Introduction

Multivitamin/mineral supplements have been used by many as a simple means to ensure adequate intake of several essential micronutrients in the hope for prophylactic benefits. Typical multivitamin/mineral supplements on the market contain about 10 vitamins and 10 minerals, such as vitamin A, vitamin C, B vitamins, vitamin E, folic acid, vitamin D, calcium, magnesium, zinc, iron among others. The following section summarizes the evidence from randomized controlled trials on the efficacy of multivitamin/mineral supplement use in the prevention of chronic disease.

Results of literature search for Key Question 1

Our literature search identified 11 articles from randomized controlled trials that addressed the efficacy of multivitamin/mineral supplements in the primary prevention of cancer, cardiovascular disease, cataract and age-related macular degeneration. Data for other diseases were lacking (Table 1). These studies used designed vitamin/mineral combinations, but not the one-a-day type of multivitamin supplements available on the United States market.

Table 1

Number of articles by key questions and disease categories.

The 11 articles documented results from 5 randomized controlled trials published from 1993 to 2005, including (1) the Linxian General Population Trial in China,^{64 65 66 67 68}, (2) the Supplementation en Vitamines et Mineraux Antioxydants (SU.VI.MAX) study in France,^{69 70} (3) the Multi-center Ophthalmic and Nutritional Eye-Related Macular Degeneration Study (MONMD) in United States veterans,⁷¹ (4) the Roche European American Cataract Trial (REACT) in the United States and United Kingdom,⁷² and (5) the Age-Related Eye Disease Study (AREDS) in the United States.⁷³

Design of randomized controlled trials

The Linxian General Population Trial (referred to as “Linxian Trial” henceforth) was a fractional factorial trial designed to determine the efficacy of 8 vitamin/mineral combinations in cancer prevention in 29,584 adults of ages 40 to 69 years from 4 Linxian communes⁶⁴ where the rates of esophageal cancer were high. Users of any vitamins were ineligible for trial participation. Vitamin/mineral supplements were combinations of the following: (A) retinol 5000 IU and zinc 22 mg, (B) riboflavin 3 mg and niacin 40 mg, (C) vitamin C 120 mg and molybdenum 30 μg, and (D) β-carotene 15 mg, α-tocopherol 30 mg, and selenium 50 μg. The combinations were AB, AC, AD, BC, BD, CD, ABCD, and placebo. The dose of each nutrient ranged from 1 to 2 times the United States RDAs. The follow up period was 1986 to 1991. At the end of the trial, 3,249 participants had eye exams,⁶⁵ and 391 participants had esophageal/gastric endoscopy examinations⁶⁸ (Appendix F ^a, Evidence Tables 1a–1c).

The SU.VI.MAX study was designed to determine the efficacy of a daily supplement of antioxidants (vitamin C 120 mg, vitamin E 30 mg, β-carotene 6 mg, selenium 100 μg, and zinc 20 mg) for the primary prevention of cancer and ischemic cardiovascular disease in 13,017 French adults (7,876 women of age 35 to 60 years, and 5,141 men of age 45 to 60 years).⁶⁹ Regular users of any of the vitamins/minerals provided in the study were ineligible for trial participation. The follow up period was 1994 to 2002. Women had higher baseline serum β-carotene levels than men. Women also had slightly higher baseline serum levels of vitamin C but lower levels of zinc and selenium. Information on self-selected supplement use was not provided (Appendix F, Evidence Tables 1a–1c).

The MONMD study was aplacebo-controlled trial conducted in 1992 to evaluate nutritional status in 71 United States veterans with dry age-related macular degeneration (AMD) and to assess the efficacy of multivitamin/mineral supplement use for 18 months on the progression of AMD and potential side effects. The daily multivitamin/mineral supplements included β-carotene 20,000 IU, vitamin E 200 IU, vitamin C 750 mg, citrus bioflavonoid complex 125 mg, quercitin 50 mg, biberry extract 5 mg, rutin 50 mg, zinc picolinate 12.5 mg, selenium 50 mcg, taurine 100 mg, n-acetyl cysteine 100 mg, l-glutathione 5 mg, vitamin B2 25 mg, and chromium 100 mcg. The study excluded people who had used vitamins in the year prior to enrollment.⁷⁴ The instruments used to measure cataract transparence were changed during the study period, but the examiners were not well instructed on how to use the new instruments⁷¹ (Appendix F, Evidence Tables 1a–1c).

REACT assessed the efficacy of a mixture of antioxidant supplements in preventing cataract progression among 297 individuals in Boston, United States and Oxford and Bradford, United Kingdom.⁷² Regular users of any vitamin supplement were also excluded. Participants took a placebo or combined β-carotene (6 mg, in the form of beadlets), vitamin C (250 mg), and all-rac α-tocopherol acetate (200 mg) 3 times per day with meals. The follow up period was 1990 to 1995. (Appendix F, Evidence Tables 1a–1c).

The AREDS study was an 11-center trial that assessed the efficacy of zinc (80 mg zinc oxide and 2 mg cupric oxide) and antioxidants (vitamin C 500 mg, vitamin E 400 IU, and β-carotene 15 mg) in the development and progression of age-related lens opacities and visual acuity loss in the United States.^{73, 75} Participants were classified into 4 AMD categories according to the size and the extent of drusen and retinal pigment abnormality in each eye, the presence of manifestations of advanced AMD, and visual acuity. Persons in AMD category 1 (n= 1,117) were assigned to antioxidant or placebo, whereas persons in AMD categories 2 to 4 (n=3,640) were assigned to placebo, antioxidants, zinc, or combined antioxidants and zinc. The follow up period was 1992 to 2001. The major limitations were the option of multivitamin use (by 66% of the participants) and self-selected use of non-study supplements (20% of participants) that contain at least one of the study nutrients (Appendixes F, Evidence Tables 1a–1c).

Similarity and heterogeneity in study design among trials

The Linxian trial was conducted in a Chinese population that was nutritionally inadequate whereas SU.VI.MAX was conducted in an apparently healthy French population. The Linxian trial and the REACT study excluded any vitamin use without specifying how recent the use was. The MONMD study excluded persons with supplement use during the year prior to enrollment. In contrast, AREDS provided Centrum^® to 66 percent of the study participants, in addition to study supplements, and SU.VI.MAX allowed use of supplements other than those under study. The Linxian trial and the SU.VI.MAX study used doses of 1–2 times RDAs. In contrast, MONMD used vitamins C and B2 at doses that were more than 10 times the RDAs; AREDS used high doses of vitamin E and zinc (10 times the RDA), and a moderate dose of vitamin C (6 times the RDA); REACT used a high dose of vitamin E. All trials employed a parallel-arm design except for the Linxian trial that used a fractional factorial design. A total of 47,289 individuals were included in this review section (Appendixes F, Evidence Tables 1a–1c; Table 2).

Table 2

Summary of randomized controlled trials on multivitamin/mineral supplements and chronic disease prevention.

Study quality

Inclusion/exclusion criteria were clearly defined in most trials. Quality of these trials was good in terms of randomization, double masking, ascertainment of trial endpoints, adherence, and use of intention-to treat approach in statistical analyses. However, there was a lack of descriptions as to whether concealment of allocation sequence was done, and whether observers independently evaluated trial outcomes. There was a paucity of data on prior supplement use, concomitant supplement use, and medication use that may have had effects on the efficacy of study supplements. None of the trials reported success of blinding and the extent of unintended crossover. Only the AREDS and REACT studies provided information on numbers and reasons for withdrawals and percents of loss-to-follow-up (Table 3).

Table 3

Assessment of the quality of randomized controlled trials on the efficacy of multivitamin/mineral supplements and single nutrients in the prevention of chronic diseases and conditions.

Cancer

The Linxian trial examined incidence of and mortality for all cancer, esophageal cancer, stomach cancer (cardia and noncardia), esophageal/gastric cardia, and other cancers.⁶⁴ After 5.25 years of follow up, no significant risk reduction by supplement use was observed for these endpoints. The only exceptions were the reductions in gastric cancer incidence (relative risk (RR) 0.84, 95% confidence interval (CI) (0.71–1.00)), cancer mortality (RR 0.87, 95% CI 0.75–1.00), especially stomach cancer mortality (RR 0.79, 95% CI 0.64–0.99) in the groups receiving β-carotene, vitamin E and selenium compared to the groups receiving other vitamin/mineral combinations,⁶⁴ and a lower non-cardia stomach cancer mortality in those receiving retinol and zinc (RR 0.59, 95% CI 0.37–0.93).⁶⁴ Reduction in cancer mortality was greater in women than in men and among those of age less than 55 years in this trial (RR 0.79, 95% CI 0.64–0.98) vs. RR 0.93, 95% CI 0.77–1.12), and (RR 0.71, 95% CI 0.55–0.92) vs. RR 0.94, 95% CI 0.80–1.11), respectively).⁶⁷ In the substudy where participants underwent endoscopy examination, there was no significant effect of β-carotene, vitamin E and selenium supplement use on worse overall diagnoses of esophageal and gastric cancer or combined cancer and dysplasia prevalence, although the odds ratios were in the protective direction ⁶⁸ (Appendixes F, Evidence Tables 1b–1e).

The SU.VI.MAX study reports no benefit on overall cancer incidence by the antioxidant supplement use in women (RR 1.04, 95% CI 0.85–1.29), but a 31 percent risk reduction (RR 0.69, 95% CI 0.53–0.91) in men.⁶⁹ As a result, there was a statistically significant interactive effect of sex and randomized group on total cancer incidence (p=.02). Women were younger than men in this trial, and generally had a healthier lifestyle as evident by higher serum β-carotene and vitamin C and fewer smokers. Among men, a moderate reduction in prostate cancer risk was observed in the antioxidant supplement group (RR 0.88, 95% CI 0.60–1.29). Further stratification analysis showed differential efficacy by baseline prostate specific antigen (PSA) level with a risk reduction among men with normal baseline PSA (≤3 μg/L) (hazard ratio (HR) 0.52, 95% CI 0.29–0.92), but not among men with elevated PSA (HR 1.54, 95% CI 0.87–2.72)⁷⁰ (Appendixes F, Evidence Table 1d, Figures 3 and 4).

Figure

Figure 3. Relative risk (RR) of total cancer, gastric cancer and esophageal cancer incidence in relation to multivitamin/mineral supplement use.

Figure

Figure 4. Relative risk (RR) of total cancer, gastric cancer and esophageal cancer mortality in relation to multivitamin/mineral supplement use.

Cardiovascular disease

The Linxian trial reported a non-significant lower risk of stroke mortality with the greatest risk reduction (RR 0.91, 95% CI 0.76–1.07) observed in those receiving β-carotene, selenium, and α-tocopherol with or without other study nutrients,⁶⁶ particularly in those receiving the combination of β-carotene, selenium, α-tocopherol, retinol and zinc (RR 0.71, 95% CI 0.50–1.00) as compared to the counterpart. There was no sex difference in the risk reduction. Hemorrhagic and ischemic stroke was not distinguished but other sources showed that approximately two-thirds of the strokes were ischemic in this population⁷⁶ (Appendixes F, Evidence Tables 1b–1e).

In the SU.VI.MAX study, no significant difference in ischemic cardiovascular disease incidence was noted between randomized groups. There was no interaction between sex and randomized groups. The cardiovascular events in women were only 22.6 percent of the events in men⁶⁹ (Appendixes F, Evidence Table 1d, Figure 5).

Figure

Figure 5. Relative risk (RR) of cardiovascular disease incidence in relation to multivitamin/mineral supplement use.

Total mortality

In the Linxian trial, total mortality was lower among those who received β-carotene, selenium, and vitamin E, but not other nutrient combinations (RR 0.91, 95% CI 0.84–0.99).⁶⁶ In the AREDS study, total mortality was 6 percent higher in the group receiving antioxidants compared to the group receiving no antioxidants, but the increase was not statistically significant.^{64, 73, 75} When limited to those participants with AMD categories 2, 3, and 4, total mortality was 19 percent and 13 percent lower in the groups receiving zinc alone or zinc combined with antioxidants respectively.⁷³ A sex difference in the relative risk for total mortality was documented in the SU.VI.MAX study (RR 0.63, 95% CI 0.42–0.93 in men and RR 1.03, 95% CI 0.64–1.63 in women)⁶⁹, but no sex or age differences were noted in the Linxian trial⁶⁷ In the REACT, 9 deaths occurred in the antioxidant group, whereas 3 deaths occurred in the placebo group. Further examination on the causes of death revealed that the deaths in the antioxidant group were due to esophagitis, sudden death, aneurysm, pulmonary fibrosis, cancer and coronary thrombosis (Appendix F, Evidence Table 1e, Figure 6).

Figure

Figure 6. Relative risk (RR) of all cause mortality in relation to multivitamin/ mineral supplement use.

Cataract and age-related macular degeneration

In the Linxian trial, there was no effect of combined vitamin E, selenium and β-carotene on nuclear cataract, cortical cataracts, or posterior subcapsular cataracts⁶⁵ (Appendix F, Evidence Table 1d).

In the MONMD study, distance acuity declined in the placebo group, but was unchanged in the multivitamin group (p=.03). The multivitamin group also had better M print acuity and fewer scotoma in left eyes in the multivitamin group (p=.07), after 12 months. There was no significant difference between randomized groups in refraction, metamorphopsia and Lens Opacities Classification System (LOCS) II readings on nuclear color, nuclear opalescence, and posterior subcapsular opacities. There was an unanticipated cortical cataractogenic effect for right eyes in the multivitamin group.⁷¹ (Appendix F, Evidence Tables 1d).

In the REACT, the primary outcome was the difference between baseline and the last visit in percentage pixel opaque (IPO) in the anteriorly-focused, retroillumination image. Secondary outcomes were posterior subcapsular cataract, nuclear cataract, cortical cataract, and nuclear color. At the end of the second year, there was a small positive effect on percent IPO in both the United States and United Kingdom groups. After the third year, the positive effects were greater in the United States group (percent pixel opaque = 0.389 vs. 2.517 in the vitamin vs. placebo group, p=.0001), but not the United Kingdom group. Unfavorable changes in all secondary outcomes were smaller in the vitamin group than the placebo group, but none was statistically significantly different⁷² (Appendix F, Evidence Table 1d).

In the AREDS study on cataract, outcome measures were cataract surgery, changes in photographic grade of nuclear, cortical and posterior subcapsular opacities, and visual acuity loss (≥15 letters). After 6 years of follow up, no appreciable difference was found in any of the outcomes between antioxidant and placebo groups⁷³ (Appendix F, Evidence Table 1d).

In the AREDS study, outcomes were rates of progression to advanced AMD and visual acuity. After an average follow up period of 6.3 years, the odds ratio (OR) (99% CI) of developing advanced AMD was 0.75 (0.55–1.03), 0.80 (0.59–1.09), and 0.72 (0.52–0.98) among individuals with zinc, antioxidants, and combined zinc and antioxidant supplementation as compared to individuals in the placebo group. Excluding individuals in AMD category 2 (extensive small drusen, nonextensive intermediate size drusen or pigment abnormalities), the OR (99% CI) of developing advanced AMD was 0.71 (0.52–0.99), 0.76 (0.55–1.05), and 0.66 (0.47–0.91) among individuals with zinc, antioxidants, and combined zinc supplementation and antioxidant supplementation, and the OR(99% CI) of having moderate visual acuity loss was 0.73 (0.54–0.99) in the group with antioxidants plus zinc, but not statistically significant for other supplementation groups⁷⁵ (Appendix F, Evidence Table 1d).

Summary

There is a paucity of data on the efficacy of multivitamin/mineral supplement use in the prevention of chronic disease in the general United States population. Limited data from the Linxian trial suggest 13 percent to 21 percent reductions in gastric cancer incidence, gastric cancer mortality, and cancer mortality by use of β-carotene, vitamin E and selenium supplements of doses 1 to 2 times RDAs. Results of total cancer incidence in the SU.VI.MAX trial in France were sex-dependent with a 31percent lower risk in men who received vitamin C, vitamin E, β-carotene, selenium, and zinc at doses near RDAs, but no risk reduction in women who appeared to have had higher fruit/vegetable intake. The antioxidants used in SU.VI.MAX did not confer benefit in preventing ischemic cardiovascular disease, whereas use of β-carotene, selenium, α-tocopherol, retinol, and zinc supplements in the Linxian trial had a moderate reduction (30%) in stroke mortality. Generalizability of these findings for the United States population is uncertain in view of the French paradox and the general nutritional inadequacy of the Linxian population. Multivitamin/mineral supplement use for 3 to 6 years had no significant benefits in preventing cataract. Zinc (of dose 10 times thhe RDA) alone or in combination with antioxidants had beneficial effects on AMD only in those with intermediate AMD in one or both eyes, or those with advanced AMD in one eye. Overall, the quality of individual articles was “medium” (Table 3). Taking into consideration the quantity, quality, and consistency of evidence, we concluded the strength of evidence on the efficacy of multivitamin/mineral supplementation was rated as “very low” for primary prevention of cancer and cardiovascular disease, and “low” for cataract and age-related macular degeneration (Table 4).

Table 4

Grading of the quality of evidence of the efficacy of multivitamins/minerals in the prevention of chronic disease.

Issues to consider

Because the most recent revisions of recommended nutrient intakes, the 1997-2004 dietary reference intakes (DRIs), include for the first time an upper level of intake, this concept has been used as a benchmark to assess the ‘safety’ of micronutrient intake. However, it is important to point out that the UL was designed to identify risk, not safety. Risk is a probabilistic, biological, objective indicator of the potential adverse effect resulting from a defined intake level. The risk associated with a given intake level is expected to be similar for comparable human populations. Safety, on the other hand, is a social, cultural and intellectual construct, and reflects the risk that a given society is willing to tolerate. This threshold varies in different cultures and societies, and can change over time. The distinction is of relevance for our review since, in the absence of standardized methods to assess risk associated with nutrient intakes, studies report these adverse or unexpected events in a variety of ways, in some cases reflecting more a subjective self-assessment of ‘safety’, and in others a more specific assessment of risk based on objective indicators, such as laboratory tests.

Very few studies have been specifically designed to assess the risk associated with different intake levels of single or multiple micronutrients. Nevertheless, many randomized controlled trials include a data collection component aimed at monitoring safety, thus providing information on adverse events in active and control groups. These data typically include a variety of endpoints, from spontaneous or elicited self-reported symptoms or events, exit surveys in participants withdrawing from the study, or objective measurements such as blood or urine tests or clinical examination. It should be noted that most randomized controlled trials reviewed in this evidence report used one or more nutrients at doses above the UL defined by the current DRIs. Besides randomized controlled trials, additional insight on risk associated with specific nutrients can be obtained from other types of studies, including case series and case reports, usually of very small sample size (often single case reports). Not surprisingly, many case reports describe the effects of very high intake levels or of unusual host conditions, thus limiting their generalizability.

The basic conditions that enhance the quality of a study in terms of determining the main health effects also apply to adverse effects: temporal association, adequate exposure, dose-response relationship, biological plausibility, and specificity, etc. In the case of safety, reversal of effects upon withdrawal may also enhance the solidity of the findings.

Review of data on the safety/risk of multivitamin/mineral supplements

We identified 8 articles that reported the adverse effects of multivitamin/mineral preparations. The 8 articles were published from 4 randomized controlled trials and 3 case reports.^{72–75, 77–81} We considered the following criteria when assessing adverse effects: (1) randomized allocation of treatment, (2) adequate sample size, (3) well-defined population, (4) defined dose and total intake of the nutrient(s) of interest, and (5) adequate duration of exposure. We used the following criteria for assessing causality: temporal relationship, lack of alternative causes, dose-response relationship, evidence of increased circulating levels of the nutrient under investigation, and response to re-challenge.

Doses were usually 2 to 10 times the RDA. For example, typical daily dosage for vitamin E doses ranged from 200 to 600 IU, vitamin A from 10,000 to 20,000 IU, and vitamin C from 75 to 750 mg. Overall, we found no consistent pattern of increased adverse events in the active group compared with the placebo group, with the exception of changes in skin color, which was common in studies in which β-carotene was part of the multivitamin preparation (Appendix F, Evidence Tables 2a–2d).

The REACT study evaluated the effects of an antioxidant vitamin combination (750 mg vitamin C, 600 mg vitamin E, and 18 mg β-carotene), given daily for 3 years. The frequency of reported side effects did not differ between intervention and control groups⁷² (Appendix F, Evidence Tables 2a–2d).

In the AREDS trial,⁷³ an antioxidant combination (400 IU vitamin E, 500 mg vitamin C, 15 mg β-carotene) and/or 80 mg zinc and 2 mg Cu, was given to healthy adults with early signs of lens opacity. The only significant effect of the antioxidant supplement was yellowing of the skin (Appendix F, Evidence Tables 2a–2d). A similar study enrolling patients with incipient macular degeneration,⁷⁵ and using a similar antioxidant combination, also found a higher percent of yellowing of the skin in the active group (8.3% vs. 6.0%, p<0.008).

The MONMD trial assigned 39 patients with macular degeneration to an antioxidant combination, with follow up of 18 months.⁷⁴ No adverse effects were reported, except for “a few cases of diarrhea,” which the authors attributed to the high ascorbic acid content of the preparation.

In a 2 by 4 factorial feasibility trial in Yunnan Province, China, where the incidence of lung cancer was extremely high, participants received combinations of retinol 25,000 IU, β-carotene 50 mg, α-tocopherol 800 IU and selenium 400 μg each day, and there were no excessive adverse effects reported for the active supplement groups. Symptoms such as broken nails and skin yellowing were generally improved in the groups receiving active supplements.⁷⁸

In the 3 trials^{66 73 69} of multivitamin supplements where mortality rates were compared between active and control groups, no adverse effects of supplementation on the outcomes were found. In fact, two trials reported lower mortality in the groups receiving multivitamin/mineral supplements. ^{66 69} Few if any studies met all or even a few of the causality criteria (Appendix F, Evidence Table 1e).

β-Carotene

Introduction

β-carotene is a major dietary carotenoid and the most abundant carotene found in nature. In the 1980s, several large clinical trials had been launched to determine the role of β-carotene in chronic disease prevention. The following section summarizes the evidence.

Results of the literature search

We identified 22 articles from randomized controlled trials that assessed the efficacy of β-carotene in the prevention of cancer, cardiovascular disease, diabetes mellitus, or age-related maculopathy. The 22 articles were published from 6 trials, the Alpha-tocopherol β-carotene Cancer Prevention Study (ATBC), the Βeta-Carotene and Retinol Efficacy Trial (CARET), the Nambour Skin Cancer Prevention Trial (NSCP), the Skin Cancer Prevention Study (SCP), the Physician's Health Study (PHS), and the Women's Health Study (WHS).^82–87

Design of randomized controlled trials

The ATBC was a 2 by 2 factorial trial of synthetic all-rac-α-tocopherol acetate (50% powder, 50 mg per day) and synthetic β-carotene (10% water-soluble beadlets, 20 mg per day) supplementation in 29,133 Finnish smokers aged 50 to 69 years.⁸⁸ Users of vitamin E, vitamin A, and/or β-carotene in excess of predefined doses were excluded. The follow up period was 1985 to 1993. Post-intervention follow up of cancer incidence and cause-specific mortality was performed from 1993 to 1999 for cancer incidence and cause-specific mortality and up to 2001 for total mortality.⁸⁹ Gastrointestinal endoscopy was performed on 1,344 men with gastritis after a median supplementation time of 5.1 years⁹⁰ (Appendix F, Evidence Tables 3a–3c).

The CARET study consisted of two pilot studies conducted in 1985 to 1988 followed by a large trial conducted from 1988 to 1991 in the United States. The first pilot study, the Asbestos Workers Pilot Study for CARET, involved 816 men with a history of asbestos exposure. ⁹¹ The second pilot study, the Smokers Pilot Study, involved 1029 men and women with a history of cigarette smoking.⁹² The full CARET study was conducted in 18,314 high-risk men and women who had a history of asbestos exposure or smoking. Participants were randomly assigned to receive either β-carotene 50 mg and retinyl palmitate 25,000 IU per day or placebo. Prior β-carotene supplement users were excluded.⁹³ The follow up period was 1985 to 1995. Post-trial follow up of cancer incidence and mortality was performed until the end of 2001⁹⁴ (Appendix F, Evidence Tables 3a–3c).

The NSCP trial was a 2 by 2-factorial trial of β-carotene 30 mg per day and daily sunscreen among 1,621 adult Australians of age 20 to 69 years.⁸⁴ No exclusion or prior supplement use was reported. The follow up period was 1992 to 1996 (Appendix F, Evidence Tables 3a–3c).

The SCP was a trial with a parallel-arm design conducted in 1,729 adults of age 85 years or less who had at least one biopsy-proven basal cell or squamous cell skin cancer at baseline. Participants were randomized to receive placebo or β-carotene (50 mg per day) during the trial.⁸⁵ No exclusion or prior supplement use was reported. The follow up period was 1983 to 1993 (Appendix F, Evidence Tables 3a–3c).

The PHS was a 2 by 2 factorial trial of β-carotene (50 mg every other day) and aspirin conducted among 22,071 apparently healthy male physicians, aged 40–84 years, in the United States. Vitamin A supplement users were ineligible for trial enrollment. The follow-up period was 1982 to 1995 ⁹⁵ (Appendix F, Evidence Tables 3a–3c).

The WHS was a 2 by 2 by 2 factorial trial conducted in 39,876 female health professionals in the United States aged 45 years or older to determine whether alternate daily use of aspirin (100 mg), β-carotene (50 mg), and vitamin E (600 IU) can prevent cancer and cardiovascular disease.⁸⁷ β-carotene supplementation was terminated after a median treatment duration of 2.1 years (range 0 to 2.7 years), primarily because of the null findings from the PHS.⁹⁵ Users of vitamin A, β-carotene, or vitamin E were ineligible for trial enrollment. Nearly 40 percent of the trial participants reported to have multivitamin supplement use at baseline. The follow up period of this trial was 1992 to 2004 (Appendix F, Evidence Tables 3a–3c).

Similarity and heterogeneity in study design among trials

Except for the ATBC and NSCP that were conducted in Finland and Australia, respectively, the 4 other trials included in this review were conducted in the United States. Except for the SCP and NSCP, prior users of β-carotene and/or vitamin A supplements were excluded. The range of follow up was 4 to 10 years. The range of daily doses was 20 mg to 50 mg. The ATBC, PHS, WHS, and NSCP used a factorial design with α-tocopherol, aspirin, aspirin with/without vitamin E, and sunscreen, respectively, as the other intervention, whereas SCP and CARET adopted a parallel-arm study design. A total of 112,564 individuals were included in this review section. They were heterogeneous populations that ranged from high-risk people with a history of asbestos exposure and cigarette smoking (ATBC, CARET) to male physicians (PHS), female health professionals (WHS), and adults in a high sun exposure area in Australia (Appendix F, Evidence Tables 3a–3c).

Study quality

The general strengths of the randomized clinical trials were large sample size, double masking and randomization, high adherence to treatment, and ascertainment of clinical outcomes. Adherence was not reported in the CARET study, although β-carotene treatment was shown to raise the median serum β-carotene levels to 12 times the baseline levels.⁹³ The success of blinding the study was not reported in the NCSP,⁸⁴ WHS,⁹⁶ and SCP.⁸⁵ The study population was incompletely described in the SCP.⁸⁵ Most of these studies did not report on participants' prior use of supplements (Table 3).

Results

Cancer. In the ATBC study, compared to those who did not receive β-carotene, participants receiving β-carotene had a higher lung cancer incidence and lung cancer mortality (RR 1.18, 95% CI 1.03–1.36; RR 1.08, 95% CI 1.01–1.16, respectively),^{97, 98} but no increased risk for gastric cancer,⁹⁰ pancreatic cancer,⁹⁹ colorectal adenomas,¹⁰⁰ prostate cancer ¹⁰¹ or colorectal cancer ¹⁰² In the 6-year post-trial follow up, the relative risk of lung cancer was 1.06 (95% CI 0.94–1.20) for β-carotene recipients versus non-recipients. The supplementation had a late effect on colorectal cancer (RR 1.88; 95% CI 1.28–2.76) 4 years after the end of supplementation, but no late effect on other cancer outcomes.¹⁰³ (Appendix F, Evidence Tables 3b–3e, Table 5).

Table 5

Summary of randomized controlled trials on beta-carotene and chronic disease.

In the PHS, β-carotene supplementation increased the risk of thyroid cancer (RR 9.5, 95% CI 2.2–40.7), and bladder cancer (RR 1.5, 95% CI 1.0–2.2), but had no effect on other malignant neoplasms^{95, 104} or non-melanoma skin cancer⁸⁶ (Appendix F, Evidence Table 3d, Table 5).

In the CARET study, the combination of β-carotene and vitamin A supplementation increased the incidence of lung cancer (RR 1.28, 95% CI 1.04–1.57)^{93, 105} and the effects persisted 6 years after the trial terminated, especially among women.⁹⁴ β-carotene supplementation had no effects on cancers such as leukemia, mesothelioma, bladder cancer, breast cancer, prostate cancer, colorectal cancer, head and neck cancer, or lymphoma¹⁰⁵ (Appendix F, Evidence Table 3d, Table 5).

In the SCP, β-carotene supplementation had no effect on cancer deaths⁸⁵ (Appendix F, Evidence Table 3d, Table 5).

In the WHS, β-carotene supplementation had no impact on the incidence of cancer⁹⁶ (Appendix F, Evidence Table 3d, Table 5).

In the NSCP trial, β-carotene supplementation had no impact upon the incidence of basal cell carcinoma or squamous cell carcinoma after 4 years of follow up⁸⁴ (Appendix F, Evidence Table 3d, Table 5).

Cardiovascular disease. The ATBC study participants who received β-carotene had a non-significant higher incidence of angina and stroke mortality during the trial,^{106, 107} and had higher mortality for a wide spectrum of cardiovascular disease during the post-trial follow up¹⁰³ (Appendix F, Evidence Table 3d, Table 5).

Participants receiving β-carotene and vitamin A in CARET had a non-significant increased risk of cardiovascular death after a mean follow up of 4 years (RR 1.26, 95% CI 0.99–1.61),⁹³ but the risk was lower (RR 1.02) 6 years after supplementation was terminated.⁹⁴

Participants in the WHS study had a non-significant higher risk for stroke (RR 1.42, 95% CI 0.96–2.10), but lower risk for myocardial infarction (RR 0.84, 95% CI 0.56–1.27)⁹⁶ (Appendix F, Evidence Table 3d, Table 5).

In the PHS, β-carotene supplementation had no effects on incidence of type 2 diabetes mellitus,¹⁰⁸ incidence of myocardial infarction, stroke and all important cardiovascular events, or cardiovascular mortality⁹⁵ (Appendix F, Evidence Table 3d, Table 5).

Cataract and age-related macular degeneration. In the ATBC trial, β-carotene supplementation had no effect on age-related cataract or age-related maculopathy^{109, 110} (Appendix F, Evidence Table 3d, Table 5).

Total mortality. β-carotene supplementation was associated with an 8 percent, 7 percent, and 5 percent increased risk of total mortality in the ATBC, WHS and SCP studies, respectively.^{85, 96, 97} Only the increase in the ATBC trial reached statistical significance (p=0.02). In the post-trial follow up on total mortality (8 years of follow up) of the ATBC trial, the relative risk of total mortality in the groups receiving β-carotene compared to the corresponding placebo groups was 1.07 (95% CI 1.02–1.12)¹⁰³ (Appendix F, Evidence Table 3e, Table 5).

Summary

In summary, β-carotene was associated with increased risk of lung cancer incidence and mortality in persons who were heavy smokers or who were regularly exposed to asbestos. β-carotene supplementation did not reduce risk of other chronic disease outcomes, including cardiovascular disease, diabetes mellitus, cataract, and maculopathy. Taking into consideration the quantity, quality, and consistency of evidence, we concluded that the overall strength of evidence regarding the effects of β-carotene on the incidence of cancer and cardiovascular disease was “moderate” and on the prevalence of cataract or age-related maculopathy was “very low” (Table 6).

Table 6

Grading of the quality of evidence of the efficacy of single nutrients in the prevention of chronic disease.

Vitamin A

Introduction

The following section summarizes the evidence from randomized controlled trials on the efficacy of vitamin A supplement use in the prevention of chronic disease.

Results of the literature search

Our literature search identified no data on the efficacy of vitamin A alone in the prevention of chronic disease. We identified 9 eligible articles that addressed the efficacy of pre-formed vitamin A, combined with zinc or β-carotene, in preventing chronic disease. Three articles were from the Linxian trial in China^64–66 in which retinyl palmitate and zinc was combined as one type of supplementation, and 5 articles were from the CARET in the United States^{92–94, 103, 105} in which retinyl palmitate and β-carotene were combined as one type of supplementation.

Design of randomized controlled trials

The designs of the Linxian and CARET trials were described in a previous section of the Results chapter, Design of Randomized Controlled Trials, for Key Question 1 and Design of Randomized Controlled Trials for Key Question 3, β-carotene, respectively (Appendix F, Evidence Tables 3a–3c).

Results

In the Linxian trial, combined vitamin A and zinc had no impact on reducing deaths from stroke,⁶⁶ mortality,⁶⁴ or esophageal or gastric dysplasia or cancer.¹¹¹

CARET used a combination of β-carotene and retinyl palmitate which increased the incidence of lung cancer (RR 1.28, 95% CI 1.04–1.57), mortality related to lung cancer (RR 1.46, 95% CI 1.07–2.00) and cardiovascular disease (RR 1.26, 95% CI 0.99–1.61).⁹³ The risk for cardiovascular disease was lower (RR 1.02) 6 years after supplementation was terminated.⁹⁴ Total mortality was higher in the group receiving retinyl palmitate and β-carotene at the end of the trial (RR 1.17, 95% CI 1.03–1.33) ⁹³, but leveled off in a post-trial follow up for 6 years (RR 1.08, 95% CI 0.99–1.17)⁹⁴ (Appendix F, Evidence Tables 3d–3e).

Summary

Available evidence from two studies in selected populations (nutritionally inadequate or exposure to asbestos and/or cigarette smoke) suggests no benefit of combinations of vitamin A and zinc or vitamin A and β-carotene for cancer or cardiovascular disease prevention. Because no trial has been conducted to assess the efficacy of vitamin A alone in the prevention of the chronic diseases listed in the Key Question 1, we drew no conclusion for vitamin A by itself.

Vitamin E

Introduction

Vitamin E is the second most commonly used dietary supplement in the United States.¹ The following section reviews the evidence on the efficacy of vitamin E supplementation in the prevention of chronic disease.

Results of literature search

Our literature search identified 16 articles (including articles containing post-trial data) that provided evidence on the efficacy of vitamin E supplements in the prevention of chronic disease. These articles were generated from 4 randomized controlled trials, the ATBC trial, the WHS, the Primary Prevention Project (PPP), and the Vitamin E, Cataract, and Age-Related Maculopathy Trial (VECAT). The predominant source of evidence (from 12 articles, including articles containing post-trial data) on this topic stems from the ATBC trial.

Design of randomized controlled trials

The designs of the ATBC trial and the WHS were described in a previous section of the Results chapter, Design of Randomized Controlled Trials, on β-carotene (Appendix F, Evidence Tables 3f–3h).

The PPP was a randomized controlled, open-labeled, 2 by 2 factorial trial designed to investigate the efficacy of vitamin E (synthetic, 300 IU per day) and aspirin (100 mg per day) for cardiovascular disease prevention.¹¹² Participants were 4,495 men and women age 50 years or older with at least one of the major well-accepted risk factors for cardiovascular disease. Long-term vitamin E users were ineligible. At the end of the trial, the percent of participants lost to follow up was 13.6 percent in the vitamin E group. (Appendix F, Evidence Tables 3f–3h).

The VECAT was designed to evaluate whether daily vitamin E supplements reduced the risk of age-related cataracts in 1,193 Australians who were 55 to 80 years old upon entry into the study and who had early or no cataract. Trial participants were randomized to receive 500 IU per day of natural vitamin E or placebo for 4 years. Approximately 27 percent of the trial participants had prior supplement use. The percent of participants lost to follow up was 25 percent, and among those who were retained in the trial, 12 percent ceased taking study supplements¹¹³ (Appendix F, Evidence Tables 3f–3h).

Similarity and heterogeneity among trials

The participants in these trials had distinct characteristics, being female health professionals in the United States (WHS), male smokers in Finland (ATBC), or Italians who might have followed a Mediterranean diet (PPP). A total of 74,697 individuals were included in these trials with 87 percent being ATBC or WHS participants. Accordingly, approximately 27 percent of these trial participants were assigned to also take aspirin and 20 percent were assigned to also take β-carotene supplements. Vitamin E supplements used in these studies included synthetic form, natural source, and natural vitamin E at doses ranging from 50 IU per day in synthetic form to 600 IU per day of natural source (Appendix F, Evidence Tables 3f–3h).

Study quality

Inclusion/exclusion criteria were clearly defined in most trials. The quality of these trials was good with respect to randomization, double masking, ascertainment of trial endpoints, adherence, and use of an intention-to treat approach in statistical analyses (see Table 3, Assessment of Quality of Studies). There was a lack of descriptions as to whether concealment of allocation sequence was performed and whether there was an unintended crossover. The WHS and PPP trials collected data on lifestyle factors and medication use. None of the trials reported success of blinding and the extent of unintended crossover. Most trials provided no information on numbers and reasons for withdrawals and percent lost to follow up.

Results.

Cancer. In the ATBC trial, synthetic α-tocopherol of 50 IU per day had no benefit on the incidence of lung cancer and gastric neoplasm,^{90, 98} lung cancer mortality, or pancreatic cancer mortality,^{97, 99} but increased colorectal adenoma incidence (RR 1.66, 95% CI 1.19–2.32)¹⁰⁰. Questions have been raised whether the finding on colorectal adenoma was due to increased rectal bleeding by α-tocopherol supplementation, leading to the increased diagnosis of polyps. In contrast to these findings, men who received α-tocopherol supplements had a non-significant protective effect on colorectal cancer development (RR 0.78, 95% CI 0.55–1.09)¹⁰² and had a 32 percent and 41 percent reduction in the incidence of, and the mortality from prostate cancer respectively.¹⁰¹ The reduction was evident for clinical prostate cancer but not for latent cancer. In the post-trial follow up, the protective effect of α-tocopherol against prostate cancer was attenuated (RR 0.88, 95% CI 0.76–1.03). The moderate protective effects of α-tocopherol on colorectal cancer during the trial was no longer evident in the 6-year post-trial follow up, and α-tocopherol had no late effects on other cancers.¹⁰³

In the WHS study, vitamin E of 600 IU on alternate days did not affect the risk of developing total invasive cancer, breast cancer, lung cancer, and colon cancer, or the risk of cancer death⁸⁷ (Appendix F, Evidence Table 3i, Table 7).

Table 7

Summary of randomized controlled trials on vitamin E and chronic disease.

Cardiovascular disease. In the ATBC trial, all-rac-α-tocopheryl acetate of 50 IU per day had a borderline effect in reducing the incidence of angina (RR 0.91 comparing alpha-tocopherol with or without beta-carotene to no alpha-tocopherol with or without beta-carotene; RR 0.97 comparing alpha-tocopherol alone to placebo), decreased the risk of cerebral infarction (RR 0.86, 95% CI 0.75–0.99), and increased the risk of subarachnoid hemorrhage (RR 1.50, 95% CI 0.97–2.32) and fatal subarachnoid hemorrhage (RR 1.81, 95% CI 0.49–1.32).¹⁰⁶ A similar increased risk in hemorrhagic stroke persisted during the post-trial follow up.¹⁰³

In the PPP, ¹⁰⁷ the evidence was inconclusive due to small numbers of events and premature stopping of the trial; there was a non-significant increased risk for main cardiovascular endpoints (cardiovascular death, non-fatal myocardial infarction and non-fatal stroke) (RR 1.07, 95% CI 0.74–1.56), but a lower risk for total cardiovascular events or diseases (RR 0.94, 95% CI 0.77–1.16)¹¹² (Appendix F, Evidence Table 3i, Table 7).

In the WHS, use of vitamin E, 600 IU every other day had no effects on fatal and non-fatal myocardial infarction and fatal and non-fatal stroke, but reduced total cardiovascular death (RR 0.76, 95% CI 0.59–0.98).^{87, 96} There was no effect of vitamin E supplementation on hemorrhagic stroke (RR 0.92, 95% CI 0.61–1.38)⁸⁷ (Appendix F, Evidence Table 3i, Table 7).

Serum lipid levels. Shekelle etal. conducted a systematic review of the effects of vitamin E on the prevention and treatment of cardiovascular disease.¹¹⁴ The review, published in April 2004, was part of a larger evidence report on the effects of vitamin C, vitamin E, and coenzyme Q10 on cardiovascular outcomes.¹¹⁵ The review included an examination of the effects of vitamin E on lipid levels. The search strategy was comprehensive and retrieved English and non-English studies from multiple electronic databases. Additional studies were obtained by hand-searching reference lists from key articles and by consulting experts in the field. Multiple synonyms for vitamin E and for clinical trials were used in the initial search, but only randomized trials in humans using clinical or important surrogate outcomes were included in the report. Two independent evaluators using a standardized form extracted study data, and quality was assessed using the Jadad scale. Both primary and secondary prevention trials were evaluated. Meta-analyses were performed whenever groups of studies were judged to be sufficiently similar (Appendix F, Evidence Table 3i, Table 7).

The Shekelle review included 84 eligible trials of the effect of vitamin E on cardiovascular outcomes. However, only four of the trials were primary prevention studies, and these were deemed to be too heterogeneous (with respect to the type of intervention) to permit meta-analysis to be performed. The individual results of these 4 studies (ATBC,¹¹⁶ PPP,¹¹⁷ SCP,¹¹⁸ and Linxian¹¹⁹) were presented by the authors in narrative form. With respect to lipid lowering, the authors stated that “the 2 large primary prevention trials (ATBC and Linxian) reported clinically insignificant (but statistically significant) changes in (lipid) outcomes,” and that “there is no evidence that vitamin E alone or in combination has a clinically or statistically significant favorable or unfavorable effect on lipids.” In their meta-analyses of all primary and secondary prevention trials on the lipid effects of vitamin E compared to placebo, they found effect sizes that were not significant for total cholesterol (effect size -0.07, 95% CI -0.31 to 0.08), low-density cholesterol (effect size -0.07, 95% CI -0.24 to 0.10), or high-density lipoprotein (effect size 0.01, 95% CI -0.21 to 0.22).¹¹⁶ A negative effect size would indicate a favorable effect of treatment (Appendix F, Evidence Table 3i, Table 7).

Cataract and age-related macular degeneration. The evidence concerning vitamin E supplements and cataract is compatible with no effect. In the VECAT trial,¹¹³ the relative risk of cataract in the vitamin E group versus the placebo group was 1.0 for any cataract (95% CI 0.8–1.4). The relative risk for specific types of cataract were 0.9 for cortical cataract (95% CI 0.5–1.6), 1.1 for nuclear cataract (95% CI 0.8–1.5), and 0.5 for posterior subcapsular cataract (95% CI 0.2–1.1)¹¹³ (Appendix F, Evidence Table 3i, Table 7).

In the ATBC trial, lens opacity was measured at the end of the trial in a random sample of 1,828 participants.¹⁰⁹ The results showed that participants randomized to the α-tocopherol group were not different from the non-α-tocopherol group with respect to the risk of having nuclear cataract (OR 0.8, 95% CI 0.4–1.4), cortical cataract (OR 0.9, 95% CI 0.6–1.4), or posterior subcapsular cataract (OR 0.9, 95% CI 0.4–1.8)¹⁰⁹(Appendix F, Evidence Table 3i, Table 7).

The same approach was used in the ATBC trial to assess the association between α-tocopherol and the end-of-trial prevalence of age-related maculopathy.¹¹⁰ The prevalence of age-related maculopathy was higher among those assigned to receive α-tocopherol supplements than in the placebo group (32% versus 25%), showing no evidence of a beneficial affect of α-tocopherol¹¹⁰(Appendix F, Evidence Table 3i, Table 7).

Total mortality. The relative risk for total mortality in the vitamin E supplement users compared to non-users was 1.04 (95% CI 0.93–1.16), 1.02 (95% CI 0.95–1.09), and 1.07 (95% 0.61–1.90) in the WHS, the ATBC, and the PPP, respectively. In the post-intervention follow up on mortality (8 years of follow up) of the ATBC trial, the relative risk of total mortality in α-tocopherol users compared to non-users was 1.01 (95% CI 0.96–1.05).¹⁰³ Investigators in the WHS reported that “the main causes of death, apart from cardiovascular and cancer deaths, were pulmonary diseases (32 vitamin E, 22 placebo) and violent deaths, excluding suicide (9 vs. 6). None of these causes of deaths was significantly related to vitamin E.” ⁸⁷ The relative risk of cardiovascular death and cancer death in the WHS was 0.76 (95% CI 0.59–9.98) and 1.12 (95% CI 0.95–1.32), respectively.⁸⁷ The VECAT documented 31 deaths (20 in vitamin E; 11 in placebo), and the authors reported “no consistent or unusual patterns were identified among the specific causes of death recorded”¹¹³ (Appendix F, Evidence Table 3j, Table 7).

Summary

Vitamin E supplements have been studied for efficacy in the primary prevention of cancer, cardiovascular disease, cataract, and age-related macular degeneration. There was a lack of effects of vitamin E supplement use in the prevention of these diseases, except for a 32 percent reduction in prostate cancer incidence, a 41 percent reduction in the prostate cancer mortality, and a 22 percent reduction in colorectal cancer in the ATBC trial. The findings on hemorrhagic stroke were conflicting between the ATBC trial and WHS trial in that the former found a higher risk with use of low-dose α-tocopherol supplements but the latter found a lower risk with use of a high dose. Taking into consideration the quantity, quality, and consistency of evidence on the efficacy of vitamin E in preventing chronic disease, we concluded that the overall strength of evidence is “very low” for cancer, “low” for the relationship to cardiovascular disease, and “moderate” for cataract (Table 6).

Folic acid and B vitamins

Introduction

The co-prevalence of dementia and low circulating levels of micronutrients among the elderly has led to the research interest in vitamin supplementation as a means to prevent dementia. In various observational studies, low circulating levels of folate and vitamin B6 have been associated with poor cognitive function, dementia, and Alzheimer's disease^120–124 and hyperhomocysteinemia.^{125, 126} The essential role of folate and the B vitamins in homocysteine metabolism has been used to explain the possible role of these vitamins in dementia.

Results of literature search

Our search revealed two systematic reviews on single or paired vitamin supplementation with B vitamin(s) or folic acid for primary or secondary prevention of dementia and cognitive decline, and 4 articles from 1 trial that addressed vitamin B2 and niacin in the prevention of chronic disease. The systematic reviews were from the Cochrane Collaboration. The review on folic acid with or without vitamin B12 was comprised of 4 randomized controlled trials. The review of vitamin B6 was comprised of 2 randomized controlled trials. The trial on vitamin B2 and niacin was the Linxian trial. No studies were found to assess the efficacy of single or paired B vitamins or folic acid supplementation for prevention of other chronic diseases.

Design of Systematic Reviews

Malouf et al. systematically reviewed the literature to “assess the efficacy of vitamin B6 supplementation in reducing the risk of developing cognitive impairment by older healthy people, or improving cognitive functioning of people with cognitive decline and dementia,”¹²⁷ and to “examine the effects of folic acid supplementation, with or without vitamin B12, on elderly healthy and demented people in preventing cognitive impairment or retarding its progress.”¹²⁸ The search strategy, data collection and analysis methods were similar in both reviews. Trials were identified from a broad database by a predefined search strategy by the Dementia and Cognitive Improvement Group. Outcomes were measured as changes in continuous rating scales from baseline where available. When the same rating scales were used across trials, the weighted mean difference was presented for pooled trials. A standardized mean difference was reported for different rating scales. Weighted estimates for odds ratio were used for binary outcomes. When duration varied greatly and the range was considered too great to combine, a separate meta-analysis was conducted for smaller time periods. If there was evidence of heterogeneity of treatment effect between trials, either only homogeneous results were pooled or a random effect model was used. There was no pooled outcome measure presented due to heterogeneity of study participants and supplements.

Study quality

Design and quality of the meta-analyses on folic acid and vitamin B6 were similar. Strengths of these systematic reviews include: clarity of review question, description and completeness of search strategy, and reproducibility of review. Limitations were due primarily to heterogeneity among studies reviewed. The authors presented standardized outcomes of cognition when possible. No attempt was made to summarize outcome measures because of the great variation in trials included.

The review on vitamin B6 supplementation ¹²⁷ reviewed 2 randomized controlled trials for primary prevention. ^{128, 129} The authors attempted to minimize heterogeneity of study subjects by extracting data on older subjects. Follow up time varied from 5 to 12 weeks. Dosages of B6 supplementation varied from 20 to 75 mg per day. Among the different trials, there were wide disparities in dosages of folic acid (750 mcg to 15 mg) and vitamin B6 (20 to 75 mg).

The review on folic acid¹³⁰ reviewed 4 randomized controlled trials for primary and secondary prevention.^{128, 131–133} The authors attempted to minimize heterogeneity of study subjects by extracting data on older subjects at the expense of decreasing sample size. Despite this, there was considerable heterogeneity in study population. One study for primary prevention involved only women.¹²⁸ The remaining 3 studies^131–133 were secondary prevention trials. Dosage of folic acid varied widely from 750 mcg to 15 mg per day. Two studies combined vitamin B12 with the folic acid supplementation and these results were combined together with those receiving folic acid alone.

Results.

Cognitive decline. Although the meta-analysis by Malouf found improvement in biochemical indicators of vitamin B6, no measurable improvement in cognition was found after short-term supplementation with vitamin B6. Although folic acid with vitamin B12 was effective in reducing serum homocysteine levels, the authors concluded that these limited studies did not support folic acid supplementation for prevention of cognitive decline.

Summary

There is limited evidence to suggest no benefit of vitamin B6, vitamin B12, or folic acid supplementation for primary prevention of cognitive decline. Taking into consideration the quantity, quality, and consistency of evidence on the efficacy of folic acid, vitamin B6 and vitamin B12 in preventing chronic disease, we concluded that the overall strength of evidence is “low” for folic acid with or without vitamin B12 and “moderate” for vitamin B6 (Table 6).

Vitamin B2 and niacin

Introduction

The following section summarizes the evidence on the efficacy of vitamin B2 and niacin supplement use in the prevention of chronic disease.

Results of the literature search

Our literature search identified 4 eligible articles from the Linxian General Population Trial that addressed the efficacy of vitamin B2 (3mg per day) and niacin (vitamin B3, 40 mg per day) in preventing cancer, cardiovascular disease or cataract. ^{64–66, 111} Data on other chronic diseases were lacking.

Design of randomized controlled trial

The design of the Linxian trial was described in a previous section of the Results chapter, Design of Randomized Controlled Trials, for Key Question 1 (Appendix F, Evidence Tables 3k–3o).

Results.

Cancer, cardiovascular disease and total mortality. In the Linxian trial, combined vitamin B2 and niacin had no impact on reducing deaths from stroke,⁶⁶ mortality,⁶⁴ or esophageal or gastric dysplasia or cancer.¹¹¹ (Appendix F, Evidence Tables 3k–3o)

Cataract. A lower prevalence of nuclear cataracts was observed in those who received riboflavin and niacin, and there was no difference between randomized groups in cortical cataracts.⁶⁵ However, a 2.64-fold increased prevalence in posterior subcapsular cataract was documented for the groups receiving riboflavin 3 mg and niacin 40 mg compared to the groups not receiving riboflavin and niacin⁶⁵ (Appendix F, Evidence Tables 3k–3o).

Summary

Data on the efficacy of vitamin B2 and niacin supplement use in the primary prevention of chronic disease are sparse and the only study was conducted in a nutritionally deprived Chinese population found no benefit of combined vitamin B2 and niacin for primary prevention of cancer, cardiovascular mortality, or cataracts.

Selenium

Introduction

Selenium functions as an antioxidant since it is essential to the antioxidant enzyme glutathione peroxidase.¹²⁹ Because selenium is involved in the biosynthesis of testosterone, another proposed mechanism involves its role in the endocrine and immune system. ^{130, 131} Selenium has also been theorized to function on the molecular level by changing carcinogen metabolism, inhibiting protein synthesis or specific enzymes, and stimulating apoptosis.¹³²The following section summarizes the evidence on the efficacy of selenium supplement use in chronic disease prevention.

Results of literature search

Our literature search identified 6 articles that provided evidence on the efficacy of selenium supplements in the prevention of cancer, cardiovascular disease. These publications were generated from 2 different trials, the Nutritional Prevention of Cancer (NPC) trial and another study. We included the NPC trial of patients with a history of non-melanoma skin cancer because the study reported on the risk of cancer other than non-melanoma skin cancer, and non-melanoma skin cancer is not a precursor of other cancers.

Design of randomized controlled trials

The NPC trial was a double-blind, placebo-controlled multi-center cancer prevention trial in 1,312 men and women to test the efficacy of selenium supplementation (200 mcg supplied as 500 mg high-selenium yeast tablets) in reducing chronic disease, specifically cancer.^133–135 Trial participants had a history of either 2 or more basal cell carcinomas (BCC) or one squamous cell carcinoma (SCC) of the skin within the prior year. Prior supplement users were eligible for enrollment. The primary outcome of interest was occurrence of a new non-melanotic skin cancer. Secondary endpoints included incidence of lung, colorectal, and prostate cancers, total mortality and cancer mortality. The total blinded treatment period was from September 1983 until January 1996. Interim analysis was published in 1996 on data from the full cohort of 1312 participants through December. 1993^{133, 134} Analyses at the end of the full, blinded treatment period in 1996 were published on total cancer outcomes, ¹³⁶ prostate cancer, ¹³⁷ and lung cancer.¹³⁵ Later analyses excluded 62 patients who had baseline blood tests more than 4 days after randomization.^135–137 Interim analysis for prostate cancer was performed on 974 male participants, accounting for a 2-year lag effect.¹³⁴ Re-analysis of prostate cancer data at the end of full, blinded treatment was done on 927 participants without a history of prostate cancer before randomization, using those individuals with a valid baseline blood draw less than 4 days after randomization. ¹³⁷ By the end of the blinded study in 1996, 35.9 percent of participants were on supplementation, 16.6 percent were off supplementation, but continuing follow up, 22.1 percent were censored for dermatological endpoints but not other endpoints, and 24.8 percent had died. We did not include one study with melanotic skin cancer recurrences in the NPC trial because it addressed secondary prevention. Full text of the articles on cardiovascular disease and colorectal cancer were published after the cutoff date of our review¹³⁸ Another study by Yu et al. was conducted in Qidong County, China and published in 1991.¹³⁹ (Appendix F, Evidence Tables 3k–3o).

Similarity and heterogeneity among trials

Participants in the NPC trial were recruited from dermatology clinics and had non-melanotic skin cancer without recent treatment for internal malignancy. Participants in the study by Yu etal. were selected to be at high risk for liver cancer because of a family history of cancer in addition to living in an area of China that has high rates of liver cancer. Both studies used 200 mcg per day of selenium as a yeast tablet.

Study quality

In the NPC Study, the study population, inclusion and exclusion criteria, flow of patients, outcome reporting and statistical analyses were well described. Well designed aspects of the study included: random assignment of patients, placebo control, confirmation of outcomes, efforts at blinding, assessment of adherence, appropriate handling of losses to follow up, reporting of statistical analyses, and intention-to-treat analysis. However, there was inadequate information reported regarding excluded patients, prior supplement use, prior and concurrent medication use, success of blinding, independent ascertainment of outcomes, unintended cross-over rates, description of supplements, and statistical power.¹³³ The study was initially designed to look at incidence of non-melanoma skin cancer, and other cancer endpoints were designated secondary outcomes 7 years after commencement of the trial.

The study by Yu et al. had inadequate data reporting on almost all aspects of the study with the exception of a fair description of supplements and assessment of adherence to supplements by biomarkers (Table 3).

Results.

Cancer. Initial interim analysis of the NPC trial through 1993 found that the selenium group had a significantly lower total cancer mortality (RR 0.5, 95% CI 0.31–0.8), total cancer incidence (RR 0.63, 95% CI 0.47–0.85), and significantly lower incidence of lung, colorectal, and prostate cancers (RR 0.56, 95% CI 0.31–1.01; RR 0.39, 95% CI 0.17–0.90; RR 0.35, 95% CI 0.18–0.65, respectively).¹³³ Cancer endpoints from the full trial period through 1996 were analyzed and had a mean follow up of 7.9 years. Selenium continued to reduce the risk of all cancers (HR 0.75, 95% CI 0.58–0.97) and prostate cancer (HR 0.51, 95% CI 0.29–0.87), lung cancer (HR 0.70, 95% CI 0.40–1.21) and colorectal cancer (HR 0.46, 95% CI 0.21–1.02), although the findings on lung cancer and colorectal cancer were not statistically significant.^{135, 137}

An interim reanalysis of 843 male patients with prostate specific antigen levels less than 4 ng/ml, taking into account a 2-year treatment lag, found that the selenium group had a significant reduction in prostate cancer (RR 0.37, p-value 0.002).¹³⁴ Subgroup analyses showed that the effect of selenium on prostate cancer was greatest in those with a baseline prostate specific antigen level less than 4 ng/ml (RR 0.35, 95% CI 0.13–0.87)¹³⁷ (Appendix F, Evidence Tables 3k–3o).

In the 2-year intervention trial with selenized yeast by Yu et al., the incidence of primary liver cancer was significantly less (p<0.05) in selenium supplemented subjects (10 of 1444; 0.69%) compared to control subjects (13 of 1030; 1.26%)¹³⁹ (Appendix F, Evidence Tables 3k–3o) (Appendix F, Evidence Tables 3k–3o).

Cardiovascular disease. Only the NPC study reported cardiovascular outcomes in the context of selenium supplementation, and there was no effect on cardiovascular disease (HR 1.03, 95% CI 0.78–1.37), stroke (HR 1.02, 95% CI 0.63–1.65), or cardiovascular mortality (HR 1.22, 95% CI 0.76–1.95) for primary prevention in those without prior cardiovascular disease.^{133, 140} (Appendix F, Evidence Tables 3k–3o).

Total mortality. Total mortality in the NPC study was reduced by 21 percent in the group receiving selenium (HR 0.79, 95% CI 0.61–1.02) as compared to placebo¹³³ (Appendix F, Evidence Tables 3o).

Summary

Evidence on the role of selenium in cancer prevention is limited, but suggests some benefit in prevention of total and prostate cancer, with the greatest benefit in men with a normal baseline prostate specific antigen level. Selenium did not significantly reduce the risk of lung or colorectal cancer. The only well-designed randomized controlled study supporting selenium supplementation for cancer prevention was done in a population with non-melanotic skin cancer. Taking into consideration the quantity, quality, and consistency of evidence on the efficacy of selenium in preventing chronic disease, we concluded that the overall strength of evidence is “moderate” (Table 6).

Calcium and vitamin D

Introduction

Supplementation with calcium, vitamin D, or both has been recommended for primary prevention of osteoporosis. Physiologically, calcium supplementation corrects for suboptimal intake or decreased intestinal absorption of calcium. Left uncorrected, secondary hyperparathyroidism develops, leading to accelerated bone resorption and ultimately to increased risk for fractures. Supplemental vitamin D optimizes intestinal calcium absorption, and it also improves neuromuscular function and reduces the recurrences of fractures. ¹⁵¹

Improvement in bone mineral density (BMD) is a marker for stronger bones and is predictive of fracture reduction.¹⁵⁰ However, fracture is the major clinical outcome of osteoporosis.

Due to the substantial amount of efficacy data on calcium/vitamin D and osteoporosis, we reviewed systematic review articles supplemented with data from recent randomized controlled trials. We also used data from randomized controlled trials meeting our inclusion criteria, that were not included in previous systematic reviews.

Results of literature search

Our search for evidence that supplemental calcium and/or vitamin D prevents osteoporosis/fractures/falls revealed 7 articles from 6 recent systematic reviews, authored by Shea et al.,^{47, 50} Mackerras and Lumley,⁵² Papadimitropoulos et al.,⁴⁹ Avenell et al.,¹⁴³ and Bischoff-Ferrari et al.^{144, 145} Two articles on osteoporosis and colorectal cancer from the Women's Health Initiative study (WHI)¹⁴⁶ were released as we prepared this report. We also identified three small relevant randomized controlled trials ^147–149 that were not included in previous systematic reviews. Using our search strategies, we identified no additional randomized controlled trials for the efficacy of calcium with or without vitamin D supplement use in the primary prevention of other chronic diseases. In 2005, AHRQ awarded a contract to the University of Ottawa's EPC to conduct a systematic review of the efficacy of vitamin D on bone density and fracture risk, but that review was not available in time for inclusion in the evidence report.

All of this literature met our criteria for calcium and vitamin D formulations and doses. For calcium, the doses were less than 2.5 grams per day, the adult UL recommended by the Food and Nutrition Board (Appendix F, Evidence Tables 3p–3r). With regard to vitamin D, our interest was in over-the-counter supplements, but some systematic reviews included studies using formulations available only by prescription. Therefore, in summarizing these previous reviews, we extracted the relevant data reported for non-prescription vitamin D₃ (cholecalciferol) and vitamin D₂ (ergocalciferol) used in doses not exceeding the UL, 2000 IU per day (Appendix F, Evidence Tables 3p–3r).

Calcium

Design of systematic reviews. Three articles from 2 systematic reviews ^{47, 50, 52} examined the efficacy of calcium on BMD. Two of the reviews ^{47, 50} by Shea etal. presented identical data, so only the more recent article ⁴⁷ was used. Shea etal. analyzed randomized controlled trials published from 1978 to 1998 investigating skeletal effects of calcium supplementation in post-menopausal women. The randomized controlled trials addressed fractures in 5 trials (n = 638) and BMD in 15 trials (n=1826) of 1 to 4 years duration in women whose mean age ranged from 46 to 72 years. The Mackerras review ⁵² evaluated 8 randomized controlled trials from 1987 to 1995. However, Mackerras etal. had a different focus, concentrating on year-by-year BMD changes in a younger group of postmenopausal women (n = 1386, mean age 51 to 66 years) (Appendix F, Evidence Tables 3p and 3q).

Quality of reviews. The strengths of the Shea review were: attention to methodologic detail (e.g., contacting authors for details of randomization and blinding) and assessment of heterogeneity of BMD results across studies with various subgroup analyses (e.g., losses to follow up, time after menopause). A major limitation of the Shea review (and also that of Mackerras whose papers were all included in the Shea review) was that conclusions were compromised by problems inherent in the original studies. These problems included small sample size, large losses to follow up, and significant heterogeneity of study populations and interventions (e.g., the variable use of vitamin D in addition to calcium in treated and control subjects) (Appendix F, Evidence Tables 3p and 3q).

Strengths of the Mackerras review were: strict attention to precision and quality control issues involving bone density measurements that are often overlooked (e.g., excluded a study that changed densitometers mid-study); rigorous analysis of BMD data (e.g., did not pool measurements from different anatomical sites and measured BMD change year by year rather than averaging total change over the treatment period); and subgroup analyses to evaluate effects of calcium on bone density independent of other potential effectors, especially vitamin D and exercise. An important weakness of the Mackerras review, in addition to those mentioned above, was lack of discrimination against poorly randomized trials. Mackerras etal. did not contact investigators for missing information.

Design of randomized controlled trials. The WHI published 2 articles comparing the effect of calcium and vitamin D with placebo for primary prevention of fractures ¹⁴⁶ and colorectal cancer ¹⁵² in healthy postmenopausal women. A subgroup of 2431 women had BMD measured at annual visits 3, 6, and 9.

Storm et al.¹⁴⁹ compared the effect of calcium supplementation versus dietary calcium intake or placebo on seasonal (i.e. winter) bone loss in healthy, older postmenopausal women (n = 60, age greater than 65 years).

Meier et al.¹⁴⁷ compared the effect of calcium and vitamin D versus no treatment on seasonal bone loss in healthy, German, community dwellers (Appendix F, Evidence Tables 3s and 3t).

Similarity and heterogeneity among randomized controlled trials. The WHI studies ^{146, 152} selected participants from multiple United States cities. WHI studies allowed personal calcium and vitamin D supplementation up to 1000 mg and 600 IU daily respectively and thus had a baseline average daily intake of 1150 mg calcium and 365 IU vitamin D as assessed by a food frequency questionnaire. Meier ¹⁴⁷ did not allow prior or personal use of calcium or vitamin D supplements, but did not assess baseline calcium or vitamin D intake at baseline. Storm limited calcium intake to less than 800 mg per day as measured by food frequency questionnaire and thus had an average baseline calcium intake of 684 mg per day (Appendix F, Evidence Tables 3s and 3t).

Quality of randomized controlled trials. Strengths of the WHI study included: double blinded, placebo-controlled study, large sample size, rigorous quality control, reporting of baseline characteristics, clearly documented protocol, appropriate analytic methods, few losses to follow-up, long follow-up, and central adjudication of outcomes. Weaknesses of the study included: possible inadequate ascertainment of all outcomes, lack of adherence to treatment regimen, high baseline intake of calcium and vitamin D (though diet and supplement use), and inadequate power, all of which may bias this study to the null.

Strengths of the Storm study¹⁴⁹ were the administration of calcium alone without vitamin D, double blinding with placebo and treatment group, description of baseline calcium intake, description and number of withdrawals, quality control and outcome ascertainment and measurement of serum 25-hydroxyvitamin D levels during the study period. Weaknesses included small sample size, poor description of adherence assessment, and clarity and appropriateness of statistical analyses.

Strengths of the Meier study¹⁴⁷ included randomization with description of baseline equivalence of groups. Weaknesses of this study include: lack of placebo-control and double blinding, unclear description of inclusion and exclusion criteria, no description of adherence, high rate of withdrawals, short supplementation time, and heterogeneity of a relatively small sample size of participants.

Results

Calcium and bone density. Both Shea et al. and Mackerras et al. reported a small positive effect of calcium in preventing bone loss. Shea et al., who averaged BMD changes across the entire treatment period, concluded that BMD at four different sites was consistently 1.5 to 2.0 percent higher after two years of treatment. In a more rigorous analysis of BMD data, Mackerras et al. found that calcium's effects occurred mainly in the first year. They concluded that BMD losses actually occurred in both treated and control groups, but that losses were relatively greater in controls (0.5–2.8% from baseline at 10 different sites) than in treated groups (with corresponding losses of only 0.1–1.1%).

The WHI¹⁴⁶ found significant cumulative dose-responsive difference in total hip BMD between patients treated with 1000 mg calcium and 400 IU vitamin D and placebo-treated patients, but no significant difference in spine BMD.

Storm et al. found that supplemental calcium alone (1000 mg per day) prevented seasonal bone loss of the greater trochanter (associated with a 25% decrease in serum 25-hydroxyvitamin D levels) and significantly increased BMD of the femoral neck by 3 percent from baseline. In contrast, seasonal bone loss occurred in placebo-treated women who had a 3 percent loss of BMD in the greater trochanter and 0.3 percent loss in the femoral neck after 2 years. Dietary calcium treated subjects (average 1000 mg per day) had a 1.5 percent loss in greater trochanter BMD and 1.8 percent loss in the femoral neck BMD. (Appendix F, Evidence Table 3r).

Meier et al. found that 500 mg calcium and 500 IU vitamin D supplementation significantly increased lumbar (+0.8%, p=.04) and femoral BMD (+0.1%, p=.05) compared to the previous year without any supplementation, which was significantly different (p=.03 for lumbar spine, and p=.05 for femoral bone) from the control group, which had a decrease in lumbar and femoral BMD¹⁴⁷ (Appendix F, Evidence Table 3u).

Calcium and fractures. Shea et al. found in calcium-treated individuals, a trend toward reduction of vertebral fractures (RR 0.79, 95% CI 0.55–1.13). There was no significant effect of calcium on non-vertebral fractures. Fracture results were consistent across studies but the strength of the conclusion was limited by the small study populations and short follow up periods (Appendix F, Evidence Table 3r).

Intention-to-treat analysis of the WHI study ¹⁴⁶ found that calcium plus vitamin D supplementation did not significantly decrease the incidence of hip fracture (HR 0.88, 95% CI 0.72–1.08), clinical spine fracture (HR 0.90, 95% CI 0.74–1.10) or total fractures (HR 0.96, 95% CI 0.91–1.02) (Appendix F, Evidence Table 3u).

Calcium and colorectal cancer. A secondary outcome of the WHI trial was colorectal cancer. ¹⁵² Intention to treat analysis found that calcium plus vitamin D supplementation did not significantly decrease the incidence of invasive colon cancer (HR 1.08, 95% CI 0.86–1.34).

Summary. The studies showed a consistent small effect of calcium on prevention of BMD loss (approximately 2%) over a period of 2 or more years in postmenopausal women. The effects occurred mainly in the first year. Calcium supplementation prevented the seasonal bone loss associated with wintertime drops in vitamin D levels. Based on very limited data, Shea also raised the possibility that calcium may reduce vertebral, but not non-vertebral, fractures. (Appendix F, Evidence Tables 3r and 3u).

Vitamin D

Four articles from 3 systematic reviews, ^{49, 143–145} and one article from the WHI ¹⁴⁶ addressed the effect of vitamin D on fractures. Vitamin D effects on BMD were also assessed in one of these reviews, by Papadimitropoulos et al.,⁴⁹ as well as in the WHI study and 2 small randomized controlled trials.^{147, 148}

Design of Systematic reviews. The most comprehensive of the systematic reviews, the Avenell study, ¹⁴³ investigated the effects of vitamin D with or without calcium on fractures. Avenell et al. analyzed 38 randomized controlled trials; 12 of these (from 1983-2005) are pertinent to our review because they involved treatment with vitamin D_3, 400–800 IU per day, in about 35,000 men and women, age 65 or more. Included among the 12 trials is the large Porthouse primary prevention trial (n=3454 women)¹⁵⁰which employed a treatment regimen of vitamin D3 (800 IU/day) and calcium (1000 mg/day). The other 26 trials were not considered in this review because they used active hydroxylated metabolites of vitamin D (Appendix F, Evidence Tables 3p and 3q).

Of the two systematic reviews by Bischoff-Ferrari et al., the first ¹⁴⁵ explored anti-fracture efficacy of vitamin D with or without calcium in older persons (8 trials, n = 9820, mean age 75 to 85 years), whereas the second ¹⁴⁴ tested the effects of vitamin D₃ on fall prevention in a similar but smaller population (3 trials, n = 613).

The Papadimitropoulos review, limited to older postmenopausal women (mean age 72 to 84 years), evaluated 25 randomized controlled trials, 10 of which we included in our review because they employed vitamin D₃ in doses of 300–2000 IU/day. Of the 10 trials, 6 measured BMD changes (n = 956), and 4 evaluated fracture prevention (n = 5780) (Appendix F, Evidence Tables 3p and 3q).

Quality of Reviews. The strengths of the Avenell et al. review included its large size and comprehensive nature that allowed independent assessment of the anti-fracture effects of vitamin D and calcium, administered separately and in combination. Also important were assessments of methodological quality for each reviewed trial (revealing a range of quality from poor to satisfactory). A weakness of the Avenell et al. study was lack of information on dropouts from both treatment and control arms of some studies, possibly causing inaccurate estimates of outcome events by the intention to treat analysis. Similar to Avenell et al., a strength of the Papadimitropoulos review was the assessment of methodologic quality of each eligible study. In addition, a priori hypotheses concerning study design, population, intervention, and methodologic quality were developed in an attempt to identify reasons for differences in results across studies. Nevertheless, both the Avenell and Papadimitropoulos reviews suffered from marked heterogeneity across the included studies.

A strength of the Bischoff-Ferrari fracture prevention review ¹⁴⁵ was the consistency of treatment across studies with regard to vitamin D₃ doses, but a problem was that calcium was also used with some patients, possibly obscuring the effects of vitamin D alone. Other problems were the small number of trials analyzed and the absence of specific large relevant studies.^{150, 151} Similar issues of scope and variability in treatment regimens apply to Bischoff-Ferrari's review on fall prevention (Appendix F, Evidence Tables 3p and 3q). ¹⁴⁴

Design of randomized controlled trials. The WHI study assessed the efficacy of vitamin D₃ (400 IU/day) with calcium (1000 mg/day) for primary prevention of fractures in healthy postmenopausal women (n = 36,282, mean age 63 years).¹⁴⁶ BMD was followed at annual visits 3, 6, and 9 in a subgroup (n = 2431) (Appendix F, evidence tables 3s–3u).

Meier etal. (519) compared the effect of supplemental vitamin D₃ (500 IU/day) plus calcium (500 mg/day) with no treatment for prevention of wintertime BMD losses in health German men and women (n=55, age range 34–75 years).

Hunter etal. ¹⁴⁸ compared the effect of vitamin D (800 IU cholecalciferol/day) with placebo in a twin-control on change in BMD in healthy postmenopausal women living in the United Kingdom over 2 years (Appendix F, Evidence Tables 3s and 3t).

Similarity and heterogeneity among randomized controlled trials. Most studies included primarily postmenopausal women. Only Meier et al.¹⁴⁷ included men in addition to postmenopausal women. There was wide variation in baseline calcium and vitamin D intake and exposure. The WHI studies ^{152, 146} selected participants from multiple United States cities. Two studies were conducted in areas of northern latitude, Germany,¹⁴⁷ and the United Kingdom, with presumably less sunlight exposure. WHI studies allowed personal calcium and vitamin D supplementation up to 1000 mg and 600 IU daily respectively and thus had a baseline average daily intake of 1150 mg calcium and 365 IU vitamin D as assessed by a food frequency questionnaire. Hunter ¹⁴⁸ did not allow vitamin D or calcium supplementation, but participants had daily baseline calcium and vitamin D intakes of 1050 mg and 135 IU respectively. Meier et al. ¹⁴⁷ did not allow prior or personal use of calcium or vitamin D supplements, but did not assess baseline calcium or vitamin D intake at baseline. Treatment intervention regimens also varied among the different studies. Three studies used both calcium and vitamin D.^{146, 147, 152} One study used only vitamin D.¹⁴⁸ Vitamin D formulation was cholecalciferol^{146, 147, 152} with dosage ranging from 400 to 500 IU daily (Appendix F, Evidence Tables 3s and 3t).

Quality of randomized controlled trials. Strengths of the WHI study included: double blinded, placebo-controlled study, large sample size, rigorous quality control, reporting of baseline characteristics, clearly documented protocol, appropriate analytic methods, few losses to follow-up, long follow-up, and central adjudication of outcomes. Weaknesses of the study included: possible inadequate ascertainment of all outcomes, lack of adherence to treatment regimen, high baseline intake of calcium and vitamin D (though diet and supplement use), and inadequate power, all of which may bias this study to the null.

Strengths of the Meier study ¹⁴⁷ include randomization with description of baseline equivalence of groups. Weaknesses of this study include: lack of placebo-control and double blinding, unclear description of inclusion and exclusion criteria, no description of adherence, high rate of withdrawals, short supplementation time, and heterogeneity of a relatively small sample size of participants.

Hunter et al. ¹⁴⁸ described inclusion/exclusion criteria, flow of patients, and baseline equivalence of patients well. Other strengths of the study included double blinding, placebo-control, and assessment of adherence. Small size of the study, nearly 20 percent withdrawal rate, and high baseline intake of calcium and vitamin D may have limited the power of the study.

Results.

Bone mineral density. The Papadimitropoulos review also analyzed BMD effects of vitamin D. Treatment with vitamin D₃ between 300 and 2000 IU/day caused only marginal positive effects of the vitamin D and calcium intervention (increases by about 1% in the femoral neck in year 5 and in the lumbar spine in year 1).

The WHI¹⁴⁶, found a mean difference in total hip BMD of 0.59 percent (p<.001) at 3 years, 0.86 percent (p<.001) at 6 years, and 1.06 percent (p=.01) at 9 years between those treated with calcium and vitamin D and placebo group. There was no significant difference in BMD in the spine.

Hunter et al.¹⁴⁸ did not find any significant difference in spine or hip BMD between those treated with vitamin D alone and control.

Meier et al.¹⁴⁷ found calcium and vitamin D supplementation significantly increased lumbar BMD (+0.8%, p=.04) and femoral BMD (+0.1%, p=.05) compared to the previous year without any supplementation, which was significantly different (p=0.03 for lumbar spine, and p=0.05 for femoral bone) from the control group that had a decrease in lumbar and femoral BMD (Appendix F, Evidence Tables 3r and 3u)

Fractures. The review by Avenell et al. included data from primary prevention trials as well as secondary prevention trials. They reported that vitamin D alone did not prevent hip, vertebral, or any non-vertebral fractures, and that vitamin D (700–800 IU per day) plus calcium (1000 mg/day) reduced hip fractures (RR 0.81, 95% CI 0.68–0.96) and non-vertebral fractures (RR 0.87, 95% CI 0.78–0.97), but the combination was no more effective than calcium alone. There was no effect on vertebral fractures. Subgroup analysis indicated that the effects on hip and non-vetabral fracture were primarily reported from studies of the incidence of fracture (3 trials, n=4242; RR 0.75, 95% CI 0.62–0.91 for hip fracture; RR 0.83, 95% CI 0.72–0.95 for non-vertabral fracture), but not recurrence of fracture (4 trials, n=6134; RR 1.02, 95% CI 0.71–1.47 for hip fracture; RR 0.93, 95% CI 0.79–1.10 for non-vertebral fracture). Another subgroup analysis showed that the effects on hip fracture were primarily reported from studies in institutionalized groups (2 trials, n=3853, RR 0.75, 95% CI 0.62–0.92), but not in community-dwelling groups (5 trials, n=6523, RR 1.01, 95% CI 0.70–1.44), whereas the effects on non-vertabral fracture were similar between the two types of populations (RR 0.85 and 0.89, respectively). Baseline mean serum 25-OH vitamin D levels (measured in 9 of the studies) were generally quite low (≤ 15 ng/mL), but levels after vitamin D supplementation were not available.

In the Papadimitropoulos review, fracture results were similar to those of Avenell et al. Comparable treatment regimens of vitamin D₃ and calcium were associated with a non-significant trend in reduction of non-vertebral fractures (RR 0.78, 95% CI 0.55 – 1.09). (Appendix F, Evidence Table 3r).

In contrast, both of the reviews by Bischoff-Ferrari et al. showed definitive positive effects of vitamin D_3, with or without calcium, on fracture reduction and prevention of falls. Analysis of all the fracture results revealed heterogeneity that was resolved by pooling studies into separate high-dose (700–800 IU/day) and low dose (≤ 400 IU/day) subgroups. Studies using the high-dose regimen showed reductions in the pooled relative risk of hip fracture (RR 0.74, 95% CI 0.61–0.88) and of non-vertebral fracture of (RR 0.83, 95% CI 0.70–0.98). In a similar analysis of the effect of vitamin D on falls, supplementation with 800 IU/day with or without calcium had a pooled odds ratio for prevention of falls of 0.78 (95% CI 0.64–0.92) (Appendix F, Evidence Table 3r).

Incidence of fractures was the primary outcome of interest in the WHI study.¹⁴⁶, Intention-to-treat analysis found that calcium plus vitamin D supplementation did not significantly decrease the incidence of hip fracture (HR 0.88, 95% CI 0.72–1.08), clinical spine fracture (HR 0.90, 95% CI 0.74–1.10) and total fractures (HR 0.96, 95% CI 0.91–1.02) in the total trial participants. A subgroup analysis of women who took at least 80 percent of study medication showed a significant risk reduction in hip fracture (HR 0.71, 95% CI 0.52–0.97) (Appendix F, Evidence Table 3u).

Colorectal Cancer. A secondary outcome of the WHI trial was colorectal cancer.¹⁵² Intention to treat analysis found that calcium plus vitamin D supplementation did not significantly decrease the incidence of invasive colorectal cancer (HR 1.08, 95% CI 0.86–1.34) (Appendix F, Evidence Table 3u).

Summary. The majority of published literature on calcium and vitamin D are studies in postmenopausal women. Review of this evidence supports improvement in BMD with calcium with or without vitamin D supplementation for postmenopausal women. The evidence also indicates that calcium supplementation was associated with a non-significant trend toward decreasing the risk of vertebral fractures. The greatest benefit of calcium supplementation was found to occur in the first year of use. There is a paucity of data on the effect of vitamin D alone on BMD. Vitamin D combined with calcium prevented hip fracture and non-vertebral fracture with the greatest benefit seen in populations with a low baseline intake of calcium and/or vitamin D. A high dose of vitamin D (700–800 IU per day) with or without calcium prevented hip fracture, non-vertebral fracture and falls. Taking into consideration the quantity, quality, and consistency of evidence on the efficacy of vitamin D and calcium, we concluded that the overall strength of evidence is “low” for calcium to prevent loss in BMD, vitamin D to prevent loss in BMD, and for vitamin D to prevent fractures, “very low” for calcium to prevent fractures, and “high” for combined calcium and vitamin D to prevent BMD loss, hip fracture or non-vertebral fracture (Table 6).

Calcium and vitamin D

In a recent Cochrane review,⁴⁸ it was concluded that studies are too different (exposure time, doses, etc) to draw general conclusions regarding the safety of calcium supplements. A case report of nephropathy with calcified lesions in a patient consuming 1g/day of calcium lactate appears to be the result of the combined use of high dose ascorbic acid (6,000mg/day) plus laxatives that led to chronic hypokalemia.⁷⁹

The calcium-vitamin D arm of the WHI study ¹⁴⁶ administered 1g of calcium carbonate and 400 IU of vitamin D daily to 18,000 postmenopausal women for 7 years. The study reported an increased risk for kidney stones in the active group (HR 1.17). No other significant differences among the study groups were observed, including gastrointestinal symptoms.

Long-term consumption of 1g or more per day of calcium may increase risk of kidney stones. It is not clear whether this finding can be generalized to premenopausal women or to men.

Vitamin A

Randomized controlled trials

A number of studies compared retinol or β-carotene supplementation with placebo. The CARET trial in smokers ⁹³ administered 25,000 IU per day of vitamin A and 30 mg per day of beta-carotene for 5 years, and reported no adverse effects other than yellowing of the skin in 0.3 percent of people in the active group. In this study,¹⁵³ the active group also exhibited a modest but significant rise in serum triglycerides. This increase remained stable after the first year of follow up, i.e., it was non-progressive (Appendix F, Evidence Tables 4a–4d).

Another study in healthy adults aged 18–54 years⁷⁷ compared the effects of 15,000 IU per day of vitamin A (4500 RE) with a group receiving only 75 IU per day, for 5 years. The only relevant finding was an increase in serum triglycerides in the high-dose group, from 1.0 at baseline to 1.30 at year 3 and 1.18 at year 5. There was no effect on liver enzymes, and no increase above defined maximal plasma retinol levels (3.49 μmol/L) (Appendix F, Evidence Tables 4a–4d).

Observational studies

The possibility that high intakes of retinol increase the risk of hip fractures, particularly in postmenopausal women, has been raised by one observational study that tracked 35 77-year-old women for 18 years. ¹⁵⁴ This study reported an increased risk of hip fractures in persons at the higher quartile of total retinol intake. However, there was no significant difference in fracture risk between users and non-users of multivitamin or vitamin A supplements. These provided around 25 percent of the total daily retinol intake, or around 400–500μg RE/day. There was no association between hip fractures and β-carotene intake, either total, from foods, or from supplements.

Another, 9-year observational study in 34,000 postmenopausal women found no significant correlation between food or supplemental retinol intake and hip or all-type fractures.¹⁵⁵

Cross-sectional studies

A cross-sectional study in 178 Swedish women¹⁵⁶ reported a significant negative correlation between dietary retinol intake and BMD. The authors attributed this finding to the very high retinol intake in Nordic countries, associated with the common use of cod liver oil and the fortification of milk with vitamin A. The potential contribution of vitamin supplements was not reported in this study. Another more recent cross-sectional assessment of 11,000 women enrolled in the WHI cohort ¹⁵⁷ found no correlation between diet-only or total retinol intake and BMD. Blood retinol levels, measured in a subsample, were not correlated with BMD either. Similarly, an analysis of data from the NHANES III survey found no correlation between serum retinyl ester concentrations and BMD.¹⁵⁸

In terms of the possible effects of total daily vitamin intake, a conservative interpretation of the limited human data may be warranted, because of the biological plausibility of a negative effect of excess vitamin A on bone. However, the data specifically linking vitamin A supplements or multivitamins containing retinol to fracture risk are very limited and insufficient to draw a definitive conclusion at this time.

Vitamin E

The VECAT study administered 500 IU of vitamin E per day to 1200 volunteers (50–88 years of age) for 4 years. ¹¹³ No difference in adverse events or mortality was identified between active and placebo groups.

Another study administered vitamin E to healthy adults, but is not discussed here because of its low sample size (n=42 total, divided in 4 arms), short follow up (6 weeks), and lack of outcome data relevant to this report.

In the WHS,⁸⁷ participants received 600 IU of vitamin E every other day. No excess adverse effects were identified in the active group, except for marginally significant increased epistaxis. Authors attributed this to a chance finding, since there was no other evidence of an adverse effect on bleeding (coagulation time, hemorrhage, hemorrhagic stroke, etc). The PPP study¹¹² administered 300 mg/d for 3.6 years, to people more than 65 yrs of age. Only bleeding and mortality were monitored, and no significant differences in these outcomes were found between active and control groups (Appendix F, Evidence Tables 4e–4g).

β-carotene

The beta-carotene arm of the WHI study ⁹⁶ administered 50 mg/day of beta-carotene to about 20,000 women for 2 years. The only adverse effect associated with treatment was yellowing of the skin.

Another randomized controlled trial ⁸⁴ followed about 400 adults for 4 years, administering 30 mg/day of beta-carotene or placebo. This study did not report specific events associated with the beta-carotene arm, but the number of withdrawals associated with self-reported adverse effects of the supplement was 65 in the active group and 64 in the placebo group.

The PHS administered 50 mg of beta-carotene on alternate days to about 11,000 participants for almost 12 years. The only significant adverse effects reported were yellowing of the skin (1700 in active vs. 1500 in placebo) and minor gastrointestinal symptoms, such as belching (275 in active vs. 124 in placebo) (Appendix F, Evidence Tables 4a–4d).

Selenium

One randomized controlled trial administered 200 μg/day of selenium for 4.5 years to 1300 patients with a history of skin cancer. ¹³³ More participants complained of gastrointestinal symptoms in the active group than in the placebo group (21 vs. 14). There were no differences in plasma selenium levels between those reporting symptoms and those who did not (Appendix F, Evidence Tables 4h–4j).

Iron

The possible adverse effect of iron supplementation in healthy children is an issue receiving intense scrutiny at this time. An early report from a small randomized trial in 40 iron-sufficient, non-anemic children showed a significant reduction in weight gain over 4 months in supplemented (3mg/kg/day) children compared to placebo ¹⁵⁹. More recent trials have not fully clarified this issue, because they targeted deficient populations and/or included other micronutrients in the intervention formulation (Appendix F, Evidence Tables 4h–4j).