Adherence to WCRF/AICR Cancer Prevention Recommendations and Risk of Postmenopausal Breast Cancer

Several diet-quality indexes have been developed on the basis of recommendations aimed at preventing chronic disease. In previous studies, most diet indexes have not predicted total cancer incidence (1–5) or total cancer mortality (6, 7), with some exceptions (8–10). However, in most cases, lifestyle indexes that add other modifiable risk factors such as body weight, physical activity, and smoking have been found to predict total cancer incidence (11–14) and mortality (10, 11, 15–19).

Because the association between modifiable risk factors and cancer outcomes differs by cancer site (20), looking at total cancer incidence or mortality may mask associations for individual cancers. Although breast cancer risk is associated with several modifiable risk factors including alcohol consumption and body weight (20, 21), evidence is mixed as to whether other individual dietary behaviors (22–25), diet indexes (13), or lifestyle indexes (13, 16) are associated with breast cancer incidence. Several diet-quality indexes have failed to predict total breast cancer risk and mortality (23, 26, 27); however, associations between diet and breast cancer may differ by menopausal status (26). Similarly, the association between body mass index (BMI) and breast cancer risk differs for pre- and postmenopausal women (20). In addition, if not all of the behaviors considered are associated with reduced risk of breast cancer, weighting each equally when assessing the association between a group of behaviors and breast cancer incidence will yield misleading results.

On the basis of a comprehensive literature review, in 2007 the World Cancer Research Fund (WCRF) and the American Institute for Cancer Research (AICR) released 8 recommendations related to body fatness, diet, and physical activity aimed at reducing incidence of the most common cancers worldwide (20). One recent study examined the association between these recommendations and total and site-specific cancer incidence and reported a statistically significant 16% reduction in risk of breast cancer in women with the highest concordance with the recommendations compared with women with the lowest concordance in the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort (12). However, combining pre- and postmenopausal women may have attenuated their results because several of the recommendations were developed on the basis of findings for postmenopausal breast cancer.

The purpose of this study is to assess whether an index of behaviors related to body fatness, physical activity, and diet based on the 2007 WCRF/AICR cancer prevention recommendations is associated with reduced breast cancer incidence in a cohort of postmenopausal women in the United States and, if so, to evaluate which recommendations are most strongly associated with reduced risk.

The Vitamins and Lifestyle (VITAL) study is a prospective cohort study designed to investigate the associations between use of dietary supplements and cancer risk and has previously been described in detail (28). Participants were eligible to join the cohort if they were between the ages of 50 and 76 and lived in one of the 13 counties in the Western Washington Surveillance, Epidemiology and End Results (SEER) cancer registry at baseline. Recruitment of women into the VITAL cohort is described below.

Using names purchased from a commercial mailing list, baseline questionnaires were mailed to 168,953 women between October 2000 and December 2002 and followed by reminder postcards. A total of 41,157 questionnaires were returned and 40,337 passed quality control checks. Overall, 30,797 women were included in our analyses after excluding women with a history of breast cancer or with incomplete baseline medical history (n = 3,164); premenopausal women (n = 1,356); those who were missing BMI (n = 1,619) or physical activity (n = 509) data; those whose food frequency questionnaires (FFQ) had less than 5 items completed on a single page (n = 2,334) or whose estimated energy consumption was less than 600 calories (n = 1,477) or more than 4,000 calories per day (n = 137); and women diagnosed with rare histologic types of breast cancer (e.g., sarcoma, lymphoma, phyllodes) postbaseline (n = 9; numbers of exclusions reported are not mutually exclusive).

The Institutional Review Board at the Fred Hutchinson Cancer Research Center approved this research.

Baseline questionnaires included medical history, self-reported height and weight, physical activity over the previous 10 years, reproductive history, medication use, and a 126-item FFQ covering diet in the year before baseline.

The FFQ was adapted from the questionnaire developed for use in the Women's Health Initiative and other studies. The University of Minnesota's Nutrition Coding Center database was used to convert food frequency information into nutrients (29). Number of servings was calculated on the basis of frequency of intake and whether respondents indicated their usual portion size was small, medium, or large compared with a provided sex-specific medium portion size of each food and beverage. Validity studies of similar FFQs have reported correlation coefficients of 0.41 for servings of fruits and vegetables compared with food records and 0.57 to 0.61 compared with dietary recalls (30, 31), 0.50 to 0.59 for measures of red and processed meat intake compared with diet records (32), and 0.89 for alcohol consumption compared with a combination of diet recalls and food records (33).

Physical activity was assessed by a one-page questionnaire that asked about walking (including usual pace) and other moderate and strenuous recreational activities done at least once per week in the previous 10 years. The questionnaire included categorical questions on minutes per day, days per week, and number of years they participated in each activity. Estimates of metabolic equivalent task (MET)-hours per week were validated against a detailed in-person interview on 10-year recreational physical activity among a subsample of VITAL respondents (age-adjusted correlation coefficient among women: 0.70; ref. 34).

Body fatness was assessed using self-reported height and weight at baseline. Estimates of overweight status on the basis of self-reported height and weight have been found to have 84% to 95% sensitivity and 96% to 99% specificity among women ages 50 to 79 (35).

The main exposures of this study were whether or not respondents met each individual WCRF/AICR cancer prevention recommendation and the number of recommendations met. The recommendations include 8 broad guidelines, with between 1 and 4 more specific personal recommendations and public health goals for each. An expert panel of 4 nutritional epidemiologists (R.E. Patterson, Alan R. Kristal, S.A.A. Beresford, and E. White) with knowledge of the VITAL cohort data made recommendations on the operationalization of 6 of the 8 recommendations. The key components of each recommendation were identified (noted in italics in Table 1) and specific cutoffs for meeting the key components were selected on the basis of information provided in the recommendations or from external sources (e.g., the World Health Organization; ref. 36). The recommendation to limit salt-preserved foods and moldy cereals and legumes was not operationalized because those exposures are not common in the food supply in the United States and data were not available in VITAL. The recommendation to meet nutritional needs through diet alone was not operationalized because, while it did not recommend using dietary supplements, it also did not recommend against supplement use. Detail about the operationalization of the remaining recommendations is below.

The recommendation to be as lean as possible within the normal range of body weight was operationalized as having a BMI of at least 18.5 but less than 25 kg/m² on the basis of height and weight reported at baseline. This range was based on that set by the World Health Organization (36) as normal weight, as suggested in the recommendation (20).

For women with missing weight at baseline but who reported BMI at age 45 (n = 685), BMI at baseline was imputed by calculating the average annual change in BMI within 18 age- and race/ethnicity-specific strata (<60, 60–69, and 70+ for age; white, Hispanic, African-American, American Indian/Alaska Native, Asian, or Pacific Islander, and other for race/ethnicity), multiplying that value by the difference between baseline age of respondents and 45, and adding the product to their BMI at age 45.

The recommendation to be physically active as a part of everyday life (equivalent to brisk walking for at least 30 minutes every day) was operationalized as engaging in moderate or fast walking and/or other moderate or strenuous activity for an average of at least 30 minutes per day, at least 5 days per week, and in at least 7 of the past 10 years. Respondents who were missing data for this constructed variable but whose physical activity responses were complete enough to estimate their MET-hours per week of walking and/or moderate/strenuous physical activity (n = 1,950) were categorized as meeting this recommendation if they engaged in an average of at least 10 MET hours of walking and/or moderate/strenuous physical activity per week (using an estimate of 4.0 METs/hour × 0.5 hours/day × 5 days/week) over the previous 10 years.

The recommendation to limit consumption of energy-dense foods and to avoid sugary drinks was operationalized as consuming a diet where the energy density of foods consumed was less than 125 kcal per 100 g on the basis of responses to the FFQ and also consuming less than 1 serving of regular (not diet) soda, fruit drinks, and/or cranberry juice per week. Fruit juices which typically do not have added sugar (e.g., orange juice) were not counted as sugary drinks. The energy density cutoff was based on a public health goal included in the WCRF/AICR recommendation that the average energy density of diets should be lowered toward 125 kcal per 100 g.

The recommendation to eat mostly foods of plant origin, specifically the personal recommendations to eat at least 5 servings of a variety of non-starchy vegetables and fruits every day and to eat relatively unprocessed grains and/or legumes with every meal, were operationalized as consuming at least 5 servings of fruits and/or non-starchy vegetables and also at least 1 serving of whole grains and/or legumes per day. Servings of fruits and non-starchy vegetables included 25 foods or food groups, adjusted by portion size and by summary questions on total numbers of fruits and vegetables eaten to reduce overestimation by participants. It excluded fruit juices and potatoes. Because the VITAL FFQ only included 5 items on whole grains (covering breads and breakfast cereals) and, therefore, failed to fully separate whole grains from other grains (e.g., brown rice from white rice), we used a cutoff of 1 serving per day rather than per meal to represent those who habitually eat whole grains and/or legumes. Legume servings included 3 items on bean dishes and one item on tofu and tempeh.

The recommendation to limit intake of red meat and to avoid processed meat was operationalized as consuming less than 18 ounces of red or processed meat per week. Red and processed meat from mixed dishes in the FFQ were also included by assuming that red or processed meat accounted for one-quarter of their weight.

The recommendation to limit alcoholic drinks was operationalized as consuming no more than one alcoholic beverage per day on average, where a drink was classified as a 12-ounce bottle or can of beer; 4-ounce glass of red, white, or rosé wine; or one shot (1.5 ounces) of liquor or one mixed drink.

Incident primary breast cancers were ascertained through December 31, 2008 by annual linkage with the Western Washington SEER cancer registry. All incident cancers diagnosed in the 13 counties of the Western Washington SEER registry (except for nonmelanoma skin cancers) are reported to SEER by all area hospitals and by offices of pathologists, oncologists, and radiotherapists. Linkage between VITAL and SEER is largely automated and based on ranking agreement between items common to both sets of data, such as Social Security number, name, and date of birth. Matches with high concordance are linked automatically. Visual inspection is used to adjudicate incomplete matches. A total of 899 incident invasive breast cancers were identified among the women eligible for this analysis in an average of 6.7 years of follow-up.

Participants who did not develop breast cancer were right-censored at the earliest of the following events: date they requested removal from the study (n = 12), date they moved out of the Western Washington SEER catchment area (n = 1,617), date of death (n = 1,318), or December 31, 2008 (n = 26,712). Because in situ breast cancers are commonly treated and not all would be expected to progress to invasive disease, women diagnosed with in situ breast cancer (n = 239) were not considered as meeting the case definition and were censored at the date of diagnosis. Moves out of the SEER catchment area were identified through linkage with the National Change of Address System and deaths occurring in Washington State were ascertained via linkage with the Washington State death file using linkage procedures similar to that for linkage to SEER.

HRs and 95% confidence intervals (CI) of incident, invasive breast cancer associated with meeting (vs. not meeting) each individual recommendation and for the number of recommendations met (vs. meeting none) were estimated using Cox proportional hazards models. We used participant age as the time scale, with participants entering the analysis at the age they completed the baseline questionnaire and exiting at age at diagnosis of invasive breast cancer or age at censoring event. Proportional hazards assumptions were examined using scaled Schoenfeld residuals. No significant (P < 0.05) deviations from proportionality were observed. P_trend values were calculated using the Wald test associated with modeling the number of recommendations met as a continuous variable. All statistical tests were two-sided.

Multivariate analyses included adjustment for potential confounders selected a priori, including known risk factors for breast cancer. These analyses included categorical variable adjustment for education, race, age at menarche, age at birth of first child, years of combined estrogen plus progestin hormone therapy use, age at menopause, receipt of a mammogram in the 2 years before baseline, and history of breast cancer in a first-degree relative using the categories in Table 2, as well as adjustment for kilocalories of average daily energy intake (continuous). Participants with missing data were treated as their own category for each potential confounder.

In additional analyses, binary variables indicating whether respondents met each of the 6 recommendations were included in one model with the potential confounders. The least-protective recommendation (the one with the largest HR) was subsequently removed, the model was run again, and the process was repeated until only the most protective recommendation remained. Each subset of recommendations identified through this process was evaluated in relation to breast cancer risk. Analyses were conducted using Stata 12.1 (StataCorp LP).

Table 1 gives the proportion of women in the VITAL cohort and those diagnosed with breast cancer during follow-up who met each recommendation. More than 70% of the women in the cohort and of cases met the recommendations to limit consumption of alcohol and of red and processed meat. Less than 40% met the recommendations on body fatness, consumption of energy dense foods and sugary drinks, plant foods, and physical activity.

The average age at baseline was 61.3 years for women in the cohort and 63.2 for cases. Table 2 gives baseline characteristics of women in the overall study population and those diagnosed with invasive breast cancer during follow-up. Women in both groups were predominantly white, and more than one-third of each group completed college. Compared with the cohort as a whole, a larger proportion of women diagnosed with breast cancer reported having at least one first-degree relative with a history of breast cancer and reported using hormone therapy for at least 5 years (P < 0.05 for both).

Table 3 gives HRs and 95% CIs associated with meeting (vs. not meeting) each individual recommendation. In covariate-adjusted analyses, meeting the recommendation to limit alcohol to no more than 1 drink per day was associated with a 37% reduction in breast cancer risk compared with consuming more than one drink per day (HR, 0.63; 95% CI, 0.54–0.75) and meeting the recommendation to consume mostly plant foods was associated with a 21% reduction in risk (HR, 0.79; 95% CI, 0.64–0.97) compared with not meeting the recommendation. Breast cancer risk was 13% lower in women whose BMI was in the normal range at baseline relative to those with BMIs above or below the normal range (HR, 0.87; 95% CI, 0.76–1.00). After also adjusting for whether respondents met each of the other recommendations, the recommendations related to body fatness and alcohol remained associated with significant reductions in breast cancer risk.

Table 4 gives HRs and 95% CIs associated with the number of recommendations met. After adjusting for potential confounders, each additional recommendation met was associated with an 11% (95% CI, 5%–16%) reduction in breast cancer risk (P_trend < 0.001). In adjusted analyses, compared with meeting no recommendations, meeting one recommendation was associated with a marginally significant 32% reduction in breast cancer risk, meeting 2 to 4 recommendations was associated with a statistically significant 44% to 48% reduction in risk, and meeting 5 to 6 recommendations was associated with a 60% reduction in risk (HR, 0.40; 95% CI, 0.25–0.65; P_trend = 0.001).

Sensitivity analyses excluding women missing BMI and physical activity data rather than imputing their values did not alter our results (data not shown).

To identify specific subsets of recommendations most associated with reduced breast cancer risk, additional analyses were conducted in which the fully adjusted model presented in Table 3 was repeated with the recommendation least associated with reduction in breast cancer risk sequentially removed. Results given in Table 5 suggest that, compared with meeting no recommendations, the reduction in breast cancer risk observed for women meeting the recommendations related to body fatness, plant foods, and alcohol intake (HR, 0.38; 95% CI, 0.25–0.58 vs. meeting none of the three) is at least as great as for meeting any combination of 5 to 6 of all of the recommendations considered (HR, 0.40; 95% CI, 0.25–0.65; Table 4). Modest additional benefit is achieved from also meeting the recommendations related to red and processed meat and physical activity.

Results appeared to be somewhat stronger for estrogen-receptor (ER)-negative breast cancer (per-recommendation HR, 0.84; 95% CI, 0.72–0.99) than for ER-positive breast cancer (per-recommendation HR, 0.90; 95% CI, 0.85–0.96). As for all breast cancer cases, body fatness, plant foods, and alcohol accounted for most of the association with reduced ER-positive breast cancer risk, where red and processed meat, body fatness, and alcohol accounted for most of the association with ER-negative breast cancer (data not shown).

In this cohort of postmenopausal women, meeting at least 5 of the WCRF/AICR recommendations was associated with a 60% reduction in breast cancer risk compared with meeting no recommendations (HR: 0.40; 95% CI, 0.25–0.65). Further analyses suggest that this association is primarily due to meeting recommendations related to body fatness, plant foods, and alcohol (HR for meeting all vs. none of these recommendations: 0.38; 95% CI, 0.25–0.58). The results based on counts of the 6 recommendations are somewhat misleading, as each recommendation does not have the same impact on breast cancer risk; rather, as the number of recommendations met increases, the likelihood that a woman has met the three key recommendations increases.

In the previous study that examined the association between concordance with the WCRF/AICR recommendations and overall and site-specific cancer incidence in the EPIC cohort, each additional point on a recommendation score was associated with a 5% (95% CI, 3%–7%) reduction in breast cancer risk and a recommendation score of 5 points or more was associated with a HR for breast cancer incidence of 0.84 (95% CI, 0.78–0.90) compared with a score of 3 points or less (12).

Although we cannot directly compare the much larger risk reduction for our highest versus our lowest recommendation group to that reported previously due to differences in operationalizing the recommendations and in defining the high and low groups, our result of an 11% decreased risk per recommendation met (HR, 0.89; 95% CI, 0.84–0.94) is more than twice as large a reduced risk as in the EPIC cohort. Several differences between our study and the EPIC study may have contributed to the different magnitude of our findings. Previous research suggests that both the strength and the direction of the association between modifiable risk factors and breast cancer differ in pre- and postmenopausal women (e.g., body fatness is inversely associated with breast cancer risk in premenopausal women, but positively associated in postmenopausal women; ref. 20). Thus, the results of the previous study may have been attenuated by including premenopausal women (33.4% of women were premenopausal at baseline; ref. 12). In addition, the EPIC study included the WCRF/AICR special recommendation that mothers should aim to breastfeed exclusively for 6 months (12), although this difference would have made the EPIC findings stronger relative to ours (breastfeeding data were not available in VITAL). Furthermore, the current study operationalized recommendations as being met or not met but, in the previous study, respondents received partial points for near-adherence (12).

To our knowledge, only two other studies of lifestyle indexes (i.e., those that include modifiable risk factors such as body weight and physical activity as well as diet) and cancer incidence have reported results for breast cancer in subanalyses (13, 16). Heidemann and colleagues examined the association between a diabetes risk score (including waist circumference, height, physical activity, age, hypertension, smoking, and intake of alcohol, red meat, whole-grain bread, and coffee) and breast cancer incidence (HR for breast cancer associated with ≥10% vs. <1% 5-year probability of developing diabetes: 1.3; 95% CI, 0.6–2.7; P_trend = 0.85); however, this estimate is based on only 9 breast cancer cases in the highest-risk group (16). In a study of the association between an index representing adherence to the Dietary Guidelines for Americans (including dietary behaviors, BMI and physical activity) and total and site-specific cancer incidence, Harnack and colleagues reported a HR for breast cancer of 0.76 (95% CI, 0.65–0.89) for women ages 55–69 in the highest quintile of adherence compared with women in the lowest quintile (13). When BMI and physical activity were removed from the index, leaving only the dietary behaviors, the association attenuated to a HR of 0.86 (95% CI, 0.73–1.00; ref. 13).

Strengths of this study include its large sample size, prospective design, and the detailed information collected at baseline that allowed us to operationalize 6 of the WCRF/AICR cancer prevention recommendations and control for several potential confounding factors. Linkage with the SEER database provided accurate and near-complete ascertainment of breast cancer outcomes. In additional analyses, we were able to identify the specific recommendations associated with reduction in postmenopausal breast cancer risk.

Another feature of our study was the use of an expert panel to operationalize the recommendations. Although this was designed to be a strength, our results might have differed had we used different interpretations. Therefore, we conducted sensitivity analyses using stricter standards for the recommendations related to body fatness (BMIs of 21–23 at baseline) and physical activity (60+ minutes/day on 5–7 days/week in all 10 years) on the basis of additional information in the recommendations. In adjusted analyses, meeting the stricter recommendations related to body fatness and physical activity was more strongly associated with reductions in breast cancer risk compared with the original classifications (HR for body fatness: 0.75; 95% CI, 0.61–0.93; HR for physical activity, 0.90; 95% CI, 0.72–1.12). In further analyses of the per-recommendation association with breast cancer risk using the stricter version of these two recommendations, each additional recommendation met was associated with a 14% reduction in breast cancer risk (HR, 0.86; 95% CI, 0.81–0.92), a somewhat stronger association compared with the original interpretation.

One limitation of our study is that those who chose to join the VITAL cohort may have had more positive health behaviors than the general population. Compared with women in Washington State, a larger proportion of VITAL participants were white, college-educated, maintained BMIs in the normal range, and consumed at least 1 alcoholic drink per day, and a lower proportion were current smokers (37). However, selection bias is not likely to affect results in a prospective study where future disease status is unknown at baseline. In addition, although analyses controlled for many potential confounding factors, residual confounding may exist due to missing or misspecified confounders. Measurement error in our assessment of meeting individual recommendations and the combined score would act to bias results toward the null. Finally, we did not have the data to include the special recommendation related to breastfeeding, which has been found to be protective against breast cancer (20). Therefore, the effect of the WCRF/AICR recommendations on breast cancer risk may be even greater than that found in this study.

This study is the first to examine adherence to the WCRF/AICR cancer prevention recommendations related to body fatness, diet, and physical activity and cancer risk in a U.S. population, and to examine the association between meeting the recommendations and risk of postmenopausal breast cancer. These results indicate that meeting the WCRF/AICR cancer prevention recommendations related to body fatness, plant foods, and alcohol is associated with substantial reductions in risk of invasive breast cancer in postmenopausal women.

No potential conflicts of interest were disclosed.

Conception and design: T.A. Hastert, A.R. Kristal, E. White

Development of methodology: T.A. Hastert, S.A.A. Beresford, R.E. Patterson, A.R. Kristal, E. White

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R.E. Patterson, A.R. Kristal, E. White

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T.A. Hastert, S.A.A. Beresford, R.E. Patterson, A.R. Kristal, E. White

Writing, review, and/or revision of the manuscript: T.A. Hastert, S.A.A. Beresford, R.E. Patterson, A.R. Kristal, E. White

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. White

Study supervision: E. White

This work was financially supported by the Biobehavioral Cancer Prevention and Control Training Program at the University of Washington funded by the National Cancer Institute R25CA92408 (to T. A. Hastert) and by the National Cancer Institute and the National Institutes of Health Office of Dietary Supplements K05CA154337 (to E. White).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.