Meagher EA, Barry OP, Lawson JA, Rokach J, FitzGerald GA. Effects of Vitamin E on Lipid Peroxidation in Healthy Persons. JAMA. 2001;285(9):1178-1182. doi:10.1001/jama.285.9.1178
Author Affiliations: Center for Experimental Therapeutics, University of Pennsylvania, Philadelphia (Drs Meagher, Barry, and FitzGerald and Mr Lawson); and the Claude Pepper Institute and Department of Chemistry, Florida Institute of Technology, Melbourne (Dr Rokach).
Context Oxidative stress may play a role in the development or exacerbation
of many common diseases. However, results of prospective controlled trials
of the effects of antioxidants such as vitamin E are contradictory.
Objective To assess the effects of supplemental vitamin E on lipid peroxidation
in vivo in healthy adults.
Design Randomized, double-blind, placebo-controlled trial conducted March 1999
to June 2000.
Setting A general clinical research center in a tertiary referral academic medical
Participants Thirty healthy men and women aged 18 to 60 years.
Interventions Participants were randomly assigned to receive placebo or α-tocopherol
dosages of 200, 400, 800, 1200, or 2000 IU/d for 8 weeks (n = 5 in each group),
followed by an 8-week washout period.
Main Outcome Measures Three indices of lipid peroxidation, urinary 4-hydroxynonenal (4-HNE)
and 2 isoprostanes, iPF2α-III and iPF2α-VI,
measured by gas chromatography/mass spectrometry and compared among the 6
groups at baseline, 2, 4, 6, and 8 weeks, and 1, 3, and 8 weeks after discontinuation.
Results Circulating vitamin E levels increased in a dose-dependent manner during
the study. No significant effect of vitamin E on levels of urinary 4-HNE or
either isoprostane was observed. Mean (SEM) baseline vs week 8 levels of iPF2α-III were 154 (20.1) vs 168 (22.3) pg/mg of creatinine for subjects
taking placebo; 165 (19.6) vs 234 (30.1) pg/mg for those taking 200 IU/d of
vitamin E; and 195 (26.7) vs 213 (40.6) pg/mg for subjects taking 2000 IU/d.
Corresponding iPF2α-VI levels were 1.43 (0.6) vs 1.62 (0.4)
ng/mg of creatinine for subjects taking placebo; 1.64 (0.3) vs 1.24 (0.8)
ng/mg for those taking 200 IU/d of vitamin E; and 1.83 (0.3) vs 1.94 (0.9)
ng/mg for those taking 2000 IU/d. Baseline vs week 8 levels of 4-HNE were
0.5 (0.04) vs 0.4 (0.05) ng/mg of creatinine for subjects taking placebo;
0.4 (0.06) vs 0.5 (0.02) ng/mg with 200 IU/d of vitamin E; and 0.2 (0.02)
vs 0.2 (0.1) ng/mg with 2000 IU/d.
Conclusions Our results question the rationale for vitamin E supplementation in
healthy individuals. Specific quantitative indices of oxidative stress in
vivo should be considered as entry criteria and for dose selection in clinical
trials of antioxidant drugs and vitamins in human disease.
Oxidative stress appears to be of fundamental relevance to diseases
as diverse as atherosclerosis, cancer, and Alzheimer disease.1- 3
However, prospective, controlled clinical trials of antioxidants present a
confused picture. For example, while administration of vitamin E appeared
to benefit patients with coronary disease in the CHAOS study,4
the HOPE and GISSI Prevenzione studies failed to detect such a benefit.5,6 Similarly, while dietary flavonoids
appear to reduce cardiovascular mortality, supplemental β-carotene and
vitamin A appear to increase the risk of death from lung cancer and heart
disease in smokers and workers exposed to asbestos.7,8
Various possibilities have been advanced to explain this discrepancy, including
differences in patient characteristics, the antioxidant content of their diets,
dose selection, and random distribution of outcomes about the mean. However,
a striking feature of these and other trials of antioxidants is the absence
of a biochemical basis for patient inclusion or dose selection.
Elaborate and diversified antioxidant mechanisms protect tissues from
oxidative damage in humans and other organisms,9- 11
and susceptibility to benefit from exogenous antioxidants in vitro is conditioned
by the degree of their depletion.12- 14
Additionally, certain vitamins with antioxidant properties can function as
prooxidants, at least in vitro.15,16
Despite these observations, nothing is known about the susceptibility of clinical
trial patients to supplementation with antioxidants or whether the doses selected
exhibit antioxidant effects. What little is known of the dose-response relationships
of antioxidants in humans is based on an ex vivo assay of the oxidizability
of low-density lipoprotein (LDL) cholesterol. Small studies using this approach
report variable effects of vitamin E supplementation.17,18
However, this assay bears an uncertain relationship to actual oxidation of
LDL in vivo, and its relationship to oxidation of other lipid substances and
to disease pathogenesis is unclear.
Reliable, quantitative indices of free radical–induced modification
of lipids, proteins, and DNA in vivo have begun to emerge only recently.2,19 Most information acquired in humans
relates to lipid peroxidation. F2isoprostanes are free radical–catalyzed
prostaglandin F2 isomers.20,21
They are chemically stable and can be measured noninvasively in urine with
precision and sensitivity using homologous internal standards and mass spectrometry.22- 25 We
used this approach to investigate the dose-response relationships of vitamin
E with lipid peroxidation in healthy volunteers.
To investigate the dose-response relationship of vitamin E with lipid
peroxidation in healthy volunteers, we performed a randomized, double-blind,
placebo-controlled study from March 1999 to June 2000. The study was approved
by the institutional review board of the University of Pennsylvania and the
General Clinical Research Center Advisory Committee and all participants provided
informed consent. Thirty volunteers, 15 men and 15 women, between the ages
of 18 and 60 years (mean [SD], 38 [12.5] years) were randomized to 1 of 6
dosing groups: placebo or vitamin E at 200, 400, 800, 1200, or 2000 IU/d for
8 weeks, followed by an 8-week washout period. There were 5 subjects in each
group. All volunteers were nonsmokers. In addition, they were all less than
120% of ideal body weight and had normal levels of vitamin E, vitamin C, selenium,
and cholesterol at screening. Exclusion criteria included intake of any vitamin
supplements within the preceding month, any medical illness, or use of any
medications known to interfere with lipid metabolism within the last month.
Participants were screened for study eligibility in the General Clinical Research
Center at the University of Pennsylvania.
Vitamin E was supplied as (d) α-tocopherol capsules. Placebo capsules
were identical in size, shape, and color. All subjects were available for
measurement at all study points. Compliance was monitored by regular capsule
counts and serum vitamin E levels. Twenty-four-hour urine collections for
isoprostanes and serum vitamin E, vitamin C, and selenium levels were measured
at baseline and at 2, 4, 6, and 8 weeks of dosing and 1, 3, and 8 weeks after
Urinary iPF2α-III and iPF2α-VI were
assayed by stable-isotope dilution gas chromatography/mass spectrometry as
previously described.22,23 Urinary
creatinine was determined using a standard automated colorimetric assay (Beckman
Synchron CX System, Beckman Instruments, Arlington Heights, Ill).
The method for urinary measurement of 4-hydroxynonenal (4-HNE)26 was adapted from one previously developed27 to measure 4-hydroxyalkenals in oxidized LDL. Briefly,
5-mL urine samples were spiked with 5 ng of d3-HNE, mixed well, and allowed
to equilibrate for 15 minutes at room temperature. Two milligrams of (2,3,4,5,6-pentafluorobenzyl)
hydroxylamine hydrochloride was added to each sample, and they were allowed
to stand for 30 minutes at room temperature. 4-Hydroxynonenal was extracted
using reverse-phase solid-extraction cartridges (C18 EC, 500 mg; International
Sorbent Technology Ltd, Mid Glamorgan, Wales) under the following conditions.
The cartridge was conditioned with 5 mL of ethanol and washed with 1.5 mL
of distilled water. A sample was loaded onto the cartridge, which was washed
with 3 mL of 60% ethanol. The cartridge was dried for 10 minutes, and the
sample was eluted with 3 mL of ethyl acetate. The sample was then dried under
a stream of nitrogen and dissolved in 1 mL of hexane. A second extraction
used straight-phase solid-extraction cartridges (100 mg of silica conditioned
with 1 mL of hexane). The sample was eluted with 1 mL of 30% ethyl acetate
in hexane, dried, and dissolved in 15 µL of dodecane. One microliter
of the sample was used for gas chromatography/mass spectrometry analysis.
The mass spectrometer was operated in the negative-ion, electron capture ionization
mode, using ammonia as the moderating gas. Ions monitored were of mass-to-mass
charge ratio 283 and 286 for 4-HNE and d3-HNE, respectively.
Serum levels of vitamin E and plasma concentrations of vitamin C were
measured by high-performance liquid chromatography.28,29
Serum selenium concentrations were measured by atomic absorption spectrometry.30
Tests of statistical hypothesis for main effects were based on a fixed
type I error rate of 5%. The study was powered to detect a 10% change in urinary
isoprostane and 4-HNE measurements. Data were initially subjected to analysis
of variance with subsequent pairwise analysis using a 2-tailed t test, as appropriate. Data are expressed as mean (SEM). To account
for potential differences among group baseline values, all successive time
values were adjusted for each subject's baseline measure: Δ(i,k) = value(i,k) −
value(i,0), for each subject i at k = 2, 4, 8, 9, 12, and 16 weeks. These adjusted measurements were
subjected to an analysis of variance appropriate for a 2-factor experiment
design with 1 repeated measure (time) and 1 nonrepeated measure (dose).
Prior to dosing, all measurements of endogenous antioxidants were within
normal limits. Serum vitamin E increased in a dose-dependent manner and reached
a steady state by 8 weeks. For example, levels at baseline vs 8 weeks of placebo
were 9.4 (2.1) vs 8.5 (0.8) mg/dL (reference range for serum vitamin E, 4.6-14.5
mg/dL). Levels rose from 8.4 (1.4) to 20.7 (2.8) mg/dL in subjects taking
200 IU/d and from 8.9 (1.0) to 52.8 (4.7) mg/dL in those taking 2000 IU/d.
Vitamin E levels declined to preintervention levels at 8 weeks following dosing
(week 16). Levels of endogenous selenium and vitamin C were unaltered by vitamin
Urinary isoprostanes were also unaltered by vitamin E therapy. For example,
the corresponding baseline vs week 8 levels of urinary iPF2α-III
were 154 (20.1) vs 168 (22.3) pg/mg of creatinine for subjects taking placebo;
165 (19.6) vs 234 (30.1) pg/mg of creatinine for those taking 200 IU/d of
vitamin E; and 195 (26.7) vs 213 (40.6) pg/mg of creatinine for subjects taking
2000 IU/d (Figure 1, A). None of
these changes attained statistical significance. Similarly, urinary iPF2α-VI levels were 1.43 (0.6) vs 1.62 (0.4) ng/mg of creatinine
for subjects taking placebo; 1.64 (0.3) vs 1.24 (0.8) ng/mg of creatinine
for those taking 200 IU/d of vitamin E; and 1.83 (0.3) vs 1.94 (0.9) ng/mg
of creatinine for those taking 2000 IU/d (Figure 1, B). None of these differences were statistically significant.
Urinary 4-HNE was also unaltered by vitamin E supplementation. For example,
levels were 0.5 (0.04) vs 0.4 (0.05) ng/mg of creatinine for subjects taking
placebo; 0.4 (0.06) vs 0.5 (0.02) ng/mg of creatinine after 8 weeks of therapy
with 200 IU/d of vitamin E; and 0.2 (0.02) vs 0.2 (0.10) ng/mg of creatinine
after 8 weeks of 2000 IU/d of vitamin E. No adverse reactions to vitamin E
supplementation were reported.
The repeated measures design of this study enabled adjustment for baseline
variation among the dosing groups. No effect was seen following subtraction
of predosing baseline values for each of the variables of interest. For example,
the absolute changes in iPF2α-III, iPF2α-VI,
and 4-HNE excretion after 8 weeks of 800 IU/d of vitamin E were −2.76
(5.35) pg/mg of creatinine, −0.01 (0.27) ng/mg of creatinine, and 0.10
(0.05) ng/mg of creatinine, respectively.
Given a nonsignificant (P = .13) second-order
interaction between dose and time for 4-HNE, we examined the main effects,
dose and time, which were each statistically nonsignificant (P = .71 and P = .13, respectively). The second-order
interaction (dose × time) for iPF2α-III also failed
to attain significance (P = .62). Dose and time were
again statistically nonsignificant (P = .31 and P = .14, respectively). Finally, the second-order interaction
(dose × time) for iPF2α-VI was also nonsignificant
(P = .92). Dose and time were also statistically
nonsignificant (P = .49 and P
= .25, respectively).
Given that studies of LDL oxidation ex vivo17,18
have suggested that the effects of vitamin E supplementation approach a maximum
at daily doses of 400 IU, we performed a subsidiary analysis that compared
the effects of the combined doses of 400, 800, 1200, and 2000 IU/d with placebo.
We performed t tests for independent samples for
4-HNE, iPF2α-III, and iPF2α-VI to compare
the 5 participants who received placebo with the 20 who received ≥400 IU/d
of vitamin E. In all cases, nonsignificant P values
were obtained: .78, .84, and .52 respectively.
Oxidative damage is widely implicated in the pathogenesis of disease.
While traditional indices of this process in integrated systems are recognized
as fallible, several novel approaches have been developed that permit quantitation
of the consequences of excessive free radical generation in vivo. Isoprostanes
are free radical–catalyzed isomers of prostaglandins, peroxidation products
of arachidonic acid.20,21 Initially
formed in situ in the phospholipid domain of cell membranes, they are cleaved
out, circulate, and are excreted in urine. They effect a range of biological
activities in vitro and may activate both membrane receptors for traditional
prostaglandins31 and peroxisomal proliferator
activated receptors in the nucleus.32 Given
the complexity of the species (up to 64 F2 isoprostanes can be
generated), we have developed highly specific assays for individual isoprostanes
rather than using semiquantitative estimates of "total isoprostanes."21 Using this approach, we have demonstrated increased
generation of isoprostanes in cigarette smokers,33
abusers of alcohol,34 and in persons with a
range of ischemia/reperfusion25,35,36
and inflammatory syndromes.37- 40
In the present study, we sought an effect of vitamin E on 2 isoprostanes,
iPF2α-III and iPF2α-VI. The former compound
may be generated either by cyclooxygenase (COX) turnover or by a free radical–dependent
mechanism.22,41 However, the COX-dependent
pathway contributes to an undetectable extent to urinary iPF2α-III, even in syndromes of COX activation.33,42
By contrast, iPF2α-VI, a more abundant entity, is formed
only as a product of lipid peroxidation.
Along with isoprostanes, we also measured 4-HNE, an independent index
of lipid peroxidation. Like the isoprostanes, urinary 4-HNE is increased in
patients with alcohol-induced liver disease. In these patients, who are deficient
in endogenous vitamin C, supplements of vitamin C reduce their elevated levels
of both urinary isoprostanes and 4-HNE.34 Similarly,
supplementation with vitamin C, but not vitamin E, reduces elevated levels
of urinary iPF2α-III in cigarette smokers, who are also selectively
depleted of vitamin C.33 We have also shown
that exogenous vitamin E alone or in combination with vitamin C reduces elevated
levels of urinary isoprostanes in patients with virally induced cirrhosis,43 the antiphospholipid syndrome,40
and obstructive pulmonary disease.37 Elevated
urinary isoprostanes can also be used to select a rational dose of an antioxidant,
such as vitamin E. For example, a dosage of vitamin E that suppresses elevated
levels of urinary iPF2α-VI retards atherogenesis in the apolipoprotein
E–deficient mouse,44 whereas a lower
dosage, selected without a biochemical rationale, fails to influence atherogenesis
in this model.45
Despite these observations, we failed to detect any impact of vitamin
E on 3 quantitative indices of lipid peroxidation. This was true when we examined
the absolute values of these indices after adjustment for interindividual
baseline variation. This was also true when we compared the values after placebo
with those after ≥400 IU/d of. vitamin E, doses at which effects on LDL
oxidizability ex vivo have been noted.17,18
We administered vitamin E over a broad dose range, incorporating the doses
used in the CHAOS, HOPE, and GISSI Prevenzione studies,4- 6
all considerably in excess of the recommended daily allowance.46
We administered the natural (d) α-tocopherol isomer of vitamin E for
a sufficient period (8 weeks) to achieve steady-state incorporation into cell
The study population included individuals assessed as consuming a diet
replete with endogenous vitamin E. Indices of endogenous antioxidant defense,
including vitamin E, were in the normal range prior to initiation of the study.
Vitamin E levels increased in a dose-dependent manner to a maximum of roughly
5-fold, thus excluding protocol noncompliance as an explanation for our results.
Although the relatively small size of this study may have precluded detection
of subtle effects of vitamin E, the study was designed to detect changes of
at least 10% in any of the parameters.
This study has implications for the evaluation of clinical trials of
antioxidants and for the widespread consumption of antioxidants by apparently
healthy individuals. First, the inclusion of patients without biochemical
evidence of increased oxidative stress in clinical trials of antioxidants
would be expected to dilute the population susceptible to benefit, even assuming
the functional importance of oxidative stress in the disease under evaluation.
This might seriously undermine the sample size calculations used in such trials,
leaving them open to a type II statistical error and outcomes reflecting random
variation about the mean. Such issues occurred for a decade in clinical trials
of aspirin in cardiovascular disease. The absence of an index of thromboxane
biosynthesis led to the inclusion of many patients in whom, in retrospect,
thromboxane-dependent platelet activation was unlikely to have been abnormal.
This, in turn, reduced the ability to detect a significant cardioprotective
effect of aspirin.50- 52
Detection of marked increases in thromboxane biosynthesis in the ischemic
episodes of unstable angina and coincident with therapeutic thrombolysis53 afforded the rationale for clinical studies that
clearly demonstrated the efficacy of aspirin in these settings.54,55
Second, incorporation of measurements such as urinary isoprostanes and
4-HNE may also be used in the rational selection of antioxidant dosages for
such trials. Several studies have indicated the prooxidant potential of antioxidant
vitamins, including vitamin E,16 and this may
have functional relevance in vivo. For example, low doses of α-tocopherol
improve endothelial function in hypercholesterolemic rabbits, but is worsened
by higher doses of the vitamin.56 We found
no evidence of a net prooxidant effect of vitamin E. It is theoretically possible
that competing prooxidant and antioxidant effects of vitamin E canceled each
other out, but if so, this was unrelated to dosage. Although we and others
have provided some information on the biochemical effects of such supplements
in vivo in diseased populations, the optimal antioxidant regimens for specific
conditions may vary.57 Such regimens should
be defined before initiating large-scale clinical trials.58
Finally, the average Western diet provides the recommended daily allowance
of vitamin E,46 and the endogenous levels in
the individuals in our study fell within the normal range. We found no evidence
of additional effects of supplementing these individuals with a range of dosages
of vitamin E on their rate of lipid peroxidation in vivo. Our findings question
the potential benefit of the reportedly widespread consumption of vitamin
E by such healthy individuals.59