Calcium intake and bone mineral density: systematic review and meta-analysis

Literature search

As part of a broader search for studies of calcium intake and health, we searched Ovid Medline and Embase in July 2013 and updated the search using Pubmed and Embase in September 2014 for randomised controlled trials of calcium, milk, or dairy intake, or calcium supplements with BMD as an endpoint. We also hand searched recent systematic reviews, meta-analyses, and any other articles included in our review for other relevant articles. Appendix 1 provided details of the searches.

Patient involvement

No patients were involved in setting the research question or the outcome measures, nor were they involved in the design and implementation of the study. There are no plans to involve patients in dissemination.

Study selection

Included studies were randomised controlled trials in participants aged >50 at baseline with BMD measured by dual energy x ray absorptiometry (DXA) or precursor technology such as photon absorptiometry. We included studies that reported bone mineral content (BMC) because BMD is obtained by dividing BMC by bone area and therefore the two are highly correlated. Studies in which most participants at baseline had a major systemic pathology other than osteoporosis, such as renal failure or malignancy, were excluded. We included studies of calcium supplements used in combination with other treatment provided that the other treatment was given to both arms (such as calcium plus vitamin K versus placebo plus vitamin K), and studies of co-administered calcium and vitamin D supplements (CaD). Randomised controlled trials of hydroxyapatite as a dietary source of calcium were included because it is made from bone and contains other minerals, hormones, protein, and amino acids in addition to calcium. One author (WL or MB) screened titles and abstracts, and two authors (WL, MB, or VT) independently screened the full text of potentially relevant studies. The flow of articles is shown in figure A in appendix 2.

Data extraction and synthesis

We extracted information from each study on participants’ characteristics, study design, funding source and conflicts of interest, and BMD at the lumbar spine, femoral neck, total hip, forearm, and total body. BMD can be measured at several sites in the forearm, although the 33% (1/3) radius is most commonly used. For each study, we used the reported data for the forearm, regardless of site. If more than one site was reported, we used the data for the site closest to the 33% radius. A single author (VT) extracted data, which were checked by a second author (MB). Risk of bias was assessed as recommended in the Cochrane Handbook.¹¹ Any discrepancies were resolved through discussion.

The primary endpoints were the percentage changes in BMD from baseline at the five BMD sites. We categorised the studies into three groups by duration: one year was duration <18 months; two years was duration ≥18 months and ≤2.5 years; and others were studies lasting more than two and a half years. For studies that presented absolute data rather than percentage change from baseline, we calculated the mean percentage change from the raw data and the standard deviation of the percentage change using the approach described in the Cochrane Handbook.¹¹ When data were presented only in figures, we used digital callipers to extract data. In four studies that reported mean data but not measures of spread,^{12 13 14 15} we imputed the standard deviation for the percentage change in BMD for each site from the average site and duration specific standard deviations of all other studies included in our review. We prespecified subgroup analyses based on the following variables: dietary calcium intake v calcium supplements; risk of bias; calcium monotherapy v CaD; baseline age (<65); sex; community v institutionalised participants; baseline dietary calcium intake <800 mg/day; baseline 25-hydroxyvitamin D <50 nmol/L; calcium dose (≤500 v >500 mg/day and <1000 v ≥1000 mg/day); and vitamin D dose <800 IU/day.

Statistics

We pooled the data using random effects meta-analyses and assessed for heterogeneity between studies using the I² statistic (I² >50% was considered significant heterogeneity). Funnel plots and Egger’s regression model were used to assess for the likelihood of systematic bias. We included randomised controlled trials of calcium with or without vitamin D in the primary analyses. Randomised controlled trials in which supplemental vitamin D was provided to both treatment groups, so that the groups differed only in treatment by calcium, were included in calcium monotherapy subgroup analyses, while those comparing co-administered CaD with placebo or controls were included in the CaD subgroup analyses. We included all available data from trials with factorial designs or multiple arms. Thus, for factorial randomised controlled trials we included all study arms involving a comparison of calcium versus no calcium in the primary analyses and the calcium monotherapy subgroup analysis, but only arms comparing CaD with controls in the CaD subgroup analysis. For multi-arm randomised controlled trials, we pooled data from the separate treatment arms for the primary analyses, but each treatment arm was used only once. We undertook analyses of prespecified subgroups using a random effects model when there were 10 or more studies in the analysis and three or more studies in each subgroup and performed a test for interaction between subgroups. All tests were two tailed, and P<0.05 was considered significant. All analyses were performed with Comprehensive Meta-Analysis (version 2, Biostat, Englewood, NJ).

Baseline characteristics

We identified 59 randomised controlled trials of calcium intake that reported BMD as an outcome.^{7 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70} Fifteen studied dietary sources of calcium (n=810 calcium, n=723 controls),^{16 17 18 19 20 21 22 23 24 25 26 27 28 29 30} and 51 studied calcium supplements (n=6547 calcium, n=5710 controls).^{7 12 13 14 15 17 19 20 21 22 26 28 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70} Table 1 shows study design and selected baseline characteristics for included studies of dietary calcium. Tables 2 and 3 show the study design and selected baseline characteristics for trials of calcium supplements, without and with additional vitamin D, respectively. Further details are in tables A-C in appendix 2. Of the 15 randomised controlled trials of dietary sources of calcium, 10 used milk or milk powder, two used dairy products, and three used hydroxyapatite preparations. Of the 51 trials of calcium supplements, 36 studied calcium monotherapy, 13 co-administered CaD, and two were multi-arm studies of both. Table 4 summarises other features of the trials. Most of them studied calcium without vitamin D in women aged <70 living in the community; the mean baseline dietary calcium intake was <800 mg/day; and most trials lasted ≤2 years. A calcium dose of >500 mg/day was used in most trials, but a higher proportion of trials of calcium supplements used a dose of ≥1000 mg/day. Table C in appendix 2 shows our assessment of risk of bias. Of the 15 trials of dietary sources of calcium, we assessed two as low risk of bias, six as moderate risk, and seven as high risk. Of the 51 trials of calcium supplements, we assessed 19 as low risk of bias, 12 as moderate risk, and 20 as high risk.

Primary analyses

Table 5 summarises the results of the meta-analyses. Increasing calcium intake from dietary sources increased BMD by 0.6-1.0% at the total hip and total body at one year and by 0.7-1.8% at these sites and the lumbar spine and femoral neck at two years (figs 1 and 2 . There was no effect on BMD at the forearm.

When we restricted the analyses to the 12 randomised controlled trials of milk or dairy products, by excluding three trials of hydroxyapatite, there was little change in the results. Calcium supplements increased BMD at all five skeletal sites by 0.7-1.4% at one year (figs 3 and 4 ), by 0.8-1.5% at two years (figs 5 and 6 ), and by 0.8-1.8% at more than two and a half years (fig 7) (range of duration of trials was three to five years).

When we used Egger’s regression model and visual inspection of funnel plots, data seemed skewed toward positive results with increased calcium intake from dietary sources or supplements in about half of analyses that included five or more studies. The asymmetry of the funnel plot was caused by more small-moderate sized studies reporting larger effects of calcium on BMD than expected, raising the possibility of publication bias. Seven multi-arm randomised controlled trials included a dietary source of calcium arm and a calcium supplement arm,^{17 19 20 21 22 26 28} which allowed a direct comparison of the interventions. There were no significant differences between groups in BMD at any site in any individual trial, and there were also no significant differences between groups in BMD at any site or any time point in the pooled analyses (table D, appendix 2). We also tested for differences between the results of the trials of dietary sources of calcium and the trials of calcium supplements by comparing the two groups in subgroup analyses (table 4). There were no differences between the groups at any time point at the lumbar spine, total hip, or total body. At the femoral neck, there were greater increases in BMD at one year in the calcium supplement trials than in the dietary calcium trials, but at two years we found the opposite—that is, greater changes with dietary calcium than with calcium supplements. At the forearm, there were increases in BMD in the calcium supplement trials but no effect in the trials of dietary sources of calcium.

Subgroup analyses

We carried out additional subgroup analyses when there were 10 or more trials in an analysis and three or more trials in each subgroup. In the trials of dietary sources of calcium, these criteria allowed analyses to be carried out only on the one year results for the lumbar spine. For the calcium supplement trials, we carried out analyses on the one year and two year results for the lumbar spine, femoral neck, and forearm results, and the one year result for total body. Table E in appendix 2 shows that there were no consistent differences between subgroups based on calcium monotherapy versus CaD, age, risk of bias, calcium dose of ≥1000 mg/day versus <1000 mg/day, calcium dose of ≤500 mg/day versus >500 mg/day, vitamin D dose, baseline dietary calcium intake, or baseline 25-hydroxyvitamin D level. We did not find enough trials to carry out subgroup analyses based on sex and residence (community versus institution).

Principal findings

Increasing calcium intake from dietary sources slightly increased bone mineral density (BMD) (by 0.6-1.8%) over one to two years at all sites, except the forearm where there was no effect. Calcium supplements increased BMD to a similar degree at all sites and all time points (by 0.7-1.8%). In the randomised controlled trials of calcium supplements, the increases in BMD were present by one year, but there were no further subsequent increases. Thus the increases from baseline at both two and over two and half years at each site were similar to the increases at one year. The increases in BMD with dietary sources of calcium were similar to the increases with calcium supplements, except at the forearm, in both direct comparisons of the two interventions in multi-arm studies and in indirect comparisons of the two interventions through subgroup analyses. The increases in BMD were similar in trials of calcium monotherapy and CaD, consistent with a recent meta-analysis reporting that vitamin D monotherapy had no effect on BMD.⁷¹ There were no differences in changes in BMD in our subgroup analyses between trials with calcium doses of ≥1000 mg/day and <1000 mg/day or doses of ≤500 mg/day and >500 mg/day, and in populations with baseline dietary calcium intake of <800 mg/day and ≥800 mg/day. Overall, the results suggest that increasing calcium intake, whether from dietary sources or by taking calcium supplements, provides a small non-progressive increase in BMD, without any ongoing reduction in rates of BMD loss beyond one year. The similar effect of increased dietary intake and supplements suggests that the non-calcium components of the dietary sources of calcium do not directly affect BMD.

Strengths and limitations of the study

The strength of this meta-analysis is its comprehensive nature. We included 59 randomised controlled trials and assessed the effects of both dietary calcium sources and calcium supplements on BMD at five skeletal sites and at three time points. The size of the review permitted a comparison of the effects on BMD of different sources of calcium—dietary sources or supplements—and also the effects in important subgroups such as those defined by dose of calcium, use of co-administered vitamin D, and baseline clinical characteristics. The results are consistent with those from an earlier meta-analysis of 15 randomised controlled trials of calcium supplements, which reported an increase in BMD of 1.6-2.0% over two to four years.⁷²

An important limitation is that BMD is only a surrogate for the clinical outcome of fracture. We undertook the review, however, because many of the subgroup analyses in the dataset of trials with fracture as an endpoint have limited power,¹⁰ and a comparison between randomised controlled trials of dietary sources of calcium and calcium supplements with fracture as the endpoint is not possible because only two small randomised controlled trials of dietary sources of calcium reported fracture data.¹⁰ Another limitation is that in 60% of the meta-analyses, statistical heterogeneity between the studies was high (I²>50%). This indicates substantial variability in the results of included trials, although this was often because of the presence of a small number of outlying results. Subgroup analyses generally did not substantially reduce or explain the heterogeneity. We used random effects meta-analyses that take heterogeneity into account, and their results should be interpreted as reflecting the average result across the group of trials.

Implications of findings

The absence of any interaction with baseline dietary calcium intake or a dose-response relation suggests that increasing intake through dietary sources or through supplements does not correct a dietary deficiency (in which case greater effects would be seen in those with the lowest intakes or the highest doses). An alternative possibility is that increasing calcium intake has a weak anti-resorptive effect. Calcium supplements reduce markers of bone formation and resorption by about 20%,^{62 65 73} and increasing milk intake also reduces bone turnover by a similar amount.⁷⁴ Suppression of bone turnover by this amount might lead to the small observed increases in BMD.

Increases in BMD of about 1-2% over one to five years are unlikely to translate into clinically meaningful reductions in fractures. The average rate of BMD loss in older post-menopausal women is about 1% a year. So the effect of increasing calcium intake is to prevent about one to two years of normal BMD loss, and if calcium intake is increased for more than one year it will slow down but not stop BMD loss. Epidemiological studies suggest that a decrease in BMD of one standard deviation is associated with an increase in the relative risk of fracture of about 1.5-2.0.⁷⁵ A one standard deviation change in BMD is about equivalent to a 10% change in BMD. Based on these calculations, a 10% increase in BMD would be associated with a 33-50% reduction in risk of fracture. Therefore, the 1-2% increase in BMD observed with increased calcium intake would be predicted to produce a 5-10% reduction in risk of fracture. These estimates are consistent with findings from randomised controlled trials of other agents. The modest increases in BMD with increased calcium intake are smaller than observed with weak anti-resorptive agents such as etidronate⁷⁶ and raloxifene.⁷⁷ Etidronate, however, does not reduce vertebral or non-vertebral fractures, and raloxifene reduces vertebral but not non-vertebral fractures.⁷⁸ In contrast, potent anti-resorptive agents such as alendronate, zoledronate, and denosumab increase BMD by 6-9% at the spine and 5-6% at the hip over three years.^{79 80 81 82} These changes are associated with reductions of 44-70% in vertebral fracture, 35-41% in hip fracture, and 15-25% in non-vertebral fractures.⁷⁸ The magnitude of fracture reduction predicted by the small increases in BMD we observed with increased calcium intake are also consistent with the findings of our systematic review of calcium supplements and fracture.¹⁰ We observed small (<15%) inconsistent reductions in total and vertebral fracture overall but no reductions in fractures in the large randomised controlled trials at lowest risk of bias and no reductions in forearm or hip fractures.

The large number of randomised controlled trials that studied increased calcium intake and BMD and the consistency of the results across different populations in studies using higher or lower doses of calcium and in studies of dietary calcium sources or calcium supplements does not reveal any obvious gaps in the evidence. Any future trials conducted should have a strong rationale as to why the results are likely to differ from the large body of existing trial evidence. It is usually recommended that anti-resorptive agents are co-prescribed with calcium and vitamin D, although randomised controlled trials of such agents have shown reductions in risk of fracture^{83 84 85} and the expected increases in BMD^{64 86 87 88} without the co-administration of calcium and vitamin D. Randomised controlled trials clarifying the role of calcium and vitamin D in individuals using anti-resorptive agents might be valuable. In subgroup analyses, we stratified trials by thresholds of baseline dietary calcium intake (800 mg/day) and 25-hydroxyvitamin D (50 nmol/L). The clinical consequences of low calcium intake or vitamin D status such as osteomalacia, however, probably occur only at much lower thresholds, and there might also be interactions between calcium intake and vitamin D status. Analyses of individual patient data would be valuable in exploring these issues further.

Conclusions

In summary, increasing calcium intake from dietary sources increases BMD by a similar amount to increases in BMD from calcium supplements. In each case, the increases are small (1-2%) and non-progressive, with little further effect on BMD after a year. Subgroup analyses do not suggest greater benefits of increasing calcium intake on BMD in any subpopulation based on clinically relevant baseline characteristics. The small effects on BMD are unlikely to translate into clinically meaningful reductions in fractures. Therefore, for most individuals concerned about their bone density, increasing calcium intake is unlikely to be beneficial.