Empirically derived dietary patterns and risk of postmenopausal breast cancer in a large prospective cohort study1–3 Ellen M Velie, Catherine Schairer, Andrew Flood, Jian-Ping He, Ravindra Khattree, and Arthur Schatzkin ABSTRACT of total fat and refined carbohydrate may increase breast cancer Downloaded from https://academic.oup.com/ajcn/article/82/6/1308/4648925 by guest on 24 September 2021 Background: Inconsistent associations have been reported between risk (2, 4, 5), and total vegetable intake—particularly cruciferous diet and breast cancer. vegetable intake (6)—phytoestrogens (4), and fruit intake (2) Objective: We prospectively examined the association between may reduce risk, but the results of recent cohort studies measur- dietary patterns and postmenopausal breast cancer risk in a US-wide ing adult diet do not support an association (3, 7–9). cohort study. The examination of dietary patterns has been suggested as an Design: Data were analyzed from 40 559 women who completed a alternative to quantify aggregate diet risk for chronic disease self-administered 61-item Block food-frequency questionnaire in (10–16). Results from studies of single nutrients and foods may the Breast Cancer Detection Demonstration Project, 1987–1998; be inconsistent because they cannot disaggregate individual ef- 1868 of those women developed breast cancer. Dietary patterns were fects of highly correlated foods and may be unable to account for defined by using principal components factor analysis. Cox propor- synergistic interactions of food combinations and constituents or tional hazard regression was used to assess breast cancer risk. other factors that may affect nutrient bioavailability (eg, cooking Results: Three major dietary patterns emerged: vegetable-fish/ practices) (10–18). Dietary patterns are also more significantly poultry-fruit, beef/pork-starch, and traditional southern. The associated with overall mortality and lowered risk of heart dis- vegetable-fish/poultry-fruit pattern was associated with higher ed- ease, high blood pressure, and possibly cancer than are single ucation than were the other patterns, but was similar in nutrient nutrients (15, 16, 18–21). intake to the traditional southern pattern. After adjustment for con- One common approach to defining empirical patterns of food founders, there was no significant association between the vegetable-fish/poultry-fruit and beef/pork-starch patterns and breast intake is exploratory factor analysis, which is a statistical method cancer. The traditional southern pattern, however, was associated used to combine correlated food items into a single exposure (12, with a nonsignificantly reduced breast cancer risk among all cases (in 22). Patterns identified by factor analysis have been shown to be situ and invasive) that was significant for invasive breast cancer reliable and reasonably valid (16, 23–26). Two primary patterns (relative hazard ҃ 0.78; 95% CI ҃ 0.65, 0.95; P for trend ҃ 0.003). have been identified in the United States through the use of this This diet was also associated with a reduced risk in women without approach. The first pattern is characterized by intake of vegeta- a family history of breast cancer (P ҃ 0.05), who were underweight bles, fruit, whole grains, low-fat dairy products, fish, poultry, and or normal weight [body mass index (in kg/m2) 25; P ϭ 0.02], or often wine; the second is characterized by intake of red meats, who had tumors positive for estrogen receptor (P ҃ 0.01) or pro- refined grains, fat, sweets, and alcohol (13, 15, 16, 27, 28). The gesterone receptor (P ҃ 0.003). Foods in the traditional southern first pattern has been associated with decreased risk of obesity, pattern associated with reduced breast cancer risk were legumes, low diabetes, colon cancer, and cardiovascular disease; the inverse mayonnaise–salad dressing intake, and possibly cabbage. has been shown for the second pattern (14–16). Conclusions: The traditional southern diet or its components are Dietary patterns have been suggested to be most useful in associated with a reduced risk of invasive breast cancer in postmeno- studying disease etiology when there is insufficient or conflicting pausal women. Am J Clin Nutr 2005;82:1308–19. 1 From the Department of Epidemiology, College of Human Medicine KEY WORDS Breast cancer, dietary patterns, nutrition, factor (EMV), and the Data Coordinating Center and Department of Epidemiology analysis, cohort study (J-PH), Michigan State University, East Lansing, MI; the Environmental Epidemiology Branch (CS) and the Nutritional Epidemiology Branch (AS), Division of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda, MD; the Department of Epidemiology, University of Minnesota, INTRODUCTION Minneapolis, MN (AF); and the Department of Mathematics and Statistics, Few modifiable risk factors for breast cancer have been iden- Oakland University, Rochester, MI (RK). 2 tified (1); therefore, potential dietary associations are of partic- Supported in part by the National Cancer Institute Cancer Prevention Fellowship Program and grant no. K07CA094984. ular interest. Associations between single nutrients and food 3 components of the diet and breast cancer have been widely stud- Address reprint requests to E Velie, Department of Epidemiology, Mich- igan State University, B601 West Fee Hall, East Lansing, MI 48824. E-mail: ied (2, 3), yet with inconsistent results (2, 4). The only well- [email protected]. established nutrition-related risk factors for postmenopausal Received March 3, 2005. breast cancer are obesity and alcohol intake (2, 4). High intakes Accepted for publication August 22, 2005. 1308 Am J Clin Nutr 2005;82:1308–19. Printed in USA. © 2005 American Society for Nutrition DIETARY PATTERNS AND BREAST CANCER 1309 evidence for diet associations, as is the case for breast cancer for the survey instrument. The frequency of consumption of each (13). Yet, only 3 published studies that we are aware of have food was multiplied by reported portion size to obtain gram examined dietary patterns and breast cancer risk (29–31). All intake, which was then multiplied by the nutrient content of that used principal components factor analysis (PCFA) to define pat- food (37). To standardize reports of food consumption into sim- terns. The first, which was conducted in Sweden, reported an ilar units, intakes were converted to medium servings by dividing increased risk associated with a “drinker” pattern (29). The sec- reported gram intake by the gram amount in a medium serving [as ond, which was conducted in northern Italy, observed a de- defined by the Block-NCI software (37)]. To also standardize creased risk associated with a raw salad vegetable pattern (30). diets to similar caloric intakes given different body sizes and The third, which was conducted among nurses in 11 northeastern physical activity levels, servings of daily food intake per 1000 US states, showed no overall association between a “prudent” or kcal were calculated by dividing the intake of each food item by Western pattern and breast cancer risk but a reduced risk asso- reported total energy intake and multiplying the result by 1000. ciated with the prudent pattern for estrogen-receptor-negative Questionnaires with reported caloric intakes of 400 or ͧ3800 tumors (31). kcal/d and those with ͧ30 skipped food items were considered Here we examine the association between dietary patterns of invalid and were excluded from the analyses (n ҃ 5080; 11% of Downloaded from https://academic.oup.com/ajcn/article/82/6/1308/4648925 by guest on 24 September 2021 food items derived from PCFA and their potential association the 46 331 eligible women who completed the questionnaire). with breast cancer risk among postmenopausal women in a large, prospective cohort study in the United States. Given evidence Covariates that family history of breast cancer (32), body mass index (BMI) Education level was obtained at the first screening visit (1973– (30), hormone receptor status (33, 34), and history of benign 1975). Information about family history of breast cancer in a breast disease (BBD) (35) may modify diet–breast cancer asso- first-degree relative, history of biopsies for BBD, self-reported ciations, we further examined potential interactions with these race, use of female hormones or oral contraceptives, age at men- factors. arche, parity, age at first live birth, and menopausal age was obtained at baseline and through annual telephone interviews SUBJECTS AND METHODS (1979–1986) and was updated through subsequent mailed follow-up questionnaires. Alcohol intake, physical activity, Study population weight, height, tobacco use, and average hours of weekday vig- Study subjects were participants in the Breast Cancer Detec- orous physical activity were reported during phase II. BMI was tion Demonstration Project (BCDDP) follow-up cohort study, defined as weight in kilograms divided by squared height in the details of which were provided previously (36). Briefly, par- meters. Region of residence was obtained from reported address ticipants were selected from 280 000 past participants in the during phase III, or if missing, phase IV, and was grouped into BCDDP breast cancer screening program conducted between Northeast (including New England and the mid-Atlantic states), 1973 and 1981 at 29 centers throughout the United States. Be- North/Midwest (including the Midwest, Mountain, and North- ginning in 1979, the National Cancer Institute (NCI) began a west regions), Southwest (including California