Evidence for Effect of Supplementation on Fracture Prevention

Among the community-dwelling populations without prior history of fractures or known vitamin deficiency or osteoporosis included in this review, we rated the strength of evidence as low for no benefit of supplementation with vitamin D alone or vitamin D with calcium on fracture prevention over 3–7 years. This is consistent with the findings of the prior review and is not surprising given that only one new study was identified. Findings were imprecise and confidence intervals in all but one study included the null effect and the absolute differences in incidence reported may not be clinically meaningful. Few studies were powered for fractures as a primary end point, and even those that were (i.e., the WHI CaD Trial) were powered based on an effect size that was nearly twice as large as what was observed in the trial. Although the primary intent-to-treat analysis in the WHI CaD Trial was a null effect, some consider the bone density changes observed, the favorable per-protocol analyses of adherent participants, and some of the favorable subgroup analyses among nonusers of supplements at baseline and among older participants as evidence of a favorable effect on bone health.⁷¹ We did not consider any subgroup analyses findings in our assessment of the strength of evidence because of the known methodologic limitations and challenges associated with interpreting subgroup findings.¹²³

We found limited evidence to draw conclusions regarding the impact of calcium alone on fracture prevention and rated the strength of evidence as insufficient. We found only two eligible studies (N=339). Only one reported a clinical fracture outcome in addition to incident morphometric vertebral fractures; the other reported only morphometric vertebral fractures. Small sample sizes and relatively rare event rates in the studies led to imprecise effect estimates.

The body of evidence on vitamin D alone is applicable to men and postmenopausal women, while the body of evidence for vitamin D with calcium and for calcium alone was limited to postmenopausal women. Daily doses of vitamin D ranged from 300 to 700 IU; one study used a 100,000 IU oral dose every 4 months (equivalent to 833 IU per day). Daily oral doses of calcium ranged from 500 to 1,600 mg. However, not enough eligible studies were identified to ascertain the influence of dose, route, or frequency on incident fractures.

We found some evidence of reporting bias for this KQ. One study (N=1,180) comparing calcium alone, vitamin D with calcium, or placebo was designed with fractures as a primary outcome and was completed in 2005, although no fracture outcomes have been published to date. Other study findings have been published.^{108, 124} Per the study author, data from the study suggested no effect on fracture incidence; however, the study was not published because of concerns related to study contamination because of participant use of alendronate, which came to market during the study (personal communication with author). The identification of other unpublished studies with null findings would increase the certainty for drawing conclusions about the lack of effect of supplementation on fracture prevention.

Evidence for Effect of Supplementation on Harms

This review focused primarily on four harms; all-cause mortality, kidney stones, CVD, and cancer. We were unable to ascertain the impact of dose, duration, or frequency on these harms because not enough eligible studies were available. Though cohort and case-control studies of supplementation were eligible for KQ 2 outcomes in this review, we excluded those identified through our search for poor quality because of many of the methodologic limitations also noted by others.^{125, 126} Further, we did not consider studies evaluating the association between serum vitamin D levels and all-cause mortality, kidney stones, CVD, or cancer incidence as this has been previously synthesized.¹⁵ Thus, the evidence for harms that we summarized on behalf of the USPSTF comes from randomized controlled trials.

The evidence for the effect of supplementation with vitamin D alone or with calcium on all-cause mortality over 3–7 years suggests no clinically meaningful harm. The absolute risk differences ranged from -1.9 percent to 0.1 percent, but findings were imprecise; thus, we assigned a low strength of evidence to this finding. This body of evidence is applicable to men and postmenopausal women. We found the evidence for calcium alone to be limited for assessing impact on all-cause mortality. The single study available suggests no effect on mortality, but was very imprecise, and only included men. Thus, the strength of evidence for calcium alone on this outcome was rated as insufficient.

The evidence for the impact of supplementation on incident kidney stones was mixed. We identified no eligible studies of vitamin D alone that reported this outcome, resulting in an insufficient strength of evidence rating. The evidence for calcium alone suggests no increased incidence of kidney stones over 2–4 years, although findings are imprecise. Further, this body of evidence was limited by lack of information on how kidney stone outcomes were ascertained; thus, we assigned a low strength of evidence to this finding. The body of evidence on vitamin D with calcium comes from three RCTs. One of the trials (the WHI CaD Trial) was large and provided reasonably precise estimates of a small harm, and when pooled with two smaller studies with nonsignificant differences, this harm persisted. %%%Thus, we assigned a moderate strength of evidence for harm.

The evidence for the effect of supplementation with vitamin D alone or with calcium suggests no clinically meaningful harm with respect to CVD outcomes over 4–7 years. The body of evidence related to vitamin D alone included three RCTs that were consistent, but the estimates of effect were imprecise. The body of evidence for vitamin D with calcium was limited to a single study in women (WHI CaD Trial) with a sufficient sample size and event rate for precise estimates. Thus, we assigned a low strength of evidence for no harm for vitamin D alone or with calcium interventions. Findings from one of the two post hoc analyses of the WHI CaD Trial suggested that trial participants assigned to supplementation with vitamin D and calcium who were not taking personal supplements at the time of randomization had a marginally increased risk of cardiovascular events relative to those who were taking personal supplements at the time of randomization.¹¹⁶ However, the post hoc analysis by the WHI CaD study authors did not report similar findings,⁹⁸ possibly because of slight differences in the way in which outcomes were specified between the two analyses.

We found the evidence limited for assessing the effect of calcium alone on CVD outcomes. The single study suggested no effect over 2 years but was limited by imprecise estimates and minimal information about outcome specification and ascertainment; thus, we rated this body of evidence as insufficient. The role of dietary and supplemental calcium on intermediate CVD outcomes (i.e., vascular calcification) and clinical CVD outcomes has been the subject of recent debate, with several analyses and meta-analyses published related to this issue in the past several years.^125-130 Most of the meta-analyses and systematic reviews on this topic included broader study populations and settings (e.g., institutionalized elderly, participants with prior history of fracture) than specified in our review. These analyses have found mixed results, some suggesting a small increased risk for CVD^{126, 128} and others suggesting no effect (either harm or benefit)^{125, 129} or inconclusive findings.¹²⁷ Several of these reviews included observational study designs. Of particular concern in using observational evidence to assess this relationship is that osteoporosis (a common indication for calcium supplement) and CVD risk factors overlap (e.g., smoking, physical activity), leading to high potential for confounding when looking at the association between calcium use and CVD events.

Last, we found the evidence for the impact of vitamin D alone and calcium alone to be limited for drawing conclusions related to the impact of supplementation on cancer incidence; thus, we rated these bodies of evidence as insufficient. Two RCTs of vitamin D alone reported inconsistent and imprecise findings; only a single study reported the impact of calcium alone and its findings were imprecise. The evidence for vitamin D with calcium supplementation over 4–7 years suggests no increased cancer incidence, but results were somewhat inconsistent; thus, we assigned a low strength of evidence to this finding. These findings are only applicable to postmenopausal women.

Limitations of the Evidence

Most studies were not powered for the fracture or harm outcomes considered in this review; thus, small sample sizes and low event rates resulted in imprecise effect estimates. Some studies, notably the WHI CaD Trial, allowed for use of personal calcium and vitamin D supplements during the study; thus, these trials could be characterized as trials of provider-directed supplementation, and some have suggested this design feature as an explanation for the null intention-to-treat analysis findings reported by the WHI CaD Trial.¹³¹

Heterogeneity in outcome specification is another limitation of this body of evidence. The specific types of fractures that were considered as contributing to “total fracture” included both traumatic and osteoporotic in most studies, and the specific sites contributing to total fractures varied across studies. Author queries were required to determine whether some studies reporting vertebral fractures were reporting clinical or morphometric fractures. Studies evaluating harms varied in specificity of definition or rigor of harm outcome ascertainment. Because harms were rarely the main study aim, little information was provided regarding how harms were defined, ascertained, or validated. Some studies relied on self-report, some on adverse event reporting during study monitoring; others relied on secondary data sources (registries, claims, death certificates) to identify cases. Although some evidence on men exists, the majority of this body of evidence is applicable to postmenopausal women, and few studies include populations that are racially representative of the U.S. population. Finally, only a few studies evaluated doses more than 800 IU per day. The evidence on calcium included doses ranging from 400 mg to 1,600 mg per day.

Because this review was narrower in scope than other published reviews of vitamin D (with or without calcium), the conclusions differ somewhat from the conclusions drawn from reviews with a broader or different scope. Bolland and Grey discuss the issue of discordant results from different meta-analyses on the same topic using vitamin D supplementation and fracture as an example.¹³² In their analysis, differences in trial selection, outcome definitions used, and analytic approaches explain the majority of differences in findings. Across a body of evidence of 25 trials, they found strong statements concluding both benefit and no benefit. Thus, it is important to consider the scope of the populations and interventions included when drawing conclusions from the body of evidence in this review to avoid inappropriate comparisons to reviews with a different scope. We contrast our findings with two recent systematic reviews below.

The 2014 Cochrane review evaluated vitamin D and vitamin D analogues for preventing fractures and, similar to our review, found no benefit for vitamin D alone; however, they concluded that vitamin D with calcium may prevent fracture.¹³³ The study populations considered in the Cochrane review included participants with osteoporosis and institutionalized participants and secondary prevention populations. The fracture benefits overall appear to be largely attributable to benefits among the high-risk populations, with little to no benefit in lower-risk populations (1 fewer hip fracture per 1,000 community-dwelling adults per year [95% CI, 0 to 2]). Like our review, the Cochrane review concluded that vitamin D with calcium was associated with increased gastrointestinal and renal disease, but did not adversely affect the risk of death.

The 2016 systematic review and meta-analysis of calcium and vitamin D supplementation conducted on behalf of the National Osteoporosis Foundation (NOF) included eight trials.^134-136 Two were conducted in institutional settings, two were exclusively secondary prevention trials, and the study population of one trial had more than 50 percent of participants with a prior fracture; thus, all these studies were out of scope for our update review. Of the remaining three trials in the NOF analysis, we used one in our sensitivity analyses because it had between 20 and 50 percent of subjects with a history of prior fracture.¹⁰⁷ The other two trials, Dawson-Hughes et al⁷⁵ and the WHI CaD trial,⁷¹ were included in our review; however, the NOF analysis used data from WHI CaD subgroups related to adherence to assigned pills and personal supplement use, not data from the intent-to-treat analysis. The NOF analysis reported an overall RR for total fracture incidence of 0.86 (95% CI, 0.75 to 0.98) and for hip fractures of 0.61 (95% CI, 0.46 to 0.82).