Description of the condition
Osteoporosis, fracture and peak bone mass
Osteoporosis is a systemic skeletal disorder characterised by low bone mass and microarchitectural deterioration of bone tissue, with a consequent increase in bone fragility and susceptibility to fracture (NIH 1993). In other words, there is both less bone and poor bone quality. The clinically important consequence of osteoporosis is fracture. Osteoporotic fractures in the elderly are both common (Johnell 2006) and costly (Burge 2007). They also have significant implications for individuals who sustain a fracture as death, loss of ability to live independently and long-term restrictions on usual activities, including self care, are all common outcomes (Pasco 2005). Adult bone mineral density (BMD) is determined by both peak bone mass (the maximum bone mass attained in a person's life) and the rate of bone loss after peak bone mass is achieved (Hansen 1991; Riis 1996). Suboptimal bone growth in childhood and adolescence appears at least as important as later bone loss in the development of osteoporosis (NIH 2001). It is estimated that a 10% improvement in peak bone mass could delay the onset of osteoporosis by 13 years (Hernandez 2003). Such an increase is equivalent to about one standard deviation higher bone density at the lumbar spine from the age of 60. With a one standard deviation decrease in lumbar spine BMD there is a 60% increase in hip fracture risk (Cummings 1993), so a BMD increase of this magnitude could translate to an estimated 50% reduction in the relative risk of hip fracture. Thus the potential long-term benefits of intervening to improve peak bone mass in children are very substantial. Benefits may also not only be restricted to later life. Childhood fractures are common, occurring at a comparable rate to fractures in older adults (Jones 2002) and, as in adults (Marshall 1996; Nguyen 1993), bone density is a risk factor for fracture in children (Clark 2006; Goulding 1998; Goulding 2001). As a result, improvements in bone acquisition in childhood are likely to produce benefits in terms of reduced childhood fractures, though this has not yet been investigated in prospective studies with fracture outcomes (Winzenberg 2013).
One potential intervention to improve bone acquisition is vitamin D.
Description of the intervention
Vitamin D is frequently used as a generic term to describe a number of specific molecules, including vitamin D
How the intervention might work
Vitamin D is essential for bone health. Its impact on bone health in adults is well accepted (ANZBMS 2005; IOM 2010). Overt vitamin D deficiency in children leads to rickets and there is increasing evidence that sub-clinical vitamin D deficiency may also affect bone mineralisation (Cheng 2003; Jones 1998; Jones 2005; Outila 2001; Zamora 1999). In children and adolescents, the available data suggest that in vitamin D deficient children at least, clinically important improvements in bone density outcomes may be achievable through the use of vitamin D supplements, though this has not yet been definitively proven (Winzenberg 2010; Winzenberg 2011a).
Generally, breast-fed infants have lower bone acquisition compared to formula-fed infants and it is thought that this is at least partly due to the low vitamin D content of breast milk compared to most formulae (Specker 2004), resulting in infants having a lower vitamin D intake and lower serum 25(OH)D levels. This bone deficit may be temporary. One three-arm randomised controlled trial (RCT) compared the effects of two different formulae, one with moderate and one with low mineral content, and breast-feeding in infants on total body bone mineral content (TBBMC). After six months, TBBMC was higher in the moderate-mineral content formula group compared to either the low-mineral content formula or breast-feeding. After six months, the infants were re-randomised to moderate or low-mineral content formula or cows' milk. TBBMC at 12 months was similar across these three groups. In the group who breast-fed for the first six months, TBBMC increases were higher than in the other groups in this second six-month period (Specker 1997). Nonetheless, because of the potential contribution of the low vitamin D content of breast milk to at least short-term deficits in bone mineralisation, it is possible that bone development in breastfeeding infants could be augmented through the use of vitamin D supplements (Winzenberg 2012).
Why it is important to do this review
This review is important because current evidence on whether vitamin D supplements are beneficial for bone density in infants is conflicting. Observational data assessing the associations between infants' vitamin D status and bone density outcomes are limited. Studies are relatively small, vary in their design and their results are inconsistent (Winzenberg 2012). In 76 term Caucasian infants distal radius BMC at 2 and 16 weeks of age was similar in infants consuming human milk alone, human milk with supplemental vitamin D or Similac, even though infants consuming human milk alone had lower serum 25(OH)D levels at 16 weeks (mean serum 25(OH)D 42, 55 and 62 nmol/l in the human milk, human with vitamin D and formula-fed groups respectively) (Roberts 1981). This null finding may have been partly due to the breast-fed infants in the study only being mildly vitamin D deficient. In a retrospective cohort study of 106 healthy prepubertal Caucasian girls who were breast-fed, vitamin D supplement use in the first six months of life was associated with higher areal bone mineral density (BMD) at the radial metaphysis, femoral neck and femoral trochanter, but not the lumbar spine or radial or femoral diaphysis when the girls were aged eight years (Zamora 1999). Supplement use was common in this study, with 86% of children receiving 400 IU/day of vitamin D3 for a median of 12 months. In 35 Korean infants aged two to five months who were born in the winter, breast-fed infants had lower serum 25(OH)D but lumbar spine BMC was similar in breast-fed and formula-fed infants (Park 1998). The correlation of 0.17 between lumbar spine BMC and serum 25(OH)D levels was not statistically significant though correlations of a similar magnitude have been reported in older children (Winzenberg 2012) and the lack of a statistically significant effect may be due to inadequate sample size. Lastly, in pre-term (n = 44, mean postnatal age 43 weeks) and full-term (n = 82, mean postnatal age 36 weeks) infants, serum 25(OH)D was not associated with lumbar spine BMD or BMC (Bougle 1998). This study did not report whether or not participants were breast-fed. There is clearly potential for significant confounding and other biases in such observational studies. However, RCT data are also limited and conflicting, variously reporting benefits at 12 weeks (Greer 1981) but not 12 months (Greer 1982), possible detrimental effects (Greer 1989) and no effect (Chan 1982) of vitamin D supplements in term breast-fed infants.
The systematic review in children and adolescents described above (Winzenberg 2010; Winzenberg 2011a) excluded studies in neonates and did not identify any studies in infants (< 12 months of age). A systematic review of RCTs in infants (including neonates) with meta-analysis is therefore important to endeavour to provide the best evidence by which determine whether vitamin D supplementation of infants will in fact result in benefits for bone acquisition which are of clinical or public health significance, or both.
- To determine the effectiveness of vitamin D supplementation of infants for improving bone mineral density in infancy, childhood or adulthood.
- To determine if any effect of vitamin D supplementation varies by baseline vitamin D status, sex, or the type or dose of vitamin D given.
Criteria for considering studies for this review
Types of studies
We will include randomised or quasi-randomised controlled trials of vitamin D supplementation compared with placebo or no intervention control, with a treatment period of at least three months.
Types of participants
We will include trials in healthy, full-term infants. If it is not specified whether infants in the study are full-term or not we will include the study, but perform a sensitivity analysis omitting those studies. If studies are performed in a mixture of full- and pre-term infants, we will include the study if data on full-term infants alone can be extracted or if > 80% of infants are full-term.
We will only include full-term infants as the impact of different feeding methods and health issues related to pre-term status on growth and bone development mean that the potential effects and role of vitamin D supplementation in this group are too different to be sensibly assessed in the same review with full-term infants.
Types of interventions
Vitamin D supplementation commencing in the neonatal period, regardless of type or dose of vitamin D supplement or method of administration, with a minimum treatment period of at least three months, compared to placebo or no intervention control. There is no upper limit to the duration of supplementation. Studies with co-interventions which are nutritional in nature (such as dairy products or calcium supplements) will be included provided co-interventions contain less than 100 IU vitamin D per day.
Types of outcome measures
While fractures in later life would be the ideal outcome measure, for intervention studies in children this would require following large numbers of subjects for decades. These studies have not been performed and to our knowledge there are no studies with childhood fracture outcomes (Winzenberg 2012). Therefore, in this review we will use bone density measures as surrogate outcomes, as is commonly seen in intervention studies in children (Gilsanz 1998) and as we have previously done in other systematic reviews in children (Winzenberg 2010; Winzenberg 2011a; Winzenberg 2006; Winzenberg 2006b).
Primary outcome measures will be: areal or volumetric bone mineral density or bone mineral content at the femoral neck, total hip, total body, lumbar spine, proximal and distal forearm, where possible taken as per cent change from baseline. Time points will be:
- the longest point after supplementation has ceased;
- at the end of supplementation;
- at the time point which is common to the largest number of studies (if there is substantial variation in the timing of measures).
Methods of measurement may include dual x-ray absorptiometry (DXA), single photon absorptiometry, dual photon absorptiometry and peripheral quantitative computerised tomography. Studies will only be included if they measure a primary outcome.
Total adverse effects and withdrawals from studies due to adverse events will also be primary outcomes, ascertained at (1) the end of supplementation and (2) at the longest point after supplementation has ceased.
We will also extract data at interim time points (i.e. time points between baseline and the end of the study) where available and analyse these as secondary outcomes if sufficient data are available.
Search methods for identification of studies
We will search the Cochrane Central Register of Controlled Trials (CENTRAL) (via The Cochrane Library), MEDLINE (via OVID 1946 to present), EMBASE (via OVID 1947 to present), CINAHL (via EbscoHost 1966 to present), AMED (via OVID 1966 to present) and ISI Web of Science (via Web of Knowledge). There will be no language restrictions. The search strategy is given for MEDLINE (via OVID) in Appendix 1 and uses the Cochrane Highly Sensitive Search Strategy for identifying randomised trials in MEDLINE: sensitivity-maximising version (2008 revision); Ovid format (Lefebvre 2011). We will adapt this as appropriate for the other databases.
Searching other resources
We will also handsearch conference abstract issues of key journals (Osteoporosis International, Journal of Bone and Mineral Research, Journal of the American Dietetic Association, Proceedings of the Nutrition Society, Journal of Nutrition) for one year prior to the date of our updated electronic search to identify recent trials that have not yet been published in full. We will examine the reference lists and ISI citations of all included studies. We will also search trial registries (the World Health Organization (WHO) international registry platform (http://www.who.int/ictrp/en/) and clinicaltrials.gov for ongoing trials.
Data collection and analysis
Selection of studies
Two review authors will independently assess all potentially relevant articles against the study inclusion/exclusion criteria. Disagreements will be resolved by consensus where possible, but if necessary a third review author will be consulted to help resolve differences.
Data extraction and management
Two review authors will also independently extract data on primary and secondary outcomes as described above (Primary outcomes; Secondary outcomes) as well as study and population characteristics. Disagreements will be resolved by consensus where possible, but if necessary a third review author will be consulted to help resolve differences. We will extract the following study and population characteristics where possible: study setting, sex, age, baseline serum vitamin D and vitamin D assay used, mode of infant feeding, baseline length, baseline weight, vitamin D intake, levels of sun exposure, ethnicity, type and dose of vitamin D given, duration of supplementation, duration of follow-up and levels of compliance. We will also extract data relevant to the assessment of risk of bias (Assessment of risk of bias in included studies).
Assessment of risk of bias in included studies
Two review authors will independently assess each trial for risk of bias, addressing randomisation, allocation concealment, blinding of those providing treatment and of participants, completeness of outcome assessment, selective reporting and other potential sources of bias as per the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). Disagreements in risk of bias assessments will be resolved by consensus where possible, but if necessary a third review author will be consulted to help resolve differences.
Measures of treatment effect
We will convert continuous primary outcome measures (i.e. bone density outcomes) to standardised mean differences (SMD) unless measurement scales are identical in which case we will use mean differences (MD). For safety outcomes, we will determine a risk ratio (RR) for the number of children with adverse events. Where possible, we will base the analyses on intention-to-treat data from the individual clinical trials, but if these are not available, in order of preference, we will use data from available data or per protocol analysis.
Unit of analysis issues
Should we identify cluster-RCTs, we will (in order of preference):
- use reported outcomes that have been appropriately adjusted for clustering in the analysis;
- perform an approximate correction for clustering by correcting effects according to an estimation of a design effect.
Dealing with missing data
Where there are missing data, if there is sufficient evidence to support an assumption of being missing at random, we will use data from available data analysis. If this is not the case, we will:
- where possible, contact the original investigators to request missing data;
- consider alternative approaches to imputing missing data, such as assuming biologically plausible outcomes, and we will undertake sensitivity analyses to test the effects of these assumptions on meta-analysis outcomes.
Assessment of heterogeneity
We will assess statistical heterogeneity of the data using a Chi
Assessment of reporting biases
We will note any incomplete reporting of outcomes, perform sensitivity analyses if required and create a funnel plot to assess potential publication bias. We will only test for funnel plot asymmetry if the review has more than 10 included studies.
Studies must report a measure of variance for outcome measures to be included in the meta-analysis or a statistic from which this can be derived. If no standard deviation is reported, we will calculate standard deviations if possible from standard errors, P values or confidence intervals (CI).
We will conduct a meta-analysis using random-effects modelling. For continuous outcomes (i.e. bone density outcomes) we will calculate standardised mean differences. For dichotomous outcomes (total withdrawals and withdrawals due to adverse events) we will determine the risk ratio. Because of the difficulty in defining a clinically significant effect for early life bone density changes, we will not present the results for the primary bone density outcomes as number needed to treat but rather as the percentage change in bone density outcomes over the duration of the study, as we did in our previous reviews of vitamin D supplementation and of calcium supplementation for bone density in children (Winzenberg 2010; Winzenberg 2013).
Where possible, the analyses will be based on intention-to-treat data from the individual clinical trials. We will grade evidence using the GRADE system as described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).
We will re-express SMDs to estimate an absolute benefit in g/cm
The major outcomes that will be presented in the 'Summary of findings' table will be: femoral neck, total hip, lumbar spine and distal forearm bone mineral density, total body bone mineral content, total adverse events and withdrawals due to adverse events.
Subgroup analysis and investigation of heterogeneity
We anticipate that there will be a limited number of studies, which will restrict our ability to perform subgroup analyses. Where sufficient data are available, we will perform the following subgroup analyses or meta-regression on the following variables in order of priority (based on the findings of our review of vitamin D in children (Winzenberg 2010; Winzenberg 2011a)) to investigate effect modification and heterogeneity: baseline vitamin D levels, dose of vitamin D administered, compliance, sex, mode of infant feeding, age and ethnicity. We will perform subgroup analyses (for compliance, sex, mode of infant feeding, age and ethnicity) in RevMan 5 (RevMan 2012). We will perform meta-regression in Stata/IC 12.1 for Windows using the metareg macro (for mean study baseline vitamin D levels and dose of vitamin D administered).
We will perform the following sensitivity analyses:
- an assessment of the impact of risk of bias (Assessment of risk of bias in included studies) on the results;
- fixed-effect modelling compared with random-effects modelling;
- an analysis of adverse event outcomes using a per protocol approach rather than an intention-to-treat approach.
Professor Jones receives a National Health and Medical Research Council Practitioner Fellowship and Associate Professor Winzenberg a National Health and Medical Research Council/Primary Health Care Research Evaluation and Development Career Development Fellowship. Dr Van der Mei receives an Australian Research Council Fellowship.
Appendix 1. MEDLINE search strategy
3. exp bone density/
4. bone loss$.tw.
5. (bone adj2 densit$).tw.
7. exp vitamin d/
8. vitamin d.tw.
9. vitamin d2.tw.
10. vitamin d3.tw.
11. exp Ergocalciferols/
13. exp Cholecalciferol/
17. dihydroxyvitamin D3.tw.
20. 6 and 19
21. randomized controlled trial.pt.
22. controlled clinical trial.pt.
25. drug therapy.fs.
30. exp animals/ not humans.sh.
31. 29 not 30
32. 20 and 31
33. limit 32 to "all child (0 to 18 years)
Contributions of authors
All authors contributed to the design of this protocol, commented on the manuscript draft and approved the final version of this protocol. All authors will contribute to interpretation of results and drafting of the review manuscript. TW will be primarily responsible for project management, data analysis (with input from GJ) and drafting of the final review manuscript. KS and IV will perform assessment of studies against the inclusion and exclusion criteria. KS and TW will extract data and undertake 'Risk of bias' assessment of included studies.
Declarations of interest
None to declare.
Sources of support
- Menzies Research Institute Tasmania, University of Tasmania, Australia.In kind support
- Department of Health and Human Services, Tasmanian Government, Australia.In kind support
- National Health and Medical Research Council, Australia.Fellowship support to two authors (GJ/TW)
- Australian Research Council, Australia.Fellowship support to one author (IV)