High-Dosage Vitamin E Supplementation and All-Cause Mortality

Full text

13. Mitchinson MJ, Stephens NG, Parsons A, Bligh E, Schofield PM, Brown MJ.
Mortality in the CHAOS trial [Letter]. Lancet. 1999;353:381-2. [PMID: 9950454]
TO THE EDITOR: There are several flaws in the meta-analysis by
Miller and colleagues (1), including erroneous interpretation of the
pooled trials of
␣
-tocopherol and lack of clarity in vitamin E nomen-
clature. The analysis includes trials from many time periods, with
different trial designs, doses and combinations, and end points that
make comparisons difficult and fallacious. Participants in several of
the trials had significant medical conditions, such as coronary artery
disease, end-stage renal disease, diabetes mellitus, Parkinson disease,
and Alzheimer disease. Given this heterogeneity in the participant
pool, we would consider it presumptuous to draw solid conclusions
of the magnitude of Miller and colleagues’, even with complex sta-
tistical tools such as meta-analysis, and to extend the observations to
normal, healthy people. In addition to clinical heterogeneity, this
analysis also suffered from heterogeneity in test nutrients. In many of
the trials, vitamin E was used alone and in combination with another
nutrient, such as
␤
-carotene. In these cases, the authors combined
the data for vitamin E alone with the data for vitamin E plus another
nutrient even when the data for the other nutrient indicated that it
was statistically significantly associated with increased mortality.
Moreover, those studies in which fewer than 10 deaths occurred were
excluded from the meta-analyses, giving an artificial weight to studies
in which more patients died—that is, those in which patients had
serious illness compared with studies of healthy individuals. A close
look at the odds ratios in Miller and colleagues’ Figure 2 does not
suggest harmful effects at a dose of 400 IU of
␣
-tocopherol, and yet
the authors concluded that harmful effects begin at a dose of 150 IU.
Vitamin E is not a single compound and exists in 8 different
isoforms in nature (4 tocopherols and 4 tocotrienols) that have dis-
tinct biopotencies, biokinetics, and cancer-preventive properties.
Food sources vary in their content of the vitamin E isoforms.
␥
-To-
copherol, the primary source of dietary vitamin E, is abundant in
plant seeds (corn, soybean, and sesame), vegetable oils, and nuts
(walnuts, pecans, and peanuts). It is not appropriate to “lump” all
the different forms of vitamin E into a single basket and call them
“vitamin E.” Natural vitamin E forms have different properties than
synthetic vitamin E does. Most of the trials cited in Miller and
colleagues’ analysis used synthetic
␣
-tocopherol. This should have
been emphasized. Dietary and supplemental sources of vitamin E
isoforms have unique properties that can influence critical pathways
involved in cancer, inflammation, cardiovascular disease (CVD), and
neurodegenerative disease. For example, mechanistic differences be-
tween the
␣
- and
␥
-tocopherols and their metabolites provide a mo-
lecular basis for the superiority of
␥
-tocopherol (2–5). Although
␣
-tocopherol has a high concentration in supplements, the primary
form of vitamin E in the diet is
␥
-tocopherol, which is present at a
concentration 2 to 4 times higher than that in
␣
-tocopherol. A high
intake of synthetic
␣
-tocopherol can lower plasma and tissue levels of
␥
-tocopherol. We believe that carefully conducted randomized studies
with long follow-up periods and well-defined end points are required to
address the potential clinical efficacy of the different isoforms of vitamin
E. We also have to be clear on the terminology that is used when we
discuss the properties and effects of the different forms of vitamin E.
Koyamangalath Krishnan, MD
The University of Texas M.D. Anderson Cancer Center
Houston, TX 77030
Sharon Campbell, PhD
William L. Stone, PhD
East Tennessee State University
Johnson City, TN 37614
Potential Financial Conflicts of Interest: None disclosed.
References
1. Miller ER 3rd, Pastor-Barriuso R, Dalal D, Riemersma RA, Appel LJ, Guallar E.
Meta-analysis: high-dosage vitamin E supplementation may increase all-cause mortal-
ity. Ann Intern Med. 2005;142:37-46. [PMID: 15537682]
2. Wagner KH, Kamal-Eldin A, Elmadfa I. Gamma-tocopherol—an underestimated
vitamin? Ann Nutr Metab. 2004;48:169-88. [PMID: 15256801]
3. Hensley K, Benaksas EJ, Bolli R, Comp P, Grammas P, Hamdheydari L, et al. New
perspectives on vitamin E: gamma-tocopherol and carboxyelthylhydroxychroman me-
tabolites in biology and medicine. Free Radic Biol Med. 2004;36:1-15. [PMID:
14732286]
4. Stone WL, Krishnan K, Campbell SE, Qui M, Whaley SG, Yang H. Tocopherols
and the treatment of colon cancer. Ann N Y Acad Sci. 2004;1031:223-33. [PMID:
15753148]
5. Campbell S, Stone W, Whaley S, Krishnan K. Development of gamma (gamma)-
tocopherol as a colorectal cancer chemopreventive agent. Crit Rev Oncol Hematol.
2003;47:249-59. [PMID: 12962899]
TO THE EDITOR: In their important meta-analysis focusing on the
potentially harmful effects of vitamin E supplementation, Miller and
colleagues (1) assumed that there may be a precise threshold level
and that larger intakes of the vitamin would progressively increase
the risk for harm. However, there may be biological heterogeneity
between population groups, meaning that persons’ characteristics
would determine whether vitamin E supplementation causes net
benefit or harm.
In the Alpha-Tocopherol, Beta-Carotene Cancer Prevention
(ATBC) Study, the effect of vitamin E on the risk for pneumonia
was significantly modified by the age of smoking initiation (P⬍
0.001) (2). Vitamin E increased pneumonia risk in those who began
smoking at age 20 years or earlier (relative risk, 1.14 [95% CI, 0.98
to 1.32]), but decreased pneumonia risk in participants who began
smoking at later ages (relative risk, 0.65 [CI, 0.49 to 0.86]). Further-
more, in the latter subgroup, the benefit was greater among those
who smoked less or quit smoking during the trial. Thus, less expo-
sure to cigarette smoking was associated with greater benefit of vita-
min E.
In the ATBC Study, the vitamin E dosage was 50 IU/d (50
mg/d of dl-
␣
-tocopheryl acetate), which is substantially less than the
threshold of 150 IU/d estimated by Miller and colleagues (1). How-
ever, from the subgroup differences described in the ATBC Study, it
seems probable that some population groups experience ill effects at
this low dosage while other persons benefit. Assuming that there is
biological heterogeneity between population groups, further studies
should characterize people who benefit from or are harmed by vita-
min E rather than just estimating a uniform threshold for harm and
presuming it is valid for everyone.
Letters
www.annals.org 19 July 2005 Annals of Internal Medicine Volume 143 • Number 2 151

Harri Hemilä, PhD, MD
University of Helsinki
FIN-00014 Helsinki, Finland
Potential Financial Conflicts of Interest: None disclosed.
References
1. Miller ER 3rd, Pastor-Barriuso R, Dalal D, Riemersma RA, Appel LJ, Guallar E.
Meta-analysis: high-dosage vitamin E supplementation may increase all-cause mortal-
ity. Ann Intern Med. 2005;142:37-46. [PMID: 15537682]
2. Hemilä H, Virtamo J, Albanes D, Kaprio J. Vitamin E and beta-carotene supple-
mentation and hospital-treated pneumonia incidence in male smokers. Chest. 2004;
125:557-65. [PMID: 14769738]
TO THE EDITOR: Miller and colleagues (1) highlighted the danger of
assuming the safety of high-dose vitamin E in the absence of defin-
itive long-term safety data. The impact of their study, however, may
be mitigated by methodologic concerns.
The first issue is the restrictive inclusion criteria stipulating that
a trial have at least 10 deaths, apparently because the authors “antic-
ipated that many small trials did not collect mortality data.” This
exclusion contradicts the raison d’être of meta-analysis, which in-
volves the statistical pooling of multiple trials that individually have
inadequate statistical power. The exclusion of at least 3 reasonably
large, well-conducted trials (2–4) of high-dose vitamin E in which
fewer than 10 deaths occurred and the inclusion only of trials meet-
ing this arbitrary cutoff would spuriously increase the power of the
meta-analysis. We would also be interested in the funnel plot analysis
to determine whether publication bias affected the study results.
Although the authors attempted to adjust for average follow-up
in their analysis, a more robust statistical treatment of the variance in
follow-up periods across included trials would be to express the sum-
mary statistic of pooled death risk as the number of deaths per
10 000 person-years (as opposed to per 10 000 persons).
Heterogeneity in the study samples may not have been fully
accounted for despite the use of the random-effects model and dos-
age differentiation. In particular, people with CVD may be a select
group at distinct risk from the effects of high-dose vitamin E. Seven
of the 8 high-dosage trials showing harmful effects of vitamin E
involve participants with vascular risk factors or those who had es-
tablished CVD. In contrast, the Deprenyl and Tocopherol Antioxi-
dative Therapy for Parkinson’s Disease (DATATOP) study and the
Alzheimer’s Disease Cooperative Study (ADCS) used megadoses of
vitamin E (2000 IU/d) in individuals with neurodegenerative disor-
ders rather than in those with CVD but did not reveal safety con-
cerns. A separate meta-analysis looking solely at neurodegenerative
diseases (including a recent study using 5000 mg of vitamin E per
day [5]) may be warranted.
Although Miller and colleagues’ study may have focused on
safety, the data ultimately challenge the advocates of high-dose vita-
min E to reexamine the evidence for benefit. Efficacy in controlled
trials ranges from minimal to modest, in contrast to the more posi-
tive results of observational studies. It is time for clinicians to return
to the drawing board and review both the safety and efficacy data for
vitamin E supplementation.
Wee-Shiong Lim, MBBS
Rajka M. Liscic, MD, PhD
Chengjie Xiong, PhD
John C. Morris, MD
Washington University School of Medicine
St. Louis, MO 63108
Potential Financial Conflicts of Interest: None disclosed.
References
1. Miller ER 3rd, Pastor-Barriuso R, Dalal D, Riemersma RA, Appel LJ, Guallar E.
Meta-analysis: high-dosage vitamin E supplementation may increase all-cause mortal-
ity. Ann Intern Med. 2005;142:37-46. [PMID: 15537682]
2. Berson EL, Rosner B, Sandberg MA, Hayes KC, Nicholson BW, Weigel-DiFranco
C, et al. A randomized trial of vitamin A and vitamin E supplementation for retinitis
pigmentosa. Arch Ophthalmol. 1993;111:761-72. [PMID: 8512476]
3. de Waart FG, Kok FJ, Smilde TJ, Hijmans A, Wollersheim H, Stalenhoef AF.
Effect of glutathione S-transferase M1 genotype on progression of atherosclerosis in
lifelong male smokers. Atherosclerosis. 2001;158:227-31. [PMID: 11500195]
4. Wluka AE, Stuckey S, Brand C, Cicuttini FM. Supplementary vitamin E does not
affect the loss of cartilage volume in knee osteoarthritis: a 2 year double blind random-
ized placebo controlled study. J Rheumatol. 2002;29:2585-91. [PMID: 12465157]
5. Graf M, Ecker D, Horowski R, Kramer B, Riederer P, Gerlach M, et al. High dose
vitamin E therapy in amyotrophic lateral sclerosis as add-on therapy to riluzole: results
of a placebo-controlled double-blind study. J Neural Transm. 2005;112:649-660.
[PMID: 15517433]
TO THE EDITOR: In a meta-analysis of selected randomized trials,
Miller and colleagues (1) found an increase in all-cause mortality
associated with high-dosage vitamin E supplementation for at least 1
year (1). Their analysis included the DATATOP study, a random-
ized, placebo-controlled trial of selegiline and vitamin E in 800 pa-
tients with early Parkinson disease (2). In the DATATOP trial, 399
of 800 participants were randomly assigned to receive 2000 IU of
vitamin E per day, the highest dosage studied in Miller and col-
leagues’ meta-analysis. Median duration of vitamin E exposure dur-
ing the randomized phase was 2.6 years. We have now accrued 13
years of follow-up and have documented 296 deaths compared with
the 137 DATATOP deaths incorporated into Miller and colleagues’
meta-analysis (3).
A slightly but not significantly higher proportion of patients
randomly assigned to vitamin E had died by 13 years after enroll-
ment (154 of 399 in the vitamin E group [39%] vs. 142 of 401 in
the placebo group [35%]; P⫽0.35). However, after adjustment for
age and sex in a logistic regression, there was no excess mortality in
the group assigned to vitamin E (odds ratio, 0.996 [95% CI, 0.72 to
1.38]; P⫽0.98).
We also constructed a Cox model to make full use of survival
information, incorporating duration of blinded exposure to vitamin
E as a time-dependent covariate. We did not observe increased mor-
tality for each additional year of exposure to vitamin E (hazard ratio,
1.05 [CI, 0.95 to 1.16]; P⫽0.31). Similar results were obtained
when both blinded and subsequent open-label tocopherol supple-
mentation were considered.
We found no evidence of increased mortality in DATATOP
related to 2.6 years of high-dosage vitamin E exposure through 13
years of observation. The DATATOP cohort was selected for ab-
sence of serious comorbid illness and was more highly educated than
the general population. These and other differences in selection may
Letters
152 19 July 2005 Annals of Internal Medicine Volume 143 • Number 2 www.annals.org