It has been hypothesized that caloric restriction can prevent cancer in humans,1 including breast cancer.2 However, the risk of breast cancer was increased in a dose–response manner in women who were exposed to a short but severe caloric restriction as children during the 1944–1945 Dutch famine.3 An earlier study of the Dutch famine also found statistically significantly increased risk of breast cancer in women exposed to dietary restriction during or shortly after the adolescent growth spurt period.4 Further studies of men and women exposed to the Dutch famine in childhood showed that total cancer risk, exclusive of breast cancer, was not affected by the famine in women5 and that famine was nonsignificantly associated with increased prostate cancer risk in men6 and decreased colon cancer risk in both men and women.7
A study examining female survivors of the German Occupation of Guernsey 1940–1945 found that women who remained in Guernsey during the Occupation experienced on an average delayed menarche and a nonsignificantly increased risk of breast cancer.8 Recent analyses of Jewish Holocaust survivors residing in Israel found significantly higher cancer risk in the exposed men and women, with the strongest effects seen for breast and colorectal cancers and for those exposed prior to adolescence.9
The starvation experienced by the population of Leningrad in 1941–1944 provides a unique opportunity to replicate and compare the results from the Dutch and other famine studies. The siege of Leningrad by German troops during World War II lasted from September 1941 to January 1944 and resulted in an acute food shortage and severe malnutrition.10 Food was rationed from the beginning of the war,11 and average daily rations during the most severe period in winter 1941/1942 provided around 300 kcal,12 compared with the daily rations of ∼700 kcal or less during the Dutch Hunter Winter3 or 1,200 kcal in occupied Guernsey.8
Material and methods
Participants in the US-USSR Lipid Research Clinics Program13–16 were selected randomly from the socially mixed Petrogradsky district of St. Petersburg (formerly Leningrad) in 1975–1977 (men) and 1980–1982 (women); the participation rate was 80%. As regards variables used in this analysis, a common protocol was used for men and women. Height and weight were measured at the lipid research clinic,14–16 and body mass index was calculated as weight/height2 and analyzed as quartiles. Cigarette smoking, alcohol consumption and social characteristics (education, ethnic group/nationality, occupation and marital status) were determined by interview. We used 2 categories of smoking (past or current smoker vs. never smoker) and 2 categories frequency of alcohol consumption (regular once a week or more often vs. less often or never). Education was classified in 3 categories as university, secondary and lower than secondary; ethnic group/nationality was defined as Russian vs. non-Russian (mainly Jewish, Belorussian and Ukrainian); occupation was defined as white-collar occupation vs. blue-collar occupation or other; and marital status was analyzed as married vs. other.
Information about whether the subjects lived in Leningrad during the siege was available for more than 99% of men and 91% of women. A total of 3,901 men born between 1916 and 1935, and 1,429 women born between 1910 and 1940, of whom 1,406 (36%) and 605 (42%), respectively, experienced the siege as children, adolescents or young adults, were followed until the end of 2005. Vital status was ascertained through contacts with the participants, relatives, neighbors, and address bureaux (population registry). Participants who emigrated or were lost before the end of follow-up in 2005 (4.5% of men and 4.7% of women) were censored at the last day of contact. Copies of death certificates and additional information from hospitals were obtained to register date of death and the underlying cause of death that a team of Russian physicians coded according to ICD 8. Data on smoking and alcohol were missing for 17% of men, all examined within the same 8-week period.
Effect of siege exposure on risk of death from cancer was studied using multivariable Cox regression in STATA,17 stratified by gender, and adjusted for age, smoking, alcohol, adult stature and social characteristics. Models with and without an interaction term for statistical interaction between siege exposure and age at exposure were compared using a likelihood ratio test. Crude analyses of siege exposure and cancer mortality were first performed in 3,901 men and 943 women born between 1916 and1935. Crude analyses were then repeated, and multivariable analyses performed in a sample restricted to 3,240 (83%) men and 932 (99%) women born 1916–1935 with no missing data on other studied variables. There were no significant differences in social characteristics or mortality between men and women for whom all data were available and those excluded from multivariable analyses.
We performed stratified analyses by year of birth in 3 categories identical to those previously used in an analysis of breast cancer in Dutch3 and Guernsey women.8 Subjects from the respective birth cohorts were aged 6–9, 10–18, and 19–25 years when exposed to the severest starvation in January 1942. Additional analyses of breast cancer death included 171 women born 1910–1915 and 315 women born 1936–1940.
Total mortality of the cohort were compared with mortality in the Russian Federation by indirect standardization, using gender-specific mortality rates for 5-year age groups (35–39, 40–44, … 80–84 years) and 5-year birth cohorts (1915–1919, 1920–1924, 1925–1929, 1930–1934) published in the Human Mortality Database.18
The initial study was based on an agreement between the US and Soviet governments, and mortality follow-up was approved by the ethics committee at the Institute of Experimental Medicine at the Russian Academy of Medical Sciences.
The cohort had significantly lower mortality compared to the national mortality rates18 in the Russian population (standardized mortality ratio 0.86; 95% CI, 0.83–0.90). In the analyzed sample of 3,901 men and 943 women born 1916–1935, a total of 404 deaths among women and 2,633 deaths among men were recorded, of which 105 (26.0%) and 687 (26.1%), respectively, were deaths from cancer (Table I). Cause of death was unknown for 3% of deceased men and 5% of women. Excluding prostate and breast cancer, cancer of lymphatic and hematopoietic tissues, pancreas and oesophagus were most common in men, and cancer of oesophagus, uterus and ovary were most common in women.
Table I. Descriptive Characteristics of Men and Women Born Between 1916 and 1935
Note that results in this column include subjects with missing data for some of the below variables, and the number of subjects varies between 3,243 and 3,901 for men and between 938 and 943 for women.
Other (ICD 8: 140–150, 152, 155–159, 164–173, 175–184, 186–209)
Mortality from breast cancer appeared to be increased in women exposed to the siege, and mortality from prostate cancer was nonsignificantly increased in exposed men. In both genders, mortality from colon cancer was nonsignificantly lower among the exposed. No consistent effect of the siege was seen for stomach cancer, respiratory cancer or other cancer sites (Table II). The association of siege exposure with breast cancer mortality in women was not confounded by social characteristics; adjustment for smoking and alcohol consumption, and adjustment for adult height and body mass index weakened the association only slightly (Table III). In the restricted sample, mortality from respiratory cancer was higher in men exposed to the siege. The association was of borderline significance in the fully adjusted model (Table III).
Table II. Hazard Ratios (with 95% CI) for Mortality from Cancer in 1975–2005 among 943 Russian Women and 3,901 Russian Men Born Between 1916 and 1935 and Exposed to the Siege of Leningrad During World War II, Compared with Subjects who did not Live in Leningrad During the Siege
Number of events
Number of events
HR, hazard ratio adjusted for exact age at examination.
Table III. Hazard Ratios (with 95% CI) for Mortality from Cancer in 1975–2005 among 932 Russian Women and 3240 Russian Men Born Between 1916 and 1935 and Exposed to the Siege of Leningrad During World War II, Compared with Subjects who did not Live in Leningrad During the Siege
Number of events
Number of events
HR1, hazard ratio adjusted for exact age at examination; HR2, hazard ratio adjusted for exact age at examination and social characteristics (education, occupation, ethnic group/nationality and marital status); HR3, hazard ratio adjusted for exact age at examination and behavioral characteristics (smoking, alcohol consumption); HR4, hazard ratio adjusted for exact age at examination and adult stature (adult height and body mass index); HR5, hazard ratio adjusted for exact age at examination, social and behavioral characteristics and adult stature.
The total number of breast cancer deaths among 1,429 women born 1910–1940 was 26. The association of siege exposure with breast cancer was most pronounced in 477 women who were between 10- and 18-year-old at the peak of starvation (HR, 9.7; 95% CI, 1.1–84.5) (Fig. 1). The age-adjusted HR for death from breast cancer in women exposed at age 1–5 years was 0.44 (95% CI, 0.05–3.68). There was only one case of breast cancer death among women aged 26–31 years at the peak of siege, and the hazard ratio could not be calculated for this cohort. The statistical interaction between age at exposure (1–9 years, 10–18 years, 19–31 years) and siege (exposed vs. unexposed) in their effect on breast cancer death was not significant (p = 0.115). There was no indication of an interaction between age at exposure to siege and mortality from other cancer types studied; however, the number of events in stratified analyses was very low (data not shown).
Women exposed to the siege at ages 10–18 were 1.7 cm (95% CI, 0.5–3.0) taller as adults. The HR for breast cancer in 469 women with complete data who were aged 10–18 at the peak of siege, adjusted for exact age at examination, was 9.9, 95% CI 1.1–86.5, and did not appear to be confounded by social or behavioral factors. Adjustments for women's adult height and body mass index did not lead to attenuation of the association either (HR, 11.1; 95% CI, 1.2–99.7, adjusted for age and adult stature). We noted a positive, statistically nonsignificant association of height with breast cancer in exposed women and the youngest cohorts, and a weak negative, statistically nonsignificant association of height with breast cancer in women not exposed to siege and the older cohorts (data not shown).
None of the main cancer groups analyzed was statistically significantly associated with exposure to siege. In analyses stratified by age at exposure, women aged 10–18 years at the peak of siege had higher risk of death from breast cancer.
Limited statistical power, misclassification and selection bias are central issues in the interpretation of our results. The cohort had lower mortality compared with national mortality rates18 in the Russian population. The fact that this is a cohort of men and women who responded positively to a call for health examination explains in all likelihood a substantial part of this difference. Lower mortality of the cohort is also consistent with documented differences in adult mortality between urban and rural areas of the Russian Federation,19 and the generally better living conditions and access to health services in St. Petersburg.
The age at start of follow-up was on an average higher in women, and we have missed deaths from cancer that occurred at younger ages. The relatively small number of deaths from cancer, particularly in women, prevented us from performing more complex analyses. It is difficult to directly address the selection processes related both to the survival of the siege period and return or immigration of the control group to the city after the war. We were, however, able to adjust our results for a number of social, demographic and behavioral factors, including smoking and alcohol consumption. There was little difference in prevalence of smoking between men exposed vs. unexposed to siege, and the suggested increase in risk of death from respiratory cancer was not substantially affected by adjustments for smoking or alcohol in our analysis. It is, however, possible that a more detailed measurement of smoking habits, and thus better adjustment, might lead to a further reduction of the association between the siege and cancer mortality in men. Women's reproductive histories and age at cancer diagnosis were not known and could not be taken into account in the analysis.
Increased breast cancer mortality in women exposed to severe starvation in Leningrad is consistent with findings from a study of Guernsey women8 in that, in both studies, the excess risk of breast cancer was seen among women exposed at 10–18 years of age. In studies of women exposed to the Dutch Hunger, the greatest risk of breast cancer was reported among women exposed during adolescent growth spurt in study by Dirx et al.,4 but in women exposed at age 2–9 years in a subsequent study by Elias et al.3 Most recently, Painter et al.20 observed more breast cancer in women born within a year before the Dutch famine who were exposed as infants when compared with women conceived after the famine, although the difference was not statistically significant.
The generally consistent finding of excess risk of breast cancer in women exposed to starvation in childhood or adolescence is not in line with certain animal experiments showing caloric restriction to prevent breast cancer,21 and is inconsistent with studies that found a decreased risk of breast cancer in Norwegian women exposed to transient energy restriction during World War II.22, 23 A recent study of dietary restriction in rodent models demonstrated inhibition of carcinogenesis as a result of long-term dietary restriction, but short-term and cyclical dietary restriction were not associated with the same protective effect and instead often had an adverse effect on carcinogenesis,24 suggesting that whether caloric restriction results in protective or deleterious effects on cancer risk may depend greatly on the pattern, length and severity of the restriction.
The Leningrad siege constituted more severe starvation than the average daily caloric intake of 2,700–2,800 kcal in Norway in 1943–1945,25, 26 the reported daily adult per capita ratios of less than 700 kcal in the Netherlands in January 19453 and the ∼1200 cal per day at the peak of the Occupation of Guernsey.8 Specific estimates of daily caloric intake by the participants in the Israel study are not available. While the food situation improved rapidly in Netherlands and Norway after the war,3, 25, 26 limitations in food availability continued to affect the population of the Soviet Union in the postwar period.27 We speculate that the apparently stronger effect of siege exposure on breast cancer in Russian women might be due to severity of exposure or, alternatively, to a more selective survival in the Russian cohort. More than one fifth of the population of Leningrad died from hunger-related causes during the period of siege,11 and it is highly unlikely that the survivors were representative of the total population. On the other hand, our results showing nonsignificantly higher risk of prostate cancer and nonsignificantly lower risk of colon cancer in survivors of the siege are markedly consistent with evidence from the Dutch famine6, 7 and the reduction in colon cancer mortality noted in birth cohorts born around World War II in Norway28 and other Nordic countries.29
Earlier analyses of data from our cohort also showed that mean adult height was somewhat increased in women who were exposed to the siege as children or adolescents,30 and that blood pressure and mortality from ischemic heart disease and stroke mortality was higher in men who experienced the siege of Leningrad around puberty.30–32 It is plausible that greater adult height among surviving women who were exposed to siege may be due to their relatively more privileged living conditions both before or during the siege, a delayed menarche (and thus longer linear growth) caused by starvation during prepubertal age, or to a selective survival of women possessing particular metabolic or physiological characteristics. In our current analysis, however, the effect of siege exposure on breast cancer was not removed by adjustments for adult height.
In contrast to our results, adult height was reduced in female survivors of the Dutch Hunger Winter.33 While investigations of the effects of war on childhood growth have indicated growth delays in children in Norway34 and other Nordic countries,35 the resulting effect on adult stature is not known. Interestingly, stronger associations of adult height with breast cancer in Norwegian women born36 or experiencing puberty37 during the war were interpreted as an indication of the importance of nutritional conditions during these periods of growth. Female survivors of the Occupation of Guernsey8 who were aged 6–14 at the beginning of the Occupation were shown to have on average delayed menarche; unexpectedly, this was associated with a higher risk of breast cancer.
A recent study of postmenopausal women in the United States showed a positive association of weight gain throughout adulthood with risk for postmenopausal breast cancer,38 suggesting that an upward trend in body mass index beginning in late adolescence/early adulthood may be associated with later postmenopausal breast cancer risk. Our earlier analyses failed to show increased body mass index in women exposed to siege,30 and the current analyses did not indicate that the association of siege with breast cancer was confounded by women's overweight or obesity.
A number of other mechanisms were proposed as an explanation for the higher breast cancer risk in women in the Dutch famine, all of which were based on the evidence from Dutch cohorts, where effects of famine on female reproductive and hormonal system were demonstrated. These include effects of starvation on age at menarche,39 insulin-like growth factor axis,40 reproductive outcomes,41 age at menopause,42 levels of sex hormones in postmenopausal women,43 neurohormonal effects and brain function.44, 45 A mechanism proposed to explain the higher breast cancer risk in women exposed to the Occupation of Guernsey is based on the interruption of differentiation of mammary epithelium in prepubertal females leading to increased likelihood of tumor development.8 Although it is likely that most cases of breast cancer in the Leningrad cohort were postmenopausal, we did not have information on age at diagnosis or women's reproductive histories in our study.
A hypothesis suggesting that a hypothalamopituitary axis might be programmed by events experienced in adolescence45 and lead to long-term changes in levels of hormones and reproductive function is of particular interest, as it also offers a plausible mechanism linking the major simultaneous exposures of the siege (psychological stress, starvation, severe cold) to hormone-associated cancer in men and to higher office-recorded systolic blood pressure in both sexes.30
Despite the spurious or borderline statistical significance of our results, we believe that the results constitute a valuable contribution to the sparse literature on potential long-term effects of severe war starvation in humans. Long- and short-term starvation are by no means things of the past, but on the contrary, persistent features in many parts of the world. The specificity and relative consistency of our findings, with those reported in the survivors of the famines in the Netherlands, Norway, Guernsey and the Holocaust, calls for further studies addressing mechanisms underlying the possible adverse effect of war starvation on breast and prostate cancer and (more speculatively) the seemingly protective effect on colon cancer.
We thank Prof. Pär Sparén and Mr. Reidar Österman for help with data collection and preparation of the dataset, Prof. Orly Manor and Dr. Bianca De Stavola for advice on statistical analysis, and Prof. Sviatoslav Plavinski and anonymous reviewers for comments on our earlier manuscripts.