Interventions to improve vitamin D status in at-risk ethnic groups during pregnancy and early childhood: a systematic review

Objective: To systematically review the literature with the primary aim of identifying behavioural interventions to improve vitamin D stores in children from at-risk ethnic groups. Design: Review based on Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. PROSPERO registration number: CRD42017080932. Health Behaviour Model and Behaviour Change Wheel framework constructs used to underpin evaluation of interventions. Methodological quality evaluated using Cochrane Risk of Bias, Cochrane ROBINS-I and NHLBI tools. Setting: Databases Cochrane Library, MEDLINE, EMBASE, CINAHL with secondary search of Google Scholar. No country limits set. Papers between January 1990 and February 2018, published in English included. Anticipating study heterogeneity, outcome measures not pre-specified and identified from individual full papers. Updated literature search November 2020. Participants: Patient or population including pregnant women, newborns and children aged under 18 years, from Asian or African ethnic groups. Results: Of 10 690 articles screened, 298 underwent full-text review, with 24 ultimately included for data extraction. All identified studies conducted a vitamin D pharmacological supplementation intervention, with two also incorporating a behavioural intervention strategy. No study explicitly defined a primary aim of evaluating a behavioural intervention, undertaken to study its effect on vitamin D supplement uptake. Conclusions: There is a need to address the paucity of data in ethnic at-risk children on how behavioural interventions ideally developed and co-produced with the community under study, affect and help improve vitamin D uptake, within the antenatal and pregnancy phase as well as during childhood.

The major cause of vitamin D deficiency is low transmission or reduced penetration of solar ultra-violet B radiation, affecting the cutaneous synthesis of vitamin D. Although found naturally in some foods such as oily fish, red meat, liver and egg yolks and fortified foods (infant formula, breakfast cereals and fat spreads), less than 10 % of vitamin D stores come from individual diets (3,16) . Often, the main source of dietary vitamin D is in the form of supplements. In women of child-bearing age, supplementation rates range between 12 and 27 % (17) , and data indicate that 63 % women of reproductive age are vitamin D-deficient (18) . The Royal College of Paediatrics and Child Health and the British Paediatric Surveillance Unit recently ascertained the national incidence of nutritional rickets in children under the age of 16 years (2) . Black and South Asian children had a 10-fold and 5-fold greater incidence of nutritional rickets, respectively, compared with other ethnic groups under 5 years of age.
Despite current national policy and guidance to ensure supplementation in high-risk groups (1,3,19,20) , uptake is not consistent and a deeper understanding of facilitators and barriers to improve uptake is required. Behavioural interventions have shown some promise in contributing to the prevention, management and treatment of various other non-communicable diseases; obesity, diabetes, chronic pain, asthma and emotional difficulties (21) . In this paper, we refer to ethnic minority groups, accepting that the term is often used interchangeably with race (22) . Race is a social construct, usually identified based on a combination of physical, cultural and behavioural attributes, whereas ethnicity, self-identified by an individual, encompasses aspects such as nationality, culture, language and religion (23) . There are known structural barriers that Black, Asian and minority ethnic (BAME) groups face when accessing healthcare or being involved in clinical research (24)(25)(26) . Lack of exposure to health promotion messages and economic disadvantage are factors contributing to health inequalities. Considering these inter-related complexities, it has been recognised that interventions that are developed with the community of interest and which are culturally sensitive and tailored may be more effective (24) . Our teams have a record of working with at-risk communities. With time, we have become increasingly aware of the need to understand how behavioural interventions may help at-risk communities to comply with public health advice, including that for vitamin D supplementation. Acknowledging this background, we undertook a systematic review to evaluate behavioural interventions implemented to improve or optimise vitamin D stores in children from at-risk ethnic groups.

Methods
Our systematic review, performed based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (27) , is registered within the PROSPERO database, CRD42017080932.
Identified interventions were planned to be evaluated using the Health Behaviour Model (HBM) (28) . The HBM (28) attempts to explain and predict health-related behaviours across four constructs: (i) perceived susceptibility (i.e. affect with known risk of deficiency); (ii) perceived severity (i.e. affect with known consequences/health outcomes of deficiency); (iii) perceived benefits (defined here as promoters for optimal vitamin D status, i.e. sufficient sun exposure, supplementation, health information access); and (iv) perceived barriers (i.e. sufficient sun exposure, supplementation, health information access) (29) .

Search question
The PICO acronym, an established model for aiding systematic reviews, guided all elements of our research question:

Data sources
The following databases were searched and verified: Cochrane Library, MEDLINE, EMBASE, CINAHL with limited secondary search on Google Scholar, in order to identify relevant studies. No country limits set. Only articles written in English published after 1990 were included. Searches were conducted in February 2018, with a limited updated search in November 2020.

Search strategy
The search strategy included terms for 'vitamin D' and 'ethnic' and terms specifying all major subgroups. Details are as follows: Term 1: Vitamin D Vitamin D * OR ricket* OR osteomalacia Term 2: Ethnic Ethnic group* OR Asia* OR Africa* OR emigrant* OR immigrants Search string: (Vitamin D * OR ricket* OR osteomalacia) AND (ethnic group* OR Africa* OR Asia* OR emigrant* OR immigrant*) A selective search strategy focusing on identifying behavioural interventions was used for updated review of abstracts February 2018-November 2020 (Additional File 1).

Eligibility criteria
We classified Asian ethnic groups as individuals of central, east, south, south-east, and western Asian origin and African as individuals of east, north, south, and western and south of the Sahara origin. Studies were included if they met the PICO inclusion criteria above, were published in English or with translation available, and were randomised controlled trials (RCT), quasi-RCT and non-RCT (beforeafter studies).
We excluded purely observational studies, non-English language, or no English version available, full-text not available, or studies classified as dissertations/abstracts/ conference pieces/editorial letter or review. We also excluded studies with no or non-extractable ethnic/demographic or vitamin D data and those focusing on adult populations.

Study selection and data extraction
Two reviewers (EA and CL) shared screening of titles and abstracts. Shortlisted abstracts underwent full-text review by two researchers (PA and AT), with conflicts resolved by a third reviewer (CL). For each study of interventions, data were extracted and classified (AT and PA) and checked (EA and NT) for publication year, characteristics of study population (sample size, mean age, ethnicity, gender and age), study design, available vitamin D data, supplementation data and intervention measured. For the updated search in November 2020, in line with our primary aim, we searched only for studies with an explicitly defined behavioural intervention. Titles and abstracts were screened by two reviewers (RP and NT) and confirmed with third reviewer (EA).

Risk of bias assessment
Methodological quality of the studies was assessed using the Cochrane Risk of Bias-2 (31) assessment tool for RCT (nineteen studies), the Cochrane ROBINS-I (32) tool for non-randomised studies of interventions (one study) and the NHLBI (33) tools for Observational Cohort and Cross-Sectional Studies (one study) or Before-After (Pre-Post) Studies with No Control Group (three studies). The studies were quality-assessed by EA, with ratings reviewed by NT, points of uncertainty discussed and final ratings agreed with NT (n 5; 22 %). Studies with a high or critical risk of bias were included but with quality ratings highlighted within the results section in order to contextualise findings.

Results synthesis
In view of the heterogeneity of studies identified in terms of methods, participants, vitamin D thresholds used, the interventions and outcomes, a narrative approach to synthesis was used following guidance developed by the University of York Centre for Reviews and Dissemination and the Economic and Social Research Council (34,35) .

Study selection
Initially, 10 690 articles were identified. Title and abstract screening returned 298 potential articles. After full-text review, 274 articles were excluded, and 24 intervention studies (including 2 with behavioural components) were included. No further behavioural studies were identified from the updated literature search in November 2020 (see Fig. 1).

Study characteristics
The majority (11) of studies included were conducted in the USA, with two further studies each in Australia, Canada, and Mongolia. One study was undertaken in each of the following: UK, Norway, Denmark, Pakistan, India, Nigeria, and Turkey. Table 1 summarises these papers.
With reference to quality appraisal, nineteen studies were appraised using Cochrane Risk of Bias for RCT. Of these, one (36) had a low risk of bias and the remaining eighteen had some concerns. One study (37) , appraised with the ROBINS-I tool, was coded as having a critical risk of bias due to confounding. Of the NHLBI appraised studies, two were rated as fair (38,39) and two as good (40,41) (see  Tables 2-5).

Behavioural interventions
All studies included conducted a vitamin D pharmacological supplementation intervention in our target population, with two that incorporated a behavioural component to their intervention strategy (38,42) in addition to the pharmacological intervention. We found no study that explicitly defined a primary aim of evaluating a behavioural intervention undertaken with the intention to study its effect on vitamin D supplement uptake.
Madar et al. (42) ( Table 2; overall risk of bias: some concerns) studied the effect of vitamin D 2 drops on serum 25-hydroxy-vitamin D (25(OH)D) in infants with immigrant origin within a cluster RCT. In total, sixty-six healthy infants of Pakistani (South Asian origin), Turkish (West Asian/ Middle Eastern) or Somali (East African) origin were recruited for the study from eight child health clinics in Oslo, Norway. The behavioural component involved multilingual translated brochures for mothers, with information provided to each ethnic group on vitamin D, its sources and instructions on how to administer the vitamin D drops, made available free of charge. Aims were to evaluate whether a free supply of a 400 iu daily dose, for 6-week old infants, together with information handouts that had been translated into Urdu, Turkish or Somali languages incorporating text and simple illustrations, improved vitamin D status, assessed at 7-week follow-up in the intervention and control group. Fifty-one (78 %) infants completed the study, with serum 25(OH)D levels significantly higher in the intervention group v. Control group (93·5 v. 72·7 nmol/l; P = 0·03). Amongst exclusively breast-fed infants at baseline, serum 25(OH)D levels increased by 32·3 nmol (P = 0·035) in the intervention group. This study concluded (42) that free supply of vitamin D drops, with translated information handouts, significantly improved the vitamin D status of healthy infants of immigrant background. Considering the Behaviour Change Wheel framework, this combined intervention strategy (if part of an explicitly stated behavioural intervention) could have been coded under criteria; education and training (translated information leaflets; administration advice) and enablement (free supply) (30) .
The second study with behavioural components centred around a nationwide prevention campaign instigated due to resurgence of vitamin D deficiency rickets in children of ethnic minority origin living in Turkey, a sunny country in West Asia/Middle East. In order to gauge the campaign's impact which included supply of vitamin D supplements as an intervention, Mutlu et al. (38) (Table 4;    Primary outcome maternal/neonatal circulating 25(OH)D at delivery; secondary outcomes 25(OH)D > or = 80 nmol/l achieved?; 25(OH)D concentration required to achieve maximal 1,25(OH)D production. Deficiency defined as total circulating 25(OH)D < 50 nmol/l (20 ng/ ml), insufficiency as ≥50 to <80 nmol/l (≥20 to <32 ng/ml), and sufficiency as ≥80 nmol/l (≥32 ng/ml).
Mean vitamin 25(OH)D levels by group at delivery and 1 month before delivery was significantly different (P < 0·0001). 4000 iu daily dose in pregnancy is safe and most effective in achieving sufficiency in all women regardless of race (P < 0·0001).
Hollis et al.
Were eligibility/selection criteria for the study population pre-specified and clearly described?
Were the participants in the study representative of those who would be eligible for the test/service/intervention in the general or clinical population of interest?
Were all eligible participants that met the pre-specified entry criteria enrolled?
Was the sample size sufficiently large to provide confidence in the findings?
Was the test/ service/intervention clearly described and delivered consistently across the study population?
Were the outcome measures prespecified, clearly defined, valid, reliable, and assessed consistently across all study participants?
Were the people assessing the outcomes blinded to the participants' exposures/ interventions?
Was the loss to follow-up after baseline 20% or less? Were those lost to follow-up accounted for in the analysis?
Did the statistical methods examine changes in outcome measures from before to after the intervention? Were statistical tests done that provided P-values for the pre-topost-changes?
Were outcome measures of interest taken multiple times before the intervention and multiple times after the intervention (i.e. did they use an interrupted time series design)?
If the intervention was conducted at a group level (e.g. a whole hospital, a community, etc.), did the statistical analysis take into account the use of individual-level data to determine effects at the group level?  (30) . The authors (38) noted that the major obstacles for use of vitamin D supplements in Turkey included limited public awareness, access to healthcare and supplement costs.

Pharmacological interventions
All studies included in this review incorporated some form of pharmacological intervention, the results of which are summarised in the following sections. Ten studies (7,37,38,(42)(43)(44)(45)(46)(47)(48) were undertaken during the pregnancy/newborn phase with ethnic minority origin as a main risk factor. One study specifically targeted a cohort with known 'at-risk status', that is, nutritional rickets (44) .  (46) also undertaken in Canada evaluated newborn infants receiving formula feeds during intensive care hospitalisation. Classified as White or non-White African American origin, the study arms included supplementation with 400 iu vitamin D or matching placebo with formula feeds. White infants achieved significantly higher mean 25(OH)D levels by time of discharge (P = 0·0003). Hollis et al. reported findings from two studies (43,48) . The 2011 US cohort (48) were women with singleton pregnancies, given varying doses of vitamin D (400/2000/4000 iu daily) from 12-16 weeks of gestation to delivery. The 4000 iu daily dose was safe and the most effective in achieving sufficiency in all women regardless of race (Caucasian; African American; Hispanic; P < 0·0001). The 2015 US study (43) included 334 mother-infant pairs with newborns of 4-6 weeks old, recruited as mothers planned to breast-feed for 6 months. Compared to a 400 iu daily dose, the higher 6400 iu daily dose showed a significant increase in maternal vitamin D levels (P < 0·0001) and was safe. Vitamin D deficiency in breastfed infants was affected by race, with African American mothers and infants having substantially lower circulating 25(OH) D levels.

Pregnancy/newborn group
Rodda et al.'s Australian study (45) recruited 78 ethnic minority pregnant women and newborns, identified on the basis of dark skin or veiling (84 % treated and 97 % control group). They were supplemented with 2000-4000 iu daily from 12 to 16 weeks' gestation to delivery. Umbilical cord serum 25(OH)D levels at delivery were higher in neonates of mothers in the treated group (81 nmol/l v. 42 nmol/l; P < 0·0001).
Yu et al.'s prospective UK study (7) recruited 180 women, of Indian Asian, Middle Eastern, Black and Caucasian origin, treated from 27 weeks' gestation to delivery. Study arms compared a single 200 000 iu dose injection, an 800 iu daily dose and no treatment. Treated mothers had significantly improved vitamin D levels (42 v. 27 nmol/l; P < 0·0001), but only a small percentage (women 30 %; babies 8 %) given supplements achieved sufficiency levels (50 nmol/l or over).
The intervention in Hossain et al.'s single-centre openlabel RCT (47) in Pakistan (South Asia) was 4000 iu vitamin D taken daily from gestational week 20 to delivery. Positive correlation was found between maternal and neonatal 25(OH)D levels (r = 0·83; P = 0·001), with improved status with supplementation (maternal 25(OH)D increased from mean 8·82 ng/dl to 18·3 ng/dl; P = 0·001). Mondal et al. (44) studied seventy-one Indian children (South Asia) with nutritional rickets, aged between 6 months and 5 years, given either a single intramuscular dose of 600 000 iu vitamin D or 10 weeks of a 60 000 iu dose taken weekly. Both were safe and effective with no difference found in efficacy on comparing diagnostic parameters for rickets (P > 0·05).
The Madar et al. (42) and Mutlu et al. (38) studies discussed above included a behavioural component in addition to the pharmacological intervention.
Lewis et al. (54) assessed five daily oral vitamin D doses (0,400,1000,2000 and 4000 iu) in Black and White children aged 9-13 years. Increases in vitamin 25(OH)D levels depended on dose taken; 12-week changes ranged between −10 and 76 nmol/l (placebo v. 4000 iu, respectively). In the highest dose group, larger vitamin D gains were seen in White compared with Black children (P < 0·01) Rajakumar et al. reported vitamin D insufficiency at baseline (41) in about half of their study group (6-10 years old, African American children). Using the 2005 definition for adults, deficiency status was defined as less than 10 ng/ml and insufficiency as between 10 and 20 ng/ml (41) . Supplementation with 400 iu daily for 4 weeks helped achieve a significant increase in levels (see Table 1), but insufficiency status still persisted in 18 % of the experimental group (41) . In a separate study, Rajakumar et al. also noted (40) that vitamin D deficiency is common amongst obese and non-obese preadolescent African American children, finding that 400 iu daily dose was not enough to raise vitamin D serum levels to 30 ng/ml (Table 1; deficiency defined as less than 20 ng/ml; insufficiency 21-29 ng/ml). In further work (49,52) , Rajakumar et al. studied African American and White children aged 8-14 years supplemented with vitamin D 3 1000 iu/d for 6 months. They reported that vitamin D 3 supplementation varied with race and was more effective and significant in Black children (49) (Table 1). In addition, the intake required to maintain concentrations at 20 ng/ml in 97·5 % of Black and White children, adjusting for race and pubertal status, was three times higher than the US recommended allowance of 600 iu/d (52) . Sachek et al.'s (53) randomised control trial recruited 604 children (Black, Hispanic, White and Asian), aged 8-15 years. They found that serum vitamin D levels increased over 6 months with all dose ranges (600, 1000 or 2000 iu daily) with the 2000 iu daily dose achieving a higher vitamin D concentration compared with the two lower doses (33·1 v. 26·3 and 27·5 ng/ml; P < 0·001). Talib et al. (55) found that adolescents (Hispanic, Black, White and Asian; 13-20 years old) required 8 weeks of high-dose colecalciferol, of at least 5000 iu/d in order to correct deficiency status, noting also that obese adolescents had poorer response to treatment (13·7 þ/− 10·7 v. 21·9 þ/− 16·9 ng/ml; P < 0·001).
Abrams et al. (36) found no significant effect of an increase in vitamin D3, with decrease in parathyroid hormone levels, on calcium absorption, which was one of their study's primary outcomes of interest. The pharmacological intervention used was supplementation with 1000 iu daily vitamin D for 8 weeks in African American, Hispanic, Asian, and White children aged between 4 and 9 years. Dong et al. (50) studied supplementation in forty-nine healthy African American children, aged 14-18 years, with 2000 iu/d v. 400 iu/d in the control group. They reported the 2000 iu daily dose as probably more effective (see Table 1) in optimising vitamin D status and counteracting progression of aortic stiffness in Black youth.
Doherty et al.'s (51) target group were African American children and young adults (5-20 years old) with or without sickle cell disease (SCD-SS). The study aim was to assess safety and efficacy of two oral vitamin D 3 daily doses (4000 iu and 7000 iu). They noted that with 12 weeks of supplementation, both doses were safe and well tolerated but deficiency status (<20 ng/ml) was eliminated only in those receiving 7000 iu/d (P < 0·05).
Andersen et al. (56) recruited adolescent girls of Pakistani (South Asia) origin living in Denmark, aged 10-15 years, to study two vitamin D doses (10, 20 mcg/d). It is difficult to comment on any significant difference in 25(OH)D levels in the group after 1 year, as numbers recruited were small (n 21). Ganmaa et al. (39,57) found that drinking ultra-hightemperature (UHT) processed milk fortified with about 100 iu vitamin D per serving improved levels in Mongolian (East Asia) children aged 9-11 years, and that correction of baseline deficiency in Mongolian children aged 12-15 years with 800 iu daily over 6 months increased growth (57) . Thatcher et al. (58) studied Nigerian (West Africa) children aged between 1 and 14 years with nutritional rickets, finding that calcium supplementation with or without vitamin D was more effective then vitamin D alone in healing active rickets (P < 0·001). (Tables 1, 6 and 7).

Research published after completion of the initial review
The updated literature review on MEDLINE, EMBASE, CINAHL and Cochrane Library databases, including secondary search of Google Scholar, undertaken for February 2018-November 2020, focused purely on identifying behavioural intervention studies. No publication with an explicitly defined behavioural intervention implemented to study vitamin D supplementation outcomes was identified (see Fig. 1). Protocols for two new systematic reviews have been published (66,67) , but there is no indication that any behavioural interventions per se will be considered.

Discussion
This systematic review included twenty-four publications for final evaluation. We had planned to analyse behavioural interventions using HBM constructs (28) and the Behaviour Change Wheel framework criteria (30) . We have utilised HBM constructs for assessment in a previous patient and public involvement study (29) .
Despite our a priori aim, we found no studies that explicitly defined and studied the effects of a behavioural intervention. Therefore, we completed a narrative evaluation of the two studies that undertook a pharmacological intervention and incorporated a behavioural component, which included translated information brochures, some support from health professionals and free supply of vitamin D supplements (38,42) . These studies did not evaluate which individual component helped improve vitamin D status in the infants, for example, whether it was the translated information brochure provided to the mothers or the free supplement supply. The combined strategies broadly meet criteria of education, training and enablement from the total of nine that underpin the Behaviour Change Wheel model (30) .  (46) Newborn 400 iu daily in addition to formula feed Positive White infants had statistically significant higher mean 25(OH)D levels at discharge then non-White infants.  (43) Newborn 4-6 weeks with breast-feeding planned for 6 months 6400 iu vitamin D/d v. 400 iu/d Positive 6400 iu vitamin D dose daily was safe. Vitamin D deficiency in breast-fed infants affected by race. Hossain et al. 2014 (47) Pregnant women and newborns 4000 iu/d from week 20 to delivery Positive There was a positive correlation between maternal and neonatal 25(OH)D levels. Lewis et al. (2013) (54) 9-13 5 oral doses tested over 12 weeks; 0, 400 iu, 1000 iu, 2000 iu, 4000 iu.

Positive
Larger gains in vitamin D levels in White v. Black population at the highest dose Madar et al. (2009) (42) 7 weeks old 400 iu vitamin D 2 daily over 7 weeks Positive Translated information brochures and free supply of vitamin D supplement. Mondal et al. (2014) (44) 0·5-5 Single i/m dose of 600 000 iu or 60 000 iu taken weekly for 10 weeks.

Positive
No difference found in efficacy of both regimens when comparing diagnostic parameters of rickets at 12 weeks. Mutlu 2011 (38) Infants ( Behavioural interventions for Vitamin D In general, there was a great deal of variability in the approaches of the studies. Several studies were undertaken in healthy populations (36,42,50,54) , including populations where the aim was to evaluate a large-scale national prevention campaign (38) or study the effect of fortification (39,57) . Other studies were undertaken in patients with medical conditions that precluded achievement of vitamin D sufficiency status (Table 7). Examples include a sickle cell anaemia -SS patient cohort (51) , infants in intensive care (46) , treatment for rickets (44,58) or studying treatment in obesity (40,53) . Variation in study outcomes is to be expected due to heterogeneity within the studies, which used different thresholds for insufficiency and deficiency status when measuring serum vitamin D levels (7,36,40,41,50,57) . Hanson et al. (46) , Rodda et al. (45) and Yu et al. (7) undertook cord blood sampling, whereas Hollis et al. (43) evaluated maternal and infant urine and blood samples. The majority of the studies aimed to assess whether a vitamin D 400 iu daily dose was sufficient, using various surrogate endpoints to rationalise and justify their conclusions, with study of higher doses in studies that included patients with research groupdefined deficiency/insufficiency status (Tables 1 and 7).
Cochrane reviews have shown that supplements taken during pregnancy could reduce the risks for pre-eclampsia, gestational diabetes and low birth weight (9,59) . The 2018 Food Standards Agency's commissioned research on use of food supplements (60) found that consumers with higher levels of education or currently working were more likely to take supplements. The most common reason cited for taking food supplements was an aspiration for a healthy lifestyle. However, consumers who may be at risk may not be adequately informed about food supplements, to include vitamin D. Michie and Tayarachakul (61) suggest that any 'information and advisory' health professional role for pregnant women should include the recommendation of 400 iu vitamin D daily; with communication, especially with vulnerable mothers, and these should include mothers of darker skinned ethnic populations, undertaken and supported by midwives or community pharmacists (61) . High-risk groups may need individualised advice and information on the need for vitamin D supplements. In addition, it may be important to address perceived medicalisation of vitamin D supplements, whereby women feel they only need to take a supplement on the recommendation or prescription from their GP (29) . Currently in the UK, patients and the public are advised to purchase vitamin D, unless eligible for the Healthy Start Scheme (19) . Even if prescribed per se, there are important issues that need to be addressed. Wan et al. (62) demonstrated an increase in prescribing of vitamin D in the UK, but their 2008-2016 database study found inconsistency between supply regimens prescribed, an absence of pre-supplementation (range 29-56 % annually), and a trend for increased prescribing of higher pharmacological treatment doses rather than maintenance doses. Global and national data (1)(2)(3)(4)(5)(6)(7)(8)(9)11,12,(16)(17)(18)20,(59)(60)(61)63) show there are still ongoing issues relating to vitamin D deficiency status and supplementation. More importantly, studies have not adequately addressed the best approach to improve uptake in high-risk groups. This area requires further work to identify the most effective behavioural interventions, which should ideally be co-developed working with the high-risk community itself. We used ethnicity as an inclusion criterion in order to consider at-risk groups of Asian or African origin. Both culture and religion can impact on uptake of supplements, but no study in this review discussed this complex relationship and effect on uptake, compliance or adherence (64,65) with vitamin D supplementation. We noted variations in study design to support improved uptake of pharmacological treatment and these observations could be used as supporting information for future behavioural interventions. These modifications included use of chewable tablets (53) , liquid (37,51) or flavoured formulations (36,53) , drops for infants (37,38,42,51) and halalcertified products (45,56) . Some of these modifications would be applicable in the case of all children, whereas some are important when considering ethnicity and culturally acceptable treatment.
The Palacios et al. (9) updated Cochrane 'interventions' systematic review aimed to evaluate the effect of vitamin D supplementation. With a focus on pregnancy only and not on at-risk ethnic minority groups per se, they assessed the evidence base for three distinct patient cohorts; vitamin D only (22 trials; 3725 pregnant women), calcium and vitamin D supplementation (9 trials; 1916 pregnant women) and vitamin D and other micronutrients (1 study; 1300 pregnant women). The studies considered pharmacological interventions evaluating risk or harm; no behavioural interventions were described. As in our review, they observed that supplementation increased serum 25(OH) D concentrations during pregnancy but with large heterogeneity in the results, possibly related to differences in vitamin D doses and methods used to assess outcomes. Another Cochrane review (59) considered variations in vitamin D supplement regimens during pregnancy, focusing on pregnancy and neonatal outcomes, but did not specifically consider at-risk ethnic minority groups and no behavioural interventions were reported.
Although we identified few publications in this systematic review, there is work that shows that milk fortification (35,53) appears to be acceptable to mothers, and chewable formulations (53) in addition to drops or liquid products (34,35,39,47) appear tolerable in children. This could have important implications for public policy on mass fortification. Variations in the available forms of supplements should be highlighted, especially to those groups less likely to present to health services.
Structural barriers will be important to overcome to ensure messaging reaches at-risk groups, including health literacy and low socio-economic status, especially where the targeted community comprises first-generation immigrants or residents. Other facets that deserve due consideration include individual communication preferences, comprehension and concerns, and research teams will need to seriously consider these with use of co-production methodologies. The NIHR INVOLVE guidance (24) defines the approach used by researchers, practitioners and the public working together on a co-produced research project as one of joint ownership, with sharing of power and responsibility from the start to the end of the project. Working together helps promote understanding by including all perspectives and skills. With close community engagement, there is an expectation that implementation of the co-designed intervention will result in better outcomes.
The two behavioural type strategies that we identified used elements of education, training and enablement as possible supportive behaviour change techniques, with the associated pharmacological intervention demonstrating achievement of significantly higher vitamin D levels. As these were mainly medical intervention trials, it is to be expected that the focus would be on participant adherence rather than on changing behaviour related to ongoing supplementation. However, bearing in mind racial/ethnic disparities, using a behavioural strategy may be important. We accepted that the search term 'interventions' would identify both pharmacological-and behavioural-type interventions, and we evaluated on the premise that there can be learning from any behavioural component incorporated within pharmacological trials as well. Future research groups could consider incorporating explicitly defined behavioural interventions, underpinned by HBM constructs and Behaviour Change Wheel categories, as a formal aspect of their intervention study. This may help address structural barriers and could be implemented, for example, during routine antenatal visits, for widespread population benefits. It is important to note the results of a 2020 study of 125 UK children with nutritional rickets, which found that over three-quarters (77·6 %) of these children were not taking vitamin D supplements (2) . Accepting that there is still a need for more data on children who did not have nutritional rickets, it is evident that current recommendations from Department of Health (1,3,19) are not reaching high-risk groups. Our aim to study behavioural interventions aiming to improve both initiation and maintenance of supplementation in pregnancy and beyond has been informed by this observation. The findings of our study, namely a lack of research into behavioural interventions and the forms of intervention that would be most effective in engaging at-risk groups, lead us to recommend this as a research priority.

Limitations of this review
Mutlu et al.'s study (38) describes a national campaign implemented in Turkey, aiming to encourage the entire population, including pregnant and nursing women and infants, to have adequate sunlight exposure. Although the primary source of vitamin D is sun exposure, the ability of the body to create and maintain sufficient levels is affected by many variables. These include socio-economic/socio-demographic status, geographical/environmental, and cultural and religious, lifestyle and dietary as well as genetic factors. Knoss et al. (68) studied people of different ethnic origins with similar practices, noting that skin pigmentation was a significant risk factor for vitamin D deficiency, irrespective of ultra-violet light exposure. A Swedish, primarily a food-based intervention study (69) , did not report ethnic origin but recruited children living in the northern and southern latitudes of the country. With children stratified by skin colour using Fitzpatrick's definition, the results showed that those with darker skin required higher vitamin D intake or supplementation. We focused on ethnicity to help identify behavioural interventions implemented, if any, but note that there may be other relevant work on sun exposure-related behavioural interventions in children. We used ethnicity as a primary search term to help identify high-risk groups, but studies reported their selected populations based on ethnic origin or alternatively used race for classification. This adds to the heterogeneity of data available. No studies considered inter-related complexities including cultural or religious beliefs, distance from country of origin, or first-or second-generation immigrant status, and this needs due acknowledgement. The primary focus for all studies was outcome of a pharmacological intervention, rather than behavioural interventions, with different thresholds used to define vitamin D deficiency and extrapolation of findings should be undertaken cautiously. We only included studies published in English, and studies where full text was available, meaning some relevant studies may have been omitted.

Conclusions
In summary, all the studies included in our systematic review, which focused on high-risk ethnic groups aged under 18 years, evaluated the effect of a pharmacological vitamin D intervention. There were no studies identified that included direct evaluation of an explicitly defined behavioural intervention. There is a need for additional rigorous high-quality and larger RCT to evaluate the effects of vitamin D supplementation in pregnancy. Equally, there is a need for research into clearly defined behavioural interventions targeted to individual ethnic groups, ideally designed with use of co-production methodologies, in order to help our understanding of how behaviour change can affect vitamin D use in antenatal care, pregnancy and childhood.