8. Caloric Restriction and Incidence of Breast Cancer, The Journal of the American Medical Association.
Copyright 2004 by the American Medical Association. All Rights Reserved.
Volume 291(10), 10 March 2004, p 1226-1230
Caloric Restriction and Incidence of Breast Cancer
Michels, Karin B. ScD, MSc, MPH; Ekbom, Anders MD, PhD
Obstetrics and Gynecology Epidemiology Center, Brigham and Women's Hospital,
Harvard Medical School, Boston, Mass (Dr Michels); Department of Epidemiology,
Harvard School of Public Health, Boston (Drs Michels and Ekbom); and Unit of
Clinical Epidemiology, Department of Medicine, Karolinska Hospital, Karolinska
Institute, Stockholm, Sweden (Dr Ekbom).
Corresponding Author: Karin B. Michels, ScD, MSc, MPH, Obstetrics and Gynecology
Epidemiology Center, Brigham and Women's Hospital, Harvard Medical School, 221
Longwood Ave, Boston, MA 02115 (email@example.com).
Context: Restricting caloric intake is one of the most effective ways to extend
lifespan and to reduce spontaneous tumor occurrence in experimental animals, but
whether similar associations hold in humans has not been appropriately studied.
Objective: To determine whether caloric restriction in early life reduces the
risk of invasive breast cancer.
Design, Setting, and Participants: Retrospective cohort study using data from
the Swedish Inpatient Registry, the Swedish Cancer Registry, the Swedish Death
Registry, and the Swedish Fertility Registry. Participants were 7303 Swedish
women hospitalized for anorexia nervosa prior to age 40 years between 1965 and
1998. Women were excluded (n = 31) if they were diagnosed with cancer prior to
their first discharge from hospitalization for anorexia nervosa.
Main Outcome Measure: Incidence of invasive breast cancer.
: Compared with the Swedish general population, women hospitalized for anorexia
nervosa prior to age 40 years had a 53% (95% confidence interval [CI], 3%-81%)
lower incidence of breast cancer; nulliparous women with anorexia nervosa had a
23% (95% CI, 79% higher to 75% lower) lower incidence, and parous women with
anorexia nervosa had a 76% (95% CI, 13%-97%) lower incidence.
Conlusions: Severe caloric restriction in humans may confer protection from
invasive breast cancer. Low caloric intake prior to first birth followed by a
subsequent pregnancy appears to be associated with an even more pronounced
reduction in risk.
RESTRICTING CALORIC INTAKE is one of the most effective ways to extend lifespan
and to reduce spontaneous tumor occurrence in experimental animals. 1,2 Caloric
restriction has an important protective role in experimental mammary carcinogenesis.
3,4 A recent meta-analysis summarized the available evidence of the effect of
energy restriction on spontaneous mammary tumors in mice. 5 The combined
estimate for the 14 included studies implied that the energy-restricted animals
developed 55% (95% confidence interval [CI], 41%-69%) fewer mammary tumors than
did those in the control groups, irrespective of the type of restricted
nutrient. 5 The authors called for dietary cohort studies to gain insight into
the effects of energy restriction on development of breast cancer in humans. 5
Reduced caloric intake in experimental animals has been found to be accompanied
by lower levels of circulating insulin, insulin-like growth factors I and II,
and epidermal growth factor, as well as by modified cellular responsiveness to
estrogens, enhanced immunologic responsiveness, alterations in cell cycle
regulation, lower rates of cellular proliferation, increased DNA repair, reduced
expression of oncogenes, and enhanced expression of tumor suppressor genes. 5-10
Energy restriction may be crucial during early life and prior to first
pregnancy, when mammary tissue is especially susceptible to carcinogenic
processes. 11,12 This hypothesis is supported by the observation that greater
height, which although genetically influenced still reflects nutritional status
and hence caloric intake during growth, is associated with an increased
incidence of breast cancer. 13,14
Energy restriction is difficult to study in humans. One marker of caloric
restriction is anorexia nervosa, an illness that occurs generally during
adolescence or early adulthood and is characterized by very low caloric intake,
low body mass index (BMI), and amenorrhea.
We conducted a retrospective cohort study in Sweden to evaluate whether women
with anorexia nervosa severe enough to require hospitalization and treatment
have a lower incidence of breast cancer than expected in the general population.
A cohort of Swedish women who had been hospitalized for anorexia nervosa between
1965 and 1998 was retrospectively formed and followed up for the occurrence of
breast cancer by linkage with the Swedish Cancer Registry, the Swedish Death
Registry, and the Emigration Registry using the National Registration Number, a
unique identifier for each resident of Sweden. We ascertained cancers diagnosed
after 1965, the date and cause of death after 1965, and date of censoring due to
emigration. The observed cancer numbers were compared with the expected cancer
incidence rates based on rates in the female Swedish population in 5-year age
intervals during the same calendar period. The study was approved by the
institutional review board of the Brigham and Women's Hospital, Boston, Mass,
and by the Office for Human Research Protection at the Karolinska Institute,
Study Population and Ascertainment of Exposure
Using the Swedish Inpatient Registry, we identified all women in Sweden who were
hospitalized and treated for severe anorexia nervosa between 1965 and 1998. The
Swedish Inpatient Registry was initiated in 1965 and is composed of hospital-provided
somatic and psychiatric medical services in the Swedish population. Each record
contains the medical procedure, discharge diagnoses coded according to the
International Classification of Diseases, seventh through ninth revisions
(ICD-7: 1965-1968; ICD-8: 1969-1987; ICD-9: 1988 forward), and the National
Registration Number. The register is complete within Swedish counties and
covered 60% of the Swedish population in 1969, 75% in 1978, 85% in 1983, and the
entire Swedish population since 1987.
Anorexia nervosa is defined in this study as having received inpatient care for
the disease. A hospitalization for anorexia nervosa was identified by discharge
diagnoses ICD-7 codes 316.99 and 784.09, ICD-8 code 306.50, and ICD-9 code 307B.
This study was restricted to a first hospitalization for anorexia nervosa.
We identified 7303 women from the Swedish Inpatient Registry who were hospitalized
for anorexia nervosa prior to age 40 years between 1965 and 1998. This anorexia
population excluded 31 women who were diagnosed with cancer prior to their first
discharge from hospitalization for anorexia nervosa.
Ascertainment of Outcomes
All first primary incident cancers diagnosed among our anorexia cohort were
ascertained through record linkage with the Swedish Cancer Registry, which
contains all diagnoses of incident cancer among Swedish residents from January
1, 1958, to December 31, 2000, and its completeness has been evaluated to exceed
95%. 15,16 The reports list the National Registration Number and the cancer
diagnosis according to the ICD-7 codes.
Assessment of Effect Modifiers
The Swedish Fertility Registry has information on every birth in Sweden among
all women born in 1925 and thereafter. We obtained information on parity status
from this registry for the members of our anorexia cohort.
Person-time of follow-up was calculated for each member of the anorexia cohort
from the date of first discharge from hospitalization for anorexia nervosa until
the diagnosis of cancer, death, emigration, or December 31, 2000 (end of
follow-up), whichever occurred first. The expected number of cancer cases in our
cohort was calculated by multiplying the 5-year age interval- and calendar
year-specific cancer rates of the general Swedish population by the person-time
accumulated in each interval of the exposed population after censoring for
death, emigration, or end of follow-up.
A standardized incidence ratio (SIR) was calculated as the ratio of the observed
number of first primary cancer cases to the number expected. 17 A 95% CI around
the SIR was constructed based on the assumption that the observed number follows
a Poisson distribution. 17
Effect modification by parity status was assessed by dividing the cohort into
nulliparous and parous women based on the parity information obtained from the
Swedish Fertility Registry. Parous women contributed person-time as nulliparae
until they gave birth.
The distribution of women hospitalized for anorexia nervosa in our cohort by age
and calendar year of discharge diagnosis is presented in Table 1. The majority
of anorexia cases (73%) were diagnosed prior to age 20 years.
Table 1. Distribution of Women Hospitalized for Anorexia Nervosa in Sweden
Between 1965 and 1998, by Age at Discharge and Calendar Year of Discharge
Among the 7303 women who were diagnosed with anorexia prior to age 40 years, we
identified 52 women who were diagnosed with any type of cancer during 96 887
person-years of follow-up between 1965 and 2000; the expected number of cancer
cases was 56.6. The SIR for all types of cancers combined was 0.92 (95% CI,
0.69-1.21) (Table 2).
Table 2. Standardized Incidence Ratios for Cancer Among Parous (n = 1942)
and Nulliparous (n = 5361) Swedish Women With a Hospital Discharge Diagnosis of
Anorexia Nervosa Between 1965 and 1998, Overall and by Parity Status
Among women who were diagnosed with anorexia prior to age 40 years, the SIR for
breast cancer was 0.47 (95% CI, 0.19-0.97) (Table 2). Seven women in this group
developed breast cancer. The expected number of breast cancer cases was 14.8;
thus, women in this group had a 53% (95% CI, 3%-81%) lower incidence of breast
cancer than the Swedish general population. Among women diagnosed with anorexia
nervosa prior to age 20 years, no case of breast cancer occurred; the number of
breast cancer cases expected in this group was 2.7. Among women diagnosed
between the ages of 20 and 29 years, 4 cases were diagnosed and 6.4 expected;
among women diagnosed with anorexia between the ages of 30 and 39 years, 3 cases
were observed and 5.7 expected.
In our anorexia cohort, 73% of women remained nulliparous throughout the
observation period. Among women who remained nulliparous, the SIR for breast
cancer was 0.77 (95% CI, 0.25-1.79) during 78 984 person-years of follow-up;
there were 5 cases diagnosed and 6.5 expected, and thus women in this group had
a 23% (95% CI, 79% increased to 75% decreased) reduced incidence of breast
cancer. The corresponding SIR value among parous women was 0.24 (95% CI,
0.03-0.87) during 17 903 person-years of follow-up; there were 2 cases diagnosed
and 8.3 expected, and thus women in this group had a 76% (95% CI, 13%-97%)
reduced incidence (Table 2).
Analyses of other hormone-dependent cancers such as uterine or ovarian cancer
did not reveal a significant inverse association, but statistical power was
limited (data not shown).
Among this Swedish cohort of women with a hospital discharge diagnosis of
anorexia nervosa prior to age 40 years, we found a significant decrease in
breast cancer incidence compared with the general Swedish female population of
comparable age and birth cohort. Because anorexia nervosa that requires
hospitalization is associated with severe caloric restriction during a prolonged
period of early life, we conclude that such starvation during adolescence and
early adulthood may impact on mechanisms crucial for development of breast
cancer. The majority of breast cancers in this cohort arose in premenopausal
women due to the age structure of the cohort.
These findings confirm observations made in rodents in which caloric restriction
has been a very effective measure to reduce cancer incidence. We are aware of
only 1 other study in which the association between anorexia nervosa and cancer
incidence was considered among humans. A study conducted in Denmark took a
similar approach, linking the Danish Psychiatric Case Register and the National
Registry of Patients to the Danish Cancer Registry. 18 The authors report an SIR
of 0.80 (95% CI, 0.52-1.18) for overall cancer incidence among 2151 women with a
hospital discharge diagnosis of anorexia nervosa based on 25 observed cases. In
this cohort, statistical power was limited to consider site-specific cancers.
The SIR observed for breast cancer was 0.8 (95% CI, 0.3-1.7), but the study
population was not restricted to women who experienced anorexia prior to age 40
years. 18 In another study, prepubertal girls who were exposed to the Norwegian
famine in World War II and who consumed an average of 22% fewer calories had a
lower subsequent rate of breast cancer than women from earlier or later birth
Among women with anorexia in our cohort who later had 1 or more children the
risk of breast cancer was reduced by 76%. Among these women, we mimic the
environment of developing countries where women experience caloric restriction
but go on to become pregnant and deliver children. Rates of breast cancer in
developing countries are considerably lower than in most affluent countries. 20
It is conceivable that caloric restriction during early periods of life is
associated with decreased development of breast parenchyma and that subsequent
differentiation of breast cells during pregnancy confers stronger protection
than among women in industrialized countries. Hence, low rates of breast cancer
in developing countries may be due to low caloric intake prior to first birth,
and a particularly pronounced protection is conferred by subsequent pregnancy.
Among women with anorexia nervosa, later fertility does not seem to be
compromised, but fecundability is reduced. 21,22 In general, women with anorexia
nervosa may have delayed childbirth and thus lower parity. In our cohort, only
26% of women gave birth after a diagnosis of anorexia nervosa, but due to the
age structure of the cohort most women are still of childbearing age. In Sweden,
the average age at first birth is currently 30 years. 23
A number of possible mechanisms may underlie our observations. Caloric
restriction may have a direct effect on breast cell growth and development.
Prolonged caloric restriction can also affect the expression of various
oncogenes and tumor suppressor genes. Reduced levels of epidermal growth factor
expression, reduced ERBB2 levels, a decrease in cyclin D1 expression, and
increased p53 and p27 expression have been found in the mammary tissue of
chronically calorie-restricted rodents. 7-9
Furthermore, caloric restriction and anorexia have been found to reduce levels
of estrogen (which may be manifested in amenorrhea) 10,24 and insulin-like
growth factor I (IGF-I). 6,25 Recent research has identified IGF-I as an
important biomarker for the prediction of breast cancer. 26,27 IGF-I is a
powerful growth hormone whose serum and tissue levels peak during adolescence.
28 Anorexia nervosa might lead to lower tissue levels of IGF-I during a crucial
phase of mammary gland development and thus affect risk of breast cancer.
Anorexia nervosa is associated with amenorrhea and consequently a reduced number
of lifetime ovulations, which might confer protection from breast cancer. Women
with anorexia often enhance weight loss by strenuous physical activity. The
epidemiologic data on the association between physical activity and risk of
breast cancer are not conclusive, but more evidence supports an inverse relation
with postmenopausal breast cancer 29,30 than with premenopausal 31,32 breast
cancer. Some studies indicate a reduced risk of breast cancer among women who
exercised heavily during their teenage years. 33,34 Heavy exercise was
associated with the strongest risk reduction in breast cancer among women with
low BMI in at least 1 study. 35
Anorexia may be associated with considerable weight fluctuations, which have
also been found to protect from breast cancer in the animal model. 36 We are
aware of only 1 epidemiologic study that has considered weight cycling and
breast cancer among humans. 37 In a population-based case-control study
conducted in the United States, weight cycling was defined as losing 20 lb (9
kg) or more and gaining at least half of the lost weight back within 1 year; 1
cycle was sufficient to qualify an individual as a weight cycler. No association
was found with breast cancer.
The biological pathways of the association between body mass and the risk of
breast cancer are complex, with opposing effects of a high BMI on premenopausal
and postmenopausal risk of breast cancer. 38,39 A number of reports have related
a high BMI during adolescence with a decreased risk of breast cancer. 40-43
Coates and colleagues 44 have reported a U-shaped relation between relative
weight in adolescence and later risk of breast cancer: women who were either
much heavier or much thinner than average were at reduced risk. Both anorexia
nervosa and obesity during adolescence may be associated with anovulation and
thus a delayed onset of menarche. Alternatively, mechanisms underlying the
protection conferred by low and high caloric intake may differ; while very low
circulating levels of estrogen and other growth hormones might be protective,
fairly high levels might induce early differentiation of breast cells. 43
Our study has a number of limitations. The number of cases observed in our
cohort is limited. Furthermore, there is the possibility of unmeasured
confounding. It is unlikely, however, that established risk factors for breast
cancer, such as family history, would be related to anorexia nervosa. While our
cohort of women with anorexia may differ from the general population in some
aspects other than total caloric intake, women with anorexia who go on to become
pregnant are more similar to the general population, which lends support to our
conclusions. Reproductive factors (eg, age at menarche, parity, age at first
birth) and anthropometric variables (eg, height, BMI) might be in the causal
pathway of the association of interest; thus, we would not want to adjust for
them. Furthermore, lower parity and a later age at first birth would place women
with anorexia at a higher risk of breast cancer. In this Swedish cohort,
exposure and outcome were ascertained with high accuracy. Neither differential
nor nondifferential misclassification of exposure or disease are probable.
Therefore, it is unlikely that our results are due to bias.
Our observations suggest an important role for caloric intake in the etiology of
breast cancer and call for further research exploring the underlying mechanisms
of this association to elucidate whether it is primarily due to direct effects
of caloric restriction on breast cell growth and development, to amenorrhea and
associated hypoestrogenism, or to a decreased level of growth factors.