ORIGINAL CONTRIBUTION Low-Fat Dietary Pattern and Risk of Invasive Breast Cancer The Women’s Health Initiative Randomized Controlled Dietary Modification Trial Ross L. Prentice, PhD; Bette Caan, DrPH; Rowan T. Chlebowski, MD; Ruth Patterson, PhD; Lewis H. Kuller, MD; Judith K. Ockene, PhD; Karen L. Margolis, MD; Marian C. Limacher, MD; JoAnn E. Manson, MD; Linda M. Parker, DSc; Electra Paskett, PhD; Lawrence Phillips, MD; John Robbins, MD; Jacques E. Rossouw, MD; Gloria E. Sarto, MD; James M. Shikany, DrPH; Marcia L. Stefanick, PhD; Cynthia A. Thomson, PhD; Linda Van Horn, PhD; Mara Z. Vitolins, DrPH; Jean Wactawski-Wende, PhD; Robert B. Wallace, MD; Sylvia Wassertheil-Smoller, PhD; Evelyn Whitlock, MD; Katsuhiko Yano, MD; Lucile Adams-Campbell, PhD; Garnet L. Anderson, PhD; Annlouise R. Assaf, PhD; Shirley A. A. Beresford, PhD; Henry R. Black, MD; Robert L. Brunner, PhD; Robert G. Brzyski, MD; Leslie Ford, MD; Margery Gass, MD; Jennifer Hays, PhD; David Heber, MD; Gerardo Heiss, MD; Susan L. Hendrix, DO; Judith Hsia, MD; F. Allan Hubbell, MD; Rebecca D. Jackson, MD; Karen C. Johnson, MD; Jane Morley Kotchen, MD; Andrea Z. LaCroix, PhD; Dorothy S. Lane, MD; Robert D. Langer, MD; Norman L. Lasser, MD; Maureen M. Henderson, MD T Context The hypothesis that a low-fat dietary pattern can reduce breast cancer risk has existed for decades but has never been tested in a controlled intervention trial. Objective To assess the effects of undertaking a low-fat dietary pattern on breast cancer incidence. Design and Setting A randomized, controlled, primary prevention trial conducted at 40 US clinical centers from 1993 to 2005. Participants A total of 48 835 postmenopausal women, aged 50 to 79 years, without prior breast cancer, including 18.6% of minority race/ethnicity, were enrolled. Interventions Women were randomly assigned to the dietary modification intervention group (40% [n = 19 541]) or the comparison group (60% [n = 29 294]). The intervention was designed to promote dietary change with the goals of reducing intake of total fat to 20% of energy and increasing consumption of vegetables and fruit to at least 5 servings daily and grains to at least 6 servings daily. Comparison group participants were not asked to make dietary changes. Main Outcome Measure Invasive breast cancer incidence. Results Dietary fat intake was significantly lower in the dietary modification intervention group compared with the comparison group. The difference between groups in change from baseline for percentage of energy from fat varied from 10.7% at year 1 to 8.1% at year 6. Vegetable and fruit consumption was higher in the intervention group by at least 1 serving per day and a smaller, more transient difference was found for grain consumption. The number of women who developed invasive breast cancer (annualized incidence rate) over the 8.1-year average follow-up period was 655 (0.42%) in the intervention group and 1072 (0.45%) in the comparison group (hazard ratio, 0.91; 95% confidence interval, 0.83-1.01 for the comparison between the 2 groups). Secondary analyses suggest a lower hazard ratio among adherent women, provide greater evidence of risk reduction among women having a high-fat diet at baseline, and suggest a dietary effect that varies by hormone receptor characteristics of the tumor. Conclusions Among postmenopausal women, a low-fat dietary pattern did not result in a statistically significant reduction in invasive breast cancer risk over an 8.1year average follow-up period. However, the nonsignificant trends observed suggesting reduced risk associated with a low-fat dietary pattern indicate that longer, planned, nonintervention follow-up may yield a more definitive comparison. Clinical Trials Registration ClinicalTrials.gov Identifier: NCT00000611 HE HYPOTHESIS THAT A LOW- fat dietary pattern can reduce breast cancer risk has existed for decades. Supported by early rodent experiments, 1 country-toSee also pp 643, 655, and 691. www.jama.com JAMA. 2006;295:629-642 country comparisons linked higher dietary fat intake to breast cancer risk.2 However, case-control and cohort studies have had mixed results. A metaanalysis 3 of 12 international case- ©2006 American Medical Association. All rights reserved. Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013 Author Affiliations and WHI Investigators are listed at the end of this article. Corresponding Author: Ross L. Prentice, PhD, Division of Public Health Sciences, Fred Hutchinson Cancer Research Center, 1100 Fairview Ave N, M3A410, PO Box 19024, Seattle, WA 98109 (rprentice @whi.org). (Reprinted) JAMA, February 8, 2006—Vol 295, No. 6 629 LOW-FAT DIETARY PATTERN AND RISK OF INVASIVE BREAST CANCER control studies reported a significant positive association between fat intake and breast cancer with relative risks of 1.00, 1.20, 1.24, 1.24, and 1.46 across total fat intake quintiles defined by one of the Canadian case-control studies. In contrast, an analysis4 of 7 Western cohort studies found no such association with relative risks of 1.00, 1.01, 1.12, 1.07, and 1.05 across energyadjusted fat intake quintiles. A recent meta-analysis, including both casecontrol and cohort studies, comparing highest and lowest fat intake categories reported a relative risk for breast cancer of 1.13 (95% confidence interval [CI], 1.03-1.25).5 Such inconsistent results may reflect limitations of the dietary assessment methods used6-8; a recent study reported a significant positive association of fat intake and postmenopausal breast cancer incidence only when diet was measured with food diaries rather than a food frequency questionnaire (FFQ) used in most analytic epidemiological studies.9 Previous randomized controlled trials have demonstrated the feasibility of achieving a dietary fat reduction among healthy postmenopausal women in a multicenter trial setting.10,11 Prior smallscale intervention trials have demonstrated reductions in serum estradiol levels among women undertaking a dietary pattern that is low in fat,12-14 and observational studies have linked low dietary fat intake both with low blood estrogen levels and low breast cancer risk.15 Observational studies of consumption ofvegetablesandfruitinrelationtobreast cancer incidence have also yielded inconsistent results.16,17 Some summary analyses report an association with vegetable intake18-20 but not fruit intake,18,19 with more evidence for association from case-control studies19,20 than from cohort Figure 1. Participant Flow in the Dietary Modification Component of the Women’s Health Initiative 373 092 Women Initiated Screening by Providing the Eligibility Screening Form METHODS 316 953 Excluded 24 473 Refused Consent 107 210 Had <32% Energy From Fat 185 270 Consent Information Not Available 56 139 Provided Consent and Met the ≥32% Energy From Fat Eligibility Criterion 7304 Excluded∗ 1668 Nutritionist Judgment/Participant Reevaluation 2163 Administrative Ineligibility 278 Ate ≥10 Meals/wk Away From Home 229 Had History of Breast Cancer 453 Other Medical Condition 48 835 Randomized 19 541 Were Assigned to Receive Low-Fat Diet 29 294 Were Assigned to Receive Usual Diet Status on 3/31/2005 17 674 Alive and Outcomes Data Submitted in Last 18 mo 663 Withdrew 254 Lost to Follow-up 950 Deceased Status on 3/31/2005 26 677 Alive and Outcomes Data Submitted in Last 18 mo 890 Withdrew 273 Lost to Follow-up 1454 Deceased 19 541 Included in Primary Analyses 29 294 Included in Primary Analyses *Categories are presented for which exclusions are known. More than 1 reason could be given for exclusion. 630 JAMA, February 8, 2006—Vol 295, No. 6 (Reprinted) Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013 studies. Similarly, meta-analyses of casecontrol studies21,22 report a marginally lower breast cancer incidence at higher whole grain consumption levels but a recent large cohort study23 found no such association. Once again, inconsistency could be due to measurement error in dietary assessment8,24 or to other sources of bias, including recall bias in casecontrol studies. The Women’s Health Initiative (WHI) began in 1992 and included a full-scale randomized controlled trial with a dietary modification intervention consisting of consumption of a reduced amount of fat (20% of energy) and of an increased amount of vegetables and fruit (ⱖ5 servings/d) and grains (ⱖ6 servings/d) referred to herein as the low-fat dietary pattern. Breast cancer and colorectal cancer incidence are the primary outcomes of the trial and coronary heart disease is the secondary outcome.25,26 This dietary modification intervention trial is the first to directly assess the health benefits and risks of promoting a low-fat dietary pattern. This article reports the principal results on the incidence of breast cancer. Study Population The design of the WHI clinical trial, including the dietary modification component, has been previously described, as have detailed eligibility criteria and recruitment methods.25-27 All women were postmenopausal and aged 50 to 79 years at screening (FIGURE 1 and TABLE 1). Special efforts were made to recruit minority women so that dietary intervention effects could be compared among selfreported racial/ethnic groups (18.6% of trial participants). Interested and eligible women were informed that they could be randomized to the dietary modification trial and/or the postmenopausal hormone therapy trial, which involved either estrogen alone (women without a uterus) or estrogen plus progestin (women with a uterus). After 1 year’s participation in the clinical trial, women were invited to consider further randomization to calcium and vi- ©2006 American Medical Association. All rights reserved. LOW-FAT DIETARY PATTERN AND RISK OF INVASIVE BREAST CANCER tamin D supplementation or placebo. The postmenopausal hormone therapy trial components were stopped early and have been reported.28,29 Major exclusions for the dietary modification trial included prior breast cancer, prior colorectal cancer, other cancer except nonmelanoma skin cancer in the last 10 years, medical conditions with predicted survival of less than 3 years, adherence or retention concerns (eg, alcoholism, dementia), or a diet at baseline with fat intake of less than 32% of total energy as estimated by the FFQ created for the WHI.30 Eligible women were randomized to the dietary modification intervention group (40%) or to the comparison group (60%) using a permuted block algorithm with blocks of size 5, 10, or 15 and stratified by clinical center and age group (50-54 years, 55-59 years, 60-69 years, 70-79 years). The randomization rate of 40% for the intervention group and 60% for the comparison group was chosen to minimize study cost at a specified level of power. The protocol and consent forms were approved by the institutional review boards for each participating institution and all women provided written informed consent. Low-Fat Dietary Pattern Intervention and Maintenance Details of the dietary modification (lowfat dietary pattern) intervention have been published.31 Briefly, the intervention was designed to promote dietary change with the goals of reducing total fat intake to 20% of total energy and increasing consumption of vegetables and fruit to at least 5 servings daily and grains to at least 6 servings daily. The intervention did not include total energy reduction or weight-loss goals. Although not a separate focus of the intervention, it was presumed that by reducing total fat to 20% of total energy the amount of saturated fat also would be reduced to about 7% of energy. The intervention group received an intensive behavioral modification program that consisted of 18 group sessions in the first year and quarterly maintenance sessions thereafter. Each group had 8 to 15 women and was led by a specially trained and certified nutritionist.25,26,31 Each participant was given her own total fat gram goal based on her height. The intervention emphasized self-monitoring techniques and introduced other individually tailored and targeted strategies, such as motivational interviewing. Comparison group participants received a copy of Nutrition and Your Health: Dietary Guidelines for Americans32 and other health-related materials but were not asked to make dietary changes. Neither group was asked to make changes in their use of dietary supplements or in other health-related behaviors. Follow-up and Data Collection Dietary intake for all participants was monitored using the FFQ, which was Table 1. Baseline Demographics of Participants in Women’s Health Initiative Dietary Modification Trial* No. (%) of Participants Intervention (n = 19 541) Comparison (n = 29 294) Age, y 50-59 7206 (36.9) 10 797 (36.9) 60-69 70-79 9086 (46.5) 3249 (16.6) 13 626 (46.5) 4871 (16.6) 15 869 (81.2) 2137 (10.9) 755 (3.9) 23 890 (81.6) 3129 (10.7) 1099 (3.8) Race/ethnicity White Black Hispanic American Indian Asian/Pacific Islander Unknown 88 (0.5) 433 (2.2) 259 (1.3) 115 (0.4) 674 (2.3) 387 (1.3) P Value† ⬎.99 .76 Family history of breast cancer‡ Gail model 5-y risk ⬎1.7% 3396 (18.3) 6812 (34.9) 4929 (17.8) 10 153 (34.7) Body mass index§ ⬍25 25-29 30-34 5072 (26.1) 6940 (35.7) 4450 (22.9) 7585 (26.0) 10 446 (35.8) 6748 (23.1) 2992 (15.4) 4378 (15.0) 12 262 (62.8) 2711 (13.9) 4568 (23.4) 18 452 (63.0) 3933 (13.4) 6909 (23.6) 14 196 (72.7) 21 299 (72.7) 2768 (14.2) 2576 (13.2) 4114 (14.0) 3881 (13.2) .92 ⱖ35 Postmenopausal hormone use, y Estrogen alone None ⬍5 ⱖ5 Estrogen plus progestin None ⬍5 ⱖ5 Mammography screening within 2 y .13 .65 .69 .36 15 729 (83.1) 23 708 (83.6) .12 Treated disease/condition Diabetes㛳 Hypertension¶ 866 (4.4) 7617 (42.5) 1336 (4.6) 11 596 (43.2) .50 .15 White blood cell count, ⫻109/L ⬍5.1 5.1-6.3 ⱖ6.4 5920 (30.3) 6752 (34.6) 6855 (35.1) 8921 (30.5) 10 179 (34.8) 10 166 (34.7) .70 *Percentages may not sum to 100% because of rounding error. †Based on a ␹2 test of association. ‡First-degree or second-degree female relative. §Calculated as weight in kilograms divided by the square of height in meters. 㛳Self-report of taking pills or insulin via injection. ¶Systolic blood pressure higher than 140 mm Hg, diastolic blood pressure higher than 90 mm Hg, self-report of taking pills to lower blood pressure. ©2006 American Medical Association. All rights reserved. Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013 (Reprinted) JAMA, February 8, 2006—Vol 295, No. 6 631 LOW-FAT DIETARY PATTERN AND RISK OF INVASIVE BREAST CANCER Table 2. Nutrient Consumption Estimates and Body Weight at Baseline and Year 1 Nutrient Consumption Estimate, Mean (SD)* Baseline Factor Mean (SD) Difference in Change Between Groups Year 1 Intervention Comparison Intervention Comparison Year 1 Year 3 Year 6 37.8 (5.1) 75.7 (34.1) 12.7 (2.5) 7.8 (2.0) 14.4 (2.3) 37.8 (5.0) 75.7 (33.6) 12.7 (2.5) 7.8 (2.0) 14.4 (2.3) 24.3 (7.5) 40.8 (21.4) 8.1 (2.8) 5.2 (1.8) 8.9 (3.1) 35.1 (6.9) 63.0 (31.0) 11.8 (2.9) 7.2 (2.1) 13.3 (2.9) −10.7† (7.0) −22.4† (31.1) −3.7† (2.9) −2.0† (2.1) −4.4† (3.0) −9.5† (7.4) −20.1† (32.0) −3.3† (3.1) −1.7† (2.2) −3.9† (3.2) −8.1† (7.8) −18.4† (33.5) −2.9† (3.3) −1.4† (2.3) −3.3† (3.4) Fat Percentage of energy Total, g Saturated, % Polyunsaturated, % Monounsaturated, % Energy, kcal Consumption per day Vegetables and fruit, servings Grains, servings Fiber, g Folate, µg Alcohol, g Weight, kg 1790.2 (710.1) 1789.4 (703.0) 1500.5 (544.2) 1593.8 (644.0) −95.8† (616.2) −92.5† (632.1) −119.9† (662.9) 3.6 (1.8) 4.7 (2.5) 15.4 (6.4) 3.6 (1.8) 4.8 (2.5) 15.4 (6.4) 5.1 (2.3) 5.1 (2.7) 18.1 (7.5) 3.9 (2.0) 4.2 (2.3) 14.9 (6.5) 1.2† (1.9) 0.9† (2.5) 3.2† (6.1) 1.3† (2.0) 0.7† (2.6) 3.1† (6.4) 1.1† (2.1) 0.4† (2.6) 2.4† (6.6) 259.2 (136.6) 4.4 (8.4) 259.3 (138.1) 4.4 (8.6) 398.5 (215.0) 4.3 (8.9) 346.1 (195.1) 4.3 (9.2) 52.3† (192.3) 0 (6.7) 62.1† (208.2) 0.1 (7.1) 45.6† (201.1) −0.1 (7.4) 76.8 (16.6) 76.7 (16.5) 74.4 (16.7) 76.3 (16.7) −2.2† (8.4) −1.3† (9.1) −0.8† (9.4) *Based on responses to the Women’s Health Initiative food frequency questionnaire. †Difference significant at P⬍.001 from a 2-sample t test. designed specifically for the WHI trial (TABLE 2). This FFQ was administered at baseline and at 1 year following randomization and thereafter to about one third of participants each year in a rotating sample. Additionally, 4-day food records were provided by all women prior to randomization. Study participants were contacted every 6 months for outcome ascertainment. Height, weight, waist circumference, and blood pressure were measured using standardized procedures at annual clinic visits. A fasting serum sample was collected at baseline and at 1 year after randomization. A 4.6% subsample (n = 2245) with an overrepresentation of minority women provided an additional 4-day food record at 1 year after randomization and 24-hour dietary recalls at 3 and 6 years after randomization. This 4.6% subsample combined with additional women who were participating in both the dietary modification trial and the hormone therapy trial yields a 5.8% subsample (n=2816) of women who provided fasting serum samples at 1, 3, and 6 years after randomization. The serum samples were centrally stored and analyzed for dietrelated biomarkers.27 Changes in levels of ␣-carotene, ␤-carotene, total carotenoids, ␣-tocopherol, ␥-tocopherol, 632 ␤-cryptoxanthin, lycopene, lutein plus zeaxanthin, retinol, glucose, insulin, total cholesterol, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, and triglycerides from baseline to year 3 were compared for the intervention group with the comparison group (TABLE 3). These measures provide an objective assessment of some aspects of the dietary changes reported by the participating women. Further independent 1% subsamples of women provided a 24-hour dietary recall annually. To examine whether a low-fat dietary pattern could influence breast cancer risk through changes in circulating hormones, serum hormone concentrations at baseline and 1 year after randomization were compared between random samples from 150 women in the intervention group and 150 women in the comparison group who were not enrolled in the WHI hormone therapy trials and who were not taking postmenopausal hormones at baseline (TABLE 4). Analyte determinations were performed at Esoterix Laboratory Services (Calabasas Hills, Calif). Baseline and follow-up samples were included in the same batches along with split duplicates. Intrabatch coefficients of variation were 7.6% JAMA, February 8, 2006—Vol 295, No. 6 (Reprinted) Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013 for estradiol, 7.3% for estrone, 8.9% for testosterone, and 5.7% for sex hormone−binding globulin. Outcome Ascertainment Women were expected under the study protocol to undergo mammography screening at baseline and every 2 years thereafter. Clinical centers made arrangements with a mammography facility or instructed women to undergo mammography screening through their usual sources of care. Details of clinical outcome definitions, documentation, and classification have been published.33 In brief, women were queried twice each year to determine whether they had been hospitalized or diagnosed with any of the clinical outcomes on a prespecified list, including breast cancer. Self-report of breast cancer was verified by medical record and pathology report review by centrally trained WHI physician adjudicators at each participating clinical center. Central adjudication and coding of histology, extent of disease, and estrogen receptor and progesterone receptor status (positive or negative per local pathology report) were performed at the clinical coordinating center using the National Cancer Institute’s Surveillance, Epidemiology, and End Results coding system. ©2006 American Medical Association. All rights reserved. LOW-FAT DIETARY PATTERN AND RISK OF INVASIVE BREAST CANCER Statistical Design and Analysis Trial design assumptions included a linear dependence of breast cancer risk on the lifetime dietary percentage of energy from fat: a 50% lower breast cancer incidence among women with a diet consisting of 20% of energy from fat compared with women with a diet consisting of 40% of energy from fat. These assumptions also specify that the risk reduction for women undertaking a lowfat dietary pattern would be achieved linearly over a 10-year intervention period.25 Adherence assumptions, motivated by preceding feasibility studies,10,11 included a 13% lower energy from fat consumption in the intervention group compared with the comparison group at 1 year after randomization, decreasing to an 11% difference by 9 years after randomization. These assumptions led to a projected 14% lower breast cancer incidence in the intervention group compared with the comparison group and to a study power of 86% for a test at the .05 level of significance at a sample size of 48 000 over a planned 9-year follow-up period. Blood analyte concentrations were analyzed by examining mean changes from baseline for log-transformed concentrations. Logarithmic transformation was used to obtain distributions that are approximately normal and differences in changes between intervention and comparison groups were assessed using t tests. Back-transformed (geometric) means and associated 95% CIs are presented herein. Event rate comparisons between the intervention group and the comparison group are based on the intent-totreat principle using time-to-event methods.34 A (2-sided) weighted log-rank test for cancer incidence and mortality was specified in the protocol with weights increasing linearly from zero at randomization to a maximum value of 1 at 10 years after randomization, and constant thereafter, to enhance trial power under design assumptions. Both weighted and unweighted log-rank tests are presented herein to assess the null hypothesis for breast cancer and for other major trial outcomes and to as- sess the simultaneous null hypothesis for breast cancer and for colorectal cancer. The time to an event for a particular outcome was defined as the number of days after randomization to the first diagnosis of the designated event (eg, invasive breast cancer). Follow-up time was censored at the time of a woman’s Table 3. Blood Biomarkers for Baseline and Year 3* Geometric Mean (95% CI) Biomarker Intervention Comparison Relative Change (95% CI)† Total carotenoids, µg/mL Baseline Year 3 0.78 (0.76-0.81) 0.75 (0.72-0.78) 0.77 (0.75-0.79) 0.72 (0.69-0.74) 1.05 (1.00-1.10) ␣-Carotene, µg/10 mL Baseline Year 3 0.59 (0.56-0.63) 0.53 (0.50-0.57) 0.59 (0.56-0.62) 0.49 (0.47-0.51) 1.10 (1.03-1.18) ␤-Carotene, µg/mL Baseline Year 3 0.22 (0.21-0.23) 0.21 (0.20-0.23) 0.22 (0.21-0.23) 0.19 (0.18-0.20) 1.09 (1.01-1.17) 14.77 (14.39-15.16) 16.77 (16.28-17.29) 15.19 (14.85-15.53) 16.81 (16.41-17.22) 1.01 (0.97-1.04) 1.76 (1.66-1.86) 1.11 (1.04-1.19) 1.71 (1.64-1.79) 1.29 (1.23-1.36) 0.85 (0.79-0.91) 0.69 (0.66-0.73) 0.80 (0.75-0.84) 0.67 (0.64-0.69) 0.73 (0.69-0.76) 1.07 (1.01-1.14) ␣-Tocopherol, µg/mL Baseline Year 3 ␥-Tocopherol, µg/mL Baseline Year 3 ␤-Cryptoxanthin, µg/10 mL Baseline Year 3 Lycopene, µg/mL Baseline Year 3 Lutein plus zeaxanthin, µg/mL Baseline Year 3 0.37 (0.36-0.38) 0.33 (0.31-0.34) 0.36 (0.35-0.37) 0.33 (0.32-0.34) 1.00 (0.94-1.05) 0.19 (0.19-0.20) 0.19 (0.18-0.19) 0.19 (0.18-0.19) 0.17 (0.17-0.18) 1.03 (0.99-1.06) Retinol, µg/mL Baseline Year 3 0.59 (0.58-0.60) 0.59 (0.58-0.60) 0.60 (0.59-0.61) 0.59 (0.58-0.60) 1.02 (1.00-1.04) 97.90 (96.42-99.41) 96.47 (95.02-97.94) 97.70 (96.63-98.77) 97.06 (95.88-98.26) 0.99 (0.97-1.00) 9.95 (9.60-10.31) 10.53 (10.12-10.97) 10.22 (9.92-10.54) 11.24 (10.88-11.61) 0.98 (0.93-1.02) Glucose, mg/dL Baseline Year 3 Insulin, µIU/mL Baseline Year 3 Cholesterol, mg/dL Total Baseline Year 3 Low-density lipoprotein Baseline Year 3 High-density lipoprotein Baseline Year 3 Triglycerides, mg/dL Baseline Year 3 220.90 (218.38-223.48) 220.90 (218.76-223.11) 211.20 (208.51-213.87) 213.60 (211.44-215.78) 0.98 (0.97-1.00) 128.40 (125.91-131.04) 129.40 (127.32-131.47) 118.70 (116.18-121.33) 122.20 (120.13-124.39) 0.97 (0.95-1.00) 58.05 (56.95-59.17) 57.65 (56.47-58.86) 56.44 (55.59-57.30) 56.20 (55.29-57.13) 138.60 (133.98-143.29) 141.10 (137.35-144.95) 142.30 (137.17-147.53) 144.60 (140.76-148.50) 0.99 (0.98-1.01) 1.00 (0.97-1.04) Abbreviation: CI, confidence interval. SI conversion factors: To convert glucose to mmol/L, multiply by 0.0555; low-density, high-density, and total cholesterol to mmol/L, multiply by 0.0259; triglycerides to mmol/L, multiply by 0.0113. *Values based on a 5.8% subsample (n = 2816). †Calculated as the ratio of year 3 to baseline geometric means and is the ratio of changes in the intervention group to the comparison group. ©2006 American Medical Association. All rights reserved. Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013 (Reprinted) JAMA, February 8, 2006—Vol 295, No. 6 633 LOW-FAT DIETARY PATTERN AND RISK OF INVASIVE BREAST CANCER last documented follow-up contact or death. Quantitative comparisons of event rates between the intervention group and the comparison group are presented as hazard ratios (HRs) and nominal 95% CIs from Cox regression34 and are stratified by age and randomization status in the hormone therapy trials. Annualized event rates also were calculated for absolute disease rate comparisons. Cumulative hazard rates were estimated by the KaplanMeier method (FIGURE 2). The HR estimates among women who were adherent to dietary goals are of particular interest. Because of the lim- Table 4. Blood Hormone Concentrations for Baseline and Year 1* Geometric Mean (95% CI) Biomarker Estradiol, pg/mL Baseline Year 1 Estrone, pg/mL Baseline Year 1 Testosterone, pg/mL Baseline Year 1 Sex hormone−binding globulin, nmol/L Baseline Year 1 Intervention Relative Change (95% CI)† Comparison 7.6 (6.6-8.8) 6.7 (5.9-7.7) 6.4 (5.6-7.4) 6.6 (5.9-7.4) 0.85 (0.72-1.00) 24.5 (21.9-27.5) 23.7 (21.3-26.5) 23.5 (21.2-26.1) 23.8 (21.5-26.3) 0.98 (0.87-1.11) 201.8 (185.0-220.2) 199.4 (182.5-217.8) 192.0 (176.3-209.1) 192.4 (177.6-208.4) 0.99 (0.94-1.05) 67.0 (60.8-73.7) 72.3 (66.4-78.8) 66.2 (61.3-71.5) 65.3 (60.3-70.8) 1.09 (1.03-1.16) Abbreviation: CI, confidence interval. *Values based on a subsample of 150 women in the intervention group and 150 women in the comparison group. Women taking postmenopausal hormone therapy or randomized in the hormone therapy trials were excluded from the sample for hormone analysis. †Change is calculated as the ratio of year 1 to baseline geometric means and is the ratio of changes in the intervention group to the comparison group. Figure 2. Kaplan-Meier Estimates of the Cumulative Hazard for Invasive Breast Cancer 0.05 Comparison Intervention 0.04 Cumulative Hazard HR, 0.91 (95% CI, 0.83-1.01) 0.03 0.02 0.01 0 0 1 2 3 4 5 6 7 8 9 94 137 89 145 46 97 33 58 Time, y Events Intervention Comparison No. at Risk Intervention Comparison 47 74 79 140 92 123 80 137 19 541 19 328 19 084 18 798 18 520 18 263 17 900 15 507 10 245 5075 29 294 28 908 28 536 28 195 27 806 27 372 26 977 23 337 15 373 7580 CI indicates confidence interval; HR, hazard ratio. 634 72 136 JAMA, February 8, 2006—Vol 295, No. 6 (Reprinted) Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013 ited reliability of individual dietary assessment, we chose to define adherence in terms of participation in trial activities. A comparison group participant was considered nonadherent and her follow-up time was censored the first time she missed an annual visit. An intervention group participant was considered nonadherent at the earliest missed annual visit, when she failed to participate in 9 or more of the firstyear intervention sessions, or when she failed to participate in 2 or more of the 4 maintenance sessions in subsequent years. To produce a fair comparison between randomization groups, each participant who continued to be adherent was included in the HR estimation procedure that used the inverse of the participant’s estimated adherence probability as a weighting factor. This method35 yields valid HR estimates among participants meeting adherence criteria provided that censoring probabilities can be accurately estimated (FIGURE 3 and TABLE 5). To control for factors that may relate to adherence, time to nonadherence (censoring) was modeled separately for the intervention group and the comparison group using Cox models,34 which included age, ethnicity, education, income, body mass index, alcohol consumption, multivitamin use, randomization into the hormone therapy trial, Gail risk score, percentage of energy from fat, vegetable, fruit, and grain consumption, physical activity, and several psychosocial variables (social support, optimism, life events, hostility, and negative emotions). The HRs were estimated in subsets of the study population defined by baseline dietary factors and by including product terms between randomization assignment and indicator variables for the subsets of interest in the Cox models. Interactions between the HRs and the baseline dietary factors were examined by testing equality of the product term coefficients (TABLE 6). These analyses also are related to adherence because women in the intervention group who had a comparatively high-fat diet ©2006 American Medical Association. All rights reserved. LOW-FAT DIETARY PATTERN AND RISK OF INVASIVE BREAST CANCER estimates also were compared across tumor characteristics using competing risk partial likelihood methods and Cox models.37 Characteristics considered included hormone receptor status, grade, and measures to determine the extent of disease (TABLE 7). SAS version 9.1.3 (SAS Institute Inc, Cary, NC) was used for these analyses. Data and Safety Monitoring Statistical monitoring boundaries were based on the O’Brien-Fleming group sequential procedures38 with additional Bonferroni correction for the 2 primary outcomes. Monitoring guidelines39 adopted by the external data and safety monitoring board involved breast cancer, colorectal cancer, coronary heart disease, and deaths from other causes, as well as a global index defined as the time to the earliest of any of these 4 outcomes. This study proceeded to its planned termination. grams.27 Baseline participant characteristics have been described.26 Briefly, participants were on average 62.3 years old, 18.6% were of minority race/ ethnicity, and the average body mass index was 29.1. Risk factors for breast cancer were closely comparable in the 2 study groups including age, prior hormone therapy use, family history, education, ethnicity, and Gail 5-year risk estimate (Table 1). Tamoxifen and raloxifene use was nonexistent at baseline and remained low and balanced throughout follow-up (eg, tamoxifen use was approximately 1.5% and raloxifene use was approximately 2.9% in either group at year 6). Participants were at moderate risk for breast cancer based on a mean (SD) Gail 5-year risk estimate of 1.7% (0.9%). Figure 3. Hazard Ratio Estimates for Invasive Breast Cancer Based on Cumulative Data Through Each Follow-up Year RESULTS 2.0 Recruitment and Baseline Characteristics Between 1993 and 1998, a total of 48 835 women (102% of goal) were randomized into the dietary modification trial: 19 541 women to the intervention group and 29 294 women to the comparison group (Figure 1). Most women were recruited to the study by population-based direct mailing campaigns and by media awareness pro- Hazard Ratio or low consumption of vegetables, fruit, or grains at baseline needed to make larger dietary changes to achieve specified dietary goals. We used 4-day food records rather than FFQs to characterize participants’ baseline diets in terms of total fat, total energy, and percentage of energy from fat because the 32% of energy from fat trial eligibility criterion in conjunction with the random measurement error yields distorted baseline FFQ estimates of these quantities. This is the same phenomenon as the regression to the mean problem. In this study, baseline FFQs overestimate the percentage of energy from fat by about 3%. To avoid a costly analysis of 4-day food records for all trial participants, the HR estimates were based on the 4-day food records of women who developed breast cancer. The resulting case-only HR estimates36 are nearly identical to those that would arise from Cox regression on the entire cohort in the circumstances (rare disease and high follow-up rates) of this trial. Technically, the logarithm of HR estimates in these analyses are obtained by logistic regression of randomization assignment on indicator variables for the baseline 4-day food record dietary categories, with a constant term of log (2/3) that acknowledges the 2 to 3 randomization ratio between the intervention group and the comparison group. The possibility of differential intervention effects across other subsets of the study population also was explored by including product terms between the randomization assignment and indicator variables for the subsets in the Cox models and by testing the equality of the product term coefficients. Because 17 interactions with baseline characteristics are reported, about 1 significant test at the ␣ level of .05 can be expected based on chance alone. Baseline factors were restricted to established breast cancer risk factors, postmenopausal hormone therapy use or randomization assignment in the hormone therapy trial, and a small number of other factors that plausibly relate to intervention efficacy. The HR 1.0 0.05 1 2 3 4 5 6 7 8 9 Overall Follow-up, y Error bars indicate 95% confidence intervals. Table 5. Risk of Invasive Breast Cancer and Other Major Clinical Outcomes No. of Cases (Annualized %) P Value Intervention Comparison HR (95% CI)* Unweighted* Weighted† Breast cancer Incidence Mortality 655 (0.42) 27 (0.02) 1072 (0.45) 53 (0.02) 0.91 (0.83-1.01) 0.77 (0.48-1.22) .07 .26 .09 .27 Total cancer Incidence Mortality 1946 (1.23) 436 (0.28) 3040 (1.28) 690 (0.29) 0.96 (0.91-1.02) 0.95 (0.84-1.07) .15 .41 .10 .22 Total mortality Global index‡ 950 (0.60) 2051 (1.30) 1454 (0.61) 3207 (1.35) 0.98 (0.91-1.07) 0.96 (0.91-1.02) .70 .16 .29 .16 Abbreviations: CI, confidence interval; HR, hazard ratio. *Proportional hazards model stratified by prevalent condition (when appropriate), age, and randomization group. †Weighted log-rank test stratified by prevalent condition (when appropriate), age, and randomization group. Weights increase linearly from zero at randomization to a maximum of 1 at 10 years. ‡Defined for a participant as the time to the earliest invasive breast cancer, colorectal cancer, coronary heart disease, or mortality from any other cause. ©2006 American Medical Association. All rights reserved. Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013 (Reprinted) JAMA, February 8, 2006—Vol 295, No. 6 635 LOW-FAT DIETARY PATTERN AND RISK OF INVASIVE BREAST CANCER Dietary Intervention Effects on Nutrients and Other Factors Table 2 provides information at baseline and at 1 year after randomization for the nutrients targeted in the intervention group as well as for other dietary variables, body weight, and factors that may be affected by participation in intervention group. These baseline variables are nearly equal between the intervention group and the comparison group. Also, differences from baseline in these factors at 1, 3, and 6 years after randomization are compared between the intervention group and the comparison group. Based on data from women who provided FFQs, the average reductions in percentage of energy from fat for the intervention group compared with the comparison group was 10.7 at year 1 and decreased to 8.1 at year 6. Compared with the comparison group, the consumption of vegetable and fruit servings in the intervention group was more than 1 serving per day greater, while the difference for the consumption of grain servings was significant but appeared to decline as the study progressed. A small reduction in energy consumption was reported in the intervention group compared with the comparison group. Women in the intervention group experienced a modest weight loss early in the trial and maintained a greater weight change from baseline throughout follow-up than women in the comparison group. Dietary differences were similar to those reported in Table 2 when assess- ment was based on a 4-day food record or 24-hour dietary recall. For example, based on 4-day food record assessments, the percentage of energy from fat was 11.3% lower and intake of fat was 26.3 g lower in the intervention group compared with the comparison group at year 1. At year 3, using 24-hour recall assessments, the percentage of energy from fat was 8.2% (19.4 g) lower in the intervention group compared with the comparison group and at year 6 was 7.5% (24 g) lower, respectively, in the intervention group. The changes from baseline to year 3 for blood markers between the intervention group and the comparison group appear in Table 3 and are based on values from the 5.8% subsample. Most blood concentration changes were Table 6. Breast Cancer Risk by Baseline Dietary Factors Year 1, Mean (SD)* Baseline Quartiles Intervention Comparison No. (%) of Breast Cancer Cases Mean (SD) Difference Between Groups Intervention (n = 655) Comparison (n = 1072) HR (95% CI)† Interaction P Value‡ Fat Percentage of energy, kcal ⬍27.9 27.9-⬍32.3 32.3-⬍36.8 ⱖ36.8 Total intake, g ⬍46.2 46.2-⬍59.8 59.8-⬍76.0 ⱖ76.0 Energy intake, kcal ⬍1391.8 1391.8-⬍1663.6 1663.6-⬍1958.7 ⱖ1958.7 Vegetables and fruit, servings/d ⬍2.3 2.3-⬍3.3 3.3-⬍4.6 ⱖ4.6 Grains, servings/d ⬍3 3-⬍4.3 4.3-⬍5.9 ⱖ5.9 18.8 (6.2) 21.0 (7.0) 21.7 (6.7) 28.6 (6.2) 31.4 (6.0) 33.5 (6.5) 9.7 (6.2) 10.4 (6.5) 11.7 (6.6) 144 (22) 186 (28) 160 (24) 222 (21) 259 (24) 283 (26) 0.97 (0.79-1.20) 1.08 (0.89-1.30) 0.85 (0.70-1.03) 23.6 (7.9) 35.8 (6.3) 12.2 (7.0) 151 (23) 291 (27) 0.78 (0.64-0.95) .04 29.5 (12.8) 45.5 (16.5) 16.1 (15.1) 128 (20) 221 (21) 0.87 (0.70-1.08) 32.9 (12.5) 35.0 (14.9) 38.9 (17.0) 57.8 (19.9) 61.9 (20.4) 73.9 (26.6) 24.9 (17.1) 27.0 (18.5) 35.0 (23.3) 176 (27) 194 (30) 143 (22) 261 (24) 300 (28) 273 (25) 1.01 (0.84-1.22) 0.97 (0.81-1.16) 0.79 (0.64-0.96) 1225 (301.5) 1376 (314.1) 1470 (298.5) 1608 (397.3) 1314 (339.1) 1541 (353.4) 1690 (386.4) 1927 (468.3) 89.1 (324.2) 164.7 (338.2) 219.3 (353.0) 318.5 (440.7) 139 (21) 164 (25) 179 (27) 159 (24) 226 (21) 290 (27) 271 (25) 268 (25) 0.92 (0.75-1.14) 0.85 (0.70-1.03) 0.99 (0.82-1.20) 0.89 (0.73-1.08) .89 3.7 (2.1) 4.6 (2.0) 5.3 (2.1) 6.5 (2.2) 2.4 (1.4) 3.3 (1.4) 4.1 (1.6) 5.5 (2.0) −1.3 (1.7) −1.3 (1.7) −1.2 (1.8) −1.0 (2.1) 155 (24) 158 (24) 144 (22) 197 (30) 259 (24) 268 (25) 264 (25) 276 (26) 0.90 (0.73-1.09) 0.88 (0.72-1.07) 0.82 (0.67-1.00) 1.08 (0.90-1.29) .07 3.7 (2.1) 4.6 (2.1) 5.4 (2.4) 6.7 (3.1) 2.8 (1.5) 3.7 (1.7) 4.5 (2.0) 5.9 (2.7) −0.9 (1.7) −0.8 (1.9) −0.9 (2.1) −0.8 (2.9) 160 (24) 171 (26) 178 (27) 145 (22) 258 (24) 242 (23) 311 (29) 256 (24) 0.94 (0.77-1.15) 1.02 (0.84-1.25) 0.85 (0.70-1.02) 0.88 (0.71-1.07) .98 .42 Abbreviations: CI, confidence interval; HR, hazard ratio. *Baseline classification and year 1 data for percentage of energy from fat, total fat intake, and total energy based on 4-day food records from the 4.6% subsample. Consumption of vegetables and fruit and grains based on food frequency questionnaires from the entire trial cohort. †Based on case-only analysis for percentage of energy from fat, total fat intake, and total energy and standard Cox regression for vegetables and fruit and grains. An unweighted proportional hazards model stratified by age and randomization group was used. ‡Test of interaction between the randomization assignment and the variable of interest. 636 JAMA, February 8, 2006—Vol 295, No. 6 (Reprinted) Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013 ©2006 American Medical Association. All rights reserved. LOW-FAT DIETARY PATTERN AND RISK OF INVASIVE BREAST CANCER minor. There was a greater reduction in levels of ␥-tocopherol in the intervention group compared with the comparison group and small positive differences in levels of ␣-carotene, ␤-carotene, and ␤-cryptoxanthin. Lowdensity lipoprotein cholesterol level was modestly decreased in the intervention group compared with the comparison group, but changes in levels of high-density lipoprotein cholesterol, triglycerides, insulin, and glucose did not differ significantly between the 2 groups. Corresponding biomarker changes were similar at the other time points considered (year 1 and year 6) and differences between the interven- tion group and the comparison group were somewhat larger at year 1 than at the later time points. Changes from baseline to year 1 in blood hormone metabolites based on the subsample of 150 women in the intervention group and 150 women in the comparison group appear in Table 4. A greater reduction in estradiol and a greater increase in sex hormone−binding globulin occurred for women in the intervention group compared with women in the comparison group. Clinical Outcomes The average follow-up time was 8.1 years in both the intervention group and the comparison group. Over the course of the trial, 4.7% of the women in the intervention group withdrew from participation or were lost to follow-up compared with 4.0% women in the comparison group (Figure 1). Frequencies of mammography screening were 87% at baseline, 92% at year 2, 91% at year 4, 89% at year 6, and 88% at year 8 in the intervention group. There were nearly identical frequencies of mammography screening in the comparison group: 87% at baseline, 92% at year 2, 92% at year 4, 90% at year 6, and 88% at year 8. Overall, 655 (3.35%) women in the intervention group and 1072 (3.66%) women in the comparison Table 7. Risk of Breast Cancer by Tumor Characteristics No. of Cases (Annualized %)* Tumor Characteristic P Value Comparison HR (95% CI) 486 (0.31) 94 (0.06) 817 (0.34) 159 (0.07) 0.89 (0.80-1.00) 0.89 (0.69-1.14) .04 .36 407 (0.26) 162 (0.10) 634 (0.27) 319 (0.13) 0.96 (0.85-1.09) 0.76 (0.63-0.92) .54 .004 .04 399 (0.25) 77 (0.05) 8 (0.01) 82 (0.05) 616 (0.26) 179 (0.08) 18 (0.01) 138 (0.06) 0.97 (0.86-1.10) 0.64 (0.49-0.84) 0.67 (0.29-1.54) 0.89 (0.68-1.17) .64 .001 .34 .41 .04 164 (0.10) 235 (0.15) 176 (0.11) 283 (0.12) 404 (0.17) 271 (0.11) 0.87 (0.72-1.05) 0.87 (0.74-1.02) 0.97 (0.80-1.18) .15 .09 .77 .63 SEER stage In situ㛳 Localized Regional Distant 178 (0.11) 475 (0.30) 148 (0.09) 8 (0.01) 263 (0.11) 789 (0.33) 243 (0.10) 12 (0.01) 1.01 (0.83-1.22) 0.90 (0.80-1.01) 0.91 (0.74-1.12) 1.00 (0.41-2.44) .93 .07 .39 .99 .79 No. of positive lymph nodes None 1-3 ⬎3 437 (0.28) 104 (0.07) 38 (0.02) 723 (0.30) 164 (0.07) 67 (0.03) 0.90 (0.80-1.02) 0.95 (0.74-1.22) 0.85 (0.57-1.27) .10 .69 .44 .91 Tumor size, cm ⬍0.5 0.5-1 ⬎1-2 ⬎2-5 70 (0.04) 156 (0.10) 249 (0.16) 97 (0.06) 132 (0.06) 279 (0.12) 373 (0.16) 190 (0.08) 0.80 (0.60-1.07) 0.84 (0.69-1.02) 1.00 (0.85-1.17) 0.76 (0.60-0.98) .13 .07 .99 .03 18 (0.01) 20 (0.01) 1.35 (0.71-2.56) .35 Estrogen receptor status Positive Negative Progesterone receptor status Positive Negative Ratio of estrogen to progesterone receptors§ Estrogen⫹/progesterone⫹ Estrogen⫹/progesterone− Estrogen−/progesterone⫹ Estrogen−/progesterone− Differential grade Good Moderate Poor ⬎5 Unweighted† Competing Risks Analysis‡ Intervention ⬎.99 .20 Abbreviations: CI, confidence interval; HR, hazard ratio; SEER, Surveillance, Epidemiology, and End Results program of the National Cancer Institute. *Numbers for some characteristics are less than the total number of invasive breast cancers because of missing tumor characteristic data. †From an unweighted proportional hazards model stratified by age and randomization group; tests whether HRs equal unity. ‡Analysis of the partial likelihoods; tests whether HRs are equal between tumor types. §The numbers of cases do not total due to missing data for receptor status. 㛳In situ breast tumors included in this portion of table only. ©2006 American Medical Association. All rights reserved. Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013 (Reprinted) JAMA, February 8, 2006—Vol 295, No. 6 637 LOW-FAT DIETARY PATTERN AND RISK OF INVASIVE BREAST CANCER group developed invasive breast cancer during follow-up. The comparison group incidence rate is slightly in excess of design assumptions. The estimated HR for invasive breast cancer is 0.91 (95% CI, 0.83-1.01). The corresponding log-rank significance level is .07. The protocol-specified weighted log-rank test significance level was .07. The cumulative hazard curves separate in favor of the intervention group after about 4 years (Figure 2). As an alternate view of these same data, the HRs for the cumulative data through each of years 1 to 9 after randomization appear in Figure 3. A simultaneous test of the null hypothesis for the primary outcomes of breast cancer and colorectal cancer had a significance level of .14 using either weighted or unweighted log-rank tests. More than half (51.6%) of the participants also enrolled in the calcium and vitamin D trial, mainly at 1 year after enrollment in the dietary modification trial. The enrollment rate was slightly lower among women in the intervention group (49.4%) than among women in the comparison group (53.1%). However, the HR estimate remained at 0.91 (95% CI, 0.83-1.01) following control for calcium and vitamin D trial enrollment and randomization assignment as timedependent covariates. The incidence rates, HRs, and weighted and unweighted log-rank tests for breast cancer mortality, total cancer (exclusive of nonmelanoma skin cancer) incidence and mortality, total mortality, and the global index appear in Table 5. For each outcome, the event rates are slightly lower in the intervention group compared with the comparison group but the differences are not significant at the .05 level. HRs for Adherent Women The HR for invasive breast cancer for the intervention group compared with the comparison group was estimated among women who participated actively in the dietary modification trial. Under these criteria, the comparison group adherence rates (estimated from a Cox regres638 sion model) were 87% at year 3, 75% at year 6, and 65% at year 9, while the corresponding intervention group adherence rates were 57%, 31%, and 19%. The inverse adherence probabilityweighted HR estimate under these adherence criteria is 0.85 (95% CI, 0.711.02). The weighting procedure aims to produce valid HR estimates even when adherence rates differ between the 2 groups. The difference in the percentage of energy from fat on the FFQ between adherent women in the intervention group and adherent women in the comparison group was 12.1% at year 1, 11.8% at year 3, 11.1% at year 6, and 10.1% at year 9. The use of more stringent adherence criteria for the intervention group (eg, ⱖ10 first-year intervention sessions, ⱖ3 maintenance sessions annually) leads to even smaller HR estimates and to 95% CIs that exclude 1. However, these estimates may be sensitive to adherence model inadequacies. HRs in Relation to Baseline Dietary Factors Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013 Breast Tumor Characteristics The grade, size, lymph node status, and stage of breast cancers occurring in the intervention group were similar to those seen in the comparison group (Table 7). The HR estimate was lower for tumors negative for the progesterone receptor than for tumors positive for the progesterone receptor (P=.04) but did not depend on estrogen receptor status. When tumors were classified by both estrogen and progesterone receptor status, there was an indication (P=.04) of HR variation with stronger evidence for a reduction in the occurrence of tumors that are positive for the estrogen receptor and negative for the progesterone receptor. Subgroup Analyses The numbers of invasive breast cancers and the HRs across quartiles of baseline dietary factors appear in Table 6. The HR estimates for fat, energy, and percentage of energy from fat are based on case-only analyses of 4-day food records at baseline and quartiles are defined by food records from the 4.6% subsample of the trial cohort, whereas the HR estimates for consumption of vegetables and fruit and grains are based on FFQs from the entire cohort. A significant (P=.04) trend in HR with baseline percentage of energy from fat is observed. Women with higher baseline percentages of energy from fat show greater evidence for a reduction in breast cancer risk. There is also a suggestive trend (P=.07) with baseline consumption of vegetables and fruit. The means and SDs for baseline dietary factors at year 1 appear in Table 6. The limited variation in the comparison group means at year 1 across these baseline categories, in conjunction with intervention group vs comparison group differences, suggests that the HR varia- JAMA, February 8, 2006—Vol 295, No. 6 (Reprinted) tion for the percentage of energy from fat may primarily reflect study adherence differences. For example, the year 1 trend from a 12.2% difference in the highest percentage of energy from fat quartile to a 9.7% difference in the lowest quartile is significant (P=.001). Invasive breast cancer HR estimates for the subgroups defined by baseline demographic, medical history, and health behavioral factors appear in TABLE 8. Two of 17 interactions were significant at the .05 level (hypertension and white blood cell count) and another was significant at the .10 level (estrogen plus progestin use). COMMENT The WHI Dietary Modification Trial is the first large-scale randomized trial to test whether adopting a low-fat dietary pattern in the middle to later decades of life reduces the risk for breast cancer. The relatively intensive dietary intervention implemented in the WHI resulted in a significant and sustained reduction in fat intake and an increase in vegetable and fruit intake. After approximately 8 years of follow-up, breast cancer incidence was 9% lower for women in the dietary intervention group compared with women in the comparison group (HR, 0.91; 95% CI, 0.83-1.01). Because incidence rates did not differ ©2006 American Medical Association. All rights reserved. LOW-FAT DIETARY PATTERN AND RISK OF INVASIVE BREAST CANCER Table 8. Breast Cancer Risk Based on Baseline Demographics, Medical History, and Health Behavior Variables No. of Participants (Annualized %) Age, y 50-59 60-69 70-79 Race/ethnicity White Black Hispanic American Indian Asian/Pacific Islander Unknown Family history of breast cancer Yes No Gail model 5-y risk, % ⬍1.25 1.25-1.74 ⱖ1.75 Mammography screening within 2 y Yes No Hypertension† Yes No Diabetes‡ Yes No White blood cell count, ⫻109/L ⬍5.1 5.1-6.3 ⱖ6.4 Postmenopausal hormone use, y Estrogen alone None ⬍5 ⱖ5 Estrogen plus progestin, y None ⬍5 ⱖ5 Randomized to estrogen alone Active Placebo Randomized to estrogen plus progestin Active Placebo Baseline postmenopausal hormone use Estrogen alone or randomized to estrogen alone Yes No Estrogen plus progestin or randomized to estrogen plus progestin Yes No Body mass index§ ⬍24.9 24.9-⬍28.2 28.2-⬍32.5 ⱖ32.5 Waist ⬎88 cm Yes No Physical activity, METs ⬍1.5 1.5-⬍6.3 6.3-⬍14.8 ⱖ14.8 Intervention (n = 655) Comparison (n = 1072) 227 (0.37) 304 (0.42) 124 (0.49) 359 (0.39) 519 (0.48) 194 (0.51) 0.95 (0.81-1.13) 0.87 (0.76-1.01) 0.96 (0.76-1.20) 560 (0.43) 55 (0.32) 17 (0.29) 2 (0.28) 12 (0.36) 9 (0.46) 936 (0.48) 73 (0.29) 26 (0.31) 2 (0.22) 24 (0.45) 11 (0.36) 0.90 (0.81-1.00) 1.09 (0.77-1.55) 0.94 (0.51-1.74) 1.24 (0.17-8.82) 0.79 (0.39-1.57) 1.26 (0.52-3.04) 140 (0.51) 488 (0.40) 247 (0.62) 771 (0.42) 0.81 (0.66-1.00) 0.95 (0.85-1.07) .19 175 (0.31) 196 (0.39) 284 (0.55) 276 (0.33) 330 (0.43) 466 (0.60) 0.95 (0.78-1.15) 0.90 (0.75-1.07) 0.90 (0.78-1.05) .47 533 (0.42) 104 (0.41) 905 (0.47) 140 (0.38) 0.89 (0.80-0.99) 1.08 (0.84-1.39) .16 262 (0.44) 349 (0.43) 464 (0.51) 498 (0.41) 0.85 (0.73-0.99) 1.04 (0.91-1.19) .05 23 (0.35) 632 (0.42) 46 (0.45) 1026 (0.45) 0.75 (0.46-1.24) 0.92 (0.83-1.02) .43 195 (0.40) 242 (0.44) 218 (0.40) 285 (0.39) 368 (0.44) 419 (0.52) 1.03 (0.85-1.23) 0.99 (0.84-1.16) 0.77 (0.66-0.91) .04 431 (0.44) 82 (0.37) 142 (0.39) 676 (0.45) 153 (0.47) 243 (0.44) 0.96 (0.85-1.08) 0.77 (0.59-1.01) 0.89 (0.72-1.09) .33 447 (0.39) 87 (0.38) 121 (0.59) 683 (0.40) 175 (0.51) 214 (0.68) 0.98 (0.87-1.11) 0.73 (0.57-0.95) 0.85 (0.68-1.06) .10 10 (0.20) 18 (0.34) 23 (0.28) 25 (0.29) 0.73 (0.35-1.53) 1.15 (0.63-2.10) .35 36 (0.46) 29 (0.39) 64 (0.54) 38 (0.36) 0.85 (0.56-1.28) 1.07 (0.66-1.73) .48 162 (0.37) 88 (0.34) 275 (0.42) 146 (0.37) 0.89 (0.73-1.08) 0.92 (0.71-1.20) .84 188 (0.53) 217 (0.41) 345 (0.64) 306 (0.39) 0.83 (0.69-0.99) 1.04 (0.88-1.24) .06 151 (0.38) 156 (0.40) 175 (0.44) 171 (0.44) 251 (0.42) 260 (0.43) 278 (0.48) 277 (0.48) 0.89 (0.73-1.09) 0.92 (0.76-1.13) 0.92 (0.76-1.12) 0.92 (0.76-1.12) 312 (0.42) 341 (0.41) 547 (0.49) 524 (0.42) 0.86 (0.75-0.99) 0.97 (0.84-1.11) .24 146 (0.43) 134 (0.40) 145 (0.41) 158 (0.45) 235 (0.46) 242 (0.49) 264 (0.49) 201 (0.39) 0.94 (0.76-1.15) 0.80 (0.65-0.99) 0.85 (0.69-1.04) 1.17 (0.95-1.44) .12 HR (95% CI) P Value for Interaction* .79 .87 .88 Abbreviations: CI, confidence interval; HR, hazard ratio; METs, metabolic equivalent units. *Unweighted proportional hazards model stratified by age and randomization group. †Systolic blood pressure higher than 140 mm Hg, diastolic blood pressure higher than 90 mm Hg, or self-report of taking pills to lower blood pressure. ‡Self-report of taking pills or insulin via injection. §Calculated as weight in kilograms divided by the square of height in meters. ©2006 American Medical Association. All rights reserved. Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013 (Reprinted) JAMA, February 8, 2006—Vol 295, No. 6 639 LOW-FAT DIETARY PATTERN AND RISK OF INVASIVE BREAST CANCER between the intervention group and the comparison group at the conventional .05 level of significance, chance provides an explanation for the modestly lower breast cancer incidence rates in the intervention group. However, interpretation of these results needs to take into account the following. There were departures from the design assumptions that likely reduced study power. In addition, there was a significant interaction between the HR for the intervention group compared with the comparison group for baseline dietary fat consumption. Women in the intervention group with a higher baseline percentage of energy from fat provided stronger evidence for breast cancer reduction than women in the comparison group. Also, the HR varied among breast cancer subtypes defined by tumor hormone receptor characteristics. Such variation would not be expected if the intervention had no effect on breast cancer risk. As noted above, there were certain departures from the original study design. Although accrual goals were met, recruitment took longer than anticipated and therefore the average follow-up at the planned trial completion date was 8.1 years, rather than the original target of 9 years. In addition, the difference in the percentage of energy from fat between the women in the intervention group and women in the comparison group was only about 70% of the design goal. Relatively few women met the dietary target of 20% of energy from fat: 31.4% at year 1 and 14.4% at year 6. Also, the differences in the consumption of vegetables and fruit and grains between the intervention and comparison groups were modest. If the WHI design assumptions are revised to take into account these departures, projections are that breast cancer incidence in the intervention group would be 8% to 9% lower than in the comparison group the trial would be somewhat underpowered (projected power of approximately 60%) to detect a statistically significant difference,40 which is consistent with the observed results. This perspective is fur640 ther supported by our analyses demonstrating that the magnitude of the breast cancer HR was consistent with original design assumptions in the subset of adherent women. The argument for some intervention effect on breast cancer risk is strengthened also by the HR variation (Table 7) according to the progesterone receptor status of the tumor and according to the combined estrogen and progesterone receptor status. These variations were detected even though the tumor classification was based on local receptor laboratory results without standardization across clinical centers. Dependence of dietary pattern associations on breast tumor hormone receptor status also has been described in a preliminary report41 from the Women’s Intervention Nutrition Study and in the Nurses’ Health Study cohort.42 The HR variations across tumor characteristics are perhaps not surprising because breast cancer is increasingly recognized as a heterogeneous diagnosis in which medical interventions43-45 are effective primarily in subgroups defined by specific biological properties.46,47 Of interest also in relation to the finding of estradiol reduction among women in the intervention group (Table 4) are clinical trial results demonstrating the effectiveness of aromatase inhibitors, such as anastrazole in the ATAC (Arimidex, Tamoxifen, Alone or in Combination) trial,48 for breast cancer treatment. Trial results may suggest that factors other than estrogen48 contribute to any effect of a low-fat dietary pattern on breast cancer risk. Potential mechanisms include influence on insulin levels,49 insulin-like growth factors,50 and markers of inflammation. The last is consistent with the suggestion that women having higher baseline white blood cell counts show greater evidence of intervention benefit (Table 8). There are a number of limitations to the WHI dietary modification trial, including the reliance on self-report methods to assess differences in dietary consumption between the intervention and JAMA, February 8, 2006—Vol 295, No. 6 (Reprinted) Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013 comparison groups. However, relative changes between randomization groups in serum levels of ␥-tocopherol are consistent with intervention participant reports of decreases in consumption of added fats and oils51 and those of the carotenoids with FFQ differences in consumption of vegetables and fruit. Also, the available data are somewhat limited for the purpose of separating any breast cancer effect resulting from dietary fat reduction from that due to increases in the consumption of vegetables and fruit and/or grains. Similarly, there is limited potential to separate out the influence of any lifestyle changes or other nontargeted dietary changes that may result from adopting a low-fat dietary pattern. Additional biomarker data are being assembled to facilitate analyses of this kind. In light of our findings, additional research on diet and breast cancer prevention could focus on those women most likely to benefit from a low-fat dietary pattern, such as those with diets habitually high in fat. The potential differential effect of a low-fat dietary pattern by tumor subtype should continue to further characterize these subtypes and encourage the exploration of underlying mechanisms. Observational studies examining associations between diet and breast cancer should consider the use of consumption estimates that are calibrated with appropriate biomarkers. In conclusion, among postmenopausal women, a low-fat dietary pattern did not result in a statistically significant reduction in the risk of invasive breast cancer over an 8.1-year average follow-up period. However, nonsignificant trends were observed suggesting a reduced risk with a low-fat dietary pattern and incidence rate differences between groups are in agreement with design assumptions on acknowledging the dietary differences achieved. Because the health implications of a low-fat dietary pattern may take years to be fully realized, longer, planned, nonintervention follow-up may yield a more definitive comparison. ©2006 American Medical Association. All rights reserved. LOW-FAT DIETARY PATTERN AND RISK OF INVASIVE BREAST CANCER Author Affiliations: Fred Hutchinson Cancer Research Center, Seattle, Wash (Drs Prentice, Patterson, Anderson, LaCroix, and Henderson); Kaiser Permanente Division of Research, Oakland, Calif (Dr Caan); Harbor-UCLA Research and Education Institute, Torrance, Calif (Dr Chlebowski); University of Pittsburgh, Pittsburgh, Pa (Dr Kuller); University of Massachusetts/Fallon Clinic, Worcester (Dr Ockene); University of Minnesota, Minneapolis (Dr Margolis); University of Florida, Gainesville/Jacksonville (Dr Limacher); Brigham and Women’s Hospital and Harvard Medical School, Boston, Mass (Dr Manson); University of Miami, Miami, Fla (Dr Parker); Ohio State University, Columbus (Drs Paskett and Jackson); Emory University, Atlanta, Ga (Dr Phillips); University of California at Davis, Sacramento (Dr Robbins); National Heart, Lung, and Blood Institute, Bethesda, Md (Dr Rossouw); University of Wisconsin, Madison (Dr Sarto); University of Alabama, Birmingham (Dr Shikany); Stanford Prevention Research Center, Stanford, Calif (Dr Stefanick); University of Arizona, Tucson/Phoenix (Dr Thomson); Northwestern University, Chicago/ Evanston (Dr Van Horn); Wake Forest University School of Medicine, Winston-Salem, NC (Dr Vitolins); State University of New York, Buffalo (Dr WactawskiWende); University of Iowa, Iowa City/Davenport (Dr Wallace); Albert Einstein College of Medicine, Bronx, NY (Dr Wassertheil-Smoller); Kaiser Permanente Center for Health Research, Portland, Ore (Dr Whitlock); Pacific Health Research Institute, Honolulu, Hawaii (Dr Yano); MedStar Research Institute and Howard University, Washington, DC (Dr Adams-Campbell); Brown University, Providence, RI (Dr Assaf ); University of Washington, Seattle (Dr Beresford); Rush University Medical Center, Chicago, Ill (Dr Black); University of Nevada, Reno (Dr Brunner); University of Texas Health Science Center, San Antonio (Dr Brzyski); National Cancer Institute, Bethesda, Md (Dr Ford); University of Cincinnati, Cincinnati, Ohio (Dr Gass); Baylor College of Medicine, Houston, Tex (Dr Hays); University of California, Los Angeles (Dr Heber); University of North Carolina, Chapel Hill (Dr Heiss); Wayne State University School of Medicine and Hutzel Hospital, Detroit, Mich (Dr Hendrix); George Washington University, Washington, DC (Dr Hsia); University of California, Irvine (Dr Hubbell); University of Tennessee Health Science Center, Memphis (Dr Johnson); Medical College of Wisconsin, Milwaukee (Dr Kotchen); State University of New York, Stony Brook (Dr Lane); University of California at San Diego, LaJolla/Chula Vista (Dr Langer); University of Medicine and Dentistry of New Jersey, Newark (Dr Lasser). Author Contributions: Dr Prentice had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Prentice, Kuller, Ockene, Margolis, Manson, Robbins, Rossouw, Van Horn, Wactawski-Wende, Wallace, Anderson, Black, Ford, Hays, Heber, Lasser, Henderson. Acquisition of data: Prentice, Caan, Chlebowski, Patterson, Kuller, Ockene, Margolis, Limacher, Manson, Parker, Paskett, Phillips, Robbins, Sarto, Shikany, Stefanick, Thomson, Van Horn, WactawskiWende, Wallace, Wassertheil-Smoller, Whitlock, Yano, Adams-Campbell, Assaf, Beresford, Black, Brunner, Brzyski, Gass, Hays, Heber, Heiss, Hendrix, Hsia, Hubbell, Jackson, Johnson, Kotchen, LaCroix, Lane, Langer, Lasser, Henderson. Analysis and interpretation of data: Prentice, Caan, Chlebowski, Patterson, Kuller, Margolis, Limacher, Manson, Rossouw, Stefanick, Thomson, Van Horn, Vitolins, Wactawski-Wende, Wassertheil-Smoller, Anderson, Black, Heber, Jackson, LaCroix. Drafting of the manuscript: Prentice, Caan, Chlebowski, Patterson, Kuller, Ockene, Thomson, Van Horn, Vitolins, Adams-Campbell, Heber. Critical revision of the manuscript for important intel- lectual content: Prentice, Caan, Chlebowski, Patterson, Ockene, Margolis, Limacher, Manson, Parker, Paskett, Phillips, Robbins, Rossouw, Sarto, Shikany, Stefanick, Van Horn, Wactawski-Wende, Wallace, WassertheilSmoller, Whitlock, Yano, Anderson, Assaf, Beresford, Black, Brunner, Brzyski, Ford, Gass, Hays, Heber, Heiss, Hendrix, Hsia, Hubbell, Jackson, Johnson, Kotchen, LaCroix, Lane, Langer, Lasser, Henderson. Statistical analysis: Prentice, Caan, Patterson, Kuller, Van Horn, Anderson, Heber. Obtained funding: Prentice, Kuller, Ockene, Manson, Paskett, Robbins, Rossouw, Stefanick, Van Horn, Wactawski-Wende, Wallace, Wassertheil-Smoller, Adams-Campbell, Assaf, Beresford, Black, Brunner, Heber, Heiss, Hendrix, Hubbell, Lane, Langer, Henderson. Administrative, technical, or material support: Prentice, Kuller, Ockene, Margolis, Limacher, Manson, Paskett, Phillips, Robbins, Rossouw, Shikany, Stefanick, Van Horn, Wactawski-Wende, Wallace, Whitlock, Yano, Anderson, Assaf, Beresford, Brunner, Brzyski, Ford, Hays, Heber, Heiss, Hendrix, Hsia, Hubbell, Jackson, Johnson, Kotchen, Lane, Langer, Lasser, Henderson. Study supervision: Prentice, Chlebowski, Patterson, Kuller, Ockene, Limacher, Parker, Robbins, Stefanick, Thomson, Van Horn, Wallace, Wassertheil-Smoller, Anderson, Assaf, Black, Brunner, Hays, Heber, Heiss, Hendrix, Hubbell, Johnson. Financial Disclosures: Dr Assaf is an employee of Pfizer. Dr Black has received research grants from Pfizer and AstraZeneca; served on the speaker’s bureau for Pfizer, Novartis, Sanofi-Aventis, Bristol-Myers Squibb, Searle, Pharmacia, and Boehringer; and served as a consultant to or on an advisory board for Myogen, Merck Sharp & Dohme, Novartis, Mylan-Bertek, Pfizer, BristolMyers Squibb, and Sanofi-Aventis. No other authors reported disclosures. Funding/Support: The Women’s Health Initiative program was funded by the National Heart, Lung, and Blood Institute of the National Institutes of Health, Department of Health and Human Services. Role of the Sponsor: The funding organization had representation on the steering committee, which governed the design and conduct of the study, the interpretation of the data, and the preparation and approval of manuscripts. The National Heart, Lung, and Blood Institute Program Office reviewed the manuscript prior to publication. WHI Investigators by Clinical Center Program Office: National Heart, Lung, and Blood Institute, Bethesda, Md (Barbara Alving, Jacques Rossouw, Linda Pottern, Shari Ludlam, Joan McGowan, Nancy Geller, Leslie Ford). Clinical Coordinating Centers: Fred Hutchinson Cancer Research Center, Seattle, Wash (Ross Prentice, Garnet Anderson, Andrea LaCroix, Ruth Patterson, Anne McTiernan, Barbara Cochrane, Julie Hunt, Lesley Tinker, Charles Kooperberg, Martin McIntosh, C. Y. Wang, Chu Chen, Deborah Bowen, Alan Kristal, Janet Stanford, Nicole Urban, Noel Weiss, Emily White); Wake Forest University School of Medicine, WinstonSalem, NC (Sally Shumaker, Ronald Prineas, Michelle Naughton); Medical Research Laboratories, Highland Heights, Ky (Evan Stein, Peter Laskarzewski); San Francisco Coordinating Center, San Francisco, Calif (Steven R. Cummings, Michael Nevitt, Lisa Palermo); University of Minnesota, Minneapolis (Lisa Harnack); Fisher BioServices, Rockville, Md (Frank Cammarata, Steve Lindenfelser); University of Washington, Seattle (Bruce Psaty, Susan Heckbert). Clinical Centers: Albert Einstein College of Medicine, Bronx, NY (Sylvia Wassertheil-Smoller, William Frishman, Judith Wylie-Rosett, David Barad, Ruth Freeman); Baylor College of Medicine, Houston, Tex ( Jennifer Hays, Ronald Young, Jill Anderson, Sandy Lithgow, Paul Bray); Brigham and Women’s Hospital, Harvard Medical School, Boston, Mass ( JoAnn Manson, J. Michael Gaziano, Claudia Chae, Kathryn ©2006 American Medical Association. All rights reserved. Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013 Rexrode, Caren Solomon); Brown University, Providence, RI (Annlouise R. Assaf, Carol Wheeler, Charles Eaton, Michelle Cyr); Emory University, Atlanta, Ga (Lawrence Phillips, Margaret Pedersen, Ora Strickland, Margaret Huber, Vivian Porter); Fred Hutchinson Cancer Research Center, Seattle, Wash (Shirley A. A. Beresford, Vicky M. Taylor, Nancy F. Woods, Maureen Henderson, Robyn Andersen); George Washington University, Washington, DC ( Judith Hsia, Nancy Gaba, Joao Ascensao); Harbor-UCLA Research and Education Institute, Torrance, Calif (Rowan Chlebowski, Robert Detrano, Anita Nelson, Michele Geller); Kaiser Permanente Center for Health Research, Portland, Ore (Evelyn Whitlock, Patricia Elmer, Victor Stevens, Njeri Karanja); Kaiser Permanente Division of Research, Oakland, Calif (Bette Caan, Stephen Sidney, Geri Bailey Jane Hirata); Medical College of Wisconsin, Milwaukee ( Jane Morley Kotchen, Vanessa Barnabei, Theodore A. Kotchen, Mary Ann C. Gilligan, Joan Neuner); MedStar Research Institute and Howard University, Washington, DC (Barbara V. Howard, Lucile Adams-Campbell, Lawrence Lessin, Monique Rainford, Gabriel Uwaifo); Northwestern University, Chicago/Evanston, Ill (Linda Van Horn, Philip Greenland, Janardan Khandekar, Kiang Liu, Carol Rosenberg); Rush University Medical Center, Chicago, Ill (Henry Black, Lynda Powell, Ellen Mason; Martha Gulati); Stanford Prevention Research Center, Stanford, Calif (Marcia L. Stefanick, Mark A. Hlatky, Bertha Chen, Randall S. Stafford, Sally Mackey); State University of New York, Stony Brook (Dorothy Lane, Iris Granek, William Lawson, Gabriel San Roman, Catherine Messina); Ohio State University, Columbus (Rebecca Jackson, Randall Harris, Electra Paskett, W. Jerry Mysiw, Michael Blumenfeld); University of Alabama, Birmingham (Cora E. Lewis, Albert Oberman, James M. Shikany, Monika Safford, Mona Fouad); University of Arizona, Tucson/Phoenix (Tamsen Bassford, Cyndi Thomson, Marcia Ko, Ana Maria Lopez, Cheryl Ritenbaugh); State University of New York, Buffalo ( Jean Wactawski-Wende, Maurizio Trevisan, Ellen Smit, Susan Graham, June Chang); University of California at Davis, Sacramento ( John Robbins, Shagufta Yasmeen); University of California, Irvine (F. Allan Hubbell, Gail Frank, Nathan Wong, Nancy Greep, Bradley Monk); University of California, Los Angeles (Howard Judd, David Heber, Robert Elashoff ); University of California at San Diego, LaJolla/Chula Vista (Robert D. Langer, Michael H. Criqui, Gregory T. Talavera, Cedric F. Garland, Matthew A. Allison); University of Cincinnati, Cincinnati, Ohio (Margery Gass, Suzanne Wernke); University of Florida, Gainesville/ Jacksonville (Marian Limacher, Michael Perri, Andrew Kaunitz, R. Stan Williams, Yvonne Brinson); University of Hawaii, Honolulu ( J. David Curb, Helen Petrovitch, Beatriz Rodriguez, Kamal Masaki, Santosh Sharma); University of Iowa, Iowa City/ Davenport (Robert Wallace, James Torner, Susan Johnson, Linda Snetselaar, Jennifer Robinson); University of Massachusetts/Fallon Clinic, Worcester ( Judith Ockene, Milagros Rosal, Ira Ockene, Robert Yood, Patricia Aronson); University of Medicine and Dentistry of New Jersey, Newark (Norman Lasser, Baljinder Singh, Vera Lasser, John Kostis, Peter McGovern); University of Miami, Miami, Fla (Mary Jo O’Sullivan, Linda Parker, Timothy DeSantis, Diann Fernandez, Pat Caralis); University of Minnesota, Minneapolis (Karen L. Margolis, Richard H. Grimm, Mary F. Perron, Cynthia Bjerk, Sarah Kempainen); University of Nevada, Reno (Robert Brunner, William Graettinger, Vicki Oujevolk, Michael Bloch); University of North Carolina, Chapel Hill (Gerardo Heiss, Pamela Haines, David Ontjes, Carla Sueta, Ellen Wells); University of Pittsburgh, Pittsburgh, Pa (Lewis Kuller, Jane Cauley, N. Carole Milas); University of Tennessee Health Science Center, Memphis (Karen C. Johnson, Suzanne Satterfield, Raymond W. Ke, Stephanie Connelly, Fran Tylavsky); University of Texas Health Science Cen- (Reprinted) JAMA, February 8, 2006—Vol 295, No. 6 641 LOW-FAT DIETARY PATTERN AND RISK OF INVASIVE BREAST CANCER ter, San Antonio (Robert Brzyski, Robert Schenken, Jose Trabal, Mercedes Rodriguez-Sifuentes, Charles Mouton); University of Wisconsin, Madison (Gloria E. Sarto, Douglas Laube, Patrick McBride, Julie MaresPerlman, Barbara Loevinger); Wake Forest University School of Medicine, Winston-Salem, NC (Denise Bonds, Greg Burke, Robin Crouse, Mara Vitolins, Scott Washburn); Wayne State University School of Medicine and Hutzel Hospital, Detroit, Mich (Susan Hendrix, Michael Simon, Gene McNeeley). Former WHI Principal Investigators and Project Officers: John Foreyt (Baylor College of Medicine); Dallas Hall, Sally McNagny, Nelson Watts (Emory University); Valery Miller (George Washington University); Robert Hiatt (Kaiser Permanente); Barbara Valanis (Kaiser Permanente); Carolyn Clifford† (National Cancer Institute); Thomas Moon (University of Arizona); Frank Meyskens, Jr (University of California); James Liu (University of Cincinnati); Marianna Baum (University of Miami); Sandra Daugherty† (University of Nevada); David Sheps, Barbara Hulka (University of North Carolina); William Applegate (University of Tennessee); Catherine Allen† (University of Wisconsin). †Deceased. Acknowledgment: We acknowledge the dedicated efforts of the dietary modification trial participants and that of project staff, particularly nutritionists involved in delivering the dietary modification intervention and Aaron Aragaki, MS, of the clinical coordinating center of the Fred Hutchinson Cancer Research Center, who contributed much to the analytical aspects of this article. REFERENCES 1. Freedman LS, Clifford C, Messina M. Analysis of dietary fat, calories, body weight and the development of mammary tumors in rats and mice: a review. Cancer Res. 1990;50:5710-5719. 2. Prentice RL, Sheppard L. Dietary fat and cancer: consistency of the epidemiologic data and disease prevention that may follow from a practical reduction in fat consumption. Cancer Causes Control. 1990;1:8197. 3. Howe GR, Hirohata T, Hislop TG, et al. Dietary factors and risk of breast cancer: combined analysis of 12 case-control studies. J Natl Cancer Inst. 1990;82: 561-569. 4. Hunter DJ, Spiegelman D, Adami HO, et al. Cohort studies of fat intake and the risk of breast cancer: a pooled analysis. N Engl J Med. 1996;334:356361. 5. Boyd NF, Stone J, Vogt KN, Connelly BS, Martin LJ, Minkin S. Dietary fat and breast cancer risk revisited: a meta-analysis of the published literature. Br J Cancer. 2003;89:1672-1685. 6. Heitmann BL, Lissner L. Dietary underreporting by obese individuals: is it specific or non-specific. BMJ. 1995;311:986-989. 7. Subar AF, Kipnis V, Troiano RP, et al. Using intake biomarkers to evaluate the extent of dietary misreporting in a large sample of adults: the OPEN study. Am J Epidemiol. 2003;158:1-13. 8. Day NE, McKeown N, Wong MY, Welch A, Bingham S. Epidemiologic assessment of diet: a comparison of a 7-day diary with a food frequency questionnaire using urinary markers of nitrogen, potassium and sodium. Int J Epidemiol. 2001;30:309-317. 9. Bingham SA, Luben R, Welch A, Wareham N, Khaw KT, Day N. Are imprecise methods obscuring a relation between fat and breast cancer. Lancet. 2003;362: 212-214. 10. Insull W, Henderson MM, Prentice RL, et al. Results of a randomized feasibility study of a low-fat diet. Arch Intern Med. 1990;150:421-427. 11. Coates RJ, Bowen DJ, Kristal AR, et al. The Women’s Health Trial Feasibility Study in minority populations: changes in dietary intakes. Am J Epidemiol. 1999; 149:1104-1112. 642 12. Prentice R, Thompson D, Clifford C, Gorbach S, Goldin B, Byar D. Dietary fat reduction and plasma estradiol concentration among healthy postmenopausal women. J Natl Cancer Inst. 1990;82:129-134. 13. Rose DP, Connolly JM, Chlebowski RT, Buzzard IM, Wynder EL. The effects of a low-fat dietary intervention and tamoxifen adjuvant therapy on the serum estrogen and sex hormone-binding globulin concentrations of postmenopausal breast cancer patients. Breast Cancer Res Treat. 1993;27:253-262. 14. Rock CL, Flatt SW, Thompson CA, et al. Effects of a high-fiber, low-fat diet intervention on serum concentrations of reproductive steroid hormones in women with a history of breast cancer. J Clin Oncol. 2004;22: 2379-2387. 15. Wu AH, Pike MC, Stramm DO. Meta-analysis: dietary fat intake, serum estrogen levels, and risk of breast cancer. J Natl Cancer Inst. 1999;91:529-534. 16. World Cancer Research Fund. Food, Nutrition and the Prevention of Cancer: A Global Perspective. Washington, DC: American Institute for Cancer Research; 1997. 17. Smith-Warner SA, Spiegelman D, Yaun SS, et al. Intake of fruits and vegetables and risk of breast cancer: a pooled analysis of cohort studies. JAMA. 2001; 285:769-776. 18. Gandini S, Merzenich H, Robertson C, Boyle P. Meta-analysis of studies on breast cancer risk and diet: the role of fruit and Vegetable consumption and the intake of associated micronutrients. Eur J Cancer. 2000; 36:636-646. 19. Riboli E, Norat T. Epidemiologic evidence of the protective effect of fruit and vegetables on cancer risk. Am J Clin Nutr. 2003;78(suppl 3):559S-569S. 20. Working IARC Group on the Evaluation of CancerPreventive Strategies. Fruit and vegetables In: IARC Handbooks of Cancer Prevention: Vol. 8. Lyon, France: International Agency for Research on Cancer; 2003. 21. Chatenoud L, Tavani A, La Vecchia C, et al. Whole grain food intake and cancer risk. Int J Cancer. 1998; 77:24-28. 22. Jacobs DR, Marquent L, Slavin J, Kischi LH. Whole grain intake and cancer: an expanded review and meta-analysis. Nutr Cancer. 1998;30:85-96. 23. Nicodemus KK, Jacobs DR Jr, Folsom AR. Whole and refined grain and risk of incident postmenopausal breast cancer (United States). Cancer Causes Control. 2001;12:917-925. 24. Kipnis V, Midthune D, Freedman LS, et al. Empirical evidence of correlated biases in dietary assessment instruments and its implications. Am J Epidemiol. 2001;153:394-403. 25. Women’s Health Initiative Study Group. Design of the Women’s Health Initiative clinical trial and observational study. Control Clin Trials. 1998;19:61109. 26. Ritenbaugh C, Patterson R, Chlebowski RT, et al. The Women’s Health Initiative Dietary Modification Trial: overview and baseline characteristics of participants. Ann Epidemiol. 2003;13(9S):S87-S97. 27. Hays J, Hunt JR, Hubbell FA, et al. The Women’s Health Initiative recruitment methods and results. Ann Epidemiol. 2003;13(9S):S18-S77. 28. Writing Group for the Women’s Health Initiative Investigators. Risks and benefits of estrogen plus progestin in healthy postmenopausal women: principal results from the Women’s Health Initiative randomized controlled trial. JAMA. 2002;288:321-333. 29. Women’s Health Initiative Steering Committee. Effects of conjugated equine estrogen in postmenopausal women with hysterectomy. JAMA. 2004;291: 1701-1712. 30. Patterson RE, Kristal AR, Tinker LF, et al. Measurement characteristics of the Women’s Health Initiative food frequency questionnaire. Ann Epidemiol. 1999;9:178-187. 31. Tinker L, Burrows E, Henry H, Patterson R, Rupp J, Van Horn L. The Women’s Health Initiative: over- JAMA, February 8, 2006—Vol 295, No. 6 (Reprinted) Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013 view of the nutrition components. In: Krummel DA, Kris-Etherton PM, eds. Nutrition in Women’s Health. Gaithersburg, Md: Aspen Publishers; 1996:510-549. 32. Nutrition and Your Health: Dietary Guidelines for Americans. 3rd ed. Washington, DC: US Dept of Health and Human Services; 1990. 33. Curb JD, McTiernan A, Heckbert SR, et al. Outcomes ascertainment and adjudication methods in the Women’s Health Initiative. Ann Epidemiol. 2003; 13(9S):S122-S128. 34. Cox DR. Regression analysis and life tables. J R Stat Soc B. 1972;34:187-220. 35. Robins JM, Finkelstein DM. Correcting for noncompliance and dependent censoring in an AIDS clinical trial with inverse probability of censoring weighted (IPCW) log-rank tests. Biometrics. 2000;56:779-781. 36. Piegorsch WW, Weinberg CR, Taylor JA. Nonhierarchical logistic models and case-only designs for assessing susceptibility in population-based casecontrol studies. Stat Med. 1994;13:153-162. 37. Kalbfleisch JD, Prentice RL. The Statistical Analysis of Failure Time Data. 2nd ed. New York, NY: John Wiley & Sons Inc; 2002:255. 38. O’Brien PC, Fleming RT. A multiple testing procedure for clinical trials. Biometrics. 1979;35:549556. 39. Freedman L, Anderson G, Kipnis V, et al. Approaches to monitoring the results of long-term disease prevention trials: examples from the Women’s Health Initiative. Control Clin Trials. 1996;17:509-525. 40. Women’s Health Initiative Study Group. Dietary adherence in the Women’s Health Initiative Dietary Modification Trial. J Am Diet Assoc. 2004;104: 654-658. 41. Chlebowski RT, Blackburn GL, Elashoff RE, et al. Dietary fat reduction in postmenopausal women with primary breast cancer: phase III Women’s Intervention Nutrition Study (WINS) [abstract]. Proc Am Soc Clin Oncol. 2005;23(16S):3S. Abstract 10. 42. Fung TT, Hu FB, Holmes MD, et al. Dietary patterns and the risk of postmenopausal breast cancer. Int J Cancer. 2005;116:116-121. 43. Early Breast Cancer Trialists’ Collaborative Group. Effects of chemotherapy and hormonal therapy for early breast cancer on recurrence and 15 year survival: an overview of the randomized trials. Lancet. 2005;365:1687-1717. 44. Smith IE, Dowsett M. Aromatase inhibitors in breast cancer. N Engl J Med. 2003;348:2431-2442. 45. Romond EH, Perez EA, Bryant J, et al. Trastuzumab plus adjuvant chemotherapy for operable HER-2 positive breast cancer. N Engl J Med. 2005;353:16731684. 46. Piccart-Gebhart MJ, Procter M, Leyland-Jones B, et al; Herceptin Adjuvant (HERA) Trial Study Team. Trastuzumab after adjuvant chemotherapy in HER-2 positive breast cancer. N Engl J Med. 2005;353:16591672. 47. Rouzier R, Perou DM, Symmans WF, et al. Breast cancer molecular subtypes respond differently to preoperative chemotherapy. Clin Cancer Res. 2005;11: 5678-5685. 48. ATAC Trialist Group. Results of the ATAC (Arimidex, Tamoxifen, Alone or in Combination) trial after completion of 5 years of adjuvant treatment for breast cancer. Lancet. 2005;365:60-62. 49. Lawlor DA, Smith GD, Ebrahim S. Hyperinsulinaemia and increased risk of breast cancer: findings from the British Women’s Heart and Health Study. Cancer Causes Control. 2004;15:267-275. 50. Toniolo P, Bruning PF, Akhmedkhanov A, et al. Serum insulin-like growth factor-I and breast cancer. Int J Cancer. 2000;88:828-832. 51. Patterson RE, Kristal A, Rodabough R, et al. Changes in food sources of dietary fat in response to an intensive low-fat dietary intervention: early results from the Women’s Health Initiative. J Am Diet Assoc. 2003;103:454-460. ©2006 American Medical Association. All rights reserved.