Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
J Natl Cancer Inst. Author manuscript; available in PMC 2009 April 20.
Published in final edited form as:
PMCID: PMC2670850

Low-Fat Dietary Pattern and Cancer Incidence in the Women’s Health Initiative Dietary Modification Randomized Controlled Trial



The Women’s Health Initiative Dietary Modification (DM) Randomized Controlled Trial evaluated the effects of a low-fat dietary pattern on chronic disease incidence, with breast cancer and colorectal cancer as primary outcomes. The trial protocol also listed ovarian cancer and endometrial cancer as outcomes that may be favorably affected by the intervention.


A total of 48835 postmenopausal women were randomly assigned during 1993–1998 to a DM intervention (n = 19541) or comparison (usual diet; n = 29294) group and followed up for an average of 8.1 years. The intervention goal was to reduce total fat intake to 20% of energy and to increase consumption of vegetables, fruits, and grains. Cancer outcomes were verified by pathology report review. We used weighted log-rank tests to compare incidence of invasive cancers of the ovary and endometrium, total invasive cancer, and invasive cancers at other sites between the groups. All statistical tests were two-sided.


Ovarian cancer risk was lower in the intervention than in the comparison group (P = .03). Although the overall ovarian cancer hazard ratio (HR) was not statistically significantly less than 1.0, the hazard ratio decreased with increasing intervention duration (Ptrend = .01). For the first 4 years, the risk for ovarian cancer was similar in the intervention and control groups (0.52 cases per 1000 person-years in the intervention group versus 0.45 per 1000 person-years in the comparison group; HR = 1.16, 95% confidence interval [CI] = 0.73 to 1.84); over the next 4.1 years, the risk was lower in the intervention group (0.38 cases per 1000 person-years in the intervention group versus 0.64 per 1000 person-years in the comparison group; HR = 0.60, 95% CI = 0.38 to 0.96). Risk of cancer of the endometrium did not differ between the groups (P = .18). The estimated risk of total invasive cancer was slightly lower in the intervention group than in the control group (HR = 0.95, 95% CI = 0.89 to 1.01; P = .10).


A low-fat dietary pattern may reduce the incidence of ovarian cancer among postmenopausal women.

The Women’s Health Initiative (WHI) was initiated in 1992 (1) and included a full-scale randomized controlled trial of a dietary modification (DM) intervention with the goals of reduced fat intake (≤20% of energy from fat) and increased intake of vegetables and fruit (≥5 servings/day) and grains (≥6 servings/day). A total of 48835 postmenopausal women aged 50–79 years were enrolled, of whom 19541 (40%) were randomly assigned to the low-fat “dietary pattern” (intervention group) and 29294 (60%) were assigned to continue their usual diet (comparison group). The DM trial was designed to test whether a low-fat dietary pattern could reduce the risk of cancer among postmenopausal women, with breast and colorectal cancers listed as primary outcomes. Based on favorable plasma cholesterol effects of the DM in preceding feasibility studies (2), coronary heart disease was listed as a secondary outcome.

Results for the designated primary and secondary outcomes were recently reported (35). For breast cancer (3), the hazard ratio (HR) for the intervention versus comparison group was 0.91 (95% confidence interval [CI] = 0.83 to 1.01). The 9% lower incidence seen in the intervention group was similar to that projected under study design assumptions, given the measured dietary differences between randomization groups. In addition, there was a statistically significant interaction (Pinteraction = .04) between baseline percent energy from fat and breast cancer risk, with women in the upper quartile of percent energy from fat at baseline (>36.8% of total energy from fat) having a larger estimated reduction in risk with the intervention (HR = 0.78, 95% CI = 0.64 to 0.95). By contrast, the hazard ratio for colorectal cancer in the intervention versus comparison group was 1.08 (95% CI = 0.90 to 1.29), with no suggestion of intervention benefit (3).

The study protocol also listed the ovary and the endometrium as cancer sites that would potentially benefit from the DM intervention, in part based on international correlation analyses (6,7). Analytic epidemiologic studies also tend to support associations between reduced fat intake and reduced risk of these cancers. For example, the 1997 international review of food, nutrition, and the prevention of cancer (8) stated that “overall, the evidence suggests that diets high in total fat may increase the risk of ovarian cancer but is, as yet, insufficient,” with a nearly identical statement for endometrial cancer. For ovarian cancer, more recent studies (914) provide mixed findings. Recent analytic epidemiology studies of endometrial cancer include some reports of positive associations with dietary fat, particularly among obese women (15,16), but those also leave the question of association unresolved.

Observational studies are also inconclusive concerning the association between dietary fat and cancers of sites other than breast, colon, rectum, ovary, or endometrium. For example, the international review (8) lists lung cancer as possibly related to total dietary fat but does not list any cancer as “convincingly” or “probably” related to dietary fat. International correlation analyses, by contrast, have suggested positive associations of dietary fat with several cancers, including cancers of the kidney, bladder, and lung (6,8), but have only hypothesis generation potential.

In this study, we compared cancer incidence rates through the end of the DM trial intervention period for invasive cancers of the ovary and endometrium. We also evaluated the DM intervention in relation to total and site-specific invasive cancer.

Subjects and Methods

Detailed accounts of the methodology of the WHI DM trial have been presented (1,35). Briefly, participating women were postmenopausal and aged 50–79 years at recruitment during 1993–1998. Interested and eligible women could be randomly assigned to one or both of the DM trial and companion trials of postmenopausal hormone therapy and had the opportunity for further random assignment into a trial of calcium and vitamin D supplementation following 1 year of clinical trial participation. DM intervention and maintenance activities continued throughout the average 8.1-year average follow-up period (3), which concluded as planned on March 31, 2005. Major DM trial exclusions included any prior breast or colorectal cancer, other cancer except nonmelanoma skin cancer within the past 10 years, medical conditions yielding predicted survival of less than 3 years, adherence or retention concerns, or a baseline diet estimated to have less than 32% of energy from fat, as assessed by the WHI food-frequency questionnaire (FFQ).


Prior knowledge

Previously, the Women’s Health Initiative Dietary Modification (DM) trial analyzed whether a low-fat diet would alter the incidence of breast cancer, colorectal cancer, chronic disease, and cardiovascular disease.

Study design

Randomized controlled trial of postmenopausal women who were assigned to their usual diet or to the DM intervention. Risks of invasive ovarian and endometrial cancer as well as total invasive cancer and invasive cancer at other sites for a period of 8.1 years were determined.


Risk for invasive ovarian cancer was similar in the two groups in the first 4 years but reduced in the subsequent 4.1 years among women in the intervention group compared with women in the comparison group. No statistically significant differences in risk were observed among the two groups for total invasive cancer or invasive endometrial cancer.


A low-fat diet may reduce the incidence of ovarian cancer in post-menopausal women.


Adjustment for multiple comparisons for the risks for the five types of cancer targeted in the trial may reduce the statistical significance of the findings.

The 40% of women assigned to a low-fat dietary pattern received an intensive behavioral modification program to assist them in achieving the previously mentioned dietary intervention goals. The intervention program included 18 group sessions in the first year and quarterly maintenance sessions thereafter. In these sessions, groups of 8–15 women were led by specially trained and certified nutritionists. As elaborated in (1), each session included both nutritional topics (e.g., fat content of food, fat budgeting, high-risk food situations, and nutritional evaluation) and behavioral topics (e.g., dietary self-monitoring, social influences on eating, group cohesiveness, and relapse prevention). All participating women provided a 4-day food record at baseline and provided FFQs at baseline and 1 year and approximately every 3 years thereafter on a rotating basis, and randomly selected subsets provided 24-hour dietary recalls every 3 years.

As previously described (35), the dietary intervention resulted in noteworthy dietary differences between randomization groups as assessed by the WHI FFQ. In particular, the percentage of energy from fat was lower in the intervention group (versus the comparison group) by 10.7% at 1 year, 9.5% at 3 years, and 8.1% at 6 years. Consumption of vegetables and fruit was higher in the intervention group by 1.2, 1.3, and 1.1 servings at 1, 3, and 6 years from random assignment, respectively, and grain consumption was higher by 0.9, 0.7, and 0.4 servings at these times. Biomarker data (35) lend support to a meaningful dietary difference between intervention and control group women, including differences in blood estradiol and in certain blood micronutrient concentrations.

Participating women were queried twice per year regarding diagnosis of any cancer other than nonmelanoma skin cancers. Cancer screening behaviors, including mammograms (which were required at least every 2 years), pap smears, and colonoscopies, were tracked throughout the intervention period and did not differ substantially between randomization groups (3,4). Cancer reports were verified by medical record and pathology report review by centrally trained physician adjudicators at each of the 40 participating clinical centers (17). Central adjudication and coding at the clinical coordinating center using the National Cancer Institute’s Surveillance, Epidemiology, and End Results coding system also took place for cancers of the breast, colon, rectum, ovary, and endometrium. For this report, 308 cancers of “other” sites classified as in situ or borderline concerning invasiveness were reviewed, along with 69 cancers having unknown tumor behavior. As a result of this review, which was conducted blinded to random assignment, 55 of these 377 cancers were classified as invasive on central review and are included in this report. Review of a small sample (n = 30) of “other” cancers classified locally as invasive provided reassurance that few invasive cancers would be reclassified as noninvasive on central review. As in previous reports, disease events are included through the final intervention visit for each participating woman, which was scheduled between October 1, 2004, and March 31, 2005.

Statistical Analysis

The statistical design and analysis methods have also been described previously (1,3). Disease incidence comparisons between the intervention and comparison groups are based on the intent-to-treat principle using time-to-event methods. A weighted log-rank test was prespecified in the WHI protocol as the primary means of comparing randomization groups in the clinical trial. For cancer outcomes, the weights were specified to increase linearly from zero at random assignment to a plateau of 1.0 at 10 or more years following enrollment. This weighting procedure was selected to increase statistical power under hypothesized intervention effects that were more pronounced toward the end of the intervention period. Overall hazard ratio estimates and nominal 95% confidence intervals from Cox regression (18) analyses are also presented. These estimates arise from proportional hazards models, and confidence intervals that exclude 1.0 correspond to unweighted log-rank tests that are statistically significant at the α =.05 level. If the hazard rates for intervention and comparison groups are not proportional, the Cox model hazard ratio can be interpreted as estimating a type of averaged hazard ratio over the study follow-up period. Tests for time trends in hazard ratio over the intervention period were carried out by including a product term between randomization assignment and time from random assignment in the Cox regression procedure. All statistical tests were two-sided.

Analyses for ovarian cancer were restricted to women having at least one ovary at baseline. Analyses for endometrial cancer were restricted to women with a uterus at baseline. Interactions between hazard ratios and baseline factors (e.g., age, race/ethnicity, body mass index [BMI]) were examined by the inclusion of product terms between the randomization assignment and baseline factor categories in the Cox regression analysis. Interaction analyses with baseline dietary factor relied mostly on FFQ data. However, baseline FFQ percent energy from fat and total fat estimates were distorted for trial enrollees due to the use of the FFQ in eligibility screening. Hence, interactions with these factors used data from baseline 4-day food records. For reasons of cost, the 4-day food records were stored but not routinely analyzed in the trial cohort. The 4-day food records of ovarian cancer patients were analyzed for this report and were used in “patient-only” analyses to examine ovarian cancer hazard ratios according to baseline percentage energy from fat and total fat. This methodology was used also in earlier reports (35) from the DM trial. In the absence of a natural categorization (e.g., decade of age, major race/ethnicity, BMI categories), baseline factors were classified into quartiles, or into tertiles if the number of disease events was small (e.g., for ovarian cancer).

The Cox regression model was also used for explanatory analyses of intervention effects. For example, both an indicator variable for intervention group assignment and a time-dependent variable for body weight change from baseline to 1 year from random assignment were included in Cox model analyses (along with baseline weight) to examine whether weight changes attributable to the intervention provided an explanation for observed intervention effects on the hazard ratio.

The time to event for a particular outcome was defined as the number of days after randomization to the first diagnosis of the designated event (e.g., invasive cancer of any site). Follow-up time was censored at the time of a woman’s last documented contact within the intervention period for the trial, or death. Ovarian cancers were classified according to disease stage and tumor histology using the National Cancer Institute’s Surveillance, Epidemiology, and End Results coding system, with some grouping of rare histologic types.

From a multiple testing perspective, results for cancers of the ovary and endometrium can be viewed in the context of comparisons for each of the five “diet-related” cancers (breast, colon, rectum, ovary, and endometrium) specified in the DM trial protocol, and results for other cancers can be interpreted in the context of the entire set of approximately 25 site-specific comparisons. Statistical significance testing was based on the weighted log-rank test; trend testing and unweighted log-rank tests provided additional information about specific comparisons.


Baseline Characteristics

The baseline characteristics of the 19541 women in the intervention group and the 29294 women in the comparison group have been described (35). Briefly, the average age of study participants was 62.3 years, 18.6% were of minority race/ethnicity, about three-quarters were overweight or obese (BMI ≥ 25 kg/m2), and more than 40% reported a history of hypertension. The follow-up period ranged from 6 to more than 11 years and averaged 8.1 years.

Ovarian Cancer

We observed a lower incidence of ovarian cancer in the intervention group than in the comparison group (P = .03, from the protocol-specified weighted log-rank test) (Table 1) among the 39954 women (n = 15657 intervention, n = 23297 comparison) without prior bilateral oophorectomy at baseline. However, the hazard ratio averaged over the entire intervention period was not statistically significantly less than 1.0 (HR = 0.83, 95% CI = 0.60 to 1.14; unweighted log-rank P = .24) (Fig. 1,A). This apparent discrepancy can be explained by variation in this hazard ratio across the intervention period. Specifically, a test for trend in hazard ratio in relation to time from random assignment was statistically significant (Ptrend = .01). Dividing the 8.1-year average trial follow-up period into the first 4 and latter 4.1 years yielded hazard ratios of 1.16 (95% CI = 0.73 to 1.84, P = .53) and 0.60 (95% CI = 0.38 to 0.96, P = .03), respectively. The absolute incidence rates in the first 4 years were 0.52 cases per 1000 person-years in the intervention group and 0.45 cases per 1000 person-years in the comparison group. The corresponding rates in the subsequent years were 0.38 and 0.64 in the intervention and comparison groups, respectively. Hence, although there was little evidence for an intervention effect on ovarian cancer risk during the first few intervention years, a stronger and nominally statistically significant risk reduction emerged in the later years. Rates of bilateral oopherectomy during follow-up did not differ between randomization groups (P = .53), and the weighted log-rank test for the difference in incidence between the intervention and comparison groups remained statistically significant (P = .04) when the follow-up period was censored at the date of surgery for women undergoing bilateral oophorectomy during trial follow-up.

Fig. 1
Cumulative hazard estimates for invasive cancers in the intervention and comparison groups in the Women’s Health Initiative Dietary Modification trial. A) Invasive ovarian cancer. B) Total invasive cancer. P = statistical significance level based ...
Table 1
Comparison of incidence of invasive ovarian cancer, invasive endometrial cancer, and other invasive cancers between intervention and comparison groups in the Women’s Health Initiative (WHI) Dietary Modification trial*

We also examined the distribution of tumor histologic type and disease stage among women who developed invasive ovarian cancer (Table 2). The numbers of women in each category were small, but there did not appear to be any noteworthy differences in stage distribution between the intervention and comparison groups within major tumor histology categories.

Table 2
Stage and histology distribution of ovarian cancer patients by randomization group in the Women’s Health Dietary Modification trial*

Endometrial Cancer

The overall incidence of cancer of the endometrium did not differ between randomization groups (HR = 1.11, 95% CI = 0.88 to 1.40; P = .18), based on 27629 women (n = 11092 intervention, n = 16537 comparison) with a uterus at baseline. No indication of an intervention effect later in the intervention period was observed. Hysterectomy rates did not differ between randomization groups during follow-up (P = .85), and results were unchanged by additionally censoring follow-up times at the date of hysterectomy.

Breast, Colorectal, and All Invasive Cancers

The incidence of breast and colorectal cancers in the intervention and comparison groups was previously reported (34) and is given in Table 1, as is the incidence of invasive cancer at sites other than breast, colorectum, ovary, and endometrium. The hazard ratio for total (invasive) cancer was 0.95 (95% CI = 0.89 to 1.01), suggestive of an intervention benefit. The statistical significance level for the total cancer comparison was P = .10 (both weighted and unweighted log-rank tests). No suggestion was observed of a trend in hazard ratio for total cancer incidence with time from random assignment (Fig. 1, B; Ptrend = .68). For completeness, we note that hazard ratios for total cancer exclusive of breast cancer (HR = 0.96, 95% CI = 0.88 to 1.05) and for total cancer exclusive of colorectal cancer (HR = 0.94, 95% CI = 0.88 to 1.00; P = .05) were similar to that shown in Table 1 for total cancer.

Ovarian and Total Invasive Cancer Risk According to Baseline Characteristics

We next examined variations in the overall hazard ratio for ovarian cancer according to the baseline characteristics of participating women and to baseline dietary variables relevant to the DM intervention (Table 3). Interactions of hazard ratios with baseline percentage of energy from fat (P = .05) and baseline total fat intake (P = .06) were suggested. Among women whose values fell in the upper tertile for these variables, based on their baseline 4-day food records, estimated intervention versus comparison group hazard ratios (and 95% confidence intervals) over the entire follow-up period were 0.58 (95% CI = 0.31 to 1.08) for percentage of energy from fat and 0.49 (95% CI = 0.25 to 0.93) for total fat.

Table 3
Women’s Health Initiative Dietary Modification trial intervention versus 95% confidence intervals (CIs) for invasive ovarian cancer by baseline characteristics and dietary factors

Hazard ratio interaction analyses were also performed for total cancer (Table 4). No interactions were statistically significant, although there was a suggestion (P = .07) of a lower hazard ratio among women with a personal history of cancer (other than nonmelanoma skin cancer) before trial enrollment. Among these women, the hazard ratio was 0.74 (95% CI = 0.57 to 0.98).

Table 4
Women’s Health Initiative Dietary Modification trial intervention versus 95% confidence intervals (CIs) for total invasive cancer by baseline characteristics and dietary factors

Weight Change in Relation to Ovarian and Total Invasive Cancer Effects

The major emphasis of the DM intervention was on dietary fat reduction, with less emphasis placed on increasing intake of vegetables, fruits, and grains (4). The DM intervention did not target a reduction in total calories, although the intervention group did experience an early modest weight loss, with an average weight difference between randomization groups of 1.9 kg at 1 year from random assignment that diminished to 0.4 kg at 7.5 years (19). To test for a role of weight loss in explaining the observed hazard ratio trends, the hazard ratios for ovarian cancer and total cancer risk in the intervention versus comparison groups were recalculated in Cox model analyses that included both baseline weight and weight change from baseline to 1 year as a time-dependent covariate. The resulting overall intervention versus comparison group hazard ratios were 0.79 (95% CI = 0.55 to 1.13) for ovarian cancer and 0.95 (95% CI = 0.89 to 1.02) for total cancer, similar to the hazard ratio values given in Table 1 from the corresponding analyses without the weight and weight change variables. Thus, for these clinical outcomes, it is likely that any observed differences in disease incidence rates between intervention and comparison groups primarily reflect differences in percentage of energy obtained from fat.

Other Cancer Sites

The “all other sites” category of Table 1 was divided according to anatomic site (Table 5). Even categories with few incident events were included for completeness. The statistical significance level for Hodgkin disease was P = .05, based on only nine patients. Otherwise, none of the sites listed had a weighted log-rank P value less than or equal to .05, and none of the 95% confidence intervals excluded 1.0, although unweighted log-rank P values (not shown) were .06 for biliary tract cancer and .08 for liver cancer.

Table 5
Comparison of incidence rates for “other” cancer sites between intervention and comparison groups in the Women’s Health Initiative (WHI) Dietary Modification trial


This report provides evidence for a reduced risk of ovarian cancer as a result of the low-fat dietary pattern intervention, along with suggestive evidence for a reduction in total invasive cancer. However, several issues need to be considered in interpreting these findings. Ovarian cancer was one of five DM protocol-specified cancers tested. The probability that a statistical significance level as extreme as the observed weighted log-rank P = .03 arises by chance when five tests are conducted could be as large as 15% using a conservative Bonferroni correction. Also, the lack of a consistent effect across the entire intervention period may detract from the certainty of an intervention effect. Furthermore, there is a possibility that the cumulative hazard estimates shown in Fig. 1 could be distorted if ovarian cancers were detected earlier in the intervention group than in the comparison group.

The following points can be made in response to these issues and in support of an ovarian cancer risk reduction as a result of the intervention. Concerning the possibility of early detection in the intervention group, we note that the evidence for an early elevation in risk is weak. For example, a test of hazard ratio equal to 1.0 during the first 4 intervention years is not statistically significant (P = .53). Also, earlier detection would have given rise to a cumulative hazard curve for the intervention group that was elevated early in the intervention period and converged to that for the comparison group some years later, a pattern quite different from the crossing cumulative hazard curves shown in Fig. 1. Finally, the distribution of ovarian cancer diagnosis by stage and histology (Table 2) does not suggest any important differential ascertainment.

On the topic of multiple testing, we note that a hazard ratio trend test as extreme as P = .01 remains statistically significant at the 5% level when the Bonferroni correction for the five “diet-related” cancer sites is performed. Hence, the observed trend in ovarian cancer hazard ratio cannot easily be attributed to chance. The hazard ratio of 0.60 (95% CI = 0.38 to 0.96, P = .03) for the latter half of the intervention period is of particular interest in the context of this statistically significant hazard ratio trend, whereas the overall hazard ratio of 0.83 (95% CI = 0.60 to 1.14) can be viewed as diluted by little or no intervention effect during the early intervention years, as anticipated in trial design.

Perhaps the strongest data in favor of an intervention effect on ovarian cancer risk derive from analyses of hazard ratios in relation to baseline percentage of energy from fat. We have previously noted (3) that women whose baseline dietary fat intakes is high achieve a larger reduction in the percentage of energy from fat than do women with lower baseline dietary fat intakes, if assigned to the dietary intervention group. The women in the highest tertile of fat intake at baseline correspondingly had smaller ovarian cancer hazard ratios than women in the lowest tertile (Table 3).

The suggestion (P = .10) of a modestly reduced total invasive cancer hazard ratio among intervention group women (HR = 0.95, 95% CI = 0.89 to 1.01) could be of some practical importance. Also, the total cancer hazard ratio interaction analyses (Table 4) suggest a lower risk of invasive cancer in the intervention versus comparison group among women having a personal history of cancer before trial enrollment. The 1-year FFQ difference in percentage of energy from fat between randomization groups was slightly larger (P = .04) for women with a personal history of cancer (11.45%) than for women without such a history (10.72%), so it is possible that differences in adherence to the dietary intervention could contribute to this suggested interaction.

The results seen in Table 5 for Hodgkin disease (P = .05, HR = 0.19), biliary tract cancer (P = .20, HR = 1.96), and liver cancer (P = .31, HR = 2.30) can readily be attributed to chance. They arise in the context of approximately 25 comparisons, each based on a small number of disease events. Also, the limited observational literature for biliary tract and liver cancer mostly (8,20), but not entirely (21), tend to suggest a positive association with dietary fat.

In summary, the DM trial indicates that a low-fat eating pattern may reduce ovarian cancer risk (P = .03), although this finding needs to be interpreted in the context of comparisons for five cancer sites. The DM trial also suggests (P = .10) a possible reduction in total invasive cancer. Ongoing nonintervention follow-up of trial participants may provide additional valuable assessment of the effects of a low-fat dietary pattern on these and other cancer incidence rates.



National Heart, Lung, and Blood Institute with investigators funded under a contract mechanism (NHLBI N01-WH-2-2110).

The authors thank the WHI investigators and staff for their outstanding dedication and commitment. A list of key investigators involved in this research follows.

A full listing of WHI investigators can be found at the following Web site:

Program Office: (National Heart, Lung, and Blood Institute, Bethesda, MD). Elizabeth Nabel, Jacques Rossouw, Linda Pottern, Shari Ludlam, Joan McGowan.

Clinical Coordinating Center: 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 (Fred Hutchinson Cancer Research Center, Seattle, WA); Sally Shumaker, Pentti Rautaharju, Ronald Prineas, Michelle Naughton (Wake Forest University School of Medicine, Winston-Salem, NC); Evan Stein, Peter Laskarzewski (Medical Research Labs, Highland Heights, KY); Steven Cummings, Michael Nevitt, Maurice Dockrell (University of California at San Francisco, San Francisco, CA); Lisa Harnack (University of Minnesota, Minneapolis, MN); Frank Cammarata, Steve Lindenfelser (Fisher BioServices, Rockville, MD); Bruce Psaty, Susan Heckbert (University of Washington, Seattle, WA).

Clinical Centers: Sylvia Wassertheil-Smoller, William Frishman, Judith Wylie-Rosett, David Barad, Ruth Freeman (Albert Einstein College of Medicine, Bronx, NY); Jennifer Hays, Ronald Young, Jill Anderson, Sandy Lithgow, Paul Bray (Baylor College of Medicine, Houston, TX); JoAnn Manson, J. Michael Gaziano, Claudia Chae, Kathryn Rexrode, Caren Solomon (Brigham and Women’s Hospital, Harvard Medical School, Boston, MA); Annlouise R. Assaf, Carol Wheeler, Charles Eaton, Michelle Cyr (Brown University, Providence, RI); Lawrence Phillips, Margaret Pedersen, Ora Strickland, Margaret Huber, Vivian Porter (Emory University, Atlanta, GA); Shirley A. A. Beresford, Vicky M. Taylor, Nancy F. Woods, Maureen Henderson, Robyn Andersen (Fred Hutchinson Cancer Research Center, Seattle, WA); Judith Hsia, Nancy Gaba, Joao Ascensao (George Washington University, Washington, DC); Rowan Chlebowski, Robert Detrano, Anita Nelson, James Heiner, John Marshall (Harbor—UCLA Research and Education Institute, Torrance, CA); Cheryl Ritenbaugh, Barbara Valanis, Victor Stevens, Njeri Karanja (Kaiser Permanente Center for Health Research, Portland, OR); Bette Caan, Stephen Sidney, Geri Bailey, Jane Hirata (Kaiser Permanente Division of Research, Oakland, CA); Jane Morley Kotchen, Vanessa Barnabei, Theodore A. Kotchen, Mary Ann C. Gilligan, Joan Neuner (Medical College of Wisconsin, Milwaukee, WI); Barbara V. Howard, Lucile Adams-Campbell, Lawrence Lessin, Monique Rainford, Gabriel Uwaifo (MedStar Research Institute/Howard University, Washington, DC); Linda Van Horn, Philip Greenland, Janardan Khandekar, Kiang Liu, Carol Rosenberg (Northwestern University, Chicago/Evanston, IL); Henry Black, Lynda Powell, Ellen Mason (Rush Medical Center, Chicago, IL);) Marcia L. Stefanick, Mark A. Hlatky, Bertha Chen, Randall S. Stafford, Sally Mackey (Stanford Prevention Research Center, Stanford University, Stanford, CA; Dorothy Lane, Iris Granek, William Lawson, Gabriel San Roman, Catherine Messina (State University of New York at Stony Brook, Stony Brook, NY); Rebecca Jackson, Randall Harris, Electra Paskett, W. Jerry Mysiw, Michael Blumenfeld (The Ohio State University, Columbus, OH); Cora E. Lewis, Albert Oberman, James M. Shikany, Monika Safford, Brian K. Britt (University of Alabama at Birmingham, Birmingham, AL); Tamsen Bassford, Cyndi Thomson, Marcia Ko, Ana Maria Lopez (University of Arizona, Tucson/Phoenix, AZ); Jean Wactawski-Wende, Maurizio Trevisan, Ellen Smit, Susan Graham, June Chang (University at Buffalo, Buffalo, NY); John Robbins, S. Yasmeen (University of California at Davis, Sacramento, CA); F. Allan Hubbell, Gail Frank, Nathan Wong, Nancy Greep, Bradley Monk (University of California at Irvine, Irvine, CA); Howard Judd, David Heber, Robert Elashoff (University of California at Los Angeles, Los Angeles, CA); Robert D. Langer, Michael H. Criqui, Gregory T. Talavera, Cedric F. Garland, R. Elaine Hanson (University of California at San Diego, LaJolla/Chula Vista, CA); Margery Gass, Suzanne Wernke (University of Cincinnati, Cincinnati, OH); Marian Limacher, Michael Perri, Andrew Kaunitz, R. Stan Williams, Yvonne Brinson (University of Florida, Gainesville/Jacksonville, FL); David Curb, Helen Petrovitch, Beatriz Rodriguez, Kamal Masaki, Santosh Sharma (University of Hawaii, Honolulu, HI); Robert Wallace, James Torner, Susan Johnson, Linda Snetselaar, Bradley VanVoorhis (University of Iowa, Iowa City/Davenport, IA); Judith Ockene, Milagros Rosal, Ira Ockene, Robert Yood, Patricia Aronson (University of Massachusetts/Fallon Clinic, Worcester, MA); Norman Lasser, Baljinder Singh, Vera Lasser, John Kostis (University of Medicine and Dentistry of New Jersey, Newark, NJ); Mary Jo O’Sullivan, Linda Parker, R. Estape, Diann Fernandez (University of Miami, Miami, FL); Karen L. Margolis, Richard H. Grimm, Donald B. Hunninghake, June LaValleur, Sarah Kempainen (University of Minnesota, Minneapolis, MN); Robert Brunner, William Graettinger, Vicki Oujevolk (University of Nevada, Reno, NV); Gerardo Heiss, Pamela Haines, David Ontjes, Carla Sueta, Ellen Wells (University of North Carolina, Chapel Hill, NC); Lewis Kuller, Jane Cauley, N. Carole Milas (University of Pittsburgh, Pittsburgh, PA); Karen C. Johnson, Suzanne Satterfield, Raymond W. Ke, Stephanie Connelly, Fran Tylavsky (University of Tennessee Health Science Center, Memphis, TN); Robert Brzyski, Robert Schenken, Jose Trabal, Mercedes Rodriguez-Sifuentes, Charles Mouton (University of Texas Health Science Center, San Antonio, TX); Gloria Sarto, Douglas Laube, Patrick McBride, Julie Mares-Perlman, Barbara Loevinger (University of Wisconsin, Madison, WI); Denise Bonds, Greg Burke, Robin Crouse, Mara Vitolins, Scott Washburn (Wake Forest University School of Medicine, Winston-Salem, NC); Susan Hendrix, Michael Simon, Gene McNeeley (Wayne State University School of Medicine/Hutzel Hospital, Detroit, MI).



Clinical Trials Registration— identifier: NCT00000611. Decisions concerning study design, data collection and analysis, interpretation of the results, the preparation of the manuscript, or the decision to submit the manuscript for publication resided with committees comprising WHI investigators that included National Heart, Lung, and Blood Institute representatives. An external Data and Safety Monitoring Committee met twice yearly throughout the study intervention period, and recommended changes, as appropriate to the sponsor. This study proceeded to its planned termination date.


(1) Women’s Health Initiative Study Group Design of the Women’s Health Initiative clinical trial and observational study. Control Clin Trials. 1998;19:61–109. [PubMed]
(2) Insull W, Henderson MM, Prentice RL, Thompson DJ, Clifford C, Goldman S, et al. Results of a randomized feasibility study of a low-fat diet. Arch Intern Med. 1990;150:421–7. [PubMed]
(3) Prentice RL, Caan B, Chlebowski RT, Patterson R, Kuller LH, Ockene JK, et al. Low-fat dietary pattern and risk of invasive breast cancer: the Women’s Health Initiative randomized controlled Dietary Modification trial. JAMA. 2006;295:629–42. [PubMed]
(4) Beresford SA, Johnson KC, Ritenbaugh C, Lasser NL, Snetselaar LG, Black HR, et al. Low-fat dietary pattern and risk of colorectal cancer: the Women’s Health Initiative randomized controlled Dietary Modification trial. JAMA. 2006;295:643–54. [PubMed]
(5) Howard BV, Van Horn L, Hsia J, Manson JE, Stefanick ML, Wassertheil-Smoller S, et al. Low-fat dietary pattern and risk of cardiovascular disease: the Women’s Health Initiative randomized controlled Dietary Modification trial. JAMA. 2006;295:655–66. [PubMed]
(6) Armstrong B, Doll R. Environmental factors and cancer incidence and mortality in different countries, with special reference to dietary practices. Int J Cancer. 1975;15:617–31. [PubMed]
(7) 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;1:81–97. [PubMed]
(8) World Cancer Research Fund . Food, nutrition, and the prevention of cancer: a global perspective. American Institute for Cancer Research; Washington (DC): 1997. [PubMed]
(9) Huncharek M, Kupelnick B. Dietary fat intake and risk of epithelial ovarian cancer: a meta-analysis of 6,689 subjects from 8 observational studies. Nutr Cancer. 2001;40:87–91. [PubMed]
(10) Pan SY, Ugnat AM, Mao Y, Wen SW, Johnson KC, Canadian Cancer Registries Epidemiology Research Group A case-control study of diet and the risk of ovarian cancer. Cancer Epidemiol Biomarkers Prev. 2004;13:1521–7. [PubMed]
(11) Zhang M, Lee AH, Binns CW. Reproductive and dietary risk factors for epithelial ovarian cancer in China. Gynecol Oncol. 2004;92:320–6. [PubMed]
(12) Lubin F, Chetrit A, Modan B, Freedman LS. Dietary intake changes and their association with ovarian cancer risk. J Nutr. 2006;136:2362–7. [PubMed]
(13) Mommers M, Schouten LJ, Goldbohm RA, Van den Brandt PA. Dairy consumption and ovarian cancer risk in the Netherlands Cohort Study on Diet and Cancer. Br J Cancer. 2006;94:165–70. [PMC free article] [PubMed]
(14) Genkinger JM, Hunter DJ, Spiegelmann D, Anderson KE, Beeson WL, Buring JE, et al. A pooled analysis of 12 cohort studies of dietary fat, cholesterol, and egg intake and ovarian cancer. Cancer Causes Control. 2006;17:273–85. [PubMed]
(15) Littman AJ, Beresford SA, White E. The association of dietary fat and plant foods with endometrial cancer (United States) Cancer Causes Control. 2001;12:691–702. [PubMed]
(16) Goodman MT, Hankin JH, Wilkens LR, Lyu LC, McDuffie K, Liu LQ, et al. Diet, body size, physical activity and risk for endometrial cancer. Cancer Res. 1997;57:5077–85. [PubMed]
(17) Curb JD, McTiernan A, Heckbert SR, Kooperberg C, Stanford J, Nevitt M, et al. Outcomes ascertainment and adjudication methods in the Women’s Health Initiative. Ann Epidemiol. 2003;13(Suppl):S122–8. [PubMed]
(18) Cox DR. Regression analysis and life tables (with discussion) J R Stat Soc. 1972;34:187–220.
(19) Howard BV, Manson JE, Stefanick ML, Beresford SA, Frank G, Jones B, et al. Low-fat eating pattern and weight change over 7 years in the Women’s Health Initiative dietary modification trial. JAMA. 2006;295:39–49. [PubMed]
(20) Zatonski WA, Lowenfels AB, Boyle P, Maisonneuve P, de Mesquita HB Bueno, Ghadirian P, et al. Epidemiologic aspects of gallbladder cancer: a case-control study of the SEARCH Program of the International Agency for Research on Cancer. J Natl Cancer Inst. 1997;89:1132–8. [PubMed]
(21) Carroll KK, Braden LM, Bell JA, Kalamegham R. Fat and cancer. Cancer. 1986;58:S1818–25. [PubMed]