Content uploaded by Mariette Gerber
Author content
All content in this area was uploaded by Mariette Gerber on Nov 17, 2015
Content may be subject to copyright.
Nutrition and Cancer, 60(2), 177–187
Copyright © 2008, Taylor & Francis Group, LLC
ISSN: 0163-5581 print / 1532-7914 online
DOI: 10.1080/01635580701649651
Dietary Factors and Breast Cancer Risk: A Case Control
Study Among a Population in Southern France
Fa¨
ıza Bessaoud and Jean-Pierre Daur`
es
Laboratoire de Biostatistiques et d’Epid´
emiologie-Institut Universitaire de Recherche Clinique,
Montpellier Cedex, France
Mariette Gerber
Centre de Recherche en Canc´
erologie—Centre R´
egional de Lutte Contre le Cancer—Val d’Aurelle Paul
Lamarque, Montpellier, France
This case-control study examined different food groups in rela-
tion to breast cancer. Between 2002 and 2004, 437 cases and 922
controls matched according to age and area of residence were in-
terviewed. Diet was measured by a validated food frequency ques-
tionnaire. Adjusted odds ratios (Ors) were computed across levels
of various dietary intakes identified by two methods: the “classical”
and the “spline” methods. Neither of the 2 methods found an asso-
ciation between total fruit and vegetable consumption and breast
cancer. Results of the 2 methods showed a nonsignificant decreased
association with cooked vegetables intake as well as legumes and
fish consumption. Whereas the spline method showed no associa-
tion, the classical method showed significant associations related
to the lowest consumption of raw vegetables or dairy products and
breast cancer risk: Adjusted OR for raw vegetable consumption
between (67.4 and 101.3 g/day) vs. (<67.4 g/day) was 0.63 [95%
confidence interval (CI) =0.43–0.93]. Adjusted OR for dairy con-
sumption between (134.3 and 271.2 g/day) vs. (<134.3 g/day) was
1.57 (95% CI =1.06–2.32). However, the overall results were not
consistent. Compared to the classical method, the use of the spline
method showed a significant association for cereal, meat, and olive
oil. Cereal and olive oil were inversely associated with breast cancer
risk. Breast cancer risk increased by 56% for each additional 100
g/day of meat consumption. Studies using novel methodological
techniques are needed to confirm the dietary threshold responsible
for changes in breast cancer risk. New approaches that consist in
analyzing dietary patterns rather than dietary food are necessary.
INTRODUCTION
Breast cancer is a major public health problem. Despite sig-
nificant advances in clinical research over the past few decades,
breast cancer remains the commonest cancer among women. Al-
though mortality from breast cancer has declined, the incidence
Submitted 11 April 2007; accepted in final form 13 August 2007.
Address correspondence to F. Bessaoud, 1 Laboratoire de Biostatis-
tiques et d’Epid´
emiologie-Institut Universitaire de Recherche Clinique,
640 Av du Doyen Gaston Giraud 34093 Montpellier Cedex 5, France.
has increased dramatically. Furthermore, the rate of breast can-
cer incidence varies widely amongst the world’s populations.
Age-adjusted breast cancer incidence rates are high in Western
countries in North America and Western Europe, for example,
whereas rates in some Asian countries are low but increasing.
The fact that breast cancer rates among migrants from low-risk
population areas reach the rates of their adoptive country (1)
suggests that exposure to environmental rather than genetic fac-
tors may play a role in breast cancer etiology. It is certain that
some of the international variations in breast cancer are due
to a variation in reproductive factors such as age at full-term
pregnancy and the number of births or are due to genetic fac-
tors. These factors do not explain all the differences in rates,
however, and imply that environmental factors such as nutrition
might also be important. In France, the number of newly diag-
nosed cases of breast cancer has increased twofold over the last
two decades. Between 1978 and 2000 alone, the incidence rate
rose by 53% (2). This increase is due largely to mass screen-
ing, but there is growing evidence that changes in environment
and dietary habits also play an important role. High dietary fat
and high consumption of meat and dairy products appear to
be partly responsible for the increase in breast cancer (3–6).
Conversely, other studies have shown that high consumption of
fruits, vegetables, and fish might protect against breast cancer
(7–9).
Although the relationship between food intake and breast
cancer was widely investigated, findings remain inconsistent, in
part because of heterogeneity of the risk factor thresholds for
which the risk of breast cancer varies. Thus, the way of such
thresholds are determined has an impact on public health point of
view. Dose-response analysis remains the key approach to epi-
demiological studies; it is usually based on a priori choice of the
threshold values because they are represented by the quintiles of
the dietary variables. In this case-control study on a population
in Southern France, we have chosen to examine the associa-
tion between diet and breast cancer using a different statistical
177
178 F. BESSAOUD ET AL
dose-response approach. This new approach provides an es-
timated threshold values related to the probability of disease
occurrence and varying according to the odds ratio (OR) of
breast cancer. Such thresholds are selected by using free knot
spline function in a logistic model (10). The spline logistic
method provides a different way of looking at the relationship
between disease and risk factors because an accurate estimation
of quantitative risk factor thresholds can be provided, and the
dose-response interpretation can be optimized.
To point out the advantage of this new method, we compared
both the “classical method” and the alternative statistical “spline
method” by examining the association between several dietary
factors responsible for increasing (e.g., dietary fat) or reducing
(e.g., dietary fiber and vegetables) the risk of breast cancer.
MATERIALS AND METHODS
Study Population
We conducted a population-based case-control study among
French women aged 25 to 85 yr. Eligible case subjects were
French-speaking women, residing in Southern France (H´
erault
only), with no former history of breast cancer. We attempted to
recruit all those patients newly diagnosed with primary breast
cancer confirmed histologically between June 2002 and Decem-
ber 2004. All histological types of breast cancer were selected
except the lobular carcinoma in situ. Patients were identified
from both surgical wards and medical information department
records of the 3 principal regional centers of care. The median
time between diagnosis and interview was 3 mo.
The control subjects were randomly selected from the list of
residents supplied by the electoral roll. For each case recruited,
two controls were matched according to both age (±1 yr) and
area of residence (rural/urban).
Among the respondents, the rate of participation was 90.8%
for cases and 71.6% for controls. A total of 437 eligible cases
and 922 eligible controls were enrolled in the study. The study
protocol was approved by the two ethical committees for stud-
ies using human subjects (the Consultative Committee on Data
Processing for Medical Research and the National Commission
on Data Processing and Rights), which advocate that all medical
information is confidential and anonymous.
Data Collection
Data were obtained by means of a structured questionnaire
administered by two trained interviewers. The first part of the
questionnaire was related to demographic characteristics, re-
productive and menstrual factors, oral contraception, hormone
replacement therapy, family history of cancer, anthropometric
factors, physical activities, and smoking history.
Women were considered to be menopausal if menstruation
had ceased naturally in the 12 mo prior to the interview or
following a bilateral oophorectomy. In addition, we estimated
the lifetime duration of ovulatory activity by subtracting the
cumulative number of months of pregnancy, breast-feeding, and
use of oral contraceptives from the number of years between
menarche and menopause (or the date of interview in the case
of premenopausal women).
The second part of the questionnaire was related to diet and
alcohol consumption. Diet was measured by a validated food fre-
quency questionnaire (11,12) organized mainly by food groups.
This food frequency questionnaire was previously used for diet
studies (13,14). It assessed the subjects’ usual diets in the pre-
vious year if their habits had not changed recently. Otherwise,
dietary habits before such a change were taken into account.
Subjects were asked to indicate their average daily, weekly, and
monthly consumption of single food items. The seasonal varia-
tion of food consumption was taken into account in this ques-
tionnaire. Intakes were quantitatively evaluated using a validated
set of photographs that represent different portions of food in
grams. Daily consumption in grams was obtained by dividing
the reported amount of food intake by the frequency of consump-
tion during a given period of time (e.g., divided by 28 if once a
month, divided by 7 if once a week). Total energy intake (kcal)
and 59 different nutrients (measured in grams per day) were
computed using a modified program (Fruit d’Or Astra-Calv´
e),
which to a large extent draws from the French food composition
R´
egal database (15). Ten individual fatty acids were added to
the initial database (INSERM Unit 476).
Statistical Analysis
Multivariate conditional logistic regression was employed to
analyze the association between breast cancer and dietary risk
factors. Odd ratios (OR) with their associated 95% confidence
intervals (CI) across exposure levels of various types of dietary
intake were calculated and adjusted for potential confounders.
Estimates were produced using the PHREG procedure in the
statistical software package SAS (Version 5.0). The Wald test
was used to test nullity of parameters.
The statistical analysis was subdivided into two distinct steps.
First, the relationship of each food with breast cancer was an-
alyzed with adjustment made only for the total energy intake.
Second, we took into account complementary adjustments by
potential confounders. These factors were chosen according to
their statistical significance (P<10%) when they were taken
separately in a univariate analysis. The potential confounders
included in the multivariate logistic model were education, par-
ity, age at first full-term pregnancy (FFTP), duration of ovu-
latory activity, breast-feeding, family history of breast cancer
(first-degree: mother and sisters), body mass index (BMI), and
physical activity (leisure, household work, occupational). This
case-control study was matched according to age with a nar-
row interval (±1 yr), so statistical adjustment for age was not
required. Furthermore, as menopausal was identified as a non-
confounder, there was no need for statistical adjustment.
For the first step of the dose-response analysis, instead of
transforming the continuous variables into ordinal variables only
DIETARY FACTORS AND BREAST CANCER RISK 179
TABLE 1
The best modelling of breast cancer occurrence based on AIC criterion
according to food intake and associated threshold values
Food intake The best model Threshold location
Total fruit (fruit and juice) Linear model
Total vegetables Linear piecewise with 1 knot 222 g/day
Raw vegetables Linear piecewise with 1 knot 99 g/day
Cooked vegetables Linear piecewise with 1 knot 105 g/day
Cereals Linear piecewise with 1 knot 44 g/day
Legumes Linear model
Fish and seafood Linear piecewise with 1 knot 23 g/day
Meats Linear model
Dairy products (milk and cheese) Linear model
Olive oil Linear piecewise with 3 knots 2, 12, 20.5 g/day
according to distribution quartiles (the sole function of which is
to form different groups of homogenous size), free knot splines
for logistic models and threshold selection were used to provide
a better adapted way of splitting the continuous variables into
several groups by threshold values varying according to the
OR of breast cancer. Furthermore, extreme thresholds (located
before the 25th quartile or after the 75th quartile) can be revealed
by this method. This statistical method was recently applied to
a clinical trial for an in vitro fertilization program and provided
convincing results (10).
Unlike logistic models that assess the linear effect of the risk
factor X on the model and for which the logit function is linear
in its parameters and supposes that a unit change in one com-
ponent of X has the same effect on the patient over the whole
range of this component, the “spline method” provides an inter-
esting way of looking at the relationship between disease and
explanatory variables by relieving the linearity assumption of
the logit function. The risk factor X enters the logistic model
via a set of linear spline functions (i.e., the degree of the spline
function is restricted to 1) that differs according to the num-
ber of knots (including the case where the number of knots is
null). A spline function without knots corresponds to a linear
function. The function that best fits the data is selected using
Akaike information criterion and Bayesian information crite-
rion (16,17). Moreover, the knots of spline function are esti-
mated for a given risk factor and can be considered as threshold
values for which the OR of breast cancer changes. Thus, the
continuous variables were transformed into ordinal variables
using these threshold values. When the best modeling corre-
sponds to the linear model—model without knots—the vari-
ables were analyzed as a continuous variables. This method was
applied to our data using the Splus package. Results are shown
in Table 1.
The best models in terms of goodness of fit, which repre-
sent the variation of breast cancer OR according to different
categories of foods, were given with their associated threshold
values.
The comparison between the different models used to fit the
OR of breast cancer, that is, the logit function, shows that the
linear function corresponds to the best model for fruit, meat,
legumes, and dairy products, whereas the linear piecewise func-
tion with one knot corresponds to the best model for vegetables,
cereals, and fish consumption. The model with 3 knots seems
to be more adapted to olive oil consumption. Thus, we used the
knot locations of each model to change the continuous variables
to discrete variables.
Tests of the linear trend, when appropriate, were conducted
by considering the variable in the conditional logistic model
to be continuous, with the Pvalue attesting the degree of sig-
nificance of the variable in the model. All the Pvalues were
two-sided, and values <0.05 were considered statistically sig-
nificant.
RESULTS
The mean age of cases was 58.1 (±11.5) yr and that of con-
trols was 57.4 (±11.4) yr, with no significant difference between
the two groups (P=0.3056). Table 2 shows the relationship
between breast cancer and established risk factors.
Parity tends to decrease the risk of breast cancer, and a first-
degree family history of breast cancer tends to increase breast
cancer risk. Education, breast-feeding, and leisure physical ac-
tivity were significantly and inversely associated to breast cancer
risk. Conversely, a late FFTP, long duration of ovulatory activ-
ity, high BMI, as well as household and occupational physical
activities significantly increased breast cancer risk.
Further results were related to the relationship between
breast cancer risk and various food intake. Levels of food
consumption were determined by 2 methods: the “classical
method” (quartile thresholds) and the “spline method” (es-
timated thresholds). Neither of the 2 methods revealed an
association between total vegetable intake (Tables 3 and 4)
or total fruit intake (Tables 3 and 5) and the risk of breast
cancer.
180 F. BESSAOUD ET AL
TABLE 2
Association between breast cancer and established risk factorsa
Variable Subgroups No. of cases No. of controls Unadjusted OR CI (95%)
First-degree family
history of breast
cancer
No 360 789 1
Yes 64 107 1.34 0.95–1.89
Parity Nulliparous 55 89 1
Parous 382 833 0.70 0.49–1.01
Age at first full
term pregnancy
<30 321 744 1
≥30or nulliparous 115 178 1.60 1.21–2.21
Breast-feeding No 233 419 1
Yes 204 503 0.71 0.56–0.89
Duration of
ovulatory activity
≤29.6yr 189 453 1
>29.6 yr 223 424 1.43 1.08–1 91
Education <Baccalaureate 302 567 1
≥Baccalaureate 134 355 0.71 0.55–0.92
BMI 18–25 281 657 1
25–30 105 204 1.17 0.89–1.55
>30 51 61 1.90 1.26–2.86
Physical activity
(leisure)
No 160 243 1
Yes 276 679 0.63 0.49–0.80
Physical activity
(household)
≤14h/week 178 249 1
>14 h/week 259 373 2.13 1.67–2.72
Physical activity
(occupation)
Sedentary job 248 618 1
No sedentary job 189 304 1.57 1.23–2.00
aAbbreviations are as follows: OR, odds ratio; CI, confidence interval; BMI, body mass index.
When the estimated threshold (spline threshold) was used to
create the groups of levels of consumption, analysis by vegetable
subgroup showed that raw vegetables were not associated with
breast cancer (Table 4), whereas when the quartile thresholds
were used, an inverse association was observed: the adjusted
OR estimates for consumption between (67.4 and 101.3 g/day)
vs the class with the lowest consumption level (<67.4 g/day)
was 0.63 (95% CI =0.43–0.93). However, classes with higher
levels of consumption were not associated with breast cancer
risk (Table 3).
Cooked vegetable consumption was not associated with
breast cancer risk (Table 3 and 4), and this result did not differ
according to the splitting method chosen to create the different
groups of levels of consumption.
With regards to consumption of legumes and dairy products,
the spline method showed that the risk of breast cancer varied
linearly according to these kind of foods (i.e., no threshold was
detected). OR per 100 g/day of legumes =0.58, 95% CI =
0.29–1.19 and OR per 100 g/day of dairy products =1.02, 95%
CI =0.95–1.10). Whereas results of the two methods agreed, in
the sense that both of them showed a nonsignificant association
between legumes consumption and breast cancer risk, the results
related to dairy products were discordant. No association with
breast cancer was observed when the spline method was used
(Table 5).
However, when the quartile thresholds were used, the
breast cancer risk associated with dairy consumption between
134.3 and 271.2 g/day was increased compared to the risk
with lower consumption (<134.3 g/day). Classes of higher
dairy consumption were not associated with breast cancer risk
(Table 3).
Whereas the classical method found no association between
cereals, fish, meat, and olive oil consumption and breast can-
cer, the spline method showed a significant association between
cereals, meat, and olive oil and breast cancer risk. For fish con-
sumption, a borderline inverse association was observed.
Cereal intake appears to be related to breast cancer risk. The
logistic spline method shows that the OR of breast cancer had
a cut point located at cereal intake of 44 g/day (e.g., 2 slices of
bread), and that a consumption of cereals above this threshold
decreased the breast cancer risk by 60%. The association is
significant (OR =0.40, 95% CI =0.23–0.68) and adjustment
for established risk factors did not invalidate this result (OR =
0.29, 95% CI =0.15–0.55). To verify whether the risk reduction
was related to fiber in the cereals, we adjusted for both total fiber
and for total energy intake, and once again the relationship was
unchanged (OR =0.40 95% CI =0.24–0.69).
Breast cancer risk associated with fish and seafood consump-
tion above 23 g/day was reduced (OR =0.77; 95% CI =0.61–
0.98) compared to the risk associated with lower consumption
DIETARY FACTORS AND BREAST CANCER RISK 181
TABLE 3
Adjusted OR and CI of breast cancer associated with food groupsa
Intake (g/day) SubgroupsbNo. of cases No. of controls % ORc(CI95)OR
d(CI95)
Total fruits 0 ≤Q≤171.96 104 23.8 236 25.61 1
171.96<Q≤271.10 103 23.6 237 25.7 0.95 0.68–1.33] 1.06 (0.72–1.57)
271.10 <Q≤402.01 102 23.3 238 25.8 0.94 0.66–1.33] 1.54 (0.70–1.59)
Q>402.01 128 29.3 211 23.9 1.26 0.89–1.78] 1.35 (0.90–2.04)
Total vegetables 0 ≤Q≤218.8 115 26.32 226 24.51 1 1
218.8 <Q≤310.3 99 22.65 240 26.03 0.79 (0.57, 1.10) 0.86 (0.59 –1.26)
310.3 <Q≤426.3 105 24.03 235 25.49 0.86 (0.60–1.17) 0.99 (0.67–1.47)
Q>426.3 118 27.00 221 23.97 0.95 (0.67–1.34) 1.10 (0.76–1.64)
Raw vegetables 0 ≤Q≤67.4 122 23.6 218 27.91 1
67.4 <Q≤101.3 92 26.9 248 21.0 0.64 (0.46–0.90) 0.63 (0.43–0.93)
101.3 <Q≤160.2 102 25.7 337 23.3 0.75 (0.54–1.04) 0.88 (0.59–1.25)
Q>160.2 121 23.6 219 27.7 0.95 (0.68–1.33) 1.04 (0.70–1.53)
Cooked vegetables 0 ≤Q≤127.14 114 26.09 224 24.31 1
127.14 <Q≤193.45 97 22.20 244 26.46 0.75 (0.54–1.05) 0.85 (0.57–1.23)
193.45 <Q≤276.19 119 27.23 221 23.97 1.01 (0.72–1.40) 1.09 (0.74–1.59)
Q>276.19 107 24.49 233 25.27 0.80 (0.57–1.13) 0.99 (0.67–1.46)
Cereals 0 ≤Q≤96.6 118 27.0 222 24.11 1
96.6 <Q≤143.9 104 23.8 236 25.6 0.80 (0.58–1.11) 0.81 (0.56–1.19)
143.9 <Q≤200.7 97 22.2 235 25.5 0.72(0.51–1.01) 0.77 (0.52–1.15)
Q>200.7 118 27.0 229 24.8 0.83 (0.58–1.19) 0.79 (0.52–1.21)
Legumes 0 ≤Q≤7.1 81 18.5 148 16.11 1
7.1 <Q≤17.9 158 36.2 344 37.3 0.81 (0.58–1.14) 0.96 (0.65–1.43)
17.9 <Q≤32.1 96 21.9 188 20.4 0.88 (0.60–1.28) 1.14 (0.72–1.77)
Q>32.1 102 23.3 242 26.3 0.69 (0.47–1.01) 0.75 (0.48–1.18)
Dairy products 0 ≤Q≤134.34 101 23.11 239 25.92 1 1
134.34 <Q≤271.17 128 29.29 211 22.89 1.39 (1.00–1.92) 1.57 (1.06–2.32)
271.17 <Q≤402.87 100 22.88 241 26.14 0.92 (0.66–1.24) 0.94 (0.64–1.40)
Q>402.87 108 24.86 231 25.05 1.00 (0.71–1.41) 1.00 (0.67–1.50)
Meats 0 ≤Q≤10.71 49 11.21 114 12.36 1 1
10.71 <Q≤21.43 302 69.11 704 76.36 0.99 (0.68–1.43) 0.97 (0.63–1.50)
Q>21.43 86 19.68 104 11.28 1.98 (1.24–3.14) 1.61 (0.92–2.79)
Fish and seafood Q ≤15.57 123 28.15 216 23.43 1 1
15.57 <Q≤25.64 108 24.71 232 25.16 0.83 (0.60–1.14) 0.80(0.68–1.40)
25.64 <Q≤41.4 102 23.34 236 25.60 0.75 (0.56–1.04) 0.86(0.53–1.24)
Q>41.4 104 23.80 238 25.81 0.69 (0.49–0.97) 0.79(0.54–1.16)
Olive oil 0 ≤Q≤5.82 120 27.46 219 23.75 1 1
5.82 <Q≤11.04 100 22.88 240 26.03 0.75 (0.56–1.04) 0.85 (0.58–1.26)
11.04 <Q≤20.03 109 24.94 232 25.16 0.88 (0.64–1.21) 1.02 (0.71–1.47)
Q>20.03 108 24.71 231 25.05 0.82 (0.59–1.14) 1.02 (0.71–1.49)
aAbbreviations are as follows: OR, odds ratio; CI95, confidence interval (95%); Q, quartile.
bLevels of food consumption were determined according to quartile distribution.
cAdjusted for total energy intake.
dAdjusted for total energy intake, education, parity, breast-feeding age at first full-term pregnancy, duration of ovulatory activity,
body mass index, physical activity, and first-degree family history of breast cancer.
(≤23 g/day). Adjustment for established risk factors attenuates
the association (OR =0.80; 95% CI =0.61–1.06). On the other
hand, adjustment for only eicosapentaenoic acid (EPA) and do-
cosahexaenoic acid (DHA) fatty acids and total energy intake
modifies this association, suggesting that the link observed be-
tween fish intake and breast cancer is potentially due to the
protective effect of Omega3 fatty acid EPA-DHA (OR =0.83;
95% CI =0.63–1.10).
182 F. BESSAOUD ET AL
TABLE 4
OR and CI of breast cancer associated with selected foods adjusted for total energy and breast cancer risk factorsa
Intake (g/day) Subgroups No. of cases % No. of controls % ORb(CI95%)OR
c(CI95%)
Total vegetables 0 ≤Q≤222 118 27% 238 25.81 1 1
Q>222 319 73% 684 74.19 0.88 (0.67–1.16) 0.98 (0.72–1.36)
Raw vegetables 0 ≤Q≤99 209 47.83% 443 48.05 1 1
Q>99 228 52.7% 479 51.95 0.98 (0.77–1.24) 1.12 (0.85–1.49)
Cooked vegetables 0 ≤Q≤105 86 19.68% 155 16.81 1 1
Q>105 351 80.32% 767 83.19 0.77 (0.57–1.05) 0.86 (0.60–1.24)
Cereals 0 ≤Q≤44 28 6.41% 31 3.36 1 1
Q>44 409 93.59% 891 96.64 0.40 (0.23–0.68) 0.29 (0.15–0.55)
Fish and seafood Q ≤23 203 46.45% 382 41.43 1 1
Q>23 234 53.55% 540 58.57 0.77 (0.61–0.98) 0.80 (0.61–1.06)
Olive oil Q ≤2 67 15.33% 77 8.35 1 1
2<Q≤11.6 169 38.67% 401 43.49 0.50 (0.34–0.72) 0.53 (0.35–0.84)
11.6 <Q≤20.5 96 21.97% 230 24.95 0.51 (0.34–0.76) 0.60 (0.37–0.96)
Q>20.5 105 24.03% 214 23.21 0.56 (0.38–0.84) 0.71 (0.44–1.14)
aAbbreviations are as follows: OR, odds ratio; CI95%, confidence interval (95%). Foods are ones for which a linear piecewise delineated
by knots was the best model.
bAdjusted for total energy intake.
cAdjusted for total energy intake, education, parity, breast-feeding age at first full-term pregnancy, duration of ovulatory activity, body mass
index, physical activity, and first-degree family history of breast cancer.
Meat intake was linearly associated with breast cancer risk.
There was an approximate 2.5-fold increase in the risk of breast
cancer with each additional 100 g/day of meat consumption
(OR =2.56; 95% CI =1.60–4.10). This association remains
significant after adjustment for established risk factors (OR =
1.95; 95% CI =1.08–3.53). Adjustment for saturated fatty
acids and total energy intake did not affect this association
(OR =2.50; 95% CI =1.56–4.03), implying that the dele-
terious effect was probably not directly related to saturated
fatty acid.
TABLE 5
OR and CI of breast cancer associated with selected foods
adjusted for total energy and breast cancer risk factorsa
Intake (g/day) ORb(CI95%)OR
c(CI95%)
100 of total fruit 1.03 (0.96–1.09) 1.02 (0.95–1.10)
100 of legumes 0.60 (0.32–1.11) 0.58 (0.29–1.19)
100 of meat 2.56 (1.60–4.10) 1.95 (1.08 –3.53)
100 of dairy products 0.99 (0.95–1.05) 1.00 (0.95–1.06)
aAbbreviations are as follows: OR, odds ratio; CI95%, confidence
interval (95%). Foods are ones for which a linear model was the best
model.
bAdjusted for total energy intake.
cAdjusted for total energy intake, education, parity, breast-feeding
age at first full-term pregnancy, duration of ovulatory activity, body
mass index, physical activity, and first-degree family history of breast
cancer.
Olive oil was inversely associated with breast cancer risk.
However, adjustment for established risk factors modifies this
association only on the third class identified by the spline method
(>20.5 g/day vs. <2 g/day; OR =0.71; 95% CI =0.44–
1.14), whereas adjustment by monounsaturated fatty acids and
total energy intake maintained a significant risk decrease in
this class (>20.5 g/day vs. <2 g/day; OR =0.29; 95% CI =
0.18–0.47).
DISCUSSION
With regards to established risk factors, our results are in line
with the findings of previous studies (18). An increased risk of
breast cancer was associated with late age of FTTP, long du-
ration of ovulatory activity, high BMI, and first-degree family
history of breast cancer. Inversely, a decreased risk of breast
cancer was associated with parity, breast-feeding, and high ed-
ucational level. In our data, leisure physical activity decreased
breast cancer risk significantly. These findings are supported by
several studies (19–21). In contrast, occupational and household
activity in our data was associated with increased risk. The asso-
ciation remained unchanged after adjustment for socioeconomic
level, which can be considered as a confounding factor. Further
investigation must be conducted to understand the underlying
process.
This case-control study examined a number of different food
groups in relation to breast cancer. We used two statistical meth-
ods that allowed the evaluation of the risk of breast cancer
through different levels of consumption of dietary factors. The
DIETARY FACTORS AND BREAST CANCER RISK 183
first method, the classical method, was based on quartile distri-
bution. The second method, the spline method, based on free
knot spline function in logistic models and threshold selection,
provides a different way of looking at the relationship between
disease and risk factors. The resulting threshold values were
estimated according to OR of breast cancer variation.
Fruit and Vegetable Consumption
In this case-control study, neither of the two methods found
an association between total fruit consumption and breast cancer
risk. Several previous studies have investigated the relationship
between fruit consumption and breast cancer (8,9,22). Whereas
some studies have shown no significant decrease in breast can-
cer risk associated with high consumption of fruit, a large scale
prospective study (23) agreed with our finding that there is no
evidence of the protective effect of fruit intake. Although fruit
consumption has been assumed to play an important role in
prevention due to the concentration of antioxidants and other
phytochemical components, its association with breast cancer
risk has not yet been demonstrated. This indicates that these
types of phytoconstituents do not play a significant role in breast
cancer prevention (24). Our negative findings on raw vegetables
support this observation. However, results from the two meth-
ods were divergent. The use of the classical method showed a
protective effect associated with low consumption of raw veg-
etables. The risk associated with consumption between 67.4 and
101.3 g/day decreased significantly compared to the risk asso-
ciated with lower consumption (<67.4 g/day). However, higher
levels of consumption were not associated with breast cancer
risk.
Results from the 2 methods agreed and showed a borderline
significant association for consumption of cooked vegetables,
but adjustment for the established risk factors largely modified
this association. As for total vegetables, no association with
breast cancer risk was found.
Previous studies on the relationship between total vegetable
consumption or that of subgroup consumption and risk of breast
cancer have produced differing results (8,9,22,25). The pooled
analysis of 8 cohort studies (9) found results in agreement with
our findings in that no evidence of the association between
intake of vegetables and risk of breast cancer was observed. A
meta-analysis of 12 case-control studies and 9 cohort studies
(8) found a protective effect of vegetables against breast cancer
when all the studies were considered together: The analysis of
case-control studies showed a significant reduction of 14% for
an increment of 100 g/day intake of vegetables; no association
was observed after analysis of the cohort studies. In the meta-
analysis of 26 published studies (22), a significant reduction in
breast cancer of 25% was associated with a high consumption of
vegetables. We cannot exclude the possibility that the protective
effect of vegetables may be due to specific subgroups of women.
There is some evidence that the protective effect of vegetables
would be stronger in women with estrogen receptor positive
tumors (26). Information relating to tumor hormone dependence
was not included in our study; hence, we did not evaluate this
link.
Dairy Products Consumption
In this study, the use of the spline method showed no asso-
ciation between dairy products and breast cancer, whereas the
classical method showed that breast cancer risk associated with
consumption between 134.3 and 271.2 g/day increased by 57%
when compared to the risk associated with consumption of less
than 134.3 g/day. However, higher levels of consumption were
not associated with breast cancer risk. Results from epidemi-
ological studies of dairy products and breast cancer have been
inconsistent so far (5,6,27–31). Most of the studies reviewed
recently (32) showed no consistent pattern of increased or de-
creased breast cancer with high consumption of dairy products.
Two opposite mechanisms have been suggested to explain the
association between dairy products and breast cancer. On one
hand, certain types of fat such as saturated fat as well as growth
factors contained in milk products have been shown to promote
breast cancer cell growth. In contrast, mechanisms based on
the cellular role of vitamin D and calcium have been thought
to protect against breast cancer development (33). The vitamin
D contained in dairy products results mostly from enrichment,
which varies from country to country. This may explain partly
the divergence between studies.
Meat Consumption
Compared to the classical method, which didn’t find an asso-
ciation between some kinds of food and breast cancer risk, the
use of the spline method found an interesting association. With
regards to the association between meat consumption and breast
cancer risk, only 6 studies (5,6,34–37) out of 10 studies (5,6,34–
41) have shown a positive association with breast cancer risk. In
our study, the risk of breast cancer increased approximately 2.5-
fold with high versus low consumption of meat. Meat has been
assumed to increase breast cancer risk through its contribution of
one or all of following factors: saturated fat, animal protein, and
potential mammary carcinogens including hetrocyclic amines,
polyaromatic hydrocarbons, and nitroso compounds (42). In our
analysis, adjustment for saturated fatty acids and total energy
intake did not modify the results. This suggests that the role
played by meat intake is due to other factors than saturated fat
and energy. Recent studies have found the relationship between
meat intake and breast cancer to be independent of total fats
and proteins (6,34,43). Conversely, methods of processing and
cooking meat may be a source of hetrocyclic amines, polyaro-
matic hydrocarbons, and nitroso compounds, all of which have
been shown to be mammary carcinogens in rodents and in some
human cell cultures (44).
Fish Consumption
In our study, the consumption of fish was associated with a
significant decrease in breast cancer risk, and this appeared to
184 F. BESSAOUD ET AL
be primarily due to EPA-DHA fatty acids. This effect, however,
was attenuated after adjustment for the established risk fac-
tors. These results are consistent with the hypothesis suggesting
that omega 3 fatty acids, specifically EPA and DHA present
in fish, reduce mammary tumor cell proliferation by increas-
ing apoptosis (45). A review of several epidemiological studies
(7) comprising 7 prospective cohort studies and 19 case control
studies examined the association between fish consumption or
marine fatty acid intake and breast cancer risk. However, re-
sults of these epidemiological studies remain inconsistent with
regards to the relationship between fish consumption and breast
cancer risk. A recent study (46) examined fish consumption and
breast cancer risk in 310,671 women aged between 25 and 70
yr at recruitment into the European Prospective Investigation
Into Cancer and Nutrition project. The results provided no ev-
idence of an association between fish intake and breast cancer
risk. Separate analysis found a slightly positive association with
fatty fish. Further investigation must be conducted to elucidate
the association between fish consumption and breast cancer risk
and especially to consider the potential negative effect of some
contaminants such as mercury and dioxin which could counter-
balance the beneficial effects (46). Recent studies (47,48) have
suggested than the real effect of Omega 3 fatty acids depends
on background levels of Omega 6 fatty acids. It is possible that
discrepancies between studies may be due in part to this hypoth-
esis. Further investigations are needed to evaluate the potential
interaction between fatty acids on the development of breast
cancer.
Cereals Consumption
Epidemiological analysis as regards to the association be-
tween cereals and breast cancer risk remains discordant (29,49–
53). Some studies (51–53) have shown that cereal-based foods
decreased breast cancer risk. In Southern France, cereals are
consumed in the form of breakfast cereals, breads, pasta, or
whole-meal breads. These products are rich in fiber. Thus,
fiber appears to be the most plausible explanation for this
effect (54,55) because it may protect against breast cancer
by inhibiting intestinal reabsorption of estrogens present in
the bile. The fact that fiber adjustment did not modify the
OR associated with cereals was an unexpected outcome and
could reflect some inaccuracy in the existing food composition
table.
Olive Oil Consumption
Consumption of olive oil has been associated with decreased
breast cancer risk in case control studies in Spain (56), Greece
(57), and Italy (58) and also in the recent longitudinal study
in a Mediterranean population of dietary determinants related
to high mammographic breast density (59). Our findings agree
with this protective effect. Unlike cardiovascular disease, the
role of monounsaturated fatty acids (MUFA) in the protective
association between olive oil and breast cancer is not evident.
MUFA were associated with an increase in breast cancer risk
when animal meat provided the bulk of MUFA. In contrast, when
olive oil was the predominant source of MUFA, risk decreased
(45). In our findings, adjustment for MUFA emphasized the
inverse association between olive oil and breast cancer risk,
indicating that it might be other constituents of olive oil that
were responsible for the observed effect. It is possible that the
antioxidant oleuropeine and secoiridoid phenolic compounds
modify Phase I and Phase II enzymes and hence have an effect on
carcinogen elimination (60). On the other hand, the consumption
of olive oil might reflect a favorable dietary pattern rich in
vegetables and poor in animal products.
This study is limited by measurement errors inherent in retro-
spective studies. Although cases and controls were interviewed
in similar conditions, we cannot avoid a possible recall bias
because cases could have modified responses concerning their
dietary habit owing to their knowledge of their disease. Further-
more, we cannot exclude the possibility of selection bias because
controls who participated were likely to have been more health
conscious.
To reduce possible measurement errors that exist in ex-
posure measurements when using a nutritional questionnaire,
photographs of portion sizes were used to estimate intake as
demonstrated by Block et al. (61). However, the nondifferential
errors of measurement of dietary intake can never be completely
avoided and could have led to the lack of association between
some foods and the risk of breast cancer. It is possible that the
borderline effects observed between breast cancer risk and sev-
eral foods, such as cooked vegetables and legumes, were the
consequences of such errors.
Although some inconsistency appeared when we compared
the classical with the spline method, the spline method brings
more convincing results: Whereas the spline method found no
association between dairy products or raw vegetables and breast
cancer, the classical method showed an association. However,
the associations shown were the results of “significance” among
1 cell of 3 considered, without an evidence of consistency of
the overall results. Inconsistent findings of the previous studies
(8,9,22,25) about the relationship between raw vegetables or
dairy products (5,6,27–31) and breast cancer didn’t allow us to
confirm the associations found.
The classical method, known for robustness, was based on
contestable thresholds because their choice was based only on
homogeneity of the groups in terms of size. The use of the spline
logistic model provided estimated threshold values, which al-
lowed the creation of “optimal groups” of exposure to the risk
factor. These groups were constructed according to the varia-
tion of breast cancer OR and thus to the probability of disease
occurrence.
Rather than to investigate the association across all the classes
defined by the classical method, the spline method assesses the
risk across generally restricted classes delineated by the values
of the explanatory variable for which the risk may be changed.
DIETARY FACTORS AND BREAST CANCER RISK 185
Such a case is illustrated by the relationship between fish con-
sumption and breast cancer. The classical method explored the
risk across 4 classes of seafood consumption and finally showed
that above 25.64 g/day of fish consumption, the risk of breast
cancer begun to vary significantly. As for the spline method, it
pointed out directly the quantities of fish consumption (i.e., 25
g/day) for which the risk may be changed significantly, allowing
to reduce the statistical analysis.
Similar observations were also noted for the relationship
between olive oil and breast cancer. In contrast to the classical
method, the spline method showed an association between olive
oil and breast cancer. The accuracy of the result found seems
plausible because this method by creating optimal groups has
improved the statistical power of analysis and thus has revealed
an association between olive oil and breast cancer.
It is important to note that use of such method does not
provide systematically threshold values: When the risk varies
linearly according to the explanatory variable, the thresholds
cannot be determined. Such a case was observed for meat con-
sumption: Whereas the classical method did not find any asso-
ciation between meat consumption and breast cancer, the spline
method showed that the risk didn’t depend on threshold val-
ues but was increased linearly according to meat consumption.
Furthermore, this association was statistically significant.
Even if the 2 methods require similar conditions of sample
size, this new method increases the statistical power of the analy-
sis. Thus, statistical differences can be detected because the risk
is explored on a reduced number of classes of the explanatory
variable.
Furthermore, extreme thresholds (located before the 25th
quartile or after the 75th quartile) can be revealed by the use of
the spline method. It is, however, necessary to point out that this
advantage can constitute a weakness if the highlighted threshold
is misleading due to extreme or aberrant data. To avoid this
methodological problem, an optimization constraint to exclude
the thresholds lower than the 5th or higher than the 95th quintile
is advised.
In conclusion, the use of the spline logistic method in
this study provides a practical tool for investigating the dose-
response relationship. Our results support the role played by
consumption of meat and dairy products in increasing breast
cancer risk and that of cereals, raw vegetables, and oil olive
in its reduction. Nonsignificant decreases were associated with
cooked vegetables as well as legumes and fish consumption. To-
tal fruit and vegetable intake seemed not to be associated with
breast cancer risk.
Although a large number of studies have been already de-
voted to the relationship between dietary factors and breast can-
cer, additional studies using novel and sophisticated method-
ological techniques are needed to elucidate the association and
confirm the dietary threshold values responsible for changes in
breast cancer risk.
Because of strong evidence revealed by several reports (62–
66) that have suggested that analysis of dietary patterns gives
more insight into the relationship between diet and cancer than
analysis of single nutrients or foods, we will be adopting this
new approach in the future.
ACKNOWLEDGMENTS
This work was supported by a grant from the National Cancer
Institute of France and the H´
erault Regional Council.
REFERENCES
1. Ziegler RG, Hoover RN, Pike MC, Hildesheim A, Nomura AM, et al.:
Migration patterns and breast cancer risk in Asian-American women. J
Natl Cancer Inst 85, 1819–1827, 1993.
2. Remontet L, Esteve J, Bouvier AM, Grosclaude P, Launov G, Menegoz F,
Exbrayat C, Tretare B, Carli PM, et al.: Cancer incidence and mortality
in France over the period 1978–2000. Rev Epidemiol Sante Publique 51,
3–30, 2003.
3. Willett WC: Diet and breast cancer. J Intern Med 249, 395–411, 2001.
4. Boyd NF, Stone J, Vogt KN, Connelly BS, Martin LJ, et al.: Dietary fat and
breast cancer risk revisited: a meta-analysis of the published literature. Br
J Cancer 89, 1672–1685, 2003.
5. Richardson S, Gerber M, and Cenee S: The role of fat, animal protein and
some vitamin consumption in breast cancer: a case control study in southern
France. Int J Cancer 48, 1–9, 1991.
6. Toniolo P, Riboli E, Shore RE, and Pasternack BS: Consumption of meat,
animal products, protein, and fat and risk of breast cancer: a prospective
cohort study in New York. Epidemiology 5, 391–397, 1994.
7. Terry PD, Rohan TE, and Wolk A: Intakes of fish and marine fatty acids
and the risks of cancers of the breast and prostate and of other hormone-
related cancers: a review of the epidemiologic evidence. Am J Clin Nutr 77,
532–543, 2003.
8. Riboli E, and Norat T: Epidemiologic evidence of the protective effect of
fruit and vegetables on cancer risk. Am J Clin Nutr 78, 559S–569S, 2003.
9. Smith-Warner SA, Spiegelman D, Yaun SS, Adami HO, Beeson WL, van
den Brandt PA, Folsom AR, Fraser GE, Freudenheim JL, et al.: Intake of
fruits and vegetables and risk of breast cancer: a pooled analysis of cohort
studies. JAMA 285, 769–776, 2001.
10. Bessaoud F, Daures JP, and Molinari N: Free knot splines for logistic mod-
els and threshold selection. Comput Methods Programs Biomed 77, 1–9,
2005.
11. Daures JP, Gerber M, Scali J, Astre C, Bonifacj C, and Kaaks R: Vali-
dation of a food-frequency questionnaire using multiple-day records and
biochemical markers: application of the triads method. J Epidemiol Biostat
5, 109–115, 2000.
12. Bonifacj C, Gerber M, Scali J, and Daures JP: Comparison of dietary as-
sessment methods in a southern French population: use of weighed records,
estimated-diet records and a food-frequency questionnaire. Eur J Clin Nutr
51, 217–231, 1997.
13. Scali J, Richard A, and Gerber M: Diet profiles in a population sample from
Mediterranean southern France. Public Health Nutr 4, 173–182, 2001.
14. Gerber MJ, Scali JD, Michaud A, Durand MD, Astre CM, Dallongeville
J, and Romon MM: Profiles of a healthful diet and its relationship to
biomarkers in a population sample from Mediterranean southern France. J
Am Diet Assoc 100, 1164–1171, 2000.
15. Favier J-C, Ireland-Ripert J, Toque C, and Feinberg M: R´
epertoire g´
en´
eral
des aliments—Table de composition 1995.
16. Akaike H: Information theory and an extension of the maximum likelihood
principle. In P. Petrov, and F. Csaki (Eds.), 1974.
17. Schwarz G: Estimating the dimension of a model. The Annals of Statistics:
461–464, 1978.
18. Key TJ, Verkasalo PK, and Banks E: Epidemiology of breast cancer. Lancet
Oncol 2, 133–140, 2001.
186 F. BESSAOUD ET AL
19. Tehard B, Friedenreich CM, Oppert JM, and Clavel-Chapelon F: Effect of
physical activity on women at increased risk of breast cancer: results from
the E3N cohort study. Cancer Epidemiol Biomarkers Prev 15, 57–64, 2006.
20. Patel AV, Callel EE, Bernstein L, Wu AH, and Thun MJ: Recreational
physical activity and risk of postmenopausal breast cancer in a large cohort
of US women. Cancer Causes Control 14, 519–529, 2003.
21. Breslow RA, Ballard-Barbash R, Munoz K, and Graubard BI: Long-term
recreational physical activity and breast cancer in the National Health and
Nutrition Examination Survey I epidemiologic follow-up study. Cancer
Epidemiol Biomarkers Prev 10, 805–808, 2001.
22. Gandini S, Merzenich H, Robertson C, and Boyle P: Meta-analysis of
studies on breast cancer risk and diet: the role of fruit and vegetable con-
sumption and the intake of associated micronutrients. Eur J Cancer 36,
636–646, 2000.
23. van Gils CH, Peeters PH, Bueno-de-Mesquita HB, Boshuizen HC, Lahmann
PH, Clavel-Chapelon F, and al, e: Consumption of vegetables and fruits and
risk of breast cancer. Jama 293, 183–193, 2005.
24. Gerber M, Boutron-Ruault MC, Hercberg S, Riboli E, Scalbert A, and Siess
MH: Food and cancer: state of the art about the protective effect of fruits
and vegetables. Bull Cancer 89, 293–312, 2002.
25. Link LB, and Potter JD: Raw versus cooked vegetables and cancer risk.
Cancer Epidemiol Biomarkers Prev 13, 1422–1435 2004.
26. Gaudet MM, Britton JA, Kabat GC, Steck-Scott S, Eng SM, Teitelbaum SL,
Terry MB, Neugut AI, and Gammon MD: Fruits, vegetables, and micronu-
trients in relation to breast cancer modified by menopause and hormone
receptor status. Cancer Epidemiol Biomarkers Prev 13, 1485–1494, 2004.
27. Landa MC, Frago N, and Tres A: Diet and the risk of breast cancer in Spain.
Eur J Cancer Prev 3, 313–320, 1994.
28. Franceschi S, Favero A, La Vecchia C, Negri E, Dal Maso L, Salvini S,
Decarli A, and Giacosa A: Influence of food groups and food diversity on
breast cancer risk in Italy. Int J Cancer 63, 785–789, 1995.
29. Missmer SA, Smith-Warner SA, Spiegelman D, Yaun SS, Adami HO,
Beeson WL, van den Brandt PA, Fraser GE, Freudenheim JL, et al.: Meat
and dairy food consumption and breast cancer: a pooled analysis of cohort
studies. Int J Epidemiol 31, 78–85, 2002.
30. Shin MH, Holmes MD, Hankinson SE, Wu K, Colditz GA, and Willett WC:
Intake of dairy products, calcium, and vitamin d and risk of breast cancer.
JNatlCancerInst94, 1301–1311, 2002.
31. Mannisto S, Pietinen P, Virtanen M, Kataja V, and Uusitupa M: Diet and
the risk of breast cancer in a case-control study: does the threat of dis-
ease have an influence on recall bias? J Clin Epidemiol 52, 429–439,
1999.
32. Moorman PG, and Terry PD: Consumption of dairy products and the risk
of breast cancer: a review of the literature. Am J Clin Nutr 80, 5–14, 2004.
33. Cui Y, and Rohan TE: Vitamin D, calcium, and breast cancer risk: a review.
Cancer Epidemiol Biomarkers Prev 15, 1427–1437, 2006.
34. Shannon J, Cook LS, and Stanford JL: Dietary intake and risk of post-
menopausal breast cancer (United States). Cancer Causes Control 14, 19–
27, 2003.
35. De Stefani E, Ronco A, Mendilaharsu M, Guidobono M, and Deneo-
Pellegrini H: Meat intake, heterocyclic amines, and risk of breast cancer:
a case-control study in Uruguay. Cancer Epidemiol Biomarkers Prev 6,
573–581, 1997.
36. Gaard M, Tretli S, and Loken EB: Dietary fat and the risk of breast cancer:
a prospective study of 25,892 Norwegian women. Int J Cancer 63, 13–17,
1995.
37. Vatten LJ, Solvoll K, and Loken EB: Frequency of meat and fish intake and
risk of breast cancer in a prospective study of 14,500 Norwegian women.
IntJCancer46, 12–15, 1990.
38. Gertig DM, Hankinson SE, Hough H, Spiegelman D, Colditz GA, Willett
WC, Kelsey KT, and Hunter DJ: N-acetyl transferase 2 genotypes, meat
intake and breast cancer risk. Int J Cancer 80, 13–17, 1999.
39. Ambrosone CB, Freudenheim JL, Sinha R, Graham S, Marshall JR, Vena
JE, Laughlin R, Nemoto T, and Shields PG: Breast cancer risk, meat con-
sumption and N-acetyltransferase (NAT2) genetic polymorphisms. Int J
Cancer 75, 825–830, 1998.
40. Matos EL, Thomas DB, Sobel N, and Vuoto D: Breast cancer in Argentina:
case-control study with special reference to meat eating habits. Neoplasma
38, 357–366, 1991.
41. Mills PK, Beeson WL, Phillips RL, and Fraser GE: Dietary habits and
breast cancer incidence among Seventh-day Adventists. Cancer 64, 582–
590, 1989.
42. Food, nutrition and the prevention of cancer: a global perspective American
Institute for Cancer Research, Washington DC, 1997.
43. Ronco A, De Stefani E, Mendilaharsu M, and Deneo-Pellegrini H: Meat, fat
and risk of breast cancer: a case-control study from Uruguay. IntJCancer
65, 328–331, 1996.
44. Snyderwine EG: Some perspectives on the nutritional aspects of breast
cancer research. Food-derived heterocyclic amines as etiologic agents in
human mammary cancer. Cancer 74, 1070–1077, 1994.
45. GerberM, Thi ´
ebaut A, Astrog P, and al. e: Dietary fat, fatty acid composition
and risk of cancer. Eur. J. Lipid. Sci. 107, 540–559, 2005.
46. Engeset D, Alsaker E, Lund E, Welch A, Khaw KT, Clavel-Chapelon F,
Thiebaut A, Chajes V, Key TJ, et al.: Fish consumption and breast cancer
risk. The European Prospective Investigation into Cancer and Nutrition
(EPIC). J Cancer 2006.
47. Larsson SC, Kumlin M, Ingelman-Sundberg M, and Wolk A: Dietary long-
chain n-3 fatty acids for the prevention of cancer: a review of potential
mechanisms. Am J Clin Nutr 79, 935–945, 2004.
48. MacLean CH, Newberry SJ, Mojica WA, Khanna P, Issa AM, Suttorp MJ,
Lim YW, Traina SB, Hilton L, et al.: Effects of omega-3 fatty acids on
cancer risk: a systematic review. Jama 295, 403–415, 2006.
49. Favero A, Parpinel M, and Franceschi S: Diet and risk of breast cancer:
major findings from an Italian case-control study. Biomed Pharmacother
52, 109–115, 1998.
50. Franceschi S, La Vecchia C, Russo A, Negri E, Favero A, and Decarli A:
Low-risk diet for breast cancer in Italy. Cancer Epidemiol Biomarkers Prev
6, 875–879, 1997.
51. La Vecchia C, Chatenoud L, Negri E, and Franceschi S: Session: whole
cereal grains, fibre and human cancer wholegrain cereals and cancer in
Italy. Proc Nutr Soc 62, 45–49, 2003.
52. Chatenoud L, Tavani A, La Vecchia C, Jacobs DR Jr., Negri E, Levi F, and
Franceschi S: Whole grain food intake and cancer risk. Int J Cancer 77,
24–28, 1998.
53. Nicodemus KK, Jacobs DR Jr., and Folsom AR: Whole and refined grain
intake and risk of incident postmenopausal breast cancer (United States).
Cancer Causes Control 12, 917–925, 2001.
54. Goldin BR, Adlercreutz H, Gorbach SL, Woods MN, Dwyer JT, Conlon
T, Bohn E, and Gershoff SN: The relationship between estrogen levels and
diets of Caucasian American and Oriental immigrant women. Am J Clin
Nutr 44, 945–953, 1986.
55. Cohen LA, Zhao Z, Zang EA, Wynn TT, Simi B, and Rivenson A: Wheat
bran and psyllium diets: effects on N-methylnitrosourea-induced mammary
tumorigenesis in F344 rats. J Natl Cancer Inst 88, 899–907, 1996.
56. Martin-Moreno JM, Willett WC, Gorgojo L, Banegas JR, Rodriguez-
Artalejo F, Fernandez-Rodriguez JC, Maisonneuve P, and Boyle P: Dietary
fat, olive oil intake and breast cancer risk. Int J Cancer 58, 774–780, 1994.
57. Trichopoulou A, Katsouyanni K, Stuver S, Tzala L, Gnardellis C, Rimm
E, and Trichopoulos D: Consumption of olive oil and specific food groups
in relation to breast cancer risk in Greece. J Natl Cancer Inst 87, 110–116,
1995.
58. la Vecchia C, Negri E, Franceschi S, Decarli A, Giacosa A, and Lipworth
L: Olive oil, other dietary fats, and the risk of breast cancer (Italy). Cancer
Causes Control 6, 545–550, 1995.
59. Masala G, Ambrogetti D, Assedi M, Giorgi D, Del Turco MR, and Palli
D: Dietary and lifestyle determinants of mammographic breast density. A
longitudinal study in a Mediterranean population. Int J Cancer 118, 1782–
1789, 2006.
DIETARY FACTORS AND BREAST CANCER RISK 187
60. L´
eger CL, and Descomps B: Huile d’olive, coproduits de l’huilerie d’oliveet
olive-fruit: donn´
ees actuelles et perspectives concernant la relation aliment
et sant´
e. Agropolis Montpellier, France, 2000.
61. Block G, Hartman AM, Dresser CM, Carroll MD, Gannon J, and Gardner
L: A data-based approach to diet questionnaire design and testing. Am J
Epidemiol 124, 453–469, 1986.
62. Terry P, Suzuki R, Hu FB, and Wolk A: A prospective study of major dietary
patterns and the risk of breast cancer. Cancer Epidemiol Biomarkers Prev
10, 1281–1285, 2001.
63. Sieri S, Krogh V, Pala V, Muti P, Micheli A, Evangelista A, Tagliabue G,
and Berrino F: Dietary patterns and risk of breast cancer in the ORDET
cohort. Cancer Epidemiol Biomarkers Prev 13, 567–572, 2004.
64. Mannisto S, Dixon LB, Balder HF, Virtanen MJ, Krogh V, Khani BR,
Berrino F, van den Brandt PA, Hartman AM, et al.: Dietary patterns and
breast cancer risk: results from three cohort studies in the DIETSCAN
project. Cancer Causes Control 16, 725–733, 2005.
65. Fung TT, Hu FB, Holmes MD, Rosner BA, Hunter, DJ, Colditz GA, and
Willett WC: Dietary patterns and the risk of postmenopausal breast cancer.
Int J Cancer 116, 116–121, 2005.
66. Ronco AL, De Stefani E, Boffetta P, Deneo-Pellegrini H, Acosta G, and
Mendilaharsu M: Food patterns and risk of breast cancer: A factor analysis
study in Uruguay. Int J Cancer 119, 1672–1678, 2006.
APPENDIX
Fruit: apricots, strawberries, cherries, plums, melons, grapes,
figs, bananas, grapefruits, tangerines, oranges, peaches, kiwis,
apples, and pears.
Cooked vegetables: vegetable soup, carrots, cabbages
(red, green), zucchinis, eggplants, peppers, celery, spinaches,
chicories, tomatoes, and green beans.
Raw vegetables: lettuces, chicories, celeriac, cabbages (red,
green, white), peppers, cucumbers, tomatoes, radishes, and car-
rots.
Cereals: bread (white, whole meal, farmhouse), toast, corn
flakes, rice, and pasta (whole or white), semolina and wheat.
Legumes: dried beans, peas, chickpeas, split peas, and broad
beans.
Meat: beef, pork, lamb, mutton, and veal.
Dairy products: milk (whole, skimmed, semi-skimmed),
cheeses (soft, hard, milk), butter, and yoghurt.
Sea foods: fish, shellfish, and mollusk.
