Dietary Fat and Protein in
Relation to Risk of
Non-Hodgkin’s Lymphoma
Among Women
Shumin Zhang, David J. Hunter,
Bernard A. Rosner, Graham A.
Colditz, Charlie S. Fuchs, Frank E.
Speizer, Walter C. Willett
Higher risk of non-Hodgkin’s lymphoma has been reported consistently in
immunocompromised patients (1–4). An
increase in dietary fat, particularly of n-6
polyunsaturated fatty acids, suppresses
immune status in several animal models
(5), but this has not been seen in humans
fed polyunsaturated fat at levels of from
3% to 13% of energy (6,7). A decrease in
total fat intake (to 25% of energy, mainly
due to a reduction in saturated fat) improved the immune status of healthy
women (6), and long-term ingestion of
large amounts of dietary protein has increased immune tolerance (8). We thus
hypothesized that higher dietary intakes
of fat and protein would increase the risk
of non-Hodgkin’s lymphoma.
In a large cohort of postmenopausal
women, the risk of non-Hodgkin’s lymphoma increased with higher intakes of
animal fat, saturated and monounsaturated fats but not with polyunsaturated
fat (9). In that study, the association with
intake of trans unsaturated fat was not
reported (9). Few studies (9–14) have examined dietary protein and foods contributing to dietary fat and protein in relation
to non-Hodgkin’s lymphoma risk, and the
findings have been inconsistent. We,
therefore, evaluated these relationships in
the Nurses’ Health Study, a large prospective cohort of U.S. women.
METHODS
Study Cohort
In 1976, a total of 121 700 female registered
nurses, aged 30–55 years living in 11 states of the
United States, completed a mailed questionnaire
about their medical history and health-related behaviors. Every 2 years, questionnaires have been sent to
cohort members to update information on potential
risk factors and to ascertain newly diagnosed cancers and other diseases. In 1980, a 61-food-item,
semiquantitative food-frequency questionnaire was
included to obtain dietary information (15). In 1984,
the food-frequency questionnaire was expanded to
116 items (16). Similar questionnaires were used to
update dietary intakes in 1986 and 1990. Through
May 31, 1994, the follow-up was 95% complete as
a percentage of potential person-years.
For the analyses presented here, women were excluded from the 1980 baseline population if they
completed a 1980 dietary questionnaire with implausible total energy intake (i.e., <500 or >3500 kcal/
day), left more than 10 food items blank, had a previous diagnosis of cancer (other than nonmelanoma
Journal of the National Cancer Institute, Vol. 91, No. 20, October 20, 1999
skin cancer), or had missing information on height
and cigarette smoking. These exclusions left a total
of 88 410 women for the analyses. The study was
approved by the Human Research Committee at the
Brigham and Women’s Hospital.
Semiquantitative Food-Frequency
Questionnaire
The validity and reliability of the food-frequency
questionnaires used in the Nurses’ Health Study
have been described elsewhere (15–18). For each
food, a commonly used unit or portion size (i.e., one
egg or one slice of bread) was specified, and women
were asked how often on average over the previous
year they had consumed that amount of each food.
There were nine possible responses, ranging from
“never” to “six or more times per day.” Nutrient
intake was computed by multiplying the frequency
response by the nutrient content of the specified portion sizes. We also asked questions on the types of
fats or oil used in the preparation of foods and at the
table. Values for the specific types of fat and protein
in the foods were obtained from the Harvard University Food Composition Database (updated on November 22, 1993), which is derived from U.S. Department of Agriculture sources (19) and
supplemented with information from manufacturers.
Total trans isomers of unsaturated fat were based on
values analyzed by Enig et al. (20) and by Slover et
al. (21). We included all trans isomers of 18-carbon
unsaturated fatty acids. The polyunsaturated fat reported in this study was the n-6 polyunsaturated linoleic acid (comprising 81% of total polyunsaturated
fat in this population). To calculate the total red meat
intake, we summed the frequencies for six categories—beef, pork, or lamb as a main dish; beef, pork,
or lamb as a sandwich or mixed dish; hamburgers;
Affiliations of authors: S. Zhang, Department of
Nutrition and Epidemiology, Harvard School of
Public Health, Boston, MA; D. J. Hunter, G. A.
Colditz, Department of Epidemiology, Harvard Center for Cancer Prevention, Harvard School of Public
Health, and Channing Laboratory, Department of
Medicine, Harvard Medical School and Brigham
and Women’s Hospital, Boston; B. A. Rosner,
Channing Laboratory, Department of Medicine,
Harvard Medical School and Brigham and Women’s
Hospital, and Department of Biostatistics, Harvard
School of Public Health; C. S. Fuchs, Division of
Medical Oncology, Dana-Farber Cancer Institute,
and Channing Laboratory, Department of Medicine,
Harvard Medical School and Brigham and Women’s
Hospital; F. E. Speizer, Channing Laboratory, Department of Medicine, Harvard Medical School and
Brigham and Women’s Hospital, and Department of
Environmental Health, Harvard School of Public
Health; W. C. Willett, Department of Nutrition and
Epidemiology, Harvard School of Public Health,
and Channing Laboratory, Department of Medicine,
Harvard Medical School and Brigham and Women’s
Hospital.
Correspondence to: Shumin Zhang, M.D., Sc.D.,
Department of Nutrition, Harvard School of Public
Health, 665 Huntington Ave., Boston, MA 02115
(e-mail: Shumin.Zhang@channing.harvard.edu).
See “Notes” following “References.”
© Oxford University Press
REPORTS 1751
Downloaded from http://jnci.oxfordjournals.org/ by guest on November 11, 2014
Background: Non-Hodgkin’s lymphoma occurs more frequently in individuals with suppressed immune status,
and some types of dietary fat and protein have been associated with decreased immune responses. In this
study, we examined the intake of specific types of dietary fat and protein in
relation to the risk of non-Hodgkin’s
lymphoma. Methods: We documented
199 incident cases of non-Hodgkin’s
lymphoma in a cohort of 88 410 women,
who were enrolled in the Nurses’
Health Study and were aged 34–60
years in 1980, during 14 years of follow-up. Relative risks of the disease
and 95% confidence intervals (95%
CIs) were calculated. All P values are
two-sided and were considered to be
statistically significant for P<.05. Results: Intake of saturated fat was associated with an increase in risk that was
not statistically significant; the multivariate relative risk for the highest versus the lowest quintiles of intake was
1.4 (95% CI = 0.7–3.0; P for trend =
.42). Intake of beef, pork, or lamb as a
main dish was associated with a statistically significantly increased risk of
non-Hodgkin’s lymphoma; the multivariate relative risk for consumption of
these meats at least once per day as
compared with less than once per week
was 2.2 (95% CI = 1.1–4.4; P for trend
= .002). Higher intake of trans unsaturated fat was also statistically significantly associated with an increased risk
of the disease; the multivariate relative
risk for the highest versus the lowest
quintiles was 2.4 (95% CI = 1.3–4.6; P
for trend = .01). Higher intake of red
meat cooked by broiling or barbecuing—but not by roasting, pan-frying,
or boiling or stewing—was associated
with an increase in risk that was not
statistically significant. Conclusions:
Greater dietary intake of certain meats
and fats was associated with a higher
risk of non-Hodgkin’s lymphoma.
These relationships and their potential
mechanisms deserve further examination. [J Natl Cancer Inst 1999;91:
1751–8]
Ascertainment of Non-Hodgkin’s
Lymphoma Cases
Incident cases of non-Hodgkin’s lymphoma (International Classification of Diseases, 8th Revision,
code No. 202) were identified by self-report on each
biennial questionnaire from 1982 through 1994.
Deaths in the cohort were identified by reports from
family members, the postal service, and a search of
the National Death Index; we estimated that 98% of
all deaths were identified (28). Women who reported
lymphomas (or their next of kin, if the subject had
died) were asked for permission to obtain hospital
records and pathology reports. Physicians who had
no knowledge of the dietary intake of the participants reviewed the records. During 14 years (representing 1 169 326 person-years) of follow-up, 199
cases of non-Hodgkin’s lymphoma were documented.
Statistical Analysis
Person-years of observation for each participant
were calculated from the date of returning the 1980
questionnaire to the date of diagnosis of nonHodgkin’s lymphoma, death, or June 1, 1994,
whichever came first. For nutrient analyses, women
were categorized by quintile of the 1980 baseline
dietary intake of specific types of fat and protein,
1752 REPORTS
after adjustment for total energy by the multivariate
nutrient density method (18). In the multivariate nutrient density method, nutrient intake was expressed
as a percentage of total energy intake, and total energy was simultaneously included in the model (18).
For each category of nutrient intake, we calculated the incidence rate by dividing the number of
non-Hodgkin’s lymphoma case patients by the number of person-years of follow-up. Relative risk (RR)
was calculated by dividing the incidence rate in an
exposure category by the corresponding rate in the
reference category. Age-adjusted RRs were calculated with the use of 5-year age categories by the
Mantel–Haenszel method (29). In multivariate
analyses by use of pooled logistic regression with
2-year time increments (30,31), we simultaneously
adjusted for age (5-year categories), smoking status
(never, past, or current smoking of one to 14 cigarettes per day or ù15 cigarettes per day), geographic
region (Northeast, Midwest, South, or California),
length of follow-up, total energy intake (quintiles),
and height in inches (<62, ù62 to <64, ù64 to <66,
ù66 to <68, or ù68). These variables were risk
factors for non-Hodgkin’s lymphoma either in this
population (age and height) or in other populations
(smoking status and geographic region). We controlled for total energy to reduce measurement error
due to general overreporting or underreporting of
food items (18). For analyses of types of fat and
protein, we adjusted as well for consumption of
other types of fat, protein, and alcohol; in such models, the effect of each macronutrient can be considered as a substitution for carbohydrate (18). We controlled for fruit and vegetable intake, since this was
inversely associated with risk of non-Hodgkin’s
lymphoma in this population. In these models, age
and smoking status were updated biennially, and total energy intake was calculated from the 1980 foodfrequency questionnaire. We also related body mass
index, defined as the weight in kilograms divided by
the square of the height in meters, to non-Hodgkin’s
lymphoma risk. We did not include body mass index
Table 1. Relative risks of non-Hodgkin’s lymphoma by age, smoking status, geographic region, height,
body mass index, and total energy intake*
No. of
case patients
Person-years
Multivariate RR†
95% CI
Age,‡ y
<50
50–54
55–59
60–64
ù65
34
39
47
43
36
473 384
239 398
228 573
153 210
74 761
1.0
2.1
2.6
3.2
5.4
Referent
1.3–3.3
1.6–4.0
2.0–5.1
3.1–9.1
Smoking status‡
Never
Past
Current, 1–14 cigarettes/day
Current, ù15 cigarettes/day
82
78
16
23
510 112
393 047
82 289
183 878
1.0
1.2
1.3
0.9
Referent
0.8–1.6
0.7–2.2
0.6–1.4
101
41
28
29
682 974
221 512
123 349
141 492
1.0
1.3
1.4
1.2
Referent
0.9–1.8
0.9–2.2
0.8–1.8
12
44
65
51
27
116 096
285 636
364 426
279 291
123 878
1.0
1.5
1.8
1.9
2.4
.01
Referent
0.8–2.9
1.0–3.4
1.0–3.6
1.2–4.7
36
49
45
48
21
249 464
284 711
235 730
268 573
125 246
1.0
1.1
1.2
1.1
1.0
.92
Referent
0.7–1.7
0.8–1.8
0.7–1.6
0.6–1.8
51
42
30
40
36
233 646
233 283
234 280
234 619
233 498
1.0
0.8
0.6
0.8
0.7
.14
Referent
0.6–1.3
0.4–0.9
0.5–1.2
0.5–1.1
Characteristic
Geographic region of residence§
Northeast
Midwest
South
California
Height, in
<62
ù62 to <64
ù64 to <66
ù66 to <68
ù68
Two-sided P for trend
Body mass index,§ weight in
kg/(height in m)2
<21.0
21.0–22.9
23.0–24.9
25.0–29.9
ù30
Two-sided P for trend
Total energy,§ kcal
<1139
1139–1392
1393–1634
1635—1954
ù1955
Two-sided P for trend
*RR 4 relative risk; CI 4 confidence interval.
†Variables were mutually adjusted for each other with additionally controlling for length of follow-up
(seven periods).
‡Updated every 2 years.
§1980 information.
Journal of the National Cancer Institute, Vol. 91, No. 20, October 20, 1999
Downloaded from http://jnci.oxfordjournals.org/ by guest on November 11, 2014
hot dogs; processed meats; and bacon—and took
into account the gram weights of servings.
Nutrient intakes calculated from the 1980 foodfrequency questionnaire were found reasonably correlated with those recorded by 173 Boston women
who kept diet diaries for four 1-week periods for
more than 1 year (15,16). The Pearson correlation
coefficients between energy-adjusted nutrient estimates from the 1980 food-frequency questionnaire
and from the four 1-week dietary records were .47
for total protein, .53 for total fat, .59 for saturated
fat, .48 for polyunsaturated fat, and .61 for cholesterol (15). The Pearson correlation coefficient between the calculated dietary intake of trans unsaturated fatty acids and the proportion of trans
unsaturated fatty acids in adipose tissue was .51
(22). The Pearson correlation coefficients between
the dietary records and the 1980 food-frequency
questionnaire were .38 for meat and .71 for margarine (17). For meat products, fish, and eggs, the
Pearson correlation coefficients between the dietary
records and the 1980 food-frequency questionnaire
ranged from .26 for hamburgers to .68 for eggs; for
dairy products, the values ranged from .49 for hard
cheese to .79 for skim milk and for butter; for sweet
and baked foods, the values ranged from .27 for rice
or pasta to .67 for peanut butter (17).
Because heterocyclic amines (HCAs) generated
during the cooking of meats may be of etiologic
significance for non-Hodgkin’s lymphoma (23–25),
we examined the method of cooking and the degree
of doneness, which have been reported to be related
to HCA levels (26,27). In the 1980 food-frequency
questionnaire, women were also asked what percentage of the time they ate their meat cooked well-done.
In the 1986 food-frequency questionnaire, we asked
participants how often they ate beef, pork, or lamb
as roasted, pan-fried, broiled, barbecued, and boiled
or stewed; there were six possible responses for each
form of cooking ranging from “never” to “five or
more times per week.”
Table 2. Relative risks (RRs) of non-Hodgkin’s lymphoma by quintiles of specific types of dietary fat, cholesterol, and dietary protein in
1980 among 88 410 women
Quintile of nutrient intake
2
3
4
5
Quintile 5,
95% CI*
Two-sided
P for trend
Total fat
No. of case patients
Intake, % energy†
Age-adjusted RR
Multivariate RR‡
Multivariate RR§
32
28.7
1.0
1.0
1.0
38
34.9
1.3
1.3
1.2
43
39.1
1.5
1.5
1.4
46
43.1
1.6
1.7
1.4
40
48.7
1.4
1.4
1.2
0.9–2.2
0.9–2.3
0.7–2.0
.10
.07
.44
Animal fat
No. of case patients
Intake, % energy†
Age-adjusted RR
Multivariate RR‡
Multivariate RR§ + adjustment for vegetable fat
34
18.6
1.0
1.0
1.0
32
24.5
1.0
1.0
0.9
46
28.9
1.5
1.5
1.3
43
33.4
1.4
1.4
1.2
44
40.1
1.4
1.4
1.2
0.9–2.2
0.9–2.2
0.7–2.1
.07
.06
.30
Vegetable fat
No. of case patients
Intake, % energy†
Age-adjusted RR
Multivariate RR‡
Multivariate RR§ + adjustment for animal fat
38
4.0
1.0
1.0
1.0
43
6.7
1.2
1.2
1.2
42
9.0
1.2
1.2
1.2
41
11.7
1.2
1.2
1.2
35
16.2
1.0
1.0
1.0
0.6–1.5
0.6–1.6
0.6–1.7
.81
.90
.89
Saturated fat
No. of case patients
Intake, % energy†
Age-adjusted RR
Multivariate RR‡
Multivariate RR§ + adjustments for monounsaturated,
polyunsaturated, and trans unsaturated fats
30
11.0
1.0
1.0
1.0
36
13.7
1.3
1.3
1.1
50
15.6
1.8
1.9
1.5
36
17.4
1.3
1.4
1.1
47
20.1
1.7
1.8
1.4
1.1–2.7
1.1–2.9
0.7–3.0
.03
.02
.42
Monounsaturated fat
No. of case patients
Intake, % energy†
Age-adjusted RR
Multivariate RR‡
Multivariate RR§ + adjustments for saturated,
polyunsaturated, and trans unsaturated fats
31
11.1
1.0
1.0
1.0
34
13.9
1.2
1.2
0.9
47
15.9
1.7
1.7
1.1
42
17.9
1.5
1.6
0.9
45
20.9
1.6
1.7
0.8
1.0–2.6
1.1–2.7
0.4–1.9
.03
.02
.63
n-6 polyunsaturated fat
No. of case patients
Intake, % energy†
Age-adjusted RR
Multivariate RR‡
Multivariate RR§ + adjustments for saturated,
monounsaturated, and trans unsaturated fats
39
2.5
1.0
1.0
1.0
42
3.4
1.1
1.1
1.0
43
4.1
1.2
1.2
0.9
31
4.8
0.9
0.9
0.6
44
6.2
1.3
1.3
0.8
0.9–2.0
0.8–2.0
0.5–1.4
.38
.49
.30
Trans unsaturated fat
No. of case patients
Intake, % energy†
Age-adjusted RR
Multivariate RR‡
Multivariate RR§ + adjustments for saturated,
monounsaturated, and polyunsaturated fats
28
1.3
1.0
1.0
1.0
37
1.8
1.4
1.4
1.3
46
2.2
1.8
1.8
1.8
34
2.6
1.4
1.4
1.4
54
3.2
2.2
2.2
2.4
1.4–3.4
1.4–3.5
1.3–4.6
.003
.002
.01
Fish omega-3 fatty acids
No. of case patients
Intake, % energy†
Age-adjusted RR
Multivariate RR‡
Multivariate RR‡ + adjustments for saturated and
trans unsaturated fats and fruit and vegetable intake
33
0.02
1.0
1.0
1.0
40
0.03
1.2
1.2
1.2
46
0.04
1.3
1.3
1.4
39
0.05
1.1
1.0
1.2
41
0.10
1.1
1.0
1.4
0.7–1.7
0.7–1.7
0.8–2.2
.90
.71
.42
Cholesterol
No. of case patients
Intake, mg/1000 kcal/day†
Age-adjusted RR
Multivariate RR‡
Multivariate RR§ + adjustments for saturated, monounsaturated,
polyunsaturated, and trans unsaturated fats
Total protein
No. of case patients
Intake, % energy†
Age-adjusted RR
Multivariate RR‡
36
133.8
1.0
1.0
1.0
37
172.3
1.1
1.1
0.9
45
201.8
1.3
1.3
1.1
39
234.6
1.1
1.1
0.9
42
296.9
1.2
1.1
1.0
0.8–1.9
0.7–1.7
0.6–1.6
.49
.75
.92
31
14.7
1.0
1.0
48
17.1
1.5
1.5
37
18.8
1.1
1.1
41
20.6
1.2
1.2
42
24.0
1.2
1.2
0.8–1.9
0.7–1.8
.77
.99
(Table continues)
Journal of the National Cancer Institute, Vol. 91, No. 20, October 20, 1999
REPORTS 1753
Downloaded from http://jnci.oxfordjournals.org/ by guest on November 11, 2014
1
Table 2 (continued). Relative risks (RRs) of non-Hodgkin’s lymphoma by quintiles of specific types of dietary fat, cholesterol, and
dietary protein in 1980 among 88 410 women
Quintile of nutrient intake
Multivariate RR‡ + adjustments for saturated, monounsaturated,
polyunsaturated, and trans unsaturated fats, alcohol, and
fruit and vegetable intake
Animal protein
No. of case patients
Intake, % energy†
Age-adjusted RR
Multivariate RR‡
Multivariate RR‡ + adjustments for saturated, monounsaturated,
polyunsaturated, and trans unsaturated fats, vegetable
protein, alcohol, and fruit and vegetable intake
2
3
4
5
Quintile 5,
95% CI*
Two-sided
P for trend
1.0
1.4
1.1
1.2
1.3
0.8–2.1
.69
29
10.7
1.0
1.0
1.0
50
13.3
1.7
1.7
1.6
34
15.2
1.1
1.1
1.0
45
17.2
1.5
1.5
1.4
41
20.6
1.3
1.3
1.2
0.8–2.1
0.8–2.0
0.7–2.1
.54
.71
.86
47
2.4
1.0
1.0
1.0
33
3.0
0.7
0.7
0.7
47
3.5
1.0
1.0
1.0
40
4.1
0.8
0.8
0.9
32
5.0
0.6
0.6
0.7
0.4–1.0
0.4–0.9
0.4–1.3
.09
.06
.49
*95% CI 4 95% confidence intervals for RRs for quintile 5.
†Values for intake are medians for each quintile of nutrient intake in 1980.
‡Multivariate models included age (5-year categories), total energy (quintiles), length of follow-up (seven periods), geographic region (Northeast, Midwest, South,
or California), cigarette smoking (never, past, or current smoking of 1–14 cigarettes per day or ù15 cigarettes per day), and height in inches (<62, ù62 to <64, ù64
to <66, ù66 to <68, or ù68).
§Additional adjustments for dietary protein (quintiles), alcohol intake (0 g/day, 0.1–4.9 g/day, 5–14.9 g/day, or ù15 g/day), and fruit and vegetable intake (<3
servings/day, 3–3.9 servings/day, 4–4.9 servings/day, 5–5.9 servings/day, or ù6 servings/day).
in multivariate analyses of types of fat and protein
because it was not a risk factor for non-Hodgkin’s
lymphoma in this poulation. In separate analyses
that incorporated repeated dietary measurements,
the incidence of non-Hodgkin’s lymphoma was related to the cumulative average intake from all available dietary questionnaires up to the start of each
2-year follow-up interval or to the most recent intake
at the start of each 2-year follow-up interval (32).
Indicator variables were used to denote any time
period for which a questionnaire was not available.
For all RRs, we calculated 95% confidence intervals
(CIs). All P values were two-tailed and were considered to be statistically significant when <.05.
Tests for trend were conducted by use of the median
values for each quintile of nutrient intake as a continuous variable for nutrient analysis or by use of the
frequency responses in servings per day as one variable for food analysis.
RESULTS
Age was strongly associated with an
increased risk of non-Hodgkin’s lymphoma, and we also observed a strong
positive association with height (Table 1).
However, smoking status, geographic region, body mass index in 1980, and total
energy intake were unrelated to risk
(Table 1). When body mass index was
updated every 2 years, which represents
the most recent status, the association became inverse (RRs for increasing categories 4 1.0, 0.8, 0.8, 0.6 and 0.7 [95% CI
4 0.4–1.2]; P for trend 4 .11).
1754 REPORTS
After adjustment for age and other potential risk factors, we observed statistically significant positive associations between the 1980 dietary intakes of
saturated, monounsaturated, and trans unsaturated fats and the risk of nonHodgkin’s lymphoma (P for trend <.05)
(Table 2). The statistically significant association for intake of trans unsaturated
fat remained after further adjustments for
other types of fat, protein, alcohol, and
fruit and vegetable intake; the multivariate RR for the highest quintile of intake
compared with the lowest quintile of intake was 2.4 (95% CI 4 1.3–4.6). The
comparable RR after additional adjustment for beef, pork, or lamb as a main
dish was 1.8 (95% CI 4 1.1–3.0). However, the positive associations for saturated and monounsaturated fats were
greatly attenuated after further adjustments for other types of fat, protein, alcohol, and fruit and vegetable intake. We
also examined trans unsaturated fat from
vegetable fat (due to partial hydrogenation of vegetable oils) and from animal fat
separately. For increasing quintiles, the
multivariate RRs for trans unsaturated
vegetable fat were 1.0, 1.7, 1.8, 1.8, and
1.9 (95% CI 4 1.2–3.1) (P for trend 4
.03) and for trans unsaturated animal fat
were 1.0, 1.4, 1.3, 1.7, and 1.4 (95% CI 4
0.8–2.2) (P for trend 4 .15); the mean
intakes as percent of energy for quintiles
of trans unsaturated vegetable fat were
0.5%, 0.9%, 1.2%, 1.6%, and 2.3% and
for trans unsaturated animal fat were
0.5%, 0.7%, 0.9%, 1.1%, and 1.3%. Dietary intakes of total fat, animal fat, vegetable fat, n-6 polyunsaturated fat, fish
omega-3 fatty acids, cholesterol, total
protein, animal protein, and vegetable
protein were not statistically significantly
related to risk (Table 2).
When we treated the 1980 nutrients as
continuous variables, the multivariate
RRs were 2.2 (95% CI 4 1.2–3.9) for an
increase of 2% in energy obtained from
trans unsaturated fat compared with the
equivalent energy from carbohydrate and
1.2 (95% CI 4 0.8–1.7) for an increase of
5% in energy from saturated fat compared
with the equivalent energy from carbohydrate. To address the potential bias that
women may change their diet because of
symptoms of non-Hodgkin’s lymphoma,
we further excluded incident cases (n 4
30 cases) occurring during the first 4
years of follow-up; the results did not appreciably change (RR 4 2.3 [95% CI 4
1.2–4.3] for trans unsaturated fat and
RR 4 1.3 [95% CI 4 0.9–2.0] for saturated fat). When nutrients were expressed
as the cumulative average intakes, the as-
Journal of the National Cancer Institute, Vol. 91, No. 20, October 20, 1999
Downloaded from http://jnci.oxfordjournals.org/ by guest on November 11, 2014
Vegetable protein
No. of case patients
Intake, % energy†
Age-adjusted RR
Multivariate RR‡
Multivariate RR‡ + adjustments for saturated, monounsaturated,
polyunsaturated, and trans unsaturated fats, animal
protein, alcohol, and fruit and vegetable intake
1
Table 3. Relative risks (RRs) of non-Hodgkin’s lymphoma by intake of foods as major sources of trans unsaturated fat in 1980
(unit of consumption given in parentheses)
Frequency of consumption
<1/mo
Beef, pork, or lamb (main dish, 6–8 oz)
No. of case patients
Multivariate RR*
95% CI†
1/wk
2–4/wk
}
}}
Margarine (teaspoon; 5 mL)
No. of case patients
Multivariate RR*
95% CI†
1–3/mo
41
1.0
Referent
26
1.0
Referent
32
1.2
0.7–1.8
5–6/wk
1/d
ù2/d
14
0.9
0.5–1.6
42
1.0
0.7–1.5
70
1.2
0.8–1.8
17
2.2
1.1–4.4
51
1.0
0.6–1.5
82
1.4
0.9–2.2
23
1.7
0.9–3.2
Cookie (1)
No. of case patients
Multivariate RR*
95% CI†
39
1.0
Referent
52
1.2
0.8–1.9
27
1.1
0.7–1.8
46
1.5
1.0–2.3
35
1.2
0.7–1.9
White bread (1 slice)
No. of case patients
Multivariate RR*
95% CI†
65
1.0
Referent
14
1.1
0.6–1.9
12
0.9
0.5–1.6
28
0.9
0.6–1.4
25
1.7
1.0–2.6
Pie, home-made (1 slice)
No. of case patients
Multivariate RR*
95% CI†
98
1.0
Referent
75
1.1
0.8–1.5
26
1.6
1.0–2.6
Cake (1 slice)
No. of case patients
Multivariate RR*
95% CI†
51
1.0
Referent
99
1.6
1.1–2.2
28
1.3
0.8–2.1
Two-sided
P for trend
.25
.002
.46
28
1.0
0.6–1.6
27
0.9
0.5–1.4
.65
.07
21
2.0
1.2–3.4
.03
*Multivariate models included age (5-year categories), total energy (quintiles), length of follow-up (seven periods), geographic region (Northeast, Midwest, South,
or California), cigarette smoking (never, past, or current smoking of 1–14 cigarettes per day or ù15 cigarettes per day), and height in inches (<62, ù62 to <64, ù64
to <66, ù66 to <68, or ù68).
†CI 4 confidence interval.
Journal of the National Cancer Institute, Vol. 91, No. 20, October 20, 1999
REPORTS 1755
Downloaded from http://jnci.oxfordjournals.org/ by guest on November 11, 2014
tivariate RR was 1.6 (95% CI 4 1.0–2.6)
for women consuming at least one serving
per week compared with those consuming
less than one serving per month. Higher
intake of cake was also statistically significantly associated with a higher risk;
the multivariate RR was 2.0 (95% CI 4
1.2–3.4) for women consuming at least
two servings per week compared with
those consuming less than one serving per
month. Intakes of cookies and white
bread, two other sources of trans unsaturated fat, were not statistically significantly related to risk. We did not find statistically significant associations for
intakes of beef, pork, or lamb as a mixed
dish or sandwich; bacon; hot dogs; processed meats; hamburgers; chicken with
skin; chicken without skin; fish; eggs;
liver; skim or low-fat milk; whole milk;
or other dairy products with the risk of
non-Hodgkin’s lymphoma (data not
shown).
Risk of non-Hodgkin’s lymphoma was
slightly lower among women who consumed meat cooked well-done; for increasing quintiles of the percent of instances of consumption as well-done, the
multivariate RRs were 1.0 (referent), 0.7,
greatly change their margarine intake between 1970 and 1980 (RRs 4 1.0, 1.2,
0.8, 1.0, and 1.2 [95% CI 4 0.8–1.9]; P
for trend 4 .33). A statistically significant increase in the risk of non-Hodgkin’s
lymphoma was seen with greater intake of
beef, pork, or lamb as a main dish; the
multivariate RR was 2.2 (95% CI 4 1.1–
4.4) for consumption at least once per day
compared with less than once per week
(Table 3) and fell to 2.0 (95% CI 4 1.0–
4.0) after further adjustment for fruit and
vegetable intake and to 1.7 (95% CI 4
0.8–3.7) after additional adjustments for
saturated and trans unsaturated fats. The
association was strengthened when we
limited analyses to women who reported
that they did not greatly change their diet
between 1970 and 1980; the comparable
multivariate RRs were 1.0 (referent), 1.4,
1.9, 2.4, and 3.2 (95% CI 4 1.4–7.3) (P
for trend 4 .003). The multivariate RRs
for increasing quintiles of total red meat
intake were 1.0 (referent), 0.9, 1.3, 1.0,
and 1.3 (95% CI 4 0.8–2.2) (P for trend
4 .25). Beef, pork, or lamb as a main
dish accounted for 68% of the total red
meat intake. Intake of homemade pie was
also related to an increased risk; the mul-
sociations were greatly attenuated; the
multivariate RRs were 1.5 (95% CI 4
0.8–3.0) for an increase of 2% in energy
obtained from trans unsaturated fat and
1.0 (95% CI 4 0.6–1.7) for an increase of
5% in energy from saturated fat. When
nutrients were expressed as the most recent intakes, the associations were further
reduced and became less than 1.0; the
multivariate RRs were 0.9 (95% CI 4
0.5–1.6) for an increase of 2% in energy
obtained from trans unsaturated fat and
0.7 (95% CI 4 0.4–1.0) for an increase of
5% in energy from saturated fat.
Margarine accounted for 22% of total
trans unsaturated fat intake; beef, pork, or
lamb as a main dish for 16%; and cookies
for 12%. Other individual meats, baked
foods, and dairy products each accounted
for less than 7%. Margarine consumption
was weakly and statistically nonsignificantly associated with risk (Table 3).
Similar weak associations were seen for
both stick margarine (RRs 4 1.0, 1.1,
0.7, 1.0, and 1.2 [95% CI 4 0.8–1.2];
P for trend 4 .39) and tub margarine
(RRs 4 1.0, 1.0, 0.8, 0.8, and 1.3 [95%
CI 4 0.9–2.0]; P for trend 4 .33) and in
analyses limited to women who did not
1.0, 0.8, and 0.7 (95% CI 4 0.5–1.0) (P
for trend 4 .07). Consumption of broiled
beef, pork, or lamb at least two to four
times per week was associated with a statistically nonsignificant increase in risk
(RR 4 1.8; 95% CI 4 1.0–3.3; P for
trend 4 .09) (Table 4). We also observed
a statistically nonsignificant increase in
risk with higher intake of barbecued beef,
pork, or lamb. However, consumption of
beef, pork, or lamb that was roasted, panfried, or boiled or stewed was not associated with risk.
DISCUSSION
Table 4. Relative risks (RRs) of non-Hodgkin’s lymphoma by method of cooking beef, pork, or lamb
in 1986
Frequency of consumption
Two-sided
P for trend
<1/mo
1–3/mo
1/wk
ù2–4/wk
Broiled
No. of case patients
Age-adjusted RR
Multivariate RR*
95% CI†
27
1.0
1.0
Referent
37
1.5
1.5
0.9–2.4
30
1.6
1.6
1.0–2.7
19
1.8
1.8
1.0–3.3
Barbecued
No. of case patients
Age-adjusted RR
Multivariate RR*
95% CI†
55
1.0
1.0
Referent
32
1.2
1.2
0.8–1.9
25
1.4
1.5
0.9–2.4
.16
.13
Roasted
No. of case patients
Age-adjusted RR
Multivariate RR*
95% CI†
30
1.0
1.0
Referent
47
0.9
0.9
0.6–1.5
37
0.9
0.9
0.5–1.4
.53
.64
Pan-fried
No. of case patients
Age-adjusted RR
Multivariate RR*
95% CI†
59
1.0
1.0
Referent
29
1.4
1.4
0.9–2.2
22
1.0
1.0
0.6–1.7
.91
.89
Boiled or stewed
No. of case patients
Age-adjusted RR
Multivariate RR*
95% CI†
52
1.0
1.0
Referent
42
1.3
1.3
0.9–2.0
17
0.9
1.0
0.5–1.7
.79
.94
.13
.09
*Multivariate models included age (5-year categories), total energy (quintiles), length of follow-up (four
periods), geographic region (Northeast, Midwest, South, or California), cigarette smoking (never, past, or
current smoking of 1–14 cigarettes per day or ù15 cigarettes per day), and height in inches (<62, ù62 to <64,
ù64 to <66, ù66 to <68, or ù68).
†CI 4 confidence interval.
1756 REPORTS
Health Study (9), we observed a positive
association between saturated fat intake
and the risk of non-Hodgkin’s lymphoma,
albeit not statistically independent of
trans unsaturated fat, and we found no
association with n-6 polyunsaturated fat.
Animal protein intake was associated
with an increased risk in the Iowa Women’s Health Study (9) and in an international correlation study (10). We did not
confirm this association in this study,
which is consistent with the finding from
a case–control study (11).
This study revealed a strong positive
association between trans unsaturated fat
and the risk of non-Hodgkin’s lymphoma.
Margarine intake, a major source of trans
unsaturated fat, had a statistically significant positive association with risk in a
case–control study (RR 4 1.4; P <.05)
(13). We found weak and nonsignificant
associations with intakes of margarine
and cookies and observed no association
with white bread. However, we found statistically significant positive associations
with intakes of pie and cake, which are
often high in trans unsaturated fat from
partially hydrogenated vegetable oils. Our
finding on trans unsaturated fat needs to
be interpreted with caution. We cannot
exclude the possibility that this was due to
chance because so many nutrients were
examined.
Milk consumption has been associated
with the risk of non-Hodgkin’s lymphoma
in two case–control studies and in a cohort study (12–14); however, this was not
seen in the Iowa Women’s Health Study
(9), a case–control study in women (11),
or our data. Hamburger intake was associated with an increased risk in the Iowa
Women’s Health Study (9), but we found
no such association. However, women in
the Iowa Women’s Health Study had
higher consumption of hamburgers (36%
of them consumed more than one per
week as compared with 24% in this study)
(9). The present finding of an approximately twofold higher risk of nonHodgkin’s lymphoma among women who
consumed a higher amount of beef, pork,
or lamb has not been reported before. Because so many foods were examined, we
also cannot exclude with confidence the
possibility that this positive result was
due to chance. However, the greater
strength of association among women
with consistent long-term consumption
suggests that this may be a true causative
factor. Although different meat products
were most strongly associated with risk in
Journal of the National Cancer Institute, Vol. 91, No. 20, October 20, 1999
Downloaded from http://jnci.oxfordjournals.org/ by guest on November 11, 2014
In this large cohort of women, intake
of beef, pork, or lamb as a main dish was
most strongly associated with the risk of
non-Hodgkin’s lymphoma. Higher intakes of trans unsaturated fat and, perhaps, saturated fat were associated with
an increased risk. Other major types of fat
or protein were not independently associated with risk.
The prospective design and high follow-up rates in this study minimize the
concern that methodologic biases explain
the findings. The estimates of dietary intakes derived from the dietary questionnaire used in this study reasonably reflect
long-term intakes of study subjects (15).
However, some misclassification of individual long-term diet still exists, which
would tend to weaken any true associations. Because patients with non-Hodgkin’s
lymphoma are usually diagnosed because
of symptoms, such as significant weight
loss, fever, and night sweats, dietary
change because of these symptoms is a
potential source of bias, which is consistent with the different results for body
mass index when using the most recent
exposure rather than the index in 1980.
The use of baseline diet rather than the
most recent diet helps to reduce this bias.
Residual confounding by nondietary factors cannot be excluded, but it is also unlikely to explain these observed findings
because controlling for potential nondietary risk factors for non-Hodgkin’s lymphoma had minimal effect on the RRs.
Other aspects of foods associated with
risk could confound associations with
types of fat.
Consistent with the Iowa Women’s
REFERENCES
(1) Filipovich AH, Mathur A, Kamat D, Shapiro
RS. Primary immunodeficiencies: genetic risk
factors for lymphoma. Cancer Res 1992;52(19
Suppl):5465s–7s.
(2) Kinlen L. Immunosuppressive therapy and acquired immunological disorders. Cancer Res
1992;52(19 Suppl):5474s–6s.
(3) Rabkin CS, Biggar RJ, Baptiste MS, Abe T,
Kohler BA, Nasca PC. Cancer incidence trends
in women at high risk of human immunodeficiency virus (HIV) infection. Int J Cancer
1993;55:208–12.
(4) Levine AM, Shibata D, Sullivan-Halley J,
Nathwani B, Brynes R, Slovak ML, et al. Epidemiological and biological study of acquired
immunodeficiency syndrome-related lymphoma in the county of Los Angeles: preliminary results. Cancer Res 1992;52 (19 Suppl):
5482s–4s.
(5) Kelley DS, Bendich A. Essential nutrients and
immunologic functions. Am J Clin Nutr 1996;
63:994S–6S.
(6) Kelley DS, Dougherty RM, Branch LB, Taylor
PC, Iacono JM. Concentration of dietary N-6
polyunsaturated fatty acids and the human immune status. Clin Immunol Immunopathol
1992;62:240–4.
(7) Kelley DS, Branch LB, Iacono JM. Nutritional
modulation of human immune status. Nutr Res
1989;9:965–75.
(8) Moreau MC, Coste M. Immune responses to
dietary protein antigens. World Rev Nutr Diet
1993;74:22–57.
(9) Chiu BC, Cerhan JR, Folsom AR, Sellers TA,
Kushi LH, Wallace RB, et al. Diet and risk of
non-Hodgkin lymphoma in older women.
JAMA 1996;275:1315–21.
(10) Cunningham AS. Lymphomas and animalprotein consumption. Lancet 1976;2:1184–6.
(11) Ward MH, Zahm SH, Weisenburger DD, Gridley G, Cantor KP, Saal RC, et al. Dietary factors and non-Hodgkin’s lymphoma in Nebraska (United States). Cancer Causes Control
1994;5:422–32.
(12) Franceschi S, Serraino D, Carbone A, Talamini
R, La Vecchia C. Dietary factors and nonHodgkin’s lymphoma: a case–control study in
the northeastern part of Italy. Nutr Cancer
1989;12:333–41.
(13) Tavani A, Pregnolato A, Negri E, Franceschi
S, Serraino D, Carbone A, et al. Diet and risk
of lymphoid neoplasms and soft tissue sarcomas. Nutr Cancer 1997;27:256–60.
(14) Ursin G, Bjelke E, Heuch I, Vollset SE. Milk
consumption and cancer incidence: a Norwegian prospective study. Br J Cancer 1990;61:
454–9.
(15) Willett WC, Sampson L, Stampfer MJ, Rosner
B, Bain C, Witschi J, et al. Reproducibility and
validity of a semiquantitative food frequency
questionnaire. Am J Epidemiol 1985;122:
51–65.
(16) Willett WC, Sampson L, Browne ML, Stampfer
MJ, Rosner B, Hennekens CH, et al. The use of
a self-administered questionnaire to assess diet
four years in the past. Am J Epidemiol 1988;
127:188–99.
(17) Salvini S, Hunter DJ, Sampson L, Stampfer
MJ, Colditz GA, Rosner B, et al. Food-based
validation of a dietary questionnaire: the
effects of week-to-week variation in food
consumption. Int J Epidemiol 1989;18:858–
67.
(18) Willett W. Nutritional epidemiology. 2nd ed.
New York (NY): Oxford University Press;
1998.
(19) U.S. Department of Agriculture, Agricultural
Journal of the National Cancer Institute, Vol. 91, No. 20, October 20, 1999
(20)
(21)
(22)
(23)
(24)
(25)
(26)
(27)
(28)
(29)
(30)
(31)
(32)
(33)
Research Service. USDA Nutrient Database
for Standard Reference, Release 10: Nutrient
Data Laboratory Home Page, http://
www.nal.usda.gov/fnic/foodcomp, 1993.
Enig MG, Pallansch LA, Sampugna J, Keeney
M. Fatty acid composition of the fat in selected
food items with emphasis on trans components. J Am Oil Chem Soc 1983;60:1788–95.
Slover HT, Thompson RH, Jr., Davis CS,
Merola GV. Lipids in margarines and margarine-like foods. J Am Oil Chem Soc 1985;62:
775–86.
London SJ, Sacks FM, Caesar J, Stampfer MJ,
Siguel E, Willett WC. Fatty acid composition
of subcutaneous adipose tissue and diet in postmenopausal US women. Am J Clin Nutr 1991;
54:340–5.
Frandsen H, Rasmussen ES, Nielsen PA,
Farmer P, Dragsted L, Larsen JC. Metabolic
formation, synthesis and genotoxicity of the Nhydroxy derivative of the food mutagen
2-amino-1-methyl-6-phenylimidazo(4,5-b)
pyridine (PhIP). Mutagenesis 1990;6:93–8.
Snyderwine EG, Schut HA, Adamson RH,
Thorgeirsson UP, Thorgeirsson SS. Metabolic
activation and genotoxicity of heterocyclic arylamines. Cancer Res 1992;52(7 Suppl):
2099s–102s.
Esumi H, Ohgaki H, Kohzen E, Takayama S,
Sugimura T. Induction of lymphoma in CDF1
mice by the food mutagen 2-amino-1-methyl6-phenylimidazo[4,5-b]pyridine. Jpn J Cancer
Res 1989;80:1176–8.
Sinha R, Rothman N, Salmon CP, Knize MG,
Brown ED, Swanson CA, et al. Heterocyclic
amine content in beef cooked by different
methods to varying degrees of doneness and
gravy made from meat drippings. Food Chem
Toxicol 1998;36:279–87.
Sinha R, Knize MG, Salmon CP, Brown ED,
Rhodes D, Felton JS, et al. Heterocyclic amine
content of pork products cooked by different
methods and to varying degrees of doneness.
Food Chem Toxicol 1998;36:289–97.
Stampfer MJ, Willett WC, Speizer FE, Dysert
DC, Lipnick R, Rosner B, et al. Test of the
National Death Index. Am J Epidemiol 1984;
119:837–9.
Rothman KJ, Greenland S. Modern epidemiology. 2nd ed. Philadelphia (PA): Lippincott–
Raven; 1998.
Cupples LA, D’Agostino RB, Anderson K,
Kannel WB. Comparison of baseline and repeated measure covariate techniques in the
Framingham Heart Study. Stat Med 1988;7:
205–22.
D’Agostino RB, Lee ML, Belanger AJ, Cupples
LA, Anderson K, Kannel WB. Relation of
pooled logistic regression to time dependent
Cox regression analysis: the Framingham
Heart Study. Stat Med 1990;9:1501–15.
Hu FB, Stampfer MJ, Rimm E, Ascherio A,
Rosner BA, Spiegelman D, et al. Dietary fat
and coronary heart disease: a comparison of
approaches for adjusting for total energy intake
and modeling repeated dietary measurements.
Am J Epidemiol 1999;149:531–40.
Bjeldanes LF, Morris MM, Timourian H,
Hatch FT. Effects of meat composition and
cooking conditions on mutagen formation in
REPORTS 1757
Downloaded from http://jnci.oxfordjournals.org/ by guest on November 11, 2014
the Nurses’ Health Study and in the Iowa
Women’s Health Study (9), they are consistent in showing that some aspects of
red meat consumption may be related to
risk of non-Hodgkin’s lymphoma.
The formation of HCAs increases with
temperature and duration of cooking;
meats cooked to a rare or medium-rare
state have markedly less mutagen content
than well-done meats (26,27,33). However, we found a weak inverse association
between a higher frequency of well-done
meat consumption and the risk of nonHodgkin’s lymphoma, which is consistent
with the finding from the Iowa Women’s
Health Study comparing women who
consumed well-done meats with those
who consumed rare or medium-rare meats
(9). Food-cooking methods also influence
the formation of HCAs; frying, broiling,
and barbecuing meats produced substantially more HCAs than did stewing,
steaming, and poaching (34). In this
study, only intake of beef, pork, or lamb
that had been broiled or barbecued was
associated with increased risk. However,
in this population, chicken was the major
source of HCAs (35), yet chicken consumption was unrelated to risk. Thus, our
findings do not fully support the hypothesis of HCAs as causative factors for nonHodgkin’s lymphoma. Our questions on
meat-cooking method and doneness are
not comprehensive and have limitations,
which may underestimate the true association of HCAs with non-Hodgkin’s
lymphoma risk.
We are unaware of other studies that
have examined the association between
height and the risk of non-Hodgkin’s lymphoma. Adult height in part reflects early
energy balance (18); its implication on
non-Hodgkin’s lymphoma deserves further investigation.
In the Nurses’ Health Study, higher intakes of beef, pork, or lamb as a main
dish, trans unsaturated fat, and saturated
fat are associated with greater risk of nonHodgkin’s lymphoma. The etiologic importance of these associations remains
unclear, and these deserve further investigations in other cohorts.
fried ground beef. J Agric Food Chem 1983;
31:18–21.
(34) Doolittle DJ, Rahn CA, Burger GT, Lee CK,
Reed B, Riccio E, et al. Effect of cooking
methods on the mutagenicity of food and on
urinary mutagenicity of human consumers
[published erratum appears in Food Chem
Toxicol 1990;28:140]. Food Chem Toxicol
1989;27:657–66.
(35) Byrne C, Sinha R, Platz EA, Giovannucci E,
Colditz GA, Hunter DJ, et al. Predictors of dietary heterocyclic amine intake in three prospective cohorts. Cancer Epidemiol Biomarkers Prev 1998;7:523–9.
NOTES
1758 REPORTS
Downloaded from http://jnci.oxfordjournals.org/ by guest on November 11, 2014
Supported by Public Health Service grant
CA40356 from the National Cancer Institute, National Institutes of Health, Department of Health and
Human Services.
We thank the participants of the Nurses’ Health
Study for their continuing outstanding dedication
and commitment to the study and Frank B. Hu, Barbara Egan, Lisa Li, Karen Corsano, Laura Sampson,
and Debbie Flynn for their advice and assistance.
Manuscript received February 21, 1999; revised
August 4, 1999; accepted August 20, 1999.
Journal of the National Cancer Institute, Vol. 91, No. 20, October 20, 1999