Download Multivitamin Use and Colorectal Cancer Incidence in a US Cohort

American Journal of Epidemiology
Copyright © 2003 by the Johns Hopkins Bloomberg School of Public Health
All rights reserved
Vol. 158, No. 7
Printed in U.S.A.
DOI: 10.1093/aje/kwg190
Multivitamin Use and Colorectal Cancer Incidence in a US Cohort: Does Timing
Matter?
Eric J. Jacobs, Cari J. Connell, Ann Chao, Marjorie L. McCullough, Carmen Rodriguez,
Michael J. Thun, and Eugenia E. Calle
From the Department of Epidemiology and Surveillance Research, American Cancer Society, Atlanta, GA.
Received for publication October 28, 2002; accepted for publication April 3, 2003.
Multivitamins contain several nutrients, including folic acid, that are hypothesized to reduce the risk of colorectal
cancer. Previous studies suggest that multivitamin use may reduce colorectal cancer risk but only after a long
latency period. The authors examined the association between regular multivitamin use (four or more times per
week) and colorectal cancer incidence among 145,260 men and women in the Cancer Prevention Study II
Nutrition Cohort. Current multivitamin use was reported on a questionnaire at enrollment in 1992–1993. All
participants had also reported multivitamin use on a questionnaire completed for a different study approximately
10 years earlier (in 1982). The authors observed 797 incident cases of colorectal cancer during follow-up from
1992 to 1997. After multivariate adjustment, regular multivitamin use at enrollment was not associated with risk
of colorectal cancer (rate ratio = 1.04, 95% confidence interval: 0.87, 1.23), whereas regular multivitamin use 10
years before enrollment was associated with reduced risk (rate ratio = 0.71, 95% confidence interval: 0.57, 0.89).
Regular multivitamin users 10 years before enrollment were at similarly reduced risk whether they were still
regular multivitamin users at enrollment or had stopped. These results are consistent with the hypothesis that
past, but not recent, multivitamin use may be associated with modestly reduced risk of colorectal cancer.
cohort studies; colonic neoplasms; colorectal neoplasms; folic acid; vitamins
Abbreviations: CI, confidence interval; CPS-II, Cancer Prevention Study II; RR, rate ratio.
Multivitamins contain several vitamins that are hypothesized to reduce colorectal cancer risk, most notably folic
acid. In 1973, the legal limit on folic acid in multivitamins in
the United States was increased from 100 µg to 400 µg (1).
The folic acid content of different multivitamin brands
increased over time (based on a review of successive
editions of the Physicians’ Desk Reference starting in 1973),
but 1985 was the first year that the majority of US multivitamin brands listed in the Physicians’ Desk Reference
included 400 µg of folic acid (2, 3). Analyses of blood
samples collected in 1982 from Physicians’ Health Study
participants showed that multivitamin users had over twice
the level of serum folate as nonusers (4). Daily use of a
multivitamin containing 400 µg of folic acid would have
more than doubled the folate intake of most US residents
(mean dietary intake, 283 µg) (5), at least until folic acid
fortification of grain started in 1996–1998 (6).
Evidence from both laboratory (7) and epidemiologic (8–
10) studies suggests that folic acid may inhibit colorectal
carcinogenesis, although epidemiologic results have not
been entirely consistent. Folic acid prevents colon carcinogenesis in several rodent models (7). In a mouse model, folic
acid supplementation was effective in preventing colonic
adenomas if started before the establishment of neoplastic
foci, whereas supplementation after the establishment of
neoplastic foci had no effect, indicating that folic acid
inhibits only early stages of colorectal carcinogenesis (11). If
folic acid also selectively inhibits early stages of colorectal
carcinogenesis in humans, we might expect a long latency
period between folic acid supplementation and any reduction
in risk of incident colorectal cancer.
In addition to folic acid, multivitamins typically contain
the US Recommended Daily Allowance of several other
vitamins, including vitamin D, which has been associated
with decreased risk of colorectal cancer in some epidemio-
Correspondence to Dr. Eric J. Jacobs, Epidemiology and Surveillance Research, American Cancer Society, National Home Office, 1599 Clifton
Road NE, Atlanta, GA 30329-4251 (e-mail: [email protected]).
621
Am J Epidemiol 2003;158:621–628
622 Jacobs et al.
logic studies (12). It should be noted that associations
between multivitamin use and disease outcomes cannot be
definitively attributed to any single nutrient.
The association between multivitamin use and colon
cancer incidence or mortality has been examined in three
cohort studies (13–15) and one case-control study (16). All
four studies found reduced risk of colon cancer with either a
long interval since first use or a long duration of use. In the
Nurses’ Health Study cohort, use of a multivitamin potentially containing 400 µg of folic acid (excluding multivitamin use before 1973) was associated with strongly reduced
risk of colon cancer incidence, but only after 15 or more
years since first use (rate ratio (RR) = 0.25, 95 percent confidence interval (CI): 0.13, 0.51) (14). A similar analysis of
colon cancer mortality in the Cancer Prevention Study II
(CPS-II) cohort found weakly reduced risk after 15 or more
years since first use of a multivitamin potentially containing
400 µg of folic acid (RR = 0.89, 95 percent CI: 0.80, 0.99)
(15). The reductions in risk associated with a long interval
since first use observed in both the Nurses’ Health Study and
the CPS-II mortality study are consistent with a long latency
period between multivitamin use and risk of colorectal
cancer diagnosis. However, because time since first use is
likely to be strongly correlated with duration of use, the
reduced risk observed in these studies could be due solely to
longer duration of use. In the Health Professionals’ Followup Study, multivitamin use of 10 or more years’ duration at
enrollment in 1986 was associated with a rate ratio of 0.74
(95 percent CI: 0.47, 1.17) for colon cancer incidence (13).
A case-control study of colon cancer incidence in the Seattle,
Washington, area examined average daily multivitamin
intake over a 10-year period ending 2 years before the diagnosis date (16). Use of one or more multivitamin tablets per
day during this period was associated with substantially
reduced risk of colon cancer incidence (RR = 0.51, 95
percent CI: 0.34, 0.77).
We examined the association between multivitamin use
and colorectal cancer incidence in the CPS-II Nutrition
Cohort, a large cohort of predominantly elderly US adults.
Participants in this cohort had reported multivitamin use at
two time points, at study enrollment in 1992–1993 and
approximately 10 years earlier (in 1982), which enabled us
to examine the role of timing of multivitamin use.
MATERIALS AND METHODS
Study cohort and follow-up
Men and women in this analysis were participants in the
CPS-II Nutrition Cohort (hereafter referred to simply as the
Nutrition Cohort), a prospective study of cancer incidence in
approximately 184,000 US adults (17). The Nutrition
Cohort, established in 1992, is a subgroup of the approximately 1.2 million participants in the CPS-II cohort, a
prospective study of cancer mortality established in 1982
(18). The CPS-II cohort was recruited by American Cancer
Society volunteers in all 50 US states, the District of
Columbia, and Puerto Rico and included men and women
aged 30 or more years at enrollment in 1982. Nutrition
Cohort participants were recruited from CPS-II cohort
participants who resided in 21 states with population-based
state cancer registries. The recruitment and characteristics of
Nutrition Cohort participants are described in detail elsewhere (17). All aspects of the CPS-II Nutrition Cohort study
are approved by the Emory University Institutional Review
Board.
At enrollment in 1992 or 1993, Nutrition Cohort participants completed a self-administered questionnaire that
included information on demographic characteristics,
medical history, and various behavioral, occupational, and
dietary factors. Usual dietary intake was assessed using a
semiquantitative 68-item food frequency questionnaire,
which is a modification of the brief Health Habits and
History Questionnaire developed by Block et al. (19).
Nutrient intakes were estimated using Diet Analysis System
version 3.8a (20), which did not account for the subsequent
fortification of the grain supply with folic acid (6). Exercise
level (measured in metabolic equivalent hours per week) was
calculated by multiplying the number of hours spent at each
of seven leisure-time physical activities by an estimated
intensity level for each activity (21) and summing across
activities. Additional risk factor information from approximately 10 years earlier, including information on multivitamin use, was available from the four-page selfadministered questionnaire participants completed in 1982 at
enrollment in the earlier CPS-II cohort.
In 1997, a follow-up questionnaire was sent to all Nutrition Cohort members to update information and to ascertain
newly diagnosed cancers. Information on use of sigmoidoscopy and colonoscopy was collected for the first time on this
questionnaire. Among living Nutrition Cohort participants,
the response rate for the 1997 follow-up questionnaire was
approximately 91 percent (17).
From the overall Nutrition Cohort population of 184,190
we excluded from this analysis participants who reported at
enrollment a history of cancer, other than nonmelanoma skin
cancer (n = 23,630), did not respond to the 1997 follow-up
questionnaire (unless known to be deceased) (n = 13,227),
did not have complete information on multivitamin use (n =
1,820), or had self-reported colorectal cancers that could not
be verified by obtaining medical records or by linkage with
state registries (n = 253). Analyses include the remaining
145,260 participants (68,934 men and 76,326 women).
This analysis includes 797 incident cases of colorectal
cancer diagnosed between enrollment in 1992 or 1993 and
August 31, 1997. Most incident cases of colorectal cancer
were identified initially through a self-report of cancer on the
1997 follow-up questionnaire (n = 595). Previous pilot work
linking self-reports of cancer from the 1997–1998 questionnaire with information from four state cancer registries indicated that the ability of our participants to self-report an
incident cancer is high (sensitivity = 0.93) (22). Self-reports
of cancer were verified by obtaining medical records (n =
463), or if medical records could not be obtained, through
linkage with state cancer registries (n = 132). An additional
202 cases of colorectal cancer were identified through automated linkage of all cohort members with the National Death
Index (23), which was current through December 31, 1998,
at the time of linkage. Participants who died between enrollment and August 31, 1997, and whose death certificates
Am J Epidemiol 2003;158:621–628
Multivitamins and Colorectal Cancer 623
listed colon or rectal cancer as a primary or contributory
cause of death were categorized as cases of colorectal
cancer. Colorectal tumors known to be not adenocarcinomas
(e.g., carcinoid tumors) were not considered cases in this
analysis because these tumors comprise only a very small
proportion of colorectal cancers and may have a different
etiology than adenocarcinomas.
Ascertainment of use of multivitamins and other dietary
supplements
The 1992–1993 Nutrition Cohort questionnaire asked
about the use of six supplements (multivitamins, vitamin A,
beta-carotene, vitamin C, vitamin E, and calcium) during the
past year. For each supplement they had used at least once a
week, participants were asked to report the number of tablets
taken. Response categories were 1–3 per week, 4–6 per
week, one per day, two per day, three per day, four per day,
or five or more per day. No information was collected on
duration of supplement use.
Because all Nutrition Cohort participants were also participants in the original CPS-II cohort, information on past
multivitamin use was available from the 1982 CPS-II questionnaire. The 1982 CPS-II questionnaire asked about
current use of four supplements (multivitamins, vitamin A,
vitamin C, and vitamin E). For each supplement, participants
were asked to report the number of times used in the last
month and the number of years of use. Participants who used
a vitamin only “occasionally” were instructed to report times
per month as “1/2.”
Statistical analysis
We used proportional hazards modeling to examine the
association between multivitamin use and colorectal cancer
incidence while adjusting for other potential risk factors
(24). We used follow-up time since enrollment in 1992–
1993 as the time-axis. We created variables for “recent”
multivitamin use based on responses to the 1992–1993
Nutrition Cohort questionnaire. Participants who reported
taking four or more multivitamin tablets per week were categorized as regular users, while participants who reported
taking 1–3 multivitamin tablets per week were categorized
as occasional users. We created separate variables for “past”
multivitamin use based on responses to the 1982 CPS-II
questionnaire. Participants reporting multivitamin use 16 or
more times per month (equivalent to four times per week in
a 4-week month) were categorized as regular users, while
participants reporting unquantified “occasional” use, or use
1–15 times per month, were categorized as occasional users.
Although the category of occasional users included participants reporting multivitamin use up to 15 times per month,
only 20 percent of these occasional users reported use more
than four times per month.
We used two approaches to model the association between
multivitamin use and colorectal cancer incidence. The first
modeling approach simultaneously included variables for
past multivitamin use (defined as multivitamin use in 1982)
and recent multivitamin use (defined as use at enrollment in
1992–1993), thereby adjusting risk estimates for past multiAm J Epidemiol 2003;158:621–628
vitamin use for recent multivitamin use and vice versa. The
second modeling approach categorized participants into five
mutually exclusive groups on the basis of combinations of
past and recent multivitamin use: 1) no multivitamin use in
either 1982 or 1992–1993, 2) occasional use only in 1982
and/or 1992–1993, 3) regular multivitamin use in 1982 but
no use or occasional use only in 1992–1993, 4) regular
multivitamin use in 1992–1993 but no use or occasional use
only in 1982, and 5) regular use in both 1982 and 1992–
1993.
We could not examine lifetime duration of multivitamin
use because the 1992–1993 Nutrition Cohort questionnaire
did not ask about duration of use. Lifetime duration of multivitamin use may be of limited relevance because multivitamins were not permitted to contain more than 100 µg of folic
acid before 1973 (1).
All proportional hazards models were adjusted for age,
sex, educational level, body mass index, exercise level,
intake of saturated fat and dietary fiber, and use of vitamin C
and calcium supplements. All covariates, except age and
vitamin C use, were modeled as dummy variables using the
exact categories shown in table 1, including the category
“unclassifiable” for some covariates. We adjusted for age
using the stratified Cox procedure with 1-year age strata
(25). We adjusted for vitamin C use both as reported in 1982
(none, occasional, regular, unknown) and as reported in
1992–1993 (none, occasional, regular, unknown). The variable for educational level was based on information from the
1982 questionnaire. All other covariates were based on
information from the 1992–1993 questionnaire.
We also examined potential confounding by race, cigarette
smoking, alcohol use, aspirin use, postmenopausal hormone
replacement therapy, vitamin E supplement use, dietary
intake of calcium, methionine, vitamin D, and folate, and
consumption of red meat, fruits, vegetables, whole grains,
and total energy. However, we did not adjust for these
factors in the final models because such adjustment had
negligible effects on results.
We examined whether the association between regular
multivitamin use and colorectal cancer incidence varied by
intake of dietary folate, dietary methionine, and alcohol,
which might be expected to modify the effect of folic acid
from multivitamins (8). Specifically, we modeled interaction
terms between either past multivitamin use (use in 1982) or
recent multivitamin use (use at enrollment in 1992) and a
dichotomous term for dietary folate (median split), dietary
methionine (median split), or alcohol (<2 drinks/day vs. ≥2
drinks per day). p values for heterogeneity of rate ratios were
calculated using the likelihood ratio statistic (26).
Because folic acid may be the multivitamin component
that reduces colorectal cancer risk, we also examined the
association between colorectal cancer and total folate intake
(from diet and multivitamins combined) at enrollment in
1992–1993. Total folate was calculated as the sum of dietary
folate (adjusted for total energy using the residuals method
(27)) and folate from multivitamins (400 µg per tablet). We
could not examine total folate intake 10 years before enrollment, because the 1982 CPS-II questionnaire did not include
sufficiently detailed dietary questions to calculate dietary
folate intake.
624 Jacobs et al.
TABLE 1. Selected demographic and colorectal cancer risk factors by multivitamin use, Cancer Prevention Study II Nutrition Cohort,
1992–1997*
Multivitamin use
Men (%)
Women (%)
Past use but Recent use but Both past and
No use†
not recent use‡ not past use§ recent use¶
(n = 38,924)
(n = 11,492)
(n = 8,422)
(n = 4,688)
Past use but Recent use but Both past and
No use
recent use
not recent use not past use
(n = 33,772)
(n = 15,606) (n = 12,461)
(n = 6,495)
Age at enrollment (years)
<50
0.1
0.0
0.0
0.1
1.9
2.9
2.0
2.1
50–59
25.6
24.0
23.7
22.2
33.5
34.5
34.2
30.4
60–69
58.4
59.5
57.6
57.7
49.4
48.5
48.4
51.0
70–79
15.3
16.0
17.9
19.1
15.2
14.0
15.4
16.5
0.7
0.5
0.8
0.9
0.1
0.1
0.1
0.1
White
97.5
97.9
97.4
97.9
97.3
97.5
97.3
98.1
Black
1.2
0.9
1.1
0.7
1.6
1.3
1.5
0.8
Other
1.3
1.2
1.5
1.4
1.1
1.2
1.2
1.1
≥80
Race
Education
Less than high school
9.3
7.4
7.6
4.3
6.0
4.2
4.7
3.2
High school graduate
21.3
16.6
17.5
14.0
34.9
30.3
30.9
27.6
Some college
25.4
26.8
26.7
24.6
29.2
33.5
32.0
33.0
College graduate
20.3
22.7
21.8
25.7
17.9
18.8
19.0
20.8
Graduate school
22.9
25.9
25.7
30.8
11.3
12.7
12.7
14.9
0.7
0.7
0.6
0.5
0.7
0.6
0.7
0.5
Unclassifiable
Body mass index (kg/m2)
<22.5
9.9
11.8
12.2
14.0
25.3
28.8
27.9
32.8
22.5–<25
23.3
26.4
26.0
28.8
24.2
24.5
25.0
25.6
25.0–<27.5
31.5
31.0
30.4
30.3
20.8
19.4
20.8
18.8
27.5–<30
18.5
17.3
17.1
15.4
11.1
10.6
10.3
9.5
≥30.0
15.4
12.3
12.9
10.4
16.9
15.1
14.5
12.1
1.5
1.1
1.4
1.1
1.7
1.6
1.4
1.2
<3.5
13.5
10.4
10.8
8.3
10.4
8.4
8.1
7.4
3.5–<4.5
27.4
27.1
25.9
25.0
32.1
30.0
28.9
27.5
4.5–<14.0
14.5
16.3
16.2
18.2
17.8
19.4
20.2
21.6
14.0–<24.5
22.4
23.8
24.0
24.5
23.0
24.6
24.8
25.3
≥24.5
20.5
21.2
22.0
23.0
14.9
16.0
16.5
17.2
1.7
1.1
1.3
1.0
1.6
1.6
1.5
Unclassifiable
Exercise (metabolic equivalents/week)
Unclassifiable
1.0
Table continues
RESULTS
Approximately 50 percent of participants included in this
analysis reported no multivitamin use, while 8 percent
reported regular multivitamin use only in the past (in 1982),
19 percent reported regular multivitamin use only recently
(at enrollment in 1992–1993), and 14 percent reported
regular multivitamin use both in the past and recently. An
additional 9 percent of participants reported only occasional
multivitamin use. Table 1 compares these groups with
respect to potential colorectal cancer risk factors, omitting
results for participants who reported only occasional multivitamin use. Most participants in this cohort, regardless of
multivitamin use, were White and middle-aged or elderly.
All categories of regular multivitamin users were much more
likely to take vitamin C and calcium supplements than were
nonusers. In addition, all categories of regular multivitamin
users were more likely than nonusers to be highly educated,
to have a lower body mass index, to exercise, and to
consume less saturated fat and more fiber.
Table 2 shows the association between incident colon,
rectal, and colorectal cancer and past regular multivitamin
use (use in 1982) and recent regular multivitamin use (use at
enrollment in 1992–1993), adjusted for multiple covariates.
This analysis examines the importance of timing of multivitamin use by comparing the separate effects of past and
recent multivitamin use. Past regular multivitamin use was
associated with decreased risk of colorectal cancer (RR =
0.71, 95 percent CI: 0.57, 0.89), whereas recent regular
multivitamin use was not (RR = 1.04, 95 percent CI: 0.87,
1.23). Results were similar when past and recent multivitamin use was modeled separately, without adjusting for
Am J Epidemiol 2003;158:621–628
Multivitamins and Colorectal Cancer 625
TABLE 1. Continued
Multivitamin use
Men (%)
Women (%)
Past use but Recent use but Both past and
No use†
not recent use‡ not past use§ recent use¶
(n = 38,924)
(n = 11,492)
(n = 8,422)
(n = 4,688)
Past use but Recent use but Both past and
No use
recent use
not recent use not past use
(n = 33,772)
(n = 15,606) (n = 12,461)
(n = 6,495)
Saturated fat (% of total energy)#
Quintile 1 (low)
18.2
22.6
24.6
27.9
18.4
22.1
24.8
28.6
Quintile 2
17.4
18.5
19.0
20.3
17.6
19.6
19.9
20.9
Quintile 3
20.6
20.9
20.3
20.2
9.9
10.6
10.6
11.0
Quintile 4
10.1
8.9
8.8
8.1
19.8
19.0
17.5
16.8
Quintile 5 (high)
23.5
19.6
18.0
16.3
24.4
19.2
18.7
16.5
Unclassifiable
10.2
9.4
9.2
7.2
10.0
9.4
8.6
6.1
Quintile 1 (low)
20.5
15.9
16.1
14.2
20.7
15.9
16.7
14.2
Quintile 2
18.2
16.7
16.7
16.0
19.2
18.1
17.2
16.9
Quintile 3
18.0
17.5
17.6
18.2
17.5
17.4
17.9
18.3
Quintile 4
17.2
19.6
19.1
20.8
17.2
19.2
19.3
20.8
Quintile 5 (high)
16.0
20.9
21.3
23.6
15.4
20.0
20.3
23.6
Unclassifiable
10.2
9.4
9.2
7.2
10.0
9.4
8.6
6.1
82.8
46.6
51.2
32.7
79.1
45.5
51.9
32.9
6.8
24.2
9.5
10.6
8.3
22.4
10.7
12.0
10.0
28.3
38.8
56.3
12.0
31.3
36.8
54.3
0.4
0.8
0.5
0.4
0.6
0.8
0.6
0.8
Fiber (g/day)**
Vitamin C supplements
Never
Past
Current
Unclassifiable
Calcium supplements
Nonuser
96.3
86.8
81.6
74.5
80.6
63.1
49.5
39.8
Use at least weekly
3.0
11.2
14.1
21.8
17.3
33.7
45.1
55.6
Unclassifiable
0.7
1.9
4.3
3.7
2.1
3.2
5.5
4.6
* Percentages adjusted to the age distribution of the entire study population. Table does not include participants who reported only occasional multivitamin use.
† No use reported either 10 years before enrollment (in 1982) or at enrollment (in 1992–1993).
‡ Regular use (≥4 times per week) 10 years before enrollment (in 1982) but no use, or occasional use only, at enrollment (in 1992–1993).
§ Regular use (≥4 times per week) at enrollment (in 1992–1993) but no use, or occasional use only, 10 years before enrollment (in 1982).
¶ Regular use (≥4 times per week) both 10 years before enrollment (in 1982) and at enrollment (in 1992–1993).
# Quintiles for men are <0.09, 0.09–<0.11, 0.11–<0.13, 0.13–<0.14, and ≥0.14. Quintiles for women are <0.08, 0.08–<0.10, 0.10–<0.11, 0.11–<0.13, and ≥0.13.
** Quintiles for men are <9.3, 9.3–<11.4, 11.4–<13.6, 13.6–<16.6, and ≥16.6. Quintiles for women are <8.0, 8.0–<9.9, 9.9–<11.8, 11.8–<14.4, and ≥14.4.
Adjusted for total energy.
multivitamin use at the other time point (RR = 0.72, 95
percent CI: 0.58, 0.89 for past regular multivitamin use; RR
= 0.95, 95 percent CI: 0.80, 1.12 for recent regular multivitamin use).
The association between regular multivitamin use and
colorectal cancer incidence was similar in men and women.
The adjusted rate ratio for past regular multivitamin use was
0.67 (95 percent CI: 0.50, 0.92) in men and 0.75 (95 percent
CI: 0.54, 1.03) in women. The adjusted rate ratio for recent
regular multivitamin use was 1.02 (95 percent CI: 0.81, 1.28)
in men and 1.06 (95 percent CI: 0.81, 1.39) in women.
The association with past regular multivitamin use
appeared to be similar for colon and rectal cancers (table 2).
Recent regular multivitamin use was not associated with
colon cancer (RR = 0.92, 95 percent CI: 0.74, 1.14) but was
associated with marginally increased risk of rectal cancer
(RR = 1.36, 95 percent CI: 1.00, 1.85), although confidence
intervals were wide because of small numbers of rectal
cancers. We also examined results by subsite within the
Am J Epidemiol 2003;158:621–628
colon, after excluding 67 cases of colon cancer with an
unknown or an overlapping subsite. Adjusted rate ratios for
past regular multivitamin use were 0.61 (95 percent CI: 0.42,
0.89) for proximal colon cancer (cecum to splenic flexure)
and 0.89 (95 percent CI: 0.58, 1.36) for distal colon cancer
(descending and sigmoid colon). Adjusted rate ratios for
recent regular multivitamin use were 0.96 (95 percent CI:
0.72, 1.28) for proximal colon cancer and 0.90 (95 percent
CI: 0.63, 1.29) for distal colon cancer.
We found no evidence that the association between regular
multivitamin use (either past or recent) and colorectal cancer
incidence varied by alcohol use or intake of dietary folate or
dietary methionine.
Analyses in this report exclude the 253 participants whose
self-report of colorectal cancer could not be verified
(Materials and Methods). Results were similar when we
included these 253 participants as cases of colorectal cancer
(for past regular multivitamin use, RR = 0.76, 95 percent CI:
626 Jacobs et al.
TABLE 2. Rate ratios and 95% confidence intervals for incident colorectal cancer associated with past
and recent multivitamin use, Cancer Prevention Study II Nutrition Cohort, 1992–1997*
Colon
Rectal
Colorectal
Past multivitamin use (use in 1982, 10 years before
enrollment)
No use
RR† (95% CI†)
1.00 (referent)
1.00 (referent)
1.00 (referent)
Cases (no.)/participants (no.)
418/94,117
165/94,117
584/94,117
Occasional use‡
RR (95% CI)
0.83 (0.62, 1.11) 0.93 (0.61, 1.43) 0.86 (0.68, 1.09)
Cases (no.)/participants (no.)
60/19,077
29/19,077
89/19,077
Regular use (≥4 times per week)
RR (95% CI)
0.72 (0.55, 0.94) 0.69 (0.46, 1.04) 0.71 (0.57, 0.89)
Cases (no.)/participants (no.)
86/32,066
38/32,066
124/32,066
RR (95% CI)
1.00 (referent)
1.00 (referent)
1.00 (referent)
Cases (no.)/participants (no.)
400/92,092
147/92,092
548/92,092
Recent multivitamin use (use in 1992–1993, at enrollment)
No use
Occasional use§
RR (95% CI)
0.82 (0.47, 1.43) 0.47 (0.15, 1.50) 0.73 (0.44, 1.19)
Cases (no.)/participants (no.)
14/5,187
3/5,187
17/5,187
Regular use (≥4 times per week)
RR (95% CI)
0.92 (0.74, 1.14) 1.36 (1.00, 1.85) 1.04 (0.87, 1.23)
Cases (no.)/participants (no.)
150/47,981
82/47,981
232/47,981
* Model simultaneously includes variables for past and recent multivitamin use. Adjusted for age, sex, education,
body mass index, exercise level, saturated fat intake, fiber intake, and use of vitamin C and calcium supplements.
† RR, rate ratio; CI, confidence interval.
‡ Reported unquantified “occasional” use or use 1–15 times per month in 1982.
§ Reported use 1–3 times per week at enrollment in 1992–1993.
0.63, 0.92; for recent regular multivitamin use, RR = 0.96,
95 percent CI: 0.82, 1.12).
Table 3 shows the association between incident colon,
rectal, and colorectal cancer and combinations of past
regular multivitamin use (use in 1982) and recent regular
multivitamin use (use at enrollment in 1992–1993), adjusted
for multiple covariates. The reduction in risk of colorectal
cancer appeared similar for participants who reported
regular multivitamin use only in the past (RR = 0.74, 95
percent CI: 0.54, 1.02) and for those who reported both past
and recent regular multivitamin use (RR = 0.73, 95 percent
CI: 0.56, 0.94), a group which likely included many long
duration multivitamin users. No reduction in colorectal
cancer risk was observed for participants who reported only
recent regular multivitamin use (RR = 1.03, 95 percent CI:
0.85, 1.25).
We also examined the association between duration of
multivitamin use reported on the 1982 CPS-II questionnaire
and colorectal cancer incidence, although it was not until
1985 that the majority of US multivitamin brands listed in
the Physicians’ Desk Reference contained 400 µg of folic
acid (2, 3). The risk of colorectal cancer appeared similar for
participants who reported 10 or more years of multivitamin
use in 1982 and for participants who reported less than 10
years of multivitamin use in 1982 (results not shown), as was
true in our previous analysis of colon cancer mortality in the
larger CPS-II cohort (15).
After adjustment for covariates, total folate intake (from
diet and multivitamins) at enrollment in 1992–1993 was not
associated with colorectal cancer risk. The adjusted rate ratio
for the highest versus the lowest quintile of total folate was
1.17 (95 percent CI: 0.89, 1.55). Total folate intake was
strongly influenced by intake of folic acid from multivitamins; 99 percent of participants in the highest quintile of
total folate were regular multivitamin users, compared with
none of the participants in the lowest quintile. Because many
multivitamin users in 1992–1993 were not past multivitamin
users, total folate intake in 1992–1993 may correlate only
weakly with past folate intake, which may be more important
in colorectal cancer etiology.
DISCUSSION
In this large prospective study, past multivitamin use was
associated with a decreased risk of colorectal cancer,
whereas recent multivitamin use was not. These results are
consistent with the hypothesis that multivitamin use may
reduce the risk of colorectal cancer but only after a long
latency period.
Am J Epidemiol 2003;158:621–628
Multivitamins and Colorectal Cancer 627
TABLE 3. Rate ratios and 95% confidence intervals for incident colorectal cancer associated with
combinations of past and recent multivitamin use, Cancer Prevention Study II Nutrition Cohort, 1992–
1997*
Colon
Rectal
Colorectal
RR‡ (95% CI‡)
1.00 (referent)
1.00 (referent)
1.00 (referent)
Cases (no.)/participants (no.)
335/72,696
121/72,696
457/72,696
RR (95% CI)
0.85 (0.62, 1.18)
0.96 (0.58, 1.60)
0.88 (0.67, 1.15)
Cases (no.)/participants (no.)
45/13,400
18/13,400
63/13,400
RR (95% CI)
0.75 (0.52, 1.08)
0.71 (0.38, 1.34)
0.74 (0.54, 1.02)
Cases (no.)/participants (no.)
34/11,183
11/11,183
45/11,183
RR (95% CI)
0.90 (0.71, 1.14)
1.39 (1.00, 1.95)
1.03 (0.85, 1.25)
Cases (no.)/participants (no.)
98/27,098
55/27,098
153/27,098
RR (95% CI)
0.65 (0.47, 0.89)
0.95 (0.60, 1.51)
0.73 (0.56, 0.94)
Cases (no.)/participants (no.)
52/20,883
27/20,883
79/20,883
No use†
Occasional use only§
Past use but not recent use¶
Recent use but not past use#
Both past and recent use**
* Adjusted for age, sex, education, body mass index, exercise level, saturated fat intake, fiber intake, and use
of vitamin C and calcium supplements.
† No use reported either 10 years before enrollment (in 1982) or at enrollment (in 1992–1993).
‡ RR, rate ratio; CI, confidence interval.
§ No regular use (≥4 times per week) but occasional use either 10 years before enrollment (in 1982) or at
enrollment (in 1992–1993).
¶ Regular use (≥4 times per week) 10 years before enrollment (in 1982) but no use, or occasional use only, at
enrollment (in 1992–1993).
# Regular use (≥4 times per week) at enrollment (in 1992–1993) but no use, or occasional use only, 10 years
before enrollment (in 1982).
** Regular use (≥4 times per week) both 10 years before enrollment (in 1982) and at enrollment (in 1992–1993).
Confounding by “health conscious” behaviors among
multivitamin users cannot be ruled out as an explanation for
the reduced risk associated with multivitamin use. In this
cohort, multivitamin users were more likely than nonusers to
practice health conscious behaviors, including exercising
more and consuming more fiber and less saturated fat.
Although our results are adjusted for exercise and dietary
factors, both exercise and dietary factors may be measured
with considerable error, potentially resulting in residual
confounding. Confounding by health conscious behaviors,
however, would not explain the pattern of results that we
observed with respect to timing of multivitamin use. Both
participants who reported past multivitamin use and those
who reported only recent multivitamin use appeared more
health conscious than never users, as shown in table 1.
However, there was no suggestion of reduced risk of
colorectal cancer among participants who reported only
recent multivitamin use.
Potential confounding by history of sigmoidoscopy and/or
colonoscopy (whether for screening or for symptoms) is of
particular concern because these procedures can result in the
detection and removal of precancerous polyps, substantially
reducing the risk of colorectal cancer (28). We could not
adjust for sigmoidoscopy or colonoscopy in this analysis
because no information about these procedures was
Am J Epidemiol 2003;158:621–628
collected at enrollment in 1992–1993. However, information
on sigmoidoscopy and colonoscopy was collected on the
follow-up questionnaire in 1997, allowing us to compare the
prevalence of these procedures by multivitamin use status.
Ever having had a sigmoidoscopy or colonoscopy was only
slightly more common among participants who reported
regular multivitamin use in 1982 (52 percent) than among
participants who reported no multivitamin use in 1982 (46
percent). Therefore, it appears unlikely that confounding by
sigmoidoscopy or colonoscopy could fully account for the
observed reduction in risk associated with past multivitamin
use. No previous study has adjusted for sigmoidoscopy or
colonoscopy, although the prevalence of sigmoidoscopy was
noted to be virtually identical by multivitamin use in the
Nurses’ Health Study analysis (14).
The strengths of this analysis include its prospective
design and relatively large size, as well as the availability of
information on multivitamin use reported 10 years before
enrollment. To our knowledge, this is the largest study of
multivitamin use and colon or colorectal cancer incidence.
The availability of information on multivitamin use
collected 10 years before enrollment allowed us to examine
the potential effects of past multivitamin use, without relying
on the ability of participants to accurately report past use.
628 Jacobs et al.
In 1996, the US Food and Drug Administration mandated
that all grain be fortified with folic acid (140 µg of folic acid
per 100 g) by January 1, 1998 (29). This level of fortification
was anticipated to increase daily consumption by approximately 100 µg of folic acid, but improvements in blood
folate levels suggest that fortification may be delivering 200
or more µg of folic acid daily (6). Although some folic acid
fortification may have begun as early as 1996, it is unlikely
that folic acid fortification could have affected results during
our study follow-up period, which ended in mid-1997.
However, if the reduction in colorectal cancer risk associated
with multivitamin use is due to folic acid, multivitamin use
may prove to have little effect on colorectal cancer risk in the
future in populations where large amounts of folic acid are
obtained from fortified foods.
The results of this study and of previous observational
studies (13–16) are consistent with an association between
multivitamin use and a modest reduction in risk of colorectal
cancer after a substantial latency period. The multivitamin
component or components that may be responsible remain
unclear. Intervention studies using intermediate markers
may be useful in clarifying the potential effects of folic acid
and other multivitamin components on different stages of
colorectal carcinogenesis. There are at least two ongoing
randomized trials in the United States examining the effect
of folic acid supplementation on recurrence of colorectal
polyps (E. Giovannucci, Harvard University, personal
communication, 2002; J. Baron, Dartmouth University,
personal communication, 2002). It may also be informative
to conduct intervention studies using markers of even earlier
stages of colorectal carcinogenesis.
In summary, our results are consistent with an association
between past, but not recent, multivitamin use and modestly
reduced risk of colorectal cancer. The timing of multivitamin
use should be examined carefully in future analyses of this
association.
REFERENCES
1. US Food and Drug Administration. Statement of general policy
or interpretation. Subchapter B—food and food products, part
121—food additives. Federal Register 1973;38:20725–6.
2. Physicians’ desk reference. 39th ed. Oradell, NJ: Medical Economics Company, 1985.
3. Physicians’ desk reference for nonprescription drugs. 6th ed.
Oradell, NJ: Medical Economics Company, 1985.
4. Ma J, Stampfer MJ, Hennekens CH, et al. Methylenetetrahydrofolate reductase polymorphism, plama folate, homocysteine, and risk of myocardial infarction in US physicians.
Circulation 1996;94:2410–16.
5. Ford ES, Ballew C. Dietary folate intake in US adults: findings
from the Third National Health and Nutrition Examination Survey. Ethn Dis 1998;8:299–305.
6. Rader J. Folic acid fortification, folate status, and plasma
homocysteine. J Nutr 2002;132(8 suppl):2466S–70S.
7. Kim YI. Folate and carcinogenesis: evidence, mechanisms, and
implications. J Nutr Biochem 1999;10:66–88.
8. Giovannucci E. Epidemiologic studies of folate and colorectal
neoplasia: a review. J Nutr 2002;132(8 suppl):2350S–5S.
9. Terry P, Jain M, Miller AB, et al. Dietary intake of folic acid
and colorectal cancer risk in a cohort of women. Int J Cancer
2002;97:864–7.
10. Flood A, Caprario L, Chaterjee M, et al. Folate, methionine,
alcohol, and colorectal cancer in a prospective study of women
in the United States. Cancer Causes Control 2002;13:551–61.
11. Song J, Sohn KJ, Medline A, et al. Chemopreventive effects of
dietary folate on intestinal polyps in Apc+/–Msh2–/– mice.
Cancer Res 2000;60:3191–9.
12. Martinez ME, Willett WC. Calcium, vitamin D, and colorectal
cancer: a review of the epidemiologic evidence. Cancer Epidemiol Biomarkers Prev 1998;7:163–8.
13. Giovannucci E, Rimm EB, Ascherio A, et al. Alcohol, lowmethionine–low-folate diets, and risk of colon cancer in men. J
Natl Cancer Inst 1995;87:265–73.
14. Giovannucci E, Stampfer MJ, Colditz GA, et al. Multivitamin
use, folate, and colon cancer in women in the Nurses’ Health
Study. Ann Intern Med 1998;129:517–24.
15. Jacobs EJ, Connell CJ, Patel AV, et al. Multivitamin use and
colon cancer mortality in the Cancer Prevention Study II
cohort. Cancer Causes Control 2001;12:927–34.
16. White E, Shannon JS, Patterson RE. Relationship between vitamin and calcium supplement use and colon cancer. Cancer Epidemiol Biomarkers Prev 1997;6:769–74.
17. Calle EE, Rodriguez C, Jacobs EJ, et al. The American Cancer
Society Cancer Prevention Study II Nutrition Cohort—rationale, study design, and baseline characteristics. Cancer 2002;
94:2490–501.
18. Stellman SD, Garfinkel L. Smoking habits and tar levels in a
new American Cancer Society prospective study of 1.2 million
men and women. J Natl Cancer Inst 1986;76:1057–63.
19. Block G, Hartman AM, Naughton D. A reduced dietary questionnaire: development and validation. Epidemiology 1990;1:
58–64.
20. Block G, Coyl L, Smucker R, et al. Health habits and history
questionnaire: diet history and other risk factors. Personal computer system documentation. Bethesda, MD: National Cancer
Institute, Division of Cancer Prevention and Control, National
Institutes of Health, 1989.
21. Ainsworth BE, Haskell WL, Leon AS, et al. Compendium of
physical activities: classification of energy costs of human
physical activities. Med Sci Sports Exerc 1993;25:71–80.
22. Bergmann MM, Calle EE, Mervis CA, et al. Validity of selfreported cancers in a prospective cohort study in comparison
with data from state cancer registries. Am J Epidemiol 1998;
147:556–62.
23. Calle EE, Terrell DD. Utility of the National Death Index for
ascertainment of mortality among Cancer Prevention Study II
participants. Am J Epidemiol 1993;137:235–41.
24. Cox DR. Regression models and life tables (with discussion). J
R Stat Soc (B) 1972;34:187–220.
25. Kleinbaum D. Survival analysis: a self-learning text. New
York, NY: Springer-Verlag, 1996.
26. Hosmer D, Lemeshow S. Applied logistic regression. New
York, NY: John Wiley & Sons, Inc, 1989.
27. Willet W, Stampfer MJ. Total energy intake: implications for
epidemiologic analyses. Am J Epidemiol 1986;124:17–27.
28. Toribara NW, Sleisenger MH. Screening for colorectal cancer.
N Engl J Med 1995;13:861–7.
29. Food and Drug Administration. Food standards: amendment of
standards of identity for enriched grain products to require
addition of folic acid. Federal Register 1996;61:8781–97.
Am J Epidemiol 2003;158:621–628