 2014). Consequences of low vitamin D include poor skeletal and extraskeletal health outcomes (Holick 2008a; Holick 2010).

Low circulating 25(OH)D serum concentration is widely regarded as the biomarker for vitamin D status (Heaney 2009), although cutoff values indicating deficiency and insufficiency are debated (Holick 2011; Ross 2011). Between 30% and 50% of children in numerous countries in Africa, Asia, Europe, and North America (Holick 2010), including geographic areas with ample sunlight and heterogeneous economic resources, have 25(OH)D less than 20 ng/mL. In the context of vitamin D deficiency, infants and young children are considered a high‐risk population, given that vitamin D intake is low during exclusive breastfeeding, and early life represents a critical period for linear growth and development of the immune system (Leroy 2014 and Shrimpton 2001; Adkins 2004 and Levy 2007, respectively). As further detailed in the next section, pleiotropic actions of vitamin D can impact skeletal, muscular, and immunological functions, all of which are related to optimal growth.

Vitamin D sources

Vitamin D can be acquired through consumption of a diet containing naturally vitamin D‐rich and fortified foods, or vitamin D supplements, or through endogenous production via skin exposure to ultraviolet irradiation (Holick 2010). In this review, we focus on vitamin D supplementation, given that it overcomes the challenges of inadequate sunlight at some geographic latitudes, as well as minimal sun exposure based on individual lifestyle decisions and limited consumption of naturally vitamin D‐rich or fortified foods (Holick 2010). Vitamin D supplements are available in two chemical forms (ergocalciferol [D2] and cholecalciferol [D3]), which differ in their side‐chain structure (Holick 2010). Vitamins D2 and D3 have been observed to increase serum 25(OH)D; although at high doses (50,000 IU), vitamin D2 appears less potent than equivalent doses of D3 in maintaining serum 25(OH)D levels (Holick 2010).

Vitamin D requirements

According to the WHO and the Food and Agriculture Organization (FAO), 200 international units (IU) is the daily recommended nutrient intake (RNI) among children younger than five years of age (WHO, FAO 2004). In the United States, the Institute of Medicine recommends that children between one and five years of age should consume a recommended dietary allowance of 600 IU per day and have an estimated average requirement (EAR) of 400 IU per day (Institute of Medicine 2011). From birth to 12 months, it is recommended that children in the United States consume an adequate intake (AI) of 400 IU per day (Institute of Medicine 2011).

No adverse effects occur at vitamin D intakes recommended by WHO and by FAO (WHO, FAO 2004). In the United States, the recommended upper limits of vitamin D consumption are based on age: 1000 IU (birth to six months), 1500 IU (six to 12 months), 2500 IU (one to three years), and 3000 IU (four to five years) (Institute of Medicine 2011). Vitamin D toxicity has been observed in a few rare cases with long‐term consumption of extreme pharmaceutical dosages (Barrueto 2005; Blank 1995; Holick 2011; Klontz 2007; Vieth 1999). Excess vitamin D may contribute to hypercalciuria, hypercalcemia, hyperphosphatemia, and kidney stones (nephrolithiasis) (Holick 2010). Vitamin D toxicity is caused primarily by excessive intestinal calcium absorption and bone resorption (Holick 2010).

Metabolism of vitamin D

Evidence from mechanistic and dose‐response studies suggests that increasing intake of vitamin D (via consumption [supplementation, dietary intake] or cutaneous synthesis) improves serum 25(OH)D concentration (Holick 2010; Holick 2011). After it enters the body, vitamin D is stored in fat or is metabolized by the liver (Holick 2010; Holick 2011). A 25‐hydroxylase (CYP27B1) in the liver converts vitamin D to 25(OH)D, which is the major circulating form (Holick 2010; Holick 2011).

Available data from dose‐response studies show that vitamin D supplementation increases serum 25(OH)D concentration, regardless of age (Heaney 2003; Holick 2008b; Holick 2010; Institute of Medicine 2011). A non‐linear response of 25(OH)D to vitamin D has been observed in murine and human models (Institute of Medicine 2011). Dosages greater than or equal to 1000 IU daily have resulted in more gradual responses (e.g. 0.95 nmol/L to 1.4 nmol/L for every 100 IU; Smith 2009), and dosages below 1000 IU daily have achieved steeper responses (e.g. approximately 2.0 nmol/L for every 40 IU; Cashman 2008; Cashman 2009) (Institute of Medicine 2011). Moreover, studies including young children with stunting have confirmed that vitamin D supplementation increases 25(OH)D (Kumar 2011). Widely ranging vitamin D supplementation dosages across studies have included daily physiological doses (200 IU to 400 IU; Fort 2016; Hollis 2015), as well as pharmacological doses (50,000 IU at birth; Moodley 2015), and even a single dose of 100,000 IU (Gupta 2016). In summary, preliminary data highlight the need for assessment of potential beneficial effects of vitamin D supplementation on stunting among children.

How the intervention might work

Cells of kidney, immune system, bone, epithelium and of other tissues in the body use 1‐OHase (CYP27R1) to metabolize 25(OH)D to the biologically active steroid hormone 1,25(OH)2D (Bikle 2014; Christakos 2016; Holick 2010). In its hormonally active form, vitamin D plays pleiotropic roles in the human body, promoting skeletal health, muscle development and growth, and immune function. We have provided further details in the following sections (Bikle 2014; Christakos 2016; Holick 2010).

1,25(OH)2D functions through genomic and non‐genomic mechanisms (Bikle 2014; Christakos 2016; Holick 2010). First, genomic effects occur through binding of 1,25(OH)2D to vitamin D receptor and retinoid X receptor, which results in a heterodimer complex that regulates gene activity (Bikle 2014; Christakos 2016; Holick 2010). At least 100 to 1250 target genes of vitamin D are known (Adams 2010; Holick 2007; Hossein‐nezhad 2013; Ramagopalan 2010; Tarroni 2012). These are directly targeted by vitamin D (via a vitamin D response element; e.g. 1,25[OH]2D has been shown to bind to vitamin D response element in the calcium‐sensing receptor gene and subsequently to modulate calcium‐sensing receptor expression [Bikle 2014; Canaff 2002; Christakos 2016; Holick 2010]). Second, "rapid" or non‐genomic responses occur extracellularly via interaction with plasma membrane vitamin D receptor (VDR) (Bikle 2014; Christakos 2016; Holick 2010). Examples of these include stimulation of intestinal calcium absorption and inhibition of apoptosis in osteoblasts (Bikle 2014; Christakos 2016; Holick 2010). This nuclear receptor has been identified in nearly all human tissues and cells assessed (Bikle 2014; Christakos 2016; Holick 2010).

Skeletal homeostasis and linear growth

Vitamin D has well‐established effects on skeletal health, including bone mineralization and maintenance (Holick 2010). Active vitamin D (1,25[OH]2D) functions in conjunction with two other hormones (parathyroid hormone and calcitonin) to maintain endocrine control of calcium and phosphorus concentrations (Holick 2010). This tight regulation of calcium and phosphorus flux (extracellular [bones, blood], intracellular) is critical for development and maintenance of bones (Holick 2010), which prevents and addresses growth faltering. Specific roles of active vitamin D include increasing intestinal calcium absorption (Christakos 2012), renal calcium reabsorption, and skeletal calcium resorption (in conjunction with parathyroid hormone) (Holick 2010).

Previous studies have demonstrated that vitamin D deficiency is associated with stunting (Holick 2010), including stunting among children (Holick 2006; Wacker 2013). Maternal vitamin D deficiency has been associated with greater risk of stunting among neonates and children (Finkelstein 2012; Toko 2016).

Muscle development and growth

Vitamin D may influence muscle mass and function, as well as related indicators (weight‐for‐height [WFH] and ‐age [WFA]). Observational studies have corroborated the link between severe vitamin D deficiency (≤ 8 ng/mL) and poor muscle health among individuals aged 10 to 65 years (Plotnikoff 2003). As an example, among infants with HIV exposure and no infection, low 25(OH)D concentration (< 10 ng/mL or ˜ 25 nmol/L) was associated with a higher incidence of wasting (hazard ratio 1.71, 95% confidence interval [CI] 1.20 to 2.43; Sudfeld 2015).

Previous studies have identified mechanisms that link vitamin D with myopathy (Bischoff‐Ferrari 2012). In vitro studies have assessed human muscle tissues and isolated VDR (Bischoff‐Ferrari 2004; Bischoff‐Ferrari 2012; Ceglia 2010; Simpson 1985), which facilitate genomic and non‐genomic effects (Haussler 1998; McDonnell 1987; Norman 2004; Vazquez 1998). Furthermore, murine models have demonstrated that deletion of VDR (via gene knockout) resulted in impaired skeletal muscle growth and muscle‐related gene expression (Bouillon 2008a; Bouillon 2008b; Endo 2003). Mice without VDR had smaller muscle fibers in all striated muscles (Endo 2003).

Immunity and other non‐communicable diseases

Vitamin D is an important immunomodulator that may affect non‐communicable diseases such as atopy and cancers (Holick 2010). Immune cells (including macrophages) express the enzyme that converts 25(OH)D to the active form, which facilitates the immunological role of 1,25(OH)2D (Holick 2007; Holick 2010). Given this information, another systematic review considered previous studies focusing on effects of vitamin D supplementation on infections among children younger than five years of age (Yakoob 2016). Separately, 25(OH)D serum concentrations were associated with atopic diseases (Jones 2015), cancers, and chronic disease indicators (blood pressure [Scragg 2007], low‐ and high‐density lipoprotein concentration ratio [Carbone 2008], and diabetes [Christensen 2016; Holick 2010]). However, findings are inconsistent and limited, especially among children.

Why it is important to do this review

Global stunting remains a critical and complex challenge in numerous geographic regions (De Onis 2013; Prendergast 2014; UNICEF 2013; UNICEF, WHO, World Bank 2017). This is reflected in the World Health Assembly nutrition target to reduce stunting by 40% among children younger than five years of age by 2025 (WHO 2012; WHO 2014a). Although stunting among children younger than five years of age has decreased from 39.7% (in 1990) to 26.7% (in 2010) (De Onis 2012), the World Health Assembly nutrition target will not be achieved at this current trajectory (De Onis 2013).

Linear growth is considered an important overall indicator of child development (De Onis 2016). Critically, many children with stunting often show minimal (if any) catch‐up growth in later life (Martorell 1994). However, nutritional interventions have been seen to allow catch‐up growth among children (Martorell 1994), especially during key developmental windows (including between birth and five years; Prentice 2013).

The systematic method of our review is intended to achieve comprehensive assessment of current evidence on effects of vitamin D supplementation on growth faltering and other health outcomes among children. This approach facilitates consideration of other modulatory factors, particularly in subgroup analyses. Given the multifactorial origin of stunting, which needs further elucidation (Stewart 2013), accounting for other factors is important. Aside from nutritional factors that affect stunting, potential influences include repeated infections, poor sanitation, household environmental contamination, mycotoxin exposure, the gut, and associated enteropathy (Casanovas 2013; Owino 2016; Semba 2016; Stewart 2013; Waterlow 1994).

Separately, an estimated one billion people have suboptimal vitamin D status (Holick 2007), which is linked to numerous skeletal and extraskeletal outcomes (Holick 2010). Despite the multitude of studies that focus on vitamin D supplementation and clinical health indicators (Ferguson 2014; Jagannath 2010), particularly among adults (Avenell 2014; Bjelakovic 2014a; Bjelakovic 2014b; De‐Regil 2016; Straube 2015), evidence regarding growth and stunting among children younger than five years of age remains unclear. In summary, it is crucial to delineate modifiable risk factors and causes, as well as effective interventions against stunting, including vitamin D supplementation.

Objectives

To assess effects of oral vitamin D supplementation on preventing and treating stunting and non‐communicable diseases among infants and children younger than five years of age.

Methods

Criteria for considering studies for this review

Types of studies

Randomized controlled trials (RCTs) and quasi‐RCTs. Quasi‐RCTs include studies that do not involve a treatment regimen assignment with simple randomization but systematically utilize another aspect of the study design (e.g. alternating assignments based on sequential study enrolment, medical record number). We will include cluster‐randomized and cross‐over trials that meet all other criteria.

Types of participants

Infants and children younger than five years of age who live in any country. We will include studies of children younger than five years of age and five years of age and older (e.g. birth to 10 years) if study authors report stratified outcomes; we will extract results among children younger than five years of age. We will include only studies with vitamin D supplementation among infants and children younger than five years; we will exclude studies that provide vitamin D supplementation only to mothers and not to their offspring.

Types of interventions

Studies reporting any of the comparisons between intervention and comparator groups (direct comparisons indicated by the same numbers) (Table 1) described below.

Open in table viewer
Table 1. Intervention and comparator groups

Comparison

Intervention group

Comparator group

1. Oral vitamin D (cholecalciferol D3, ergocalciferol D2, calcitriol) supplementationa

1a. No intervention

1b. Placebo

2. Other micronutrientsb, including oral vitamin D (cholecalciferol D3, ergocalciferol D2, calcitriol) supplementationa

2. Other micronutrientsb not including vitamin D

aAny form, including capsules, soft gels, liquids, or powders.
bComparisons will include intervention and comparator groups with the same combination of various micronutrients, to isolate the effect of vitamin D.

Interventions

Oral vitamin D (cholecalciferol D3, ergocalciferol D2, calcitriol) supplementation (Table 1). We will include any form of oral consumption of vitamin D (such as capsules, soft gels, liquids, and powders) and will exclude alternative administration of vitamin D (e.g. intravenous injection, food fortification, dietary intake of vitamin D‐rich foods). We will consider studies providing other micronutrients that include oral vitamin D, but only if intervention and control groups differ by vitamin D supplementation. We will document key differences across interventions (including treatment dosage, duration, and frequency) during data extraction.

Comparators

Study participants who receive placebo or no intervention (Table 1). Additionally, for studies with other micronutrient supplementation that include vitamin D as the intervention, we will use the same other micronutrients without vitamin D as the reference group.

Types of outcome measures

Primary outcomes

  1. Linear growth (reported continuously in centimeters)

  2. Height (or length)‐for‐age (reported continuously as WHO z‐score, i.e. HAZ [or LAZ]; WHO 2006)

  3. Stunting (categorical outcome defined as HAZ [or LAZ] more than two SDs below the reference WHO standard; WHO 2006)

  4. Adverse effects relevant to excessive vitamin D (as categorical outcomes)

    1. Hypercalciuria

    2. Hypercalcemia

    3. Hyperphosphatemia

    4. Kidney stones

Secondary outcomes

  1. Weight‐for‐age (reported as a continuous WHO z‐score, i.e. WAZ; WHO 2006)

  2. Weight‐for‐height (reported as a continuous WHO z‐score, i.e. WHZ; WHO 2006)

  3. Vitamin D status (based on serum 25[OH]D concentration [nmol/L]; considered as continuous and categorical variables, according to current recommended cutoffs from the Institute of Medicine and the Endocrine Society [United States]). Usage of a wide spectrum of vitamin D assay instruments, including immunoassays (e.g. radioimmunoassays) and chromatographic methods (e.g. liquid chromatography‐tandem mass spectrometry)

  4. Rickets (as defined by trialists)

  5. Atopic diseases (i.e. asthma, including recurring wheeze, dermatitis, and/or rhinitis; as defined by trialists)

  6. Other non‐communicable disease outcomes (i.e. bone health, number of fractures, bone mineral density, any type of cancer, type 1 and type 2 diabetes mellitus, insulin resistance, and other autoimmune disorders; as defined by trialists)

Search methods for identification of studies

Electronic searches

We will search international and regional electronic databases and trial registers as listed below.

  1. Cochrane Central Register of Controlled Trials (CENTRAL; current issue) in the Cochrane Library, which includes the Cochrane Developmental, Psychosocial and Learning Problems Specialised Register.

  2. PubMed National Library of Medicine (www.ncbi.nlm.nih.gov/pubmed).

  3. Embase Ovid (1980 onwards).

  4. Cumulative Index to Nursing and Allied Health Literature EBSCO (CINAHL; 1982 onwards).

  5. Centre for Agriculture and Biosciences International (CABI): CAB Abstracts and Global Health Web of Science (1973 onwards).

  6. Conference Proceedings Citation Index ‐ Science Web of Science (CPCI‐S; 1990 onwards).

  7. Conference Proceedings Citation Index ‐ Social Science & Humanities Web of Science (CPCI‐SS&H; 1990 onwards).

  8. Science Citation Index ‐ EXPANDED Web of Science (SCI‐EXPANDED; 1970 onwards).

  9. Social Sciences Citation Index Web of Science (SSCI; 1970 onwards).

  10. Cochrane Database of Systematic Reviews (CDSR; current issue), part of the Cochrane Library.

  11. Database of Abstracts of Reviews of Effects (DARE; current issue), part of the Cochrane Library.

  12. IBECS (ibecs.isciii.es).

  13. Latin American and Caribbean Health Sciences Literature (LILACS; lilacs.bvsalud.org/en).

  14. Scientific Electronic Library Online (SciELO; www.scielo.br).

  15. Pan American Health Library (PAHO; www1.paho.org/english/DD/IKM/LI/library.htm).

  16. WHO Library (WHOLIS; dosei.who.int).

  17. Western Pacific Region Index Medicus (WPRO; www.wprim.org).

  18. Index Medicus for the South‐East Asia Region (IMSEAR; imsear.hellis.org).

  19. Indian Medical Journals (IndMED; indmed.nic.in).

  20. WHO International Clinical Trials Registry Platform (ICTRP; apps.who.int/trialsearch).

  21. EU Clinical Trials Register (www.clinicaltrialsregister.eu/ctr‐search).

  22. Epistemonikos (limited to systematic reviews; www.epistemonikos.org).

  23. Scopus (limited to conference papers; www.scopus.com).

We present our search strategy for PubMed in Appendix 1. We will modify this search strategy for other databases, as appropriate. We will apply no restrictions regarding publication year, language, country, or region.

In the event of finding additional search terms during the searching and screening process, we will update our electronic search strategies and will report these in the completed review.

Searching other resources

We will search the reference lists of publications (including trials, reviews, meta‐analyses, reports) identified through our searches of electronic databases, and will consider any relevant trials included in these reference lists. Additionally, we will attempt to obtain information on relevant ongoing and unpublished trials by contacting other entities such as the WHO, the United Nations Children’s Fund (UNICEF), Nutrition International (formerly Micronutrient Initiative), the International Micronutrient Malnutrition Prevention and Control Programme (IMMPaCt) from the US Centers for Disease Control and Prevention (CDC), and the Vitamin D Workshop Group.

Data collection and analysis

Selection of studies

We will design a customized data extraction form for use during screening of studies for this review. Using this form, two review authors (SH, EY) will independently screen studies identified by the electronic search (Electronic searches). Initially, SH and EY will consider the title and abstract of each record to decide whether studies meet the inclusion and exclusion criteria of this review (Criteria for considering studies for this review). For records that are not excluded, SH and EY will subsequently review full‐text reports to assess eligibility. We will contact study authors if clarifications are necessary, or if full‐text reports are not available. SH and EY will resolve discrepancies through discussion and, if necessary, through consultation with a third review author (SM). We will present the study selection procedure in a PRISMA diagram (Moher 2009).

Data extraction and management

From studies deemed eligible for inclusion (Criteria for considering studies for this review), two review authors (SH, EY) will independently extract onto data extraction templates information on study design (including intervention, participants, trial identification numbers, if available), results, and adverse events. Data extraction forms will be specifically designed for this review. We will pilot these templates on a small subset of studies and will modify them, if necessary, before commencing data extraction. We will enter study data into Review Manager 5 (Review Manager 2014).

We will contact study authors for additional information, if necessary. SH and EY will resolve disagreements through discussion or through consultation with a third review author (SM). For this review, we will aggregate study design details and findings from any duplicate or companion documents, as well as from multiple publications on a single study.

Assessment of risk of bias in included studies

For each study, we will rate risk of bias as high, low, or unclear across the domains listed below, based on criteria provided in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2017). We will present our findings in a 'Risk of bias' table and summary figure.

Selection bias

Each participant in studies included in this review should have an equal chance of assignment to study arms. To assess selection bias in terms of random sequence generation, we will identify the method used to generate the allocation sequence in sufficient detail to assess whether it could produce comparable groups. The random allocation sequence should not be known before randomisation. To measure risk of bias in terms of allocation concealment, we will consider the method used to conceal the allocation sequence in detail, to determine whether participants were (or could have been) aware of the allocation before or during enrollment.

Criteria for low risk of bias will involve referring to a random number table and using a computer random number generator. Examples of low risk of bias include central allocation and sequentially numbered units (drug containers or opaque and sealed envelopes) of the intervention that are identical in appearance. We will consider studies with inadequate concealment of allocation before assignment to have high risk of selection bias. Examples of inadequate concealment (e.g. a systematic, non‐random approach) include an open random allocation schedule and assignment by date of birth or case record number. When information is insufficient to permit a judgement of low or high risk of bias, we will categorize studies as having unclear risk of bias.

Performance bias

We will consider all measures and methods used to keep study participants and personnel blinded from intervention or comparator allocation, as well as information related to effectiveness of blinding. We will consider a study to have low risk of performance bias if it appears that lack of blinding does not influence the outcome, or if blinding was ensured and was not broken. If study participants or personnel, or both, may have been aware of assignment of the intervention or comparator among participants (e.g. no blinding, incomplete blinding, broken blinding), causing the outcome to be influenced by lack of blinding, we will consider the study to have high risk of performance bias. We will perform this assessment for each main outcome or class of outcomes. We will consider studies to be at unclear risk of bias when information is insufficient to permit a judgement of low or high risk of bias.

Detection bias

We will identify all measures used to blind those assessing each main outcome or class of outcomes from knowledge of which intervention a participant may have received, as well as whether intended blinding was effective. We will categorize a study as having high risk of detection bias if outcome assessors knew of the allocated interventions, leading to influence on the outcome by way of no blinding or broken blinding. We will consider a study to be at low risk of bias if blinding was ensured and was unlikely to have been broken. We will categorize studies as having unclear risk of bias if information is insufficient to permit a judgement of high or low risk of bias.

Attrition bias

For each study, we will consider the completeness of described outcome data for each main outcome or class of outcomes, including loss to follow‐up (i.e. attrition) and exclusions from analysis. We will state explicitly whether attrition and exclusions were reported, along with numbers in each intervention group as compared with total sample size, reasons for attrition and exclusion, and reasons for any re‐inclusions in analyses as performed by the review authors. We will consider no missing outcome data, reasons for any missing outcome data unlikely related to true outcomes, and balanced missing data across groups to indicate low risk of bias. We will consider the quantity, nature, or handling of incomplete outcome data, including missing data likely to be related to the true outcome or an 'as‐treated' analysis done with departure of the intervention received from that assigned at randomization to indicate high risk of attrition bias. We will categorize studies as having unclear risk of bias when we find insufficient information to permit a judgement of low or high risk.

Reporting bias

Among all studies in this review, we will evaluate the possibility of selective outcome reporting (Sterne 2011). Criteria for low risk of bias include that the study protocol is available and that prespecified outcomes have been reported in some way. We will assign high risk of reporting bias if not all of the study’s prespecified primary outcomes were reported; primary outcome(s) were reported via methods not prespecified; outcomes were not prespecified; outcomes were not reported completely and thus cannot be entered into the meta‐analysis; or investigators failed to include results for a key expected outcome. We will categorize studies as having unclear risk of bias when information is insufficient to permit a judgement of low or high risk of bias.

Other sources of bias

In addition to the sources of bias described above, we will assess studies for risk of any other type of bias. We will categorize a study as having high risk of bias if we detect at least one important risk of bias. Examples include bias related to a specific study design and claims of fraudulence. We will consider studies to have low risk of bias if it appears they do not have other sources of bias; we will consider studies to have unclear risk of bias if information is inadequate to permit a judgement of high or low risk.

Measures of treatment effect

Continuous outcomes

We will report continuous outcomes as mean differences (MDs) with corresponding 95% CIs (Deeks 2011). Specifically, these will include primary (linear growth; HAZ or LAZ) and secondary (WAZ, WHZ, serum 25[OH]D) outcomes. If trialists use different scales to measure the same continuous outcome across studies, we will use standardized mean differences (SMDs) with 95% CIs, when possible (Deeks 2011).

Categorical outcomes

For categorical outcomes, we will present data as measures of association (risk, rate, odds ratios with corresponding 95% CIs; Deeks 2011). These will include primary (stunting, adverse effects [hypercalciuria, hypercalcemia, hyperphosphatemia, kidney stones]) and secondary (vitamin D status, rickets, atopic diseases, other non‐communicable disease measures) outcomes.

Unit of analysis issues

For each study included in this review, we will document the unit of randomization during data extraction. For example, the unit of randomization could include individual children or clusters (households, communities, schools, classes). We will also consider whether individuals undergo more than one intervention, as in a cross‐over trial, and whether trialists report multiple observations for the same outcome(s), including repeated measurements or recurring events.

Cluster‐randomized trials

Among cluster‐randomized trials, we will account for randomization of study participant groups by conducting analysis at the cluster level. We will calculate effect estimates (with respective standard errors [SEs]) by using the generic inverse variance method presented in Review Manager 2014 (Higgins 2011). Depending on analyses of included studies, we will conduct approximately correct analyses, when possible (Higgins 2011).

Crossover trials

If suitable data are available, we will consider a paired analysis of continuous data from cross‐over trials. Specifically, we will assess data from a two‐period, two‐intervention cross‐over trial by using a paired t‐test to evaluate the difference between two measurements (subtracting the control measurement from the experimental measurement) for each study participant (Higgins 2011). For studies with potential carry‐over effects, we will consider only the first period of trial intervention follow‐up (Higgins 2011).

Studies with more than two treatment groups

When studies include more than two intervention groups, we will combine groups to perform a single pair‐wise comparison. Specifically, we will combine all relevant experimental groups into one group, and all relevant control intervention groups into a second group. Our approach to meta‐analysis will be based on information provided in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).

Dealing with missing data

As necessary, we will contact study authors to obtain unreported data that are of key interest for our review. We will not impute data that are missing and that we are unable to obtain. From each study, we will document the missingness of key data and study participant information (including loss to follow‐up) in 'Risk of bias' tables. Examples of unreported data include means and SDs of study participant subgroups. We will record attrition as part of the 'Risk of bias' assessment. Loss to follow‐up data may include additional information regarding attrition and treatment adherence; or data on study participants who did not complete the trial or follow the protocol.

We will consider all outcomes based on the intention‐to‐treat approach, when possible. In summarizing across studies, for every outcome, the denominator will represent the total number of study participants randomized to a treatment regimen (minus any participants with missing outcomes).

Assessment of heterogeneity

We will quantify statistical heterogeneity across studies by using forest plots, Chi2 (significance of α [alpha] = 0.10) testing, I2 (≥ 75%) statistics, and Tau2 values (Deeks 2011). We will also consider critical differences between study designs (including study population characteristics) and risk of bias. In the event that we observe substantial heterogeneity, we will consider performing prespecified subgroup analyses to gain a better understanding of the differences (Subgroup analysis and investigation of heterogeneity). For outcomes with substantial heterogeneity (according to our assessments), we will not report a pooled estimate.

Data synthesis

Among comparable studies in this review (including similar outcomes and populations), we may conduct a meta‐analysis to estimate summary measures across studies. Specifically, these will include studies with outcomes reported on the same scale (or as values that can be converted or standardized). For each outcome of interest, we will consider reporting both continuous and categorical values across studies; we will convert to either continuous or categorical values to facilitate comparability as recommended by the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2011).

We will conduct meta‐analysis via Review Manager 2014 and will utilize the inverse variance method. Additionally, we will conduct a random‐effects meta‐analysis to account for differences across study designs (including intervention dosages, durations, and frequencies, as well as study populations). We will anticipate heterogeneity of reported time points (by reporting endpoint data, change from baseline data, etc.). In the event that studies are too few or study data cannot be pooled, we will provide a narrative description of trial results.

We will extract and report the following information from each study.

  1. Age at time of intervention.

  2. Supplementation provided (frequency [daily, weekly]).

  3. Pharmacological versus physiological doses.

  4. Form of vitamin D supplemented: vitamin D2 versus D3.

  5. Dosage: low versus high dosages.

  6. Duration of supplementation.

  7. Dose per body weight.

  8. Serum 25(OH)D concentration at study baseline.

  9. Geographic latitude (between Tropics of Cancer and Capricorn, compared with north of Tropic of Cancer and south of Tropic of Capricorn).

  10. Season at start of supplementation or data collection for each outcome (spring, summer, fall, winter).

  11. Race or ethnicity of participants.

  12. Baseline HAZ or LAZ.

Summary of findings

For each primary outcome, two review authors (SH and EY) will use the GRADE approach to rate the quality of evidence as high, moderate, low, or very low, according to the presence of the following factors: within‐study risk of bias and limitations due to study design, directness of evidence, assessment of heterogeneity between studies, precision of effect estimates, and risk of publication bias (GRADEpro 2014; Guyatt 2011). For example, we will assign high quality to evidence from RCTs and will decrease this by one quality rating for each factor present, up to a maximum of three levels for all factors. In the event of disagreement, we will consult a third review author (SM or JP‐R), who will facilitate consensus through discussion. We will assess risk of bias for each included study using Cochrane's 'Risk of bias' tool (Higgins 2017); see Assessment of risk of bias in included studies. We will present quality ratings for primary outcomes in a GRADE 'Summary of findings' table, which will include the outcomes of interest.

We will create the 'Summary of findings' table using GRADEpro 2014 or Review Manager 2014. As a brief summary, tables will include information on the following: comparisons of interest; study design (population, intervention, comparison, outcome [PICO]; location; follow‐up duration); and outcome measurements. Additionally, for each primary outcome, we will include the evidence quality rating, as assessed through the GRADE approach (Guyatt 2011). We will include in table footnotes a rationale for the GRADE quality rating.

Subgroup analysis and investigation of heterogeneity

We plan to conduct the following subgroup analyses.

  1. Age at time of intervention (birth to 6 months of age versus 7 months to 12 months of age, 13 months to 36 months of age, and 37 months to 59 months of age).

  2. Frequency of supplementation (daily versus intermittent versus other).

  3. Serum 25(OH)D concentration at study baseline (current cutoff levels recommended by the Institute of Medicine and the Endocrine Society).

  4. Geographic latitude (between Tropics of Cancer and Capricorn, compared with north of Tropic of Cancer and south of Tropic of Capricorn).

  5. Season at start of supplementation or data collection for each outcome (spring, summer, fall, winter).

  6. Baseline HAZ or LAZ.

All subgroup analyses will include categories for studies with unknown, unreported, or mixed data; we will conduct these analyses on primary outcomes for which we find more than three studies contributing data.

Sensitivity analysis

We will perform sensitivity analyses to explore the influence of specific factors (listed below) on effect size.

  1. Restricting analysis to published studies.

  2. Accounting for risk and impact of bias (including removing studies at high risk of bias).

  3. Determining the influence of studies with longer intervention durations or greater sample sizes.

  4. Considering the influence of methods (including study design, such as differences between cluster‐randomized trials, cross‐over trials, other intervention trials) and sufficiency of allocation concealment and blinding status, along with percentage of attrition.

  5. Use of filters such as imputation, language of publication, source of funding (industry versus other), and country.

As specific examples, for outcomes with an adequate number of comparable studies (≥ 10), we will create funnel plots in Review Manager 2014. We will perform statistical tests (including Egger's test) to assess asymmetry of funnel plots and as indicators of bias (Egger 1997).

Table 1. Intervention and comparator groups

Comparison

Intervention group

Comparator group

1. Oral vitamin D (cholecalciferol D3, ergocalciferol D2, calcitriol) supplementationa

1a. No intervention

1b. Placebo

2. Other micronutrientsb, including oral vitamin D (cholecalciferol D3, ergocalciferol D2, calcitriol) supplementationa

2. Other micronutrientsb not including vitamin D

aAny form, including capsules, soft gels, liquids, or powders.
bComparisons will include intervention and comparator groups with the same combination of various micronutrients, to isolate the effect of vitamin D.

Figures and Tables -
Table 1. Intervention and comparator groups