Smoothed relative risk curves and their 95% pointwise confidence intervals in calendar time. ATBC, Alpha-Tocopherol, Beta-Carotene Cancer Prevention.
Relative risk estimates for comparisons during the trial and during the 6-year posttrial follow-up. The whiskers span the 95% confidence intervals. The size of the squares is proportional to the proportion of all deaths (diamonds) due to each cause-specific group. ATBC, Alpha-Tocopherol, Beta-Carotene Cancer Prevention.
Incidence of Cancer and Mortality Following α-Tocopherol and β-Carotene SupplementationA Postintervention Follow-up. JAMA. 2003;290(4):476-485. doi:10.1001/jama.290.4.476
Context In the Finnish Alpha-Tocopherol, Beta-Carotene Cancer Prevention (ATBC)
Study, α-tocopherol supplementation decreased prostate cancer incidence,
whereas β-carotene increased the risk of lung cancer and total mortality.
Postintervention follow-up provides information regarding duration of the
intervention effects and may reveal potential late effects of these antioxidants.
Objective To analyze postintervention effects of α-tocopherol and β-carotene
on cancer incidence and total and cause-specific mortality.
Design, Setting, and Participants Postintervention follow-up assessment of cancer incidence and cause-specific
mortality (6 years [May 1, 1993-April 30, 1999]) and total mortality (8 years
[May 1, 1993-April 30, 2001]) of 25 563 men. In the ATBC Study, 29 133
male smokers aged 50 to 69 years received α-tocopherol (50 mg), β-carotene
(20 mg), both agents, or placebo daily for 5 to 8 years. End point information
was obtained from the Finnish Cancer Registry and the Register of Causes of
Death. Cancer cases were confirmed through medical record review.
Main Outcome Measures Site-specific cancer incidence and total and cause-specific mortality
and calendar time-specific risk for lung cancer incidence and total mortality.
Results Overall posttrial relative risk (RR) for lung cancer incidence (n =
1037) was 1.06 (95% confidence interval [CI], 0.94-1.20) among recipients
of β-carotene compared with nonrecipients. For prostate cancer incidence
(n = 672), the RR was 0.88 (95% CI, 0.76-1.03) for participants receiving α-tocopherol
compared with nonrecipients. No late preventive effects on other cancers were
observed for either supplement. There were 7261 individuals who died by April
30, 2001, during the posttrial follow-up period; the RR was 1.01 (95% CI,
0.96-1.05) for α-tocopherol recipients vs nonrecipients and 1.07 (95%
CI, 1.02-1.12) for β-carotene recipients vs nonrecipients. Regarding
duration of intervention effects and potential late effects, the excess risk
for β-carotene recipients was no longer evident 4 to 6 years after ending
the intervention and was primarily due to cardiovascular diseases.
Conclusions The beneficial and adverse effects of supplemental α-tocopherol
and β-carotene disappeared during postintervention follow-up. The preventive
effects of α-tocopherol on prostate cancer require confirmation in other
trials. Smokers should avoid β-carotene supplementation.
Epidemiological studies suggest that low intake or having a low serum
concentration of antioxidants is associated with elevated risk of cancer.1,2 In the 1980s, 4 large randomized trials
were initiated to assess whether antioxidants prevent cancer, cardiovascular
disease, or both. In the Nutrition Intervention Trial I conducted in China
among 29 584 men and women aged 40 to 69 years, there was a 21% (95%
confidence interval [CI], 1%-36%) reduction in stomach cancer mortality and
a 13% (95% CI, 0%-25%) reduction in total cancer mortality in response to
5 years of supplementation with a combination of vitamin E, β-carotene,
The Alpha-Tocopherol, Beta-Carotene Cancer Prevention (ATBC) Study assessed
the effect of supplemental (1) α-tocopherol only, (2) β-carotene
only, (3) α-tocopherol plus β-carotene, or (4) placebo on the incidence
of lung cancer and other cancers among 29 133 male smokers aged 50 to
69 years.4 After a median of 6.1 years of follow-up,
there was a 17% (95% CI, 2%-33%) higher incidence of lung cancer and 8% (95%
CI, 1%-15%) higher total mortality among participants who received β-carotene
compared with nonrecipients.5,6 Supplementation
with α-tocopherol had no effect on lung cancer incidence but did reduce
the incidence of prostate cancer by 34% (95% CI, 14%-48%).7 Of
note, the original values for lung and prostate cancer during the trial period
differ slightly from those presented herein because since the original reports
were published,6,7 2 cases of
lung cancer (1 individual diagnosed during the intervention period in the α-tocopherol-only
group and 1 in the placebo group) and 2 cases of prostate cancer diagnosed
during the intervention period (both individuals received β-carotene
only) have been reported. Cases of carcinoid tumor previously reported were
excluded from the present analysis (see below). The values herein represent
current up-to-date data.
The Beta-Carotene and Retinol Efficacy Trial (CARET) assessed the combination
of β-carotene and retinyl palmitate compared with placebo in 18 314
men and women aged 45 to 74 years, who were at high risk for lung cancer because
of cigarette smoking and/or occupational asbestos exposure.8 The
results were similar to the ATBC Study: 28% (95% CI, 4%-57%) higher lung cancer
incidence and 17% (95% CI, 3%-33%) higher total mortality in the group that
received combination supplementation compared with those who received placebo
after an average of 4 years.
In contrast, the Physicians' Health Study did not demonstrate an effect
of β-carotene supplementation on lung cancer or overall mortality among
22 071 men aged 40 to 84 years after an average follow-up of 12 years.9 However, only 11% of the Physicians' Health Study
participants were current smokers, thus, the study population was at substantially
lower risk for lung cancer compared with the participants in ATBC and CARET.
Similarly, in a smaller controlled trial, the Skin Cancer Prevention Study,
which involved 1805 men and women younger than 85 years, β-carotene supplementation
had no effect on overall mortality.10
The intervention phase of the ATBC Study ended in April 1993. Because
of the unexpected finding of increased lung cancer incidence from the trial,
we were interested in observing patterns of disease in the postintervention
period to determine the duration of the intervention effects and to observe
potential late effects from the intervention. The study cohort has been monitored
through the use of national registry data on mortality and cancer incidence.
We report herein the posttrial findings for these events, reflecting 8 years
of follow-up for total mortality and 6 years of follow-up for cancer incidence
and cause-specific mortality. Findings from the intervention period for site-specific
cancer incidence and for total mortality, the details of which have been previously
reported,5- 7,11- 14 are
also presented herein to facilitate interpretation of the temporal relationships.
The design of the ATBC Study has been described in detail elsewhere.4 Briefly, this randomized, double-blind, placebo-controlled
chemoprevention trial was conducted in Finland between 1985 and 1993, with
the primary objective of evaluating the effect of α-tocopherol and β-carotene
supplementation on the incidence of lung cancer and other cancers in 29 133
male smokers aged 50 to 69 years. The cohort was screened from a source population
(n = 290 406) living in 14 adjoining areas in southwestern Finland. Recruitment
began in April 1985 and continued through June 1988.
Individuals who were eligible and willing to participate in the trial
were randomly allocated to 1 of the 4 treatment regimens: α-tocopherol
only, β-carotene only, α-tocopherol plus β-carotene, or placebo.
The formulation of the study agents was synthetic dl-α-tocopheryl
acetate (50% powder) and synthetic β-carotene (10% water-soluble beadlets).
The daily doses were 50 mg of α-tocopherol and 20 mg of β-carotene.
Randomization was performed in blocks of 8 in each of the 14 study areas.
The ATBC Study received approval from the institutional review board (IRB)
of the National Public Health Institute of Finland and from the IRB of the
National Cancer Institute (NCI). All participants provided written informed
consent prior to randomization.
The trial continued until April 30, 1993, with the trial cohort followed
up for cancer incidence and total mortality through national registries thereafter.
At the end of the intervention phase, participants were informed by individual
letter that the trial results and the available other knowledge did not indicate
that α-tocopherol and β-carotene should be used for cancer prevention.
Diet and use of supplements were not monitored during the postintervention
follow-up because there was no active clinic-based follow-up. However, dietary
assessment had been performed for all participants at baseline and repeated
for approximately 800 randomly selected men annually to follow-up possible
dietary changes during the intervention. No dietary changes were observed.
Inquiry was made regarding use of nontrial supplements at every follow-up
visit (which occurred 3 times annually during the intervention phase). Nurses
instructed participants to discontinue use of high-dose vitamin E or β-carotene
We report herein the follow-up of those 25 563 participants still
alive April 30, 1993, regarding incident cancers and cause-specific deaths
through April 30, 1999 (6-year posttrial follow-up), and overall mortality
through April 30, 2001 (8-year follow-up). The follow-up times are different
because information on death is promptly available but official information
on cause of death and identification and review of cancer cases is delayed.
Because of the time required for study review of diagnosis of cancer cases,
we allowed 2 years shorter follow-up time for cancer incidence and cause-specific
mortality data than for the total mortality data so as to have as final and
complete data as possible for cancer and cause-specific mortality.
The study was approved by the IRBs of the National Public Health Institute
of Finland and the US NCI. The Ministry of Social Affairs and Health of Finland
authorized the ATBC Study to obtain data from national health registers and
sources such as hospitals and laboratories, extending to the end of 2002.
The NCI's IRB approved the posttrial follow-up without stipulating additional
informed consent because participants would not be contacted or given invitations
for further visits. Per NCI IRB requirements, the ATBC Study must report posttrial
results to the IRB of the National Public Health Institute at regular intervals;
the post-trial follow-up was evaluated by this board most recently in March
Ascertainment of cancer cases during the intervention phase has been
described.4 During the posttrial follow-up,
most cancer cases were identified via the Finnish Cancer Registry which provides
almost 100% case coverage.15 Some cancer cases
were identified also through death certificates and the National Hospital
Discharge Register. About 0.8% of cases were identified through death certificates
and the National Hospital Discharge Register and were unknown to the Finnish
Cancer Registry at the time of our most recent ascertainment of cases reported
for the analysis herein.
The medical records of all potential cancer cases were collected from
the hospitals and pathology laboratories. Two oncologists independently reviewed
the records of all reported prostate, stomach, colorectal, and pancreatic
cancers as well as cancers recorded as having an unknown primary site. In
the case of disagreement between 2 oncologists, a third oncologist reviewed
the documents and assigned the final diagnosis. Disagreement on cancer diagnosis
was uncommon (eg, of 672 postintervention prostate cancer cases, the oncologists
disagreed on only 1 case [0.15%]). In addition, a pathologist reviewed the
histopathologic and cytological specimens from these cases.
One study physician reviewed centrally the medical records of cases
of other cancer sites to confirm the cancer diagnoses. During review of the
cancer cases that arose during the intervention period, it became evident
that the cancer diagnosis was seldom different between the 2 reviewers. Thus,
for posttrial review, the decision was made to have 2 oncologists independently
review only cases with unknown primary sites reported by the Finnish Cancer
Registry and sites for which there was interest in more detailed data (eg,
stage for prostate cancer).
In this report, we include results for cancers of lung, prostate, urinary
tract (renal pelvis, ureter, and bladder), colon and rectum (excluding cancers
of anal canal), stomach, pancreas (excluding endocrine tumors), and kidney.
Other cancers (excluding nonmelanoma skin cancer) are combined.
In situ carcinomas were included among the urinary tract cancers (8
posttrial cases) but not for other organ sites (14 cases). Carcinoid tumors
are excluded because of the difficulty in ascertaining their malignancy (16
posttrial cases). The 16 postintervention cases were distributed into trial
groups as follows: 6 cases were in the α-tocopherol-only group, 4 in
the β-carotene-only group, 3 in the group receiving α-tocopherol
plus β-carotene, and 3 in the placebo group. A lung carcinoid tumor from
the intervention period in the α-tocopherol plus β-carotene group
was excluded from the analysis herein but was included in a prior report.6 A previously reported stomach carcinoid tumor in the α-tocopherol-only
group from the intervention period was also excluded.14
Deaths were identified from the Register of Causes of Death. Specific
causes were derived from the official underlying cause of death.
The present numbers of lung, prostate, and stomach cancer cases during
intervention differ slightly from those reported earlier6,7,14 because
we excluded carcinoid tumors and identified some new cases diagnosed during
the intervention period. Since the original reports,6,7 2
cases of lung cancer (1 individual diagnosed during the intervention period
in the α-tocopherol-only group and 1 in the placebo group) and 2 cases
of prostate cancer (both diagnosed during the intervention period and had
received β-carotene only) have been reported. Prior publications6,14 included 1 case of a stomach carcinoid
tumor from the intervention period in the α-tocopherol-only group and
a lung carcinoid tumor from the intervention period in the α-tocopherol
plus β-carotene group. Both cases were excluded from the present analysis.
In the analysis of cancer incidence, follow-up continued from the date
of randomization until the first occurrence of a specific cancer, death, or
the administratively defined end of follow-up (ie, April 30, 1999). In the
analyses of cause-specific and total mortality, follow-up continued until
death or the end of follow-up (April 30, 1999, for cause-specific mortality
and April 30, 2001, for total mortality). In all analyses, censoring was assumed
to be independent of the end point. Some participants had more than 1 type
of tumor. All cases were reviewed centrally by 1 or 2 ATBC Study physicians
who confirmed that the individual had more than 1 malignancy. A participant
was counted only once in the cancer-specific analyses (ie, only the first
occurrence of an organ-specific cancer was included), but if he had 2 different
organ-specific cancers, he was included in both analyses (eg, in the analyses
of both lung and prostate cancer).
We divided the follow-up period into 4 intervals to demonstrate temporal
changes in the effects of α-tocopherol and β-carotene: (1) trial
period (April 1985-April 30, 1993); (2) posttrial period 1 (May 1, 1993-April
30, 1996); (3) posttrial period 2 (May 1, 1996-April 30, 1999); and (4) posttrial
period 3 for the analysis of total mortality only (May 1, 1999-April 30, 2001).
Within these periods, crude rates per 10 000 person-years were calculated
in each of the 4 groups and according to the 2 × 2 factorial design.
Crude relative risk (RR) point estimates and their 95% CIs for these periods
were obtained using a Poisson regression model.16 Statistical
analyses were performed using S-PLUS version 3.4 for Unix (MathSoft Inc, Seattle,
To estimate the calendar time-specific RRs for lung cancer incidence
and total mortality, we calculated smoothed RR estimates and their 95% pointwise
CIs using a generalized additive model.17 We
first divided calendar time into monthly intervals with the exception that
we combined calendar time until April 1986 for the first interval because
the risk sets were small at this earliest phase of the recruitment period
that started in 1985. The monthly rates were treated as Poisson responses.
For each target time point, 40% of all monthly observations nearest to the
target were used to define a neighborhood for which a weighted linear curve
was used to estimate the RR at the target point. The weights for the monthly
observations around the target point were calculated from a tricube kernel
centered at the target point. The 40% neighborhood was chosen after the examination
of smoothed curves with different values because fluctuation of RRs for the
lower values made the overall interpretation difficult and a great deal of
detail was lost for the higher values.
Cause-specific data on deaths are presented in mutually exclusive cause-of-death
categories based on the following International Classification
of Diseases, Eighth Revision (ICD-8), International
Classification of Diseases, Ninth Revision (ICD-9), and International Classification of Diseases, 10th Revision (ICD-10) codes:
lung cancer (ICD-8 and ICD-9 162; ICD-10 C33-C34); other cancer (ICD-8 140-161 and 163-207; ICD-9 140-161 and 163-208; ICD-10 C00-C32 and C37-C97); ischemic heart disease (ICD-8 and ICD-9 410-414; ICD-10 I20-I25); hemorrhagic stroke (ICD-8 430-431; ICD-9 430-432; ICD-10 I60-I62); nonhemorrhagic cerebrovascular disease (ICD-8 432-438; ICD-9 433-438; ICD-10 I63-I69); other cardiovascular disease (ICD-8 390-404, 420-429, and 440-458; ICD-9 390-405,
415-429, and 440-459; ICD-10 I00-I15, I26-I52, and
I70-I99); respiratory disease (ICD-8 and ICD-9 460-519; ICD-10 J00-J99); and other
causes (ICD-8 000-139, 208-389, and 520-999; ICD-9 001-139, 210-389, and 520-999; ICD-10 A00-B99, D00-H95, and K00-Y98).
Of the original randomized ATBC Study cohort, 25 563 men were still
alive at the beginning of the posttrial follow-up in May 1993. At that time,
their average age was 63.5 years, and similar across the 4 supplementation
groups (range of means, 63.4-63.6 years). At the beginning of the trial in
1985-1988, they reported smoking an average of 20.4 cigarettes daily and having
smoked for 35.5 years. Of the individuals in posttrial follow-up, 79% had
their last trial follow-up visit in the winter of 1992-1993. At this visit,
75% were still current smokers (range, 74.6%-75.4% across the 4 study groups)
who smoked a mean of 18.1 cigarettes daily (range, 17.9-18.2 cigarettes).
Similarly, 4.8% reported taking nontrial supplements containing vitamin E
and 0.6% reported taking supplements containing β-carotene. The average
daily dose from supplement intake was 20 mg for vitamin E and 7 mg for β-carotene.
Incidences and RRs of site-specific cancers in the 4 study groups for
the trial period are shown in Table 1 and
for the posttrial periods 1 and 2 are shown in Table 2. Table 3 presents
the respective RRs according to supplementation with α-tocopherol and β-carotene.
During the 6-year posttrial period, 1037 incident lung cancer cases
were ascertained among the 25 283 participants who had no lung cancer
diagnosed when the posttrial period began. No significant overall difference
in lung cancer incidence was observed during the posttrial period between α-tocopherol
recipients and nonrecipients (RR, 1.03; 95% CI, 0.91-1.16) or between β-carotene
recipients and nonrecipients (RR, 1.06; 95% CI, 0.94-1.20). Although not statistically
significant, α-tocopherol appeared to reduce lung cancer risk slightly
during the first 3 posttrial years, whereas it appeared to increase the risk
somewhat during the later years (Table 3). The elevated risk of lung cancer in the β-carotene group
observed during the trial continued during the first posttrial period, although
statistically nonsignificant (Table 3).
The smoothed calendar time-specific RRs for the α-tocopherol recipients
compared with nonrecipients were below 1.0 for the last years of the intervention
period and the first posttrial years, but increased somewhat with longer follow-up
(Figure 1A). For the β-carotene
recipients compared with nonrecipients, the RRs increased during the intervention
but declined thereafter, falling below 1.0 approximately 4 years posttrial
There were 672 incident prostate cancer cases during the 6-year posttrial
follow-up among the 25 390 participants who had no prostate cancer diagnosed
by May 1, 1993. No statistically significant difference in prostate cancer
incidence was observed between α-tocopherol recipients and nonrecipients,
although reduced risk was suggested (RR, 0.88; 95% CI, 0.76-1.03). β-carotene
showed no effect postintervention (RR, 1.06; 95% CI, 0.91-1.23).
The incidences of other cancers were low compared with those of lung
cancer and prostate cancer, and thus their RR estimates were relatively imprecise
(Table 1, Table 2, and Table 3).
The risk of colorectal cancer during the follow-up was elevated for β-carotene
recipients compared with nonrecipients (RR, 1.44; 95% CI, 1.09-1.90), but
the elevation occurred only during the second posttrial period (Table 3).
Of the 25 563 participants still alive at the beginning of the
posttrial follow-up (May 1, 1993), 7261 (28%) died by April 30, 2001. Table 4 shows the rates and RRs of mortality
in the 4 regimen groups both for the trial period and for the 8-year posttrial
follow-up, divided into 3 periods (ie, 3 + 3 + 2 years). Table 5 presents the respective RRs according to supplementation
with α-tocopherol and β-carotene. During the 8-year posttrial follow-up,
the relative mortality was 1.01 (95% CI, 0.96-1.05) among α-tocopherol
recipients compared with nonrecipients and 1.07 (95% CI, 1.02-1.12) among β-carotene
recipients compared with nonrecipients.
The smoothed calendar time-specific relative mortality rate of α-tocopherol
recipients was similar to that of nonrecipients throughout the postintervention
period (Figure 2A ).
The higher mortality rate of the β-carotene recipients compared with
that of nonrecipients evident by the end of intervention returned toward the
null approximately 4 to 6 years later (Figure
A total of 5298 deaths were identified during the 6-year posttrial follow-up
from May 1993 to April 1999. Of these, 28.8% (n = 1524) were attributed to
ischemic heart disease, 17.8% (n = 942) to lung cancer, 17.1% (n = 905) to
other cancers, 8.2% (n = 433) to respiratory disease, 4.9% (n = 261) to nonhemorrhagic
cerebrovascular disease, 2.6% (n = 139) to hemorrhagic stroke, 7.0% (n = 372)
to other cardiovascular disease, and 13.6% (n = 722) to other causes. Figure 3 presents RR estimates by cause of
death in the 2 group comparisons separately for the trial period and for the
6-year posttrial follow-up. The excess mortality due to hemorrhagic stroke
observed for α-tocopherol recipients during the trial was also present
during the posttrial period (81 vs 58 deaths; RR, 1.40; 95% CI, 1.00-1.96).
However, of the 23 excess cases of fatal hemorrhagic stroke, 19 cases (83%)
appeared in the α-tocopherol plus β-carotene group and only 4 cases
(17%) in the α-tocopherol-only group. In addition, half of the excess
cases occurred only during the sixth posttrial year. Most of the excess deaths
in the β-carotene group were from cardiovascular disease.
The primary aim of the ATBC Study was to determine whether supplementation
with α-tocopherol or β-carotene would reduce the incidence of lung
cancer in male smokers. At the conclusion of the 6 year (median) intervention,
we observed no overall effect of α-tocopherol on lung cancer incidence,
whereas β-carotene increased the rate by 17%.6 This
increased risk appeared approximately 4 years after starting β-carotene
supplementation and is shown in the present analysis to disappear within a
similar timeframe postintervention. The CARET study used a combination of β-carotene
and vitamin A in smokers and asbestos-exposed workers and showed a time lag
of about 18 months to increased lung cancer incidence during its supplementation
period.8 These temporal effects suggest that β-carotene
in some way accelerated the progression and led to earlier clinical diagnosis
of more advanced latent lung tumors. Those effects also argue against a promotional
effect of β-carotene on earlier phases of lung carcinogenesis. Whether
the lower rates of lung cancer we observed after 4 years of stopping β-carotene
supplementation merely reflect fewer lung cancer diagnoses within the now
diminished pool of at-risk subclinical cases, or the possibility that β-carotene
may exert long-term preventive effects on earlier phases of lung carcinogenesis,
may become evident during extended follow-up of this and other trial cohorts.
The mechanisms by which β-carotene affects the development of lung
cancer have not yet been determined. Experimental data, however, suggest that β-carotene
does not induce genotoxic effects per se.18 The
most widely suggested hypothesis is that components of cigarette smoke in
the presence of the relatively high oxygen tension in the lung combine to
induce oxidation of β-carotene, resulting in a prooxidant effect.19 However, a study involving human bronchial epithelial
cells found no direct prooxidant effect of the smoke-induced β-carotene
Other mechanisms by which β-carotene–smoke interactions could
increase lung carcinogenesis have been recently reported. Ferrets given β-carotene
supplements (equivalent to 30 mg/day in humans) and exposed to cigarette smoke
had a strong proliferative response in their lung tissue in addition to diminished
retinoid signaling resulting from suppression of retinoic acid receptor β
gene expression and overexpression of activator protein 1.21 The
effect appeared dose-dependent because a physiological dose compared with
low intake of β-carotene (equivalent to 6 mg/day and 2.1 mg/day) had
no potentially detrimental effects and afforded weak protection against smoke-induced
lung damage.22 Paolini et al23,24 reported
that β-carotene in the rat lung produced a powerful booster effect on
several carcinogen-metabolizing P450 enzymes, and that this was associated
with the generation of oxidative stress. These effects may predispose to cancer
through bioactivation of tobacco smoke procarcinogens and oxidation of β-carotene
to a prooxidant.23,24 However,
these experiments require cautious interpretation because they are based on
exposure of nonmalignant bronchial epithelial cell lines or lung tissue, whereas
the findings from the ATBC Study and CARET suggest that β-carotene advances
the progression of lung cancer in a chronically high-risk organ with abnormal
cells and/or latent tumors already present.
The α-tocopherol supplementation in the ATBC Study reduced the
incidence of prostate cancer by 34%, with the preventive effect observed approximately
18 months after the start of intervention.7 This
effect was present throughout the 6-year posttrial follow-up evaluated herein,
but was substantially attenuated even within the first 3 posttrial years.
This suggests that α-tocopherol prevents the progression of prostate
cancer in a later phase of carcinogenesis, and that this effect is transient,
diminishing fairly rapidly following cessation of supplementation.
Although laboratory experiments point to several potential mechanisms
for a cancer preventive effect of α-tocopherol,25 observational
studies have provided only weak support for the vitamin E prostate cancer
hypothesis thus far.26 In the Heart Protection
Study, a randomized placebo-controlled trial involving more than 15 000
men aged 40 to 80 years, a 9% nonsignificant decrease in risk of prostate
cancer with a daily combination of vitamin E (600 mg), vitamin C (250 mg),
and β-carotene (20 mg) for 5 years was observed.27 The
Selenium and Vitamin E Cancer Prevention Trial (SELECT) was recently launched
to assess the effects of up to 12 years of supplementation with selenium and
vitamin E in the prevention of prostate cancer among 32 400 men in North
America.28 The modest but statistically nonsignificant
increase in risk of prostate cancer among β-carotene recipients in the
ATBC Study was reduced in the early posttrial period and absent in the later
posttrial years. These observations, together with the finding of no effect
of β-carotene in the CARET and Physicians' Health Study, collectively
reflecting over 1300 prostate cancer cases, suggest that β-carotene does
not affect the development of clinical prostate cancer.
Neither α-tocopherol nor β-carotene supplementation had significant
overall effects on other major cancers during the postintervention period,
with the exception of a late increase in colorectal cancer incidence in the β-carotene–supplemented
group. Although this may be a chance finding, an effect on early colorectal
carcinogenesis, including possible progression of adenomatous polyps, cannot
be excluded. It is also possible that late posttrial differences may potentially
be influenced by competing causes of death.
Total mortality during the intervention period was 8% higher among men
who received β-carotene compared with those who did not, and excess mortality
was also observed during the posttrial follow-up. The higher trial period
mortality was due to coronary heart disease and lung cancer, whereas during
the posttrial follow-up it was due primarily to cardiovascular causes, including
coronary heart disease, cardiomyopathy, hypertensive heart disease, stroke,
and aortic rupture. We do not know of a common mechanism that would explain
how β-carotene might increase mortality across this diverse spectrum
of cardiovascular diseases. Our assessment of the duration of increased mortality
following β-carotene supplementation suggests that the excess mortality
lasted for 4 to 6 years, roughly the equivalent time it took for the excess
to become evident following the initiation of the intervention.
Supplementation with α-tocopherol increased hemorrhagic stroke
mortality by 45% during the intervention. An elevated risk (40%) was observed
also in the posttrial period. Half of this excess occurred in the sixth posttrial
year, however, and the annual numbers of fatal hemorrhagic stroke cases were
small, approximately 20. We do not know of a hypothesis that would explain
such a late effect; thus, the posttrial finding is likely due to chance. The
possibility that α-tocopherol inhibits platelet function, which is consistent
with the increased risk of hemorrhagic stroke during the intervention period
and its disappearance shortly after stopping α-tocopherol supplementation,
however, cannot be excluded. On the other hand, in the Heart Protection Study
daily supplementation with a combination of vitamin E, vitamin C, and β-carotene
had no effect on the incidence of hemorrhagic stroke when compared with placebo
(51 vs 53 cases, respectively) but no data on fatal hemorrhagic stroke were
In conclusion, large-scale controlled trials have not produced consistent
evidence for the efficacy of α-tocopherol or β-carotene in the
prevention of cancer. There is, however, consistent evidence that β-carotene
supplementation in smokers increases the risk of lung cancer and total mortality.
The cumulative experience of nearly 16 years and nearly 350 000 person-years
of observation during the intervention and postintervention follow-up of participants
in the ATBC Study suggests a symmetry in the effect of β-carotene on
these events, with the disappearance of risk occurring within the time it
became evident. Furthermore, the posttrial follow-up did not reveal any late
preventive effects on cancer.
Thus, the recommendations made at the time our initial trial results
were reported remain appropriate: the possible preventive effect of α-tocopherol
on prostate cancer requires confirmation from other trials before public health
recommendations can be made for vitamin E. Also, β-carotene supplementation
should be avoided by smokers since it may be harmful to them.