Question 1A (Part 1). Are There Specific Concentrations of Serum 25(OH)D That Are Associated With Established Vitamin D Deficiency Rickets in Infants and Young Children?

Overview of Relevant Studies

For the purposes of this review, infancy is defined as term birth to 12 months, and young children from one to five years of age. Studies that enrolled older children were included if the majority of children were in the above age groups. For studies on established rickets in infants and young children, 13 studies met our inclusion criteria and assessed the association between serum 25(OH)D and rickets.⁷⁷–⁸⁹ Of the 13 studies, there was one RCT,⁷⁷ four before-after studies⁷⁸–⁸¹ and eight case-control studies.⁸²–⁸⁹ For the RCT, bone health outcomes included improvement in the signs and symptoms of rickets, and serum PTH levels.⁷⁷ The twelve observational studies included rickets as the bone health outcome,⁷⁸–^{84, 84}–⁸⁹ and seven of the 12 studies included assessment of serum PTH,^{78, 79, 82, 84, 87, 88} as summarized in Table 1. In all studies, children were diagnosed with rickets using clinical and radiological criteria. No studies included BMD, BMC, or fractures as outcomes.

Table 1

Serum 25(OH)D Levels in Established Rickets in Infants and Young Children.

Study characteristics including country and type of vitamin D assay are summarized in the Table 1. All studies except for one case-control study with nine participants⁸² were conducted outside of North America. The North American study was conducted at a northern latitude (Canada, U.S. Midwest). Each study examined serum 25(OH)D concentrations at diagnosis and some included followup measurements during treatment.⁷⁸–^{81, 86, 87} Six studies used an RIA assay for serum 25(OH)D assays,^{77, 83}–^{86, 89} six studies used a CPBA method,⁷⁸–^{82, 87} and one study used an HPLC technique.⁸⁸ We report, in this section, baseline measurements at diagnosis or pre-treatment.

Population characteristics. Children with rickets ranged in age from as young as two months up to 14 years, with most children between 24 and 36 months. In the studies that reported ethnicity, virtually all children were non-white except for two subjects in the one North American study.⁸² The sample sizes ranged from nine⁸² to 123 participants,⁸⁴ with an average of 41. In 12 of the 13 studies, gender was mixed.

Outcome characteristics. For all studies, the diagnosis of rickets was ascertained by radiographic and clinical evidence.⁷⁷–^{87, 89} Serum PTH was measured in seven studies using either RIA or chemiluminescent immunoassays.^{78, 79, 82, 84, 87}–⁸⁹ No study evaluated BMC, BMD or fractures.

Study quality. The study quality of the RCT,⁷⁷ four before-after and eight case-control studies ranged from poor to fair with the RCT scoring 1/5 on the Jadad scale (in relation to randomization for treatment).

Qualitative synthesis of individual study results. Six studies reported a mean^{77, 78, 80, 85} or median^{79, 88} serum 25(OH)D concentration < 27.5 nmol/L associated with rickets. These studies included measurements by RIA,^{77, 85} CPBA⁷⁸–⁸⁰ or HPLC.⁸⁸ Five studies reported that children with rickets had a mean 25(OH)D concentration above 27.5 nmol/L (range of means 36 – 50 nmol/L),^{82, 84, 86, 87, 89} and the other two studies reported at least some children with serum levels above this value.^{81, 83} While 25(OH)D assays differed across the studies, these results suggest that the serum 25(OH)D concentration associated with rickets may be much higher than previously thought. In one study, deficient dietary calcium was the etiology for rickets⁸³ whereas in another study, a mean dietary calcium intake of < 300 mg/d did not alter the Odds Ratio (OR) for rickets.⁸⁴ Given the uncertainty of the dietary calcium measurement, it remains unclear whether the specific concentration of serum 25(OH)D consistent with rickets is confounded by dietary calcium.

In the studies that reported serum PTH, values in children with rickets were elevated above the normal range.^{78, 79, 82, 84, 87, 89} One study confirmed a negative relation of PTH with 25(OH)D concentrations (r = -0.70),⁸² when cases and controls were analyzed together.

The majority of studies included in this review were from developing countries where dietary calcium intake is low. Low dietary calcium can confound 25(OH)D status and is a major limitation of the studies since some cases of rickets may be attributable to a calcium deficiency. Another limitation is the paucity of studies in children with rickets in North America. The specific concentrations of serum 25(OH)D associated with rickets in North America is uncertain, given the lack of studies in populations with dietary calcium intake similar to North American diets, as well as the different methods used to determine 25(OH)D concentrations. A better understanding of the inter-relationship between 25(OH)D concentrations, calcium and rickets would improve the specific values of 25(OH)D to be used as a biomarker in the diagnosis and treatment of rickets. Only studies of established rickets were included, and other RCTs have evaluated specific 25(OH)D concentrations in relation to the development of rickets. In a rickets prevention study in China, Specker et al. found that 25(OH)D concentrations above 30 nmol/L appeared to prevent rickets in infants with or without vitamin D deficiency at birth.⁹⁰

View in own window

Summary. Circulating 25(OH)D levels associated with established rickets in infants and young children

Quantity: Six studies (one RCT, three before-after and two case-control studies) reported mean or median 25(OH)D concentrations < 30 nmol/L in children with rickets whereas the other studies reported mean or median values above 30 nmol/L and up to 50 nmol/L. In seven of eight case-control studies, serum 25(OH)D values were lower in the children with rickets compared with controls.

Quality: The study quality of the RCT, four before-after and eight case-control studies ranged from poor to fair (with the RCT scoring 1/5 on the Jadad scale).

Consistency: There is fair evidence for an association between low serum 25(OH)D and established rickets, regardless of assay type (RIA, CPBA, HPLC). There is inconsistent evidence to determine if there is a threshold concentration of serum 25(OH)D above which rickets does not occur.

Question 1A (Part 2). Are Specific Circulating Concentrations of 25 Hydroxyvitamin D [25(OH)D] Associated With Bone Health Outcomes in Infants?

Overview of Relevant Study Characteristics and Results

Infancy is defined by the Institute of Medicine as including two subcategories: birth to 6 months and 6 to 12 months.⁴ Seven studies included infants 12 months or younger and assessed the association between serum 25(OH)D and bone health outcomes.⁹¹–⁹⁷ Of the studies, there were three RCTs, two in breast-fed infants^{92, 93} and one in formula-fed infants,⁹¹ and four case-control studies.⁹⁴–⁹⁷

For the three RCTs, bone health outcomes included BMC^{92, 93} and serum PTH levels⁹¹–⁹³ (Table 2). No RCTs reported results of BMD or evaluated fracture incidence. Four observational studies reported BMC,⁹⁵–⁹⁷ BMD,^{96, 97} fractures⁹⁴ or PTH (Table 2).⁹⁴–⁹⁶

Table 2

Serum 25(OH)D and Bone Health Outcomes in Infants.

Study characteristics. Of the three RCTs, two were conducted in the U.S.^{92, 93} Both of these trials randomized human milk-fed infants to receive vitamin D₂ supplementation (400 IU/d) or placebo. One U.S. RCT was six months in duration,⁹² and the other was 26 weeks long at which time the placebo group were started on supplementation, and both groups were followed until 52 weeks.⁹³ The RCT by Zeghoud et al. was three months in duration, and randomized infants to receive either 500 or 1000 IU/d D₂.⁹¹ The 25(OH)D assays varied, with two studies using a CPBA method^{91, 93} and one using HPLC.⁹²

None of the four case-control studies were conducted in North America (Table 2). Outcomes were assessed at birth in three studies^{94, 95, 97} and at two to five months of age in the other.⁹⁶ One study measured circulating 25(OH)D by CPBA,⁹⁴ two studies used HPLC,^{95, 96} and the fourth study⁹⁷did not report the method.

Population characteristics. For the three RCTs, the age at enrolment was within a few days of birth.⁹¹–⁹³ The sample sizes ranged from 18 to 80 infants, without a predominance of male or female gender. In all three studies,⁹¹–⁹³ participants had to be healthy and free of conditions known to affect calcium metabolism. Mean vitamin D and calcium intake were not reported in any of the studies, although maternal behavior related to breast feeding was reported in all studies. Baseline 25(OH)D concentrations are summarized in Table 2.

For the case-control studies, three studies evaluated infants at birth or within the first few days of birth,^{94, 95, 97} and one study evaluated infants at two to five months of age.⁹⁶ The sample sizes ranged from 21 to 82 infants with sub-categorization as to ethnicity,⁹⁴ term born,⁹⁷ season of birth,⁹⁵ or feeding type.⁹⁶ In all case-control studies, participants had to be healthy and free of conditions known to affect calcium and bone metabolism. Data on dietary vitamin D or calcium intake plus exposure to sunshine were only relevant for the study that evaluated two to five month old infants,⁹⁶ and these data were not reported.

Covariate/confounders. No relevant covariates or effect modifiers were controlled for in the RCTs. In one RCT, baseline 25(OH)D concentrations were used to divide the study cohort into three subcategories⁹¹ (Table 2). Seasonal effects were examined in one study.⁹² For case-control studies, matching on gestational age at birth and gender was not reported. Only one study adjusted for weight when evaluating the relation between 25(OH)D and whole body BMC.⁹⁵

Outcome characteristics. For the RCTs, BMC of the distal radius was measured by single photon absorptiometry,^{92, 93} and PTH was measured using RIA.⁹¹–⁹³

For the case-control studies, BMC (whole body or spine) and BMD were measured using dual-energy x-ray absorptiometry (DXA).⁹⁵–⁹⁷ PTH was measured using RIA techniques.⁹⁴–⁹⁶ Although all studies used RIA techniques to measure PTH, these may have varied in antibody specificity and measurement of PTH fragments.⁹⁸

One case-control study reported fracture incidence⁹⁴ although the methodology was not reported.

Study quality. For the RCTs, one trial each scored 1/5,⁹¹ 3/5⁹³ and 4/5⁹² on the Jadad scale. The four case-control studies were of fair quality.

Qualitative synthesis of individual study results. Of the two RCTs measuring BMC of the distal radius, one study showed transient elevation in BMC at 12 weeks of age in the supplemented group (with serum 25(OH)D concentrations of 95 nmol/L) compared to the placebo group (with 25(OH)D concentrations of 50 nmol/L).⁹³ However, by 26 weeks there was no significant difference in BMC between the placebo and vitamin D₂ supplemented infants who continued to have higher serum 25(OH)D levels. In a second trial by Greer,⁹² no difference in BMC was observed at 3 months in vitamin D₂ supplemented or unsupplemented human milk-fed infants despite 25(OH)D concentrations of 97 nmol/L in the intervention group compared to 39 nmol/L in the control group. At six months, the control group had higher absolute BMC and was also noted to have higher levels of the (unsupplemented) D₃ isoform. However, the change in BMC from 1.5 to 6 months was not significantly different in the two groups.

Two case-control studies measured BMC and BMD of the lumbar spine (L1–4).^{96, 97} One study observed a negative correlation between 25(OH)D (levels ranging from 10 to 292 nmol/L) and spine BMC and BMD at birth but no relation was observed in regression analyses that included postnatal age and serum calcium.⁹⁷ The other study⁹⁶ did not find a difference in spine BMC at two to five months of age when a group of human milk-fed infants with an average 25(OH)D serum level of 40 nmol/L were compared with a group of formula-fed infants with an average 25(OH)D of 73 nmol/L. 8/18 infants in the human milk-fed group and 1/17 in the formula-fed group had a serum 25(OH)D level < 28 nmol/L; there was no correlation of BMC with serum 25(OH)D concentration. The one study that measured whole body BMC reported a positive relation between 25(OH)D and BMC.⁹⁵ The values for 25(OH)D in this study were on average 27 nmol/L for winter born and 75 nmol/L for summer born who had eight percent higher whole body BMC at birth.

Overall, for BMC measurements reflecting mainly cortical bone, including whole body and radial assessments, two of three studies showed a positive association between 25(OH)D concentrations with BMC, one measuring whole body BMC and one showing a transient increase in distal radial BMC at 12 but not 26 weeks.^{93, 95} Of the two studies examining predominantly trabecular bone (lumbar spine),^{96, 97} one showed a negative correlation between 25(OH)D and BMC and BMD at birth that was not evident after using multiple regression; ⁹⁷ the other did not demonstrate any association.

Of the two RCTs reporting PTH levels, one study did not observe differences in PTH between vitamin D₂ supplemented and non supplemented infants at 1.5 to six months of age.⁹² Both groups were characterized by mean serum 25(OH)D levels above 30 nmol/L (measured by HPLC). At all timepoints, 25(OH)D values were higher in the supplemented group (range of means from 75.6 to 97.2 nmol/L compared to means of 39.4 to 58.8 nmol/L in the unsupplemented group). In the other RCT, PTH declined in all groups from birth to three months of age while 25(OH)D concentrations increased to at least 46 nmol/L (measured by CPBA).⁹¹ In that study, all neonates who had abnormally high PTH had serum 25(OH)D < 30 nmol/L. In a case-control study, serum PTH was not different among winter and summer born infants with mean serum 25(OH)D of 27 and 75 nmol/L respectively (measured by HPLC).⁹⁵ Similarly, human milk-fed infants with a mean 25(OH)D concentration of 40 nmol/L did not have different serum PTH values than formula-fed infants with a mean 25(OH)D concentration of 73 nmol/L (measured by HPLC).⁹⁶ Lastly, Asian infants had significantly higher PTH concentrations and lower 25(OH)D concentrations of 5 to 20 nmol/L (mean 6, SD 4) when compared to Caucasian infants characterized by serum 25(OH)D concentrations of 9 to 39 nmol/L (mean 15, SD 5) (measured by CPBA).⁹⁴ Overall, these five studies suggest that PTH is inversely associated with serum 25(OH)D concentrations at lower 25(OH)D concentrations but there was inconsistent evidence for a threshold that may exist somewhere above 27 nmol/L (measured by CPBA). Variable evidence for a threshold may be in part due to the different assays used, both to measure serum PTH and serum 25(OH)D.

Of the studies examining a relation between 25(OH)D and bone health outcomes, most had small sample sizes and the baseline 25(OH)D was variable ranging from deficient values around the limitation of detection to values above 27 nmol/L. In studies with repeated measurements, the baseline 25(OH)D was not considered as an effect modifier in evaluating the relation between 25(OH)D and bone health outcomes. The three included RCTs used vitamin D₂ supplementations and therefore conclusions cannot be drawn regarding supplementation with the D₃ isoform. Lastly, a definitive conclusion as to whether a specific concentration of 25(OH)D is associated with an elevated PTH (secondary hyperparathyroidism) is not possible given the evidence put forth to date. Additional studies are required to define a threshold concentration of 25(OH)D below which serum PTH levels rise. This will require not only standardization of 25(OH)D assays but also PTH assays.⁹⁸

View in own window

Summary. Serum 25(OH)D levels and bone health outcomes in infants

Quantity: Of the two RCTs examining BMC, one demonstrated no benefit of higher serum 25(OH)D on radial bone mass while the other showed a transient increase of BMC compared to the unsupplemented group at 12 weeks but not 26 weeks. Of the three case-control studies, whole body BMC was positively related to and lumbar spine negatively related to serum 25(OH)D concentrations. Based on two RCTs and three case-control studies, a rise in PTH was either not observed with 25(OH)D concentrations above 27–30 nmol/L or occurred at a lesser rate than at lower values, suggesting a threshold value may exist somewhere above 27 nmol/L.

Quality: The three RCTs were of fair to high quality (two of the three RCTs had a Jadad score of ≥ 3/5) and the four case-control studies were of fair quality.

Consistency: There is inconsistent evidence for an association between a specific concentration of serum 25(OH)D and the bone health outcome BMC in infants. Overall, there is fair evidence that PTH is inversely associated with serum 25(OH)D concentrations at lower 25(OH)D concentrations, but there was inconsistent evidence for a threshold that may exist somewhere above 27 nmol/L (measured by CPBA).

Question 1A (Part 3). Are Specific Circulating Concentrations of Serum 25 Hydroxyvitamin D [25(OH)D] Associated With Bone Health Outcomes in Older Children and Adolescents?

Definition of study populations. The Institute of Medicine defines early childhood as ages 4 though 8 years, and puberty/adolescence as ages 9 through 13 years, and 14 through 18 years.⁴ Grouping by age for the purpose of this report were based on the study populations. In this section, children six years of age or older who had not yet entered puberty were included, and adolescence (marked by the onset of puberty) was defined by the presence of at least Tanner Stage 2 for sexual development.⁹⁹ The age groups in the included studies for this section were: 6–10 years,¹⁰⁰ age 9 years,¹⁰¹ 8 – 10 years,¹⁰² 9 –15 years,¹⁰³ 15–16 years,¹⁰⁴ 10 – 17 years,¹⁰⁵ and 10 – 18 years.¹⁰⁶

Study characteristics. Three studies that included older children (one RCT,¹⁰² one prospective cohort¹⁰¹and one before-after study¹⁰⁰) assessed the association between serum 25(OH)D concentrations and bone health outcomes.

Four studies in adolescents assessed the association between 25(OH)D levels and bone health outcomes.¹⁰³–¹⁰⁶ There were two cohort studies,^{103, 104} one case-control study¹⁰⁶ and one RCT.¹⁰⁵ The first cohort evaluated the association between serum 25(OH)D levels and lumbar spine and femoral neck BMD/bone mineral apparent density (BMAD) at baseline and 3 years.¹⁰³ The second cohort study evaluated the seasonal variation in serum 25(OH)D concentrations and its relation to intact (i) PTH levels over an 18 month period.¹⁰⁴ El Hajj Fuleihan¹⁰⁵ evaluated the effect of low (1,400 IU/week) and high (14,000 IU/week) dose vitamin D₃ on areal BMD and BMC of the lumbar spine, hip, forearm, and total body and body composition. Marwaha¹⁰⁶ evaluated 25(OH)D concentrations in 5,137 children and adolescents (aged 10–18 years) from Northern India and the association with serum PTH, ionized calcium and BMD of the forearm and calcaneus, with stratification by upper and lower socioeconomic status.

Bone health outcomes - ascertainment. For the studies on older children, PTH was measured by an immunoradiometric assay that detects the mid-region of the molecule,¹⁰² and distal radial BMC was measured by single-photon absorptiometry (SPA).¹⁰² Javaid¹⁰¹ measured whole body and lumbar spine BMC and areal BMD by DXA, and calculated an apparent volumetric BMD at nine years of age in relation to maternal third trimester 25(OH)D status. Rajakumar¹⁰⁰ evaluated the association between serum 25(OH)D concentrations, serum PTH and markers of bone turnover.

For adolescents, lumbar spine BMD, femoral BMD, and lumbar spine bone mineral apparent density (BMAD) was measured by DXA¹⁰³ and iPTH by immunoradiometric assay.¹⁰⁴ Fuleihan measured areal BMD and BMC at the lumbar spine, hip and forearm, and total body and lean body mass by DXA.¹⁰⁵ Marwaha¹⁰⁶ evaluated forearm and calcaneal BMD using peripheral DXA and PTH with an immunoradiometric assay.

There were no studies that assessed the association between serum 25(OH)D concentrations and fractures in older children or adolescents.

For assessment of 25(OH)D levels, different methods were used depending on the study. These included radioimmunoassay or radioimmunometric methods in three studies,^{101, 103, 106} and CPBA in three studies.^{100, 104, 105}

Population characteristics. For older children, ages ranged from eight to ten years in two studies with mixed gender.^{101, 102} Included subjects were aged 6 – 10 years in the Rajakamar study who exhibited a combination of pre- and early pubertal status (33/42 pre-pubertal Tanner stage I).¹⁰⁰ Eligibility criteria for two studies required that participants be healthy, without co-morbidities.^{100, 102} The prospective cohort study by Javaid did not state whether children with co-morbidities were excluded. The mean dietary intake of calcium/vitamin D was reported in two studies.^{100, 101}

For adolescents, subjects ranged in age from nine to 16 years.¹⁰³–¹⁰⁶ All patients were at least Tanner Stage 2 for pubertal development with the exception of the Marwaha study which did not report pubertal status. However, the patients in the latter study were 10–18 years of age and it is anticipated that the majority were at least Tanner Stage 2 puberty. The studies involved either female,^{103, 105} male,¹⁰⁴ or mixed genders.¹⁰⁶ Participants were reported as healthy, without known co-morbidities, in two of four studies.^{103, 104} The mean dietary intake of calcium/vitamin D was reported in three studies.^{100, 103, 104} Additional characteristics are summarized in Table 3.

Table 3

Serum 25(OH)D Levels and Bone Health Outcomes in Older Children and Adolescents.

Confounders/effect modifiers. In the studies on older children, Javaid adjusted for the age of the child at the time of the BMC measurement due to the strong association between age and whole body BMC.¹⁰¹ Since bone size can affect the BMD results, volumetric BMD at the lumbar spine was calculated. For adolescents in the 25(OH)D-BMC/BMD cohort study,¹⁰³ adjustments were made for the time to followup, and regression analyses were performed to determine covariates for BMD and BMC. El-Hajj Fuleihan¹⁰⁵ made adjustments for lean mass and bone area, and did exploratory subgroup analyses on pre and post menarcheal girls in their analysis of vitamin D status in relation to BMD and BMC. Marwaha¹⁰⁶ adjusted BMD for both height and weight.

Study quality. On the Jadad scale, one RCT scored 3/5¹⁰² and one scored 4/5¹⁰⁵ indicating both were of high quality. The overall study quality for the observational studies was fair. Limitations included failure to adjust for relevant confounders or other sources of bias, and higher numbers of participants lost to followup.

Qualitative synthesis of individual study results. In a study of pre-pubertal Finnish girls, 400 IU vitamin D₂, increased serum 25(OH)D levels (measured by RIA) compared with placebo but did not impact mid-region PTH or distal radial BMC (SPA) after 13 months.¹⁰² Radial BMC was not adjusted for bone size in this study.

In the before-after study by Rajakumar,¹⁰⁰ baseline vitamin D status (measured by CPBA with deficiency defined as a serum 25(OH)D < 25 nmol/L (10 ng/ml) and insufficiency defined as ≤ 50 nmol/L) was negatively correlated with PTH (but not associated with baseline serum calcium, phosphorus, albumin, or 1,25-(OH)₂D). Serum PTH remained stable at levels of 25(OH)D around 75 nmol/L. There were no significant differences between the vitamin D insufficient and sufficient groups with regard to gender, weight, height, BMI and skin pigmentation. The mean (SD) daily dietary vitamin D intake was 277 (146) IU (mean intakes of 233 in the insufficiency group and 318 IU in the sufficient group were not significantly different). Dietary calcium intake was significantly higher in the sufficient group.

Javaid¹⁰¹ reported that low serum 25(OH)D concentrations (measured by RIA) in mothers during late pregnancy were weakly but significantly associated with reduced whole body (r = 0.21, p<0.01) and lumbar spine (r = 0.017, p = 0.03) age-adjusted BMC (DXA-Lunar DPX-L). Bone mass in children of mothers who were vitamin D deficient (25(OH)D < 28 nmol/L) during pregnancy was significantly lower compared to children born to vitamin D sufficient mothers. Reduced umbilical venous calcium also predicted reduced childhood bone mass (p = 0.0286). Whether this observation is mediated, totally or in part, through an effect on bone size and/or muscle mass is not clear. Maternal vitamin D status was positively associated with whole body and spine BMC in the offspring, and neither childhood height nor lean mass was associated with maternal 25(OH)D levels. Adjustment for childhood height did not significantly weaken the relation between maternal vitamin D status and whole body BMC. In contrast, volumetric BMD of the lumbar spine (which corrects for bone size) was not associated with maternal vitamin D status. Milk intake and physical activity at age nine were not significant determinants of bone mass although these findings do not rule out the possibility that factors such as UV exposure, diet and other lifestyle characteristics may have affected bone mass. When socioeconomic status was adjusted for, it did not change the association substantially. The type of postnatal feeding in the first three months also did not affect bone mass.

For girls age 9 – 15 years, the three year cohort study (N = 171) by Lehtonen-Veromaa evaluated the relation between baseline 25(OH)D levels (measured by RIA) and the change in lumbar spine (r = 0.35, p < 0.001) and femoral neck BMD (r = 0.32, p < 0.001). Baseline 25(OH)D also correlated with the change in LS BMAD (size-corrected form of BMD) (r = 0.35, p < 0.001) and FN BMAD (r = 0.24, p < 0.002). The difference in the percent increase from baseline in lumbar spine BMD (adjusted for the followup period) between those with low 25(OH)D levels (<20 nmol/L) and those with higher 25(OH)D levels was four percent. The difference in lumbar spine BMD was 12.7, 13.1 and 16.7 percent for the lowest, middle and highest 25(OH)D tertiles, respectively.¹⁰³

In another cohort (N = 175) of French teenage boys, there was a significant negative correlation between serum iPTH and 25(OH)D levels (measured by CPBA), with a plateau in PTH demonstrated at 25(OH)D levels of 83 nmol/L and above.¹⁰⁴ At this level of 25(OH)D, the iPTH reached a plateau at 2.48 pmol/L.

El-Hajj Fuleihan¹⁰⁵ found a significant association between baseline serum 25(OH)D levels (measured by CPBA) and baseline BMD at the lumbar spine (r=0.16, p=0.033), femoral neck (r = 0.17, p = 0.028), and radius (r = 0.24, p = 0.002) (DXA-Hologic 4500). There was also a significant association between baseline serum 25(OH)D levels and baseline radius BMC (r = 0.16, p = 0.033). The mean baseline serum 25(OH)D was 35 nmol/L (14 ng/ml). In post hoc analyses, there were negative correlations between baseline serum 25(OH)D levels and percent change in lumbar spine BMD (r = -0.16, p = 0.044) or subtotal body BMD (r = -0.20, p = 0.009) over one year. Significant negative associations were found between baseline serum 25(OH)D levels and percent change in spine, femoral neck and radius BMC.

After vitamin D supplementation for one year, total hip BMC increased in the high dose (14,000 IU/wk) group (pre- and post-menarcheal girls combined) but there were no significant changes in BMC or BMD at other skeletal sites. In an exploratory subgroup analysis in pre-menarcheal girls alone (N = 34), total body lean tissue mass increased in both supplementation groups. Lumbar spine areal BMD was significantly increased in the low dose (1,400 IU/wk) group, and trochanter BMC was increased in both the high and low dose groups. The magnitude of the treatment effect was not significant after adjusting for both bone area and lean tissue mass. The authors acknowledge a limitation of DXA in evaluating areal BMD and BMC is the lack of consensus on how best to adjust for bone size. In postmenarcheal girls, there were no differences in changes in lean mass, BMD or BMC amongst the three groups. In boys (data not shown), the authors reported there was no consistent positive effect of vitamin D supplementation on lean mass, BMD or BMC.

Marwaha¹⁰⁶ showed that children with a lower socioeconomic status had significantly lower 25(OH)D concentrations (measured by RIA) and mean BMD (unadjusted for bone size) for the forearm and calcaneus (DXA-PIXI-1.34) was higher in the upper socioeconomic group. There was a significant negative correlation between serum immunoreactive PTH and 25(OH)D concentrations (r = -0.202, p < 0.001). PTH concentrations only increased at 25(OH)D concentrations below 12.5 nmol/L. There was no significant correlation between the mean serum concentration of 25(OH)D and BMD in both groups.

Overview of Relevant Study Characteristics and Results

Five observational studies evaluated the association between vitamin D status and bone health outcomes in mothers, or their offspring. One prospective study¹⁰¹ involved the analysis of the bone status by DXA at nine years of age in 198/596 previously studied offspring and the results of this study are summarized in the section on children (Section 1A part 3). The remaining four studies provided data on changes in vitamin D status during pregnancy, and the effect of maternal vitamin D status during pregnancy on outcomes of birth gestation or size. All studies included serum 25(OH)D measurements and other markers of calcium homeostasis. Study characteristics and 25(OH)D assays are outlined in Table 4.

Table 4

Serum 25(OH)D Levels and Bone Health Outcomes in Pregnant or Lactating Women.

The time of assessment of vitamin D status, the assay method for 25(OH)D and bone health outcomes varied across studies which precluded quantitative synthesis of results.

Qualitative Synthesis of Individual Study Results

Maternal vitamin D status. In the study of non-European minority women from South Wales,¹¹¹ 50 percent of the women were vitamin D deficient at the first antenatal visit, using a criterion of serum 25(OH)D < 20 nmol/L. Vitamin D supplementation (800–1600 IU) D during pregnancy normalized vitamin D status in 60 percent of the deficient group. In the study in Saudi Arabia of 40 Asian women,¹⁰⁹ serum 25(OH)D declined significantly from baseline (about 11 weeks gestation) to the third trimester (mean of 31.4 wk of gestation) and remained low through to 6 weeks post-delivery. However, at all timepoints, mean serum 25(OH)D concentrations were within the normal range of a reference group of non-pregnant women (N = 280) who were healthy and non-lactating, suggesting that although serum levels decline during the end of the third trimester, they do not differ extensively from those of the non-pregnant state. None of the pregnant women were classified as having subclinical vitamin D deficiency (25(OH)D < 20 nmol/L). In the study¹¹⁰ in primarily Caucasian women in Australia, serum 25(OH)D was similar at recruitment (11 weeks of gestation) and at the beginning of the third trimester of pregnancy (28–32 weeks of gestation) but there were significant differences between mean values in winter versus summer months. The percent who were vitamin D deficient (9–10 percent as defined by 25(OH)D < 28 nmol/L) was significantly greater in winter than summer.

One cohort study assessed vitamin D status postpartum and in relation to breast-feeding.¹⁰⁸ There was a non-significant trend to a decline in vitamin D status in the initial 2–4 months and the pattern was not influenced by the season of birth. Vitamin D status was not influenced by the duration of breast-feeding. The percent of women who were vitamin D deficient was not provided but based on the mean values, some of the women would have had 25(OH)D values less than 20 nmol/L. Data on vitamin D intake or sun exposure were not provided.

Vitamin D status and bone health outcomes. In the cohort study by Sowers, bone mineral density of lumbar spine and femoral neck was measured in 115 mothers with different breast-feeding practices during the postpartum period and vitamin D status was not associated with changes in BMD of the femur or spine.¹⁰⁸ Women were recruited during the third trimester, lumbar spine BMD was measured at two weeks, 6, 12 and 18 months postpartum and femoral neck at two weeks, two, four, six, 12 and 18 months. Serum PTH and the other calciotrophic hormones were not associated with changes in femoral or lumbar spine BMD, suggesting that 25(OH)D, PTH and 1,25-(OH)₂D do not explain the calcium mobilization and bone turnover that occurs during lactation.¹⁰⁸

In the before-after study in pregnancy,¹¹¹ serum 25(OH)D did not appear to correlate with serum PTH concentrations, with 65/80 women with low 25(OH)D having PTH in the normal range.

In a prospective cohort study on 40 Asian women (280 non-pregnant controls),¹⁰⁹ serum 25(OH)D levels negatively correlated with intact PTH (r = -0.62, 0<0.001). In this study, serum osteocalcin, a bone formation marker was below the reference range observed in non-pregnant women, and declined in the second trimester compared to the first, but then rose to within or above the reference range at term and 6 weeks postpartum. This suggests changes in bone turnover do occur during early pregnancy, irrespective of normal vitamin D status.

In the prospective cohort study by Morley there was no association between baseline maternal 25(OH)D concentrations and measures of infant size at birth.¹¹¹ There was an inverse association between maternal log₂ 25(OH)D and log₂ PTH. Using the maternal 25(OH)D concentrations at 28–32 weeks, the mean gestational length was significantly shorter (0.7 weeks, 95% CI -1.3,-0.1 weeks) in the vitamin D-deficient mothers compared to mothers with 25(OH)D concentrations over 28 nmol/L. This association was not altered by inclusion of log₂ PTH, serum calcium and albumin concentrations. Infants born to mothers who were vitamin D deficient at 28–32 weeks gestation, had lower mean knee-heel length (-2.7 mm) compared to infants born to mothers who were not vitamin D deficient, after adjusting for gestation length.¹¹⁰ Further non-parametric smooth regression analysis and adjustment of confounders suggested the possibility of a linear association when 25(OH)D levels were below 30–40 nmol/L, but there was no association at higher 25(OH)D levels. Low maternal 25(OH)D levels were associated with a negative impact on long bone growth and the authors postulated that maternal PTH may affect fetal growth via an affect on 1,25-(OH)₂D production.¹¹⁰

Study quality. There were no RCTs identified that evaluated the association between serum 25(OH)D concentrations and bone health outcomes in pregnant and lactating women. The before-after study¹¹¹ was poorly designed, lacked detail regarding the duration and compliance with the vitamin D supplements, and the analyses were incomplete. A limitation of the included studies was failure to adjust for all relevant covariates. Only one six-week cohort study was considered to be of good quality, since it included an age-matched non-pregnant cohort with control values for all biochemical measurements (N = 280) and provided six serial measures with no attrition during followup.¹⁰⁹ The cohort study conducted during lactation,¹⁰⁸ was of good quality as it included six serial biochemical measures, four measures of spinal BMD and six of femoral neck BMD throughout lactation, and adjusted for a number of covariates. The one study in which the primary outcome was size of offspring at birth was judged to be of fair quality due to loss of followup of over 20 percent.¹¹⁰

Overview of Relevant Studies

This section summarizes the evidence from the studies that investigated the association between serum 25(OH)D concentrations and bone health outcomes in postmenopausal women and/or elderly men. The discussion focuses on observational studies and only the few (vitamin D supplementation) RCTs that specifically investigated the association of serum 25(OH)D with one or more bone health outcomes are discussed. The majority of RCT data are presented in Question 3. Tables 5–8 summarize the studies included in this section, including the vitamin D assays used.

Table 5

Studies Reporting Serum 25(OH)D Levels and Bone Health Outcomes in Postmenopausal Women and Older Men.

Table 6

Serum 25(OH)D Levels and Fractures in Postmenopausal Women and Older Men.

Table 7

Serum 25(OH)D Levels and Falls and/or Performance Measures in Postmenopausal Women and Older Men.

Table 8

Serum 25(OH)D Levels and BMD/BMC in Postmenopausal Women and Older Men.

For the prospective cohorts, assessment of study quality was based on a number of factors including how representative the cohort was, the method of ascertainment of the outcome, whether key confounders were adjusted for in the analysis, the adequacy of followup, size of the study and whether the main objective was to evaluate the association between serum 25(OH)D and bone health outcomes. For the case-control studies, study quality was evaluated based on whether methods were used to minimize sample bias: for example, similar sampling of cases and controls, matching on relevant variables and the use of population based controls or more than one control group.

Study characteristics. A total of 41 studies (42 records) evaluated the association between serum 25(OH)D concentrations and bone health outcomes in postmenopausal women and elderly men. Of these 41 studies, 10 were RCTs,¹¹²–¹²¹ 14 were single prospective cohorts,¹²²–¹³⁵ and 17 were case-control studies (18 records).^{29, 136}–¹⁵² One publication was companion paper,^{146, 147} and we refer to the primary record with the most relevant data in the results.¹⁴⁶ Study characteristics such as population, sample size, duration of followup, country, and 25(OH)D assays are summarized in Tables 6–8.

Variability in the measurement and reporting of serum 25(OH)D and bone health outcomes, along with differences in populations precluded formal meta-analysis. The results are reported by bone health outcome: fractures, bone mineral density (BMD), falls and performance measures.

Overview of Relevant RCTs

When evaluating the effect of food fortification on circulating 25(OH)D concentrations, it is important to acknowledge the potential confounding effect generated by the food source, the assay used to measure 25(OH)D and potential differences in the bioavailability and/or metabolism of vitamin D₂ versus vitamin D₃. Most studies in this review used dairy products as the source of fortified food. There is potential for study contamination through altered intake of other nutrients such as calcium, phosphate and acid load that can affect bone and mineral homeostasis.

Study characteristics. A total of 13 RCTs, 12 parallel design,^{116, 155}–¹⁶⁵ and one factorial design,¹⁶⁶ studied the effect of dietary sources of vitamin D on circulating 25(OH)D concentrations. Two of the 13 trials did not provide the vitamin D content of the dietary source and were excluded.^{116, 162} Therefore, the following summary includes a total of 11 trials (Table 9).¹⁵⁵–^{161, 163}–¹⁶⁶

Table 9

Serum 25(OH)D Levels and Fortified Foods.

Within the included trials, there were a total of 697 subjects in the vitamin D dietary intervention groups and 584 in the control groups for a total of 1,281 subjects.¹⁵⁵–^{161, 163}–¹⁶⁶

Population characteristics. All trials were in adults. Two trials studied young adults,^{158, 160} one included young women,¹⁶⁴ three involved postmenopausal women,^{155, 157, 159} one included elderly men,¹⁶³ and the remaining four studied elderly individuals of both genders.^{156, 161, 165, 166} Four out of the six trials that included both males and females provided the gender breakdown^{156, 158, 165, 166} and the percentage of females ranged from 51¹⁶⁵ to 83¹⁵⁸ percent. The ethnicity of the study population was reported in four trials, ^{155, 157, 159, 163} and BMI was also reported in four trials.^{155, 163, 164, 166} The vitamin D dietary intake was evaluated at baseline in three trials^{161, 164, 166} and sunlight exposure was assessed in three studies.^{156, 158, 166} The studies did not provide an assessment of skin type of participants. Sunlight exposure was assessed in only three of the 11 trials although several others excluded subjects who had recent or planned exposure to higher-than-usual levels of sunshine. Methods of ascertainment included a sunlight exposure score during the summer in a subsample,¹⁵⁸ the percentage of participants who were outside daily during sunny period and the percentage who avoided sunlight¹⁶⁶ and an outdoor score to reflect the average exposure to sunlight per day per season.¹⁵⁶ Results showed that sunlight exposure did not predict post therapy serum 25(OH)D in the total sub-sample,¹⁵⁸ that there was no significant difference in sunlight exposure between groups at baseline¹⁶⁶ or during the study.¹⁵⁶ Participants were community-dwelling in all of the included trials.¹⁵⁵–^{161, 163}–¹⁶⁶

Interventions and comparators. The vitamin D dietary interventions included fortified milk,¹⁵⁵–^{159, 163} nutrient dense fruit and dairy based products,¹⁶⁶ high vitamin D diet,¹⁶⁵ fortified orange juice,¹⁶⁰ fortified cheese,¹⁶¹ and fortified bread.¹⁶⁴ The RCT with a factorial design had two other intervention groups that included an exercise program and a combined program of exercise and nutrient dense products.¹⁶⁶

The type of vitamin D administered within the described vitamin D dietary interventions was vitamin D₃ in eight trials,^{155, 157}–^{161, 163, 164} and was not specified in three.^{156, 165, 166} The vitamin D content was 200 – 1,000 IU. Seven trials also specified the calcium content within the dietary intervention.¹⁵⁵–^{160, 163}

The comparators within the included trials were as follows: usual diet or no intervention,^{155, 157, 163, 165, 166} unfortified liquid milk,^{156, 158} fortified milk with a lower dose of calcium but same dose of vitamin D compared to intervention group,¹⁵⁹ unfortified orange juice,¹⁶⁰ unfortified cheese or no cheese,¹⁶¹ and regular wheat bread or regular wheat bread and a vitamin D₃ supplement.¹⁶⁴

The duration of the intervention ranged from three weeks¹⁶⁴ to 24 months.^{155, 157, 163}

Compliance was reported in four trials and was reported to be greater than 85 percent.^{155, 156, 161, 163}

Study quality. Six out of the 11 trials had a methodological quality score of ≥ 3/5 on the Jadad scale (Table 9).^{156, 157, 159}–^{161, 163} Ten trials reported the percent lost to followup,¹⁵⁵–^{159, 161, 163}–¹⁶⁶ and of these, only one reported losses greater than 20 percent.¹⁶⁶ In all trials, the description of allocation concealment was unclear.¹⁵⁵–^{161, 163}–¹⁶⁶

Intention-to-treat analysis. One trial carried out an intention-to-treat analysis,¹⁶⁵ eight trials did not,¹⁵⁵–^{160, 163, 164, 166} and the type of analysis was unclear in one trial.¹⁶¹

Study Selection for Meta-Analysis

Meta-analysis was conducted to quantify the effects of dietary sources with vitamin D with/without calcium versus placebo or calcium on serum 25(OH)D levels. Seven of the 11 included trials that reported (or provided sufficient data to calculate) the absolute change in total 25(OH)D or 25(OH)D₃ concentrations were included in the meta-analysis.^{155, 156, 158, 160, 164}–¹⁶⁶ The other four RCTs were excluded due to insufficient data required to calculate the change in 25(OH)D levels,^{157, 163} between group differences in baseline 25(OH)D levels,¹⁶¹ or the intervention and control groups receiving equal amounts of vitamin D.¹⁵⁹

Quantitative Data Synthesis

Combining all seven trials that investigated the effect of food fortification or dietary sources of vitamin D (with/without calcium) versus control was not possible due to heterogeneity of the treatment effect (I² = 79.2 percent). However, the individual weighted mean differences (WMD) demonstrated a clear trend toward a significantly higher absolute change in serum 25(OH)D in the treatment group versus control (Figure 3).^{155, 156, 158, 160, 164}–¹⁶⁶ Potential sources of heterogeneity are the different 25(OH)D assays used (two studies each used HPLC, RIA or CPBA, and one study did not report the assay), the dietary vehicles used, study populations, the type or dose of vitamin D (unclear in one trial¹⁶⁵), and the outcome employed (i.e., total 25(OH)D versus 25(OH)D₃).

Figure

Figure 3. Forest Plot on the Effect of Dietary Sources of Vitamin D (with/without calcium) vs. Control on Absolute Change in Total Serum 25(OH)D or 25(OH)D₃.

Combined data from two trials (N = 275) that were similar in the dietary vehicle used (fortified skim milk), population studied (postmenopausal women and young adults), dose of vitamin D (400 and 480 IU daily), type of vitamin D (D₃), 25(OH)D assay (RIA), and outcome (total 25(OH)D) demonstrated a significantly higher absolute change in serum 25(OH)D (WMD 15.71, 95% CI 12.89, 18.53, heterogeneity I² = 0 percent) in the treatment group^{155, 158} (Figure 4). Similarly, a significantly higher percent change in serum 25(OH)D was demonstrated in the treatment group (WMD 19.13, 95% CI 15.32, 22.95). However, heterogeneity of the treatment effect was high (I² = 54.1 percent).^{155, 158} The study by McKenna et al. demonstrated a decrease in 25(OH)D levels in both groups as a result of seasonal decline. However, food fortification reduced the degree of seasonal decline in the treatment group.¹⁵⁸

Figure

Figure 4. Forest Plot on the Effect of Vitamin D₃ Fortified Skim Milk (with calcium) vs. Control on Absolute Change in Total Serum 25(OH)D.

In an attempt to explain the heterogeneity found in the overall analysis, the following subgroups were analyzed: (1) younger versus older individuals; (2) all trials that administered 400 IU/day (the most common dose); (3) the use of total 25(OH)D versus 25(OH)D₃ and (4) the type of vitamin D assay (RIA, HPLC versus CPBA). The subgroup analysis that included studies of younger individuals demonstrated a significant absolute increase in 25(OH)D levels (4 trials, N = 323, WMD 17.02, 95% CI 12.49, 21.56, heterogeneity I² = 44.4 percent).^{155, 158, 160, 164} However, combining trials within all of the other subgroup analyses was not possible as the heterogeneity of the treatment effect was high. A meta-regression to further explore heterogeneity was not carried out due to the limited number of trials with sufficient data.

Publication Bias. We were not able to evaluate the possibility of publication bias given the limited number of trials with sufficient data required to conduct such an investigation.

Qualitative Data Synthesis

Results from the four trials^{157, 159, 161, 163} that were excluded from the quantitative analysis are described below.

Daly et al. (2006) explored the effect of fortified milk (800 IU vitamin D₃ plus 1000 mg of calcium) versus no additional milk in older Caucasian, ambulatory men (mean age 62 years) over a two year period. Serum 25(OH)D was increased in the milk supplementation group relative to controls (27 percent, p<0.001). Baseline characteristics did not differ between groups.¹⁶³

Johnson et al. (2005) investigated the effects of vitamin D fortified cheese (600 IU D₃ daily) on serum 25(OH)D versus unfortified cheese or no cheese for two months in older men and women.¹⁶¹ Serum 25(OH)D measured at the beginning of the study demonstrated a significant difference between the fortified cheese versus control groups. Overall compliance with consumption of 85 grams of cheese per day was high (96.2 percent) with no difference between groups. Results demonstrated that, despite a significantly higher total vitamin D dietary intake in the fortified cheese versus the two control groups (unfortified cheese and no cheese groups), the end of study serum 25(OH)D decreased by a mean of 6 (SD 2) nmol/L (p<0.001) in the fortified cheese group. While not a clinically significant decrease, the authors speculated that this decrease reflected the higher baseline serum 25(OH)D in the fortified cheese group.¹⁶¹

Lau et al. (2001) investigated the benefits of milk supplementation (240 IU D₃ plus 800 mg Ca) in postmenopausal Chinese women over a two year period.¹⁵⁷ At 12 months, serum 25(OH)D was higher in the milk supplementation group compared to baseline (p<0.05). Baseline and followup serum 25(OH)D for the control group, a comparison of serum 25(OH)D between the intervention and control group, and participants' sunlight exposure and vitamin D intake were not reported.¹⁵⁷

Palacios et al. (2005) assessed the effect of consuming milk enriched with calcium and vitamin D (1,200 mg Ca plus 228 IU D₃) versus milk with lower calcium content but the same amount of vitamin D (900 mg Ca plus 228 IU D₃) daily for six months in healthy postmenopausal women. Serum 25(OH)D₃ increased from baseline in those women who consumed the milk enriched with calcium (which also contained phosphorus and lactose) even thought the amount of vitamin D was similar (p <0.001). The calcium enriched milk group had significantly higher serum 25(OH)D₃ at the end of study than the non-enriched group (p = 0.007). These results led the authors to speculate that calcium may affect the absorption of vitamin D. However, compliance was not measured. The participants' sunlight exposure and vitamin D intake were also not reported.¹⁵⁹

Dose response of serum 25(OH)D to dietary interventions. The positive direction of the treatment effect of dietary interventions with foods fortified with vitamin D is consistent. Based on our synthesis of the data from the individual trials, the treatment effect may be dependent on baseline serum 25(OH)D levels (Table 10). Those trials with low baseline 25(OH)D levels (i.e., < 50 nmol/L)^{156, 160, 164}–¹⁶⁶ consistently demonstrated a greater percent increase in 25(OH)D levels at the end of study compared to trials with higher baseline 25(OH)D levels (i.e., > 50 nmol/L).^{155, 157}–^{159, 161} Observations from such indirect comparisons need to be interpreted cautiously due to differences in baseline characteristics of the study populations, the bioavailability of the vitamin D in the various food sources and the different measures of serum 25(OH)D used.

Table 10

Absolute and % Change in Serum 25(OH)D for the Intervention Group inSupplementation Trials (grouped by vitamin D dosages < 400 IU vs. ≥ 400 IU/d).

Summary

Despite the possibility of study contamination by altered intake of other nutrients contained within the different food sources that affect bone and mineral homeostasis, food sources enriched with vitamin D in the form of milk, orange juice or other dairy and fruit based products (i.e., yogurt, custard and fruit juice) significantly improved vitamin D status in vitamin D deficient, insufficient or sufficient populations including young adults, postmenopausal women and elderly men. This was demonstrated by a significant rise in serum 25(OH)D in individuals that received vitamin D enriched dietary interventions compared to controls on an individual trial basis,¹⁵⁵–^{160, 163}–¹⁶⁶ and by combining trials that permitted a quantitative analysis.^{155, 158}

Increases in serum 25(OH)D from vitamin D enriched dietary interventions may depend on baseline 25(OH)D levels as well as vitamin D dose. However, this observation is based on indirect comparisons of the individual trials and should be interpreted with caution. It was not possible to determine if results vary with age, BMI and ethnicity given the limited data available and the between trial differences in terms of population characteristics, dietary interventions and measurement of serum 25(OH)D levels.

Overview of Relevant RCTs

Study characteristics. Eight randomized trials evaluated the effect of ultraviolet exposure on serum 25(OH) D concentrations.¹⁶⁷–¹⁷⁴

Within these eight parallel design trials, there were a total of 337 subjects with 197 subjects in the intervention group and 140 subjects in the comparator groups. Four trials evaluated the effect of natural sun exposure,^{168, 169, 171, 172} and four trials evaluated the effect of artificial UV exposure^{167, 170, 173, 174} on circulating 25(OH)D concentrations.

Population characteristics. There were seven trials in adult populations and one in infants.¹⁷² Three trials involved younger or middle-aged adults^{169, 170, 174} and four trials included older adults.^{167, 168, 171, 173} The percentage of females ranged from 17¹⁷⁰ to 100 percent,¹⁶⁷ and one trial had only male participants.¹⁷⁴ In the trial in infants, 55 percent were female.¹⁷²

Body Mass Index was not reported in any of the trials. Skin type was reported in two trials: Matsuoka¹⁷⁰ in which all individuals were skin type III (i.e., sometimes burn, always tans) and Falkenbach included skin types II (i.e., always burns, sometimes tans) and III.¹⁷⁴ Another trial reported that skin pigmentation varied from fair to medium.¹⁶⁸

Vitamin D intake. One trial reported daily dietary vitamin D of 3.1 nmol or 48 IU¹⁶⁸ and another estimated dietary intake of 100 IU of vitamin D plus 1,000 mg of calcium per day.¹⁶⁷ Dietary intake was not reported in the remaining six trials.¹⁷⁰–¹⁷⁵

Vitamin D deficiency. In four of the eight trials, the proportion of subjects with vitamin D deficiency at baseline (< 30 nmol/L) was reported.¹⁶⁷–^{169, 172} In two trials of elderly nursing home residents, 93 percent of subjects were vitamin D deficient (<30 nmol/L) in one trial,¹⁶⁷ and 50 percent in the other trial.¹⁶⁸ In contrast, in a trial on community-dwelling adults in Australia, only 10 percent were vitamin D deficient.¹⁶⁹ In the infant trial,¹⁷² 20 percent of infants were deficient and 11 percent were diagnosed with rickets. Baseline concentrations and type of vitamin D assay are presented in Table 11.

Table 11

Effect of UV Exposure on Serum 25(OH)D Levels.

Interventions. In the four trials that used solar exposure,^{168, 169, 171, 172} the dose was one minimal erythemal dose (MED) in one trial,¹⁶⁸ and a geometric mean of 138 J/m² in another trial.¹⁶⁹ In two trials, the exact dose was not reported but described as 2 hours of sunshine per day with face and hands exposed¹⁷² or 15 versus 30 minutes with head, neck and arms exposed.¹⁷¹ All trials were conducted in southern latitudes, except for the infant trial.¹⁷² In the four trials that used artificial UV,^{167, 170, 173, 174} the description of the dose was as follows: (1) one suberythematous dose of 27 mJ/cm² to the whole body,¹⁷⁰ (2) 1/2 MED at doses from 30 to 140 mJ/cm²;¹⁶⁷ (3) high energy versus low energy UV-B to provide suberythematous doses,¹⁷⁴ and (4) a dose of 160 mJ/cm² per week.¹⁷³

The frequency of UV exposure was a single exposure in one trial,¹⁷⁰ one¹⁷³ to three times per week,¹⁶⁷ ten times over a 12 day period,¹⁷⁴ and daily in four trials.^{168, 169, 171, 172} The duration of the intervention varied from a single exposure,¹⁷⁰ to 12 days in one trial,¹⁷⁴ 28 days in two trials,^{171, 172} and 12 weeks in three trials.^{167, 168, 173} Marks et al. used sunscreen as the intervention.¹⁶⁹

Ascertainment of UV exposure. Three of the four trials that used natural sun exposure reported the method of ascertainment of UV-B exposure. Ho et al. used a sunshine diary to record minutes outdoors per day and used the average weekly UV score for September to October.¹⁷² Lovell used UV sensitive polysulphone badges and readings on a UV meter coupled to a sensor.¹⁶⁸ Marks also used polysulphone film badges in addition to a sun exposure and clothing diary.¹⁶⁹

Comparators. In four trials, the comparator was a placebo.^{169, 171}–¹⁷³ Two trials included a comparator arm of vitamin D₃ 400 IU¹⁶⁷ or two dosages of vitamin D₃; 289 IU or 867 IU.¹⁶⁸ The two remaining trials used lower energy UV-B,¹⁷⁴ or UV-B with 50,000 IU vitamin D₂ versus vitamin D₂ alone as comparators.¹⁷⁰

Compliance. Compliance was reported in only two trials.^{167, 174} In the Chel trial¹⁶⁷ three patients in the UV-B group did not complete the treatment and in the other trial¹⁷⁴ one subject did not comply with treatment.

Study quality. Study quality scores on the Jadad scale ranged from 1 to 4 out of a possible 5, with all except two trials having a score of less than 3.^{169, 171} A description of trial withdrawals was adequately reported in six of the trials.¹⁶⁷–^{169, 172}–¹⁷⁴ In all eight trials, the description of allocation concealment was unclear. One challenge with trials of UV exposure is the difficulty of blinding study participants to the intervention.

Type of analysis. Three trials performed an intention-to-treat analysis.^{170, 171, 174} In five trials an intention-to-treat analysis was either not performed or the type of analysis was unclear.¹⁶⁷–^{170, 173}

Qualitative data synthesis. Quantitative synthesis of the trials of UV exposure and serum 25(OH)D was not possible due to the heterogeneous study populations, the interventions (e.g., length and area of exposure, and dose) and lack of complete data.

Outcomes. Followup serum 25(OH)D or 25(OH)D₃ concentrations were evaluated in six trials^{167, 168, 171}–¹⁷⁴ (Table 11). The change in serum 25(OH)D concentrations from baseline was significant in all of the six trials.

Reid (1986) compared the effect of sun exposure in 15 Caucasian older men and women living in residential homes in New Zealand. The subjects were randomized into three groups of five each; controls who did not change their daily routine and the two intervention groups (outside daily for either 15 or 30 minutes for four weeks). Body surfaces exposed included head, neck, legs and forearms. Mean baseline serum 25(OH)D concentrations were different across groups: 35 nmol/L (15 minute group); 60 nmol/L (30 minute group), and; 60 nmol/L (control group). Serum 25(OH)D increased in both the 15 and 30 minute groups, however the increase (18.5 nmol/L) was only significant in the 30 minute group.¹⁷¹

Lovell (1988) studied the effect of sun exposure in Caucasian elderly nursing home residents in Australia compared to vitamin D₃ (either 289 IU or 867 IU/day) over a three month period. The median increase (11.0 nmol/L) in serum 25(OH)D concentrations was significant after the second month of treatment in the UV-B group and the lower dose vitamin D group and after the first month, with 867 IU vitamin D₃.¹⁶⁸

In Asian breast-fed infants aged one to eight months who were not receiving supplemental vitamin D, Ho (1985) assessed the effect of two hours of sunshine per day for two months (face and hands uncovered) versus the usual amount of sunshine. Infants in the intervention group received 115 minutes of sunshine per day compared to controls who received an average of 63 minutes. There was a significant increase in serum 25(OH)D in the treatment group, but not in the infants receiving usual sunshine exposure. Serum 25(OH)D concentrations correlated with UV exposure scores, even after adjusting for age. The estimated UV score needed to maintain serum 25(OH)D at 27.5 nmol/L was 24 minutes per day with only the face uncovered.¹⁷²

Marks et al. (1995) conducted a seven-month RCT in Australia of daily sunscreen use (SPF of 17) compared to placebo in 113 subjects over age 40 years. Participants were recruited from a random sample of a trial designed to evaluate the effect of regular sunscreen use in subjects with solar keratoses. Sunscreen was applied daily to the head, neck, forearms and dorsum of each hand. The mean baseline serum 25(OH)D₃ was 54.2 nmol/L. When the results were stratified by age, serum 25(OH)D₃ increased less in subjects over 70 years in the sunscreen group (7.4 nmol/L ) versus those younger than 70 years (15.9 nmol/L) but the differences were not significant. Overall serum 25(OH)D₃ concentrations increased by the same amount in the sunscreen and non-sunscreen groups with a difference of 0.99 nmol/L (95% CI -7.0, 5.0). Nine out of 11 subjects with serum 25(OH)D₃ below the reference range had values within the reference range by the end of the study. The absence of a difference between groups may have been due to incomplete compliance with sunscreen use.¹⁶⁹

In a 12 week trial, Toss (1982) studied the effect of artificial UV exposure on 42 elderly nursing home residents compared to vitamin D₂ 450 IU plus calcium 600 mg daily, calcium alone, or placebo. Front and back were exposed to UVR for 1 minute each, then 2 minutes and followed by ten treatments of 3 minutes each. The mean UV total dose was 160 mJ/cm². There were significant increases in serum 25(OH)D in both the UV group (end of study 25(OH)D was 59 nmol/L) and in the vitamin D₂ group (42 nmol/L), compared to no change in serum 25(OH)D in the control and calcium groups.¹⁷³

Chel (1998) investigated the effect of artificial UV-B irradiation in 45 elderly females in The Netherlands. The majority of subjects were vitamin D deficient (<30 nmol/L). Subjects were randomized to receive UV-B (one-half MED) three times per week, 400 IU vitamin D₃ or placebo for 12 weeks. Six areas of 4 cm² were irradiated with UV-B doses increasing from 30 to 140 mJ/cm², and individual doses were adjusted according to skin sensitivity as determined by the MED. After 12 weeks, the median serum 25(OH)D concentrations increased to 60 nmol/L in both the UV-B (increase of 42 nmol/L) and vitamin D₃ (increase of 37 nmol/L) groups (p<0.001).¹⁶⁷

Falkenbach (1992) evaluated the effect of artificial high energy (less emission in range of 300 nm) versus low energy, shorter wavelength UV-B in healthy young men (N=24) in Germany, during the winter. Both treatment groups were treated ten times over a 12-day period in a solarium. The initial exposure was three minutes and increased by 10 percent with each session to achieve suberythemal doses, using both ventral and dorsal irradiation. Baseline serum 25(OH)D₃ concentrations were higher (115–124 nmol/L) than in other trials which may reflect younger age of subjects. Fasting serum 25(OH)D₃ concentrations measured three days after the last exposure increased significantly in both groups and remained elevated for four weeks, in the low energy, shorter wavelength UV-B group (Table 11). Serum PTH concentrations were significantly decreased in this group.¹⁷⁴

Matsuoka (1992) evaluated if administration of vitamin D₂ interfered with the release of vitamin D₃ from the skin after exposure to UV-B light. A total of eighteen subjects were randomized to receive oral 50,000 IU vitamin D₂ alone, 50,000 IU vitamin D₂ followed by UV-B exposure 12 hours later or UV-B alone. UV-B was given as a single dose to the whole body at a suberythematous dose of 27 mJ/cm². Total serum 25 (OH)D concentrations (measured by CPBA) did not increase significantly in any group. Vitamin D₃ concentrations (measured by HPLC) increased significantly after UV-B treatment (increase of 27.5 nmol/L). A similar increase in vitamin D₃ was observed when UV-B exposure was preceded by vitamin D₂, suggesting that elevated serum vitamin D₂ does not interfere with release of vitamin D₃ from the skin.¹⁷⁰

Overview of Relevant RCTs

Study characteristics. A total of 74 RCTs in 81 published reports evaluated the effect of vitamin D supplementation on circulating 25(OH)D concentrations.^{60, 61, 90}–^{93, 102, 105, 112}–^{115, 117}–^{121, 167, 168, 176}–^{185, 185}–²³⁶ Within the trials, five had the following companion publications: Greer⁹³ had one companion¹⁹³; Grados¹⁹¹ had two companion papers^{190, 237}; Dawson-Hughes¹⁸⁴ had one companion¹⁸⁵; Schaafsma¹²¹ has one companion²²¹; and Sorva²²⁴ had two companion papers.^{225, 226} For each trial in this section we refer to the primary publication (Table 12).

Table 12

RCTs on Vitamin D Supplementation and Serum 25(OH)D Levels.

Sixty-nine studies were parallel design randomized trials.^{60, 61, 90}–^{93, 102, 105, 112}–^{115, 117}–^{121, 167, 168, 176}–^{184, 186}–^{190, 192, 194}–^{197, 199}–^{207, 209}–^{215, 217}–^{220, 222, 224, 227, 229}–²³⁶ Four were crossover trials,^{198, 216, 223, 228} and one a factorial trial.²⁰⁸

Baseline BMI was reported in nineteen trials and ranged from 24.8¹⁹⁹ to 32.8 kg/m².¹⁹⁶

Study quality. Five trials^{112, 115, 203, 210, 238} received a rating of 5/5 on the Jadad scale, 13 trials received a rating of 4/5^{92, 113, 119}–^{121, 178, 184, 190, 192, 206, 219, 223, 228} and 17 trials were rated 3/5.^{102, 114, 117, 177, 180, 183, 193, 197}–^{200, 215, 216, 218, 222, 229, 231} Thirty-nine trials received a Jadad score of ≤2/5.^{60, 61, 90, 91, 93, 118, 167, 168, 176, 179, 181, 182, 186}–^{189, 194}–^{196, 201, 202, 204, 205, 207, 209, 211}–^{214, 217, 220, 224, 227, 230, 232}–²³⁶ These ratings indicate that more than half of the studies were of lower quality (Table 12).

Interventions. Vitamin D₃ alone was the intervention in 29 trials.^{60, 61, 105, 113, 119, 167, 168, 186}–^{189, 194, 195, 198, 200, 203, 206, 208}–^{210, 216, 223, 230}–²³⁶

Twenty-six trials used vitamin D₃ combined with calcium as the intervention.^{113, 114, 117, 118, 121, 177, 178, 180, 181, 183, 184, 187, 190, 192, 197, 199, 200, 202, 207, 213, 215, 218, 219, 222, 224, 228}

Fifteen trials used vitamin D₂ alone as the intervention.⁹⁰–^{93, 102, 112, 115, 120, 176, 179, 196, 211, 212, 214, 227} and the type of vitamin D was not stated in four trials.^{168, 204, 217, 220}

Three trials had separate vitamin D₂ and vitamin D₃ arms.^{61, 229, 230}

Qualitative data synthesis. Baseline serum 25(OH) D concentrations were reported in 61 trials.^{60, 102, 105, 112}–^{115, 117, 119}–^{121, 167, 168, 177}–^{181, 184, 187}–^{190, 192, 194}–^{210, 212, 214}–^{220, 222}–^{224, 227}–^{230, 232}–²³⁶

Twenty-one trials examined the efficacy of vitamin D supplements in vitamin D deficient populations (mean serum 25(OH)D ≤ 30 nmol/L),^{112, 114, 119, 167, 179, 180, 189, 190, 197, 199, 207, 209, 210, 214, 218, 220, 222, 224, 227, 235, 236}and three other trials had a subgroup of patients who were vitamin D deficient (≤ 30 nmol/L).^{90, 91, 202}

Vitamin D assay. The majority of trials (N = 42) used a competitive binding protein assay to measure serum 25 (OH)D concentrations.^{60, 91, 93, 102, 105, 112, 113, 118, 119, 121, 168, 176, 178}–^{184, 190, 194}–^{196, 198}–^{200, 202, 204}–^{207, 209}–^{211, 214, 215, 220, 224, 227, 232, 235, 236}

Twenty-nine trials used an immunoassay method.^{61, 90, 114, 115, 117, 120, 167, 177, 186}–^{189, 192, 197, 201, 203, 208, 212, 213, 216}–^{218, 222, 223, 228, 230, 231, 233, 234} and three trials used HPLC.^{92, 219, 229} No trials reported using liquid chromatography-tandem mass spectrometry to measure serum 25(OH)D concentrations.

The qualitative results are presented by age group and additional details are presented in Table 12. For the vitamin D₃ (+/- calcium) versus placebo or calcium trials that provided adequate data, the results of quantitative synthesis are presented after the qualitative section. We did not conduct quantitative analyses of vitamin D₂ versus placebo due to the smaller number of trials, heterogeneity of trials and lack of adequate data.

Meta-analysis of Trials of Oral Vitamin D₃(+/- Calcium) on Serum 25(OH)D Concentrations

Study selection. As summarized above, 44 RCTs investigated the effect of oral vitamin D₃ supplementation (+/- calcium) versus no treatment, placebo or calcium on serum 25(OH)D concentrations.^{60, 61, 105, 113, 114, 117, 119, 121, 167, 168, 177, 178, 180, 181, 183, 184, 186, 187, 189, 190, 194, 195, 197, 199, 200, 202, 203, 206}–^{210, 213, 215, 216, 218, 219, 223, 224, 228, 230}–^{232, 235}

Seventeen trials administered oral vitamin D₃ supplements with or without calcium versus no treatment, placebo or calcium on an intermittent or daily basis and presented sufficient data to combine results of the absolute change in serum 25(OH)D concentrations.^{60, 105, 113, 177, 181, 184, 189, 194, 195, 199, 200, 202, 207, 216, 218, 219, 224} Due to a significant and unexplained difference in baseline serum 25(OH)D levels between the treatment and control groups, we excluded the study by Riis et al.²¹⁹ A total of 16 trials were therefore included in the meta-analysis. Two trials^{60, 105} included more than one treatment arm with different doses of vitamin D₃ and one placebo group, so we used results from only one treatment group (i.e., 1,000 IU/day⁶⁰ and 2,000 IU/day¹⁰⁵) in all analyses. The study by Heaney et al.⁶⁰ warrants discussion as multiple measurements of serum 25(OH)D were taken over time. A compartment model was used to derive a monotonic form for concentration as a function of time. This model was fitted to each individual's data to extrapolate an estimate of the equilibrium (asymptotic) 25(OH)D concentration. The estimates from the Heaney study differ from the other included studies that did not require extrapolation.

The effect of vitamin D ₃ supplementation (+/- calcium) versus placebo or calcium on 25(OH)D concentrations. Combining the 16 trials with a random effects model demonstrated large heterogeneity of treatment effect, (I² = 97.7 percent). However, the point estimates for each trial consistently favored vitamin D₃.^{60, 105, 113, 177, 181, 184, 189, 194, 195, 199, 200, 202, 207, 216, 218, 224} (Figure 5a).

Figure

Figure 5a. The Effect of Vitamin D₃ Supplementation (+/- calcium) vs. Placebo or Calcium on Absolute Change in 25(OH)D Concentrations.

We conducted subgroup and sensitivity analyses and a meta-regression on dose to explore potential sources of heterogeneity.

Subgroup analyses were conducted in an attempt to explain heterogeneity and included: (1) dosage of vitamin D₃ (i.e., grouped by ≤ 400 versus. > 400 IU/day), (2) study population (i.e., older institutionalized, older community-dwelling versus younger community-dwelling individuals), (3) frequency of administration (i.e., intermittent versus daily vitamin D₃), (4) assays used (i.e., CPBA versus RIA and HPLC), and (5) study quality (high quality studies defined by a Jadad score ≥ 3). Other potential explanations for the heterogeneity are the potency of the vitamin D supplement and whether 25(OH)D₃ or total 25(OH)D was measured. Only one trial⁶⁰ assessed 25(OH)D₃ and the potency of the vitamin D supplement was measured in only two trials.^{60, 183}

Subgroup Analyses

(1) Dose. To examine the effect of dose, the daily dose was derived for the two studies that used an intermittent dose of vitamin D₃.^{105, 194} The trials were classified by dose (i.e., (< 400 IU/day),^{189, 199} versus (≥ 400 IU/day)).^{60, 105, 113, 177, 181, 184, 194, 195, 200, 202, 207, 216, 218, 224}

Combined results of two trials using < 400 IU/day demonstrated a significant increase in serum 25(OH)D levels [N = 136, WMD 11.36 (95% CI 8.56, 14.15), heterogeneity I² = 0 percent].^{189, 199} Combined results of trials that used doses ≥ 400 IU was not possible due to large heterogeneity of the treatment effect (WMD varied from 17.6 to 52.6) (I² = 96.0 percent). The weighted mean differences ranged from 17.6 to 69.5 (Figure 5b).

Figure

Figure 5b. The Effects of Vitamin D₃ Supplementation (with/without calcium) vs. Placebo or Calcium on Absolute Change in 25(OH)D Levels by Dose.

(2) Study Population. To explore the effect of age and health status of the study participants, the trials were classified as follows: (1) community-dwelling younger adults,^{60, 105, 177, 194, 195, 216} (2) community-dwelling older adults,^{113, 184, 189, 195, 199, 202, 218} and (3) elderly institutionalized individuals.^{181, 200, 202, 207, 224} Two studies reported results for two different populations.^{195, 202} Combining the trials by the defined subgroups was not possible due to heterogeneity of the treatment effect and did not explain the overall heterogeneity (community-dwelling younger adults: heterogeneity I² = 85.8 percent; community-dwelling older adults: heterogeneity I² = 97.0 percent; elderly institutionalized individuals: I² = 89 percent).

Baseline vitamin D status of the study populations were categorized as either vitamin D deficient at baseline (i.e. serum 25(OH)D levels < 30 nmol/L)^{189, 199, 202, 207, 218, 224} or serum 25(OH)D > 30 nmol/L.^{60, 105, 113, 177, 181, 184, 194, 195, 200, 202, 216} Results demonstrated that combining of trials was not possible due to heterogeneity of the treatment effect (vitamin D deficient: heterogeneity I² = 98.1 percent versus not vitamin D deficient: heterogeneity I² = 96.3 percent) (Figure 5c).

Figure

Figure 5c. The Effects of Vitamin D₃ Supplementation (with/without calcium) vs. Placebo or Calcium on Absolute Change in 25(OH)D Levels by Vitamin D Status.

When we combined data from two trials^{207, 224} that had similar population characteristics (age, institutionalized participants, vitamin D deficiency) and dose (880 –1000 IU), the increase in serum 25(OH)D compared to control was 51.2 nmol/L (95% CI 45.5, 57), I² = 0.

(3) Vitamin D assay. To explore the impact of different assays, the included trials were divided into three groups as defined a priori: RIA,^{177, 189, 216, 218} CPBA ^{60, 105, 113, 181, 184, 194, 195, 199, 200, 202, 207, 224} or HPLC. None of the included studies used HPLC. Combining was not possible due to heterogeneity of the treatment effect (RIA: heterogeneity I² = 93 percent versus CPBA: heterogeneity I² = 97.5 percent).

Other subgroup analyses conducted but not presented here included (1) baseline 25(OH)D levels by classifying those with 25(OH)D levels as deficient and (2) compliance. These analyses did not reduce the heterogeneity and therefore did not permit pooling of the results.

Sensitivity analyses. The sensitivity analyses included: (1) study quality and, (2) loss to followup. Allocation concealment was not explored, since only one study reported adequate allocation concealment.

The included studies were divided into high (quality score ≥ 3 on the Jadad scale)^{105, 113, 177, 184, 199, 200, 216, 218} versus low quality subgroups.^{60, 181, 189, 194, 195, 202, 207, 224} However, combining was not possible due to heterogeneity of the treatment effects (high quality: heterogeneity I² = 93.7 percent versus low quality: heterogeneity I² = 98.2 percent).

The effect of loss to followup was explored by grouping the trials into those that reported a loss of over 20 percent^{181, 207} versus less than 20 percent.^{105, 113, 177, 184, 189, 194, 195, 199, 202, 218, 224} Combining trials was not possible due to heterogeneity of the treatment effects (loss to followup over 20 percent: heterogeneity I² = 95.3 percent versus less than 20 percent: heterogeneity I² = 97.2 percent).

Meta-regression on dose. A meta-regression of the 16 trials (a weighted linear mixed effects model estimated by REML), N = 1376, was conducted to estimate the extent to which dose of vitamin D₃ explained the heterogeneity of the treatment effects. Results demonstrated a significant association between the daily dose of oral vitamin D₃ on serum 25(OH)D concentrations and the regression coefficient [beta=0.016 (95% CI 0.007,0.032), p = 0.042] suggesting that if the dose of vitamin D₃ increases by 1 IU, the serum 25(OH)D concentrations can be expected to increase by 0.016 nmol/L. The estimated between-study variance (tau-squared) was reduced from 393.6 to 222.9. See Figure 5d for a graphical representation of the treatment effect versus daily dose.

Figure

Figure 5d. 25(OH)D Treatment Effect vs. Daily Oral Vitamin D₃ Dose.

The effect of oral vitamin D ₃ with/without calcium supplementation on serum concentrations of serum PTH. The effect of vitamin D supplementation on serum PTH was assessed in 14 of the 16 trials.^{60, 113, 177, 181, 184, 189, 194, 195, 199, 200, 207, 216, 218, 224}

Vitamin D supplementation significantly decreased PTH concentrations in nine trials (four of which were in vitamin D deficient populations)^{60, 113, 181, 184, 189, 207, 216, 218, 224} or was sufficient to maintain serum iPTH levels, in spite of seasonal effects, in one trial.¹⁹⁴ Nine trials used a vitamin D₃ dose of ≥ 700 IU.^{60, 113, 181, 184, 194, 207, 216, 218, 224} Explanations for the failure to observe a decrease in serum PTH include that the vitamin D dose may have been too low for a population with low baseline 25(OH)D concentrations,¹⁹⁹ or that serum 25(OH)D may have been above the threshold where further changes in PTH would occur. In addition, PTH is modulated by other factors such as calcium intake.¹⁹

View in own window

Summary. Quantitative Analysis

Seventeen trials of vitamin D3 provided sufficient data to conduct a quantitative analysis. The treatment effect of oral vitamin D3 supplementation increases with increasing doses. Combining trials by different clinical and methodological characteristics did not change the direction of this effect nor did it reduce the heterogeneity found. Meta-regression results demonstrated a significant association between dose and serum 25(OH)D levels (p = 0.04). The meta-regression (exploratory only) results suggested that 100 IU of vitamin D3 will increase the serum 25(OH)D concentrations by 1–2 nmol/L. This suggests that doses of 400–800 IU daily may be inadequate to prevent vitamin D deficiency in at-risk individuals. Vitamin D3 doses of 700 IU daily or more significantly and consistently decreased serum concentrations of PTH in vitamin D deficient populations.

Given the limitations in the measurement of 25(OH)D concentrations and the lack of standardization and calibration, it is difficult to suggest precise recommendations for adequate intakes, especially since optimal levels of serum 25(OH)D have not been defined.

Overview of Relevant RCTs

Study characteristics. A total of 17 randomized trials evaluated the effect of supplemental vitamin D (with or without calcium) versus control (calcium, placebo or no treatment) on bone mineral density. Of these 17 trials, 16 were parallel design RCTs of either supplemental vitamin D₂ or D₃ ¹¹⁷–^{120, 180, 181, 183, 184, 197, 203, 204, 213, 237, 241}–²⁴³ and one was a crossover trial of vitamin D₃.²¹⁶ Treatment duration varied from one¹⁸³ to seven years,²⁴³ and most trials were less than three years in duration. Three articles^{190, 191, 237} were companion papers and we refer to the primary publication²³⁷ when discussing the results provided in either paper.

Study population. The majority of trials included postmenopausal women. Only one trial included premenopausal women,²¹⁶ and one trial included women who were recently postmenopausal.²⁴² Only two trials included older men > 60 years.^{184, 213} Thirteen trials included community-dwelling individuals.^{117, 118, 120, 183, 184, 203, 204, 213, 216, 237, 241}–²⁴³ Two trials had populations of ambulatory elderly subjects living in either nursing homes or seniors' apartments,^{180, 181} and one trial included women living in homes or apartments for the elderly.¹¹⁹ Harwood included women living in the community who had sustained a hip fracture and were admitted to hospital.¹⁹⁷ One trial enrolled postmenopausal African-American women.¹¹⁷

Interventions. The majority of the trials used oral vitamin D₃, and two trials administered vitamin D₂ (Table 13).^{120, 197} Harwood also included an oral vitamin D₃ arm.¹⁹⁷ The daily dose of vitamin D₃ ranged from 300 IU²⁴² to 2,000 IU.¹¹⁷ Aloia et al. administered 800 IU vitamin D₃ for two years followed by 2,000 IU daily for one year. Five trials used a dose of 800 IU vitamin D₃,^{180, 181, 197, 203, 216} four trials used a daily dose less than 800 IU but greater than or equal to 400 IU.^{118, 119, 183, 184, 204, 213, 241, 243} One trial used 300 IU vitamin D₃.²⁴² Doses of vitamin D₂ ranged from 10,000 IU orally per week¹²⁰ to an annual injection of 300,000 IU.¹⁹⁷

Table 13

Effect of Vitamin D₂ or D₃ on BMD by Site in Individual Trials.

Fourteen trials had treatment arms that combined vitamin D with calcium,^{117, 118, 180, 181, 183, 183, 184, 197, 204, 213, 237, 241}–²⁴³ and three trials administered vitamin D alone.^{119, 203, 216}

Daily calcium dosages ranged from 377 mg in one trial,¹⁸³ 500 mg in three trials^{118, 184, 213} 1,000 mg in four trials,^{120, 237, 241, 243} to 1,200 mg or more in three trials.^{180, 181, 204}

Dietary vitamin D intake: nine trials estimated the mean baseline daily dietary vitamin D intake^{117, 118, 180, 183, 184, 203, 237, 241, 243} which ranged from 40 IU¹⁸⁰ to 202 IU.¹⁸⁴ (Table 13)

Comparators. Comparators included calcium in five trials,^{117, 120, 183, 204}low dose vitamin D₃ (100 IU) plus calcium in one trial,¹¹⁸ and placebo in 11 trials.^{119, 180, 181, 184, 197, 203, 213, 216, 237, 241}–²⁴³

Compliance. Compliance with vitamin D was reported in eleven trials and the compliance rates (compliance defined as > 80% of supplementation taken) were over 80 percent in seven of the eleven trials.¹¹⁷–^{119, 180, 184, 203, 237} One study reported an adherence score as ‘excellent’ but did not provide a percentage score,²⁰⁴ and another reported a compliance rate (compliance defined as > 70% of supplementation taken) in 83–84%.¹⁸¹ Another study gave supplements in the presence of a nurse to ensure compliance but did not specifically report a rate.¹⁸⁰ The WHI trial reported a rate of adherence (> 80% of assigned medication taken) of 60 – 63 percent in the first three years of followup and 59% at end of study.²⁴³

Study quality. The overall quality score on the Jadad scale ranged from 1 (lowest) to 5 (highest). Four trials received a score of ≤ 2.^{118, 181, 204, 213} Thirteen trials received a score of ≥ 3 consistent with high quality.^{117, 119, 120, 180, 183, 184, 197, 203, 216, 237, 241}–²⁴³ Two trials adequately reported the allocation concealment.^{117, 203} Fourteen trials reported losses to followup with seven reporting losses over 20 percent.^{119, 180, 181, 184, 197, 204, 237}

Type of analysis. Six trials reported an intention-to-treat analysis.^{117, 180, 181, 184, 242, 243}

25 (OH) D levels. Thirteen trials reported baseline serum 25(OH) D levels.^{117, 119, 120, 180, 181, 184, 197, 203, 204, 213, 216, 237, 242} Fifteen trials reported followup or change in 25(OH)D levels.¹¹⁸–^{120, 180, 181, 183, 184, 197, 203, 204, 213, 216, 237, 242} Of the fifteen trials reporting 25(OH)D, six used an RIA assay,^{117, 120, 197, 203, 213, 216} one used a chemiluminescent immunoassay²⁴³ and eight studies used a CPBA (at least two^{184, 204} of which were the Nichols Advantage Assay).

Vitamin D-deficient populations. Mean baseline 25(OH)D concentrations were ≤ 30 nmol/L in three trials.^{180, 197, 237} Ooms reported median 25(OH)D of 27.0 and 25 nmol/L in treatment and placebo groups, respectively,¹¹⁹ and the mean 25(OH)D concentrations were just over 30 nmol/L in another trial.²¹³

BMD by region of interest. Fourteen trials assessed effect of vitamin D on lumbar spine BMD,^{117, 118, 120, 183, 184, 197, 203, 204, 213, 216, 237, 241}–²⁴³ twelve assessed femoral neck BMD,¹¹⁸–^{120, 180, 181, 184, 197, 213, 237, 241}–²⁴³ five trials evaluated total hip BMD,^{117, 197, 203, 204, 243} eight assessed total body BMD,^{117, 118, 183, 184, 203, 204, 237, 243} and five assessed a forearm site.^{117, 119, 120, 180, 241}

Ascertainment of BMD. BMD was assessed by DXA using Hologic machines in nine trials,^{117, 180, 181, 197, 203, 204, 213, 216, 243} Lunar technology in four trials,^{118, 183, 184, 242} Norland in three trials,^{119, 120, 241} and either Lunar, Hologic or Norland in one trial.²³⁷ One trial used one of three densitometers, Lunar, Hologic or Norland and standardized the results.²³⁷

Individual trial results for lumbar spine, femoral neck and total body BMD are summarized in Table 13. Three trials evaluated BMD in a subpopulation of the total trial population.^{180, 181, 243}

Data Synthesis

Six trials did not provide data in a format that would permit pooling.^{197, 203, 213, 216, 237, 243} One was a crossover trial,²¹⁶ and one trial evaluated the effect of vitamin D₃ on postmenopausal twins, in which one member of each twin pair was randomized to vitamin D₃ and the other to placebo and intra-pair differences analyzed.²⁰³ In four trials, adequate data were not provided within the published paper.^{197, 213, 237, 243}

In the twin pair (mean age 58.7 years) trial by Hunter et al., there was no significant difference in BMD at the lumbar spine with or without supplementation over a two year period and during that time, there was a mean one percent loss at the total hip.²⁰³

Patel (2001), in a two year crossover trial, evaluated whether vitamin D₃ prevented seasonal changes in BMD in healthy women (mean age 47.2 years).²¹⁶ Vitamin D₃ had no overall effect on lumbar spine, femoral neck or total body BMD. Treatment effect coefficients of lumbar spine BMD were not significantly different from zero in either the low (baseline serum 25(OH)D < 60 nmol/L) or high vitamin D (baseline serum 25(OH)D > 80 nmol/L) groups. The authors concluded that the women in this study were too replete to demonstrate seasonal changes in BMD and that vitamin D supplements did not have significant effect on BMD.

In a two year trial, Meier (2004) evaluated the effect of six months of 500 IU of daily vitamin D₃ plus 500 mg of calcium in healthy adults (male mean age 60.6 years and female mean age 54.1 years) during the winter to determine if supplements prevented seasonal bone loss. In the vitamin D₃ and calcium treated subjects, the lumbar spine and femoral neck BMD increased in the second year compared to the first year, versus controls who continued to lose BMD.²¹³

In the Women's Health Initiative trial (N = 36,282), a subgroup of 2,431 women from three of 40 centers had BMD measured (lumbar spine, total hip and total body). Women were randomized to either vitamin D₃ 400 IU plus 1,000 mg of calcium daily or placebo. Non-significant differences in lumbar spine and total body BMD were reported, with results in favour of the vitamin D₃ and calcium treated group. The BMD at the total hip was 1.06 percent higher compared to the control group after an average of seven years of treatment (p<0.001).²⁴³

Harwood et al. compared BMD changes of the lumbar spine and hip with injectable vitamin D₂ 300,000 units (± calcium), vitamin D₃ 800 IU/day (± calcium) or no treatment in women who had sustained a hip fracture. Differences in BMD for vitamin D treated versus control group ranged from 1.1 to 3.3 percent at femoral neck, 2.5 to 4.6 percent at the trochanter, and 2.1 to 4.6 percent at the total hip, with greater effects seen with oral vitamin D₃ plus calcium.¹⁹⁷

Grados (2003) compared vitamin D₃ 800 IU with calcium 1,000 mg per day in 192 elderly women in France. All women had 25(OH)D concentrations below 30 nmol/L with mean concentrations of 18.25 nmol/L which increased to 56 nmol/L after treatment. After one year, there was a median increase of 2.98% at the lumbar spine in the treatment group versus -0.21 in placebo and a 1.19% increase at the femoral neck versus -0.83% in placebo group. There was a significant increase in BMD at the total body and the trochanter compared to placebo.^{190, 237}

In a two year trial, Cooper evaluated the effect of oral 10,000 IU vitamin D₂ weekly plus calcium 1,000 mg versus calcium alone, and did not find a significant difference in annual change of the lumbar spine, femoral neck or forearm BMD between the two groups.¹²⁰

For meta-analyses, given that calcium alone increases bone density, BMD results from similar sites and treatment durations were combined in the following groups: (1) vitamin D₃ alone, (2) vitamin D₃ plus calcium versus placebo, and (3) vitamin D₃ plus calcium versus calcium. Due to variable reporting, and differences in treatment arms, quantitative pooling was limited.

The combined results by BMD site are presented in Table 14. Eleven trials provided data that allowed quantitative analysis.¹¹⁷–^{120, 180, 181, 183, 184, 204, 241, 242}

Table 14

Combined Results of Effect of Vitamin D3 on BMD.

Oral vitamin D ₃ plus calcium versus placebo. Comparing vitamin D₃ plus calcium to placebo, there were significant increases in BMD at the lumbar spine after one year with a combined estimate from two trials (N = 507) of 1.40 percent (95% CI 0.84, 1.97).^{184, 237, 241} Significant increases at the femoral neck^{180, 184, 237, 241} were observed with a combined estimate of 1.37 percent (95% CI 0.24, 2.50) from three trials after one year. The heterogeneity of treatment effect varied from low to moderate depending on the site (Table 14).

Oral vitamin D ₃ versus placebo. The combined estimates of trials that evaluated BMD of the lumbar spine²⁴² or forearm¹¹⁹ were not significant with vitamin D₃ alone, although in both trials the dose of vitamin D₃ was 300 or 400 IU daily. In the trial by Ooms, there was a significant increase in femoral neck BMD with 400 IU vitamin D₃ versus placebo over two years.¹¹⁹

Oral vitamin D ₃ plus calcium versus calcium. The combined results of trials, including the trial on African American women, that compared vitamin D₃ plus calcium vs. calcium did not demonstrate a significant effect on BMD of the lumbar spine, total hip, forearm or total body.^{117, 204}

Effect of baseline 25(OH)D concentrations and BMD response to vitamin D. Four trials assessed the effect of baseline serum 25(OH)D and BMD response to either vitamin D₃ or D₂.¹¹⁷–¹²⁰ One trial had a population that was vitamin D deficient (median serum 25(OH)D 25–27 nmol/L by CPBA) and reported that the effect of vitamin D₃ on femoral neck BMD was independent of baseline 25(OH)D concentrations.¹¹⁹ The other studies, one of which included African American women, did not report an association between baseline serum 25(OH)D concentrations and changes in BMD.

Overview of Relevant RCTs

Study characteristics. Fifteen randomized trials evaluated the effect of either vitamin D₂ or D₃ (combined with or without calcium) on incident fractures. Thirteen trials were parallel design RCTs,^{180, 181, 184, 197, 210, 218, 231, 242}–²⁴⁷ and two were factorial trials.^{248, 249} Duration ranged from one to seven years. Table 15 provides trial characteristics.

Table 15

OR (95% CI) for Total Fractures from Individual RCTs of Vitamin D.

Thirteen trials randomized individual participants and the overall number of participants in the intervention arms was 32,092, with 32,491 participants in the control or placebo groups. Two trials randomized participants using a cluster design (cluster randomization refers to randomization by group, e.g., a residential unit). The combined sample size of the two cluster randomized trials was 6,719 in the intervention groups and 4,071 in the control groups.^{247, 249} Porthouse et al. changed treatment allocation from unequal to equal during the trial so there are two entries for this study with different denominators: an equally randomized group (1:1 ratio) (study A) and an unequally randomized group (2:1 ratio in favor of the control) (study B).²⁴⁴

Population characteristics. Two trials were classified as secondary prevention trials as all participants had a history of fractures.^{197, 248} Four other trials reported a baseline fracture prevalence that ranged from 10.7 to 26 percent.²⁴²–^{244, 249}

Seven trials included only postmenopausal females,^{180, 181, 197, 218, 242}–²⁴⁴ and eight trials included both older males and postmenopausal females.^{184, 210, 231, 245}–²⁴⁹ Of these eight trials, the percentage of females ranged from 25²³¹ to 95 percent.²⁴⁶ There were no trials in women of reproductive age.

Nine trials included community-dwelling participants.^{184, 218, 231, 242}–^{245, 248, 249} One trial included community-dwelling participants living independently in apartments.²¹⁰ Four trials included cohorts of participants living in residential homes.^{180, 181, 246, 247} One trial was conducted with hospitalized participants who had been community-dwelling prior to admission.¹⁹⁷

Interventions. Eleven RCTs allocated participants to oral vitamin D₃ with dosages ranging from 300 to 800 IU/day. Harwood allocated participants to either oral vitamin D₃ arm or injectable vitamin D₂ arms.¹⁹⁷ Six trials used an oral dose of 800 IU vitamin D₃ per day^{180, 181, 197, 218, 244, 248} one trial administered 700 IU D₃,¹⁸⁴ and four trials a dosage of ≤ 400 IU vitamin D₃ daily.^{210, 242, 243, 249}

Two trials used daily oral vitamin D₂ with dosages equivalent to 1,000 or 1,100 IU, respectively.^{246, 247}

Two trials used an injectable preparation of either vitamin D₂ or D₃. Harwood used a single dose of 300,000 IM vitamin D₂ ¹⁹⁷ and another trial used an annual dose of 300,000 IU vitamin D₃.²⁴⁵

Calcium supplementation as a co-intervention ranged from 500 mg in two trials^{184, 242} to 1,000 mg in five trials^{197, 243, 244, 248, 249} to 1200 mg/day in three trials.^{180, 181, 218}

Porthouse et al. had high baseline levels of dietary calcium intake in both the intervention (1,075 mg) and control groups (1,084 mg), and provided all participants with information on dietary calcium and vitamin D.²⁴⁴ Jackson also had a high mean baseline intake of calcium in both intervention and control groups (1,150 mg).²⁴³

Comparators. Seven trials compared oral or injectable vitamin D to placebo or control.^{197, 210, 231, 243, 245, 247, 248} Seven trials compared a combination of vitamin D plus calcium to placebo.^{180, 184, 197, 243, 244, 248, 249} Four trials compared vitamin D plus calcium versus calcium alone.^{218, 242, 246, 248}

Compliance. Compliance with vitamin D was reported in eleven trials and was greater than 80 percent in five trials.^{180, 181, 210, 218, 242} Compliance was less than 80 percent in six trials.^{184, 231, 243, 243, 244, 248} In the three largest trials, the compliance ranged from 55 to 63 percent.^{243, 244, 248}

Study quality. One trial had a quality score of 2/5 on the Jadad scale.¹⁸¹ Ten trials had a score of ≥ 3/5,^{180, 184, 197, 210, 231, 242, 244}–^{246, 248} and of these, two trials had the maximum score of five.^{210, 248}

Eight trials had losses to followup greater than 20 percent.^{180, 181, 184, 197, 210, 231, 246, 248}

Two trials provided an adequate description of allocation concealment,^{210, 248} and allocation concealment was unclear in the remaining trials.

Type of analysis. Twelve trials reported an intention-to-treat analysis,^{180, 181, 184, 210, 231, 242}–^{244, 246}–²⁴⁹ and in three trials, an efficacy analysis was conducted or the type of analysis was unclear.^{197, 218, 245}

Fracture outcomes. Three RCTs provided data on vertebral fractures,^{231, 243, 248} twelve trials on non-vertebral fractures,^{180, 181, 184, 197, 210, 218, 231, 242}–^{244, 247, 248} and fourteen trials provided data on either total or hip fractures.^{180, 181, 184, 197, 210, 218, 231, 242}–^{244, 246}–²⁴⁹

Ascertainment of fractures. Ascertainment of fractures varied with some trials using self-report (± x-ray confirmation) or administrative data^{197, 210, 231, 244, 246, 249} and other trials verifying fractures by x-rays.^{180, 181, 184, 218, 242, 243, 248} One trial used several sources including self-report, physician verification, and administrative databases.²⁴⁸ Vertebral fractures were ascertained only by questionnaire in one trial²³¹ and confirmed by x-rays in two trials.^{243, 248}

25(OH)D concentrations. Eleven trials reported baseline 25(OH)D concentrations.^{180, 181, 184, 197, 210, 218, 242, 243, 247}–²⁴⁹ In six trials, 25(OH) concentrations were measured in a sub-sample of the total trial population.^{181, 242, 243, 247}–²⁴⁹

Vitamin D deficiency. Mean baseline serum 25 (OH)D concentrations below 30 nmol/L were reported in five trials.^{180, 197, 210, 218, 242}

Eleven trials reported followup or change in mean 25(OH) D concentrations.^{180, 181, 184, 197, 210, 218, 231, 242, 247}–²⁴⁹ Serum 25(OH)D concentrations were not reported in three trials.²⁴⁴–²⁴⁶ (See Table 16.)

Table 16

OR (95% CI) from Individual RCTs Included in the Meta-Analysis on the Effects of Vitamin D on Fall Risk.

Quantitative Data Synthesis

We conducted a meta-analysis of the 13 randomized trials that provided adequate data on fracture outcomes. Two entries (Study A and B) from Porthouse et al. are presented since the allocation changed from unequal to equal during the trial.²⁴⁴

Included in the meta-analysis is the Women's Health Initiative (WHI, 2006) trial on calcium plus vitamin D₃ (400 IU). The WHI trial was the largest primary prevention trial and involved 36,282 postmenopausal women (mean age of 62.4 years). Women enrolled in the WHI HRT and dietary modification trials were invited to participate in the calcium and vitamin D trial. A unique feature of this trial was that over 50 percent of women were current users of hormonal replacement therapy (HRT) and the rate of use of other osteoporosis medications was one percent. In this trial, the overall risk reduction in hip fractures with vitamin D plus calcium was not significant compared to placebo (12 percent, 95% CI -8 to 28). In subgroup analyses of women over age 60 years, and in women who were compliant, there was a significant reduction in hip fractures compared to placebo [≥ 60 years (21 percent, 95% CI 2–36); compliant women (29 percent, 95% CI 3–48)].²⁴³

Total fractures. Combined results from 13 trials (N=58,712) that used either oral vitamin D ₃ or D ₂ +/- calcium versus calcium or placebo resulted in a non-significant reduction in total fractures [(OR 0.90, (95% CI 0.81, 1.02), p=0.09)] with a I² of 48 consistent with moderate heterogeneity of treatment effect (Figure 7).

Figure

Figure 6. Forest Plot: Effect of vitamin D₃ + Calcium vs. Placebo on Femoral Neck BMD at 1 year.

Figure

Figure 7. Forest Plot Comparing Risk of Total Fractures with Vitamin D₂ or D₃ +/- Calcium vs. Placebo or Calcium.

Combined results from three trials (N=7,939) of vitamin D ₃ alone versus placebo were not consistent with a significant reduction in total fractures [(OR 0.98, 95% CI, 0.79–1.23), p=0.08, I²=61 consistent with high heterogeneity].^{210, 231, 248}

Combined results of three trials of vitamin D ₃ plus calcium versus calcium (N=2,997)^{218, 242, 248} resulted in a non-significant reduction in total fractures [(OR 0.92, 95% CI 0.74–1.25), I²=10.2 percent].

Combined results of seven trials of vitamin D ₃ plus calcium versus placebo (n=46,072)^{180, 181, 184, 197, 243, 244, 248} were consistent with a non-significant reduction in total fractures [OR 0.87, 95% CI 0.76–1.00, p=0.05, I²=43 percent] (Figure 8).

Figure

Figure 8. Forest plot Comparing the Risk of Total Fractures with Vitamin D₃ Combined With Calcium vs. Placebo.

Non-vertebral fractures. Combined results from three trials (n=7,939)^{210, 231, 248} of vitamin D ₃ alone versus placebo were not consistent with a significant reduction in non-vertebral fractures [(OR, 0.99, 95% CI, 0.83–1.17), p = 0.89, I² = 27.6 percent].

Combined results from seven trials (N = 46,074),^{180, 181, 184, 197, 243, 244, 248} of vitamin D ₃ plus calcium versus placebo were consistent with an OR of 0.87 (95% CI 0.75–1.00, p = 0.05), and a I² of 44 percent.

Hip fractures. Combined results of three trials (N=7,939)^{210, 231, 248} of vitamin D ₃ versus placebo were not consistent with a significant reduction in hip fractures [OR 1.11, 95% CI 0.86– 1.44, I² = 0].

The combined results of three trials of vitamin D ₃ plus calcium versus calcium (N=2,997)^{218, 242, 248} were not consistent with a significant reduction in hip fractures [OR 0.91, 95% CI 0.61– 1.36, I² = 0].

Combined results from seven trials (n=46,072)^{180, 181, 184, 197, 243, 244, 248} of vitamin D ₃ plus calcium versus placebo were consistent with a non-significant effect, although the point estimate favoured vitamin D [OR 0.83, 95% CI 0.68–1.00, p=0.05, I²=16.2 percent] (Figure 8).

Vertebral fractures. The combined OR from three trials (n=44,260) with oral vitamin D₂ or D₃ (+/- calcium) versus placebo or calcium for vertebral fractures was 0.88 (95% CI 0.73– 1.07), I²=0.^{231, 243, 248}

Results of Trials not Included in the Quantitative Synthesis

Larsen²⁴⁹ was a factorial cluster-randomized trial that did not appear to control for the effect of clustering in their per protocol analysis, so the results were not combined with the other trials.

Larsen administered 400 IU vitamin D₃ with 1,000 mg calcium daily versus placebo and reported a significant reduction in total fractures [RR 0.84 (95% CI 0.72, 0.98), p<0.025]. When results were presented by gender, females had a decreased fracture risk [RR 0.81 (95% CI 0.68–0.95), p<0.01].²⁴⁹

Andersen et al. administered an annual injection of 300,000 IU of vitamin D₃ versus placebo and did not report a significant reduction in hip fractures [HR 1.48 (95% CI 1.01–2.17)] or for any fracture [HR 1.10 (95% CI 0.94–1.29), p = 0.23)]. The results were similar in both males and females. Complete data were not provided.²⁴⁵

Subgroup and Sensitivity Analyses

To explore the heterogeneity of treatment effect we conducted subgroup analyses by: residential status (community-dwelling versus institutional), dosage, and 25(OH)D concentrations for the outcome of total fractures. Combining the three trials of vitamin D₂/D₃ plus calcium versus placebo or calcium in institutionalized populations^{180, 181, 246} resulted in a significant reduction in total fractures [OR 0.73 (95% CI 0.61–0.88), I² = 0] versus a non-significant reduction when combining nine trials of community-dwelling participants [OR 0.95, (95% CI 0.86, 1.05) I² = 23.4].^{184, 197, 210, 218, 231, 242}–^{244, 248}

When exploring heterogeneity of the seven trials of vitamin D₃ and calcium versus placebo by residence, the combined OR for two trials^{180, 181} in elderly populations in institutions was significant [OR 0.69 (95% CI 0.53, 0.90), I² = 0] (Figure 9).

Figure

Figure 9. Forest Plot Comparing Risk of Hip Fractures with Vitamin D₃ +/- Calcium vs. Placebo by Setting.

Subgroup analysis by dosage, (i.e., combining trials ≥ 800 IU of vitamin D versus those trials using < 800 IU/day) did not explain the heterogeneity of treatment effect.

In sensitivity analyses, we explored the heterogeneity of treatment effect by combining: (1) trials with high versus low study quality, (2) trials with over 80 percent compliance versus those with less than 80 percent compliance, and (3) trials that adequately reported allocation concealment compared to trials in which allocation concealment was not reported or was unclear. None of these analyses had a significant impact on the heterogeneity of treatment effect.

Effect of 25(OH)D concentrations on fracture risk. Eleven trials evaluated baseline serum 25(OH)D concentrations and five trials had low baseline serum 25(OH)D concentrations (<30 nmol/L).^{180, 197, 210, 218, 242} One trial that reported a significant reduction in fracture risk,¹⁸¹ had a mean baseline 25(OH)D concentration of 40 nmol/L.

Followup serum 25(OH)D concentrations (≥ 74 nmol/L) were reported in three trials that reported a significant reduction in total fractures.^{181, 184, 231}

Combining the results from four trials of vitamin D₃ ^{180, 181, 184, 231} that had end of study 25(OH)D concentrations of ≥74 nmol/L was consistent with a significant reduction in total fractures [OR 0.73 (95 % CI 0.63–0.85), I² = 0] compared to a non-significant reduction when combining results of trials with end of study 25(OH)D concentrations of < 74 nmol/L.

Publication bias. An evaluation of publication bias, using the method by Begg et al.²⁵⁰ suggested the possibility of bias, with a lack of smaller trials that failed to find an effect of vitamin D on fracture reduction.

Overview of Relevant RCTs

Study characteristics. A total of 14 trials in 16 published reports evaluated the effect of vitamin D on falls and of these, 12 were RCTs with a parallel design,^{114, 115, 180, 184, 185, 197, 218, 231, 244, 246, 247, 252} and four used a factorial design.^{208, 248, 249, 253}

Three trials used cluster randomization^{247, 249, 253} and the remaining trials randomized by individual patient.^{114, 115, 180, 184, 185, 197, 208, 218, 231, 244, 246, 248, 252} Porthouse et al. randomized patients in an equally randomized group in a 1:1 ratio (referred to as “study A”) as well as, an unequally randomized group in a 2:1 ratio in favor of the control group (referred to as “study B”).²⁴⁴

Bischoff-Ferrari et al. (2006)¹⁸⁵ was identified as the companion paper to the primary publication Dawson-Hughes et al. (1997)¹⁸⁴ and Larsen et al.(2005)²⁵³ was identified as companion paper to Larsen et al. (2004).²⁴⁹ We refer to the primary publications of each trial when discussing the results. Table 16 summarizes characteristics of the included trials.

Within the 12 RCTs, a total of 5,445 participants received the intervention and 5,212 received the control or placebo.^{114, 115, 180, 184, 197, 208, 218, 231, 244, 246, 248, 252} In the two cluster randomized trials, 6,719 participants received the intervention and 6,603 received control.^{247, 249}

Population characteristics. A total of six trials included postmenopausal women only (i.e., greater than or equal to 95 percent of the participants were female)^{114, 180, 197, 218, 244, 246} whereas the remaining eight trials included a combination of postmenopausal women and elderly men.^{115, 184, 208, 231, 247}–^{249, 252}

Seven trials included community-dwelling residents^{115, 184, 218, 231, 244, 248, 249} and seven included participants who lived in residences with varied levels of assisted care.^{114, 180, 197, 208, 246, 247, 252}

Interventions. Eleven trials used oral vitamin D₃,^{114, 180, 184, 197, 208, 218, 231, 244, 248, 249, 252} two trials used oral vitamin D₂,^{246, 247} and two used a single intramuscular injection of vitamin D₂.^{115, 197}

Six trials had an intervention arm of oral vitamin D plus calcium,^{180, 184, 197, 244, 246, 248} and Harwood et al. had an injectable D₂ treatment arm with and without calcium.¹⁹⁷

Comparators. Seven trials compared vitamin D with placebo or control,^{115, 197, 208, 231, 247, 248, 252} and one trial compared vitamin D with calcium.²⁴⁸ Of the trials that used a combination of vitamin D plus calcium, the comparator was placebo in five trials^{180, 184, 197, 244, 248} and calcium in four trials.^{114, 218, 246, 248}

Compliance. Ten of the 14 trials reported the compliance rate with taking vitamin D.^{114, 115, 180, 184, 208, 218, 231, 244, 246, 248} The method of assessment varied from direct observation by a study nurse,^{114, 115, 180, 208} self-report questionnaires,^{231, 244, 248} to pill counts.^{184, 218, 246} In six of the ten trials, compliance rates were over 80 percent,^{114, 115, 180, 184, 208, 218} and less than 80 percent in the four other trials.^{231, 244, 246, 248} In the three largest trials, the compliance rates were 55,²⁴⁴ 63,²⁴⁸and ₇₆ ²³¹ percent, respectively.

Study quality. Eleven of 12 RCTs had a quality score of three or more on the Jadad scale.^{114, 115, 180, 184, 197, 208, 218, 231, 244, 246, 248}The two factorial-designed trials received 1/5 and 2/5 on the Jadad scales, respectively.^{247, 249} Seven trials reported losses to followup of over 20 percent^{114, 180, 184, 197, 231, 246, 248} Two trials provided an adequate description of allocation concealment,^{208, 248} and in all other trials, the description of allocation concealment was unclear.^{114, 115, 180, 184, 197, 218, 231, 244, 246, 252}

Type of analysis. Ten trials reported an intention-to-treat analysis,^{114, 115, 180, 184, 231, 244, 246}–²⁴⁹ whereas four trials used an available case analysis in which the data were analyzed for every participant in whom the outcome of falls was obtained.^{197, 208, 218, 252}

Fall outcomes. Thirteen RCTs reported the number of individuals with falls,^{114, 115, 180, 184, 197, 208, 218, 231, 246}–^{249, 252} and the data was provided by the authors for one trial.²⁴⁴

Definition of falls. Seven trials included a definition for falls, all of which were a variation on “unintentionally coming to rest at a lower level or on the ground.”^{114, 115, 184, 218, 246, 249, 252}

Ascertainment of falls. Different methods were used to ascertain the number of individuals with falls, and these included the use of postcards with followup visits,¹⁸⁴ questionnaires,^{218, 231, 244, 248} fall diaries with/without followup visits,^{115, 208, 246, 252} followup visits only,^{180, 197} hospital contacts,²⁴⁹ and record keeping by geriatric care staff.^{114, 247}

25(OH)D levels. Ten out of the 14 trials reported baseline 25(OH) D levels,^{114, 115, 180, 184, 197, 208, 218, 247}–²⁴⁹ seven trials reported the end of study 25(OH)D values^{114, 115, 197, 231, 247}–²⁴⁹and two reported the change in 25(OH)D from baseline.^{208, 218} Three trials evaluated baseline and followup 25(OH) D levels in a sub-sample only.²⁴⁷–²⁴⁹ For vitamin D assay, baseline and end of study 25(OH)D levels (intervention group only) in the included trials refer to Table 16.

Quantitative Data Synthesis

Meta-analyses were conducted using data from the 12 RCTs to explore the effect of oral/injectable vitamin D with/without calcium on the risk of falls.^{114, 115, 180, 184, 197, 208, 218, 231, 244, 246, 248, 252} Data from the two cluster randomized trials^{247, 249}were not included in the quantitative analyses with trials that randomized individual patients. Refer to Tables 16 and 17 for a summary of the results.

Table 17

OR (95% CI) from Combined RCTs Included in the Meta-Analysis on the Effects of Vitamin D on Fall Risk.

Oral vitamin D alone. Combined data from four trials (N = 5,958) of oral vitamin D ₃ versus placebo did not demonstrate a statistically significant reduction in the risk of falls [OR 1.03 (95% CI 0.91–1.17), heterogeneity I² = 0 percent).^{208, 231, 248, 252}

Only one trial looked at the effect of oral vitamin D ₃ versus calcium (N = 2,654), and the results did not demonstrate a statistically significant reduction in falls [OR 1.19 (95% CI 0.96 – 1.47)].²⁴⁸

Combined data from four trials (N = 7269) of oral vitamin D ₃ versus placebo or calcium did not demonstrate a significant reduction in the risk of falls [OR 1.05 (95% CI 0.93–1.19), heterogeneity I² = 0 percent).^{208, 231, 248, 252}

Oral vitamin D with calcium. Combined data from five trials (N = 7,056) of oral vitamin D ₃ with calcium versus placebo showed a statistically significant reduction in the risk of falls [OR 0.85 (95% CI 0.76–0.96), heterogeneity I² = 0 percent].^{180, 184, 197, 244, 248}

Combined data from four trials (N = 3,512) of oral vitamin D ₂ /D ₃ with calcium versus calcium demonstrated a significant reduction in the fall risk [OR 0.81 (95% CI 0.68–0.97), heterogeneity I² = 0 percent].^{114, 218, 246, 248}

Combined data from eight trials (N = 9,262) of oral vitamin D ₂ /D ₃ with calcium versus placebo or calcium demonstrated a significant reduction in the risk of falls [OR 0.84 (95% CI 0.76–0.93), heterogeneity I² = 0 percent].^{114, 180, 184, 197, 218, 244, 246, 248} Refer to Figure 10 for forest plot.

Figure

Figure 10. Forest Plot Comparing the Risk of Falls Between Vitamin D₂/D₃ with Calcium vs. Controls (placebo or calcium).

Oral vitamin D with or without calcium. Combined data from 11 trials (N = 13,888) of oral vitamin D ₂ /D ₃ with and without calcium versus placebo or calcium did not demonstrate a significant reduction in the risk of falls [OR 0.92 (95% CI 0.85–1.00), heterogeneity I² = 0 percent).^{114, 180, 184, 197, 208, 218, 231, 244, 246, 248, 252}

Injectable vitamin D. Combined data from two trials (N = 214) of injectable vitamin D ₂ versus placebo did not show a statistically significant reduced fall risk [OR 0.31 (95% CI 0.04–2.12)]. However, heterogeneity of the treatment effect was high (I² = 78.4 percent).^{115, 197} Possible explanations include differences in the study populations (elderly women post-hip fracture versus ambulatory elderly men and women with unreported fall histories) and dose of the vitamin D₂ injection (300,000 IU versus 600,000 IU of vitamin D₂).

A small trial (N = 73) of injectable D ₂ with calcium versus placebo did not demonstrate a significant reduction in the risk of falls in the treatment group [OR 0.37 (95% CI 0.12–1.12)].¹⁹⁷

Combined data from two trials (N = 250) of injectable vitamin D ₂ with or without calcium versus placebo did not show a statistically significant reduction in falls [OR 0.42 (95% CI 0.13–1.33)]. However, heterogeneity of the treatment effect was high (I² = 67.6 percent).^{115, 197} See above for possible explanations.

There were no trials that compared the effects of injectable vitamin D with or without calcium to calcium alone.

Oral or injectable vitamin D with or without calcium. Combined data from nine trials (N = 11,895) of vitamin D ₂ /D ₃ (oral or injectable) with or without calcium versus placebo did not demonstrate a significant reduction in the risk of falls [OR 0.91 (95% CI 0.81–1.01), heterogeneity I² = 24.4 percent].^{115, 180, 184, 197, 208, 231, 244, 248, 252}

Combined data from four trials (N = 4,855) of vitamin D ₂ /D ₃ (oral or injectable) with and without calcium versus calcium also did not demonstrate a significant reduction in the risk of falls [OR 0.88 (95% CI 0.70–1.10), heterogeneity I² = 28.8 percent).^{114, 218, 246, 248}

Combined data from all 12 trials (N = 14,101) of vitamin D ₂ /D ₃ (oral or injectable) with and without calcium versus placebo or calcium demonstrated a borderline significant reduction in fall risk [OR 0.89 (95% CI 0.80–0.99), heterogeneity I² = 23.2 percent) (refer to Figure 11).^{114, 115, 180, 184, 197, 208, 218, 231, 244, 246, 248, 252}

Figure

Figure 11. Forest Plot Comparing the Risk of Falls Between Oral or Injectable Vitamin D₂/D₃ with/without Calcium vs. Controls (placebo or calcium).

Publication bias. A funnel plot (OR versus precision [1/standard error]) of the 12 RCTs that investigated the effect of oral or injectable vitamin D with/without calcium versus placebo or calcium on fall incidence indicates possible asymmetry that was confirmed statistically (intercept 0.27 (90% CI 0.19 to 0.35), p = 0.0001), suggesting the possibility of bias although other potential causes of asymmetry exist (Figure 12).

Figure

Figure 12. Treatment Effect vs. Precision from Individual RCTs of the Effect of Oral Vitamin D with/without Calcium on Fall Risk.

We conducted separate subgroup and sensitivity analyses to ascertain whether the ‘overall’ treatment effect observed in our earlier analyses was influenced by various clinical or methodological characteristics respectively.

Subgroup and Sensitivity Analyses

Subgroup analyses were conducted as follows: (1) dose of vitamin D (less than or ≥ 800 IU/day; (2) setting (community-dwelling versus institutional participants); (3) study duration (≤ versus > one year, and; (4) gender (postmenopausal women versus a mixed population). The sensitivity analyses included: (1) ascertainment of falls (adequate definition and method of ascertainment versus inadequate or not reported); (2) compliance (less than versus greater than 80 percent); (3) allocation concealment (adequate versus unclear) and; (4) loss to followup (less than versus greater than 20 percent).

Combining six trials (N = 4,942) that included postmenopausal women only demonstrated a significant reduction in falls [OR 0.80 (95% CI 0.66–0.98)]. However, the heterogeneity of treatment effect was moderate (I² = 44.8 percent) (Figure 13).^{114, 180, 197, 218, 244, 246} However, combining trials by dose, setting and study duration did not demonstrate a significant reduction in falls.

Figure

Figure 13. Forest Plot of Comparing the Risk of Falls between Oral or Injectable Vitamin D₂/D₃ with/without Calcium vs. Controls (placebo or calcium) Grouped by Study Population i.e. (more...)

For the sensitivity analyses, combining results from ten RCTs (N = 8,566) in which the allocation concealment was unclear demonstrated a significant reduction in falls [OR 0.85 (95% CI 0.76–0.96), heterogeneity I² = 23.2 percent] ((Figure 14).^{114, 115, 180, 184, 197, 218, 231, 244, 246, 252} Lastly, combining the six RCTs (N = 1,833) in which falls and ascertainment were adequately defined demonstrated a significant reduction in falls [OR 0.79 (95% CI 0.65–0.96), heterogeneity I² = 0 percent].^{114, 115, 184, 218, 246, 252}

Figure

Figure 14. Forest Plot of Comparing the Risk of Falls between Oral or Injectable Vitamin D₂/D₃ with/without Calcium vs. Controls (placebo or calcium) Grouped by Reports of Allocation (more...)

Results of Trials not Included in the Quantitative Synthesis

Both Larsen et al.²⁴⁹ and Law et al.²⁴⁷ were not included in the meta-analysis as they were cluster randomized trials. Larsen et al. compared 400 IU vitamin D₃ plus 1,000 mg calcium carbonate daily to placebo and a multivariate analysis, including age, marital status and intervention program, demonstrated a 12 percent reduction in fall risk in those females who followed the calcium plus vitamin D program (RR 88, 95% CI 0.79–0.98). However, the effect of clustering was not controlled for in their analysis.²⁴⁹ Law et al. compared 100,000 IU of vitamin D₂ every three months (equivalent to 1,100 IU daily) and did not find a significant reduction in fall risk in elderly people in care homes after adjusting for age, sex, length of time in trial and the cluster randomization of the trial (RR 1.09, 95% CI 0.95–1.25).²⁴⁷

Do Benefits of Vitamin D Supplementation on Falls Vary with Baseline Serum 25(OH)D Levels?

We were not able to quantitatively analyze if the effect of vitamin D supplementation on fall risk varies with baseline 25(OH)D levels as only four out of the 14 trials reported adequate data^{115, 180, 197, 218} Three of the trials evaluated the effect of oral vitamin D₃ (800 IU/day) and calcium,^{180, 197, 218} and two evaluated the effect of vitamin D₂ in a single injection (300,000 IU or 600,000 IU) with/without calcium on falls.^{115, 197} The 25(OH)D assays used were either RIA^{115, 197, 218} or CPBA.¹⁸⁰ Differences in the type of vitamin D administered (D₂ versus D₃), route of administration (oral versus injectable), vitamin D dosage and 25(OH)D assays used in these four trials limit a direct comparison. Refer to Table 16 for baseline 25(OH)D levels, the assays used and OR (95% CI) of the trials.

Overview of Relevant Studies

Potential consequences of vitamin D toxicity include hypercalcemia, renal stones and soft tissue and vascular calcification. Clinical symptoms associated with hypercalcemia include nausea, vomiting, increased thirst and depression. Serum concentrations of 25(OH)D above 220 nmol/L have been associated with hypercalcemia.²⁷³ Hypercalciuria can be associated with vitamin D toxicity and may contribute to the development of nephrolithiasis, although other factors such as low urinary citrate and hyperoxaluria also predispose to renal stones.²³⁴

Randomized trials that reported safety outcomes by intervention group were included in this section of the report.

Study characteristics. A total of 22 randomized controlled trials (RCTs) (in 23 published reports) reported if vitamin D supplementation resulted in toxicity.^{77, 105, 112}–^{114, 117, 118, 178, 180, 181, 184, 191, 197, 202, 207, 209, 212, 233, 234, 236, 243, 248} Twenty-one were parallel design RCTs,^{77, 105, 112}–^{114, 117, 118, 178, 180, 181, 184, 191, 197, 202, 207, 209, 212, 233, 234, 236, 243} and one RCT used a factorial design.²⁴⁸ Two publications reported the results of more than one study in each record.^{233, 236} The Vieth publication (2004) included two trials and we refer to each as Study A and Study B respectively.²³³ Zeghoud et al. included two studies, only one of which was an RCT.²³⁶ Study characteristics are summarized in Table 18.

Table 18

Reported Safety Outcomes by Intervention Group (RCTs).

Population characteristics. Within the 22 included RCTs, there were a total of 47,802 subjects. Only two trials^{243, 248} had large sample sizes, with the majority of remaining studies having sample sizes of less than 100 participants. There were a total of 25,562 participants within the intervention group and 22,240 participants within a comparator, control, or placebo group. Seven of the 22 trials included both males and females,^{77, 112, 184, 209, 233, 234, 248} thirteen included only females,^{105, 114, 117, 118, 178, 180, 181, 191, 197, 202, 207, 212, 243} one included only males,¹¹³ and one trial with infants did not specify the gender.²³⁶

Two trials included infants, healthy term neonates enrolled at birth in one study⁷⁷ and infants 3 to 36 months of age (mean age 10.6 months, SD 6.1)who were diagnosed with vitamin D deficient rickets in the other.²³⁶ One trial included healthy (pre- and post-menarchal) female children aged 10 to 17 years.¹⁰⁵ Two studies included predominantly middle-aged populations (mean age 41.6 and 38.8 years (range 18–56 years) in one study and mean age 53 and 55 years (range not reported) in the other study).^{233, 234} Seventeen studies included older adults.¹¹²–^{114, 117, 118, 178, 180, 181, 184, 191, 197, 202, 207, 209, 212, 243, 248} The precise definition of an older population varied in the studies (e.g., postmenopausal women; individuals 65 years or older including mean ages ranging from 7th to the 9th decade). The adult populations were described as participants from long-term geriatric care facilities, nursing homes or homes for the aged in five studies^{112, 114, 181, 207, 209} or community-dwelling participants in ten studies.^{113, 117, 178, 180, 184, 197, 202, 233, 234, 248}

Ascertainment of toxicity. Ascertainment of toxicity was reported in most trials. The most commonly reported laboratory measure of calcium homeostasis was serum calcium (either total or ionized).¹¹²–^{114, 117, 178, 181, 181, 184, 191, 197, 202, 202, 207, 209, 209, 212, 236, 248, 274} In most trials, hypercalcemia was defined as a total serum calcium level above 2.7–2.8 mmol/L. Thresholds used to define hypercalciuria varied across studies. For example, hypercalciuria was defined as a mean urinary calcium-creatinine ratio <1.0 when calcium and creatinine are measured in mmol (or ≤ 0.37 when measured in mg) in a randomly collected sample or as a 24-hour urinary calcium excretion value with variable thresholds of 6.25–10 mmol/day.^{180, 191, 234} Criteria used to ascertain the outcome of renal stones were not clearly reported in all trials.

Interventions. Nineteen trials used oral vitamin D₃,^{77, 105, 113, 114, 117, 118, 178, 180, 181, 184, 191, 202, 207, 209, 233, 234, 236, 243, 248} and three trials used vitamin D₂.^{112, 197, 212}

Seven trials had intervention arms of one or more doses of oral vitamin D.^{77, 105, 112, 209, 233, 234, 236} Fifteen had one or more arms of vitamin D with calcium.^{113, 114, 117, 118, 178, 180, 181, 184, 191, 197, 202, 207, 212, 243, 248}

Comparators. Twelve trials compared vitamin D with placebo^{105, 112, 117, 180, 181, 184, 191, 243, 248} or control.^{197, 202, 207} Five studies had a comparator arm of calcium.^{113, 114, 178, 212, 248} Six trials used another dose of vitamin D as the comparator.^{77, 118, 209, 233, 234, 236}

Study quality. Twelve studies received a rating of ≥ 3 on the Jadad scale.^{105, 112}–^{114, 117, 178, 180, 184, 191, 197, 243, 248} Eleven studies were described as double-blind,^{105, 112}–^{114, 117, 178, 180, 184, 191, 234, 248} and of those, nine adequately conducted the blinding.^{105, 112}–^{114, 117, 178, 180, 191, 248} In the majority of trials (N = 19), allocation concealment was unclear^{77, 105, 112, 114, 118, 178, 180, 181, 184, 191, 197, 202, 207, 209, 212, 233, 234, 236, 243} whereas three studies provided an adequate description.^{113, 117, 248}

Study withdrawals were adequately reported in 12 of the 22 studies.^{112, 113, 117, 118, 181, 184, 191, 197, 207, 236, 243, 248} Of these trials, eight reported losses to followup of over 20 percent.^{112, 180, 181, 184, 191, 207, 209, 233}

Qualitative Synthesis

Infants. Two trials reported toxicity outcomes in infant populations.^{77, 236} In one study, 56 infants with vitamin D deficient rickets (mean age 10.7 months) were randomized to receive a single oral dose of 150,000, 300,000 or 600,000 IU of vitamin D⁷⁷ The other study included 30 healthy neonates with low baseline serum 25(OH)D (< 25 nmol/L) who were randomized at birth to receive either a single oral dose of 200,000 IU vitamin D₃ or 100,000 IU at birth, three and six months of age.²³⁶ The latter study also reported on an earlier cohort of 30 non-randomized infants who were treated with 600,000 IU.

In the two trials, no serum calcium values were reported within the hypercalcemia range for the 100,000 and 150,000 IU doses. The Cesur trial reported eight cases of hypercalcemia (two in the 300,000 and six in the 600,000 treatment arms). Zeghoud et al. did not report any episodes of hypercalcemia during the RCT. However, an oral dose of 600,000 IU vitamin D₃ resulted in a significant increase in serum calcium concentrations 2 weeks later (p>0.005), with no change in serum calcium in infants receiving a lower vitamin D dose (200,000 IU). Mean serum calcium concentrations in the 100,000 and 200,000 IU dose were significantly lower than serum calcium after an oral dose of 600,000 IU of vitamin D₃. No withdrawals were reported in the trials of infant populations.^{77, 236}

Children. One trial examined the safety of vitamin D₃ in healthy female children who received either weekly 1,400 IU (200 IU per day) or 14,000 IU (2,000 IU/day) of vitamin D₃, or placebo.¹⁰⁵ The authors reported that two subjects in the placebo group had serum calcium levels above the upper limit of normal at one year versus no subjects in the intervention groups. Three subjects (1.5 percent) in the 2,000 IU/day group had serum 25(OH)D levels over 250 nmol/L (256.4, 400.8, and 485.5 nmol/L), but none had concomitant hypercalcemia. There were 11 withdrawals out of 168 participants (16 percent). However, withdrawal rates did not differ by treatment arm. One girl in the low dose vitamin D arm dropped out due to glomuerulonephritis which was thought to be secondary to a post-streptococcal infection.

Adults. Two small trials by Vieth examined the safety of vitamin D₃ in women of reproductive age or middle aged men.^{233, 234} The populations included either healthy men and women²³⁴ or endocrine outpatients.²³³ Neither trial had a placebo or control group.²³³ In one trial, subjects were randomized to either 600 IU or 4,000 IU of vitamin D₃ daily.²³³ The second trial by Vieth et al. compared 1,000 IU to 4,000 IU of vitamin D₃ daily.²³⁴ The authors did not report if subjects with a history of renal stones were excluded.

Seventeen efficacy trials examined the safety of vitamin D in older adults.¹¹²–^{114, 117, 118, 178, 180, 181, 184, 191, 197, 202, 207, 209, 212, 243, 248} Fourteen trials used vitamin D₃ as the intervention,^{113, 114, 117, 118, 178, 180, 181, 184, 191, 202, 207, 209, 243, 248} and three trials used vitamin D₂.^{112, 197, 212} Vitamin D doses ranged from 400 to 10,000 IU daily.²¹² Six trials included a treatment arm of either vitamin D₂ or D₃ alone,^{112, 113, 197, 202, 209, 248} and thirteen had a treatment arm with vitamin D combined with calcium.^{114, 117, 118, 178, 180, 181, 184, 191, 197, 207, 212, 243, 248}

Six trials used an immunoassay method to measure 25(OH)D,^{114, 117, 197, 209, 212, 243} ten used CPBA,^{112, 113, 118, 178, 180, 181, 184, 191, 202, 207} and one trial used HPLC.²⁴⁸

Exclusion criteria that were reported in the published trials are summarized in Table 17. Five trials excluded subjects with a history of hypercalcemia,^{114, 180, 191, 209, 243} seven trials excluded subjects with renal insufficiency,^{112, 114, 118, 180, 184, 191, 209} seven excluded subjects with primary hyperparathyroidism or other disorders of bone metabolism,^{113, 114, 117, 118, 178, 184, 191} and three trials excluded subjects who had a history of kidney stones.^{184, 209, 243} Most trials excluded subjects who had taken medications known to affect bone metabolism.

Hypercalcemia. Thirteen trials reported hypercalcemia as an outcome.¹¹²–^{114, 178, 180, 181, 191, 197, 207, 209, 233, 234, 248} In three trials, cases of hypercalcemia were reported in the vitamin D arm that were thought to be due to unmasking of underlying primary hyperparathyroidism.^{180, 181, 207} Six trials reported that there were no cases of hypercalcemia in either arm of the study.^{113, 114, 178, 197, 233, 234}

Twelve trials that compared vitamin D alone or vitamin D plus calcium to placebo or calcium reported on the outcome of hypercalcemia.¹¹²–^{114, 117, 178, 180, 181, 191, 197, 207, 209, 248} Supplemental calcium carbonate or citrate doses ranged from 500 mg^{118, 184, 212} to 1,200 – 1,500 mg per day.¹¹⁷ Combining the results from the twelve trials that had either calcium or placebo as a comparator resulted in a Peto odds ratio of 1.58 (95% CI 0.9, 2.77), p = 0.11 and I² = 0.5 percent. There were a total of 50/10,535 cases of hypercalcemia with 31/5410 (0.6 percent) in the vitamin D (+/- calcium) and 19/5125 (0.4 percent) in the placebo or calcium arm. Excluding cases that were due to underlying primary hyperparathyroidism, resulted in a Peto Odds Ratio of 1.4 (0.76, 2.5). Most cases of hypercalcemia were reported to be asymptomatic.

Hypercalciuria. Ten trials provided data on hypercalciuria within the adult populations.^{113, 117, 118, 178, 180, 184, 191, 209, 212, 234} Vitamin D doses ranged from 700 IU vitamin D₃/day¹¹⁸ to 10,000 IU vitamin D₂/day.²¹² Seven trials had calcium carbonate 500–1,000 mg as a co-intervention^{113, 117, 178, 180, 184, 191, 212} In six trials^{113, 117, 118, 180, 184, 212} (N = 1190) that had calcium or placebo as a comparator, there were total of eighteen cases of hypercalciuria reported, 13 in the vitamin D arms and 5 in placebo/control (Peto OR of 1.78 (95% CI 0.68, 4.7), p = 0.24 and I² = 0). In one trial, all four cases of hypercalciuria were reversed by lowering the calcium supplementation from 500 mg to 250 mg/day.¹¹⁸ In another trial in elderly women receiving 800 mg of vitamin D₃ plus 1,000 mg of calcium, 20 percent had higher 24-hour urine calcium to creatinine ratios in the intervention group.¹⁹¹