Low-risk diet for breast cancer in Italy

Abstract

To define a low-risk diet for breast cancer in Italy, a multicentric case-control study of 2569 incident cases of breast cancer and 2588 controls from Italy was analyzed, A logistic regression model was applied to the estimated intake of five macronutrients and used to compute a diet-related risk score (RS). The pattern of macronutrient and food group intake across RS deciles was defined, The mean of diet-related RSs across subsequent risk deciles ranged from 0.83 to 1.44. Total energy intake first decreased slightly, from the first to the second decile, and then increased, mostly in the last three risk deciles. Intake of starch increased in absolute and relative terms, whereas saturated fat intake rose in absolute terms but remained stable as a proportion, A relative decline was observed for unsaturated fat and sugars, with a hint, however, of U-shape effect, From a food group viewpoint, there was a marked increase in the intake of bread and cereal dishes, cakes and desserts, and refined sugar across subsequent deciles, whereas the consumption of vegetables, olive and seed oils, and fruit decreased.

Full text

Vol. 6. 875- 879. November 1997 Cancer Epidemiology, Biomarkers & Prevention 875
Low-Risk Diet for Breast Cancer in Italy’
Silvia Franceschi,2 Carlo La Vecchia, Antonio Russo,
Eva Negri, Adriano Favero, and Adriano Decarli
Servizio di Epidemiologia. Centro di Riferimento Oncologico, 33081 Aviano
[S. F.. A. R.. A. Fl; Istituto di Ricerche Farmacologiche ‘Mario Negri,” 20157
Milan [C. L. V.. E. NI: Istituto di Statistica Medica e Biometria, Universit#{224} di
Milano, 20133 Milan [C. L. V., A. Dl: and Istituto Nazionale Tumori. 20133
Milan [A. Dl. Italy
Abstract
To define a low-risk diet for breast cancer in Italy, a
multicentric case-control study of 2569 incident cases of
breast cancer and 2588 controls from Italy was analyzed.
A logistic regression model was applied to the estimated
intake of five macronutrients and used to compute a diet-
related risk score (RS). The pattern of macronutrient and
food group intake across RS deciles was defined. The
mean of diet-related RSs across subsequent risk deciles
ranged from 0.83 to 1.44. Total energy intake first
decreased slightly, from the first to the second decile, and
then increased, mostly in the last three risk deciles.
Intake of starch increased in absolute and relative terms,
whereas saturated fat intake rose in absolute terms but
remained stable as a proportion. A relative decline was
observed for unsaturated fat and sugars, with a hint,
however, of U-shape effect. From a food group viewpoint,
there was a marked increase in the intake of bread and
cereal dishes, cakes and desserts, and refined sugar
across subsequent deciles, whereas the consumption of
vegetables, olive and seed oils, and fruit decreased.
Introduction
Breast cancer rates have been relatively low in Mediterranean
countries as compared to most other Western ones (1). The
traditional Mediterranean diet is relatively rich in carbohy-
drates, vegetables, and fruits, but the total fat intake, in pro-
portional terms, is not particularly low. Olive oil, however,
predominates among seasoning fats (2).
Studies of the relationship between diet and breast cancer
have been focused thus far on defining the specific role of
various food groups, macronutrients (chiefly fat), and micro-
nutrients (3). In a large case-control investigation from six
different Italian areas, from which this report derives, breast
Received 1 1/4/96; revised 3/6/97; accepted 3/12/97.
The costs of publication of this article were defrayed in part by the payment of
page charges. This article must therefore be hereby marked advertisement in
accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
IThis work was conducted within the framework of the Consiglio Nazionale delle
Ricerche (Italian National Research Council) Applied Projects “Clinical Appli-
cations of Oncological Research” (contracts 96.00701 .PF39, 96.00759.PF39, and
95.00345.PF39) and with a contribution from the Italian Association for Research
on Cancer.
2To whom requests for reprints should be addressed, at Servizio di Epidemio-
logia. Centro di Riferimento Oncologico. Via Pedemontana Occidentale, 33081
Aviano (PN), Italy. Phone: 39-434-659232; Fax: 39-434-659222.
cancer risk was positively associated with the intake of bread
and cereal dishes, pork and processed meats, and refined sugar
(4). Conversely, intakes of milk, poultry, fish, potatoes, and raw
vegetables were negatively associated with risk (4). Oils and
fats did not increase cancer risk, whereas olive oil and selected
seed oils seemed to provide some protection (5). Among major
energy sources, high starch intake was associated with an
increased risk, whereas monounsaturated and polyunsaturated
fats were associated with a decreased risk (6). High intake of
several micronutrients, particularly 3-carotene, vitamin E, and
calcium, seemed protective against breast cancer (7). These
findings are in agreement with other investigations from South-
em Europe (8, 9).
Rarely have multivariate approaches been used to identify
low-risk dietary patterns (10). Preventive strategies, however,
would benefit from understanding the overall dietary pattern,
which could reduce breast cancer occurrence. It is therefore of
interest to identify, in terms of macronutrients and food sources,
the type of diet related to the lowest breast cancer risk.
Materials and Methods
Data were derived from a case-control study of breast cancer,
carried out in six Italian geographical areas, whose design and
methods have been described elsewhere (4-7). Briefly, cases
were 2569 women with incident histologically confirmed breast
cancer, ages 23-74 years (median age, 55 years). Controls were
2588 women, ages 20-74 years (median age, 56 years), hos-
pitalized for acute, nonneoplastic, non-hormone-related condi-
tions (i.e., 22% traumas, 33% other orthopedic diseases, 15%
acute surgical conditions, I 8% eye diseases, and I 2% other
miscellaneous diseases). Between 1991 and 1994, trained in-
terviewers administered a validated food frequency question-
naire ( 1 1 ,12) including questions on 78 foods and recipes
grouped into six sections and specific questions on individual
fat intake pattern. Italian food composition tables, appropriately
checked and supplemented with other published data ( 1 3), and
information from the manufacturers were used to estimate the
intake of selected nutrients.
Food items and recipes were grouped into 16 food groups:
milk; bread and cereal dishes; soups; eggs: poultry; red meat;
pork and processed meat; fish; cheese; raw vegetables; cooked
vegetables; potatoes; citrus fruits; other fruits; cakes and des-
serts; and refined sugar. For each subject, the weekly frequency
of intake of each food group was computed and translated into
grams/day by multiplying the intake frequency by individual
portion size. For refined sugar, it corresponded to four tea-
spoons. Estimates of intake in grams of olive oil, seed oils, and
butter used as a condiment were derived from the frequency and
self-assessed quantity of use as well as from estimates of fat
included in various foods and recipes (4-6).
Data Analysis. We fitted a multiple logistic regression equa-
tion that included terms for age in quinquennia. center, major
correlates of breast cancer risk in our study (i.e. ,years of
education, parity, menopausal status, and quintile of alcohol
consumption), and five macronutrients (starch, sugar. protein.

I
3000 -
2500
 2000
S
.
. 1500
U
1000
500
0’ 0.83 0.93 0.98 1.01 1.05 1.08 1.12 1.17 1.23 1.44
DI -relc#{225}edrIsk sre
I
UI
U I U I
PA
I
I
II
I
Stact
 Sugcrs
0Protdn
 ScttxctedFct
 Urssdtrcted Fc*
876 Low.Risk Diet for Breast Cancer
3The abbreviation used is: RS, risk score.
Fig. 1. Total energy and proportional intake of five macronutrients by decile of breast cancer diet-related RS (Italy, 1991-1994).
Table 1 Intake of macronu talents and f ood groups (i n grams) by decile of breast cancer diet-related RS (I taly. 1991-1994)
Macronutrient or food group
Mean dally intake (g)
Decile of breast cancer diet-related risk score
1 2 3 4 5 6 7 8 9 10
Starch 107 114 122 134 134 143 154 169 185 241
Sugars 116 98 96 97 92 94 92 95 99 113
Protein 90 83 83 84 82 84 83 88 91 105
Saturated fat 25 24 24 24 24 25 24 27 28 33
Unsaturated fat 46 39 37 37 35 35 34 37 37 41
Bread and cereal dishes 124 134 145 161 163 175 188 205 229 295
Soups 73 75 76 79 72 75 79 84 83 78
Eggs 13 13 12 13 13 12 12 12 13 14
Poultry 72 57 55 54 50 50 47 48 47 46
Red meat 77 69 67 65 63 62 63 67 67 69
Pork and processed meat 27 25 24 24 23 23 23 23 23 26
Fish 45 38 33 34 32 30 30 28 27 26
Cheese 44 46 48 48 49 54 51 55 60 68
Milk 218 185 189 184 176 193 162 180 187 220
Raw vegetables 196 162 153 150 134 132 129 127 121 121
Cooked vegetables 132 114 113 110 103 100 96 102 100 102
Potatoes 39 40 38 40 41 43 39 40 43 47
Citrus fruits 137 120 1 13 1 10 108 102 98 89 93 91
Other fruits 421 330 327 321 293 275 278 260 281 276
Cakes and desserts 39 36 38 44 44 47 54 60 66 102
Refined sugar 90 80 85 94 88 93 92 102 98 130
saturated fat, and unsaturated fat; Ref. 14). Unsaturated fats
were computed by adding polyunsaturated fat and monounsat-
urated fat. For each woman (j; case or control), a diet-related
RS3 (RS) was defined by disentangling estimated parameters
for each macronutrient, i, from the logistic equation as follows:
RS =exP(±biZv)
where Z, is the intake (in kcal) and b1 is the regression coef-
ficient of macronutrient ifor woman j. Women were then
classified by decile of RS distribution of cases and controls
combined.
To validate our diet-related RS, we used the jackknife
method (15, 16). Each subject was removed in turn from the
dataset, and the logistic regression model was carried out on the
remaining (n -1) subjects. The new set of logistic regression
parameters was used to compute an independent RS for the
subject left out of the dataset. A Pearson’s correlation coeffi-
cient of 0.9994 was found between the RS we used and those
derived from the jackknife method.
(1) The mean composition of the diet of cases and controls
combined was evaluated in each RS decile in terms of macro-
nutrients, food groups, and types of fat used as a condiment.
Results
Fig. 1 gives total energy intake and the proportion of energy
from starch, sugars, protein, and saturated and unsaturated fats,
according to the decile of diet-related RS. RS ranged from 0.83
to 1 .44. After a decline from the first to the second decile and
a plateau at approximately 1800 kcal/day, total energy in-

-U- Bread and cereals
-.- Fruits
-A- Vegetables
-y- Refined sugars
-1k- Beef and pork
-..- Cheese
-+- Cakes and desserts
40 -
35 -
30 -
25 -
20 -
15 -
10 -
5-
+
V
V
-- -. ±   
--I’1 I‘I‘I‘I#{149} I#{149} I#{149} I#{149} I#{149} I
0.83 0.93 0.98 1.01 1.05 1.08 1.12 1.17 1.23 1.44
Cancer Epidemiology, Biomarkers & Prevention 877
4,
4,
U)
0)
C
4,
U)
>,
0
C
4,
0#{149}
4,
(4
C
Diet -related risk score
Fig. 2. Intake frequency (servings/week) of selected food groups by decile of breast cancer diet-related RS (Italy, 1991-1994).
creased in the last three deciles, most notably in the last risk
decile, to 2515 kcal/day. Starch intake approximately doubled
in both absolute (from 447 to 1003 kcal/day) and relative terms,
from 24% of total energy in the RS lowest decile to 40% in the
highest one. Saturated fat intake increased with increasing RS
decile (from 228 to 297 kcal/day) but remained stable as a
proportion (12%). A relative decline was seen from the lowest
to the highest RS decile for sugars (from 23 to 17%) and
unsaturated fats (from 22 to 15%). Thus, the lowest RS decile
showed a specific pattern in which the intake of sugars and
unsaturated fat predominated, accounting together for 45% of
energy intake.
Table 1 gives the intake in grams/day of major macronu-
trients and food groups by RS decile. The absolute intake of
starch and, to a lesser extent, saturated fat increased across
subsequent risk deciles. The pattern was less clear for other
macronutrients. Intakes of sugars and unsaturated fat showed a
U-shaped distribution. There was an increase in bread and
cereal dishes, cheese, cakes and desserts, and refined sugar
across subsequent RS deciles, particularly in the last one. Con-
versely, the consumption of different types of vegetables and
fruits tended to decrease substantially. Also poultry and fish
consumption declined across RS deciles. Thus, the lowest risk
decile was characterized by especially high intakes of vegeta-
bles and fruits. Risk patterns were similar in pre- and postm-
enopausal women (data not shown).
To further describe a low-risk diet, Fig. 2 gives the mean
number of average servings/week for selected food groups by
decile of dietary-related RS. Women with a RS below 1.00
tended to eat fruits and vegetables more or equally often than
bread and cereal dishes. In the highest RS deciles, and again,
especially in the last one, a rise in the consumption of bread and
cereal dishes, cakes and desserts, and refined sugar was evident.
Fig. 3 shows daily intake of different types of fat as a
condiment. The intake of olive oil and seed oils fell substan-
tially, whereas butter consumption rose slightly across subse-
quent RS deciles.
Discussion
The presented approach attempts to provide new insights on the
relationship between dietary habits and breast cancer risk, with
special reference to a Mediterranean population. It goes beyond
reliance on odds ratios for specific dietary components (4-7),
allowing a combined evaluation of the influence of nutrients
and food sources. Because of the persisting ignorance on cancer
mechanisms, vegetable and fruit intake was preferred to micro-
nutrient intake as an indicator of the intake of potentially
beneficial compounds.
It is clear that an increase in total energy intake is a
component of the three highest RS deciles, most notably of the
last one, whereas no material difference was observed between
the second and seventh decile. The predominance of starch and
hence bread and cereal dishes (i.e. ,more than 80% of starch

40 -
35 -
30 -
-u-Olive oil
-.- Seed oils
-A---- Butter 1
25 -
20 -
15-
10 -
5-
0
S
A- A-
#{149} - I‘I
1.17 1.23 1.44
878 Low.Risk Diet for Breast Cancer
4,
0
U,
E
4,
0)
4,
4,
C
,ll’I’l’l’l’ 
0.83 0.93 0.98 1.01 1.05 1.08 1.12
Diet -related risk score
Fig. .1. Daily intake (grams/day) of fat as a condiment by decile of breast cancer diet-related RS (Italy. 1991-1994).
intake in study women) increased appreciably across subse-
quent RS deciles. In a population such as the Italian one, in
which refined cereals, virtually all from wheat. are consumed,
high starch intake is an indicator of a diet rich in energy but
potentially poor in beneficial micronutrients (4, 7). Increased
intakes of cakes and desserts and refined sugar were other
features of the highest risk deciles, chiefly of the last one.
Interestingly, refined cereals as well as refined sugar can pro-
duce glycemic overload, compensatory increase of blood insu-
un, and, in the long term, insulin resistance (17). These condi-
tions may lead to cellular growth promotion in the breast via
specific hormones or growth factors (18).
Conversely, the intake of sugars, unsaturated fat, and, on
the food side, fruits and vegetables declined across RS deciles.
Saturated fat constituted the macronutrient whose intake, as a
proportion of total energy intake, was the most stable (12%)
across risk deciles. Still, women in the highest risk decile ate
about a third more saturated fat than those in the lowest one.
This is compatible with an adverse effect of saturated fat intake
on breast cancer beyond a certain threshold, which may or may
not be reached in a specific population (18).
The definition of a low-risk diet for breast cancer in terms
of foods or food groups is of special relevance from a practical
viewpoint. In the lowest-RS decile, women consumed 26 serv-
ings/week of fruits and 20 servings/week of vegetables. Serv-
ings of bread and cereal dishes were consumed somewhat less
frequently, i.e., 16 servings/week. Conversely, 37 servings/
week of bread and cereal dishes were reported by women in the
highest risk decile, as compared to 18 servings/week of fruits
and 13 servings/week of vegetables.
Olive oil was the chief source of unsaturated fat and
markedly predominated among seasoning fats in the lowest risk
deciles. Fruit intake, the major source of sugars in our popu-
lation, showed a similar curve. However, after some plateau,
the absolute intake of unsaturated fat and sugars increased again
in the three highest risk deciles. Therefore, the specific effect of
some macronutrients (e.g., sugars) may vary depending on their
food sources (e.g. ,fruits in low-risk women, but refined sugar
in high-risk ones).
The use of the same dataset to construct the diet-related RS
and to evaluate the risk differentials associated with it is open
to criticism. The low-risk diet presented here awaits confirma-
tion from independent data from a similar Southern European
population. Furthermore, because RS deciles were derived from
the comparison of cases and controls, this study shares the
strengths and limitations of hospital-based case-control studies
(19). Although case recruitment was not population based,
cases were identified in the major public hospitals of the area
under surveillance, reducing the scope for selection bias. With
reference to the control group, only acute conditions unrelated
to known or likely risk factors for breast cancer or to dietary
modifications were included. Furthermore, a separate compar-
ison of cases with major diagnostic categories of controls
produced mutually consistent results. Conversely, the hospital-
based design kept refusals to a minimum and probably im-
proved the comparability of diet recall by cases and controls.
The food frequency questionnaire had been proven reproduci-
ble (1 1) and valid (12).
In conclusion, the presented model is easily transferable
into preventive measures, because it points to an overall
protective dietary pattern (chiefly, high vegetable and oil
intake) that may help in reconciling some uncertainties and

Cancer Epidemiology, Biomarkers & Prevention 879
discrepancies of previous studies (e.g. ,with respect to the
role of fat; Ref. 20).
Acknowledgments
We thank Dr. Renato Talamini and Olinda Volpato for study coordination and
Luigina Mci for editorial assistance.
References
1. Levi, F., Lucchini, F., and La Vecchia, C. Worldwide patterns of cancer
mortality. 1985-1989. Eur. J. Cancer Prey., 3: 109-143. 1994.
2. Trichopoulou, A. Olive oil and breast cancer. Cancer Causes Control, 6:
475-476, 1995.
3. Hunter, D. J., and Willett, W. C. Diet, body build, and breast cancer. Annu.
Rev. Nutr., 14: 393-418, 1994.
4. 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.
5. 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.
6. Franceschi, S., Favero, A., Decarli, A., Negri, E., La Vecchia, C., Ferraroni,
M., Russo, A., Salvini, S., Amadori, D., Conti, E., Montella, M., and Giacosa, A.
Intake of macronutrients and risk of breast cancer. Lancet, 347: 135 1-1356, 1996.
7. Negri, E., La Vecchia, C., Franceschi, S., D’Avanzo, B., Talamini, R.,
Parpinel, M. T., Ferraroni, M., Filiberti, R., Montella, M., Falcini, F., Conti, E.,
and Decarli, A. Intake of selected micronutrients and the risk of breast cancer. Int.
J. Cancer, 65: 140-144, 1996.
8. Martin-Moreno, J. M., Willett, W. C., Gorgojo, L., Benegas, J. R., Rodriguez-
Artalejo, F., Fernandez-Rodriguez, J. C., Maisonneuve, P., and Boyle, P. Dietary
fat, olive oil intake and breast cancer risk. Int. J. Cancer, 58: 774-780, 1994.
9. 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. NatI. Cancer Inst.. 87:
110-116, 1995.
10. Katsouyanni, K., Trichopoulos, D., Boyle, P.. Xirouchaki, E.. Trichopulou.
A., Lisseos, B., Vasilaros, S.. and McMahon, B. Diet and breast cancer: a
case-control study in Greece. Int. J. Cancer, 38: 815-820, 1986.
11. Franceschi, S., Negri, E., Salvini, S., Decarli, A., Ferraroni, M., Filiberti, R.,
Giacosa, A.. Talamini, R., Nanni, 0., Panarello, G., and La Vecchia, C. Repro-
ducibility of an Italian food frequency questionnaire for cancer studies: results for
specific food items. Eur. J. Cancer, 29A: 2298-2305, 993.
12. Decarli, A., Franceschi, S., Ferraroni, M., Gnagnarella, P., Parpinel, M. T.. La
Vecchia, C., Negri, E., Salvini, S., Falcini, F., and Giacosa. A. Validation of a
food-frequency questionnalre to assess dietary intakes in cancer studies in Italy.
Results for specific nutrients. Ann. Epidemiol.. 6: 1 10-1 18, 1996.
13. Salvini, S., Gnagnarella. P., Parpinel, M. T., Boyle, P., Decarli, A., Ferraroni,
M., Giacosa. A., La Vecchia, C., Negri, E., and Franceschi, S. The food corn-
position database for an Italian food frequency questionnaire. J. Food Compos.
Anal., 9: 57-71, 1996.
14. Baker, R. J., and Nelder, J. A. The Generalized Linear Interactive Model
System, Release 3. Oxford, United Kingdom: Numerical Algorithms Group.
1978.
15. Efron, B. The Jackknife, the Bootstrap and Other Re-Sampling Plans. Phil-
adelphia: Society for Industrial and Applied Mathematics, I 982.
16. Tomasson, H. Risk scores from logistic regression: unbiased estimates of
relative and attributable risk. Stat. Med.. 14: 1331-1339, 1995.
17. Kaaks, R. Nutrition, hormones, and breast cancer: is insulin the missing link?
Cancer Causes Control, 7: 605-625, 1996.
18. Wynder, E. L., Cohen, L. A., and Rose, D. P. Dietary fat and breast cancer:
where do we stand on the evidence? J. Clin. Epidemiol.. 47: 217-222. 1994.
19. Breslow, N. E., and Day, N. E. Statistical Methods in Cancer Research. I. The
Analysis of Case-Control Studies. IARC Sci. PubI. No. 32. Lyon, France: IARC,
1980.
20. Decarli, A., Favero, A., La Vecchia, C., Russo, A., Ferraroni, M., Negri, E.,
and Franceschi, S. Macronutrients, energy intake, and breast cancer risk: 1mph-
cations from different models. Epidemiology, 8: 425-428, 1997.