Low-Fat Dietary Pattern and Breast Cancer Mortality in the Women's Health Initiative Randomized Controlled Trial

Abstract

Purpose Earlier Women's Health Initiative Dietary Modification trial findings suggested that a low-fat eating pattern may reduce breast cancers with greater mortality. Therefore, as a primary outcome-related analysis from a randomized prevention trial, we examined the long-term influence of this intervention on deaths as a result of and after breast cancer during 8.5 years (median) of dietary intervention and cumulatively for all breast cancers diagnosed during 16.1 years (median) of follow-up. Patients and Methods The trial randomly assigned 48,835 postmenopausal women with normal mammograms and without prior breast cancer from 1993 to 1998 at 40 US clinical centers to a dietary intervention with goals of a reduction of fat intake to 20% of energy and an increased intake of fruits, vegetables, and grains (40%; n = 19,541) or to a usual diet comparison (60%; n = 29,294). Results In the dietary group, fat intake and body weight decreased (all P < .001). During the 8.5-year dietary intervention, with 1,764 incident breast cancers, fewer deaths occurred as a result of breast cancer in the dietary group, which was not statistically significant (27 deaths [0.016% per year] v 61 deaths [0.024% per year]; hazard ratio [HR], 0.67; 95% CI, 0.43 to 1.06; P = .08). During the same period, deaths after breast cancer (n = 134) were significantly reduced (40 deaths [0.025% per year] v 94 deaths [0.038% per year]; HR, 0.65; 95% CI, 0.45 to 0.94; P = .02) by the dietary intervention. During the 16.1-year follow-up, with 3,030 incident breast cancers, deaths after breast cancer also were significantly reduced (234 deaths [0.085% per year] v 443 deaths [0.11% per year]; HR, 0.82; 95% CI, 0.70 to 0.96; P = .01) in the dietary group. Conclusion Compared with a usual diet comparison group, a low-fat dietary pattern led to a lower incidence of deaths after breast cancer.

Fig 1.

Participant flow for analyses of deaths as a result of and after breast cancer during the 8.3- and 8.5-year (mean and median) dietary intervention period and during the 16.5- and 16.1-year (mean and median) cumulative follow-up period in analyses that incorporated all 48,835 randomly assigned participants. Vertical arrows indicate the span of conducted National Death Index searches. WHI, Women’s Health Initiative.

Fig 2.

Breast cancer results by study period. Forest plot and summary statistics of the dietary modification influence on invasive breast cancer incidence, deaths as a result of breast cancer, and deaths after breast cancer by study period (dietary intervention period, 8.5 years; cumulative follow-up [intervention + postintervention] period, 16.1 years). (*) Includes deaths as a result of any cause that occurred after breast cancer diagnosis during the intervention period only. (†) Includes deaths as a result of any cause that occurred after breast cancer diagnosis during the intervention or postintervention period.

Fig 3.

Dietary modification influence on deaths as a result of and after breast cancer during the 8.5-year (median) dietary intervention period. (A) Kaplan-Meier cumulative hazard estimates for death as a result of breast cancer (breast cancer followed by death attributed to the cancer) during the 8.5-year (median) dietary intervention period and (B) Kaplan-Meier cumulative hazard estimates for death after breast cancer (breast cancer followed by death as a result of any cause) during the dietary intervention period among all 48,835 trial participants, with 1,764 breast cancers measured since random assignment. Summary statistics are from a Cox proportional hazards regression model stratified by age-group and random assignment in the hormone therapy trials. The P value corresponds to a two-sided score (log-rank test). HR, hazard ratio.

Fig 4.

Dietary modification influence on deaths as a result of and after breast cancer during the 16.1-year (median) cumulative follow-up. (A) Kaplan-Meier cumulative hazard estimates for death as a result of breast cancer during the 16.1-year (median) cumulative follow-up and (B) Kaplan-Meier cumulative hazard estimates for death after breast cancer during the 16.1-year (median) cumulative follow-up among all 48,835 trial participants, with 3,030 breast cancers measured since random assignment. Summary statistics are from a Cox proportional hazards regression model stratified by age-group, random assignment in the hormone therapy trials, and study period (time dependent). The P value corresponds to a two-sided score (log-rank test). HR, hazard ratio.

Fig 5.

Subgroup analysis by a forest plot of hazard ratios (HRs), dietary intervention versus comparison, for death after breast cancer by select subgroups of participants with an invasive breast cancer diagnosis that occurred anytime throughout the cumulative follow-up period. Annualized percentages are shown for all subgroups except percent energy from fat, where the numbers of deaths with 4-day food records were reported. Because of expense, baseline food records were scored for only breast cancers that occurred during the intervention period. For this analysis HRs were estimated in a case-only analysis by using logistic regression of random assignment stratified by the four ordinal groups of dietary fat. For rare outcomes (eg, < 5% incidence during study follow-up), case-only analyses provided HRs and corresponding 95% CIs that were essentially equivalent to those that would arise if the subgroups were available for the entire randomly assigned cohort. To illustrate, the corresponding case-only estimate for the overall effect on death after breast cancer (HR, 0.80) and the usual Cox proportional hazards regression estimate that includes deaths as a result of any cause that occurred after breast cancer diagnosis during the intervention period only (HR, 0.81) are provided. P values correspond to a test of the interaction between randomly assigned group and subgroup. (*) Includes deaths as a result of any cause that occurred after breast cancer diagnosis during the intervention or postintervention periods. (†) Among participants who did not report a prior hysterectomy. Ever used estrogen plus progestin includes random assignment into the Women’s Health Initiative conjugated equine estrogens plus medroxyprogesterone acetate trial. (‡) Among participants who reported a prior hysterectomy. Ever used estrogen alone includes random assignment into Women’s Health Initiative conjugated equine estrogens alone trial. §Includes deaths as a result of any cause that occurred after breast cancer diagnosis during the intervention period only. ‖Includes deaths as a result of any cause that occurred after breast cancer diagnosis during the intervention period only. Estimates were strictly based on participants who had baseline 4-day food records available (98%; 505 = 167 + 338 of 516 = 174 + 342). (¶) Baseline quartiles on the basis of 4-day food records of a 4.6% random subsample. BMI, body mass index.

Fig A1.

Study design. The arrows identify the two analysis periods (at the end of the 8.3- and 8.5-year [mean and median] dietary intervention period and after 16.5 and 16.1 [mean and median] years of cumulative follow-up). NDI, National Death Index; WHI DM, Women’s Health Initiative Dietary Modification.