Vitamin D for the Prevention of Osteoporotic Fractures

Transcription

1 TITLE: Vitamin D for the Prevention of Osteoporotic Fractures AUTHOR: Jeffrey A. Tice, MD Assistant Professor of Medicine Division of General Internal Medicine Department of Medicine University of California San Francisco PUBLISHER: California Technology Assessment Forum DATE OF PUBLICATION: February 16, 2011 PLACE OF PUBLICATION: San Francisco, CA 1

2 VITAMIN D FOR THE PREVENTION OF OSTEOPOROTIC FRACTURES A Technology Assessment INTRODUCTION The California Technology Assessment Forum (CTAF) was asked to assess the evidence for the use of vitamin D to prevent osteoporotic fractures. There has been increasing interest and controversy surrounding vitamin D deficiency in the United States and the use of vitamin D supplements to prevent a myriad of diseases. 1-5 The Institute of Medicine reviewed the new literature and updated their dietary recommendations about dietary intake of vitamin D and calcium in The primary use of vitamin D has been in promoting healthy bones and preventing osteoporosis. However, several recent randomized trials have questioned the utility of vitamin D for the prevention of osteoporotic fractures, 7-9 and the most recent large trial reported a significant increase in the risk of both falls and fractures in patients randomized to receive supplemental vitamin D. 10 BACKGROUND Osteoporotic fractures People with poor bone quality and low bone mass, usually assessed by bone mineral density (BMD) are said to have osteoporosis. Osteoporosis itself is asymptomatic, but it increases an individual s risk for fractures. The most common fracture sites associated with osteoporosis are the hip, the distal forearm (wrist), and the vertebrae (spine). Hip fractures have the greatest impact on patients quality of life: half of patients have permanent limitations to their physical activity; 20% require long-term care in a nursing home, and up to 20% die in the year following the fracture In the United States, the lifetime risk for a hip fracture is approximately 18% in Caucasian women and 6% in men. 16 Similarly, the lifetime risk for a fracture of the distal forearm is approximately 16% in Caucasian women and 5% in men. Finally, the lifetime risk for at least one vertebral fracture diagnosed by a physician is approximately 16% in Caucasian women and 2.5% in men. 16 Moreover, the majority of vertebral fractures are not diagnosed clinically, but still may cause significant pain and limitations to physical activity. 17 Approximately nine million women and three million men in the United States have osteoporosis. In 2005, there were approximately two million fractures due to osteoporosis with annual direct costs exceeding $17 billion. 18 This includes approximately 300,000 hip fractures and 550,000 vertebral fractures, but does not include indirect costs associated with fractures 2

3 such as long-term rehabilitation programs or nursing home placement. 18 Bone density, an important risk factor for fracture, increases as children grow and reaches its peak at about age 30 years and declines thereafter. Osteoporosis is usually defined as a BMD at least 2.5 standard deviations below the average bone density for a 30 year old of the same sex. The number of standard deviation units from peak bone mass is called the T-score. Dual x-ray absorptiometry (DXA), a special form of low dose x-rays, is the preferred method to measure BMD. DXA measurements of BMD are usually done at the hip and lumbar spine as those are the most common sites for osteoporotic fractures. As described above, the majority of osteoporotic fractures occur in women. Other important risk factors for osteoporosis include older age, race (Caucasians and Asians are at increased risk), family history of hip fracture, early menopause, smoking, drinking three or more alcoholic beverages per day, limited physical activity, falls, low body weight, corticosteroid use, and hyperthyroidism. 19 Everyone loses bone mass as they age, but bone loss accelerates for women at the time of menopause. Estrogen therapy has been shown to slow the bone loss associated with menopause and to reduce a woman s risk for osteoporotic fractures. Parathyroid hormone analogs, bisphosphonates, nasal calcitonin, and selective estrogen receptor modulators are other classes of medications that can reduce a woman s risk for fractures. 20 Less expensive alternatives may include calcium and vitamin D. Vitamin D Vitamin D is the most important cause of nutritional rickets worldwide. Rickets is a disease of children characterized by stunted growth and soft bones that lead to bowing of the legs and arms. It results from inadequate mineralization of the bones primarily due to inadequate vitamin D, although inadequate calcium and phosphorous in the diet can also cause the disease. This led to the classification of vitamin D as a vitamin. As the biology of vitamin D and bone physiology was understood in greater detail, it became clear that vitamin D is better thought of as a pro-hormone. The activated form of vitamin D requires the addition of two hydroxyl groups. The liver adds the first hydroxyl group to form 25-hydroxyvitamin D (25(OH(D). The kidney adds the second hydroxyl group to form 1,25-dihydroxyvitamin D or calcitriol. Calcitriol is the biologically active form of vitamin D. Calcitriol acts on the intestines, kidneys, and bones to increase plasma levels of calcium and phosphate, which are required for bone mineralization. It primarily promotes calcium and phosphate absorption throughout the intestines. In combination with parathyroid hormone (PTH), calcitriol promotes resorption of 3

4 the bone matrix, which releases calcium and phosphate into the circulation. Calcitriol also slightly increases renal reabsorption of calcium. These activities cause plasma calcium levels to rise, which in turn decreases PTH secretion. Thus, the actions and levels of vitamin D, calcium and PTH are intimately linked. Many of the trials described below randomized patients to a combination of vitamin D and calcium and report the effect of supplementation on PTH levels. PTH levels should fall, thus decreasing bone resorption that is stimulated by high levels of PTH. Calcium and vitamin D also play a role in muscle function. 21 Recent meta-analyses of randomized trials report that vitamin D supplementation reduces the risk of fall in the elderly Since most fractures follow falls, a reduction in falls should translate into a reduction in fractures. The primary source of vitamin D in the human body is de novo synthesis of the vitamin by the skin when exposed to the ultraviolet B radiation from sunlight. This natural form of vitamin D, made by the skin, is called cholecalciferol, and is also known as vitamin D3. A synthetic form, derived from plant sterols, is called ergocalciferol or vitamin D2. Vitamin D2 has been commonly used in dietary supplements in the United States, although recent controversy about the bioequivalence of vitamins D2 and D3 has led to more frequent use of vitamin D Serum 25(OH)D is considered the best measure of overall vitamin D status. It reflects both dietary intake of vitamin D and endogenous production of vitamin D by the skin. Low serum 25(OH)D levels are associated with osteoporosis and predict future fractures The cutpoint for a normal serum 25(OH)D level is controversial. There is consensus that individuals with levels below 20 ng/ml are deficient in vitamin D. However, many experts argue that optimal levels are above 30 mg/ml. 26 Serum 25(OH)D levels tend to fall in populations during the winter compared to the summer months because of the shorter hours of daylight. 36 Additionally, in the developed world, individuals spend less time outside in sunlight and when outside are encouraged to keep their skin covered or to use sun block in order to prevent skin cancer. This limits the body s ability to naturally produce vitamin D. In order to prevent rickets, dairy products have been fortified with vitamin D. But many individuals consume very little dairy in their diet. Older individuals in particular have limited sun exposure and minimal vitamin D in their diet. Serum 25(OH)D levels clearly decrease with aging in the United States. 36 Thus, low 25(OH)D levels are very common, especially in the elderly, and may contribute to the increased incidence of fractures in this population. Recent data suggest that 35% to 40% of Americans over the age of 50 have vitamin D levels below 20 ng/ml and approximately 80% have levels below 30 ng/ml. 6 The Institute of Medicine (IOM) recently released updated recommendations about vitamin D intake. 6 The 4

5 Recommended Daily Allowance (RDA) represents the intake requirement that meets the needs of 97.5% of the population. The new RDA levels are 600 International Units (IU) per day for people ages 1 through 70 and 800 IU per day for people over the age of 70. Both were increased by 200 IU per day compared with their 1997 recommendations. The new RDA should be sufficient to maintain serum 25(OH)D levels above 20 ng/ml without significant sun exposure or dietary intake. The Tolerable Upper Intake Level (UL) is the intake level above which the potential for harm increases. The new UL is 4000 IU per day, up from 2000 IU per day. Low 25(OH)D levels are common and have been shown to predict bone loss, osteoporosis, and fractures. Vitamin D supplementation is also inexpensive and relatively safe. Thus, supplementation with vitamin D may be an important therapy to prevent fractures. However, excitement about the efficacy of high doses of other vitamins for the prevention of chronic diseases of aging has been tempered by the results of large clinical trials demonstrating either no benefits (folic acid, vitamin C) or increased mortality and net harm (vitamin E, beta carotene). 37 This assessment summarizes the randomized trial literature on vitamin D supplementation to prevent fractures. TECHNOLOGY ASSESSMENT (TA) TA Criterion 1: The technology must have final approval from the appropriate government regulatory bodies. The Food and Drug Administration (FDA) has regulatory responsibility for dietary supplements (including vitamins). However, a different set of regulations is used than that used for foods and drugs. Under the Dietary Supplement Health and Education Act of 1994 (DSHEA) the dietary supplement manufacturer is responsible for ensuring the supplement is safe prior to bringing it to market. The manufacturer is also responsible to ensure that the product label information is truthful. The FDA is responsible for post-marketing monitoring to ensure that unsafe products are removed from the market and that labeling is accurate. The Federal Trade Commission (FTC) regulates advertising of dietary supplements. There are some preparations of Vitamin D, both Calcitriol and Ergocalciferol which have been approved by the FDA for prescription use only. These preparations may be injectable, oral or nasal spray. TA Criterion 1 is met. 5

6 TA Criterion 2: The scientific evidence must permit conclusions concerning the effectiveness of the technology regarding health outcomes. The Medline database, Embase, Cochrane clinical trials database, Cochrane reviews database and the Database of Abstracts of Reviews of Effects (DARE) were searched using the key words ergocalciferol, cholecalciferol, vitamin D, vitamin D2, or vitamin D3. The results were crossed with the results from a search on fracture. The search was performed for the period from 1945 through January The bibliographies of systematic reviews and key articles were manually searched for additional references. References were also solicited from the manufacturers and local experts. The abstracts of citations were reviewed for relevance and all potentially relevant articles were reviewed in full. This review focuses on the effect of vitamin D on osteoporotic fractures as well as potential harms including hypercalcemia, kidney stones, renal function, and death. The search identified 996 potentially relevant studies (Figure 1). After elimination of duplicate and nonrelevant references including studies of fractures in young women and studies of activated forms of vitamin D, the search identified multiple publications describing 25 randomized or pseudo-randomized trials that reported osteoporotic fractures as one of their outcomes. 7-10,38-61 The search included pseudo-randomized trials 38,40,47 in which allocation to the intervention or control arm was primarily determined by date of birth as well as cluster randomized trials 53,56, even though both designs may introduce some selection bias. Sensitivity analyses were performed eliminating these trials from the meta-analyses. These 25 trials randomized 87,577 participants who developed 1,520 hip fractures, 480 clinical vertebral fractures, and 8,790 fractures in total. One trial, the Women s Health Initiative, contributed 36,282 participants, 374 hip fractures, and 4,260 total fractures. 8 6

7 Figure 1: Selection of studies for inclusion in review 996 potentially relevant references screened 196 abstracts for assessment 476 duplicate citations excluded 324 excluded: not randomized; reviews, abstracts only; other interventions 87 studies for full text review 109 studies excluded (Editorials, reviews, abstracts, no clinical outcomes) 62 studies excluded: young military recruits; no fracture outcomes; activated vitamin D interventions 25 studies included in assessment: 25 RCTs Level of Evidence: 1 through 5. TA Criterion 2 is met. TA Criterion 3: The technology must improve net health outcomes. The primary harm arising from osteoporosis is fracture. Hip fractures have the largest impact on patient s quality of life including long-term disability, loss of mobility, nursing home placement, and a significantly increased risk of death. Vertebral fractures can result in significant back pain, at times requiring hospitalization, as well as depression, reduced quality of life, and reduced respiratory function. However the majority of vertebral fractures are clinically silent. For that reason, the primary outcome reported in many trials is non-vertebral fractures. This assessment will summarize the outcomes in clinical trials that report data on hip fractures, non-vertebral fractures, symptomatic vertebral fractures, and total fractures. 7

8 The data from the randomized trials are described in Tables 1 through 5 below. For trials with multiple arms or a factorial design, either the vitamin D groups were combined and compared to the untreated controls or mutually exclusive groups were directly compared in two separate rows of the Tables. For example, the RECORD trial has been divided into two sub-studies: the results from the vitamin D only group were compared to those from the group that received neither supplemental vitamin D or calcium (Grant 2005a RECORD) and the results from vitamin D plus calcium group were compared to the calcium only group (Grant 2005b RECORD). Table 1 summarizes the methodological characteristics of the trials that primarily determine the quality of the trials. Table 2 summarizes the study characteristics including the size of the study, details of the intervention and control, the length of follow-up, and the primary outcomes. Table 3 describes the characteristics of the patients that reflect the inclusion and exclusion criteria of the trials. Table 4 describes the primary outcomes of the trials and Table 5 describes the reported adverse events. 8

9 Table 1: Randomized trials of vitamin D2 or D3 to prevent fractures Methodological Quality Study Randomization Allocation Groups comparable Outcome assessment Follow-up > Intention to treat Quality concealment blinded 80% analysis Inkovaara 1983a Pseudo By birthdate Yes Yes Yes Yes Yes Poor Inkovaara 1983b Heikinheimo 1992 Pseudo By birthdate Pseudo By birthdate Yes Yes Yes Yes Yes Poor Yes Yes No Poor Chapuy 1992, 1994 Yes Small difference in Yes Yes Fair height (153 versus 154 cm, p=0.003) Lips 1996 Yes Yes Yes Yes Yes Yes Good Dawson Hughes 1997 Yes Yes Yes Yes Yes Yes Good Komulainen 1998 Yes Yes Yes No Yes Yes Fair Pfeifer 2000 Yes Yes Yes Yes Fair Chapuy 2002 Yes Yes Yes Yes Fair Meyer 2002 Pseudo By birthdate Yes Yes Yes Yes Poor Bischoff 2003 Yes Yes Yes Yes No Yes Fair 9

10 Study Randomization Allocation Groups comparable Outcome assessment Follow-up > Intention to treat Quality concealment blinded 80% analysis Trivedi 2003 Yes Yes Yes Yes Yes Yes Good Avenell 2004 Yes No Yes No Yes Yes Fair Harwood 2004 Yes Yes No, differences in No Yes Poor baseline vitamin D and fractures Larsen 2004 Yes No, Vitamin D group Yes Yes Yes Poor older (p<0.001), others at p<0.001 Flicker 2005 Yes Yes No, more prior hip Yes Yes Yes Fair fractures in Vitamin D group Grant 2005a RECORD Yes Yes Yes Yes Yes Yes Good Grant 2005b RECORD Yes Yes Yes Yes Yes Yes Good Porthouse 2005 Yes Yes No Yes Yes Fair Jackson 2006 WHI Yes Yes Yes Yes Yes Good Law 2006 Yes Yes No Yes Yes Fair Lyons 2007 Yes Yes Yes Yes Yes Fair 10

11 Study Randomization Allocation Groups comparable Outcome assessment Follow-up > Intention to treat Quality concealment blinded 80% analysis Smith 2007 Yes Yes Yes Yes No Yes Fair Prince 2008 Yes No, vitamin D group Yes Yes Yes Fair shorter in height (p<0.05) Pfeifer 2009 Yes Yes Yes Yes Yes Fair Salovaara 2010 Yes No Yes No Yes Yes Fair Sanders 2010 VITAL D Yes Yes Yes Yes Yes Yes Good 11

12 Table 2: Randomized trials of vitamin D2 or D3 to prevent fractures - Study Characteristics Study N Location Randomization D2/D3 Vitamin D dose, IU Dose Frequency Calcium, mg Control Adherence FU, mo 1 endpoint Inkovaara 1983a Inkovaara 1983b Heikinheimo Finland 1 center 88 Finland 1 center 799 Finland 1 center Chapuy 1992, 3270 France nursing homes Lips Netherlands Dawson Hughes 1997 Komulainen center 389 United States 1 center 232 Finland 1 center Pfeifer Germany 1 center Chapuy France 55 apartment for elderly Meyer Norway 51 nursing homes Pseudo D Daily 0 Placebo 12 Pseudo D Daily 1000 Calcium 1000 Pseudo D2 150, , Annually 0 No placebo 41 Total fracture Simple D3 800 Daily 1200 Placebo 83% 18 Hip fracture Simple D3 400 Daily 0 Placebo 85% 42 Hip fracture Simple D3 700 Daily 500 Placebo 93% 36 Factorial with hormone therapy D3 300 Daily 93 Calcium BMD Simple D3 800 Daily 1200 Calcium % 12 Body sway Simple D3 800 Daily 1200 Placebo 95% 24 Pseudo D3 400 Daily 0 Placebo 62% 24 Hip fracture 12

13 Study N Location Randomization D2/D3 Vitamin D dose, IU Dose Frequency Calcium, mg Control Adherence FU, mo 1 endpoint Bischoff Switzerland 2 hospitals Trivedi England No center Avenell Scotland 1 center Harwood England 1 center Larsen Denmark 1 town Flicker Australia Grant 2005a RECORD Grant 2005b RECORD Porthouse 2005 Multiple 2675 United Kingdom 21 hospitals 2617 United Kingdom 21 hospitals 3314 England 107 general practices Simple D3 800 Daily 1200 Calcium 1200 Simple D3 100,000 Every 4 months 3 Falls 0 Placebo 80% 60 Simple D3 800 Daily 100 No placebo 85% 12 Recruitment Simple, 4 groups D2 D3 300, Annually Daily ± No placebo 12 BMD Cluster D3 400 Daily 1000 No placebo < 56% 42 Osteoporotic fractures Simple D Daily 600 Calcium % 24 Falls Factorial D3 800 Daily 0 Placebo 80% 45 Osteoporotic fractures Factorial D3 800 Daily 1000 Placebo + calcium % 45 Osteoporotic fractures Simple D3 800 Daily 1000 No placebo 63% 25 All fractures Jackson 2006 WHI North America Multicenter Simple D3 400 Daily 1000 Placebo 60% 84 Hip fracture Law United Kingdom 118 care homes Cluster D2 100,000 Every 3 months 13 0 No placebo 98% 10 Falls

14 Study N Location Randomization D2/D3 Vitamin D dose, IU Dose Frequency Calcium, mg Control Adherence FU, mo 1 endpoint Lyons Wales 497 care homes Smith United Kingdom 111 practices Prince Australia 1 center Pfeifer Germany and Austria 2 centers Salovaara 3195 Finland 2010 Populationbased Simple D2 100,000 Every 4 months 0 Placebo 80% 36 Fractures Simple D2 300,000 Annually 0 Placebo 98% 36 Nonvertebral fractures Simple D Daily 1000 Placebo + 86% 12 Falls calcium 1000 Simple D3 800 Daily 1000 Calcium 87% 20 Falls 1000 Simple D3 800 Daily 1000 No placebo < 86% 36 Any fracture Sanders 2010 VITAL D 2256 Australia 1 center Simple D3 500,000 Annually 0 Placebo 97% 36 Any fracture 14

15 Table 3: Randomized trials of vitamin D2 or D3 to prevent fractures - Patient Characteristics Study N Age, years Female, % Prior Fx Baseline 25(OH)D level, ng/ml Inkovaara 1983a Baseline BMD Institutionalized (I) or living in the community (C) I Inkovaara 1983b Heikinheimo I % C 40% I Chapuy 1992, ng/ml I Lips Hip fracture excluded 11 C Dawson Hughes 1997 Komulainen g/cm 2 at femoral neck g/cm 2 at femoral neck C C Pfeifer Osteoporotic 10 C fractures excluded Chapuy g/cm 2 at femoral neck C Meyer % prior hip fx 20 I 15

16 Study N Age, years Female, % Prior Fx Baseline 25(OH)D level, ng/ml Bischoff None in prior 3 months Baseline BMD Institutionalized (I) or living in the community (C) 12 I Trivedi Not excluded C Avenell %: 92% in past 3 months Harwood % hip fx w/surgery in past 7 days C g/cm 2 at femoral neck Larsen Not excluded 14 C C Flicker % Excluded if less than 10 or greater than 36 Grant 2005a RECORD Grant 2005b RECORD %, most in past 1 month %, most in past 1 month 15 C 15 C Porthouse % C I Jackson 2006 WHI % after age g/cm 2 at total hip C Law I 16

17 Study N Age, years Female, % Prior Fx Baseline 25(OH)D level, ng/ml Baseline BMD Institutionalized (I) or living in the community (C) Lyons I Smith % 23 C Prince C Pfeifer C Salovaara % g/cm 2 at femoral neck C Sanders 2010 VITAL D % since age C 17

18 Table 4: Randomized trials of vitamin D2 or D3 to prevent fractures - Study outcomes Study Inkovaara 1983a Inkovaara 1983b Heikinheimo 1992 Chapuy 1992, 1994 Vitamin D, n Control, n Lips Hip fracture, % 7.3% 9.4% 4.9% 6.7% 4.5% Vertebral fracture, % 2.3% 1.3% Non-vertebral fracture, % 10% 13% 10% Total fracture, % 2.2% 7.1% 0% 2.4% 16% 22% 25(OH)D, ng/ml during follow-up BMD, % change +2.7% -4.6% PTH, pg/ml Dawson Hughes 1997 Komulainen Pfeifer % 0% 0.5% 0.9% 1.7% 0% 9.5% 5.9% 13% 9.5% 13% 4.3% % -0.4% -4.1% -4.4% Chapuy Meyer % 6.9% 11% 8.8% 9.0% 25% 29% 12% % -2.4% 48 ~45, estimated from graph ~ % 13%

19 Study Vitamin D, n Control, n Bischoff Hip fracture, % 3.2% Vertebral fracture, % Non-vertebral fracture, % Total fracture, % 25(OH)D, ng/ml during follow-up 26 BMD, % change PTH, pg/ml Trivedi Avenell % 1.6% 1.8% 1.3% 2.1% 7.5% 9.0% 8.8% 11% 6.1% by heel ultrasound Harwood Larsen % 8.0% 14% 6.4% g/cm 2 at total hip 0.63 g/cm Flicker % 7.9% Grant 2005a RECORD Grant 2005b RECORD Porthouse % 3.1% 3.5% 3.7% 0.6% 0.3% 0.1% 0% 0.2% 11% 16% 15% 14% 14% 4.4% 16% 15% 14% 14% 4.4% Jackson 2006 WHI Law % 1.0% 1.1% 1.4% 1.0% 1.1% 4.6% 11% 11% 3.6% 4.6% 12% 12% 30 Increase at year % greater in Vitamin D group (p<0.001) 46 to % 2.6% 19

20 Study Vitamin D, n Control, n Lyons Hip fracture, % 6.5% Vertebral fracture, % 0.2% Non-vertebral fracture, % 12% Total fracture, % 12% 25(OH)D, ng/ml during follow-up 32 BMD, % change PTH, pg/ml Smith Prince % 1.4% 0.9% 0.5% 12% 6.5% 5.9% 13% 2.6% NS different 60 Non-significant decrease from baseline or control 151 Pfeifer % 5.8% Salovaara % 4.5% 11% 4.9% Sanders 2010 VITAL D % 0.6% 0.8% 5.1% 5.8% 14% 11% 22 ~28 Estimate from graph ~16 20

21 Table 5: Randomized trials of vitamin D2 or D3 to prevent fractures - Potential harms Study Vitamin D, n Control, n Inkovaara 1983a 45 Hypercalcemia, % Kidney stones, % Kidney dysfunction, % 2 0 GI distress, % Death, % Inkovaara 1983b Heikinheimo Chapuy 1992, Lips Dawson Hughes Komulainen Pfeifer Chapuy No significant effect Meyer

22 Study Vitamin D, n Control, n Bischoff Hypercalcemia, % Kidney stones, % Kidney dysfunction, % 0 GI distress, % Death, % Trivedi Avenell Harwood Larsen Flicker Grant 2005a RECORD Grant 2005b RECORD Porthouse <1%, no difference between groups <1%, no difference between groups <1%, no difference between groups <1%, no difference between groups <1%, no difference between groups <1%, no difference between groups No association with vitamin D Use of calcium associated with GI symptoms Jackson 2006 WHI Law

23 Study Vitamin D, n Control, n Lyons Hypercalcemia, % Kidney stones, % Kidney dysfunction, % GI distress, % Death, % Smith Prince Pfeiffer Salovaara Sanders 2010 VITAL D BMD I C Fx NS Not reported Bone mineral density Living in an institutional setting Living independently in the community Fracture Not significant 23

24 Comments on the trials The quality of the randomized trials was uneven. Four of the trials used pseudo-randomization by birthdate to allocate patients to either vitamin D or the control group which may have introduced some selection bias as the staff and investigators could predict a potential participant s allocation based on their birthdate. Another two trials used cluster randomization, but one had a small number of clusters and neither adjusted for clustering effects in their primary analysis. Seven of the trials did not use a placebo in the control group, effectively unblinding the study. Despite randomization, five of the trials had significant differences between the intervention and control groups. In addition, eleven of the trials did not blind their outcome assessment or did not report blinding for the assessment of study outcomes. There was significant heterogeneity in the patient populations studied, the interventions used, and the length of follow-up. Some trials included only individuals living in nursing homes, assisted living facilities, and other institutional settings. Other trials recruited individuals who lived independently in the community. The average age of participants ranged from 53 years in one trial that focused on recently menopausal women to 85 years in several of the trials that randomized patients in institutional settings. The proportion of women ranged from 24% in one study to 100% in eleven of the studies. Some trials excluded patients with prior fractures while others specifically studied patients with recent osteoporotic or hip fractures. Not all studies reported baseline vitamin D levels or bone density measurements and those that did primarily reported on a small sample of the participants. The average 25(OH) vitamin D levels in the studies ranged from 9 ng/ml to 30 ng/ml. Similarly, the average bone density at the femoral neck in the studies ranged from 0.57 g/cm 2 to 0.94 g/cm 2. Thus, some of the trials studied older institutionalized individuals with severe vitamin deficiency, osteoporosis, and prior fractures while others studied relatively young individuals with normal vitamin D levels and no prior fractures. Interventions included vitamin D2 and D3 at doses ranging from 300 IU per day to 500,000 IU annually. This dose range is much greater than the range typically tested in drug trials. Some of the trials included calcium as part of the intervention and some included calcium as part of the control intervention. Differences in the location of studies may also have influenced the results because the United States and Canada are the only countries that fortify dairy products with vitamin D. That is why vitamin D levels are typically much lower in studies performed in European countries compared to studies performed in the United States and Canada. Finally, in some trials the intervention lasted for only a few months with follow-up for less than 12 months. Other trials continued the intervention and follow-up for five years or more. If there is a time lag between the initiation of vitamin therapy and any beneficial effects on fracture incidence or adverse events, there may be heterogeneity in the outcomes based on length of followup. Because of the potential for effect modification in subgroups based on the patient populations studied and the differing interventions, meta-regression was used to assess important subgroup effects. Meta- 24

25 regression is a statistical technique that can help identify whether any of the differences in the patient populations studied or the study methods explain heterogeneity in the results of the different trials. 25

26 Primary meta-analysis All trials were included in the primary meta-analysis. Separate meta-analyses were performed for the following outcomes: all fractures (total fractures or non-vertebral fractures if total fractures were not reported), hip fractures, and vertebral fractures. Because of the heterogeneity in the patient populations and interventions, random effect models were used for all meta-analyses. Figure 2 is the Forest plot for all fractures. The diamonds represent the point estimates for the relative risk (RR) for fracture of vitamin D compared to the control intervention from each of the studies. The width of the horizontal lines represents the 95% confidence interval for each point estimate. A relative risk less than 1 (to the left of the vertical line) indicates that there was a lower incidence of fractures in the vitamin D group compared to the control group. The summary estimate indicates that there was a significant 7% overall reduction in fractures for participants in the vitamin D groups (RR 0.93, 95% CI 0.86 to 1.00, p = 0.04). Only four of the 26 studies had point estimates greater than 1. However 42% of the total variation across studies is due to heterogeneity rather than chance (I 2 = 42%, p=0.013). This suggests that results of some of the studies are so different that they should not be combined in a meta-analysis because they are likely measuring different biological phenomena. For instance, the 95% confidence intervals for the Chapuy trial (0.61 to 0.90) and for the Sanders trial (0.99 to 1.54) do not overlap at all. Potential explanations for the heterogeneity will be explored in the next section (Meta-regression results and stratified analyses). 26

27 Figure 2: Meta-analysis for all fractures 27

28 Figure 3 presents the results of the same analysis for hip fractures. Fewer studies reported hip fracture outcomes. Overall, there was a non-significant trend towards fewer hip fractures in the vitamin D groups (RR 0.97, 95% CI 0.86 to 1.09, p = 0.60). Eight of the 18 studies had point estimates greater than 1. There was less heterogeneity in the hip fracture outcomes (I 2 = 15%, p=0.272). Figure 3: Meta-analysis for hip fractures 28

29 Figure 4 summarizes the results for studies reporting non-vertebral fractures. Similar to hip fractures, there was a non-significant trend towards a 5% reduction in non-vertebral fractures (RR 0.95, 95% CI 0.88 to 1.02, p = 0.17). Only three of the 17 studies had point estimates greater than 1. However there was borderline significant heterogeneity in the studies reporting non-vertebral fracture outcomes (I 2 = 39%, p=0.053). Figure 4: Meta-analysis for non-vertebral fractures 29

30 Only seven studies reported vertebral fracture outcomes and all were clinically apparent fractures, not fractures detected solely by x-rays. Figure 5 summarizes the results. There was a larger, 18% reduction in vertebral fractures, but it was not statistically significant (RR 0.82, 95% CI 0.57 to 1.18, p = 0.29). Two of the seven studies had point estimates greater than 1. Even though the p-value for heterogeneity in the studies was not significant, the percentage of the variation beyond that expected by chance was modestly large (I 2 = 31%, p=0.19). Figure 5: Meta-analysis for vertebral fractures Thus, for all fracture outcomes there was a trend towards benefit, but also significant heterogeneity in the trials contributing to the summary estimate of the effect. In the next section, meta-regression was used to identify factors that strongly contribute to that heterogeneity. Meta-regression results and stratified analyses to understand heterogeneity Potential sources of heterogeneity in the results of these randomized trials arise from both the patient 30

31 populations studied and the differences in the therapies given to the intervention and control groups. Metaregression was performed to assess whether the average age of the study participants ( 75 years versus > 75 years), independence of the participants (community living versus institutionalized), average baseline 25(OH)D level (< 18 ng/ml versus 18 ng/ml), vitamin D dose ( 400 IU per day versus > 400 IU), vitamin D type (D2 versus D3), dose frequency (daily versus every three months or less), calcium supplementation for the vitamin D group (yes versus no), or calcium supplementation for the control group (yes versus no) significantly explained any of the heterogeneity in the results of the different studies. Only one of these characteristics reached statistical significance in the meta-regression: the use of calcium supplementation in the active control group reached statistical significance (p=0.009). Thus, the association between vitamin D and fracture in studies that gave a combination of vitamin D and calcium to the intervention group differed significantly from the association between vitamin D and fracture in studies that did not give supplemental calcium to the intervention group. Importantly, all of the studies that studied the combination of vitamin D plus calcium used daily dosing. However, meta-regression evaluating daily dosing versus intermittent dosing without calcium dosing in the model did not reach statistical significance (p = 0.21). Whether the participants were drawn from independently living individuals in the community or from nursing homes or assisted living institutions was borderline significant in the meta-regression model (p = 0.07) and will be explored further below. The metaregression did not suggest any significant heterogeneity in fracture outcomes based on the age of participants, vitamin D dose, vitamin D2 versus D3, baseline 25(OH)D levels, or calcium supplementation for the controls (p > 0.40 for each), although the power to detect an effect was low. Low power was particularly problematic for the baseline 25(OH)D level as most studies did not measure or report baseline levels. Furthermore, the individual data on patient characteristics such as age, sex, and 25(OH) D level are required to fully explore heterogeneity based on individual patient characteristics. One of the published meta-analyses 62 described in the section below (Comparison with published meta-analyses) obtained individual level data for participants in seven of the largest trials and should be considered the more definitive evaluation for individual patient characteristics. Stratified analyses help to make the results of the meta-regression analyses more clinically meaningful. The meta-analysis for all fractures stratified by the use of calcium supplements found the following. There was a significant reduction in fractures in the vitamin D group when the intervention included calcium supplementation (RR 0.95, 95% CI 0.92 to 0.99). There was not even a trend towards benefit from vitamin D when calcium supplementation was not included (RR 1.02, 95% CI 0.95 to 1.10). This suggests that vitamin D does not reduce fractures without concomitant supplementation with calcium. 31

32 Even though the meta-regression coefficients were not significant, when stratified by dosing, there was a significant reduction in fractures for the daily vitamin D group (RR 0.94, 95% CI 0.94 to 0.98), but not for vitamin D groups with therapy given every three, four or twelve months (RR 1.01). Again, it is difficult to separate this effect from that of calcium because all of the studies using intermittent vitamin D dosing did not give supplemental calcium to the vitamin D group. Similarly, there was a suggestion that studies that recruited participants from populations that had average baseline 25(OH)D levels below 18 ng/ml found benefit from vitamin D therapy (RR 0.90, 95% CI 0.83 to 0.98) while those with higher average baseline 25(OH)D levels did not find benefit (RR 0.99, 95% CI 0.94 to 1.04). However the test for interaction by baseline 25(OH)D level was not significant (p = 0.67). There was little power to detect an interaction because few studies measured baseline 25(OH)D levels and those that measured 25(OH)D only did so for a small subgroup of patients in each trial. Furthermore, only group average values could be used for stratification. Thus studies with low average 25(OH)D levels included participants with individual values greater than 18 ng/ml and studies with high average 25(OH)D levels included participants with individual values lower than 18 ng/ml. Thus, the result of our analysis stratified by 25(OH)D level must be interpreted cautiously. 32

33 Meta-analysis results for trials of daily vitamin D plus calcium stratified by institutional setting Based on the meta-regression, the original meta-analyses were repeated for studies of daily vitamin D plus calcium stratified by whether or not the participants were sampled from an institutional setting or not. Figure 6 summarizes the updated meta-analysis for all fractures. For studies that sampled participants living in institutional settings, daily vitamin D plus calcium reduced all fractures by 26% (RR 0.74, 95% CI 0.62 to 0.88, p = 0.001) without any unexplained heterogeneity beyond that expected by chance (I 2 = 0%, p for heterogeneity = 0.85). For studies that sampled participants living in the community, the summary relative risk for daily vitamin D plus calcium reduced the incidence of all fractures by 9% (RR 0.91, 95% CI 0.84 to 0.98, p = 0.019) with minimal unexplained heterogeneity (I 2 = 13%, p for heterogeneity = 0.32). If all of these studies are combined, there was a 14% reduction in all fractures (p = 0.001) with moderate, though nonsignificant heterogeneity. Figure 6: Stratified meta-analysis for all fractures 33

34 34

35 Figure 7 summarizes the updated meta-analysis for hip fractures. For studies that sampled participants living in institutional settings, daily vitamin D plus calcium reduced hip fractures by 26% (RR 0.74, 95% CI 0.56 to 0.97, p = 0.032) without any unexplained heterogeneity beyond that expected by chance (I 2 = 0%, p for heterogeneity = 0.42). For studies that sampled participants living in the community, daily vitamin D plus calcium reduced the incidence of hip fractures by 15% (RR 0.85, 95% CI 0.72 to 1.01, p = 0.059) with no unexplained heterogeneity (I 2 = 0%, p for heterogeneity = 0.82). If all studies are combined, there was a statistically and clinically significant 18% reduction in hip fractures (p = 0.007) with no heterogeneity (I 2 = 0%, p for heterogeneity = 0.83). Figure 7: Stratified meta-analysis for hip fractures 35

36 Figure 8 summarizes the updated meta-analysis for non-vertebral fractures. For studies that sampled participants living in institutional settings, daily vitamin D plus calcium reduced non-vertebral fractures by 26% (RR 0.74, 95% CI 0.62 to 0.89, p = 0.001) without any unexplained heterogeneity beyond that expected by chance (I 2 = 0%, p for heterogeneity = 0.86). For studies that sampled participants living in the community, daily vitamin D plus calcium reduced the incidence of non-vertebral fractures by a nonsignificant 5% (RR 0.95, 95% CI 0.88 to 1.02, p = 0.174) with minimal heterogeneity (I 2 = 0%, p for heterogeneity = 0.37). If all studies are combined, there was a statistically and clinically significant 12% reduction in non-vertebral fractures (p = 0.018), but there was moderate heterogeneity (I 2 = 40%, p for heterogeneity = 0.09). Figure 8: Stratified meta-analysis for non-vertebral fractures 36

37 Figure 9 summarizes the updated meta-analysis for vertebral fractures. None of the studies that sampled participants living in institutional settings reported vertebral fractures separately. For studies that sampled participants living in the community, daily vitamin D plus calcium reduced the incidence of vertebral fractures by 10% (RR 0.90, 95% CI 0.74 to 1.1, p = 0.27) with no unexplained heterogeneity (I 2 = 0%, p for heterogeneity = 0.40). Figure 9: Stratified meta-analysis for vertebral fractures In summary, daily vitamin D plus calcium significantly reduced the risk of incident fractures for individuals living in either an institutional or community setting. The reduction was greater for individuals in institutions than for patients in the community and was stronger for hip fractures than for vertebral fractures or other non-vertebral fractures. The results did not change during sensitivity analyses that dropped each study individually and dropped the poor quality studies. Harms High levels of vitamin D can cause hypercalcemia and if prolonged can result in soft tissue calcification with eventual damage to the kidneys and heart Recently, there has been increased concern that even high 37

38 normal levels of calcium in patients receiving calcium supplements with or without vitamin D may increase the risk for heart attacks and death from heart disease. 69,70 Table 5 summarizes the harms reported in the randomized trials. Hypercalcemia was rare (<1%), but occurred more frequently in patients randomized to vitamin D therapy. Kidney stones were also uncommon, but there was a statistically significant increase in kidney stones observed in the Women s Health Initiative study (2.5% versus 2.1%, p < 0.05). There were no differences in kidney function between the two groups in any of the studies. There were trends towards more gastrointestinal symptoms (constipation, bloating) in the vitamin D groups, but the differences were not significant, even in the Women s Health Initiative. Cardiovascular disease outcomes were not reported separately in the trials. The mortality outcomes are described in the next section. Thus, the only clear harm from vitamin D therapy was the excess in kidney stones in a trial that compared 400 IU of vitamin D3 plus 1000 mg of calcium to placebo. 38

39 Mortality Mortality outcomes are also summarized in Table 5. Mortality in the trials varied from 0% to 42% reflecting the large differences in the age and frailty between the populations studied in the trials of vitamin D therapy. Figure 10 summarizes the mortality data for all of the clinical trials. There was a trend towards an overall 6% reduction in total mortality (p = 0.17), but there was significant unexplained variation between the studies (I 2 = 67%, p for heterogeneity < 0.001). Figure 10: Meta-analysis for all-cause mortality in all studies of vitamin D therapy Figure 11 summarizes the mortality outcomes for the studies of daily vitamin D plus calcium. For studies that sampled participants living in institutional settings, daily vitamin D plus calcium did not significantly reduce mortality (RR 0.95, 95% CI 0.81 to 1.10, p = 0.48), although there was no unexplained heterogeneity 39

40 beyond that expected by chance (I 2 = 0%, p for heterogeneity = 0.81). For studies that sampled participants living in the community, daily vitamin D plus calcium reduced mortality by a non-significant 19% (RR 0.81, 95% CI 0.63 to 1.05, p = 0.11), but there was significant heterogeneity (I 2 = 81%, p for heterogeneity < 0.001). If all studies are combined, there was trend towards a reduction in mortality (RR 0.84, 95% CI 0.69 to 1.02, p = 0.078), but there was significant heterogeneity (I 2 = 40%, p for heterogeneity = 0.09). Figure 11: Stratified meta-analysis for all cause mortality for studies of daily vitamin D and calcium There was a non-significant trend towards decreased total mortality in all of the meta-analyses, including sensitivity analyses that left out each of the studies individually, left out the poor quality studies, left out the pseudo-randomized trials, and left out the cluster randomized trials. The heterogeneity of the outcomes remains unexplained, but there may be subgroups with a significant mortality benefit from vitamin D therapy. Meta-regression did not identify any significant explanations for the heterogeneity in the mortality outcomes. However, it is reasonable to conclude that there is no important increase in total mortality from vitamin D 40

41 therapy with or without calcium in the doses used in these trials. Comparison with published meta-analyses There have been a large number of meta-analyses attempting to make sense of the complicated and conflicting randomized trial data on vitamin D and fractures. 20,62,71-80 None of the meta-analyses included the two most recently published clinical trials. 9,10 Several of the early meta-analyses suggested that effective fracture prevention required doses of vitamin D of at least 700 to 800 IU per day. 72,76,78 The updated Cochrane meta-analysis and a recent patient level meta-analysis did not find heterogeneity by dose and concluded that 400 to 800 IU per day was sufficient. 62,77 The recent meta-analyses also concluded that vitamin D plus calcium was more effective than vitamin D alone and that vitamin D without calcium did not prevent fractures. 62,73,77,80 The Cochrane meta-analysis included 45 trials with 84,585 participants. 77 Their review included activated forms of vitamin D, such as calcitriol and alfacalcidol, in addition to vitamin D2 and vitamin D3. They found that vitamin D alone was not effective at preventing total fractures (RR 1.01, 95% CI 0.93 to 1.09) or hip fractures (RR 1.15, 95% CI 0.99 to 1.33). However, vitamin D with calcium was effective at reducing hip fractures (RR 0.84, 95% CI 0.73 to 0.96). There was a trend towards greater reduction in hip fractures for participants living in institutions compared to those living in the community, but the difference was not statistically significant (p=0.24). They also concluded that the harms from vitamin D plus calcium therapy are low, but caution should be used when considering supplementation for individuals with a history of kidney stones, kidney disease, or high serum calcium. A second meta-analysis published in 2010 by the vitamin D Individual Patient Analysis of Randomized Trials (DIPART) is instructive because the investigators obtain patient level data for 68,517 participants in seven of the largest randomized trials. 62 This allowed them to perform more precise subgroup analyses based on individual patient characteristics rather than the average characteristics of participants in the trials. The average age of the participants was 70 years and ranged from 47 to 107 years. The majority of the participants were women (85%). They found that vitamin D alone at 400 to 800 IU per day was not effective at preventing fractures. However, vitamin D plus calcium reduced total fractures (HR 0.92, 95% CI 0.86 to 0.99, p=0.025) and probably hip fractures (HR 0.84, 95% CI 0.70 to 1.01, p=0.07). There were statistically significant interactions in subgroup analysis for daily versus intermittent vitamin D, oral versus parenteral vitamin D, and supplemental calcium versus no calcium. There were no significant interactions by age, sex, use of hormone therapy, prior hip fracture, prior vertebral fracture, any other previous fracture, or vitamin D dose. There was a trend towards less efficacy of vitamin D therapy in the subgroup receiving 41

42 bisphosphonate treatment (p=0.06). The DIPART group also calculated the number needed to treat (NNT) to prevent one fracture. For the entire population studied 213 individuals needed to receive vitamin D with calcium for three years to prevent one fracture. For patients at higher risk for fracture the NNT was lower. For example, for individuals over the age of 70, then NNT dropped to 111 and for patients who had suffered a prior fracture, irrespective of their age, the NNT was 82. Thus, the published meta-analyses generally agree with the findings of the meta-analysis in this assessment. Vitamin D alone does not appear to be effective for fracture prevention, but vitamin D plus calcium is effective. The trials demonstrating efficacy all used daily dosing of vitamin D at 400 to 800 IU per day plus calcium at doses mainly between 1000 and 1200 mg/day. The majority of those trials used vitamin D3 rather than D2, though it is not clear from the trial data that vitamin D3 offers any advantages over vitamin D2. It remains uncertain whether treatment at less frequent intervals with larger vitamin D doses is effective when used in combination with daily calcium because none of the published trials that treated at intervals of three months or longer included calcium supplementation. The risk of harms from vitamin D plus calcium therapy is low, but individuals with kidney stones, renal insufficiency, or hypercalcemia should be treated cautiously. Ongoing trials of vitamin D There are a many ongoing trials of vitamin D indexed on ClinicalTrials.gov. The largest is the Vitamin D and Omega-3 Trial (VITAL), which is randomizing 20,000 men and women using a factorial design to 2000 IU of vitamin D3 and/or one gram of omega-3 fatty acids from fish oil. The primary research question is whether these supplements can reduce the risk of developing cancer, heart disease, and stroke in people who do not have a prior history of these illnesses. TA Criterion 3 is met. TA Criterion 4: The technology must be as beneficial as any established alternatives. There are a number of established therapies to prevent fractures in patients with established osteoporosis. These include PTH analogs, bisphosphonates, selective estrogen receptor modulators, and estrogens. 20 There are no head to head trials comparing vitamin D to any of the above therapies. However, most of the pivotal randomized trials that demonstrated the efficacy of these interventions treated both the intervention 42