Proof of evidence

In medical science, the evidence relies on randomised controlled trials (RCT), which are able to prove causal relationships between nutrient intake and bone mineral accrual. With a delicate design, a dose-responsive effect can also be explored⁽ Reference Viljakainen, Natri and Karkkainen ^{6 ,} Reference Ma, Huang and Yang ^{7 )}. Highest in the hierarchy are meta-analyses⁽ Reference Winzenberg, Shaw and Fryer ^{8 –} Reference Huncharek, Muscat and Kupelnick ^{10 )} and systematic reviews, which offer an overall picture of the state of evidence and may therefore modify the focus of research.

RCT still represent a simplified approach, as they typically focus on a single, specific nutrient. RCT have several limitations, such as focusing on a selected population. Especially in children and adolescents, only single centre studies are reported, and it is not clear how generalisable the results are into other populations. Another challenge is confounders and being able to control for these in the study design. A confounder such as physical activity is shown to enhance bone mineral accrual independently⁽ Reference Baxter-Jones, Kontulainen and Faulkner ^{11 )}, as well as acting together with calcium⁽ Reference Specker and Binkley ^{12 )}. Likewise, the pubertal stage may vary across age groups. The timing of accelerated growth in terms of peak height velocity occurs during the mid-pubertal stage in girls, but somewhat later in boys⁽ Reference Molgaard, Thomsen and Michaelsen ^{1 ,} Reference Bailey, Martin and McKay ^{4 ,} Reference Ma, Huang and Yang ^{7 )}, while weight status modifies both puberty and the bone mass accretion itself⁽ Reference Vandewalle, Taes and Van Helvoirt ^{13 )}. In addition, compliance with the study protocol could affect the results. A dilution effect will be marked with poor compliance and both intention-to-treat and per-protocol analysis should be reported. Without an attempt to follow compliance, an RCT should not be considered high-quality, since it is likely to give false-negative findings⁽ Reference Kehoe, Chheda and Sahariah ^{14 )}. The final concern is drop-outs, for which the researcher should be prepared from the outset. According to ethical guidelines, subjects are allowed to cancel their participation whenever, but researchers are eager to know the reason. This could reveal something about the acceptance of the dose or product. We are interested in long-term effects, as they mimic lifestyles that are beneficial for bone throughout the life. So far, only a few RCT have been able to answer this question and in fact many of them do not plan to look at long-term effects.

Calcium

In 2008, a meta-analysis by Huncharec et al.⁽ Reference Huncharek, Muscat and Kupelnick ^{10 )} identified twenty-one original RCT and summarised their findings on the effect of dairy/calcium supplementation on bone in children. Increased dietary calcium/dairy products modestly but significantly increased total body and lumbar spine bone mineral content in children. The effect was most profound in children with low baseline intakes. More recent RCT have focused on populations with low habitual calcium intakes and these have reinforced the message: calcium supplementation enhances bone mass augmentation in children and adolescents⁽ Reference Ma, Huang and Yang ^{7 ,} Reference Lambert, Eastell and Karnik ^{15 ,} Reference Yin, Zhang and Liu ^{16 )}.

An elegant continuum for this was provided by a supplementation study of Gambian boys with extremely low habitual calcium intakes (<300 mg/d). Prepubertal boys were randomised to receive either a placebo or 1000 mg calcium from supplements for 12 months⁽ Reference Ward, Cole and Laskey ^{17 )}. The group started to follow the long-term effects 1 year after ending the trial. Of forty boys in each arm, twenty five and  twenty nine completed the 12-year follow-up. The key finding was that although the peak velocity occurred earlier in the supplemented group, the total amount of bone accrued was not affected. In line with earlier study⁽ Reference Lambert, Eastell and Karnik ^{15 )}, the effect of supplementation on bone was short-lived, which highlights constant adaptation to the prevailing circumstances.

Vitamin D

In 2010, a combined meta-analysis and systematic review was undertaken on the effects of vitamin D on bone mineral accrual⁽ Reference Winzenberg and Jones ^{9 )}. At the time, the authors identified six high-quality vitamin D RCT to be included. The challenge was that these were based on rather diverse populations in many aspects: Pakistani immigrants in Denmark, Finnish girls, Saudi Arabian teens and Chinese girls; enrolment age varied between 10 and 17 years; vitamin D doses varied from 3·3 μg/d and 350 μg week and baseline 25-hydroxy vitamin D concentration 7·3–49·5 nm/l; and only one study included boys. Changes in bone mineral accrual were measured with dual-energy absorptiometry from different sites. The results from individual studies were optimistic, but still not conclusive: the effect of vitamin D supplementation on changes in bone mineral density was without effect or positive. The review summarised that ‘vitamin D supplementation is unlikely to be beneficial for children and adolescents with normal vitamin D level,’ although the greatest effect sizes were noticed in those with higher than deficient vitamin D status (25-hydroxy vitamin D > 25 nm/l)⁽ Reference Viljakainen, Natri and Karkkainen ^{6 ,} Reference El-Hajj Fuleihan, Nabulsi and Tamim ^{18 ,} Reference Cheng, Lyytikainen and Kroger ^{19 )}.

Based on the meta-analysis⁽ Reference Winzenberg and Jones ^{9 )} children with severe vitamin D deficiency could be optimal targets for vitamin D supplementation, as clinically relevant improvements in bone mineral accrual are anticipated. But in very poor, underprivileged societies, severe vitamin D deficiency coexists with multiple nutrient deficiencies, e.g. stunting of growth, anaemia and protein malnutrition⁽ Reference Ekbote, Khadilkar and Chiplonkar ^{20 )}, and focusing entirely on vitamin D may not be sufficient. Conversely, supplementing Bangladeshi women for 12 months with vitamin D solely or in combination with micronutrients improved bone mineral density in the femur similarly in both groups⁽ Reference Islam, Shamim and Viljakainen ^{21 )} while effects on overall wellbeing were not evaluated.

Protein

Dietary protein intake is shown to be associated with bone health, at least in elderly subjects. Optimal linear growth requires adequate dietary intake of protein, and protein may play a role in bone mineral accrual, but the scientific evidence is scarce. The association of protein intake with bone status was evaluated in a longitudinal study including 229 German children⁽ Reference Alexy, Remer and Manz ^{22 )}. Growth and multiple lifestyle factors were followed annually for four preceding years. Bone and muscle parameters were assessed using peripheral quantitative computed tomography in the study. A positive association was observed between dietary intake of protein and several bone characteristics in the forearm: cortical cross-sectional area, periosteal circumference, bone mineral content, and stress and strain index. Interestingly, an inverse association was marked between potential renal acid load and bone mineral content and cortical area, which suggest that the type of protein source is also meaningful. These observations have been confirmed in more recent studies⁽ Reference Zhang, Ma and Greenfield ^{23 )}.

Food pattern

Until recently, much of the research exploring the role of nutrition on peak bone mass has concentrated on single nutrients. While this approach has provided data on the effects of various vitamins and minerals on bone, it also has limitations. Human subjects do not eat single nutrients, but foods, and the RCT described here do not answer the question as to which are good dietary sources of nutrients. In addition, in epidemiological studies, separating the effect of a single nutrient has proved to be problematic, and in most cases the effect of a single nutrient may be too relatively small to be detected in epidemiological studies. This gives rise to conflicting results and confusion for policy-makers. Food or dietary pattern-oriented studies avoid some of the fundamental challenges of studies assessing the role of single nutrients. Such studies also provide practical information for the public and may be more directly utilised when revising dietary guidelines. Dietary pattern is a rather stable measurement compared with the day-to-day variation in certain nutrient intakes. Approaches that apply foods or dietary patterns could teach us novel things about bone health.

An example of the approach is milk. Milk contains over thirty nutrients, including calcium, phosphorus, protein, magnesium and potassium, that could be considered as bone agents, and these may have additive or synergistic effects on bone. What evidence is there? A number of studies have looked at the association between milk consumption and bone mineral accrual. One of the earliest RCT was performed in Sheffield, UK⁽ Reference Cadogan, Eastell and Jones ^{24 )}. A total of 82 girls aged 12 years took part in this RCT. For randomly chosen girls, milk was delivered at home. At baseline, the intake of milk was relatively low, only 150 ml/d, but it increased by 300 ml during the 18-month trial. Bone mineral density increased in both groups, but more in the milk group: 9·6 v. 8·5 %. Serum concentrations of insulin-like growth factor I increased more in the milk group compared with the control group (35 v. 25 %)⁽ Reference Cadogan, Eastell and Jones ^{24 )}.

Similarly, in a prospective study, milk intake was associated with higher axial growth in a 6-year follow-up of American girls⁽ Reference Berkey, Colditz and Rockett ^{25 )}. Further evidence from a Copenhagen study cohort of 17-year old boys and girls suggested that bone mineral accretion is associated with both total protein and milk protein intakes, but less so with meat protein intake⁽ Reference Budek, Hoppe and Ingstrup ^{26 )}, which implies that milk is an optimal protein source and synergistic effects are likely to occur. How about elimination of milk from the diet? In a Polish case–control study of 5–20-year-old children and adolescents, milk allergy was associated with an increased risk of fractures in girls⁽ Reference Konstantynowicz, Nguyen and Kaczmarski ^{27 )}. Similarly, in a New Zealand study, milk-avoiding children (n 50) had a higher incidence of fractures (twenty fractures) than a birth cohort of 1000 children from the same area (eight fractures)⁽ Reference Goulding, Rockell and Black ^{28 )}. In addition, higher hip fracture incidence has been reported in elderly subjects consuming at most one portion of milk per week in the Framingham Osteoporosis Study⁽ Reference Sahni, Mangano and Tucker ^{29 )}. The decrease in hip fracture risk was 40 % in the milk-favouring group; this is a major effect that could not be achieved with any single nutrient.

Longitudinal studies with dietary assessment contain data on habitual use of foods and are able to create food patterns that describe the variability, quality and choices of individuals. Compared with nutrient intakes, this gives broader insight and allows new strategies for preventing osteoporosis. There are several ongoing prospective studies, e.g. the Framingham Osteoporosis Study, Women's Health Initiative, Canadian Multicentre Osteoporosis study, Kuopio Osteoporosis Study, but these have focused on older populations and are considered as secondary prevention studies for osteoporosis. The findings from the Canadian Multicentre Osteoporosis study have been appealing: a diet rich in vegetables and fruits, evidently a nutrient-dense diet, prevents bone loss and low-energy fractures in Canadian women, but an energy-dense-nutrient-poor diet seems not to accelerate it⁽ Reference Langsetmo, Hanley and Prior ^{30 )}. Future research should focus on younger age groups to identify the life-course determinants for osteoporosis during prenatal, infancy, childhood and adolescence that would help to maximise peak bone mass. Diet and physical activity are the modifiable factors, but being able to describe these in more detail and the opposite as well, being able to identify lifestyle risk factors for impaired bone mineral accrual is important. In addition, the interaction of lifestyle factors with genetic factors has remained unclear. This would give us tools to recognise subjects at risk more easily and readily, and furthermore to be able to educate them (Fig. 1).

Fig. 1. Life-course determinants of osteoporosis and future perspectives.

Conclusion

Bone mass accrual is affected by multiple factors, of which nutrition is a modifiable one. Recent RCT have confirmed the effects of single nutrients on bone health, but the effects may be too small to be detected in epidemiological studies. Food or dietary pattern is a rather stable measurement compared with day-to-day variation in certain nutrient intakes. Typically dietary patterns evolve during growth, but remain stable from late adolescence. The effects of a whole diet are larger compared with a single nutrient. Research on recognising dietary patterns that benefit bone mineral accrual is warranted side-by-side with RCT to broaden our understanding of nutrition and bone health. We are interested in long-term effects, and cohorts with longitudinal follow-up are needed to provide real-life data on nutrition and bone health.