AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 73:525-537 (1987) Cross-Cultural Correlations of Childhood Growth and Adult Breast Cancer MARC S. MICOZZI Cancer Prevention Studies Branch, Blair 6A01, National Cancer Institute, 9000 Rockuille Pike, Bethesda, MD 208924200 KEY WORDS Anthropometry, Nutrition, Diet, Growth ABSTRACT International differences in breast cancer incidence and mortality, and studies on Japanese migrants to the United States, point to the importance of environmental factors, including diet and nutrition, in the etiology of breast cancer. Some studies have suggested that dietary patterns in early life are important to the long-term risk of breast cancer. Given that human growth is partially a function of early dietary intake, cross-cultural correlations between breast cancer rates and anthropometric variables measured at different times in childhood provide additional information about the association of early nutrition and cancer. In this study, the associations between food consumption and anthropometric variables, and childhood growth patterns (attained size at age) and adult breast cancer rates, were considered. Data from cross-sectional growth studies conducted during the years 19561971 on children aged 6-18 years were obtained for age-specificstature, sitting height, weight, triceps skinfold thickness, arm and chest circumferences, and biacromial and biiliac diameters. National food consumption data were obtained from the United Nations Food and Agriculture Organization (FA01 and socioeconomic status indicators from the United Nations Children’s Fund (UNICEF). Cancer incidence data for the years 1972-1977 were obtained from regional cancer registries reported by the International Agency for Research on Cancer (IARC), and mortality data for 1978 were obtained from national cancer registries around the world. Significant correlations were seen between national food consumption data and childhood growth (attained size at age); between cancer incidence and agespecific stature (r = 0.68), weight (r = 0.59), triceps skinfold thickness (r = 0.78), and biacromial width (r = 0.84); and between mortality and age-specific stature (r = 0.77), weight (r = 0.75), and biacromial width (r = 0.78). In general, the correlation coefficients of the observed anthropometric variables with breast cancer increase with increasing age and become highly significant at ages 13-14 years, reflecting cumulative childhood nutritional intake. Breast cancer has historically been the most frequent cancer in women (115,000 new cases per year), and is the leading cause of cancer death in women (36,000 deaths per year) in the United States. International differences in breast cancer incidence and mortality rates (Tables 1, 2), and studies on migrating Japanese-American populations, point to the potential importance of environmental factors in the etiology of breast cancer in women. The Japanese-American mi- 0 1987 ALAN R. LISS, INC grant studies further suggest that dietary acculturation early in life is a critical variable for long-term breast cancer risk (Buell and Dunn, 1965; Buell, 1973; Dunn, 1977). Human epidemiologic studies show high correlations between national levels of breast cancer and per capita consumption of several Received May 7,1986; revision accepted October 20, 1986 M.S. MICOZZI 526 TABLE I . Cross-national comparison of breast cancer incidence rates Country Years of study Senegal Singapore (Malay) Japan Poland (rural) India Singapore (Chinese) Hungary (rural) Puerto Rico Romania Czechoslovakia Hong Kong (Chines6?) Colombia Yugoslavia Spain Poland (urban) Jamaica Finland Norway Australia Denmark France Sweden West Germany Brazil U.S. (Asian) Italy Britain Israel New Zealand Canada U.S. (black) Switzerland U.S. (white) 1969 1973-1977 1973-1977 1973-1977 1973-1975 1973-1977 1973-1977 1963 1974-1978 1973-1977 1974-1977 ?972-?976 1973-1976 1973-1977 1973-1977 1973-1977 1971-1976 1973-1977 1973-1977 1968-1972 1975-1977 1973-1977 1973-1977 1973 1972-1977 1976-1977 1973-1977 1972-1976 1972-1976 1973-1977 1973-1977 1964 1973-1977 Breast cancer incidence per 100,000 population (age-standardizedrates)' 11.8 14.7 17.5 17.7 21.2 21.9 29.2 29.5 30.1 30.3 31.1 33.2 34.2 36.5 36.5 39.0 40.1 49.6 53.2 54.4 54.5 55.2 55.7 56.2 57.3 57.6 58.4 59.9 62.6 63.2 67.2 76.1 83.7 'Incidence data from Waterhouse e t al. (1982) macronutrients, including dietary fat, protein, and total calories (Armstrong and Doll, 1975; Gaskill et al., 1979; Gray et al., 1979). These ecologic studies make use of national food consumption, or food disappearance, data that are only indirect indicators of nutrition and do not distinguish patterns by age, sex, or other demographic characteristics within the population. Case-control studies of diet and breast cancer have obtained dietary data on individuals but have generally focused attention on current nutritional patterns in breast cancer cases (Phillips, 1975; Miller et al., 1978; Lubin et al., 1981). However, few or no data are available on the relationships of a single assessment of nutritional status at one point in time to longterm nutritional status or to the multistage process of tumor development (Micozzi, 1985). The role of nutrient intake in the development of cancer appears to depend on the timing, duration, and magnitude of dietary exposures. One interpretation (Miller, 1977) of the overall pattern of epidemiologic evidence on diet and breast cancer is consistent with the view that dietary intake in early life is a critical variable. Based on analysis of age-incidence curves for breast cancer in different populations around the world, the likely conclusion is that the risk factors that have caused increased breast cancer rates in some populations in recent years are basically determined before the age of appreciable incidence is reached (Bjarnason et al., 1974). Either exposure to relevant risk factors is confined to younger women or, if the exposure is widely spread in the population (as with diet), then only the young are susceptible. A mechanism for the effects of early nutritional patterns on the long-term risk of breast cancer is suggested by the two-stage model of carcinogenesis proposed by Moolgavkar et al. (1980). Nutrients, like hormones, may influence the risk of breast cancer by their effects on the growth of normal, conneoplastic tissue. The promotional effects of macronutrients on breast cancer development have been demonstrated in a n animal tumor model (Ross et al., 1983). Early breast secretory ac- 527 CHILDHOOD GROWTH AND ADULT BREAST CANCER tivity may also be stimulated by diet (Petrakis et al., 1981), and aspects of breast biology may be influenced by nutrient intake in early life (DeSouza et al., 1974;Berg, 1975). If nutrient intake in early life is an important determinant of the long-term risk of breast cancer, then case-control studies on elderly cancer patients may not be the most appropriate way to test hypotheses about the relationship between diet and breast cancer. Given the limitations of indirect dietary assessment (Block, 19821, measurement of nutrition-mediated variables that reflect growth (and can be determined in adults) may be meaningful for the assessment of past nutritional patterns. Anthropometry provides a reliable means of assessing past, as well as current, nutritional patterns (Johnston, 1981, 1983). Increased intake of macronutrients during the growth period is associated with increased stature and lean body mass and accelerated rates of maturation. The former are reflected in increased height and body size in adults, the latter by early age at menarche in women. Overnutrition in childhood may also be related to increased childhood and adult fatness. Fatter children are taller, heavier, and developmentally advanced compared to the lean (Garn et al., 1982; Roche, 1984; Himes and Roche, 1985). These tendencies begin early in life, and excessive macronutrient consumption during the postnatal period may lead to increased muscle growth, fat deposition, or, probably, both (Hahn and Koldovsky, 1966; Haymond et al., 1974). Thus increased macronutrient intake during growth and development may eventually lead to an adult whose size is close to the maximum for that genotype (Stini, 1978). Cross-sectional growth data from different populations are based on physical measurements taken in children that are related to nutrition in childhood. As an alternate means of nutritional assessment, childhood growth (attained size at age) can be used to test hypotheses about the relations of early nutrition to the long-term risk of breast cancer. Based on these observations, a hypothesis was developed that attained size at age during childhood, which partially reflects nutritional exposures, is directly related to the risk of breast cancer in adult life among different populations (Micozzi, 1985). To perform an initial test of the hypothesis in an ecologic model, a data base was constructed consisting of reliable, age-specific growth data. TABLE 2. Cross-national comparison of breast cancer mortality rates (1978-1979) Breast cancer deaths per 100,000population (age-standardized rates)' Country Japan Hong Kong Singapore Yugoslavia Romania Spain Poland Bulgaria Greece U.S. (black) Finland Israel Australia Italy Canada New Zealand France Hungary Norway Austria Sweden Netherlands West Germany Switzerland Denmark Belgium Great Britain U S . (white) 6.5 9 .o 10.2 15.2 15.6 17.8 18.0 18.9 20.1 21.2 22.6 24.0 24.0 28.1 28.1 29.1 30.3 32.3 32.5 33.7 34.5 37.2 37.3 39.5 41.3 42.0 47.6 52.1 'Mortality data from Kurihara et a1 (1984) MATERIALS AND METHODS Growth data were obtained from cross-sectional studies that collected anthropometric dimensions on children aged 6-18 years (Eveleth and Tanner, 1976). Not all studies included data on all ages in this range (Table 3). Growth data were obtained only on populations for whom breast cancer incidence and/ or mortality data were also available. The criteria of Waterlow et al. (1977)were considered for selection of each growth study included in the analysis: The population studied was not clinically malnourished; the samples were cross-sectional; and sampling procedures were defined and reproduceable; at least 200 individuals were included at each age interval, and age groups were presented in 1-year intervals; and measurements were carefully made and recorded by observers trained in anthropometric techniques, using equipment of well-tested design and calibrated at frequent intervals. The year of performance of all studies selected was sometime during the period 1956-1971 to minimize the effects of secular trend on relative ranking of population growth patterns. Size at age 18 years is taken as adult size (Meredith, 1971). 528 M.S. MICOZZI TABLE 3. Human growth studies in data base Population Sample (size)' Location Argentina (white) Urban (c. 100) La Plata Australia (white) Austria (white) Belgium (white) Brazil (white) Bulgaria (white) Urban (c. 1,500) Urban (c. 3,000) Urban (500) Urban (c. 500) Urban (c. 400) Sydney Vienna Brussels Sao Paolo Sofia Canada (white) Urban (c. 100) Montreal Costa Rica (white) Urban (c. 250) National Cuba (white) Urban (c. 60) Havana Czechoslovakia (white) Denmark (white) Finland (white) Urban (c. 3,000) Prague Urban (c. 40) Urban (740) Copenhagen Helsinki France (white)' Germany, West (white) Great Britain (white) Greece (white) Urban (165) Urban (c. 1,000) Paris Hamburg Urban (c. 1,000) London Urban (c. 250) National Guatemala (white) Urban (c. 45) Hong Kong (Chinese) Hungary (white) India (Asian) Israel (white) Urban (c. 500) Urban (c. 200) National (c. 10,000) National (c. 20) Guatemala City Nation a 1 Zged National Jews Italy (white) Italy (white) Japan (Asian) Urban (c. 250) Rural (c. 500) National (all) Naples Sardinia National Jamaica (black) Malaysia (Asian) Urban (c. 25) Rural (c. 40) Kingston Muar Netherlands (white) National (c. 1,000) National New Zealand (white) National (c. 800) National Norway (white) Poland (white) Poland (white) Urban (c. 6,000) Urban (c. 300) Rural (c. 160) Oslo Warsaw Warsaw Philippines (Asian) National (c. 400) National Puerto Rico (white) Rumania (white) Senegal (black) Singapore (Malay) Singapore (Chinese) Spain (white) Urban (c. 100) Urban (c. 100) Urban (c. 300) Urban (c. 200) Urban (c. 200) Urban (100) San Juan National Dakar National National Madrid Sweden (white)' Switzerland (white) U.S. (white)' U S . (black)' U.S. (Asian) U.S.S.R. (white) Yueoslavia (white) " Urban (c. 360) Urban (c. 130) Urban (c. 300) Urban (c. 150) Urban (c. 30) Urban (c. 200) Rural (c. 100) National Basle Philadelphia Philadelphia Los Angeles Moscow Lika 'At each year of age. 'Longitudinal study. Authors Years of study Ages (years) 1971 6-12 1970 1962 1960-1961 1968-1969 1963 6-18 6-10 6-18 6-12 6-18 1969-1970 6-16 1963-1969 6-18 1963-1964 7-18 1971 6-18 Andersen, 1968 BackstromJarvinen, 1964 Sempe et al., 1971 city of Hamburg, 1962 Tanner et al., 1966 1968 1959-1960 8-18 6-18 1960-1971 1960 6-17 6-18 1965 6-18 Valaoras and Laros, 1969 Sabharwal et al., 1966 Chang, 1969 Farkas, 1966 Indian Council, 1972 Shiloh and Yekutiel, 1958 Tatafiore, 1970 Pinna, 1961 Tokyo Dept. Health, 1970 Ashcroft et al., 1966 Wadsworth and Lee, 1960 Wieringen et al., 1971 New Zealand Dept. Health, 1971 Iversen, 1962 Charzewska, 1973 Panek and Piasecki. 1971 Natl. Coord. Center, 1965 Knott, 1963 Cristescu, 1969 Masse, 1969 Wong et al., 1972 Wong et al., 1972 Garcia-Almansa et al., 1969 Ljung et al., 1974 Heimendinger, 1964 Krognian, 1970 Krogman, 1970 Kondo and Eto. 1975 Vlastovskv et al.. nd Gavrilovi:, 1971 1963-1966 6-12 1961-1965 7-17 1961-1965 1958-1959 1956-1965 1956 6-18 6-17 6-18 6-11 1963 1961 1963-1970 6-18 6-12 6-17 1964 1960 6-18 6-12 1964-1966 6-18 1969 6-15 Cusminsky and Lozano, 1974 Jones et al., 1973 Stracker, 1964 Twiesselmann, 1969 Marcondes et al., nd Kadanof and Mutafov, 1969 Demirjian et al., 1972 Villarejos et al., 1971 LaskaMierzejewska, 1967 Prokopec et al., 1973 1959-1960 1971 1960 7-18 10-18 6-18 1963-1964 6-17 1962 1963-1966 1960-1962 1972 1972 1968 7-17 11-16 6-15 6-14 6-14 6-14 1964-1971 1956-1957 1956- 1966 1956-1966 1971 1969-1970 1971 10-16 6-18 7-17 7-17 6-17 6-1 7 7-15 CHILDHOOD GROWTH AND ADULT BREAST CANCER Growth data are multidimensional, and anthropometric dimensions selected for analysis represent indices of linear growth, frame size, and lean body mass as well as absolute and relative body fatness. Anthropometric variables recorded from each study included age-specific stature, weight, sitting height, triceps skinfold thickness, arm and chest circumferences, and biiliac (pelvic width) and biacromial (shoulder width) diameters. For this analysis, all reported measurements were rounded to three significant figures in appropriate units. All studies recorded height and weight; not all studies recorded all the remaining variables. The set of populations for whom complete anthropometric data were available was biased to include proportionately more European countries, which were more tightly clustered on one part of the curve. In this respect, the ability to demonstrate any relations with breast cancer would be as strong as or stronger than results demonstrated with the complete data set across the full range of variation. National food consumption data were collected from the Food and Agriculture Organization (FAO), Rome, National Food Disappearance Tables. Annual per capita food consumption data on total amounts of foods and nutrients, and calories from foods, were obtained for each country for both the earliest available period (1961-1965 annual average) and for the latest (1977).These data were taken as representative of overall adult food consumption patterns, since no samples are distinguished by age or sex. Food consumption data could also not be differentiated for urban vs. rural populations. Indicators of socioeconomic status (SES) were obtained from the United Nations Children’s Fund (UNICEF, 1984, 1985) for all countries from which populations were included in this analysis. SES indicators were obtained from both 1960 and 1981 when possible. Infant mortality rates (per 1,000 live births), gross national product, average life expectancies in men and women, total fertility rates (total children per woman), and percentage share of household income held by the lower 40% and the upper 20% of SES class. Data on average age of menarche in different populations were obtained as tabulated by Eveleth and Tanner (1976) and included studies conducted during the period 19591973. Thus the menarche data were collected on approximately the same cohort of women 529 for whom growth data were obtained, over a sufficiently narrow time period, t o minimize the effects of secular trend. Age-standardized breast cancer incidence rates (Waterhouse et al., 1982)for 32 populations and mortality rates (Kurihara et al., 1984) for 35 countries were obtained for which acceptable growth data were also available (Tables 1-3). Age-specific breast cancer incidence rates allow probable comparison of premenopausal and postmenopausal breast cancer. Anthropometric data and breast cancer incidence data were both identified as to whether data originated from urban or rural populations. Thus, in the correlations between these two variables, populations are independently identified and examined by urban or rural status. In some countries, data were available on both urban and rural populations, and the two sets of paired observations were examined independently in the correlation analysis. When both anthropometric and cancer data were available for subpopulations within a country, these populations were treated separately. For example, U.S. whites, U.S. blacks, and U.S. Asians have separate growth data and cancer data, as do Polish urban and Polish rural populations (Tables 1-3). For many other countries, anthropometric data are available on rural populations, but not cancer incidence data; these populations could not be included in this analysis. For breast cancer mortality and food consumption, only national data are available by country. Therefore, the same national food consumption and breast cancer mortality data were matched to each representative population. Since not all populations had anthropometric data available at all ages between 6 and 18 years, analyses were also performed, in each case limited to the set of populations for whom anthropometric data were available at all ages over the range. Since national breast cancer mortality data were available for a subset of 20 countries from 1964, and 35 countries from 1978, analyses were also performed only on the subset of countries having data available for both 1964 and 1978 to allow observation of any time trends. RESULTS AND DISCUSSION Anthropometric variables in children showing significant correlations to both breast cancer incidence and mortality rates were height, weight, and biacromial width 530 M.S. MICOZZI TABLE 4. Correlation of breast cancer rates with age-specific mean stature in 32 populations' Age (years) 6 7 8 9 10 11 12 13 14 15 16 17 18 N2 Premenopausal (BCI40) Incidence Postmenopausal (BCI70) Total (BCIALL) Mortality N BCMALL 25 30 30 30 32 32 32 31 31 28 25 22 14 0.61 0.51 0.53 0.57 0.53 0.54 0.52 0.58 0.80 0.61 0.63 0.65 0.67 0.58 0.48 0.47 0.52 0.48 0.48 0.47 0.55 0.58 0.60 0.65 0.70 0.72 0.54 0.41 0.46 0.49 0.49 0.48 0.47 0.53 0.54 0.56 0.56 0.57 0.68 23 29 29 29 31 30 30 28 28 26 24 21 15 0.49 0.38 0.56 0.56 0.53 0.45 0.56 0.65 0.62 0.75 0.77 0.74 0.72 Ip < 0.01 for all values reported. 'N, no. of paired observations. TABLE 5. Correlation of breast cancer rates with age-specific mean weight in 32 populations' Age (vears) N Premenopausal (BCI40) 6 7 8 9 10 11 12 13 14 15 16 17 18 25 30 30 30 32 32 32 31 31 28 25 22 15 0.62 0.59 0.59 0.52 0.48 0.51 0.49 0.57 0.65 0.59 0.45 0.50 0.57 Incidence Postmenopausal (BCI70) 0.51 0.48 0.50 0.43 0.39 0.42 0.38 0.46 0.55 0.57 0.45 0.50 0.61 Mortality Total (BCIALL) 0.56 0.49 0.53 0.45 0.45 0.47 0.44 0.53 0.59 0.54 0.37 0.44 0.59 N BCMALL 22 28 28 28 30 29 29 27 27 25 23 20 15 -_ 0.48 0.39 0.46 0.42 0.37 0.37 0.42 0.43 0.50 0.58 0.69 0.68 0.75 'p < .01for all values reported. (Tables 4-6). Triceps skinfold thickness and upper arm circumference were significantly associated with breast cancer incidence but not with mortality (Tables 7,8). In Tables 48, age-specific breast cancer incidence rates a t ages 40 (BCI40) and 70 (BC170) years are provided as representative of premenopausal and postmenopausal women, respectively, in addition to overall age-adjusted rates (BCIALL). Since most growth studies did not record all anthropometric measurements, sample size was small and correlations were not significant over most ages for chest circumference, sitting height, and biiliac (pelvic) width. Among measures of frame size, biacromial width (Table 6), but not biiliac width, was significantly correlated to breast cancer rates. In addition, mean sitting height showed significant associations to overall age-adjusted breast cancer incidence rates, but not mortality rates, at ages 7-10 years only (r = 0.47-0.62). The correlation coefficient of breast cancer rates with height, weight, triceps skinfold thickness, and biacromial diameter increase with increasing age (Tables 4-6). The correlation coefficients did not change appreciably when breast cancer mortality data from 1964 were used as compared to the mortality data from the same set of populations from 1978. As is shown in the tables, anthropometric dimensions were not available a t every age in different populations. The trend for increasing correlation coefficients between anthropometric dimensions and breast cancer rates did not diminish when analysis was limited to populations for whom data were available at all ages between 6 and 18 years. In Table 7, it can be seen that childhood fatness is as highly correlated to premenopausal as to postmenopausal cancer rates. 531 CHILDHOOD GROWTH AND ADULT BREAST CANCER TABLE 6. Correlation o f breast cancer rates with age-specific mean biacromial width (shoulder) in 17 populations Age bears) 6 7 8 9 10 11 12 13 14 15 16 17 18 N Premenopausal (BCI40) 14 16 14 14 15 15 16 15 15 12 10 8 7 NS1 0.69 0.71 0.71 0.71 0.66 0.66 0.74 0.77 0.76 NS 0.63 0.75 Incidence Postmenopausal (BCI70) NS 0.72 0.72 0.73 0.74 0.70 0.72 0.80 0.78 0.84 NS NS NS Mortality Total (BCIALL) NS 0.70 0.71 0.71 0.73 0.68 0.68 0.78 0.77 0.81 NS NS NS N BCMALL 12 17 13 15 16 16 17 16 16 14 12 10 8 NS NS 0.44 0.40 0.45 NS NS NS 0.41 0.47 0.67 0.78 0.71 INS, Not significant. Otherwise, p < ,005 TABLE 7. Correlation of breast cancer rates with agespecific mean triceps skinfold thickness in 16 populations Age (years) 6 7 8 9 10 11 12 13 14 15 16 17 18 N 16 16 16 15 16 16 16 15 14 11 10 8 5 Premenopausal (BCI40) Incidence Postmenopausal (BCI70) 0.55 0.57 0.67 0.61 0.59 0.66 0.53 0.65 0.72 0.70 0.75 0.58 0.83 0.46 0.45 0.54 0.50 0.48 0.59 0.49 0.58 0.63 0.66 0.58 0.38 0.78 Mortality Total (BCIALL) N BCMALL 0.56 0.53 0.67 0.60 0.60 0.66 0.55 0.68 0.75 0.74 0.77 0.64 0.78 12 14 14 13 14 14 14 12 11 9 9 7 4 0.52 NS1 NS NS NS NS NS NS NS NS NS NS NS INS, Not significant (p > 0.10). Otherwise, p < 0.03 The correlations with age-specificbiacromial width (an index of frame size) and with upper arm circumference (reflecting both body fatness and lean body mass) in Tables 6 and 8, respectively, are significant over a greater range of ages with premenopausal (age 40 years) than with postmenopausal (age 70 years) breast cancer incidence rates. Different anthropometric dimensions reflect linear growth, frame size, lean body mass, and/or body fatness. Both indices of absolute size (stature, sitting height, frame size, lean body mass) and fatness (triceps skinfold thickness) are correlated with increased breast cancer risk. Given that growth is partially a function of early dietary intake, these cross-national correlations between cancer rates and anthropometric dimensions at different points during childhood (attained size at age) provide additional information about the relationship of increased macronutrient intake to breast cancer risk in human populations. The correlations between breast cancer and each of the anthropometric dimensions is highest for the oldest age group and increases with each year of age. Dietary differences between populations are cumulatively reflected in the growth curve by greater size differences at older, rather than younger, ages. The trend that correlations increase with older age lends further support t o the associations between diet and cancer. The strength of the correlations between breast cancer and height and weight increases with increasing age (Table 9). These correlation coefficients markedly increase beginning after age 13 years. Age 13years is also the time in life when Japanese-American migrants (Froelich, 1970; Kondo and Eto, 1975) begin to show growth rates signifi- 532 M.S. MICOZZI TABLE 8. Correlation of breast cancer rates with age-specific mean arm circumference in 12 populations Age (years) N 6 7 8 9 10 11 12 13 14 15 16 17 9 9 8 8 9 9 10 10 10 8 7 6 18 5 Premenopausal (BCI40) Incidence Postmenopausal (BCI70) NS' Mortalitv Total (BCIALL) NS NS 0.66 0.80 0.71 0.58 0.58 0.56 0.52 0.66 0.59 0.41 . 0.55 0.72 0.62 0.50 0.49 0.52 0.46 0.61 NS NS NS NS ~~ 0.70 NS NS NS NS NS NS NS NS NS NS NS NS NS N BCMALL 9 0.67 11 NS NS ~~ 10 NS 10 11 11 12 11 11 10 9 8 6 NS NS NS NS NS NS NS NS 0.79 INS, Not significant (p > 0.10).Otherwise, p < 0.05. TABLE 9. Cross-national correlation of age-specific anthropometric measures with age-adiwted breast cancer mortalitv Age (years) 6 7 8 9 10 11 12 13 14 15 16 17 18 Stature (N)1 Correlation between breast cancer mortality and Weight (N) Biacromial width (N) 0.49 (23) 0.38 (29) 0.56 (29) 0.56 (29) 0.53 (31) 0.45 (30) 0.56 (30) 0.65 (28) 0.62 (28) 0.75 (26) 0.77 (24) 0.74 (21) 0.72 (15) 0.48 (22) 0.39 (28) 0.46 (28) 0.42 (28) 0.37 (30) 0.37 (29) 0.42 (29) 0.43 (27) 0.50 (27) 0.58 (25) 0.69 (23) 0.68 (20) 0.75 (15) NS2 (12) NS (17) 0.44 (15) 0.40 (15) 0.45 (16) NS (16) NS (17) NS (16) 0.41 (16) 0.47 (14) 0.67 (12) 0.78 (10) 0.71 (8) 'N, No. of paired observations. 'NS, Not significant. Otherwise, p < 0.05 cantly increased over those of their Japanese counterparts living in Japan (Tokyo Department of Maternal and Child Health, 1970) and over earlier generations of JapaneseAmerican migrants to California (Greulich, 1957). The correlation coefficients between breast cancer and biacromial diameter also became significant after age 13 years and increase with increasing age (Table 9). When the growth patterns of Japanese girls in Japan (Tokyo Dept. Health, 1970) and earlier generations of Japanese-American girls in California (Greulich, 1957) are compared with those of later generations of JapaneseAmerican girls in California (Froelich, 1970; Kondo and Eto, 19751, it is also seen that there is some decrease in the gap between their growth patterns at older ages (see Eveleth and Tanner, 1976, Fig. 102, p. 135). This observation may be explained by the later average age of menarche among populations of Japanese girls with less early growth, which allows some continued growth at later ages. The relationships of cancer with age at menarche were consistent with previous reports but were not as strongly or as significantly correlated (Table 10) as were relationships with anthropometric variables. Some correlations with various socioeconomic indicators were not significant. For those that were significant, the correlation coefficients generally were not greater than those for anthropometric variables. Fertility rates were inversely correlated with breast cancer rates, corresponding to the established protective effect of early age at first pregnancy, and the known association between early age at birth of the first child and high parity. The correlation coefficients be- 533 CHILDHOOD GROWTH AND ADULT BREAST CANCER TABLE 10. Correlation of breast cancer rates with various socioeconomic indicators in 32 populations N BCI40 BCI70 BCIALL N BCMALL Infant mortality (per 1,000 live births) 31 1960 31 1981 GNP U.S. dollars 31 1981 -0.54 -0.49 0.71 -0.45 -0.40 0.67 -0.50 -0.45 0.73 32 32 32 -0.65 -0.49 0.56 31 31 0.56 0.51 0.47 0.43 0.52 0.46 32 32 0.67 0.34 31 31 0.55 0.50 0.47 0.42 0.51 0.46 32 32 0.66 0.37 Average life expectancy (years) Males 1960 1981 Females 1960 1981 Average age of menarche (years) 1959-1973 Total fertility (children per woman) 1960 1981 Percentage share of household income % Income of lower 40% 6 . S . dollars) % Income of higher 20% 29 -0.41 -0.37 -0.39 26 NS 30 30 -0.43 - 0.46 -0.35 -0.40 -0.38 -0.44 32 32 -0.54 -0.61 18 NS' NS NS 20 0.57 18 NS NS NS 20 -0.67 INS, Not significant. Otherwise, p < 0.05 tween cancer rates and average life expectancy were smaller than those between cancer rates and anthropometric or nutritional variables. Although cancer is more common in aged populations, the overall ageadjusted cancer rates show that longevity does not explain the geographic pattern of breast cancer. The growth and cancer relationships hold up independently when socioeconomic status is taken into account i n multivariate analyses (not shown). The growth and nutrition relationships also explain observations in different populations among whom cancer patterns cannot be explained by socioeconomic factors. For example, in New Zealand, Maori women have greater childhood growth, greater adult body size, lower SES (New Zealand Department of Health, 1971)and higher breast cancer rates (Henderson, 1979) than white women. Several studies on Samoans in American Samoa (Borrie, 1967; Hoadley, 1980; Dines et al., 1980; Taylor and Zimmet, 1981; Wood and Gans, 1981, 1984) are generally consistent with the observations that Samoan women may have greater childhood growth and adult body size, although socioeconomic indicators are lower. The body size and body composition of Samoan children have been related to infant feeding patterns; further, the relations between infant feeding and childhood fatness were not affected by family income (Bindon, 1984a). In studies of 330 Samoan adults living in American Samoa and Hawaii (Bindon, 198413) and 2,657 Samoan adults in American Samoa, Western Samoa, and Hawaii (Bindon and Baker, 1985), obesity and relative micronutrient deficiency were related to modernity of residence or occupation. Other data (Government of American Samoa, 1975; Crews, 1985)are consistent with high rates of breast cancer in American Samoan women. Thus American Samoan women appear to represent another example of a population who have increased childhood nutrition and growth, greater adult body size, lower SES, and high age-specific breast cancer rates as compared to other populations. Hawaiian women are also generally larger and have lower SES but higher breast cancer rates than whites over most age groups (Horm et al., 1984). The same situation may apply among Kanak women in French New Caledonia (Garenne, 1985). This ecologic model indicates that childhood height, weight, and biacromial width are highly correlated with breast cancer rates 534 M.S. MICOZZI in adult women among different populations. In the one case-control study on anthropometry and breast cancer (Brinkley et al., 1971), height, sitting height, weight, biacromial width, and biacromial-biiliac ratio were greater among breast cancer cases than among controls. The current study did not permit complete characterization of sitting height or biiliac width with respect to adult breast cancer rates. However, there is a n indication that increased sitting height was significant over some ages and that biiliac diameter was less significant than biacromial diameter. Both these observations may have implications for nutritional patterns and the timing of age at menarch vis-a-vis upper body (shoulder width and sitting height) vs. lower body (pelvic width and leg length) development. These ecologic analyses are consistent with a role for childhood nutrition in determining adult breast cancer rates in different populations. The correlations between food disappearance data and attained size at age, and between food disappearance data and adult breast cancer rates, were also significant. However, these results are not presented in support of the hypothesis because of the indirect, unvalidated nature of food disappearance data and the inability to distinguish patterns of consumption by sex or age, although they nonetheless demonstrate “tracking” of nutritional patterns through life cross-culturally. A more productive and relevant approach to further testing of the hypothesis, recognizing the above limitations, is to construct multivariate models for the prediction of breast cancer that incorporate various measures of childhood nutrition (attained size at age) with measures of adult nutrition (taken as food disappearance data). Preliminary findings with these models show a highly significant contribution of childhood nutrition to explanatory models using a wide range of relevant adult nutritional variables (Micozzi, 1986). Adult nutrition (food disappearance data) does not account for the crosscultural variability in breast cancer rates, which can be explained by measures of early nutrition in the same populations. Although there is general tracking of comparative nutritional patterns among different human populations, the multivariate analyses showed that the variability in breast cancer rates among different populations cannot be explained primarily by adult food consumption but that indicators of childhood nutrition (anthropometric variables) have independent significance in the models. Measures of childhood fatness, frame size, and height were significantly correlated to both premenopausal and postmenopausal breast cancer rates in adult women in the correlation analysis. Adult fatness is independent of height and frame size and is a recognized risk factor only among older, postmenopausal women (when a distinction is made; see Micozzi, 1985). Childhood fatness is correlated with increased height, frame size, and lean body mass, all of which are indicators of increased nutrition during childhood and all of which appear to be risk factors for adult breast cancer a s well. These ecologic models remain consistent with a role for childhood nutrition, as reflected by age-specific anthropometric variables, in determining adult breast cancer rates. The hypothesis of early nutrition and breast cancer can be further tested using anthropometric data in individual adults (which reflect early nutritional patterns). Individual anthropometric data were prospectively obtained on adult women, for whom subsequent breast cancer outcome was determined, in the US. National Health and Nutrition Examination Survey (NHANES) epidemiologic follow-up Study. Of all the variables studied among this population, breast cancer cases had significantly greater mean height and frame size (elbow width) than noncases (Micozzi, 1986). Adult height and frame size are generally determined by age 18years and are not influenced by adult nutrition. In this analysis, indicators of adult nutritional patterns (weight, body mass indices, body fatness) were not associated with breast cancer. These findings are consistent with a role for increased nutrition during childhood in determining the long-term risk of breast cancer in human populations. Thus, beyond the ecologic models presented, these results are consistent with the hypothesis using data collected on individuals. These observations lend further support to a n association between increased macronutrient intake during childhood and the subsequent risk of breast cancer in adulthood. LITERATURE CITED Andersen, E (1968) Skeletal maturation of Danish schoolchildren in relation to height. Sexual development and social conditions. Aarhus: Universitatsforlaget. Armstrong, B and Doll, R (1975) Environmental factors and cancer incidence and mortality in different coun- CHILDHOOD GROWTH AN13 ADULT BREAST CANCER 535 tries, with special reference to dietary practices. Int. J. Farkas, GY (1966) Die Anderung der wichtigsten KorCancer 15~617-631. permasse der Kinder von Szeged (Sudungarn) Zwischen dem 3 und 18 lebensjahre. Acta Biol. (Szeged) 12:l-2. Ashcroft, MT, Heneage, P, and Lovell, HA (1966)Heights and weights of Jamaican schoolchildren of various eth- Froelich, J W (1970)Migration and plasticity of physique nic groups. Am. J. Phys. Anthropol. 24:35-44. in the Japanese-Americans of Hawaii. Am. J. Phys. Backstrom-Jarvinen, L (1964) Heights and weights of Anthropol. 22:429-442. Finnish children and young adults. Ann. Paediatr. Garcia-Almansa, A, Fernandez, MD, and Palacios MaSupplement 23,116 pp. teos, J M (1969) Patrones de crecimiento de 10s ninos espanoles normales. Rev. Clin. Espanola 113:45-48. Berg, JW (1975) Can nutrition explain the pattern of international epidemiology of hormone-dependent can- Garenne, M (1985)Office de la Recherche Scientifique et cers? Cancer Res. 35:3345-3350. Technique Outre-Mer (ORSTOM), Republic of France, Personal communication. Bindon, JR (1984a) The body build and body composition of Samoan children: Relationships to infant feeding Garn, SM, Ryan, AS, and Higgins, MW (1982) Implicapatterns and infant weight-for-length status. Am. J. tions of fatness and leanness. Am. J. Phys. Anthropol. Phys. Anthropol. 63:379-388. 57:191. Bindon, JR (1984b) An evaluation of the diet of three Gaskill, SP, McGuire, WL, Osborne, CK, and Stern, MP groups of Samoan adults: Modernization and dietary (1979) Breast cancer mortality and diet in the United adequacy. Ecol. Food Nutr. f4:105-115. States. Cancer Res. 39:3628-3637. Bindon, JR and Baker, PT (1985) Modernization, migra- Gavrilovic, A (1971) The anthropometrical research of tion and obesity among Samoan adults. Ann. Hum the first and second generation of the descendants of Biology 12:67-76. people from Lika settled in Vojvodina. Srpsko Biolosko Drustvo, Novi Sad, 80 pp. (in Yugoslav with English Bjarnason, 0, Day N, Snaedel, G, and Tulinius, H (1974) summary). The effect of year of birth on the breast cancer ageincidence curve in Iceland. Int. J. Cancer 13:689-696. Government of American Samoa (1975) Health Care Chart Book. American Samoa: Department of Medical Block, G (1982)A review of validations of dietary assessServices. ment methods. Am. J. Epidemiol. f15:492-505. Borrie, WD (1967) Malthusian reflections on the South Gray, GE, Pike, MC, and Henderson, BE (1979) Breast cancer incidence and mortality rates in different counPacific. Trans. R. Soc. New Zealand 2r19-29. tries in relation to known risk factors and dietary Brinkley, D, Carpenter, RG, and Haybittle, JL (1971) An practices. Br. J. Cancer 39:l-7. anthropometric study of women with cancer. Br. J. Prev. SOC. Med. 25:65-75. Greulich, WW (1957) A comparison of the physical growth and development of American-born and native Buell, P (1973) Changing incidence of breast cancer in Japanese children. Am. J. Phys. Anthropol. 15r489Japanese-American women. J. Natl. Cancer Inst. 515. 5fr1479-1483. Buell, P and Dunn, J (1965) Cancer mortality of Japa- Hahn, 0 and Koldovsky, 0 (1966) Utilization of Nutrients During Postnatal Development. London: Pernese Isei and Nisei of California. Cancer 18:656-664. gamon Press. Chang, KSF (1969) Growth and Development of Chinese Children and Youth in Hong Kong. Hong Kong: Uni- Hamburg, City of (1962) Die Schulkinder-Messung und Wagung in Mai/Juni Freie und Hansestadt Hamburg versity of Hong Kong. Gesundheitsbehorde, Medizinalstatistik (mimeograph). Charzewska, J (1973) Normal values of body height and weight in Warsaw children. Roczniki Panstwowego Haymond, MW, Karl, IE, and Pagliari, AS (1974) Increased gluconeogenic substrates in the small-for-gesZakladu Higieny 24:617-625 (in Polish with English tational-age infant. N. Engl. J. Med. 291:322. summary). Crews, DE (1985)Mortality, Survivorship and Longevity Heimendinger, J (1964) Die Ergebnisse von Korpermessungen a n 5000 Basler Kindern von 2-18 Jahren. Helv. in American Samoa 1950 to 1981. PhD dissertation, Paediatr. Acta Vol. 19, Suppl. 13. Department of Anthropology. Critescu, M (1969) Aspecte ale cresterii si dezvoltarii Henderson, BE (1979) Discussion of the hormonal basis of breast cancer. Second Symposium on Epidemiology adolescentilor din Republica Socialista Romania. Buand Cancer Registries in the Pacific Basin. Natl. Cancharest: Editura Academiei Republicii Socialiste cer Inst. Monogr. 53:192-193. Romania. Cusminsky, M and Lozano, GA (1974) Investigacion del Himes, JH and Roche, AF (1985) Subcutaneous fat and stature: relationships from infancy to adulthood. Ann. crecimiento y desarrollo del nino de 4 and 12 anos. Hum. Biol. l2[Suppl. 1]:55. Ministerio de Bienestar Social, La Plata. Pennsylvania Hoadley, JS (1980) Aid, politics and hospitals in western State University, State College, PA. Samoa. World Dev. 8:443-455. Demirjian, A, Jenicek, M, and Dubuc, MB (1972) Les normes staturo-ponderales de I’enfant urbain canadien Horm, JW,Asire, AJ, Young, JL, and Pollack, ES (1984) SEER Program: Cancer Incidence and Mortality in the francais d’age scolaire. Can. J. Public Health 63:14United States 1973-1981. NIH Publ. No. 85-1837, 30. USPHS, Bethesda, MD: National Cancer Institute. DeSouza, I, Morgan, L, Lewis, UJ, Raggatt, PR, Salih, H, and Hobbs, JR (1974) Growth hormone dependence Indian Council of Medical Research (1972) Growth and among human breast cancers. Lancet 2:182-184. physical development of Indian infants and children. Technical Report Series No. 18, New Delhi: ICMR. Dines, DR, Anderson, NE, and Gorman, DF (1980) The nutritional status of children in western Samoa. J. Iversen, I (1962) Beretningfra avdelingfor skollelegevesTrop. Pediatr. 26:95-99. enfor skolearet 1959-60. In Beretning fra Oslo helserad for aret 1960. Oslo: J Chr Gunderson, pp. 128Dunn, JE (1977) Breast Cancer among American Japa134. nese in the San Francisco Bay area. In Epidemiology and Cancer Registries in the Pacific Basin-I. Natl. Johnston, FE (1981) Physical growth and development and nutritional status: Epidemiological consideraCancer. Inst. Monogr. 47:157-160. tions. Fed. Proc. 40:2583-2587. Eveleth, PB and Tanner, JM (1976) Worldwide Variations in Growth. Cambridge: Cambridge University Johnston, FE (1983) The uses of anthropometry. Acta Med. Auxol. 15:69-74. Press. 536 M.S. M ICOZZI Jones, DL, Hemphill, W, and Meyers, ESA (1973) Height, weight and other physical characteristics of New South Wales children. Part I. Children aged five years and over. New South Wales: Department of Health G, 96543-aK5705. Kadanof, D and Mutafov, S (1969) Uber das Wachstumpstempo und die korperliche Entwicklung von Kindern und Jugendlichen von 3 bis 18 Jahren. Z. Morphol. Anthropol. 61.258-271. ‘ondo, S and Eto, M (1975) Physical growth studies on Japanese-American children in comparison with native Japanese. In: Proceedings of Meeting for Review and Seminar of the U.S.-Japan Cooperative Research es, (JIBP Synthesis, Vol. 1). Tokyo: University of Tokyo Press, pp. 13-45. Knott, VB (1963) Stature, leg girth, and body weight of Puerto Rican private school children measured in 1962. Growth 27: 157-174. Krogman, WM (1970) Growth of the head, face, trunk, and limbs in Philadelphia white and Negro children of elementary and high school age. Monogr. SOC.Res. Child Dev. 35:l-80. Kurihara, M, Aoki, K, and Tominaga, S (1984) Cancer Mortality Statistics in the World. Nagoya, Japan: University of Nagoya Press. Laska-Mierzejewska, T (19671 Desarrolla y maduracion de 10s ninos y jovenes Habaneros. Materialy i Prace Antropologiczne 74:9-64. Ljung, B, Brucefors, A, and Lindgren, G (1974) The secular trend in physical growth in Sweden. Ann. Hum. Biol. 1:245-256. Lubin, JH, Burns, PE, Blot, WJ, Ziegler, RG, Lees, AV, and Fraumeni, JF (1981) Dietary factors and breast cancer risk. Int. J. Cancer 28.685689. Marcondes, E, Berquo, ES, Yunes, J, Luongo, J, Martins, JS, Zacchi, MAS, Levy, MSF, and Hegg, R (nd) Estudo antropometrico de criancas brasileiras de zero a doze anos de idade. Anais Nestle 84:l-200. Masse, G (1969)Croissance et developpement de l’enfant a Dakar. Biometr. Hum. 4.13-23. Meredith, HV (1971) Worldwide somatic comparisons among contemporary human groups of adult females. Am. J. Phys. Anthropol. 34.89-132. Micozzi, MS (1985) Nutrition, body size and breast cancer. Yearbook Phys. Anthropol. 28:175-206. Micozzi, MS (1986) Childhood Nutrition, Growth and Development: Relation to the Long-Term Risk of Breast Cancer in Human Populations. PhD dissertation in Biomedical Anthropology, University of Pennsylvania, Philadelphia. Ann Arbor: University Microfilms Int Miller, AB (1977) Role of nutrition in the etiology of breast cancer. Cancer 39:2704-2708. Miller, AB, Kelly, A, Choi, NW, Matthews, V, Morgan, RW, Munan, L, Burch, JD, Feather, J, Howe, GR, and Jain, M (1978) A study of diet and breast cancer. Am. J. Epidemiol. 107:499-509. Moolgavkar, SH, Day, NE, and Stevens, RG (1980)Twostage model for carcinogenesis: Epidemiology of breast cancer in females. J. Natl. Cancer. Inst. 65559-569. National Coordinating Center (1965) The study and development of Filipino children and youth. Bulletin No. 1,Series 1965, Quezon City, Philippines: NCC. New Zealand Department of Health (1971) Physical development of New Zealand schoolchildren, 1969. Special Report No. 38, Health Services Research Unit, Wellington: Department of Health. Panek, S and Piasecki, E (1971)Nowa Huta: Integration of the population in the light of anthropological data. Materialy i Prace Antropologiczne 80.1-249. Petrakis, NL, Ernster, VL, Sacks, ST, King, EB, Schweitzer, TKJ, Hunt, TK, and King, MC (1981)Epidemiology of breast fluid secretion: Association with breast cancer risk factors and cerumen type. J. Natl. Cancer Inst. 67:277. Pinna, P (1961) Rilievi anthropometrici nei bambini di Sassari fra un mese e dodici anni. Ann. Ital. Pediatr. 24~30-53. Phillips, RL (1975) Role of life-style and dietary habits in risk of cancer among Seventh-Day Adventists. Cancer Res. 35:3515-3522. Prokopec, M, Suchy, J, and Titbachova, S (1973)Results of the third whole-state investigation of the youth in 1971 (Czech countries). Cesk. Pediatr. 28:341-346 (in Czech with English summary). Roche, AF (1984)Anthropometric methods: New and old. What they tell us. Int. J. Obesity 8.509-523. Ross, MH, Bras, G, and Lustbader, ED (1983)Diet, body weight and tumor susceptibility. 28th Scientific Report, Philadelphia: Institute for Cancer Research, Fox Chase Cancer Center, pp. 18-20. Sabharwal, KP, Morales, S, annd Mendez, J (19661 Body measurements and creatinine excretion among upper and lower socio-economicgroups of girls in Guatemala. Hum. Biol. 38:131-140. Sempe, P, Sempe, M, and Pedron, G (1971)Croissance et Maturation Osseuse. Paris: Theraplix. Shiloh, A, and Yekutiel, M (1958) Weights and heights of Israeli children. Acta Med. Oriental. 17.17-23. Stini, WA (1978) Early nutrition, growth, disease and human longevity. Nutr. Cancer 1.3-39. Stracker, OA (1964)Die gegenwartigen Korpermasse der Kinder und Jugendlichen. Wiener Med. Wochenschr. 244r816-818. Tanner, JM, Whitehouse, RH, and Takaishi, M (1966) Standards from birth t o maturity for height, weight, height velocity and weight velocity: British children 1965. Arch. Dis. Child. 41:454-471,613-635. Tatafiore, E (1970) Aggiornamento dei dati medi napoletani di Deso e statura. Infanzia 20t17-32. Taylor, RH and Zimmet, PZ (1981) Obesity and diabetes in western Samoa. Int. J. Obesity 5:367-376. Tokyo Department of Maternal and Child Health (1970) Physical Status of Japanese Children in 1970. Tokyo: Institute of Public Health. Twiesselmann, F (1969) Development Biometrique de 1’Enfant a 1’Adulte. Brussels: Presses Universitaires de Bruxelles. United Nation’s Children’s Fund (UNICEF) (1984)World Statistics on Children. First edition, New York: UNICEF. United Nation’s Children’s Fund (UNICEF) (1985) The State of the World’s Children. London: Oxford University Press. Valaoras, V and Laros, K (1969) Biometric characteristics of Greek pupils in elementary schools. IATRIKI 15266-276 (in Greek with English summary). Villarejos, VM, Osborne, JA, Payne, FJ, and Arguedes, JA (1971) Heights and weights of children in urban and rural Costa Rica. Environ. Child Health 27:31-43. Vlastovsky, VG, Grachera, GS, Minkina, VA, and Schevchenko, LG (nd) Unpublished data on Moscow children 1969-1970. Wadsworth, GR and Lee, TS (1960) The height, weight, and skinfold thickness of Muar schoolchildren. J. Trop. Pediatr. 6.48-54. Waterhouse, J, Muir, C, Shanmugaratnam, K, and Powell, J (1982) Cancer Incidence in Five Continents, Vol. N.Lyons: IARC Scientific Publ. No. 42. CHILDHOOD GROWTH AND ADULT BREAST CANCER Waterlow, JC, Buzina, R, Keller, W, Lane, JM, Nichaman, M Z , and Tanner, JM(1977)The presentation and use of height and weight data for comparing the nutritional status of groups of children under the age of 10 years. Bull. WHO 55,489-498. Wieringen, JC van, Wafelbakker, F, Verbrugge, HP, and de Haas, JH (1971) Growth Diagrams 1965, Netherlands. Groningen: Walters-Noordhoff Publishing. Wong Hock Boon, Tye Cho Yoke, and Quek Kai Miew 537 (1972) Anthropometric studies on Singapore children. I. Heights, weights and skull circumference on preschool children. J. Singapore Paediatr. Soc. 14,6849. Wood, CS and Gans, LP (1981) Hematological status of reproductive women in Samoa: An analysis of biometric data. Hum. Biol. 53:268-279. Wood, CS and Gans, LP (1984)Aspects of child health in western Samoa: A medical anthropological view. J. Trop. Pediatr. 30:104-110.