Dietary Reference Intakes for Calcium and Vitamin D (2011)

The first step in the decision-making process associated with the development of Dietary Reference Intakes (DRIs) is the identification of potentially useful measures—indicators—that reflect a health outcome associated with the intake of the nutrient. As described in Chapter 1, this is classically referred to as hazard identification, the first step of risk assessment. The available data are examined to determine their relevance and validity as well as strengths and limitations for elucidating a relationship between the health outcome of interest (including chronic disease risk) and the intake of the nutrient.

In considering reference values for calcium and vitamin D, there are challenges in organizing a data review to examine these nutrients independently, because they act in concert and are often administered together in experimental studies. To the extent possible, the independent effects of these nutrients were explored and taken into account; when this was not possible or not appropriate, the combined effect was considered. This chapter reviews evidence for calcium and vitamin D jointly to avoid redundancy. Evidence related to potential indicators for adverse effects of excess intake of calcium and vitamin D is reviewed separately in Chapter 6.

The array of potential health outcomes to be considered for these two nutrients was identified using five sources:

1. Agency for Healthcare Research and Quality (AHRQ) evidence report issued in 2007 (Cranney et al., 2007), hereafter referred to in this chapter as AHRQ-Ottawa without a reference citation; and

2. AHRQ evidence report issued in 2009 (Chung et al., 2009), hereafter referred to in this chapter as AHRQ-Tufts without a reference citation;

3. The Institute of Medicine (IOM) report Dietary Reference Intakes for Calcium, Phosphorus, Magnesium, Vitamin D, and Fluoride (IOM, 1997);

4. Literature searches conducted by the committee;

5. Publicly available input from stakeholders either through written submissions to the committee or as presented during the information gathering workshop.

As outlined in Chapter 1, the ARHQ analyses are highly relevant to DRI development. Evidence-based systematic reviews have been identified as a useful tool for the purposes of dietary reference value development (Russell et al., 2009), and the work of this committee was enhanced by the availability of these two high-quality evidence reports from AHRQ. The approach used, questions asked, data search criteria, and the detailed results from the AHRQ-Ottawa and AHRQ-Tufts can be found in Appendixes C and D.

In sum, the focus of AHRQ-Ottawa was on the:

• Association of specific circulating 25-hydroxyvitamin D (25OHD) concentrations with bone health outcomes in children, women of reproductive age, postmenopausal women, and elderly men;

• Effect of vitamin D dietary intake (fortified foods and/or supplements) and sun exposure on serum 25OHD levels;

• Effect of vitamin D on bone mineral density (BMD) and fracture or fall risk; and

• Identification of potential harms associated with vitamin D exposures above current reference intakes.

The AHRQ-Tufts evidence report analyzed data related to calcium and vitamin D with respect to a broader spectrum of health outcomes. AHRQ-Tufts also served to update and expand AHRQ-Ottawa. Specifically, AHRQ-Tufts focused on the:

• Relationship between vitamin D and growth, cardiovascular disease (CVD), body weight, cancer, immunological outcomes, bone health, all-cause mortality, hypertension/blood pressure, and BMD and bone mineral content (BMC); and

• Relationship between calcium and growth, CVD, body weight, and cancer.

Neither AHRQ report reviewed calcium alone as a factor in bone health.

A key component of systematic reviews of scientific literature is a specification of the quality of the available data. The AHRQ grading system is summarized in Box 4-1. In the case of the systematic analysis carried out by AHRQ-Ottawa, the Jadad scale (Jadad et al., 1996) was used for quality assessments of randomized controlled trials (RCTs). The Jadad scale is a validated scale designed to assess the methods used to generate random assignments and double blinding. The scale also scores whether there is a description of dropouts and withdrawals by intervention group. Jadad scores range from 1 to 5, and a total score of 3 and above indicates studies of higher quality. Further, to assess the quality of the observational studies, a grading system adapted from R. P. Harris et al. (2001) was used. In the case of the AHRQ-Tufts analysis, a three-category grading system (“A,” “B,” or “C”) was adapted from the AHRQ Methods Reference Guide for Effectiveness and Comparative Effectiveness Reviews (AHRQ, 2007). This system defines a generic grading system that is applicable to each type of study design including interventional and observational studies; it is summarized in Box 4-1.

The committee’s literature search identified relevant evidence outside the scope of, or not included in, the two AHRQ reports as well as newer data available after the cutoff date of the AHRQ-Tufts analysis in 2009. The nature of the literature search is outlined in Appendix E. The literature base that was included in the 1997 report of the IOM committee tasked with DRI development for calcium and vitamin D (IOM, 1997) was also considered. Additionally, information gathered as part of a public workshop and several open committee sessions (see Appendix J) and a white paper requested by the committee (Towler, 2009) were taken into account.

Through use of the five data sources listed above, health outcomes of potential interest were identified. They are listed alphabetically in Table 4-1 and are grouped by general outcome. In addition, there is the possibility of intermediate variables that are not validated biomarkers of effect for health outcomes, but which may have the potential to be useful in the development of DRIs. Two such variables were considered: serum 25OHD concentrations and levels of parathyroid hormone (PTH).

Within the scientific and clinical literature, there is a general hierarchy of study design. The lowest form of evidence is the idea or opinion,

followed, in ascending order, by case reports, case series, case–control studies, cohort studies, and, finally, the highest form of evidence, the randomized, controlled, double-blind trial (Croswell and Kramer, 2009). Only the RCT can show a causal relationship between an intervention and an outcome. Observational evidence can show only associative links, not causality. The highest level of observational evidence is the cohort study—a large, population-based, prospective investigation to compare an exposed group with an unexposed group. However, the cohort study does not reach the level of evidence of an RCT, because the intervention is not a random or chance event; rather it is the choice of the investigator (Croswell and Kramer, 2009). Nested case–control studies are a type of cohort study and were considered at that level of evidence; in some literature, populations from RCTs were evaluated as a cohort (adjusting for treatment assignment or limiting the analysis to the control group) and thus are at the same level of evidence as other observational research.

A summary of the strengths and weaknesses of the various types of observational studies and RCT studies is shown in Table 4-2. Flaws, biases, and confounding effects are an inevitable aspect of any study design, and the strength of a study therefore depends on the ability of the investigator to control such methodological obstacles. In addition, even well-designed studies can be weakened by complications such as loss to follow-up, missing outcomes, subject non-compliance, and a biased selection process (Baker and Kramer, 2008).

In addition to its consideration of the AHRQ analyses, the committee conducted searches of several online bibliographical databases, including Medline, Science Direct, and WorldCat/First Search. Evidence searches were carried out to identify relevant RCTs in support of a causal relationship between vitamin D and/or calcium and the health outcome under consideration, and these were weighted as the strongest type of evidence for development of a DRI. The second tier of evidence considered was observational to support associative relationships between vitamin D and/or calcium and a health outcome. Further examination was carried out to determine the quality of the observational evidence and whether the results were in agreement with RCT outcomes for a specific indicator. Potential confounders were also taken into account. Figure 4-1 shows the committee’s ranking of evidence by the strength of the study design. In the figure, RCTs prevail over observational and ecological studies as the strongest evidential support and were therefore necessary for a health outcome indicator to be further considered for DRI development. When the totality of evidence, including causal evidence, was supported by concordance

between RCTs and high-quality observational evidence and had strong biological plausibility, the committee gave further consideration to a potential indicator for development of a DRI. When observational evidence failed to support the findings of RCTs, the indicator’s validity for consideration was reevaluated, and a decision to give further consideration was made on the balance of the totality of evidence.

For each potential indicator discussed in this chapter, the review of evidence included consideration of the analytical approach, study population, and research protocol design and the overall quality of the evidence for each study reviewed. The introductory statement for each indicator includes ecological studies. Observations made from such studies require caution in their interpretation because the outcome measures are not known at the individual level, and inferring individual characteristics or

FIGURE 4-1 Ranking study designs: Ranking is shown in descending order of quality from top to bottom; the length of bars is arbitrary and indicates the relative strength of a study design.

relationships from group-level measures would be fallacious. Ecological studies, however, can contribute important information in more than an exploratory manner. Where it was relevant or needed in the absence of human studies, evidence for biological plausibility was included in the review as gleaned from experimental animal and mechanistic studies. The observational evidence reviewed included cross–sectional, case–control, and cohort (prospective and retrospective) studies. As pointed out previously, the strongest evidence among observational studies is from the cohort study. This study design offers an advantage over the case–control design in that it allows for observation of the incidence of a health outcome or the rate at which the health outcome develops in association with vitamin D or calcium intake or status in the population under study. In case–control studies, cases are included without identifying the entire “exposed” and “unexposed” populations from which they were derived, thus inferences drawn about a health outcome related to vitamin D or calcium intake or status are less reliable using this type of design.

As a tool to aid in the review process, the committee developed evidence “maps” for each indicator to provide an overarching view of the balance of relevant evidence from ecological and biological plausibility studies, observational studies, systematic reviews, and RCTs (including trials where the indicator was a primary outcome as well as other evidence from trials where the indicator was a secondary or non-pre-specified outcome). These served largely as an organizing tool and are included in Appendix F.

The organizational construct of the maps did not allow distinctions between studies relative to the quality of the study design; however, this was considered by the committee in the overall evaluation of data.

The nature of the data surrounding each potential indicator is described below, beginning with a brief statement about the condition under consideration, followed by a summary of the evidence for ecological and biological plausibility studies, observational studies, systematic reviews from the two AHRQ reports, and additional evidence not covered in the AHRQ reviews. Each indicator is then evaluated in a summary discussion of the utility of the evidence for DRI development.

Owing to the importance of a variety of acute and chronic diseases as public health concerns and the accumulating data focused on the hypothesis that vitamin D and/or calcium may have an impact on disease risk, it was crucial that this committee consider a wide spectrum of indicators for DRI development. After reviewing the available data, including recent systematic reviews from AHRQ and other literature, the committee chose to focus on areas where the research database is most compelling and the indicator is of public health concern within the context of DRI development. The following discussions review the roles of vitamin D and calcium in the reduction of risk for the health indicators identified in Table 4-1.

The entirety of evidence for each indicator that was reviewed by the committee cannot be presented in detail here, and the following discussions are a summary of relevant evidence. In drawing its conclusions about an indicator, the committee evaluated the strengths and weaknesses of the studies considered for each indicator, including an examination of the methods used for measuring an indicator, its relevance to total intake and functional or physiological outcomes, and the strength of the study design. This approach is summarized in Box 4-2.

As the second leading cause of death in the United States, cancer is a major public health concern. Cancer encompasses a wide range of malignancies with many variations in etiology and pathogenesis. Thus, the committee considered not only total cancer, but also specific malignancies in which vitamin D and/or calcium have been examined for an interaction thought to play a role.

Cancer is a disease in which genetically damaged cells within a tissue experience uncontrolled growth and invasion with subsequent spread to other host organs. The metastatic spread leads to dysfunction of vital organs causing significant morbidity and culminating in death. An expanding

array of experimental studies examining cells in culture and rodent models of cancer are providing evidence that vitamin D may have an impact on carcinogenesis at several organ sites (Deeb et al., 2007; Welsh, 2007; Davis, 2008). In parallel, epidemiological investigations of diverse approaches are examining the role of vitamin D in human cancer (WCRF/AICR, 2007; Yetley et al., 2009). In contrast, very few randomized and controlled prospective intervention trials with vitamin D targeting cancer as the primary outcome have been undertaken, leaving major gaps in understanding of causal relationships. Although more challenging to study in vitro, studies of dietary calcium in rodent models have also suggested a potential role in cancer risk; there are, as discussed below, experimental and clinical studies providing evidence in support of calcium as a modulator of carcinogenesis, particularly in the colon and rectal mucosa.

Cancer represents hundreds of different histopathologically distinct types of malignancy derived from virtually all organs and tissues. Investi-

gations into the cellular defects contributing to the carcinogenic process indicate that cancers, regardless of tissue origin, share in a specific set of defective biological processes (Hanahan and Weinberg, 2000) that enhance cell proliferation, survival, invasion, and metastasis. Although cancer studies initially suggested the possibility of a tissue-specific gene expression signature unique to a cancer type, it is now appreciated that multiple different mutational patterns contribute to the heterogeneity in biology and response to intervention among humans with cancer.

Biological plausibility Serum 25OHD levels are determined by both dietary intake and endogenous synthesis in the skin upon exposure to ultraviolet B (UVB) light. UVB exposure is often used as an indirect estimate of endogenous production of vitamin D in ecological studies of cancer incidence patterns. Several investigators associated lower UVB exposure with higher cancer mortality beginning decades ago (Apperley, 1941) and continuing with improved methods of estimating exposure (Boscoe and Schymura, 2006), as reviewed by IARC (2008). However, a large literature suggests that increasing latitude cannot be equated with decreasing vitamin D status, and cancer risk factors (exposure to UVB or other forms of ionizing radiation) vary with latitude. Importantly, an opposite gradient is well established for skin cancers, with a greater risk among populations residing in areas of high sun exposure (IARC, 1992). In general, ecological studies based upon estimated UVB exposure, vitamin D status, and cancer risk have many potential biases due to methodological considerations making causal biological inferences, particularly at the level of the individual, impossible.

Systematic reviews and meta-analyses Assessment of total cancer risk has been the subject of systematic reviews, including IARC (2008), WCRF/AICR (2007), and AHRQ-Tufts. Several studies, including those reviewed in AHRQ-Tufts, were examined by the committee in detail. Three intervention trials that examined total cancer as an outcome were identified from these reviews; these trials were originally designed to assess fracture risk, and none included total cancer as a pre-specified primary outcome (see Table 4-3). In both the Trivedi et al. (2003) and Lappe et al. (2007) osteoporosis trials cancer risk was determined from a secondary analysis of safety data that relied upon subjects notifying the investigators of the new diagnosis. Neither trial indicated a significant reduction in cancer incidence with vitamin D supplementation, whether given alone (Trivedi et al., 2003) or in combination with calcium and compared with calcium supplementation alone (Lappe et al., 2007). In the Lappe et al. (2007) trial, however, logistic regression analysis showed a significant reduction in risk for all cancers in the vitamin D plus calcium treatment group when

compared with the placebo group. Notably, the investigators could not exclude that cancers had been present at baseline or that cancers remained unnoticed at the end of the study. Moreover, the analysis of the multitude of outcomes in safety data raises the possibility of chance results that seem to be statistically significant but are the result of multiple comparisons being made within one data set.

Observational evidence in AHRQ-Tufts included a large 12-year prospective study of a cohort from the Third National Health and Nutrition Examination Survey (NHANES III) that examined associations between serum 25OHD levels and total cancer mortality as well as specific cancer mortalities. Serum 25OHD levels were found to be associated with gender, educational level, and race/ethnicity, but not with season/latitude. No interaction was detected, however, between serum 25OHD level and total cancer mortality (Freedman et al., 2007). In one frequently cited study included in the AHRQ-Tufts review, Giovannucci et al. (2006) prospectively examined a large cohort from the Health Professionals Follow-up Study (HPFS) for 14 years for multiple determinants of vitamin D, including diet, supplements, skin pigmentation, adiposity, and geography, and their associations with cancer mortality. This study found that each incremental increase in serum 25OHD level of 25 nmol/L was associated with a 17 percent reduction in total cancer incidence and a 29 percent reduction in total cancer mortality. Each of the determinants considered was found to influence plasma 25OHD levels among older men. These results should be viewed with caution, however, because of heterogeneity in serum 25OHD levels that is not accounted for by the variables used in the study, which included intakes based on self-administered semiquantitative food frequency questionnaires and self-reported weight and physical activity levels.

Taken together, the studies reviewed by AHRQ-Tufts, IARC (2008), and WCRF/AICR (2007) as a whole are not supportive of a role for vitamin D, with or without calcium in reducing risk for cancer.

Additional evidence from randomized controlled trials In addition to the trials identified in AHRQ-Tufts, a secondary analysis of data from the Women’s Health Initiative (WHI) trial examined the effect of combined supplementation of vitamin D and calcium (400 International Units [IU] of vitamin D and 1,000 mg of elemental calcium) on various health outcomes including cancer mortality (Lacroix et al., 2009). The results, with an average of 7 years of follow-up, indicated a non-significant trend toward reduction in risk for cancer mortality among postmenopausal women.

Observational studies One additional large cohort study, not included in the AHRQ reviews, was identified that examined serum 25OHD levels and risk for cancer mortality. This study examined cancer mortality among patients referred for coronary disease after a median of 7.75 years and

found a significant correlation between low serum 25OHD level defined as less than 25.5 nmol/L, and increased cancer mortality. No associations were detected, however, between calcitriol level and cancer mortality (Pilz et al., 2008). In total, the observational studies reviewed suggest that the association between 25OHD level and risk of death from all cancers is generally weak when considered over a broad range of serum 25OHD levels because of variability in outcomes between the studies reviewed. However, there may be a stronger association between low serum 25OHD levels and cancer risk. The evidence reviewed was not strong enough to conclude that associations between cancer mortality were dependent on latitude, or race/ethnicity.

The role of calcium in cancer risk was examined in one large prospective cohort study over 7 years of follow-up (Park et al., 2009). Calcium intake was found not to be related to total cancer risk in men, but a non-linear reduction in total cancer incidence in women was reported. A decreased cancer risk was found for calcium intakes up to approximately 1,300 mg/day, although no additional risk reduction was observed for higher intakes. Taken together, the heterogeneity among outcomes exhibited in these studies and the discrepancy in outcomes between observational and randomized trial evidence do not support a relationship between vitamin D or calcium and total cancer risk.

Concluding statement The totality of the available evidence from RCTs and observational association studies for a relationship between vitamin D and/or calcium and the risk for either incidence of or mortality from all cancers does not support the use of cancer mortality as an indicator for DRI development. The interpretation of the evidence reviewed is limited by the small number of studies identified and lack of consistency in associations between vitamin D intake or serum 25OHD levels and all cancer mortality. Interpretation is further complicated by the absence of large-scale RCTs examining total cancer risk as a pre-specified primary outcome. Given the lack of consistent evidence on associations between vitamin D intake or serum 25OHD level and total cancer, and the paucity of evidence on cancer as a primary outcome of vitamin D or calcium intervention in randomized trials, as well as inconsistency between findings in the available research for an effect of vitamin D or calcium supplementation or status on reducing risk for cancer, the committee could not draw a conclusion about the utility of the evidence for this indicator to support DRI development.

Risk for breast cancer is largely defined by reproductive endocrinology, with increased risk for those with early age of menarche, late menopause, no pregnancy, later age of first pregnancy, shorter duration of lactation, the

use of postmenopausal hormonal supplementation (Fentiman, 2002; Velie et al., 2005; Narod, 2006; Parsa and Parsa, 2009; Dietel, 2010). Dietary-related factors have been extensively reviewed with alcoholic drinks, adult attained height, and adult weight gain likely contributing to risk and with physical activity showing some benefit (WCRF/AICR, 2007). These characteristics must be considered when evaluating other putative breast cancer risk factors.

Biological plausibility The influence of the active form of vitamin D (calcitriol) on breast cancer cells in vitro is well characterized and includes anti-cancer effects such as cell cycle inhibition, reduced proliferation, enhanced sensitivity to apoptosis, and induction of differentiation markers (Welsh, 2004), which are likely mediated by the vitamin D receptor (VDR) (Matthews et al., 2010). A shortcoming in applying results from cell culture studies to risk for disease, however, is that the dose of calcitriol necessary to achieve tumor inhibition in vivo is frequently associated with hypercalcemic toxicity (Welsh, 2004; Matthews et al., 2010). Novel genomic approaches have begun to elucidate the gene expression signature of vitamin D in breast cancer cells and the mammary glands of mice (Matthews et al., 2010). Many of the genes identified show a consensus vitamin D response element (VDRE) in their promoter elements, indicating that they are specific targets of the vitamin D receptor (VDR¹) complex (Swami et al., 2003; Matthews et al., 2010). Since the discovery of polymorphisms in the Vdr gene, a search for associations of mutations with breast cancer has been undertaken, but with indeterminate results (Bertone-Johnson, 2009; McKay et al., 2009). An inverse association has been postulated between mammographic density, a putative breast cancer risk factor, and serum 25OHD levels in premenopausal women (Berube et al., 2004; Brisson et al., 2007). The role of dietary calcium intake and in breast cancer risk, however, is less well studied, and the potential biological mechanisms of action are not understood.

Systematic reviews and meta-analyses AHRQ-Tufts did not find any qualified systematic reviews that evaluated associations between vitamin D and calcium intake or serum 25OHD levels and risk for breast cancer. Three observational studies of sufficient methodological quality were identified that examined the relationship between 25OHD levels and breast cancer risk. A prospective cohort study described above for total cancer mortality reported that women whose 25OHD levels were in a higher stratification, were at significantly lower risk for breast cancer. There were, however,

only eight women in the higher stratification and a linear trend analysis was not significant (Freedman et al., 2007). A nested case–control study using data from the Nurses’ Health Study (NHS) (Bertone-Johnson et al., 2005) found no significant relationship between higher plasma 25OHD concentrations and decreased risk for breast cancer overall, except when the population was restricted to women over 60 years of age. Another nested case–control cohort study of postmenopausal women participating in the Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial also found no evidence supporting the hypothesis that higher plasma 25OHD concentrations were associated with reduced risk of breast cancer in this cohort (Freedman et al., 2008).

Analysis of the results of RCTs reviewed in AHRQ-Tufts found no significant effect of supplementation with both vitamin D and calcium on breast cancer incidence and no association between the intervention and risk for death from breast cancer. Subjects with lower baseline 25OHD levels were found to be at increased risk for breast cancer; however, the association was not significant after adjusting for body mass index and physical activity (Chlebowski et al., 2008).

A meta-analysis of evidence from observational studies carried out by IARC (2008) evaluated associations between serum vitamin D levels and cancer. The analyses for breast cancer risk indicated no significant or consistent associations. A literature review and all-inclusive meta-analysis of published studies of heterogeneous quality individually examined the impact of estimated vitamin D intake, circulating 25OHD levels, and calcium intake on breast cancer risk (Chen et al., 2010). Their analysis suggests an inverse relationship between risk and level of vitamin D intake, serum 25OHD₃ level, and calcium intake.

Additional evidence from randomized controlled trials WHI was used as a data source in an 8-year follow-up study for risk of benign proliferative breast disease, a putative premalignant condition associated with increased risk of subsequent cancer (Rohan et al., 2009). This study identified an association between risk for breast cancer and baseline age but found no effect of supplemental calcium and vitamin D intervention on reducing risk for breast cancer.

Observational studies Several case–control and cohort studies conducted subsequent to the systematic reviews were identified that examined associations between dietary and supplemental intake of vitamin D and calcium and risk for breast cancer, and these have shown mixed results. Rossi et al. (2009), a large case–control study in Italy, found an inverse association between vitamin D intake and risk for breast cancer at intakes of 188 IU/day or greater, suggesting a threshold effect; however, when risk was cal-

culated in the upper three deciles compared with the lower seven deciles the significant difference was attenuated. A population-based case–control study of women ages 25 to 74 years in Canada compared vitamin D and calcium intake from food alone or from food and supplements. When intake above 400 IU of vitamin D per day was compared with no intake, a reduced risk was found. Calcium supplement intake alone, however, did not correlate with reduced risk, although a significant inverse trend was identified (Anderson et al., 2010). Two studies were identified that examined associations between dairy intake and risk of breast cancer. Shin et al. (2002) analyzed data from the NHS 1980 cohort for dairy intake and incident breast cancer. Over 16 years of follow-up, a significant inverse association was found for premenopausal women consuming low-fat dairy products and breast cancer risk. No association was found for calcium and vitamin D intake and postmenopausal breast cancer risk. Supplemental calcium intake had no linear association and supplemental vitamin D intake a weak but non-significant association with breast cancer risk in both premenopausal and postmenopausal women. Using a similar study design, McCullough et al. (2005), in an analysis of participants from the Cancer Prevention Study II Nutrition Cohort, found that two or more daily servings of dairy products were inversely associated with breast cancer risk; however, no association was found for either calcium or vitamin D supplementation. Women with dietary calcium intakes above 1,250 mg/day had lower breast cancer risk than women with intakes at or below 500 mg/day. Altogether, these observational studies were of lower quality and thus not considered as strong support for an association between vitamin D and risk for breast cancer, and they were not well supported by randomized trial evidence.

Concluding statement In summary, although experimental studies are suggestive of a role for vitamin D in breast biology, a review of the available evidence from both RCTs and observational studies of associations between vitamin D and calcium and risk of breast cancer shows a lack of consistency between study outcomes and insufficiently strong evidence to support DRI development. Both retrospective and prospective studies do not show consistent associations between estimated vitamin D intake or 25OHD status and breast cancer risk. A paucity of RCTs of vitamin D, calcium, or both with breast cancer as a primary outcome further limited the strength of the evidence.

Foods, nutrients, and physical activity all interact in a complex array of mechanisms to influence colorectal cancer risk. There is convincing evidence that physical activity protects against colorectal cancer, whereas

red and processed meat, body fatness, and alcohol may increase the risk (WCRF/AICR, 2007). The committee’s review of studies on vitamin D and calcium and risk for colorectal cancers and possible protective benefits identified for calcium and vitamin D was inconclusive.

Biological plausibility A major role of the active form of vitamin D is to enhance calcium absorption by the intestine, and the molecular and cell biology has been well defined (Song and Fleet, 2007; Xue and Fleet, 2009). The VDR and the vitamin D converting enzyme, 1α-hydroxylase, are both expressed in the colon and rectum (Cross et al., 1997; Holt et al., 2002). Vitamin D has been reported to act on colonic epithelial and cancer cells to regulate growth factor and inhibitor expression and signaling pathways, including modulation of the cell cycle, sensitivity to apoptosis, and enhancement of cellular differentiation (Harris and Go, 2004; Yang et al., 2007). Many rodent models of colon carcinogenesis suggest that there is an increased risk for colon cancer associated with vitamin D deficiency; and a decreased risk associated with supplementation (Harris and Go, 2004; Yang et al., 2008; Newmark et al., 2009). However, few studies were identified that examined vitamin D over a range of dose levels. A recent review of findings from the Vdr-null mouse model indicates an increase in hyperplasia of the distal colonic epithelium and greater deoxyribonucleic acid (DNA) damage in vitamin D–deficient compared with wild-type mice (Bouillon et al., 2008). The independent role of calcium in modulating colon cancer risk is also under investigation. Although intracellular calcium plays a key role in cell biology and influences growth control processes that may be related to carcinogenesis, serum calcium is tightly regulated over a wide range of intakes. Thus, the potential mechanisms by which serum calcium levels could mediate risk for colon cancer may be through indirect effectors in metabolic pathways involved in tumorigenesis.

Systematic reviews and meta-analyses

Colorectal cancer The AHRQ-Tufts systematic review considered evidence for associations between 25OHD levels and risk for colorectal cancer mortality or incidence. One RCT found no significant difference between colorectal mortality or incidence and supplementation with vitamin D in an elderly population. One cohort study was identified that found an inverse association between high serum 25OHD levels and risk for colorectal cancer mortality, and two nested case–control studies in women found an inverse trend between serum 25OHD level and colorectal cancer incidence. Two nested case–control studies in men and three in both men and women found no significant associations between serum 25OHD level and risk of colorectal cancer.

The IARC (2008) meta-analysis found a significant protective effect for

serum 25OHD level against risk for colorectal cancer that correlated with each 2.5 nmol/L increase, although there was significant between-study heterogeneity. The results did not significantly differ by gender, mean population age, or cancer subsite (colon or rectum). The review noted that, based on multiple studies of circulating 25OHD and colorectal cancer risk, individuals in the high quartile or quintile of 25OHD level had about half the risk of colorectal cancer as did those in the lowest group. In another systematic review of studies examining associations between serum 25OHD levels and colorectal cancer, Bischoff-Ferrari et al. (2006a) concluded that the protective effect of 25OHD for decreased risk of colorectal cancers began at 75 nmol/L, and optimal levels were between 90 and 100 nmol/L. In contrast to these findings, the AHRQ-Ottawa systematic review reported that the studies reviewed were too inconsistent to permit conclusions to be drawn about specific serum 25OHD levels that conferred a decrease in risk.

Colorectal adenomas/polyps The AHRQ-Tufts systematic review considered evidence for associations between 25OHD levels and risk for colorectal adenomas. Colorectal adenomas or polyps are precursor lesions for colon cancer, and a number of investigations focused on the influence of vitamin D or calcium on the incidence of these surrogate markers for human colon carcinogenesis. A meta-analysis by Wei et al. (2008) of seven studies suggested that at the upper quintiles of circulating 25OHD levels there was a significant decrease in risk for colorectal adenoma. In parallel, these authors conducted a meta-analysis of vitamin D intake and colorectal adenoma risk in seven cohort and five case–control studies and found a marginally significant (11 percent) decreased risk among persons with high compared with low vitamin D intakes. The cut-points for the highest category of vitamin D intake varied between studies, with about one-third of the studies reporting cut-points of approximately 600 IU/day, one-third reporting cut-points between 250 and 600 IU/day, and one-third reporting cut-points of below 250 IU/day.

Stronger evidence has accumulated for a role of dietary calcium. The AHRQ-Tufts analysis identified four good quality cohort studies that evaluated the association between calcium intake and risk for colorectal adenoma. Two of these studies recruited men and women with a history of previous colorectal adenoma. One study found a significant inverse association between total calcium intake and colorectal adenoma recurrence after an average of 3.1 years of follow-up (highest [> 1,279 mg/day] vs. lowest [< 778 mg/day]) intake, whereas another found no significant association. Among two studies of healthy women without a history of colorectal adenoma one found a significant inverse association between total calcium intake and colorectal adenoma (highest vs. lowest intake, whereas the other found a borderline significant trend (highest [median, 1,451 mg/day] vs. lowest [median, 584 mg/day] intake. A Cochrane systematic review identified two randomized trials that found that calcium supplementation

reduced the incidence of recurrent colorectal adenoma (Weingarten et al., 2008). Overall, the evidence is suggestive that vitamin D and probably calcium may reduce the risk of this intermediate endpoint for colorectal cancer, but the available data are not sufficient to allow a definitive assessment of the effects of vitamin D, calcium, and their interactions on risk for new or recurrent colorectal adenomas.

Additional evidence from randomized controlled trials The committee did not identify any additional relevant RCTs assessing vitamin D or calcium intake and risk for colorectal cancer or adenomas.

Observational studies The European Prospective Investigation into Cancer and Nutrition (EPIC) study has recently reported data on more than 1,200 colorectal cancer cases and an equal number of controls (Jenab et al., 2010). In this report, serum concentrations lower than the pre-defined mid-level concentrations of 25OHD (50 to 75 nmol/L) were associated with higher colorectal cancer risk. Jenab et al. (2010) also reported that higher 25OHD concentrations of 75 to less than 100 nmol/L and 100 nmol/L and higher were associated with a decreased risk. No other relevant observational studies were identified outside the AHRQ reviews. Although this evidence was largely in agreement with the IARC (2008) findings and Bischoff-Ferrari et al. (2006a), the committee did not consider it convincing enough to outweigh the conclusions from both AHRQ reviews.

Concluding statement Taken in aggregate, epidemiological studies examining associations between vitamin D status and colorectal cancer incidence generally support an inverse association, although the shape of the dose–response relationship curve over a wide range of vitamin D intake remains very speculative. The biological plausibility is supported by data from cell culture and rodents, with additional support from surrogate biomarker studies in humans. There remains a paucity of prospective randomized intervention studies, and those available have not shown a significant relationship at this time. Thus, the data are insufficient for the committee to utilize colon cancer as an outcome for establishment of vitamin D DRIs. The data for an effect of dietary calcium on colorectal cancer risk are also highly suggestive of a protective effect, but there are not sufficient data available on dose–response relationships to utilize colorectal cancer as a health outcome for DRI development.

Prostate cancer risk is strongly associated with aging and is clearly dependent upon prolonged exposure to testosterone. Unlike breast cancer

in women, however, where specific reproductive events define risk, further characterization of the relationship has been challenging. Specific dietary and nutritional hypotheses, including a role for vitamin D and calcium, have been proposed but evidence supporting these relationships is not conclusive.

Biological plausibility Studies in vitro document that prostate cancer and prostate epithelial cells in culture respond to calcitriol with anti-proliferative effects and that calcitriol stimulates cell differentiation (Washington and Weigel, 2010). Evidence indicates that these effects, as for epithelial cells of other tissue origins, are mediated by the VDR expressed on prostate cells (Kivineva et al., 1998; Thorne and Campbell, 2008). Gene expression array studies provide evidence that calcitriol induces a pattern of gene expression that inhibits growth factor signaling and cell cycle progression, promotes differentiation, and is anti-inflammatory and anti-angiogenic (Krishnan et al., 2004; Peehl et al., 2004; Kovalenko et al., 2010). The role of dietary calcium intake in prostate cancer risk is less well studied, with inconsistent results, and the potential biological mechanisms of action are highly speculative.

Systematic reviews and meta-analyses The AHRQ-Tufts systematic review found no qualified systematic reviews assessing associations between serum 25OHD levels and incidence of prostate cancer. Among observational studies reviewed, 8 of 12 nested case–control studies found no association between baseline serum 25OHD levels and risk for prostate cancer, and only 1 (C-rated) (Ahonen et al., 2000) reported a significant association between baseline serum 25OHD levels below 30 nmol/L and higher risk of prostate cancer, compared to those with levels greater than 55 nmol/L. Further, the effect appeared to be stronger for men younger than age 52 at entry into the study. A meta-analysis by Huncharek et al. (2008) of 45 observational studies on dairy and milk intake and risk of prostate cancer showed no significant association between dietary intake of vitamin D and prostate cancer risk.

Additional evidence from randomized controlled trials No relevant RCTs that were not reviewed by AHRQ were identified for vitamin D or calcium intervention and risk for prostate cancer.

Observational studies Three observational studies not included in either AHRQ-Ottawa or AHRQ-Tufts were identified as potentially relevant to prostate cancer as a health indicator for vitamin D and calcium. Schwartz and Hulka (1990) suggested that vitamin D deficiency was a causative

factor in prostate cancer based upon the observation that the prevalence of vitamin D deficiency increases with age and is greater in those with dark-pigmented skin types and northern European populations, coupled with the observation that mortality rates for prostate cancer appear to be inversely related to sun exposure. However, a more recent case–control analysis of data from the Alpha-Tocopherol, Beta-Carotene Cancer Prevention (ATBC) Study, which examined male smokers, found no association; including any age-related associations to support a relationship between serum 25OHD levels and incidence of prostate cancer (Faupel-Badger et al., 2007). A potential procarcinogenic effect of higher dietary calcium was suggested by the HPFS (Giovannucci et al., 1998), which reported that calcium intake from the diet or from diet and supplements was independently associated with risk of locally advanced or metastatic prostate cancer especially when intakes exceeded 2,000 mg per day. The potential role for calcium as a risk factor for prostate cancer is discussed in detail in Chapter 6.

Because of the complexity of assessing vitamin D exposure over time relative to prostate cancer risk, high-quality evidence from observational studies was limited. The results of the HPFS, the only large, prospective cohort study identified for calcium, are not supported by evidence from available RCTs. Therefore, the evidence from human studies is insufficient to permit the committee to draw conclusions about a role for vitamin D and/or calcium in reducing prostate cancer risk.

Concluding statement Overall experimental data indicating that cultured prostate epithelial and prostate cancer cells respond to vitamin D via the VDR suggest a role for vitamin D in prostate cancer. However, associational studies of vitamin D status and risk of prostate cancer have provided mixed results, and randomized controlled clinical trials of substantial quality examining incidence or mortality have not been reported. Thus, there are insufficient data to permit the committee to draw a conclusion about the utility of the evidence for this indicator to support DRI development.

CVD broadly describes a range of diseases affecting the heart and blood vessels. Diseases that fall under the umbrella of CVD comprise coronary artery disease, myocardial infarction, stroke/cerebrovascular disease, peripheral artery disease, atherosclerosis, hypertension, arrhythmias, heart failure, and other vascular disorders. CVD is a public health concern because it is associated with an enormous burden of illness, disability, and mortality. CVD and hypertension were considered as potential indicators

based on proposed hypotheses that vitamin D alone or in combination with calcium may help to prevent CVD or hypertension. Calcium has also been implicated independently as a nutrient related to reducing risk for development of CVD. Limited data were available for this indicator in the 1997 DRI report (IOM, 1997); however, additional experimental animal and observational studies for both vitamin D and calcium and CVD have been published in the interim.

Vitamin D has been linked to decreased risk for CVD. Ecological studies suggest that there is higher cardiovascular mortality during the winter and in regions with less average exposure to UVB radiation from sunlight (Zittermann et al., 2005). Various biological mechanisms have been proposed in support of this hypothesis. Experimental animal studies failed to demonstrate an effect of vitamin D on risk for hypertension (Li et al., 2004) and increased thrombogenicity (Aihara et al., 2004). In rodents, administration of calcitriol or its analogues enhances vascular reactivity (Hatton et al., 1994). In support of the hypothesis that biological activation of vitamin D is relevant to cardiovascular function, the Vdr-null mouse has been used to model CVD.

High dietary calcium intake may help to reduce CVD risk through its roles in decreasing intestinal absorption of lipids and increasing lipid excretion, lowering blood cholesterol levels, and promoting calcium influx into cells.

The AHRQ-Tufts report identified one RCT and four relevant observational studies for vitamin D and cardiovascular outcomes. The RCT (Trivedi et al., 2003) found no statistically significant difference in incidence of cardiovascular events and deaths for subjects treated with 100,000 IU of vitamin D every 4 months over 5 years of follow-up. Among the observational studies reviewed, the Framingham Offspring Study found a significant association between low serum 25OHD levels and incident CVD (T. J. Wang et al., 2008). However, a closer look at the individuals with the highest serum 25OHD levels suggests that there was no additional reduction in risk at levels greater than 75 nmol/L and that the dose–response relationship may be U-shaped above 75 nmol/L. In the HPFS, Giovannucci et al. (2008), using a nested case–control design, found a significant association between low (< 37.5 nmol/L) serum 25OHD levels and incident myocardial infarction. A study using data from NHANES III, however,

found no significant association between serum 25OHD levels and cardiovascular mortality overall, although individuals with the lowest 25OHD levels experienced a significant increase in total mortality compared with those with the highest levels. Echoing the findings for incident CVD in the Framingham Offspring Study, a closer examination of the highest 25OHD levels suggested a U-shaped dose–response relationship, with increased total mortality at both the lowest and highest 25OHD levels in this cohort (Melamed et al., 2008). The fourth observational study reported in AHRQ-Tufts (Marniemi et al., 2005) also failed to find an association between serum 25OHD levels and total CVD incidence, although it did find that vitamin D intake predicted a decreased risk for stroke. With the exception of one case–control study, the overall findings from the observational studies reviewed reaffirm a lack of significant association between 25OHD level and CVD risk and that higher 25OHD levels may incur an increased risk for CVD. From this, AHRQ-Tufts concluded that the evidence was insufficient to support a relationship between vitamin D or calcium and risk for CVD.

A recent meta-analysis of randomized trials using calcium supplements (without vitamin D) suggested that calcium supplementation was associated with an increase in the risk of myocardial infarction (Bolland et al., 2010a). However, another recent meta-analysis that included CVD as a secondary outcome found a slightly reduced, but not significant, risk for CVD with vitamin D supplementation, no association with calcium supplementation, and no association with a combination of vitamin D plus calcium supplementation (Wang et al., 2010).

No new RCTs were identified that examined CVD as a pre-specified primary outcome although several trials analyzed CVD as a secondary treatment outcome, and the findings of secondary outcome studies were not supportive of a reduction in CVD risk for either vitamin D or calcium. Among the additional RCTs reviewed outside the AHRQ reviews examining CVD as a secondary outcome (Hsia et al., 2007 [400 IU vitamin D₃/1,000 mg calcium]; Major et al., 2007 [400 IU vitamin D/1,200 mg calcium]; Margolis et al., 2008 [400 IU vitamin D₃/1,000 mg calcium]; Prince et al., 2008 [1,000 IU vitamin D₂/1,000 mg calcium]; Manson et al., 2010 [400 IU vitamin D₃/1,000 mg calcium]), none found a significant treatment-related effect of vitamin D on risk of CVD (see Evidence Map in Appendix F). In a 5-year study of calcium intake and risk for CVD in New Zealand, Bolland et al. (2008) found that women taking 1,000 mg of elemental calcium had a significantly higher risk (compared to placebo) for myocardial infarction and a composite CVD endpoint of myocardial infarction, stroke, and sud-

den death. However, when unreported events identified from a national database were added to the analysis, the increased cardiovascular risks in the calcium group were no longer statistically significant. A bone density trial, also conducted in New Zealand, assessing self-reported composite vascular events among men was also not significant for an interaction between 1,000 mg of calcium daily and CVD outcomes (Reid et al., 2008). The results of these trials are in agreement with the null findings of the AHRQ-Tufts review described above. The additional clinical trials reviewed did not show a statistically significant causal relationship between either vitamin D or calcium and decreased cardiovascular risk, and reductions in risk that were noted in some trials were not well supported by data analyses. Therefore, the totality of the evidence does not support an interaction between either vitamin or calcium and risk for CVD. Adverse cardiovascular effects associated with excess calcium intake were also noted, and these are discussed further in Chapter 6.

In addition to the clinical trials reviewed, including those from AHRQ, several observational studies were identified that examined a role for vitamin D and/or calcium in reducing CVD risk. Two large, prospective cohort studies were identified. In one study of individuals at high risk of CVD, among coronary angiography patients followed for more than 7 years, those with the lowest serum 25OHD levels had significantly higher total mortality and cardiovascular mortality compared with those with the highest levels (Dobnig et al., 2008). Melamed et al. (2008) assessed 25OHD levels and prevalence of peripheral artery disease using data from NHANES 2001 to 2004. This study found a graded association between levels of 25OHD up to 29.1 nmol/L and levels of 29.2 nmol/L and above. In a trend analysis, a statistically significant difference was found between the lower 25OHD levels compared with the higher levels.

A number of small cohort studies were identified that evaluated serum 25OHD or calcitriol levels in patients at risk for various CVD indicators compared with control subjects who were free of CVD indicators. Watson et al. (1997) assessed calcitriol levels in subjects at high risk for developing coronary heart disease compared with asymptomatic individuals and found a significant inverse association between calcitriol and amount of vascular calcification in both groups, although the difference was greater in the at-risk group. Poole et al. (2006) compared serum 25OHD levels in a small group of patients admitted for a first stroke with those of healthy controls and found that serum 25OHD levels were significantly lower among stroke patients. Zittermann et al. (2003) compared both 25OHD and calcitriol levels against serum levels of biomarkers indicative of congestive heart

failure in a small group of patients admitted for treatment and in free-living controls. The study found a significant difference in biomarker levels between treated patients compared with controls for both 25OHD and calcitriol levels.

One small case–control study was identified that determined the relationship between serum 25OHD levels and risk for myocardial infarction in at-risk patients compared with normal controls. In this study, Scragg et al. (1990) found that serum 25OHD levels were significantly lower in myocardial infarction cases than in controls and that the difference was greater (but not significantly so) during the winter.

Although these studies together provide evidence for lower serum 25OHD levels in individuals with CVD, whether the low serum 25OHD levels are sufficient to predict risk for CVD has not been clearly established. Additional evidence indicates that low serum 25OHD levels are associated with risk factors for CVD—specifically, increased carotid arterial thickness (Targher et al., 2006)—and apparent CVD in patients with type 2 diabetes (Cigolini et al., 2006; Chonchol et al., 2008). Additionally, some studies suggest a positive association between vitamin D intake and CVD risk factors associated with other chronic conditions, including hypertension (Krause et al., 1998; Pfeifer et al., 2001; Forman et al., 2007; L. Wang et al., 2008; Wang et al., 2010), impaired glucose tolerance or type 2 diabetes (Liu et al., 2005; Pittas et al., 2006, 2007a; Mattila et al., 2007), and inflammation (Timms et al., 2002; Schleithoff et al., 2006; Shea et al., 2008).

Risk of incident hypertension in relation to dietary vitamin D intake has been evaluated in three large prospective study cohorts; NHS 1, NHS 2, and the HPFS for 8 years and longer. Women in NHS 1 and NHS 2 (a younger cohort) showed no association between vitamin D intake and risk for incident hypertension. Likewise, among men from the HPFS no association was found between vitamin D intake and risk for incident hypertension. Al-Delaimy et al. (2003) also found no association between calcium intake, vitamin D intake, or total dairy intake and risk for total ischemic heart disease in men enrolled in the HPFS. Similarly, no association was found when the cohort was analyzed for calcium supplement intake, although an inverse association was identified between calcium intake among supplement users compared with nonusers and fatal ischemic heart disease only.

In contrast to the intake studies, in a prospective study, Forman et al. (2007) found inverse associations between incidence of hypertension and measured serum 25OHD levels in a larger cohort in the HPFS and in women from a larger cohort in NHS.

In summary, three of four large, prospective cohort studies reviewed found associations between serum 25OHD levels and risk for CVD. Among the many smaller observational studies of lower quality that were identified,

most did not find a significant positive association between vitamin D and calcium intake and risk for CVD. Taken together, this observational evidence was strong enough to support a relationship between serum 25OHD levels and incident disease, but not a conclusion that higher serum 25OHD levels were associated with a lower risk for CVD. Additionally, the review of randomized trial evidence does not support a causal relationship between vitamin D intake and risk for CVD.

Review of the available evidence, from both RCTs and observational studies on associations between vitamin D and calcium intake and risk for CVD shows that although observational evidence supports a relationship between serum 25OHD levels and the presence of CVD, it does not show a relationship with risk for developing CVD, and evidence was not found for a causal relationship between vitamin D intake and development of disease. Given the lack of statistically significant evidence supporting associations between vitamin D intake or serum 25OHD level and risk for CVD and the lack of evidence on CVD as a primary outcome of treatment in RCTs with vitamin D and/or calcium, the committee could not draw an inference about the efficacy of this indicator to support DRI development.

Type 2 diabetes is a blood glucose disorder characterized by insulin resistance and relative insulin deficiency. Metabolic changes that accompany chronic elevated blood glucose levels frequently lead to functional impairment in many organ systems, particularly the cardiovascular system, which contributes to substantially increased risk of morbidity and mortality.

Metabolic syndrome is a condition characterized by a constellation of metabolic risk factors, including abdominal obesity, atherogenic dyslipidemias, elevated blood pressure, insulin resistance, prothrombotic state, and proinflammatory state (e.g., elevated C-reactive protein).

Individuals with metabolic syndrome are at increased risk of coronary heart disease, stroke, peripheral vascular disease, and type 2 diabetes. Adiposity is a component of both type 2 diabetes and metabolic syndrome, which may have an impact on vitamin D status. Since the release of the 1997 DRIs (IOM, 1997), a number of studies have been published on relationships between vitamin D with or without calcium and type 2 diabetes and metabolic syndrome. The committee recognized that obesity can be a confounder to vitamin D analysis. However, as it is a component of the health outcome and because of the prevalence of both obesity and meta-

bolic syndrome in the general population, this indicator was considered as a candidate for DRI development.

Vitamin D was first implicated as a modulator of pancreatic endocrine function and insulin synthesis and secretion in studies using rodent models more than three decades ago (Norman et al., 1980; Clark et al., 1981; Chertow et al., 1983). Since then, the role of calcitriol in the synthesis and secretion of insulin and regulation of calcium trafficking in β-islet cells as well as its effects on insulin action have been established in both rodent models and in vitro cell culture models (Frankel et al., 1985; Cade and Norman, 1986; Faure et al., 1991; Sergeev and Rhoten, 1995; Billaudel et al., 1998; Bourlon et al., 1999). These findings stimulated observational and intervention studies examining the role of vitamin D and calcium in type 2 diabetes and metabolic syndrome in humans.

Neither AHRQ-Ottawa nor AHRQ-Tufts included type 2 diabetes or metabolic syndrome in its systematic review, although AHRQ-Tufts did include body weight as a health outcome and found no effect of vitamin D or calcium on changes in body weight. A systematic review and meta-analysis by Pittas et al. (2007b) included a large body of observational evidence and six intervention studies (four small short-term and two long-term studies) of vitamin D supplementation, one study using combined vitamin D and calcium supplementation and five studies using calcium alone or dairy supplementation. The results from these trials were largely negative; among the short-duration vitamin D trials, three studies reported no effect, and one reported enhanced insulin secretion but no improvement in glucose tolerance following vitamin D supplementation. In one study included in the review, however, the relationship was statistically significant only when non-Hispanic blacks were excluded from the meta-analysis.

Overall, the evidence reviewed from the intervention studies did not support a role for vitamin D alone, although vitamin D in combination with calcium supplementation may have a role in preventing type 2 diabetes in populations already at risk. The observational evidence in the review included cross–sectional and case–control studies in which serum vitamin D and calcium levels were determined from individuals in a population with established glucose intolerance. Similar confounding and a lack of adjustment for confounders limited the cohort studies. Thus, the one meta-analysis that included both observational and intervention studies could

not be considered as supportive for a relationship between either vitamin D or calcium and the health outcomes of diabetes or metabolic syndrome.

Two randomized trials were identified that evaluated the effect of vitamin D supplementation with or without supplemental calcium on markers of glucose tolerance as a primary outcome and four additional trials were identified that evaluated glucose metabolism as a secondary outcome. A trial in New Zealand that examined the effect of supplementation with 4,000 IU of vitamin D₃ per day for 6 months on insulin resistance in non-diabetic overweight South Asian women found a significant improvement in insulin sensitivity compared with those in the placebo group after 6 months (von Hurst et al., 2010). Among women who had low serum 25OHD levels at the beginning of the study, those who achieved a serum 25OHD level above 80 nmol/L at 6 months had significant improvement in insulin sensitivity. In contrast, sub-analysis of data from the Randomised Evaluation of Calcium and/Or vitamin D (RECORD) trial examining the association between incidence of self-reported development of type 2 diabetes or initiation of treatment for type 2 diabetes and supplementation with 800 IU of vitamin D₃ and 1,000 mg of calcium in an elderly population found no association (Avenell et al., 2009a). Zittermann et al. (2009), in a weight loss trial evaluating the effect of supplemental vitamin D on markers of CVD in overweight adults as a primary outcome, found no significant difference for an effect on glucose metabolism. Jorde et al. (2010), in a 1-year trial in Norway with overweight or obese subjects, found no change in measures of blood glucose in vitamin D–supplemented subjects compared with control subjects, but they did identify an unexpected and significant increase in systolic blood pressure in the supplemented group compared with controls. Without further analysis, however, it is not possible to determine whether the increase in blood pressure was related to 25OHD levels in blood. A trial in India evaluated the effect of short-term vitamin D supplementation on homeostasis model assessment and oral glucose insulin sensitivity in healthy, centrally obese men (Nagpal et al., 2009). In an intention-to-treat analysis, the difference was not significant. Overall, higher waist-to-hip ratios and lower baseline serum 25OHD levels were significant predictors of improvement in oral glucose insulin sensitivity. A posthoc analysis of a trial testing the effects of long-term supplementation with 700 IU of vitamin D and 500 mg of calcium daily on health, including associations between combined supplementation and changes in fasting glucose levels, found that subjects with impaired fasting glucose who followed the supplementation regimen for 3 years had a significantly lower rise in fasting glucose levels and less insulin resistance compared with

placebo controls (Pittas et al., 2007a). Although the findings of this study are in agreement with a previous secondary analysis of data from the NHS cohort (see below: Pittas et al., 2006), the study is limited by the small number of outcomes measured compared with the total cohort; thus, an unintended bias cannot be ruled out. In addition, the study was designed for skeletal outcomes as the primary analysis. When the totality of the evidence was considered, the negative findings from the clinical trials for an effect of vitamin D or calcium on risk for type 2 diabetes together with the lack of significant evidence from either the AHRQ reviews or the meta-analysis by Pittas et al. (2007b) compelled the committee to conclude that there was not sufficient evidence to establish a causal relationship.

Low serum 25OHD levels have been implicated in metabolic syndrome, abdominal obesity, and hyperglycemia.

In a prospective cohort analysis of data from NHS, women were followed for 20 years to examine associations between vitamin D and calcium intake and risk for type 2 diabetes (Pittas et al., 2006). A significant inverse association was found between total vitamin D intake and calcium intake and risk for type 2 diabetes. A separate analysis of the association between risk for type 2 diabetes and dairy food consumption found that women who consumed three or more dairy servings per day were at lower risk compared with those who consumed less than one dairy serving per day. These findings suggest that risk for type 2 diabetes is associated with vitamin D or dairy food intake. A small cohort study in obese and overweight individuals found that in addition to a significant inverse association between serum 25OHD level and weight and waist circumference there was a weak inverse relationship with hemoglobin A1c. However, no association between serum 25OHD level and any other indicators of type 2 diabetes or metabolic syndrome were observed (McGill et al., 2008).

In other observational evidence reviewed, a cross–sectional survey of Polynesian and white adult populations in New Zealand found a significantly lower serum 25OHD level in subjects with newly diagnosed diabetes and impaired glucose tolerance compared with controls. In addition, among the control groups, the native New Zealand populations (Maori and Pacific Islanders) were found to have significantly lower serum 25OHD levels compared with Europeans. The authors speculated that the low serum 25OHD level in the native populations explained, in part, the higher prevalence of diabetes in those groups (Scragg et al., 1995). Isaia et al. (2001), in a cross–sectional study in Italy, found that postmenopausal women diagnosed with type 2 diabetes had significantly higher body mass indexes (BMIs), lower activity scores, higher prevalence of serum 25OHD

levels below 12.5 nmol/L, and lower dietary calcium intake compared with controls. In summary, these observational studies fail to provide conclusive support of a relationship between vitamin D intake and risk for either type 2 diabetes or metabolic syndrome because of the lack of consistency among studies, the paucity of high-quality large cohort studies, and the lack of strength for an association between vitamin D status and incidence of type 2 diabetes or metabolic syndrome.

The available evidence from observational studies of the associations between vitamin D and calcium and risk for type 2 diabetes or metabolic syndrome and secondary analyses from RCTs on markers of glucose tolerance proved insufficiently strong to support DRI development. The association studies linking lower serum 25OHD levels to increased risk for type 2 diabetes may be confounded by overweight and obesity, which not only predispose individuals to type 2 diabetes, but also cause lower serum 25OHD levels as a result of sequestration in fat and possibly other mechanisms. Although both retrospective and prospective studies tend to support an inverse association between serum 25OHD levels and type 2 diabetes, these studies are limited by the study design and cannot show a causal relationship. Evidence from RCTs on the effect of vitamin D supplements on incident diabetes or markers of glucose homeostasis is variable, and few RCTs showing significant results were identified. Taken together, the evidence in support of a role for vitamin D as a modulator of pancreatic endocrine function and insulin synthesis and secretion is not conclusive and therefore is not sufficient to support glucose tolerance as an indicator for DRI development.

The committee considered falls and physical performance as independent indicators. However, because of the integration of these indicators in the literature reviewed by the committee, the evidence for both indicators is examined together in this section.

The risk of falling is a major concern among the elderly, because falls can lead to fracture and long-term disability or death in this population. Vitamin D is necessary for normal development and growth of muscle fibers, and vitamin D deficiency may adversely affect muscle strength. Muscle weakness and pain (myopathy) are characteristics of rickets and osteomalacia and contribute to poor physical performance (Prineas et al., 1965; Skaria et al., 1975; Yoshikawa et al., 1979). Thus vitamin D-deficiency muscle weakness and the implications of poor muscle tone suggest a re-

lationship between serum 25OHD level and risk for falling and/or poor physical performance in susceptible populations.

Experimental evidence suggests that vitamin D exerts its effect on muscle tissue via the VDR, but it may also use other pathways. In vitro and in vivo experiments provide evidence to support calcitriol regulation of calcium uptake by muscle, which, in turn, controls muscle contraction and relaxation, synthesis of muscle cytoskeletal proteins involved in muscle contraction, and muscle cell proliferation and differentiation (reviewed in Ceglia, 2008). Because intracellular calcium levels control the contraction and relaxation of muscle, thus affecting muscle function, it is possible that calcium intake may also affect risk for falls and poor physical performance (reviewed in Ceglia, 2008). However, the topic is not considered in more detail here because of the lack of observational and RCT data on the relationship between calcium intake and physical performance.

The AHRQ-Ottawa systematic review identified a total of 14 RCTs in addition to five prospective cohort studies and one case–control study that examined vitamin D and risk for falls in postmenopausal women and elderly men. The evidence between the RCTs and observational studies was discordant. Overall the review reported that the evidence for an association between low serum 25OHD levels and risk of falls and measures of physical performance among postmenopausal women and elderly men was inconsistent and rated the evidence as “fair.” The AHRQ-Tufts systematic review identified three additional RCTs (Bunout et al., 2006; Burleigh et al., 2007; Lyons et al., 2007), but these studies did not find a significant effect of vitamin D supplementation on reducing risk of falls or poor performance in the elderly and were given a “C” rating. No additional observational evidence was found for this indicator in the AHRQ-Tufts review.

A meta-analysis reported in AHRQ-Tufts, which included the AHRQ-Ottawa RCTs, highlighted the inconsistency of findings from RCTs on the effect of vitamin D treatment on reduction in risk or prevention of falls. A smaller meta-analysis by Bischoff-Ferrari et al. (2004a) examined RCTs in elderly populations for evidence of a reduction in risk for falls with “vitamin D”; however, only three studies used vitamin D, and the other two studies used calcitriol/1α-hydroxycholecalciferol. Some of the studies identified in this meta-analysis were also included in the AHRQ-Tuft analysis. In contrast to the AHRQ-Tufts analysis, Bischoff-Ferrari et al. (2004a) found, from pooled results, a significant reduction in risk of falling among sub-

jects treated with vitamin D compared with those treated with calcium or placebo. This disparity in findings is best explained by the small numbers in the Bischoff-Ferrari et al. (2004a) analysis and the fact that none of the vitamin D studies pooled by Bischoff-Ferrari et al. (2004a) was individually significant.

A meta-analysis published in 2009 by Bischoff-Ferrari et al. (2009a) examined fall prevention based on supplemental intake and serum 25OHD concentrations. From this analysis of the eight RCTs (n = 2,426 subjects) that met the inclusion criteria, the authors concluded that supplemental vitamin D intake (700 to 1,000 IU/day) reduced the risk of falling among older subjects by 19 percent and that serum 25OHD concentrations less than 60 nmol/L may not reduce the risk of falling. This meta-analysis as conducted has major limitations.

First, the stated inclusion/exclusion criteria and their application are problematic. As stated by the authors, to be included in the primary analysis, the trial design had to be double-blinded and the assessment of falls had to be a primary or secondary endpoint defined at the onset of the trial. The study had to include a definition of falls and how they were assessed, and falls had to be assessed for the entire trial period. Studies using patients with Parkinson’s disease, organ transplant recipients, or stroke patients were excluded as were trials using intramuscular injection of vitamin D. Of concern is the fact that some studies that met the inclusion/exclusion criteria were omitted, and at least one study that failed to meet the criteria was included. The Broe et al. (2007) study, did not have a secondary analysis that pre-specified falls as an outcome, was never powered to examine the incidence of falls; with the wide confidence interval due to small sample size the results are questionable. This study influenced the analysis considerably; other than the work of Pfeifer et al. (2009), it was the single largest contributor to the effect. The work of Law et al. (2006) was excluded because it was a cluster randomization design instead of individual randomization; however, such a design does not appear to violate the authors’ stated criteria. It was also excluded because the dose of oral vitamin D (50,000 IU) was given every 3 months; however, the serum 25OHD increased from 45 to 75 nmol/L, indicating an adequate therapeutic level. Had the Law et al. (2006) study been included, in which 44 percent of the vitamin D–treated group and 43 percent of the control group were fallers (not significantly different), the overall results would have been negative.

Second and more importantly, Figure 3 as reported in Bischoff-Ferrari et al. (2009a) is inappropriately presented. The figure is intended to demonstrate fall prevention with dose of vitamin D and achieved serum 25OHD concentrations. Specifically, the figure is a meta-regression analysis of the relative risk (RR) against vitamin D dose or serum 25OHD concentration. However, the meta-regression appears to be incorrectly carried out, or the

authors used assumptions that were not specified in the methods section of their publication. In their analysis, the dependent variable appearing in the graph is RR (linear scale of 0 to 2.5); however, log(RR) is typically used in this type of meta-regression, which is a weighted linear regression with each study being the unit of analysis. Even when RR is to be reported, a meta-regression of log(RR) against the predictor variable should be carried out and then retransformed back to the RR scale, in which case the line will be curvilinear instead of straight. Carrying out a meta-regression analysis using untransformed RR in the linear scale assumes an exponential relationship of the dose with effect. Moreover, the predictor variable (x-axis) is equally spaced for data points, but the data points are not equally spaced according to the vitamin D doses (or serum 25OHD concentrations). In the top panel of Figure 3 in Bischoff-Ferrari et al. (2009a), the dose intervals between the data points range from 0 IU (two 400 IU studies; two 800 IU studies) to 100 IU (between 600 IU and 700 IU) to 200 IU (between 200 IU and 400 IU, 400 IU and 600 IU, 800 IU and 1,000 IU). Two data points were also composed of multiple trials “collapsed” into a single data point for two levels (800 IU of vitamin D₃; 1,000 IU of vitamin D₂) of the predictor variables. This introduces considerable uncertainty as to the appropriateness of the location of the regression line. If the measurement intervals had been appropriately and evenly spaced, it is very likely that the conclusion of the analysis would have been that no significant relationship was demonstrated.

The importance of the limitations of this study becomes clear when the data are reanalyzed in the appropriate statistical manner. As shown in Figures 4-2 and 4-3, no dose–response relationship between vitamin D intake and risk of falls is evident. For this analysis, which used the STATA program, analyses were repeated by fitting a random effects meta-regression with the log(RR) of sustaining at least one fall as the response variable and the daily dose of vitamin D supplementation or the mean achieved 25OHD serum concentration in the vitamin D supplementation arm as the predictor variable (both predictor variables are continuous variables). Specifically, the results do not show a significant dose–response relationship between the risk of sustaining at least one fall and the daily dose of vitamin D supplementation or achieved 25OHD serum concentration (beta coefficient = −0.0005 ± 0.0003 and = −0.0087 ± 0.0056 standard error [SE] respectively; relative risk reduction = 0.95 for risk of falls per 100 IU/day and increased in dose of vitamin D, p = 0.13; relative risk reduction = 0.92 for risk of falls for every 10 nmol/L increase in 25OHD level, p = 0.17). Both analyses had significant heterogeneity across studies (I² = 47 percent, p = 0.05; I² = 54 percent, p = 0.03, respectively). Further, a non-linear dose–response relationship was explored by adding a quadratic term of the predictor variable to the model. The result suggests that a U-shaped curve

FIGURE 4-2 Relative risk of falls and vitamin D supplementation doses: Correct meta-regressions with continuous predictors showing non-significance.

NOTE: Relative risk reduction is 0.95 (95% confidence interval [CI] 0.89 to 1.02; p = 0.13) per 100 IU/day difference (increase) in dose.

better describes the relationship between the risk of sustaining at least one fall and the achieved serum 25OHD concentrations.

As discussed above, among RCTs that tested for effects of vitamin D with and without calcium on reduction in risk for falls, no consistent outcome was found. As described above, the data related to falls are questionable, and among muscle performance studies one included 12 subjects who were post-stroke patients and a second included 16 subjects for whom the study was only 8 weeks in length.

Two recent studies published after the AHRQ-Tufts analysis was completed have failed to show efficacy in reducing falls. A randomized but not placebo-controlled trial examined the effect of either 800 or 2,000 IU of vitamin D per day combined with enhanced or standard physiotherapy on the rate of falls and hospital re-admission following hip fracture in free-living adults with a mean age of 84 years (Bischoff-Ferrari et al., 2010). Neither of the two dosages of vitamin D₃ reduced the rate of falls or improved

FIGURE 4-3 Relative risk of falls and mean achieved serum 25OHD concentrations: Correct meta-regressions with continuous predictors showing non-significance.

NOTE: Relative risk reduction is 0.92 (95% confidence interval [CI] 0.80 to 1.05; p = 0.17) per 10 nmol/L difference (increase) in mean achieved 25OHD concentration.

strength or function compared with physiotherapy. Another study (Sanders et al., 2010) that examined the incidence of falls and fractures in elderly women treated with 500,000 IU of vitamin D₃ annually for 3 years found a significant increase in falls and fractures in the treatment group compared with the placebo group. Notably, the increased incidence of falls was significant in the treatment group by 3 months following administration of the supplemental vitamin D. Further, as described in Chapter 6, the authors of this study concluded that levels of 65 nmol/L were not consistent with reduced rates of fall or fractures.

When this committee considered the totality of evidence for causality pertinent to the relationship between vitamin D and incidence of or risk for falls, it became clear that the greater part of the causal evidence indicated no significant reduction in fall risk related to vitamin D intake or achieved level in blood. Table 4-4 illustrates the range of clinical trial data assessing changes in fall incidence or risk for falls with varying levels of vitamin D treatment that were taken into account. Of the 18 studies considered, including several studies identified in Bischoff-Ferrari et al. (2009a), only 4 (Pfeifer et al., 2000; Harwood et al., 2004; Flicker et al.,

2005; Broe et al., 2007) found a significant effect of vitamin D on fall incidence. The only two significant studies for fallers are Pfeifer et al. (2000, 2009), although Pfeifer et al. (2000) was a 2-month study and administered calcium with the vitamin D placebo.

Observational studies have long suggested an association between a higher serum 25OHD level and a lower risk of falls in elderly persons; however, when analyzed as a whole in the AHRQ reviews, there was no consistency between study findings. Snijder et al. (2006), a study of elderly subjects participating in the Longitudinal Aging Study Amsterdam, a prospective cohort study, was not included in the AHRQ reviews. This study found that a low serum 25OHD level (< 25 nmol/L) was independently associated with an increased risk of falling for subjects who experienced two or more falls compared with those who did not fall or fell once; however the study outcome does not affect the discordant findings among observational studies identified in the AHRQ reviews.

Most observational studies of associations between serum 25OHD levels and physical performance have been cross–sectional, which limits causal inference. A cross–sectional study of 4,100 older adults from NHANES III found higher serum 25OHD concentrations associated with better lower-extremity function (Bischoff-Ferrari et al., 2004b). Much of the improvement occurred at concentrations ranging from 22.5 nmol/L to approximately 40 nmol/L, but some improvement was also seen from 40 to 94 nmol/L (the top of the reference range). Results were similar in men and women, three racial/ethnic groups (whites, African Americans, and Mexican Americans), active and inactive persons, and those with high and low calcium intakes. A study of Dutch adults 65 years of age and older found that serum 25OHD concentrations below 20 nmol/L were significantly associated with poorer physical performance at baseline and a greater decline in physical performance over a 3-year period (Wicherts et al., 2007). Another cross–sectional study of healthy post-menopausal women found that serum 25OHD level was significantly associated with physical fitness indexes, including balance, handgrip strength, androidal fat mass, and lean mass (Stewart et al., 2009). Finally, a cross–sectional study of 60 men and women with heart failure (mean age of 77 years) found a significant association between serum 25OHD level and 6-minute walk distance and frailty status (Boxer et al., 2008). Taken together, however, this evidence is weakened by the cross–sectional study design, does not provide strong support for an association between serum 25OHD level and physical performance, and does not contradict the findings of the AHRQ reviews.

A problem in a number of the RCTs is that falls rather than fallers are analyzed; consequently, individuals who fell more than once were also counted more than once in the primary outcome analysis. The studies generally did not have the statistical power to detect a significant difference in the number of fallers but relied on repeat fallers to achieve the desired number of total falls. Moreover, the meta-analyses described above combined data from the few trials in which fallers were counted with data from trials in which falls were the outcome. By comparison, the U.S. Food and Drug Administration (FDA) mandated primary outcomes in osteoporosis and cardiovascular trials are the number of individuals with fractures or cardiovascular events rather than the number of events. It remains uncertain whether a reduction in the number of falls can be used to infer that the number of fallers would be significantly reduced.

The committee’s review of the available evidence, including the results from RCTs and observational associations between vitamin D with or without calcium and risk for falls and poor physical performance, indicates a lack of sufficiently strong evidence to support DRI development. A limited review of observational data outside of the AHRQ reviews found some support for an association between 25OHD levels and physical performance. However, high-quality observational evidence from large cohort studies was lacking. Additionally, although the cross–sectional studies were more supportive of an association between high serum 25OHD levels and reduced risk for falls, evidence from RCTs in particular showed outcomes that varied in significance and thus did not support the observational findings or a causal relationship. The evidence was also not consistently supportive for a role for vitamin D combined with calcium in reduction of risk for falls.

Overall, data from RCTs suggest that vitamin D dosages of at least 800 IU/day, either alone or in combination with calcium, may confer benefits for physical performance measures. Although high doses of vitamin D (i.e., ≥ 800 IU/day) appear to provide greater benefit for physical performance than low doses (i.e., 400 IU/day), evidence is insufficient to define the shape of the dose–response curve for higher levels of intake. Thus, the outcome of physical performance is appropriate for identifying Estimated Average Requirements (EARs) of vitamin D, with or without calcium, in adults above the age of 50, but cannot be used to define the shape of the dose–response curve at higher levels of intake.

Vitamin D has been reported to modulate immune functioning in cell culture and animal models. Vitamin D, specifically its active form, calcitriol,

is a regulator of both adaptive and innate immune responses. However, its role is complex and not fully understood. Many factors influence the specific effect of vitamin D on immune function including its target cells; the nature of the immune challenge and response (either autoimmune or anti-infective); the status and availability of calcium, depending on the tissue or cell type; the physiological, differentiated or activated stage of the tissue or cell type; and the expression and polymorphisms of the genes for the Vdr and 1α-hydroxylase.

Asthma is a chronic lung disease that manifests as inflammation in bronchial tissue. The disease is characterized by recurrent periods of wheezing, chest tightness, shortness of breath, and coughing and may be accompanied by comorbidities, such as eczema or atopic dermatitis. Diet has long been linked to asthma and allergic disease. Dietary sodium and magnesium intakes were implicated as risk factors for asthma in the 1980s and 1990s (Burney, 1987; Britton et al., 1994a, b). Dietary lipids have also been hypothesized to contribute to increased prevalence of asthma (Black and Sharpe, 1997). More recently, vitamin D has been linked to asthma incidence in the developing fetus and in young children (Litonjua and Weiss, 2007).

Biological plausibility Genetic studies mapping the Vdr gene in animal models of asthma suggest that Vdr polymorphism is linked with expression of asthma. In humans, Poon et al. (2004) compared Vdr genetic variants between members of a family-based cohort (223 families of 1,139 individuals) with and without asthma. Their analysis found significant associations between six polymorphisms in the Vdr gene and clinical diagnosis of asthma. Wjst (2005) conducted genotyping on 951 individuals from pedigrees that had at least two asthmatic children to determine whether transmission of Vdr polymorphism was associated with asthma in the children. Preferential transmission of candidate polymorphisms in asthmatic children could not be confirmed; however, the authors did hypothesize the possibility of transmitting a protective effect to unaffected offspring based on their finding of a low probability of an unaffected phenotype in an affected cohort.

Systematic reviews and meta-analyses No systematic reviews or meta-analyses were identified for this indicator.

Additional evidence from randomized controlled trials No RCTs were identified for this indicator.

Observational studies A few studies were identified that examined a genetic linkage between vitamin D and risk for asthma or related conditions, and these were discussed above. Additional observational studies examined the relationship between perinatal serum 25OHD levels and risk of asthma in offspring. Devereux et al. (2007) examined associations between vitamin D intake during pregnancy and risk for childhood wheezing in a large prospective cohort study; a significant inverse association was found between maternal intake of vitamin D from diet and supplements and symptoms of wheezing at 2 and 5 years of age, although there was no significant association with diagnosed asthma at 5 years of age. In another study, Camargo et al. (2007) assessed the relationship between maternal dietary intake of vitamin D during pregnancy and risk of wheezing in children in Project Viva, a large prospective cohort study examining prenatal factors and pregnancy and child health outcomes. Overall, higher maternal intake of vitamin D during pregnancy was significantly associated with a lower risk for recurrent wheezing in the offspring at 3 years of age when compared with the lowest maternal vitamin D intake. Other associated symptoms of respiratory infection and eczema, however, were not significantly associated with maternal vitamin D intake. Similarly, Hypponen et al. (2004) found in a large prospective cohort study in Finland, that the prevalence of atopy and allergic rhinitis in subjects at age 31 years was higher among those who received regular vitamin D supplementation as infants than among those who did not; however, this study relied on retrospective recall of supplementation by the mother.

These large prospective cohort studies support an association between maternal or infant vitamin D intake and risk related to symptoms of asthma, particularly wheezing, but not with diagnosed disease. Several other observational studies have examined associations between 25OHD level in blood and risk for asthma. Gale et al. (2008), in a small prospective cohort study, found a five times increased risk for asthma at 9 years among children whose mothers’ serum 25OHD level was below 27.5 nmol/L, compared with those whose mothers had levels above 75 nmol/L. The small size of the cohort that was followed to 9 years of age (178 subjects) was a limitation to the reported finding. In contrast, a larger analysis of NHANES data found no association between serum 25OHD level and sensitization to allergens (Wjst and Hypponen, 2007). The study did identify an increased prevalence of allergic rhinitis across levels of 25OHD, although unrecognized confounding may account for the association.

In a cross–sectional study examining associations between vitamin D status and markers of allergic or asthmatic response, Brehm et al. (2009) found that serum vitamin D levels below 75 nmol/L were identified in 28 percent of children and that inflammatory markers (immunoglobulin E

[IgE] and eosinophil count) were significantly and inversely associated with vitamin D status. These lower quality observational studies largely support the associations with symptoms of asthma identified in the larger cohort studies but do not support an association between 25OHD level in blood and diagnosed asthma.

Although genetic studies support a possible biological mechanism for a functional role of vitamin D in development of asthma there are no RCTs to demonstrate a causal role.

Autoimmune diseases such as multiple sclerosis (MS), rheumatoid arthritis (RA), inflammatory bowel disease (IBD), and lupus are characterized by abnormal T cell response to self, resulting in inflammatory reactions in peripheral tissues. Models of autoimmune diseases support a role for vitamin D in regulating the T helper 1 (Th1) immune response, an integral component of immune tolerance with regard to recognition of self (reviewed in Cantorna and Mahon, 2005; Szodoray et al., 2008). Recent genomic analyses for polymorphisms in the Vdr gene suggest that single nucleotide polymorphisms identified in individuals with type 1 diabetes could negatively modulate calcitriol synthesis and thereby play a detrimental role in autoimmune response and subsequent manifestation of the disease (Israni et al., 2009).

Diabetes (type 1) Type 1 diabetes is a chronic disease resulting from loss of β-cell function in the pancreas. The disease is characterized by diminished or absent insulin production and loss of control of blood glucose. Emerging evidence for an association between low vitamin D status and increased risk for type 1 diabetes comes from experimental animal, ecological, and observational studies; however, no intervention trials using supplemental vitamin D (not analogues) were identified to provide causal support for a relationship.

Biological plausibility Experimental animal, ecological, and observational evidence support a relationship between vitamin D status and risk for type 1 diabetes, although treatment protocols and dosages vary. Ecological evidence has suggested a link between type 1 diabetes risk and limited erythemal UVB exposure in Newfoundland (Sloka et al., 2009, 2010). Mohr et al. (2008) plotted incidence rates for type 1 diabetes by latitude in an ecologic study comparing geographical distribution, estimated UVB exposure and disease incidence and, using a polynomial analysis to best fit the data points, determined that the incidence of type 1 diabetes was greater at higher latitudes.

In nonobese diabetic (NOD) mice, which are genetically predisposed to develop insulitis and type 1 diabetes, disease developed earlier when the mice were fed vitamin D–deficient diet and reared in the absence of UV light (Giulietti et al., 2004). However, type 1 diabetes was not prevented when NOD mice were treated with a supraphysiological dose of vitamin D, beginning from conception and continuing to 10 weeks of age (Hawa et al., 2004). Additionally, in a study in which NOD mice were cross bred with mice null for the Vdr gene, the rate of disease presentation did not differ from that in mice carrying only the NOD mutation, even though immune abnormalities were aggravated by the absence of the VDR (Gysemans, 2008). These results indicate that severe vitamin D and UV deficiency can increase the risk of type I diabetes in a genetically predisposed animal, yet neither vitamin D nor the absence of the Vdr gene affects the onset of type 1 diabetes.

Systematic reviews and meta-analyses Neither AHRQ-Ottawa nor AHRQ-Tufts included type 1 diabetes as a health outcome in its systematic reviews. Another recent systematic review and meta-analysis of five observational, four case–control and one cohort study (no RCTs were found) assessed whether vitamin D supplementation of infants reduced risk for type 1 diabetes later in life (Zipitis and Akobeng, 2008). The meta-analysis of data from the four case–control studies revealed a significant 29 percent reduction in risk for type 1 diabetes among vitamin D–supplemented infants compared with controls, which was further supported by the cohort study. The authors also cited evidence for a dose–response effect based on studies indicating reduced likelihood for developing diabetes among subjects who received regular vitamin D supplements, whereas subjects who developed rickets early in life were more likely to develop diabetes. A limitation of this meta-analysis is that two of the studies included had study designs that relied on delayed retrospective recalls by the mothers of vitamin D–supplemented infants. Additionally, no other meta-analyses were identified that either support or refute the findings of Zipitis and Akobeng (2008).

Additional evidence from randomized controlled trials No RCTs were identified for this indicator.

Observational studies No additional observational evidence that was not included in the systematic reviews and meta-analysis was identified for consideration.

Inflammatory bowel and Crohn’s disease IBD is a group of conditions of chronic inflammation that usually involve the distal portion of the ileum. In Crohn’s disease, inflammation spreads to the colon and upper gastrointestinal tract and causes local abscesses, scarring, and bowel obstruction; the condition is also characterized by diarrhea, cramping, and loss of ap-

petite and weight. Vitamin D status has been linked to IBD in association studies of sun exposure and in genetic studies through down-regulation of the Th1-mediated immune response.

Biological plausibility Ecological studies have linked vitamin D, particularly 25OHD levels, to a number of autoimmune diseases. A connection between seasonal vitamin D status and risk for Crohn’s disease was proposed by Peyrin-Biroulet et al. (2009), based largely on ecological evidence for an association between low 25OHD levels in blood and other autoimmune diseases. The effect of seasonal variation on serum 25OHD levels in patients with Crohn’s disease, compared to matched controls found that mean serum 25OHD was lower in Crohn’s patients despite having vitamin D intake from foods and supplements and sunlight exposure similar to those of matched controls (McCarthy et al., 2005). Genetic evidence in humans and in animal models provides some support for a biological association between polymorphisms in the Vdr and susceptibility to IBD and Crohn’s disease. In a human study, a linkage analysis, used to identify the TaqI polymorphism in the Vdr gene, suggested that the variant may be a candidate for conferring susceptibility to IBD (Simmons et al., 2000). Animal model studies in both vitamin D–deficient and Vdr null mice suggested that the risk of developing IBD is increased in several respects: spontaneous occurrences are increased, the disease is more severe, and the disease is more easily provoked in response to agents that induce IBD or bacterial infections transferred from an affected animal (reviewed in Bouillon et al., 2008).

Systematic reviews and meta-analyses The AHRQ-Tufts systematic review found no RCTs for immune function clinical outcomes and no evidence for IBD or Crohn’s disease. Thus, the evidence was insufficient for further analysis in the systematic review. No meta-analyses were identified for this indicator.

Additional evidence from randomized controlled trials No RCTs were identified for this indicator.

Observational studies Two observational studies were identified that evaluated 25OHD levels in patients with Crohn’s disease and/or IBD. A cross–sectional assessment of serum 25OHD levels in children and young adults with IBD living in Boston found that prevalence of low 25OHD status (≤ 38 nmol/L) averaged 34.6 percent overall, with higher prevalence in winter compared with summer (Pappa et al., 2006). A small population-based cohort of patients with Crohn’s disease and ulcerative colitis in Scandinavia found a prevalence of 25OHD levels below 30 nmol/L in 27 percent of those with Crohn’s disease and 15 percent of those with ulcerative colitis. In addition, patients with Crohn’s disease had lower mean serum 25OHD levels compared with those with ulcerative colitis or the reference population (Jahnsen et al., 2002). The study design and poor con-

trols characteristic of these observational studies diminish the reliability of their findings. Another confounding problem is that vitamin D is absorbed with fat in the terminal ileum, and this is the area that is most inflamed in Crohn’s disease (and can become inflamed in ulcerative colitis). Consequently, low 25OHD levels can be expected to occur as a consequence of the inflammatory condition. The question not answered by these studies is whether low 25OHD levels can predispose individuals to the conditions.

Multiple sclerosis MS is a chronic disease of the central nervous system that manifests as numbness in the limbs or, in more severe cases, paralysis or loss of vision. The progress, severity, and specific symptoms of MS are unpredictable and vary among individuals. The disease is an autoimmune response directed against myelin. Damaged myelin forms scar tissue (sclerosis), which impairs nerve impulse conduction, producing the variety of symptoms associated with the disease.

Biological plausibility Similar to findings with other autoimmune-related diseases, low solar exposure, latitude, and polymorphisms in the Vdr gene have been implicated in susceptibility to MS (Partridge et al., 2004; Dwyer et al., 2008; Sloka et al., 2008; Dickinson et al., 2009). However, whether a lack of sun exposure is causally related to MS cannot be shown. Findings from animal models are not consistent. In a mouse model, vitamin D deficiency accelerated development of autoimmune encephalomyelitis (the murine model of MS in humans), whereas treatment with calcitriol reduced it (Cantorna et al., 1996). In contrast, a subsequent study, using a mouse model null for the Vdr gene, found that the Vdr null mice were protected from development of the disease compared with wild-type mice (Meehan and DeLuca, 2002). A recent genetic study in humans evaluating associations between specific Vdr gene polymorphisms (Apal and Taq1) and serum 25OHD levels in healthy adults compared with those with MS, found no relationship between mutations in Apal and Taq1 and incidence of MS (Smolders et al., 2009). Taken together, neither ecological studies nor genetic studies in animal models and humans show consistency in finding a significant relationship between serum 25OHD level and presence of MS.

Systematic reviews and meta-analyses The AHRQ-Tufts systematic review found no RCTs for immune function clinical outcomes and no evidence for MS related to vitamin D. In a recent review paper of observational studies on the effects of vitamin D on incidence and severity of MS, Smolders et al. (2008) concluded that there was no strong direct evidence supporting the ability of vitamin D to modulate MS or influence risk for the disease. Their review included observational evidence in humans linking low serum 25OHD levels with incidence of MS in white American adolescents; associations between lower circulating levels of 25OHD after onset of MS;

associations between skin pigmentation and lower disability scores in females; congruence of geographical distribution of MS with geographical distribution of low vitamin D levels; associations between seasonal variation in birth and in disease severity with the seasonal variation of low vitamin D levels; associations between remission and pregnancy (when calcitriol levels increase); and variations in risk associated with polymorphisms in the Vdr gene. In addition to the lack of positive evidence, the authors raised concerns about the safety of calcitriol treatment for MS because of the dose-dependent risk for hypercalcemia identified with calcitriol treatment in animal models. No meta-analyses were identified for this indicator.

Additional evidence from randomized controlled trials No RCTs were identified for this indicator.

Observational studies Observational studies in humans have also failed to show a consistent association between serum 25OHD levels and MS. A small longitudinal study of 23 MS patients and 23 controls found no differences in circulating 25OHD levels, no difference in seasonal variation, and comparable rates of vitamin D deficiency or insufficiency based on serum 25OHD levels between MS patients and controls (Soilu-Hanninen et al., 2008). Interestingly, in this study, serum 25OHD levels were significantly lower during relapse episodes, whereas serum levels of intact PTH were significantly higher than in remission periods in MS patients. A prospective nested case–control study in military personnel reported that, for white subjects, serum 25OHD levels were inversely related to the risk of MS and that effect was even greater when serum 25OHD levels were low in individuals under 20 years of age (Munger et al., 2006). Overall, serum 25OHD levels were lower in black and Hispanic compared to white subjects and 25OHD levels were more frequently in the range of 25 to 40 nmol/L in MS patients compared to controls. These findings, however, were unrelated to risk for MS (Munger et al., 2006). In a small population-based case–control study of individuals living at latitudes of 41 to 43°S (similar to New York City and Boston), van der Mei et al. (2007) found that serum 25OHD levels below 25 nmol/L were moderately associated with MS, compared with levels above 40 nmol/L. With more consistent serum 25OHD levels and less seasonal variability, there was an association with less disability.

Taken together, these observational studies show widely variable outcomes for associations between serum 25OHD levels and MS and such associations are not supported by meta-analyses. In addition, the lack of causal evidence further diminishes the likelihood for a relationship between vitamin D and MS.

Rheumatoid arthritis RA is a chronic disease characterized by systemic inflammation that may affect many tissues and organs, but particularly the joints. In RA inflammatory synovitis of the joints can progress to destruction of the articular cartilage and ankylosis. RA can also produce diffuse

inflammation in the lungs, pericardium, pleura, and sclera, as well as nodular lesions under the skin. This progressive disease can result in chronic pain, loss of function, and eventual disability.

Biological plausibility In experimental studies, Tetlow and Wooley (1999) found that the VDR was strongly expressed in cells associated with rheumatoid lesions, including macrophages, synovial fibroblasts, and chondrocytes, but weakly or not at all in normal articular cartilage tissue, suggesting an up-regulation of VDR-mediated activity in tissues affected by RA. Smith et al. (1999) found that cultured human synovial fibroblasts, but not human articular chondrocytes, when treated with the inflammatory cytokine, interleukin 1 (IL-1), followed by calcitriol, indicated inhibition of expression of the matrix metalloproteinases associated with RA. In a mouse model of RA, treatment with calcitriol decreased arthritis symptoms induced by injection with bovine collagen and halted the progression of arthritis after arthritic lesions were apparent (Cantorna et al., 1998). Together, this evidence is suggestive of an immunomodulatory role for vitamin D in expression of arthritic changes in some, but not all, cell types associated with RA.

Systematic reviews and meta-analyses The AHRQ systematic reviews found no RCTs for immune function clinical outcomes related to RA and no evidence that RA was related to vitamin D. No meta-analyses were identified for this indicator.

Additional evidence from randomized controlled trials No RCTs were identified for an effect of vitamin D and/or calcium on risk for RA.

Observational studies A number of studies have been conducted to determine whether serum 25OHD level and incidence of RA are associated. In a prospective cohort study, a small subset of subjects from the Iowa Women’s Health Study were followed to determine if dietary vitamin D intake (primary outcome) and/or calcium intake (secondary outcome) were associated with incident RA (Merlino et al., 2004). No significant associations were found for dietary (not supplemented) vitamin D intake and risk for RA, although the association was significant for daily supplemental intakes of 400 IU or more compared with less than 400 IU. No association was found between calcium intake and risk for incident RA. A cross–sectional analysis of women with RA living in Brazil, found a significant correlation between higher mean serum calcium level and normal BMD compared with calcium levels in women with osteopenia, although no significant difference was found between calcium and vitamin D intake and BMD (Sarkis et al., 2009).

With no large prospective cohort studies and no clinical trials to support a relationship between vitamin D and/or calcium and RA, along with a paucity of other observational evidence, the committee could not conclude that either vitamin D or calcium is related to risk for RA.

Systemic lupus erythematosus Systemic lupus erythematosus (SLE) is a chronic generalized connective tissue disorder characterized by skin eruptions, arthralgia, arthritis, leukopenia, anemia, visceral lesions, neurological manifestations, and lymphadenopathy. It has been proposed that vitamin D plays a role in maintenance of immune homeostasis, and recent studies have linked SLE to vitamin D deficiency although a causal relationship has not been established.

Biological plausibility Mouse models of SLE have been reported to produce high levels of IgG2a immune cells that are implicated in the pathogenesis of lupus (Slack et al., 1984). However, owing to the complexity of the disease, experimental animal studies have not shown consistent outcomes on questions regarding the role for vitamin D in preventing or alleviating manifestations of the disease. Administration of calcitriol in a murine SLE model using a treatment protocol of daily dosing with a low-calcium diet for 4 weeks followed by dosing every other day for 18 weeks resulted in attenuation of symptoms of SLE, including reduced dermatological lesions. All SLE mice in the study developed proteinuria by 20 weeks; however, among those treated with vitamin D, lower urinary protein/creatinine ratios indicated reduced levels of proteinuria (Lemire et al., 1992). In contrast to these findings, when Vaisberg et al. (2000) injected SLE-prone mice with vitamin D₃, they found a worsening of the histopathological effects of SLE in the kidney. Upon examining serum levels of calcitriol and 25OHD in SLE patients compared with unaffected controls, Muller et al. (1995) found lower levels of calcitriol but not 25OHD, but were unable to speculate on a cause for the difference. In humans, vitamin D has been proposed to modulate maturation and induction of interferon alpha-(IFN-α)–mediated monocyte differentiation into dendritic cells that are activated in SLE. The findings of Ben-Zvi et al. (2010) suggest that such a role is likely via vitamin D–mediated inhibition of over-expression of IFN-regulated genes in cultured monocytes from both normal and SLE patients following exposure to an activation factor. Altogether, however, there is a lack of consistency in study outcomes between animal models and human experimental studies, and thus findings are not supportive of a biological role for vitamin D in SLE.

Systematic reviews and meta-analyses The AHRQ systematic reviews found no RCTs for immune function clinical outcomes related to SLE and no evidence for a relationship between SLE and vitamin D. No meta-analyses were identified for this indicator.

Additional evidence from randomized controlled trials No RCTs were identified that examined a role for vitamin D or calcium in reducing risk for or manifestations of symptoms of SLE.

Observational studies Epidemiological evidence to support an association between vitamin D status and incidence of SLE shows variability in

the levels of 25OHD associated with SLE. Kamen et al. (2006), in a small subset of a larger population-based cohort, found that 22 out of 123 SLE patients across race, age, and gender groups had 25OHD levels below 25 nmol/L and that African Americans had levels significantly lower than whites (40 nmol/L compared with 78 nmol/L [p = 0.04]). In a small pilot study of 50 subjects (25 per group), Huisman et al. (2001) found that 50 percent of female SLE patients had 25OHD levels below 50 nmol/L. The associations between vitamin D status and incidence of SLE identified in these studies are not borne out by evidence from a prospective cohort study of dietary factors and risk for developing SLE. An analysis of a small subset of women participating in NHS over a period of 22 years found no association between vitamin D (or calcium) intake assessed with a food frequency questionnaire and risk for developing SLE or RA (Costenbader et al., 2008). In a cross–sectional survey, Ruiz-Irastorza et al. (2008) found that 75 percent of patients with SLE had serum 25OHD levels below 75 nmol/L and 15 percent had levels below 25 nmol/L, and 25OHD levels in blood in patients with SLE were not responsive to calcium and vitamin D treatment. Thus, it is not clear whether therapeutic treatment would have any effect on disease manifestation. The few relevant studies identified for review, and the lack of uniformly significant findings between studies, which may be a result of the small study populations (< 200 participants), are not sufficient to permit the committee to draw a conclusion about an association between SLE and vitamin D intake or 25OHD levels in blood.

Tuberculosis Pulmonary tuberculosis (TB) is a granulomatous infection in which hypercalcemia occurs in a subset of patients (Sharma et al., 1972; Abbasi et al., 1979; Need and Phillips, 1979). The increased production of immune and inflammatory cells in patients with TB correlates with increased serum levels of calcitriol (Adams et al., 1989) and with calcitriol in pleural fluid (Cadranel et al., 1994). Treatment of alveolar macrophages with IFN-γ appears to stimulate synthesis of calcitriol (Koeffler et al., 1985; Reichel et al., 1987). Although vitamin D has been used as a therapeutic agent in the management of TB (Martineau et al., 2007a), treatment of individuals with active TB with supplemental vitamin D exacerbates or reveals hypercalcemia (Sharma, 1981).

Biological plausibility Vitamin D may be an important factor in innate immunity in the upper respiratory tract (reviewed in Bartley, 2010). Although calcitriol does not have direct anti-bacterial activity, it induces anti-tubercular actions in cultured monocytes and macrophages (Chan et al., 1994). Recent evidence, stemming from the molecular cloning of 1α-hydroxylase (Monkawa et al., 2000), supports macrophages as the source

of 1α-hydroxylase that converts 25OHD to calcitriol and stimulates the mobilization of calcium seen in inflammatory diseases (Inui et al., 2001; Yokomura et al., 2003; Karakelides et al., 2006). Animal models provide additional evidence that calcitriol levels increase in tubercular infection. Rhodes et al. (2003) identified a transient increase in calcitriol levels following infection of cattle with bovine mycobacterium, but only among animals that went on to develop TB. This increase in activated vitamin D that accompanied infection suggests a role for vitamin D in the host immune reactivity to TB infection. In a mouse model, calcitriol increased production of nitric oxide, an endogenously produced anti-infective compound, suggesting a bactericidal mechanism for vitamin D in TB-infected animals (Waters et al., 2004). Other more recent studies have identified the peripheral cellular conversion of 25OHD to calcitriol as a mechanism for rapid local induction of anti-microbial peptides as a direct mechanism for killing TB and staphylococcal bacteria (Liu et al., 2006; Schauber et al., 2006). Immune responses to increased 25OHD levels may vary among individuals as a result of the genetic expression of Vdr polymorphisms. For example, Selvaraj et al. (2008) found that allelic variations in the Cdx-2 polymorphism were associated with either resistance or susceptibility to TB bacteria; however, more research is needed to better understand the genetic relationship between Vdr polymorphisms and TB susceptibility.

Systematic reviews and meta-analyses The AHRQ reviews did not identify relevant evidence for TB as an outcome. Nnoaham and Clarke (2008) systematically reviewed and meta-analyzed the relationship between low serum 25OHD levels and TB among diverse community- and hospital-based population groups in both developed and developing countries. The meta-analysis was restricted to studies that compared serum 25OHD levels in TB patients not on a treatment regimen with healthy matched controls. Seven observational studies were included in the meta-analysis, which found a 70 percent probability that a person without TB would have a higher serum 25OHD level than a person with TB. Whether this association predicts low serum 25OHD as a risk factor for active TB cannot be concluded from the analysis; however, the findings would increase the strength of similar findings from other high-quality observational studies or from a larger meta-analysis.

Additional evidence from randomized controlled trials Even though treatment of individuals with active TB with supplemental vitamin D can lead to hypercalcemia, vitamin D has been used as a therapeutic agent in the management of the disease. A small double-blind RCT of 131 individuals who had come into close contact with patients with TB (or TB contacts) in the United Kingdom (UK) measured the ability of a single oral dose of 100,000 IU of vitamin D to inhibit growth of recombinant mycobacteria,

an indicator of TB exposure, grown in vitro and detected using the Barger-Lux (BCG-lux) assay, from the study subjects (Martineau et al., 2007b). Compared with placebo, subjects who received the vitamin D treatment showed a significant increase in serum 25OHD levels for 6 weeks without hypercalcemia, as well as inhibition of in vitro growth of mycobacteria. Another small intervention trial examined the effect of UVB exposure on TB in recent Asian immigrants to the UK who have a higher TB prevalence than indigenous populations (Yesudian et al., 2008). Anti-microbial activity, expressed as response to the BCG-lux assay, varied between the subjects, showing a small transient decrease but no significant change following UVB exposure. Another RCT testing the effect of vitamin D supplementation on the clinical course of patients with TB also found no significant change in clinical outcome or reduction in mortality in patients treated with 100,000 IU of vitamin D three times over 8 months (Wejse et al., 2009). Only one of the two studies reviewed showed a significant effect of vitamin D treatment on 25OHD levels in TB patients, and none of the clinical trials showed a significant effect of vitamin D on clinical outcome. Thus, evidence from RCTs does not support a reduction in TB infections with vitamin D treatment.

Observational studies Because TB is rare in industrialized countries, particularly the United States, most observational studies are on populations from developing countries that have immigrated to developed countries. Gibney et al. (2008), in a retrospective analysis of hospitalized patients, found that low serum 25OHD levels in patients born in sub-Saharan African countries who immigrated to Australia were predictive of any form of TB infection as well as current and past infection. Another analysis of a small study of patients reporting to a TB clinic prior to treatment (Sita-Lumsden et al., 2007) found a statistically significant difference in serum 25OHD levels between TB patients and their contacts and the greatest difference among those patients with the lowest 25OHD levels. Although there was no difference in dietary intake of vitamin D between TB patients and their contacts, the TB patients did demonstrate a stronger correlation between dietary intake and measures of vitamin D in serum. Sun exposure did not differ between patients and their contacts. Strachan et al. (1995), in a case–control study of Asian immigrants to the UK who had a diagnosis of active TB, found a trend of increasing risk of TB correlated with a decreasing frequency of consumption of meat or fish and an 8.5-fold increased risk for TB among lactovegetarians compared with daily meat or fish eaters. Many observational studies of an effect of vitamin D on susceptibility to TB are confounded by endogenous production of calcitriol in infected individuals (Adams et al., 1989; Cadranel et al., 1994) and thus must be cautiously interpreted. Nevertheless, the few small studies identified support the findings

of the meta-analysis from Nnoaham and Clarke (2008) for higher serum 25OHD levels in TB patients; however, these studies did not uniformly find significant associations between vitamin D intake and risk for TB.

Influenza and upper respiratory infections Influenza is an acute contagious viral infection characterized by inflammation of the respiratory tract, fever, chills, and muscular pain. Upper respiratory infections are most commonly viral infections characterized by inflammation of the respiratory tract. Vitamin D has been hypothesized to act through the immune system to prevent influenza infections.

Biological plausibility Environmental observations of seasonal variation in serum 25OHD levels and occurrence of influenza have been proposed as an indicator to show a correlation between vitamin D and risk for influenza (Cannell et al., 2006; Hayes, 2009). In an animal study (Underdahl and Young, 1956), mice deficient in vitamins A, D, and E and inoculated with influenza had the same intensity of influenza infection as those mice replete in vitamins A, D, and E, showing no effect of vitamin D status on reducing influenza infection. This finding contrasts with earlier work by Young et al. (1949), which suggested that vitamin D could reduce the susceptibility of mice to influenza.

Systematic reviews and meta-analyses The AHRQ studies did not identify any relevant studies for influenza as a health outcome. There were no meta-analyses identified for this indicator.

Additional evidence from randomized controlled trials Available data from RCTs do not provide strong support for a role for vitamin D in reducing susceptibility to influenza infection. A small RCT testing the effect of 1,200 IU of vitamin D supplementation per day for 4 months found that for influenza A as the primary outcome, occurrence between days 1 and 30 was not significantly different between the vitamin D group and placebo, but between days 31 and 60, influenza A occurred significantly less often in the vitamin D than in the placebo group; between day 61 and the end of the study, the occurrence of influenza A was not significantly different between the vitamin D and placebo groups (Urashima et al., 2010). Analysis of other related secondary outcomes showed no significant difference for influenza B, influenza-like illness (negative in rapid influenza diagnostic tests), non-specific fever, gastroenteritis, pneumonia, hospital admission, or absence from school. Overall, the absolute reduction in influenza A cases was offset by a similar increase in the number of influenza B cases. These may be chance findings, however, as a result of confounding by the loss of subjects. In another 3-month prospective double-blind RCT (Li-Ng et al., 2009), even though 73 percent of subjects supplemented with 2,000 IU of vitamin D₃ daily achieved a serum 25OHD level above 75 nmol/L, no benefit of supplementation was seen for either prevention of self-reported

upper respiratory infections or a decrease in their severity. This evidence from RCTs in both children and adults shows no causal role for vitamin D in either reducing or preventing influenza.

Observational studies Only one study of 16 patients was identified in which plasma 25OHD levels were measured in children undergoing tympanosotomy tube placement. The study showed that 50 percent of the children had 25OHD levels less than 50 nmol/L and 31 percent had levels between 52.5 and 72.5 nmol/L (Linday et al., 2008). The authors concluded from this finding that there was a possible relationship between vitamin D and susceptibility to bacterial infection and influenza. This small observational study is not adequate to support a relationship between vitamin D and one outcome related to influenza infection and no additional evidence was found to verify any causal or associative relationship between vitamin D and influenza.

The committee’s review of the results of large cohort studies showed support for a positive association between vitamin D intake and reduction in symptoms associated with asthma but not with diagnosed disease. Other observational evidence of lower quality was found to largely support an association between risk for asthma and 25OHD level in blood, but associations have not been shown for diagnosed asthma. The lack of causal evidence and the lack of observational data demonstrating a relationship between vitamin D and diagnosed asthma led the committee to conclude that development of a DRI for this indicator is not supported by the totality of the evidence reviewed.

Emerging observational evidence in humans and experimental studies in animals inversely links vitamin D measures to risk of autoimmune disorders such as type 1 diabetes, MS, and IBD as well as infectious diseases such as TB (Maruotti and Cantatore, 2010). However, even though animal models indicate plausibility for a mechanistic role for vitamin D in autoimmune or anti-microbial function, results from RCTs as well as from observational associations between vitamin D and calcium and risk for either autoimmune or infectious diseases show a lack of consistency. Although both retrospective and prospective studies tend to support an inverse association between serum 25OHD levels and autoimmune and infectious diseases, these studies are limited in their interpretation owing to confounding effects that require further verification. The evidence available from RCTs is of limited utility because of the small size of the trials, inconsistency in measured outcomes, and lack of dose–response data. Overall, the evidence was not consistently supportive of a causal role for vitamin D combined with calcium or for vitamin D alone in reducing risk for developing auto-

immune or infectious diseases. In the absence of verifiable dose–response data from RCTs a conclusion about asthma, autoimmune, or infectious diseases as indicators for DRI development cannot be reached.

Emerging evidence is suggestive of a role for vitamin D in neuropsychological functioning, including a range of diseases from autism to Alzheimer’s.

Autism is a neurodevelopmental disorder of unknown etiology that manifests as repetitive behaviors, social withdrawal, and communication deficits. A number of factors are implicated in development of the disorder, including genetic (Abrahams and Geschwind, 2008) and environmental (Deth et al., 2008) factors.

Biological plausibility Although mechanistic studies using animal models tend to support an association between vitamin D intake during pregnancy and subsequent development of autism, these studies are limited in their interpretation and extrapolation to humans. There are some animal models suggesting a mechanism whereby vitamin D may influence the development of autism (reviewed in McGrath et al., 2004). These experiments demonstrate that pre-natal deprivation of vitamin D₃ results in gross abnormalities in fetal rat brains at birth. Feron et al. (2005) subsequently reported that vitamin D deprivation and associated disruptions in brain development seen in rat pups at birth persisted into adulthood.

In humans, Vdr gene polymorphisms have been proposed as a possible link to psychiatric diseases, including autism. Yan et al. (2005) analyzed the coding sequences and splice junctions of 100 patients with schizophrenia and, in a pilot study within the same population, 24 patients with autism. The frequency of the sequence variants identified, however, was not significantly different from that of sequence variants found in control subjects.

Systematic reviews and meta-analyses The AHRQ reviews did not identify evidence to support autism as a relevant health outcome for vitamin D. No meta-analyses were identified for this indicator.

Additional evidence from randomized controlled trials No RCTs were identified for this indicator.

Observational studies No large, prospective cohort studies were identified that examined associations between either vitamin D intake or 25OHD

levels in blood and risk for autism. Three lower-quality observational studies were identified for further consideration: one retrospective chart review, one small cohort from a developing country, and one cross–sectional study. A recent study in Sweden that retrospectively reviewed serum 25OHD levels from medical records of out-patients receiving psychiatric care, including autism, suggested a high prevalence of vitamin D insufficiency in psychiatric outpatients (Humble et al., 2010). Although the study did not include matched controls, comparisons made with previously published samples from healthy Swedish populations suggested that the prevalence of vitamin D insufficiency was greater in the population receiving psychiatric care. The study, however, did not take into account dietary intake of vitamin D. Fernell and Gillberg (2010) tested the association between serum 25OHD levels and prevalence of autism in an analysis of a small cohort of mothers of children with autism from Somalia and Sweden, but they found no statistically significant differences in serum 25OHD levels between either group of mothers and controls.

Herndon et al. (2009), in a cross–sectional study, examined associations between vitamin D and calcium in dairy foods and autism. This study found that children with autism spectrum disorders consumed less calcium and fewer servings of dairy foods compared with children with typical development. Interpretation of this evidence for an association between vitamin D measures and risk for autism, however, is confounded by other potential factors that could influence vitamin D measures.

Concluding statement Owing to the lack of causal evidence from RCTs and a paucity of evidence, as well as a lack of data from large, prospective cohort studies and inconsistent findings for an association between vitamin D and incidence of autism from largely cross–sectional observational studies, autism was not considered further as an indicator for DRI development.

Loss of cognitive function in the form of dementia is frequently associated with aging. Between the ages of 60 and 85, the prevalence of dementia in the general population increases from 1 to 40 percent (Bolla et al., 2000). Dementia is classified into four major subtypes: Alzheimer’s disease, Lewy body dementia, frontotemporal dementia, and vascular dementia (Bolla et al., 2000; Grossman et al., 2006). Vitamin D has been hypothesized to confer neuroprotective effects and reduce the risk for developing dementia (Buell and Dawson-Hughes, 2008; McCann and Ames, 2008).

Biological plausibility Vitamin D has been proposed to prevent cognitive decline, and plausible biological mechanisms support this hypothesis. Vitamin D may protect against cognitive decline by promoting vascular health

through anti-inflammatory or other pathways, but may also have direct neuroprotective effects (Buell and Dawson-Hughes, 2008; McCann and Ames, 2008). Rodent models show morphological and biochemical effects of vitamin D on brain tissue. Early experiments on rat brain revealed that vitamin D-deficiency reduced vitamin D–dependent enzyme activity in the cerebral context, including non-sodium-mediated glucose transport, that was restored when rats were treated with calcitriol (Stio et al., 1993). Subsequent work in mouse models demonstrated that developmental vitamin D deficiency had a negative effect on brain development, as manifested by changes in brain size and shape and ventricular size and reduced nerve growth factor expression (McGrath et al., 2004; Feron et al., 2005), as well as effects on brain function and exploratory behavior (Harms et al., 2008). When adult offspring of dams deprived of vitamin D during pregnancy underwent learning tests, they displayed impaired learning at 30 weeks of age but not at 60 weeks (de Abreu et al., 2010).

VDR and 1α-hydroxylase are found throughout the brain. Vitamin D affects gene and protein expression in brain tissue, including expression of neurotropins and glial cell–derived neurotrophic factor (Naveilhan et al., 1996; Sanchez et al., 2002). A battery of 40 different tests in Vdr-null mice showed them to have normal cognitive function but abnormal muscle and motor behavior, although the abnormal neuromuscular function may be due to hypocalcemia rather than a direct effect of loss of the VDR (reviewed in Bouillon et al., 2008). In neurological tissue, vitamin D modulates certain calcium-binding proteins, including calbindin-D28K, parvalbumin, and calretinin, which are important for brain function (de Viragh et al., 1989; Alexianu et al., 1998). In addition, calcitriol down-regulates the expression of calcium channel currents in rat hippocampal cells (Brewer et al., 2006), stimulates neurogenesis in human neuroblastoma cells (Moore et al., 1996; Taniura et al., 2006), and may affect other pathways (Garcion et al., 1997; Baas et al., 2000; Brown et al., 2003; Obradovic et al., 2006). Vitamin D restriction results in unfavorable structural and biochemical changes in the brain (Ko et al., 2004; Feron et al., 2005). However, experiments in rats rendered vitamin D deficient by dietary and UV radiation restriction and in Vdr null mice have not consistently shown learning impairments, although the data are sparse (Becker et al., 2005; Minasyan et al., 2007). Calcium independently of or in concert with vitamin D is involved in many physiological processes related to neural functioning, and disturbed calcium homeostasis is also characteristic of neurodegenerative disorders (Canzoniero and Snider, 2005; Mattson, 2007; Toescu and Verkhratsky, 2007).

Systematic reviews and meta-analyses The AHRQ reviews did not identify sufficient evidence to support cognition (or cognitive decline) as a relevant health outcome for vitamin D. No meta-analyses were identified for this indicator.

Additional evidence from randomized controlled trials In the WHI trial, a subset of participants completed a cognitive test battery, but results of analyses examining the vitamin D intervention’s effect on cognitive function are not yet available.

Observational studies Numerous observational studies that examined associations between vitamin D, serum 25OHD level, or calcium and cognitive function were identified as potentially relevant to DRI development. The greatest number of studies, however, were cross–sectional. No large prospective cohort studies were identified for review, although two analyses of data from large population-based annual surveys and several small cohort studies were included.

Low serum 25OHD levels have been associated with decreased cognitive function in various population groups. A cross–sectional analysis of 752 women 75 years of age and older in the Epidémiologie de l’Ostéoporose (EPIDOS) study found that participants with vitamin D deficiency (serum 25OHD level < 25 nmol/L) had twice the odds of cognitive impairment as other participants (Annweiler et al., 2010). In a population-based cross–sectional study, Lee et al. (2009) examined associations between serum 25OHD level and cognitive function and mood among adult men in a European population. In a spline regression model, significant associations were found between slower information processing and serum 25OHD levels below 35 nmol/L in men ages 40 years and older. In contrast to these findings, a more recent cross–sectional study found that among 1,604 men up to 65 years of age in the Osteoporotic Fractures in Men (MrOS) Study, there were no associations between serum 25OHD level and cognitive impairment, even after adjusting for age, race/ethnicity, education, and other potential confounders (Slinin et al., 2010). This study also examined vitamin D measures as a predictor of subsequent cognitive decline over a mean of 4.6 years of follow-up and found only a borderline significant trend across the first three quartiles of serum 25OHD levels, (≤ 49.75 nmol/L, 50.0 to < 62.75 nmol/L, and 62.75 to < 74.5 nmol/L, respectively), compared with the fourth quartile (≥ 74.5 nmol/L); serum 25OHD level did not predict decline on a timed test of executive function.

In a cross–sectional study of 318 older individuals (mean age 74 years) receiving home health care services, those who received a neurological exam and cranial magnetic resonance imaging (MRI), a lower serum 25OHD level (< 50 nmol/L) was associated with at least twice the odds for all-cause dementia, Alzheimer’s disease, and stroke, as well as increased white-matter hyperintensity volume and prevalence of large-vessel infarcts (Buell et al., 2010). Among three age groups (adolescent, adult, and elderly) examined from NHANES III, no association was found between high serum 25OHD levels and learning or memory, and only the elderly population group was found to have an inverse association between 25OHD level

and performance on a task of learning and memory. Within the elderly population group, those in the highest quintile for serum 25OHD level were also the most impaired; thus, the results fail to confirm the hypothesis that serum 25OHD level enhances performance in learning and memory (McGrath et al., 2007).

There are few observational studies on calcium and cognitive function. In a cross–sectional study on Korean adults 60 years of age and older, a positive association was found between calcium intake and score on the Mini-Mental State Examination for Koreans (MMSE-K) in women but not in men after adjustment for age (Lee et al., 2001). In contrast, another study in Portuguese adults more than 65 years of age found no association between calcium intake and MMSE score after 8.5 months of follow-up (Velho et al., 2008). Using a cross–sectional analysis, Wilkins et al. (2009) found, as expected, lower serum 25OHD levels among the African American population compared with the white population, and poorer cognitive performance among African Americans with the lowest 25OHD levels compared with those with higher levels. Similarly, in a cross–sectional analysis in a British population of adults ages 65 years or older with serum 25OHD levels reported in quartiles, Llewellyn et al. (2009) found a greater risk for impaired cognitive performance among persons in the lowest (8 to 30 nmol/L) compared with the highest quartile (66 to 170 nmol/L). Even though the committee identified a large number of observational studies that evaluated associations between vitamin D and calcium and cognitive function, these were predominantly lower quality cross–sectional studies or small cohort studies, and their results were mixed. No causal evidence was found to support experimental evidence for biological plausibility and the relatively weak observational evidence. The committee took into account the generally lower quality of the study designs in its interpretation of the findings and in drawing conclusions about outcomes associated with this indicator.

Depression is a disease with characteristic signs and symptoms that interfere with the ability to work, sleep, eat, and enjoy once-pleasurable activities. These signs and symptoms include loss of interest in activities; a persistently sad or anxious mood; feelings of hopelessness, pessimism, guilt, worthlessness, or helplessness; social withdrawal; fatigue; sleep disturbances; difficulty in concentrating or making decisions; unusual restlessness or irritability; persistent physical problems that do not respond to treatment; and thoughts of death or suicide or suicide attempts. Depressive disorders include major depressive disorder, dysthymic disorder, psychotic depression, postpartum depression, and seasonal affective disorder, with

major depressive disorder and dysthymic disorder being the most common.² Whether there is a functional relationship between measures of serum vitamin D or intake and mood or depression has not been determined.

Biological plausibility Seasonal affective disorder occurs more often at northern latitudes, and the etiology is presumed to be due, at least in part, to lack of sunlight exposure. In turn, lack of sunlight exposure causes low serum 25OHD levels unless the diet is adequate in vitamin D. Investigators have pursued the hypothesis that the low serum 25OHD levels are a cause of seasonal affective disorder, although it must be considered that the lack of sunlight may independently cause both seasonal affective disorder and low serum 25OHD levels without a direct link between them.

Systematic reviews and meta-analyses The AHRQ reviews did not identify sufficient evidence to support depression as a relevant health outcome for vitamin D. No meta-analyses were identified for this indicator.

Additional evidence from randomized controlled trials One RCT was identified that evaluated effects of vitamin D supplementation on depressive symptoms. Jorde et al. (2008) gave either 20,000 or 40,000 IU of vitamin D₃ or a placebo treatment weekly for 1 year to men and women ages 21 to 70 years, living in Norway. Symptoms of depression were evaluated using the Beck Depression Inventory (BDI), and serum 25OHD level and BMI were measured. Participants whose serum 25OHD levels were below 40 nmol/L had significantly higher BDI scores, indicating a higher incidence of depressive disorder, compared with those whose serum 25OHD levels were 40 nmol/L and above after 1 year. Both treatment groups indicated significant improvement in BDI score compared with placebo.

Results of other randomized trials testing the effects of vitamin D on a subtype of depression that occurs during the winter months have been mixed. Three small, short-term trials examining effects of treatment with vitamin D for seasonal affective disorder reported that vitamin D improves mood (Lansdowne and Provost, 1998; Gloth et al., 1999; Vieth et al., 2004), but a larger, longer-term trial found no effect (Thys-Jacobs et al., 1998). Vieth et al. (2004) treated adults with serum 25OHD concentrations below 61 nmol/L with the equivalent of either 4,000 or 600 IU of vitamin D per day for 3 months over two consecutive winters and found evidence of a significant difference in measures of improved well-being at the higher compared with the lower dose. Lansdowne and Provost (1998) assigned healthy adults to 5 days of treatment with either 400 or 800 IU of vitamin

D or placebo and found that both vitamin D doses increased positive affect and decreased negative affect compared with placebo. Gloth et al. (1999) assigned 15 people to either 100,000 IU of vitamin D or broad-spectrum light therapy for 1 month and found an increase in serum 25OHD level was significantly associated with improvement in depressive symptoms. However, Harris and Dawson-Hughes (1993) randomized 250 middle-aged and older women to treatment with 400 IU of vitamin D per day of vitamin D plus 377 mg of calcium per day or to calcium alone for 1 year and found no treatment-related changes in seasonal mood as assessed by the Profile of Mood States (POMS) questionnaire.

Observational studies Among four cross–sectional studies on small population groups (n < 50) that evaluated associations between serum 25OHD level and evidence for clinical diagnosis of depression in women (Michelson et al., 1996; Herran et al., 2000; Eskandari et al., 2007) or men and women (Schneider et al., 2000), only Eskandari et al. (2007) found a significant association between serum 25OHD level and diagnosis of depression and Michelson et al. (1996) found a significant association with calcitriol level. Another large population-based cross–sectional study among middle-aged and elderly Chinese also found, after controlling for confounders and geographic location, no significant associations between serum 25OHD level (grouped by tertile) and symptoms of clinical depression (Pan et al., 2009). In contrast to the cross–sectional studies, Hoogendijk et al. (2008) found, in a cohort study in the Netherlands, a significantly lower mean serum 25OHD level (47.5 nmol/L) among individuals with both major and minor depression compared with a mean level of 55 nmol/L among those who did not have depression.

Although some observational studies support an association between low measures of vitamin D exposure and risk for cognitive impairment or changes in mood, results have been inconsistent, and the majority of studies were cross–sectional in study design, including possible selection bias or other confounding factors that diminish the quality ranking of the studies. In addition, few or no clinical trials were identified to support biological plausibility. As a result of the many shortcomings in study design and quality of observational evidence and the paucity of high-quality evidence from RCTs identified by the committee, the findings for neuropsychological indicators are inconclusive. The committee’s review of the available evidence for either associations or a causal relationship between vitamin D and calcium and risk for cognitive disorders shows a lack of sufficient evidence to support DRI development.

Preeclampsia is a serious condition in which hypertension and proteinuria arise in pregnancy. It can affect both the mother and the unborn child. Pregnancy-induced hypertension is a transient hypertension without proteinuria that occurs during pregnancy. Pregnancy is a type of immunological challenge, and women with some autoimmune diseases, particularly type 1 diabetes and RA, are at increased risk for developing preeclampsia (Evers et al., 2004; Wolfberg et al., 2004). Clinical observations have noted that urinary calcium excretion is low in women with preeclampsia, whereas it is elevated in women during normal pregnancy. Calcium intake has been examined relative to reducing the risk of preeclampsia.

Vitamin D metabolism may be altered under conditions of preeclampsia (August et al., 1992), when calcitriol level is low and hypocalciuria is present, but it is unclear whether these are causes or consequences of preeclampsia. The placenta and deciduas both express 1α-hydroxylase and activate 25OHD in vitro. Calcitriol regulates immunomodulatory cytokine production in cultured decidual cells (Evans et al., 2006) and placental trophoblasts (Diaz et al., 2009). However, the specific role of vitamin D in vivo is less clear. Its actions in vitro may provide some clues as to its physiological relevance, but these hypotheses need to be examined rigorously in future studies.

As mentioned above, it has long been observed that urinary excretion of calcium is increased during pregnancy, and hypercalciuria may result. In contrast, women with preeclampsia often have hypocalciuria. This observation has prompted a number of investigations to test whether low calcium intake predisposes a pregnant woman to both hypocalciuria and preeclampsia, although a biological mechanism to explain how low calcium intake would cause the preeclampsia has not been clearly elucidated.

The AHRQ-Tufts analysis identified a single nested case–control study (rated B for methodological quality) that evaluated the association between serum 25OHD concentration and the risk of preeclampsia (Bodnar et al., 2007). The researchers found a significant association between preeclampsia and serum 25OHD concentrations when the serum values were less than 37.5 nmol/L early in pregnancy.

A 2007 systematic review of evidence incorporated 12 RCTs (15,528 women) to examine the relationship between calcium supplementation and preeclampsia prevention (Hofmeyr et al., 2007). Calcium supplementation reduced overall hypertension in 11 of the studies reviewed and incidence of preeclampsia in 12 of them. There was also a significant effect for calcium among women at high risk, which was greatest among those with lower baseline calcium intakes.

Additional RCTs not included in the review from Hofmeyr et al. (2007), although of lower-quality study design, also reported similar results suggesting that there may be no effect from daily calcium supplementation when dietary calcium intake is already adequate (Hofmeyr et al., 2006 [1,000 mg/day]; Villar et al., 2006 [2,000 mg/day]; Hiller et al., 2007 [1,800 mg/day]; Kumar et al., 2009 [500 mg/day]).

A prospective non-randomized clinical trial of the effect of vitamin D (0.5 mg/day) and calcium (312 mg/day) supplementation in women at risk for preeclampsia found that the incidence of preeclampsia was 10.9 percent lower in treated women than in controls (Ito et al., 1994). However, women who had the highest level of angiotensin II, a marker for preeclampsia, also had the lowest incidence of preeclampsia; thus, the role of calcium and vitamin D supplementation in preventing of preeclampsia is not clear from this trial (Ito et al., 1994). A small randomized trial of pregnant women in India supplemented with 600,000 IU of vitamin D in the seventh and eighth months of pregnancy compared with controls found no significant difference in incidence of preeclampsia, although a significant reduction in systolic blood pressure of 8 mmHg was seen in the vitamin D–supplemented group (Marya et al., 1987). The relationship of this small decrease in blood pressure to pregnancy-induced hypertension is unclear.

Findings from observational studies have shown mixed results. In addition to the nested case–control study reviewed in AHRQ-Tufts, the committee identified one large prospective cohort study, a large retrospective cohort study, and two small case–control studies examining vitamin D intake or serum 25OHD level and risk for preeclampsia, as well as one small case–control study of serum calcitriol level and pregnancy-induced hypertension (Lalau et al., 1993). In the large prospective study, Haugen et al. (2009) examined associations between risk for preeclampsia and intake of vitamin D from diet and supplements. This study found that women

who developed preeclampsia did not have a lower intake of vitamin D from foods compared with women without preeclampsia, but they did have a significantly lower intake of vitamin D from supplements, although the dose of the supplement was not correlated with prevalence of preeclampsia. Risk for preeclampsia was reduced by 31 percent in women who achieved a total vitamin D intake from food plus supplements between 300 and 400 IU/day, and the minimum combined intake of vitamin D needed for a protective effect was 200 IU/day. Hypponen et al. (2007) retrospectively examined the use of vitamin D supplements in infants of women with previously diagnosed preeclampsia in the Northern Finland Birth Cohort of 1966. The female children of mothers who had preeclampsia had a greater prevalence of preeclampsia in their own pregnancies, but vitamin D supplementation was significantly associated with a lower subsequent risk of developing preeclampsia. In contrast to these findings, two case–control studies, one in the United States (Seely et al., 1992) and one in Denmark (Frolich et al., 1992), found no significant difference in serum 25OHD levels between women with preeclampsia and those without, even though serum calcium levels were significantly lower in the women with pregnancy-induced hypertension or preeclampsia, respectively. In a small case–control study, women with pregnancy-induced hypertension had lower total and free serum calcitriol levels than did normotensive women during pregnancy (Lalau et al., 1993), but serum 25OHD levels and vitamin D intake were not measured.

Overall, two observational studies identified associations between supplementary vitamin D and incidence of preeclampsia, but data on associations between serum 25OHD level and preeclampsia were not conclusive. Similarly, only one observational study reported an association of pregnancy-induced hypertension with lower serum total and free calcitriol levels (Lalau et al., 1993), and no placebo-controlled RCTs were identified that examined a causal relationship between vitamin D and preeclampsia or pregnancy-induced hypertension.

Calcium supplementation has not been shown to have an effect on the incidence of preeclampsia in normal women meeting calcium requirements, but it may be of benefit in cases of low calcium intake. Associations between serum 25OHD level (as well as calcitriol level) and the onset of preeclampsia have not been well studied and a mechanism of action is unclear. Thus, because of the lack of a causal relationship and the inconsistent results in the observational studies for both vitamin D and calcium the committee concluded that neither preeclampsia nor pregnancy-induced hypertension can be considered as an indicator for DRI development.

Neither AHRQ-Ottawa nor AHRQ-Tufts addressed maternal non-skeletal outcomes beyond preeclampsia and pregnancy-induced hypertension. However, other non-skeletal outcomes may include maternal events such as cesarean section, obstructed labor, and vaginosis. Regarding fetal outcomes, so-called developmental programming of health outcomes in the offspring may focus on immune-related outcomes such as type 1 diabetes and atopic eczema, which has been included above in autoimmune response, as well as measures of BMD and skeletal development, discussed below in the Skeletal Health section. Infant birthweight is also of interest.

One observational study has reported an increased risk of approximately 65 percent and 26 percent of bacterial vaginosis in women with serum 25OHD levels below 20 nmol/L and below 50 nmol/L, respectively, compared with those with serum 25OHD levels of 75 nmol/L (Bodnar et al., 2009). Two observational studies reported conflicting results for the association of serum 25OHD levels with maternal delivery (cesarean/obstructed), with one finding an inverse association (Merewood et al., 2009) and the other finding no relationship with serum 25OHD levels. Overall, insufficient evidence makes maternal Cesarean delivery/obstructed labor uninformative for DRI development.

Regarding infant birthweight, AHRQ-Tufts discussed two RCTs (Mallet et al., 1986 and Maxwell et al., 1981; quality graded B and C, respectively) and reported no effect of supplemental vitamin D during pregnancy on offspring’s birthweight or length, and also one RCT (Marya et al., 1988; quality graded C) that reported an increased birthweight in vitamin D–supplemented pregnant women with low dietary intakes of vitamin D. One additional RCT published after the AHRQ-Tufts report (Yu et al., 2009) also reported no effect of vitamin D on infant birthweight.

Brooke et al. (1980) reported on an RCT that involved 126 women treated with either a dose of vitamin D or a placebo during pregnancy. The intended dose for the treated group was 1,000 IU/day, but it appears that a higher dose was administered (10,000 IU/day) as the achieved cord level of 25OHD was 138 nmol/L for the treated group versus 10 nmol/L. In any case, no change in birthweight was evidenced. In a smaller study of 40 pregnant women, Delvin et al. (1986) showed no effect on birthweight comparing 1,000 IU/day versus a placebo.

Several observational studies have also examined this relationship, again with conflicting results. In a nested case–control study, a U-shaped relationship was found only in white women, with an increased probability (2.4 to 3.9) of small-for-gestational-age measures in those women in the lowest (21 to 58 nmol/L) and the highest (90.7 to 245.0 nmol/L) quartiles of serum 25OHD levels (Bodnar et al., 2010). In the same study, despite

lower serum 25OHD levels in black women, no relationship was found between small-for-gestational-age measures and maternal serum 25OHD levels. In a prospective cohort observational study, mean birthweight was lowest in infants born to women with 25OHD levels below 30 nmol/L, intermediate in those born to women with 25OHD levels 30 to 50 nmol/L, and highest in those born to women with 25OHD levels above 50 nmol/L (Leffelaar et al., 2010). Birthweight was 60 g lower for the infants of women who consumed less than 200 IU of vitamin D per day during pregnancy compared to those who consumed 200 IU of vitamin D or more per day, and there was a significant linear trend for increased birthweight from lowest to highest quintile of intake (Scholl and Chen, 2009). Morley et al. (2006) enrolled 475 women in a study that compared maternal serum 25OHD levels during pregnancy to offspring birth size. No relationship was reported. Other observational studies reported no effect of vitamin D on birthweight (Brunvand et al., 1998; Gale et al., 2008; Farrant et al., 2009). The relationship, however, has not been tested in a sufficiently powered clinical trial. Corrections for differences in gestational length and other potentially confounding factors are not usually possible with associational studies. The available evidence for non-skeletal outcomes is limited and presently conflicting among both RCTs and observational studies, precluding the ability to find these data useful at this time for DRI development.

Skeletal health, referred to commonly as bone health, is manifested by desirable growth and maintenance of skeletal tissue, including bones and teeth. The use of bone health outcomes as reflective of calcium and vitamin D requirements is long-standing. Bone health served as an indicator for determining the calcium and vitamin D DRIs in 1997, when nutrient reference values for these nutrients were last reviewed (IOM, 1997). Since that time, additional studies have added to the scientific understanding of the relationships between calcium and vitamin D and bone health and are reflected to a large extent in AHRQ-Ottawa and AHRQ-Tufts, completed in 2007 and 2009, respectively.

Bone health is a concern throughout the life span. Initially, it comprises skeletal development during the times of gestational development and growth in infancy, childhood, and adolescence; this is followed by bone maintenance in adulthood. Menopause and aging result in bone loss. Various measures and health conditions are relevant to considerations of bone health; these include BMC/BMD, calcium balance, rickets/osteomalacia, and fracture risk. The latter is particularly germane to older adults. Further, while not health outcomes, measures of serum 25OHD concentrations and of circulating PTH levels have been incorporated into

studies as intermediates related to bone health. Finally, although physical performance and the incidence of falls are defined by some as a component of bone health, these measures are reviewed separately in this report.

Compared with other potential indicators, this health outcome is characterized by a sizable number of RCTs as well as numerous observational studies from large cohorts. However, many of the studies evaluated calcium and vitamin D in combination, and there are relatively few studies that have evaluated the effects of calcium alone without vitamin D supplementation or vice versa. Given the nature of the available data and the need to integrate information to develop a set of measures reflective of bone health as a potential indicator for developing DRIs, this section is organized differently from those for other potential indicators. The understandings that link calcium and vitamin D to bone health and provide the basis for biological plausibility for an effect on bone health have been described in Chapters 2 and 3 and are therefore not repeated here. The many observational studies are briefly summarized.

Given the depth and breadth of the available AHRQ systematic analyses for the topic of bone health, the AHRQ analyses are considered in a single section. The two AHRQ analyses systematically reviewed, first, the published literature on the relationship between bone health and vitamin D (often in combination with calcium) (AHRQ-Ottawa) and, second, the relationship between bone health and vitamin D alone or vitamin D in combination with calcium (AHRQ-Tufts). Neither of the two AHRQ analyses considered calcium alone in relation to bone health. These two analyses have been described at the beginning of this chapter, and specific information about the studies included in AHRQ can be found in Appendixes C and D. Relevant information has been summarized and included in the tables presented below.

AHRQ included only minimal information about reproductive outcomes, and therefore the literature related to skeletal health during pregnancy and lactation is highlighted. AHRQ also included only minimal information about PTH level, a measure some investigators relate to bone health, so PTH level as a potential measure for bone health is examined separately.

The final component of this section integrates the available data on the basis of bone accretion, bone maintenance, and bone health. The preliminary step of specifying the utility of serum 25OHD level as a marker as well as the relationship between calcium absorption and serum 25OHD level provides the opening discussions for the integration section.

The observational studies surrounding bone health are myriad. However, as is the case with the evidence hierarchy, the basis for the relation-

ship between the two nutrients and bone health is more appropriately explored by examining evidence from controlled interventions, although data from observational studies can lend support and offer confirmatory input. Observational data regarding calcium intake and bone health are mixed regarding the finding that a range of increasing calcium intakes above deficiency levels are associated with improved bone mass and reduced fracture risk. These studies are confounded by an array of variables that have an impact on measures of bone density.

Regarding serum 25OHD concentrations and bone health, the AHRQ-Ottawa analysis concluded that observational studies suggested a correlation between higher serum 25OHD concentrations and increased BMC for older children and adolescents. For postmenopausal women and elderly men, observational studies reviewed in AHRQ-Ottawa provided fair evidence to support an association between serum 25OHD level and BMD or changes in BMD at the femoral neck. This analysis noted that the observational data overall were discordant with the results from available RCTs. Newer observational studies for the most part are consistent with older observational studies with respect to a relationship between low serum 25OHD levels and outcomes such as bone loss, fractures, or osteomalacia (Cauley et al., 2008; Looker and Mussolino, 2008; van Schoor et al., 2008; Ensrud et al., 2009; Bolland et al., 2010b; Cauley et al., 2010; Melhus et al., 2010). However, there are confounders related to such studies, including age, calcium intake, and social situation.

Bone mineral content/bone mineral density: Serum 25OHD AHRQ conducted its analyses for serum 25OHD concentrations on the basis of certain age and gender groups, as presented below.

Infants Overall, AHRQ-Ottawa, for which some studies included combinations of calcium and vitamin D, has reported that there is inconsistent evidence for an association between serum 25OHD concentrations and BMC measures in infants. Of the two RCTs examining BMC (Greer et al., 1982; Zeghoud et al., 1997), one demonstrated no significant benefit of higher serum 25OHD concentrations on radial bone mass, whereas the other showed a transient increase of BMC compared with the unsupplemented group at 12 weeks, but not at 26 weeks. Based on case–control studies (Okonofua et al., 1986; Bougle et al., 1998; Namgung et al., 1998; Park et al., 1998), greater whole-body BMC was related to higher serum 25OHD levels. Data are summarized in Table 4-5. AHRQ-Tufts found no additional RCTs for infants published in the period since the completion of the AHRQ-Ottawa review.

Children and adolescents For children and adolescents, there was fair evidence from AHRQ-Ottawa of an association between serum 25OHD levels and baseline BMD and change in BMD or BMD indexes. However, the results from the RCTs (Ala-Houhala et al., 1988; El-Hajj Fuleihan et al.,

2006) did not confirm a consistent benefit on BMD or BMC across skeletal sites and age groups. Some studies included combinations of calcium and vitamin D.

There were seven studies in older children and adolescents (two RCTs, three cohort studies, one case–control study, and one before-and-after study) that evaluated the relationship between serum 25OHD concentrations and BMC or BMD (see Table 4-6). In older children, there was one RCT, one prospective cohort study, and one before-and-after study. The RCT (Ala-Houhala et al., 1988) did not find an association between serum 25OHD concentrations and distal radial BMC. Two of three non-RCT studies found a positive association between baseline serum 25OHD concentrations and BMC or BMD. The effect of bone size and muscle mass on these outcomes in relation to baseline serum 25OHD concentrations was not reported.

One RCT with children and adolescent girls (El-Hajj Fuleihan et al., 2006) demonstrated a significant relationship between baseline serum 25OHD concentrations and baseline BMD of the lumbar spine, femoral neck, and radius. However, only high dose supplementation with 14,000 IU of vitamin D₃ per week increased BMC of the total hip.

AHRQ-Tufts identified two RCTs available after the AHRQ-Ottawa analysis, both rated C, that compared the effect of vitamin D supplementation alone on BMC in healthy girls between 10 and 17 years of age (El-Hajj Fuleihan et al., 2006; Andersen et al., 2008). Both RCTs were rated C because the results were not adjusted for important potential confounders, such as height, bone area, lean mass, sun exposure, and pubertal status. One RCT (Andersen et al., 2008) analyzed 26 healthy girls, who were Pakistani immigrants primarily living in the Copenhagen area of Denmark (latitude 55°N). Girls were randomly assigned to receive either a daily dose of 400 or 800 IU of vitamin D₃ or placebo for 1 year. The mean baseline dietary calcium intake was 510 mg/day, and the serum 25OHD concentration was 11 nmol/L. At the end of the study, there were no significant differences in whole-body BMC changes between groups receiving the two doses of vitamin D₃ (400 or 800 IU/day) and the placebo group. A second RCT (El-Hajj Fuleihan et al., 2006) analyzed 168 healthy girls living in the Greater Beirut area, Lebanon (latitude 33°N). Girls were randomly assigned to receive either weekly oral vitamin D doses of 1,400 IU (equivalent to 200 IU/day) or 14,000 IU (equivalent to 2,000 IU/day) or placebo for 1 year. The mean baseline dietary calcium intake was 677 mg/day, and the 25OHD concentration was 35 nmol/L. At the end of the study, there were no significant differences in whole-body BMC changes between either the low-dose vitamin D group (200 IU/day) or the high-dose vitamin D group (2,000 IU/day) and the placebo group. The same findings were seen when analyses were restricted to either premenarcheal or postmenarcheal girls.

Postmenopausal women and elderly men Overall, regarding serum 25OHD and bone density measures, AHRQ-Ottawa, which included some studies that combined calcium and vitamin D, reported discordance between the results from RCTs and the majority of observational studies; the authors attributed this as likely due to the impact of confounders relative to observational data as a general matter. Nineteen studies (see Table 4-7) evaluated the association between serum 25OHD levels and BMD. Of these, six studies were RCTs. One RCT (Ooms et al., 1995b) reported an association between serum 25OHD concentrations and BMD or bone loss, whereas the other five RCTs (Dawson-Hughes et al., 1995; Storm et al., 1998; Schaafsma et al., 2002; Cooper et al., 2003; Aloia et al., 2005) and three cohort studies did not. Four cohort studies found a significant association between 25OHD concentrations and bone loss, which was most evident at the hip sites, but the evidence for an association between 25OHD concentrations and lumbar spine BMD was weak. Six case–control studies suggested an association between 25OHD concentrations and BMD, and the association was most consistent at the femoral neck. A forest plot showing the effect of vitamin D plus calcium supplementation (versus placebo) for femoral neck BMD at 1 year is shown in Figure 4-4. Overall, significant increases at the femoral neck were observed with a combined estimate as reported in Table 4-7 of 1.37 percent (95% confidence interval [CI]: 0.24–2.50) from three trials after 1 year.

Based on the results from the observational studies, there is fair evidence to support an association between serum 25OHD levels and BMD or changes in BMD at the femoral neck. Specific circulating concentrations of 25OHD below which bone loss at the hip was increased ranged from 30 to 80 nmol/L.

AHRQ-Tufts identified two more recent RCTs, one that combined calcium with vitamin D and one that did not. The first, an A-quality RCT (Zhu et al., 2008a), compared the effect of vitamin D₂ supplementation on hip BMC in 256 elderly women between 70 and 90 years of age. All elderly women in this trial had normal physical functioning. They were randomly assigned to receive either vitamin D₂ (1,000 IU/day) plus calcium (1,200 mg/day) supplement or calcium (1,200 mg/day) supplement alone for 1 year. The mean baseline dietary calcium intake was 1,097 mg/day, and the mean serum 25OHD concentration was 44.3 nmol/L. Total hip BMD increased significantly in both groups, with no difference between the vitamin D₂ plus calcium and calcium alone groups (hip BMD change: vitamin D, +0.5 percent; control, +0.2 percent).

The second, a B-quality RCT (Andersen et al., 2008), analyzed 89 healthy adult women and 83 healthy adult men separately. The participants were Pakistani immigrants living in the Copenhagen area of Denmark (latitude 55°N). Women and men were randomly assigned to receive either a daily dose of 400 or 800 IU vitamin D₃ or placebo for 1 year. For women,