Abstract
Background: Soy isoflavones have antiestrogenic and anticancer properties but also possess estrogen-like properties, which has raised concern about soy food consumption among breast cancer survivors.
Objective: We prospectively evaluated the association between postdiagnosis soy food consumption and breast cancer outcomes among US and Chinese women by using data from the After Breast Cancer Pooling Project.
Design: The analysis included 9514 breast cancer survivors with a diagnosis of invasive breast cancer between 1991 and 2006 from 2 US cohorts and 1 Chinese cohort. Soy isoflavone intake (mg/d) was measured with validated food-frequency questionnaires. HRs and 95% CIs were estimated by using delayed-entry Cox regression models, adjusted for sociodemographic, clinical, and lifestyle factors.
Results: After a mean follow-up of 7.4 y, we identified 1171 total deaths (881 from breast cancer) and 1348 recurrences. Despite large differences in soy isoflavone intake by country, isoflavone consumption was inversely associated with recurrence among both US and Chinese women, regardless of whether data were analyzed separately by country or combined. No heterogeneity was observed. In the pooled analysis, consumption of ≥10 mg isoflavones/d was associated with a nonsignificant reduced risk of all-cause (HR: 0.87; 95% CI: 0.70, 1.10) and breast cancer–specific (HR: 0.83; 95% CI: 0.64, 1.07) mortality and a statistically significant reduced risk of recurrence (HR: 0.75; 95% CI: 0.61, 0.92).
Conclusion: In this large study of combined data on US and Chinese women, postdiagnosis soy food consumption of ≥10 mg isoflavones/d was associated with a nonsignificant reduced risk of breast cancer–specific mortality and a statistically significant reduced risk of recurrence. One of the studies included in the After Breast Cancer Pooling Project, the Women's Healthy Eating & Living Study, was registered at clinicaltrials.gov as NCT00003787.
INTRODUCTION
Soy constituents may possess antiestrogenic and other anticancer properties (1–3). In addition, soy food consumption has been associated with a lower risk of developing breast cancer (4–6). However, soy isoflavones also may have estrogen-like properties, including the ability to bind to estrogen receptors (ERs)4 in the breast and stimulate cellular proliferation (7, 8). Furthermore, experimental studies suggest that soy isoflavones may interact with tamoxifen therapy (2), with some studies showing a potential benefit of combined dietary isoflavone intake and tamoxifen therapy use on the inhibition of breast tumor growth (9, 10), whereas other studies have reported a reduction in the anticancer effects of tamoxifen on breast tissue (11, 12). These data have raised concern regarding soy food consumption among breast cancer survivors (2, 8, 13).
Epidemiologic data on the association of postdiagnosis soy food consumption with breast cancer outcomes are sparse. Three reports to date [2 from the United States (14, 15) and 1 from China (16)] suggest that postdiagnosis soy food intake may improve prognosis among breast cancer survivors. Most women in the US studies did not consume soy food daily (only 10% consumed ≥10 mg isoflavones/d). In contrast, most Chinese women consumed soy food daily (90% consumed ≥10 mg isoflavones/d). These substantial differences in soy consumption prevented a direct comparison and thus a general recommendation. In addition, the US studies had limited power to evaluate associations by potential effect modifiers (eg, hormone-receptor status and tamoxifen use) or across a wide range of intakes. To overcome these limitations, we conducted a pooled analysis to investigate soy food intake and breast cancer outcomes by using the After Breast Cancer Pooling Project (ABCPP), an international collaboration of prospective studies of breast cancer survivors (17).
SUBJECTS AND METHODS
After Breast Cancer Pooling Project
The ABCPP design and methods were described previously (17). Briefly, the ABCPP includes pooled and harmonized data on 18,314 breast cancer survivors from 4 prospective cohort studies recruited from multiple US sites and Shanghai, China. Three of the cohorts recruited only breast cancer patients: the Shanghai Breast Cancer Survival Study (SBCSS) (16), the Life After Cancer Epidemiology (LACE) Study (18), and the Women's Healthy Eating & Living (WHEL) Study (19). The fourth cohort consists of patients with breast cancer diagnosed in the Nurses’ Health Study (20). Institutional review board approval was obtained for each cohort study. The Nurses’ Health Study was excluded from the current study because data on soy isoflavone intake were unavailable for this cohort.
Soy food assessment
Soy food intake was assessed with validated food-frequency questionnaires (FFQs), mainly focusing on dietary intake after diagnosis (16, 18, 19). The SBCSS FFQ was specifically designed to capture nutrient and major food intake among Chinese women living in Shanghai, including consumption of soy foods commonly consumed in Shanghai (tofu, soy milk, fresh soy beans, and other soy products) (21). The LACE study used the Fred Hutchinson Cancer Research Center FFQ, which is an adaptation of the 95-item Health Habits and Lifestyle Questionnaire developed by Block and colleagues at the National Cancer Institute (22), and a separate soy food–specific FFQ, which included 14 soy food items selected based on their soy contribution to dietary intake of isoflavones (14). The WHEL study used the Arizona FFQ, which is a modification of the food frequency component of the Block National Cancer Institute Health Habits and Lifestyle Questionnaire (19, 22). The SBCSS and the LACE study derived soy isoflavone intake based on the isoflavone content of soy food items included in isoflavone nutrient databases (14) or food-composition tables (23). The WHEL study derived isoflavone intake based on the USDA isoflavone database (15), which includes the isoflavone content for soy food and other dietary sources, although the major sources of dietary isoflavones are soy foods (24). Isoflavones were available at baseline and updated twice for the WHEL study and the SBCSS and once for the LACE study.
The validity of the soy food assessment in the SBCSS was evaluated in a validation study among 200 participants of the Shanghai Women's Health Study (21). Participants completed a baseline FFQ, multiple 24-h dietary recalls twice per month consecutively for 12 mo, and a second FFQ at the end of the study. Pearson correlation coefficients between intakes derived from the FFQ and the averages of the multiple 24-h recalls were 0.49 for soy food and 0.41–0.66 for other major food groups. The correlation coefficient between the first and second FFQs was 0.37 for soy food (21). Although the FFQs used in the US cohorts were validated for major nutrients (25–27), soy food intake has not been specifically validated for the US FFQs. The FFQs used in each study to assess soy food intake are available on request.
This pooled analysis includes additional years of follow-up/events compared with the previously published reports (14–16) and includes findings for all-cause mortality, breast cancer–specific mortality, and recurrence. The previous LACE publication only investigated breast cancer recurrence and did not report on mortality outcomes, and none of the previous studies reported findings for breast cancer–specific mortality. Furthermore, the previous reports from the US cohorts used baseline data only and did not consider updated soy food information.
Clinical information assessment
For the current analysis, treatment information was available for the first primary breast cancer diagnosis only. For the SBCSS, primary treatment data and tumor characteristics [eg, tumor stage and ER/progesterone receptor (PR) status] were collected via self-report at the baseline interview (mean of 6.5 mo after diagnosis) and verified by medical record review (16). For the WHEL study, tumor characteristics were abstracted from pathology reports, and primary treatment data were self-reported at baseline (mean of 23.4 mo after diagnosis) and confirmed by medical record review (19). For the LACE study, primary treatment information and tumor characteristics were available through electronic databases for Kaiser Permanente Northern California members (∼83% of the cohort) and medical charts of non–Kaiser Permanente Northern California participants (18) and were collected at cohort enrollment (mean of 22.5 mo after diagnosis).
Cohort follow-up and outcome ascertainment
Each cohort followed participants to ascertain outcomes of interest for the current study [total mortality, breast cancer–specific mortality, and recurrences (defined as recurrence/metastasis or development of new primary breast cancer)]. Detailed methods were previously published for the ABCPP (17) and each cohort: SBCSS (16), WHEL (19), and LACE (18). For the SBCSS, outcomes were obtained via in-person interviews 18, 36, and 60 mo after diagnosis, supplemented by record linkage to the Shanghai Vital Statistical Registry. Medical records were not obtained to verify recurrences in the SBCSS; however, research has shown recurrence data based on self-report to be valid (28). For the WHEL study, outcomes were obtained via semi-annual telephone contact and clinic visits through the end of the trial (June 2006), with all reported events confirmed by medical records review (29). Follow-up for deaths only has continued through linkage to death indexes. For the LACE study, outcomes were ascertained on a semi-annual basis via mailed surveys until 5 y after diagnosis and yearly thereafter, and medical records were obtained to verify any reported breast cancer outcomes (18). Medical records were also checked monthly for Kaiser Permanente Northern California members for new events, and death indexes were searched for women lost to follow-up. Cause of death information was missing for 38 women in the ABCPP.
Statistical analysis
Women with missing isoflavone data (n = 665), stage IV tumors (n = 28), or death/loss to follow-up before the soy food assessment (n = 4) were excluded, which resulted in a final analytic sample of 9514 women. An additional 8 women with no information on disease-free survival were excluded from the recurrence analysis. With the use of study-specific deciles, overlap between soy isoflavone intakes in the United States and China occurred at about the 90th percentile for US women (10 mg/d), approximately the 10th percentile for Chinese women. We present pooled results by using both study-specific deciles and common cutoffs in milligrams per day (<4, 4–9.99, and ≥10). The common cutoffs were based on the 85th (4 mg/d) and 90th (10.0 mg/d) percentiles of the isoflavone distribution of the US cohorts and the 10th percentile for the Chinese cohort.
We used delayed-entry Cox proportional hazards regression models to estimate study-specific adjusted HRs and 95% CIs. The entry date was the date of the first measure of soy food intake after diagnosis. The exit date was date of death (or recurrence for the recurrence analysis) or date of last contact (ie, date of last follow-up survey or last registry linkage, whichever was most recent). For the main analyses, isoflavone intake was based on baseline data. We conducted secondary analyses treating isoflavones as time-dependent variables by using all available soy food information. The Q statistic was used to test for heterogeneity in risk estimates across studies (30). If heterogeneity was present, we conducted a pooled analysis with study-specific HRs with the use of inverse-variance weights in random-effects models (31). Otherwise, we conducted a pooled analysis by use of the combined data and Cox regression models stratified by study.
Clinical predictors, sociodemographic characteristics, and lifestyle-related factors assessed at baseline and considered as potential confounders (Table 1) included the American Joint Committee on Cancer (6th edition) TNM stage, ER/PR status, chemotherapy, radiotherapy, hormonal therapy, education, race-ethnicity, first-degree family history of breast cancer, menopausal status, parity, recreational physical activity in metabolic equivalent hours per week (32), smoking, cruciferous vegetable intake, and BMI (33). Final models included all potential confounders, with the exception of family history of breast cancer, which did not have a statistically significant association with isoflavones. Missing covariate data were included via indicator variables, except when the missing data for a covariate were very small (n < 10) by study; then, these women were excluded (a total of 30 women were excluded, all from the US cohorts).
TABLE 1.
Cohort overview, baseline participant characteristics, and soy isoflavone intake after diagnosis by study: After Breast Cancer Pooling Project (n = 9514)1
| SBCSS (n = 4856) | WHEL (n = 2729) | LACE (n = 1929) | |
| Study design | Prospective cohort | Prospective follow-up of participants of an RCT | Prospective cohort |
| Location | Shanghai, China | 7 sites in the Western United States | California, Utah, and WHEL sites |
| Median duration of follow-up (y) | 4.2 | 9.5 | 8.6 |
| Time of diagnosis | 2002–2006 | 1991–2000 | 1997–2000 |
| Time of recruitment | 2002–2006 | 1995–2000 | 2000–2002 |
| Age (y) | 53.5 ± 10.02 | 51.3 ± 8.9 | 58.6 ± 10.8 |
| Deaths (n) | 466 | 392 | 313 |
| Deaths from breast cancer (n) | 405 | 307 | 169 |
| Recurrences (n) | 533 | 504 | 311 |
| Race-ethnicity [n (%)] | |||
| Non-Hispanic white | NA | 2328 (85.3) | 1574 (81.6) |
| Non-Hispanic black | NA | 105 (3.9) | 86 (4.5) |
| Asian | 4856 (100) | 83 (3.0) | 117 (6.1) |
| Hispanic | NA | 149 (5.5) | 114 (5.9) |
| Other | NA | 64 (2.4) | 38 (2.0) |
| Education [n (%)] | |||
| <High school | 2258 (46.5) | 21 (0.77) | 93 (4.8) |
| High school | 1830 (37.7) | 390 (14.3) | 426 (22.1) |
| Some college/technical school | 432 (8.9) | 829 (30.4) | 717 (37.2) |
| ≥College graduate | 336 (6.9) | 1489 (54.6) | 693 (35.9) |
| Stage [n (%)] | |||
| I | 1679 (34.6) | 1060 (38.8) | 915 (47.4) |
| II | 2208 (45.5) | 1252 (45.9) | 804 (41.7) |
| III | 739 (15.2) | 417 (15.3) | 210 (10.9) |
| Missing | 230 (4.7) | 0 | 0 |
| ER/PR status [n (%)] | |||
| ER+/PR+ | 2431 (50.1) | 1692 (62.0) | 1302 (67.5) |
| ER+/PR− | 631 (13.0) | 324 (11.9) | 281 (14.6) |
| ER−/PR+ | 359 (7.4) | 111 (4.1) | 36 (1.9) |
| ER−/PR− | 1341 (27.6) | 545 (20.0) | 300 (15.6) |
| Missing | 94 (1.9) | 57 (2.1) | 10 (0.52) |
| Menopausal status [n (%)]3 | |||
| Premenopausal | 2374 (48.9) | 1387 (50.8) | 410 (21.3) |
| Postmenopausal | 2482 (51.1) | 1263 (46.3) | 1250 (64.8) |
| Missing/unclear | 0 | 79 (2.9) | 269 (14.0) |
| Chemotherapy [n (%)] | |||
| No | 380 (7.8) | 807 (29.6) | 840 (43.6) |
| Yes | 4476 (92.2) | 1922 (70.4) | 1089 (56.5) |
| Radiotherapy [n (%)] | |||
| No | 3269 (67.3) | 1036 (38.0) | 720 (37.3) |
| Yes | 1587 (32.7) | 1693 (62.0) | 1209 (62.7) |
| Hormonal therapy [n (%)] | |||
| No | 2298 (47.3) | 850 (31.2) | 388 (20.1) |
| Yes | 2542 (52.4) | 1879 (68.9) | 1541 (79.9) |
| Missing | 16 (0.3) | 0 | 0 |
| Tamoxifen use among women with ER+ breast cancer [n (%)] | |||
| No | 991 (32.1) | 377 (18.6) | 137 (8.7) |
| Yes | 2077 (67.6) | 1655 (81.5) | 1446 (91.4) |
| Missing | 6 (0.2) | 0 | 0 |
| BMI [n (%)] | |||
| <18.5 kg/m2 | 141 (2.9) | 28 (1.0) | 19 (1.0) |
| 18.5–24.99 kg/m2 | 3004 (61.9) | 1160 (42.5) | 713 (37.0) |
| 25.0–29.99 kg/m2 | 1436 (29.6) | 838 (30.7) | 639 (33.1) |
| ≥30 kg/m2 | 275 (5.7) | 703 (25.8) | 513 (26.6) |
| Missing | 0 | 0 | 45 (2.3) |
| Physical activity [n (%)] | |||
| None | 1720 (35.4) | 379 (13.9) | 282 (14.6) |
| T1 (MET-h/wk) | 841 (17.3) | 463 (17.0) | 377 (19.5) |
| T2 (MET-h/wk) | 1589 (32.7) | 938 (34.4) | 609 (31.6) |
| T3 (MET-h/wk) | 706 (14.5) | 893 (32.7) | 655 (34.0) |
| Missing | 0 | 56 (2.1) | 6 (0.3) |
| Smoking status [n (%)] | |||
| Never | 4730 (97.4) | 1458 (53.4) | 1032 (53.5) |
| Past | 94 (1.9) | 1128 (41.3) | 762 (39.5) |
| Current | 32 (0.7) | 126 (4.6) | 135 (7.0) |
| Missing | 0 | 17 (0.6) | 0 |
| Cruciferous vegetable intake, g/d [n (%)] | |||
| Q1 | 1214 (25.0) | 699 (25.6) | 476 (24.7) |
| Q2 | 1213 (25.0) | 672 (24.6) | 463 (24.0) |
| Q3 | 1213 (25.0) | 675 (24.7) | 467 (24.2) |
| Q4 | 1216 (25.1) | 683 (25.0) | 466 (24.1) |
| Missing | 0 | 0 | 57 (3.0) |
| Parity [n (%)] | |||
| Nulliparous | 197 (4.1) | 615 (22.5) | 318 (16.5) |
| 1 birth | 953 (19.6) | 410 (15.0) | 229 (11.9) |
| 2 births | 1610 (33.2) | 947 (34.7) | 612 (31.7) |
| 3 births | 1122 (23.1) | 486 (17.8) | 391 (20.3) |
| ≥4 births | 974 (20.1) | 256 (9.4) | 379 (19.7) |
| Missing | 0 | 15 (0.6) | 0 |
| Isoflavones (mg/d) | |||
| Baseline | 45.9 ± 38.3 | 2.6 ± 7.9 | 4.1 ± 11.9 |
| Mean time since diagnosis (mo) | 6.5 (3–11) | 23.4 (1–48) | 22.5 (11–38) |
ER, estrogen receptor; ER–, estrogen receptor–negative; ER+, estrogen receptor–positive; LACE, Life After Cancer Epidemiology Study; MET-h, metabolic equivalent task hours; NA, not applicable; PR, progesterone receptor; PR–, progesterone receptor–negative; PR+, progesterone receptor–positive; Q, quartile; RCT, randomized controlled clinical trial; SBCSS, Shanghai Breast Cancer Survival Study; T, tertile; WHEL, Women's Healthy Eating and Living Study.
Mean ± SD (all such values).
Menopausal status at diagnosis for WHEL and LACE; within 6 mo of diagnosis for SBCSS.
We investigated potential effect modification of the associations of baseline isoflavone intake and breast cancer outcomes by ER status, menopausal status, BMI, and tamoxifen use. Multiplicative interactions were tested by using the −2 log likelihood ratio test statistic, comparing models with and without the interaction terms. The proportional hazards assumption was tested by examining interaction terms for each covariate and survival time. The assumption of proportional hazards was violated for 3 covariates (stage, ER/PR status, and radiotherapy use); hence, these 3 variables were treated as time-dependent variables in final Cox models. All analyses were performed by using SAS (version 9.2; SAS Institute). Tests of statistical significance were 2-sided, and P values <0.05 were considered statistically significant.
RESULTS
After a mean follow-up of 7.4 y, 1171 deaths and 1348 recurrences were documented. Most deaths were attributed to breast cancer (77.8%), 7.7% to cardiovascular disease, 6.2% to other malignancies, and 8.4% to other causes. Cohort characteristics and baseline participant characteristics, including postdiagnosis soy isoflavone intake, are shown by study in Table 1. The mean (±SD) postdiagnosis isoflavone intake was substantially higher for Chinese women (45.9 ± 38.3 mg/d) than for US women (3.2 ± 9.8 mg/d).
The sociodemographic and lifestyle factors by soy isoflavone intake among US and Chinese breast cancer survivors, separately, are shown in Table 2. Both US and Chinese breast cancer survivors with higher isoflavone intakes were more likely to exercise and consume more cruciferous vegetables. US women with high isoflavone intakes were more likely to have a lower BMI, be highly educated, and be current nonsmokers. Chinese women with high isoflavone intakes were more likely to have a higher BMI, but neither smoking status nor education was associated with isoflavone intake.
TABLE 2.
Baseline sociodemographic characteristics and lifestyle factors by dietary soy isoflavone intake by country: After Breast Cancer Pooling Project (n = 9514)1
| Quintile of soy isoflavone intake (mg/d) | ||||||||||
| China | United States | |||||||||
| Q1 | Q2 | Q3 | Q4 | Q5 | Q1 | Q2 | Q3 | Q4 | Q5 | |
| Age (y) | 54.5 ± 10.42 | 53.5 ± 10.0 | 53.6 ± 10.2 | 52.8 ± 9.6 | 52.8 ± 9.73 | 54.9 ± 10.7 | 55.6 ± 10.7 | 55.0 ± 10.5 | 53.4 ± 10.2 | 52.7 ± 9.64 |
| Education (%) | ||||||||||
| <High school | 49.8 | 47.0 | 45.9 | 45.6 | 44.2 | 4.3 | 2.7 | 2.9 | 1.4 | 1.0 |
| High school | 34.7 | 37.9 | 38.4 | 36.7 | 40.7 | 26.6 | 20.1 | 18.1 | 11.5 | 11.3 |
| Some college/technical school | 9.6 | 8.5 | 8.4 | 9.7 | 8.2 | 37.2 | 36.4 | 34.9 | 31.3 | 26.1 |
| ≥College graduate | 5.9 | 6.6 | 7.2 | 8.0 | 6.9 | 31.9 | 40.8 | 44.1 | 55.8 | 61.74 |
| Menopausal status (%) | ||||||||||
| Postmenopausal | 54.2 | 51.2 | 51.0 | 49.8 | 49.3 | 57.9 | 62.1 | 57.6 | 50.1 | 46.74 |
| BMI (%) | ||||||||||
| <18.5 kg/m2 | 3.9 | 3.4 | 2.1 | 3.2 | 2.0 | 0.4 | 0.8 | 1.0 | 0.9 | 2.1 |
| 18.5–24.99 kg/m2 | 63.1 | 64.5 | 62.8 | 62.8 | 56.1 | 35.2 | 37.2 | 38.6 | 42.3 | 49.8 |
| 25.0–29.99 kg/m2 | 27.4 | 28.0 | 29.4 | 28.7 | 34.4 | 31.9 | 31.7 | 33.1 | 32.9 | 30.5 |
| ≥30 kg/m2 | 5.6 | 4.1 | 5.8 | 5.4 | 7.54 | 32.5 | 30.4 | 27.4 | 23.9 | 17.74 |
| Physical activity (MET-h/wk) | 5.8 ± 9.4 | 6.0 ± 8.4 | 7.9 ± 10.8 | 7.8 ± 9.7 | 9.8 ± 12.54 | 11.0 ± 14.0 | 12.2 ± 16.4 | 14.1 ± 16.8 | 16.9 ± 17.2 | 19.1 ± 18.14 |
| Cruciferous vegetable intake (g/d) | 50.9 ± 36.1 | 55.5 ± 37.0 | 60.1 ± 35.7 | 66.6 ± 44.1 | 78.6 ± 51.34 | 23.8 ± 29.7 | 29.6 ± 35.0 | 32.6 ± 33.5 | 39.4 ± 43.5 | 51.9 ± 56.54 |
| Smoking status (%) | ||||||||||
| Never | 97.8 | 97.5 | 97.6 | 96.8 | 97.2 | 51.7 | 54.6 | 52.2 | 53.9 | 55.9 |
| Past | 1.7 | 1.7 | 1.8 | 2.6 | 2.1 | 39.3 | 38.9 | 41.4 | 42.6 | 41.4 |
| Current | 0.5 | 0.8 | 0.6 | 0.6 | 0.7 | 9.1 | 6.6 | 6.4 | 3.5 | 2.74 |
The table excludes missing data by characteristic. P values were derived by using a chi-square test for categorical variables and by using the Kruskal-Wallis test for continuous variables. MET-h, metabolic equivalent task hours; Q, quintile.
Mean ± SD (all such values).
P < 0.05.
P < 0.01.
Multivariable-adjusted HRs for isoflavone intake in association with breast cancer outcomes from a pooled analysis of individual cohort data (tests for heterogeneity were not statistically significant) are shown in Table 3. With the use of common cutoffs, consumption of ≥10 mg isoflavones/d was associated with a nonsignificant reduced risk of all-cause mortality (HR: 0.87; 95% CI: 0.70, 1.10) and breast cancer–specific mortality (HR: 0.83; 95% CI: 0.64, 1.07) and with a statistically significant reduced risk of recurrence (HR: 0.75; 95% CI: 0.61, 0.92). With the use of study-specific deciles, HRs for isoflavone intake in association with all-cause mortality ranged from 0.78 for the second highest decile to 0.82 for the highest decile, compared with the lowest decile; however, most point estimates were not statistically significant. This suggested inverse association was slightly stronger for breast cancer–specific mortality, and, for women in the highest decile of isoflavone intake, compared with the lowest decile, risk of breast cancer death was reduced by 29% (HR: 0.71; 95% CI: 0.52, 0.97). Higher intake of isoflavones was associated with a reduced risk of recurrence; women in the highest decile had a 36% reduced risk of recurrence (HR: 0.64; 95% CI: 0.50, 0.82) compared with the lowest decile. Results from sensitivity analyses excluding new breast primaries or limited to distant recurrences were similar.
TABLE 3.
Multivariable HRs for postdiagnosis baseline isoflavone intake in association with all-cause mortality, breast cancer–specific mortality, and breast cancer recurrence: After Breast Cancer Pooling Project (n = 9514)1
| All-cause mortality | Breast cancer–specific mortality | Breast cancer recurrence | |||||
| Isoflavones | Cohort | Events | HR (95% CI) | Events | HR (95% CI) | Events | HR (95% CI) |
| n | n | n | n | ||||
| Study-specific deciles | |||||||
| 1 | 949 | 140 | 1.00 (reference) | 108 | 1.00 (reference) | 165 | 1.00 (reference) |
| 2 | 952 | 113 | 0.78 (0.60, 1.00) | 82 | 0.74 (0.56, 0.99) | 123 | 0.72 (0.57, 0.91) |
| 3 | 953 | 126 | 0.91 (0.71, 1.17) | 86 | 0.80 (0.60, 1.07) | 140 | 0.86 (0.68, 1.08) |
| 4 | 950 | 113 | 0.78 (0.60, 1.00) | 79 | 0.69 (0.52, 0.93) | 133 | 0.75 (0.60, 0.95) |
| 5 | 951 | 118 | 0.83 (0.65, 1.07) | 90 | 0.81 (0.61, 1.07) | 120 | 0.69 (0.54, 0.88) |
| 6 | 953 | 119 | 0.89 (0.69, 1.14) | 92 | 0.86 (0.65, 1.15) | 134 | 0.78 (0.62, 0.99) |
| 7 | 952 | 117 | 0.83 (0.65, 1.07) | 88 | 0.77 (0.58, 1.03) | 138 | 0.77 (0.61, 0.99) |
| 8 | 952 | 118 | 0.94 (0.73, 1.21) | 92 | 0.86 (0.65, 1.15) | 139 | 0.83 (0.66, 1.05) |
| 9 | 953 | 105 | 0.85 (0.65, 1.10) | 89 | 0.87 (0.65, 1.16) | 142 | 0.85 (0.67, 1.07) |
| 10 | 949 | 102 | 0.82 (0.63, 1.08) | 75 | 0.71 (0.52, 0.97) | 114 | 0.64 (0.50, 0.82) |
| P-trend | 0.6354 | 0.4974 | 0.0759 | ||||
| Common cutoffs (mg/d)2 | |||||||
| <4.0 | 4,131 | 635 | 1.00 (reference) | 431 | 1.00 (reference) | 730 | 1.00 (reference) |
| 4.0–9.99 | 613 | 73 | 1.04 (0.80, 1.36) | 60 | 1.09 (0.81, 1.48) | 92 | 0.99 (0.78, 1.25) |
| ≥10.0 | 4,770 | 463 | 0.87 (0.70, 1.10) | 390 | 0.83 (0.64, 1.07) | 526 | 0.75 (0.61, 0.92) |
HRs were derived from delayed-entry Cox proportional hazards regression models and adjusted for age at diagnosis, estrogen receptor/progesterone receptor status, TNM stage, chemotherapy, radiotherapy, hormonal therapy, smoking, BMI, exercise, cruciferous vegetable intake, parity, menopausal status, study, race-ethnicity, and education.
Based on the distribution of daily isoflavone intake in the US cohorts (4.0 mg is approximately the 85th percentile, and 10.0 mg is the 90th percentile); 10.0 mg is the approximate cutoff for the lowest decile of isoflavone intake in the Chinese cohort.
We examined the association of isoflavone intake by country (China, n = 4856; United States, n = 4658) (Table 4). In China, consumption of ≥10 mg/d was associated with a marginally significant reduced risk of recurrence (HR: 0.69; 95% CI: 0.47, 1.01; P = 0.06), compared with <4 mg/d. In the United States, consumption of ≥10 mg/d was associated with a statistically significant reduced risk of recurrence (HR: 0.76; 95% CI: 0.58, 0.99), compared with <4 mg/d. For the US cohorts, similar results were found after the exclusion of 200 Asian American women (Table 4). Note that although 90% of Chinese breast cancer patients consumed ≥10 mg soy isoflavones/d, the inverse association between isoflavone intake and recurrence was also apparent when study-specific deciles were applied (<9.4, 9.4–<16.6, 16.6–<23.0, 23.0–<29.7, 29.7–<36.8, 36.8–<45.4, 45.4–<56.1, 56.1–<69.5, 69.5–<92.6, and ≥92.6 mg/d). The multivariable HRs (95% CIs) across increasing deciles (compared with the lowest decile) were 0.68 (0.40, 0.99), 0.67 (0.46, 0.98), 0.64 (0.44, 0.93), 0.61 (0.41, 0.90), 0.82 (0.58, 1.16), 0.70 (0.49, 1.01), 0.63 (0.44, 0.92), 0.82 (0.57, 1.17), and 0.58 (0.40, 0.86).
TABLE 4.
Multivariable HRs for postdiagnosis baseline isoflavone intake in association with all-cause mortality, breast cancer–specific mortality, and breast cancer recurrence by country: After Breast Cancer Pooling Project (n = 9514)1
| All-cause mortality | Breast cancer–specific mortality | Breast cancer recurrence | |||||
| Isoflavones (mg/d) | Cohort | Events | HR (95% CI) | Events | HR (95% CI) | Events | HR (95% CI) |
| Shanghai, China (SBCSS, n = 4856) | |||||||
| <4.0 | 198 | 20 | 1.00 (reference) | 19 | 1.00 (reference) | 28 | 1.00 (reference) |
| 4.0–10.0 | 328 | 36 | 1.07 (0.62, 1.86) | 32 | 0.98 (0.56, 1.75) | 41 | 0.91 (0.56, 1.47) |
| ≥10.0 | 4330 | 410 | 0.84 (0.54, 1.33) | 354 | 0.75 (0.47, 1.20) | 464 | 0.69 (0.47, 1.01) |
| US cohorts (LACE and WHEL) | |||||||
| All US women (n = 4658) | |||||||
| <4.0 | 3933 | 615 | 1.00 (reference) | 412 | 1.00 (reference) | 702 | 1.00 (reference) |
| 4.0–9.99 | 285 | 37 | 1.02 (0.73, 1.42) | 28 | 1.10 (0.74, 1.62) | 51 | 1.03 (0.77, 1.37) |
| ≥10.0 | 440 | 53 | 0.93 (0.69, 1.24) | 36 | 0.84 (0.59, 1.19) | 62 | 0.76 (0.58, 0.99) |
| Non-Asian US Women (n = 4458) | |||||||
| <4.0 | 3810 | 607 | 1.00 (reference) | 408 | 1.00 (reference) | 692 | 1.00 (reference) |
| 4.0–9.99 | 250 | 34 | 1.00 (0.71, 1.42) | 26 | 1.06 (0.71, 1.59) | 41 | 0.87 (0.63, 1.20) |
| ≥10.0 | 398 | 48 | 0.89 (0.66, 1.20) | 33 | 0.80 (0.55, 1.15) | 57 | 0.74 (0.56, 0.97) |
HRs were derived from delayed-entry Cox proportional hazards regression models and adjusted for age at diagnosis, estrogen receptor/progesterone receptor status, TNM stage, chemotherapy, radiotherapy, hormonal therapy, smoking, BMI, exercise, cruciferous vegetable intake, parity, menopausal status, study, race-ethnicity (where applicable), and education. Baseline isoflavone intake based on the distribution of daily isoflavone intake in the US cohorts (4.0 mg is approximately the 85th percentile, and 10.0 mg is the 90th percentile); 10.0 mg is the approximate cutoff for the lowest decile of isoflavone intake in the Chinese cohort. LACE, Life After Cancer Epidemiology Study; SBCSS, Shanghai Breast Cancer Survival Study; WHEL, Women's Healthy Eating and Living Study.
As shown in Table 5, the inverse association for isoflavones and recurrence appeared to be slightly stronger among women with ER-negative breast cancers [HR for intake of ≥10 mg/d (compared with <4 mg/d): 0.64; 95% CI: 0.44, 0.94]. Among postmenopausal women, the HR for intake of ≥10 mg/d (compared with <4 mg/d) was as follows: 0.64 (95% CI: 0.48, 0.87). However, tests for interaction were not statistically significant.
TABLE 5.
Multivariable HRs for postdiagnosis baseline isoflavone intake by ER status, menopausal status, and tamoxifen use in association with all-cause mortality, breast cancer–specific mortality, and breast cancer recurrence: After Breast Cancer Pooling Project (n = 9514)1
| Isoflavones (mg/d) | All-cause mortality | Breast cancer–specific mortality | Breast cancer recurrence | |||
| Event | HR (95% CI) | Event | HR (95% CI) | Event | HR (95% CI) | |
| Stratified by ER status2 | ||||||
| ER+ | ||||||
| <4.0 | 488 | 1.00 (reference) | 307 | 1.00 (reference) | 537 | 1.00 (reference) |
| 4.0–9.99 | 52 | 1.10 (0.81, 1.49) | 39 | 1.17 (0.82, 1.68) | 63 | 1.02 (0.77, 1.35) |
| ≥10.0 | 231 | 0.91 (0.69, 1.20) | 188 | 0.93 (0.67, 1.28) | 285 | 0.81 (0.63, 1.04) |
| P-trend | 0.54 | 0.69 | 0.11 | |||
| ER− | ||||||
| <4.0 | 142 | 1.00 (reference) | 119 | 1.00 (reference) | 185 | 1.00 (reference) |
| 4.0–9.99 | 20 | 0.85 (0.50, 1.46) | 20 | 0.85 (0.49, 1.48) | 28 | 0.86 (0.54, 1.36) |
| ≥10.0 | 215 | 0.81 (0.54, 1.23) | 187 | 0.67 (0.43, 1.05) | 226 | 0.64 (0.44, 0.94) |
| P-trend | 0.35 | 0.07 | 0.02 | |||
| P-interaction | <0.01 | 0.06 | 0.12 | |||
| Stratified by menopausal status3 | ||||||
| Premenopausal | ||||||
| <4.0 | 185 | 1.00 (reference) | 170 | 1.00 (reference) | 295 | 1.00 (reference) |
| 4.0–9.99 | 28 | 1.21 (0.78, 1.87) | 24 | 1.03 (0.64, 1.65) | 37 | 0.91 (0.63, 1.32) |
| ≥10.0 | 206 | 1.11 (0.77, 1.60) | 188 | 0.97 (0.66, 1.43) | 257 | 0.93 (0.69, 1.26) |
| P-trend | 0.59 | 0.88 | 0.64 | |||
| Postmenopausal | ||||||
| <4.0 | 409 | 1.00 (reference) | 233 | 1.00 (reference) | 381 | 1.00 (reference) |
| 4.0–9.99 | 42 | 1.04 (0.73, 1.48) | 33 | 1.21 (0.79, 1.84) | 48 | 1.06 (0.76, 1.48) |
| ≥10.0 | 255 | 0.84 (0.61, 1.14) | 201 | 0.78 (0.54, 1.14) | 266 | 0.64 (0.48, 0.87) |
| P-trend | 0.26 | 0.16 | <0.01 | |||
| P-interaction | 0.11 | 0.46 | 0.40 | |||
| Among cases with ER+ breast cancer4 | ||||||
| No tamoxifen use | ||||||
| <4.0 | 66 | 1.00 (reference) | 42 | 1.00 (reference) | 89 | 1.00 (reference) |
| 4.0–9.99 | 15 | 1.37 (0.76, 2.46) | 12 | 1.54 (0.78, 3.02) | 15 | 0.99 (0.56, 1.75) |
| ≥10.0 | 82 | 0.98 (0.65, 1.47) | 69 | 1.16 (0.71, 1.90) | 100 | 0.79 (0.55, 1.14) |
| Tamoxifen use | ||||||
| <4.0 | 422 | 0.83 (0.63, 1.09) | 265 | 0.96 (0.68, 1.35) | 448 | 0.78 (0.62, 0.99) |
| 4.0–9.99 | 37 | 0.85 (0.56, 1.29) | 27 | 1.05 (0.63, 1.74) | 48 | 0.80 (0.56, 1.16) |
| ≥10.0 | 149 | 0.74 (0.52, 1.07) | 119 | 0.84 (0.54, 1.31) | 184 | 0.63 (0.46, 0.87) |
| P-interaction | 0.68 | 0.44 | 0.99 | |||
HRs were derived from delayed-entry Cox proportional hazards regression models and adjusted for the following covariates (as applicable): age at diagnosis, estrogen receptor/progesterone receptor status, TNM stage, chemotherapy, radiotherapy, hormonal therapy, smoking, BMI, exercise, cruciferous vegetable intake, parity, menopausal status, study, race-ethnicity, and education. Based on the distribution of daily isoflavone intake in the US cohorts (4.0 mg is approximately the 85th percentile, and 10.0 mg is the 90th percentile); 10.0 mg is the approximate cutoff for the lowest decile of isoflavone intake in the Chinese cohort. ER, estrogen receptor; ER–, estrogen receptor–negative; ER+, estrogen receptor–positive.
Excludes women with missing ER status (n = 123).
Excludes women with missing or unclear menopausal status (n = 348).
Among women with ER+ tumors (n = 6690). These results are from a joint-effects analysis, where the reference group is women who did not receive tamoxifen therapy and consumed <4 mg soy isoflavones/d.
We investigated the joint effects of tamoxifen use and soy isoflavones on breast cancer outcomes (Table 5). Among women with ER-positive tumors, tamoxifen users with higher isoflavone intake had a reduced risk of breast cancer recurrence, compared with women who did not use tamoxifen and consumed the lowest amount of isoflavones. Specifically, women who consumed ≥10 mg isoflavones/d (compared with <4.0 mg/d) had a 37% reduced risk of recurrence (HR: 0.63; 95% CI: 0.46, 0.87). The inverse association among nonusers of tamoxifen was not significant. However, the test for multiplicative interaction was not statistically significant. Associations of isoflavones and breast cancer outcomes were not modified by BMI (data not shown).
We conducted an analysis by using available, updated isoflavone data for the SBCSS (2 additional measures at a mean of 1.5 and 3 y after diagnosis), WHEL (2 additional measures at a mean of 3 and 6 y after diagnosis), and LACE cohorts (1 additional measure at a mean of 7 y after diagnosis), treating isoflavone intake as time-dependent variables in Cox regression models. Heterogeneity in associations by study was found for recurrence (P = 0.003), but not mortality (P value = 0.914); hence, the pooled analysis was conducted with study-specific HRs with the use of inverse-variance weights in random-effects models (31). Use of the updated data showed that the point estimates for the association of soy isoflavones and breast cancer outcomes were generally similar to those based on baseline data, but were not statistically significant (data not shown).
DISCUSSION
In this pooled analysis of 9514 breast cancer survivors from both the United States and China, we found that soy food consumption after diagnosis, in the amount of ≥10 mg soy isoflavones/d, was associated with a nonsignificant reduced risk of all-cause and breast cancer-specific mortality, and a statistically significant reduced risk of recurrence. Analyses using study-specific deciles showed that women in the highest decile of soy isoflavone intake (compared with the lowest decile) had a 29% statistically significant reduced risk of breast cancer–specific mortality and a 36% reduced risk of recurrence. The inverse association between soy food intake and recurrence was seen among Chinese, US, and US non-Asian breast cancer survivors.
Two meta-analyses of soy food and risk of primary breast cancer have suggested that the protective effect of soy food may be limited to Asian women (5, 6). Asian women typically consume substantially more soy food throughout their lifetime than do non-Asian women (34, 35). This difference was observed in the ABCCP, in which US women consumed a mean of 3.2 mg isoflavones/d, whereas Chinese women consumed a mean of 45.9 mg isoflavones/d.
In addition to large differences in soy food consumption, we also found that the associations of sociodemographic and select lifestyle factors with soy food consumption differ between US and Chinese women. Specifically, whereas a higher intake of isoflavones was associated with regular exercise and a higher consumption of cruciferous vegetables among breast cancer survivors from both the United States and China, a higher intake of isoflavones was associated with lower BMI and nonsmoking among US breast cancer survivors only. Despite these differences, we found that the inverse association of soy food intake and recurrence was present among both US and Chinese women. In the US cohorts, the association remained after Asian American women were excluded. Furthermore, we found that the effect of soy food on breast cancer outcomes was consistent regardless of the use of study-specific decile cutoffs or common cutoffs.
In analyses stratified by ER status, we found that the statistically significant inverse association of soy isoflavone intake and recurrence appeared only among women with ER-negative breast cancer. One potential explanation for our finding is that the protective effect is more apparent among women with ER-negative breast cancer because they have a poorer prognosis than do women with ER-positive breast cancer (36, 37). However, note that the interaction was not statistically significant.
Data from laboratory studies suggest that soy isoflavones may interact with tamoxifen therapy and potentially reduce or even inhibit the anticancer effects of tamoxifen on breast tissue (11, 12), which has raised concern regarding soy food consumption among women who take tamoxifen. We found a statistically significant inverse association for isoflavone intake and recurrence only among tamoxifen users. The multiplicative interaction between tamoxifen use and isoflavone intake, however, was not statistically significant. The lack of a statistically significant association among nonusers of tamoxifen may have been because of the smaller sample size in this subgroup. Note that the proportion of Chinese women who received tamoxifen was lower (68% in the SBCSS) than that of US cohorts (82% used tamoxifen in WHEL and 91% used tamoxifen in LACE)—a difference most likely attributed to different clinical practices and the cost of adjuvant therapy. Regardless, our findings do not support the hypothesis that soy food interferes with the efficacy of tamoxifen therapy.
A key strength of our study was that rather than conducting a meta-analysis based on published available statistics, we had access to individual data from members of all 3 cohorts. This allowed for a comprehensive evaluation of soy food intake in association with all-cause mortality, breast cancer–specific mortality, and breast cancer recurrence [previous reports did not include all-cause mortality (14) or breast cancer–specific mortality (14–16)] across a wide exposure range by using standardized statistical methods, common categorization, and standardized covariate adjustments by cohort and country. Furthermore, the large sample size allowed for evaluation of potential interactions between soy food and menopausal status, tumor characteristics, and tamoxifen use with greater statistical power than has been possible for individual cohorts.
Limitations of this study should be considered. First, as noted above, a higher intake of isoflavones was associated with lifestyle-related factors, including regular exercise and higher consumption of cruciferous vegetables (among all survivors) and lower BMI and nonsmoking status (among US survivors only). We carefully adjusted for all the abovementioned factors in our analyses. However, we cannot rule out the possibility of residual confounding as a result of unmeasured factors associated with a healthy lifestyle. Nevertheless, the inconsistent patterns of lifestyle factors associated with soy food intake between US and Chinese women would only introduce noise and thus lower the statistical power of this study. Second, the timing of study enrollment and soy food assessment differed across cohorts. For the LACE and WHEL cohorts, enrollment and baseline soy food assessment occurred within a mean of 2 y after diagnosis (range: 1–48 mo). In the SBCSS, patients were enrolled within a mean of 6 mo after diagnosis (range: 3–11 mo). To adjust for varying times of enrollment and exposure assessment, delayed-entry Cox models were used (38). In the SBCSS, soy food intake at baseline was more likely to be influenced by cancer treatment and, therefore, to be lower than subsequent measures made 18 and 36 mo after diagnosis; hence, the association of soy food and breast cancer outcomes was underestimated for the Chinese women in the ABCPP. Third, we lacked information on length of tamoxifen therapy, preventing a more in-depth analysis of the joint effects of soy food and tamoxifen use. Fourth, few women used aromatase inhibitors (8%) because most of the diagnoses were made before 2005 (39); hence, the role of this therapy in our findings could not be evaluated. Fifth, whereas the SBCSS used an FFQ that has been shown to have good validity and reliability in assessing usual soy food intake, the validity of the soy food questions on the FFQs used in the US studies have not been evaluated. It is possible that the differences in the estimated absolute amount of soy food across the studies in the ABCPP may have been due, at least in part, to the differences in questionnaire design. However, we found no evidence for heterogeneity between the studies in analyses using study-specific or common cutoffs, and the results were comparable. Sixth, given the increasing trend of using soy protein powder in commercially available foods in the United States (40), misclassification of isoflavone intake may have attenuated the results. Future studies with biomarkers to estimate dietary isoflavones are needed. Finally, it is important to note that WHEL consisted of cases who originally participated in a dietary intervention, which reported a significant change in diet for the intervention compared with the control group (eg, increased vegetables and fiber), although no effect of the intervention on recurrence was found (29). We adjusted for intervention status, which did not change our findings (data not shown).
In conclusion, in the largest study to date on the influence of soy food on breast cancer outcomes, and the first study of combined data on US and Chinese women, we found that postdiagnosis soy food intake equivalent to ≥10 mg isoflavones/d was associated with a nonsignificant reduced risk of all-cause and breast cancer–specific mortality and a statistically significant reduced risk of breast cancer recurrence among both US and Chinese breast cancer survivors.
Acknowledgments
We thank the participants and research staff of the Shanghai Breast Cancer Survival Study, the Life After Cancer Epidemiology Study, and the Women's Healthy Eating & Living Study and Bethanie Rammer for assistance with the preparation of this manuscript.
The authors’ responsibilities were as follows—XOS, BJC, WYC, and JPP: conceived and designed the study; SJN, ZC, BJC, WYC, MLK, SWF, WL, YZ, WZ, and XOS: collected and assembled the data; SJN and ZC: performed the statistical analysis; SJN, BJC, WYC, ZC, MLK, SWF, WZ, JPP, and XOS: helped interpret the data; SJN, BJC, WYC, MLK, SWF, WZ, and XOS: drafted the manuscript; and XOS: 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. All authors reviewed the manuscript and approved the final version. The authors had no conflicts of interest to disclose.
Footnotes
Abbreviations used: ABCPP, After Breast Cancer Pooling Project; ER, estrogen receptor; FFQ, food-frequency questionnaire; LACE, Life After Cancer Epidemiology; PR, progesterone receptor; SBCSS, Shanghai Breast Cancer Survival Study; WHEL, Women's Healthy Eating & Living.
REFERENCES
- 1.Setchell KD, Cassidy A. Dietary isoflavones: biological effects and relevance to human health. J Nutr 1999;129:758S–67S [DOI] [PubMed] [Google Scholar]
- 2.Messina MJ, Loprinzi CL. Soy for breast cancer survivors: a critical review of the literature. J Nutr 2001;131:3095S–108S [DOI] [PubMed] [Google Scholar]
- 3.Hilakivi-Clarke L, Andrade JE, Helferich W. Is soy consumption good or bad for the breast? J Nutr 2010;140:2326S–34S [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Trock BJ, Hilakivi-Clarke L, Clarke R. Meta-analysis of soy intake and breast cancer risk. J Natl Cancer Inst 2006;98:459–71 [DOI] [PubMed] [Google Scholar]
- 5.Wu AH, Yu MC, Tseng CC, Pike MC. Epidemiology of soy exposures and breast cancer risk. Br J Cancer 2008;98:9–14 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Dong JY, Qin LQ. Soy isoflavones consumption and risk of breast cancer incidence or recurrence: a meta-analysis of prospective studies. Breast Cancer Res Treat 2011;125:315–23 [DOI] [PubMed] [Google Scholar]
- 7.Taylor CK, Levy RM, Elliott JC, Burnett BP. The effect of genistein aglycone on cancer and cancer risk: a review of in vitro, preclinical, and clinical studies. Nutr Rev 2009;67:398–415 [DOI] [PubMed] [Google Scholar]
- 8.Velentzis LS, Woodside JV, Cantwell MM, Leathem AJ, Keshtgar MR. Do phytoestrogens reduce the risk of breast cancer and breast cancer recurrence? What clinicians need to know. Eur J Cancer 2008;44:1799–806 [DOI] [PubMed] [Google Scholar]
- 9.Constantinou AI, Lantvit D, Hawthorne M, Xu X, van Breemen RB, Pezzuto JM. Chemopreventive effects of soy protein and purified soy isoflavones on DMBA-induced mammary tumors in female Sprague-Dawley rats. Nutr Cancer 2001;41:75–81 [DOI] [PubMed] [Google Scholar]
- 10.Tanos V, Brzezinski A, Drize O, Strauss N, Peretz T. Synergistic inhibitory effects of genistein and tamoxifen on human dysplastic and malignant epithelial breast cells in vitro. Eur J Obstet Gynecol Reprod Biol 2002;102:188–94 [DOI] [PubMed] [Google Scholar]
- 11.Ju YH, Doerge DR, Allred KF, Allred CD, Helferich WG. Dietary genistein negates the inhibitory effect of tamoxifen on growth of estrogen-dependent human breast cancer (MCF-7) cells implanted in athymic mice. Cancer Res 2002;62:2474–7 [PubMed] [Google Scholar]
- 12.Liu B, Edgerton S, Yang X, Kim A, Ordonez-Ercan D, Mason T, Alvarez K, McKimmey C, Liu N, Thor A. Low-dose dietary phytoestrogen abrogates tamoxifen-associated mammary tumor prevention. Cancer Res 2005;65:879–86 [PubMed] [Google Scholar]
- 13.Messina M, McCaskill-Stevens W, Lampe JW. Addressing the soy and breast cancer relationship: review, commentary, and workshop proceedings. J Natl Cancer Inst 2006;98:1275–84 [DOI] [PubMed] [Google Scholar]
- 14.Guha N, Kwan ML, Quesenberry CP, Jr, Weltzien EK, Castillo AL, Caan BJ. Soy isoflavones and risk of cancer recurrence in a cohort of breast cancer survivors: the Life After Cancer Epidemiology study. Breast Cancer Res Treat 2009;118:395–405 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Caan BJ, Natarajan L, Parker B, Gold EB, Thomson C, Newman V, Rock CL, Pu M, Al-Delaimy W, Pierce JP. Soy food consumption and breast cancer prognosis. Cancer Epidemiol Biomarkers Prev 2011;20:854–8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Shu XO, Zheng Y, Cai H, Gu K, Chen Z, Zheng W, Lu W. Soy food intake and breast cancer survival. JAMA 2009;302:2437–43 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Nechuta SJ, Caan BJ, Chen WY, Flatt SW, Lu W, Patterson RE, Poole EM, Kwan ML, Chen Z, Weltzien E, et al. The After Breast Cancer Pooling Project: rationale, methodology, and breast cancer survivor characteristics. Cancer Causes Control 2011;22:1319–31 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Caan B, Sternfeld B, Gunderson E, Coates A, Quesenberry C, Slattery ML. Life After Cancer Epidemiology (LACE) study: a cohort of early stage breast cancer survivors (United states). Cancer Causes Control 2005;16:545–56 [DOI] [PubMed] [Google Scholar]
- 19.Pierce JP, Faerber S, Wright FA, Rock CL, Newman V, Flatt SW, Kealey S, Jones VE, Caan BJ, Gold EB, et al. A randomized trial of the effect of a plant-based dietary pattern on additional breast cancer events and survival: the Women's Healthy Eating and Living (WHEL) Study. Control Clin Trials 2002;23:728–56 [DOI] [PubMed] [Google Scholar]
- 20.Colditz GA, Hankinson SE. The Nurses’ Health Study: lifestyle and health among women. Nat Rev Cancer 2005;5:388–96 [DOI] [PubMed] [Google Scholar]
- 21.Shu XO, Yang G, Jin F, Liu D, Kushi L, Wen W, Gao YT, Zheng W. Validity and reproducibility of the food frequency questionnaire used in the Shanghai Women's Health Study. Eur J Clin Nutr 2004;58:17–23 [DOI] [PubMed] [Google Scholar]
- 22.Block G, Hartman AM, Dresser CM, Carroll MD, Gannon J, Gardner L. A data-based approach to diet questionnaire design and testing. Am J Epidemiol 1986;124:453–69 [DOI] [PubMed] [Google Scholar]
- 23.Yang YX, Wang GY, Pan XC. Chinese food composition tables 2002. Beijing, China: China Beijing University Medical Press, 2002 [Google Scholar]
- 24.Horn-Ross PL, Barnes S, Lee M, Coward L, Mandel JE, Koo J, John EM, Smith M. Assessing phytoestrogen exposure in epidemiologic studies: development of a database (United States). Cancer Causes Control 2000;11:289–98 [DOI] [PubMed] [Google Scholar]
- 25.Patterson RE, Kristal AR, Tinker LF, Carter RA, Bolton MP, Agurs-Collins T. Measurement characteristics of the Women's Health Initiative food frequency questionnaire. Ann Epidemiol 1999;9:178–87 [DOI] [PubMed] [Google Scholar]
- 26.Martínez ME, Marshall JR, Graver E, Whitacre RC, Woolf K, Ritenbaugh C, Alberts DS. Reliability and validity of a self-administered food frequency questionnaire in a chemoprevention trial of adenoma recurrence. Cancer Epidemiol Biomarkers Prev 1999;8:941–6 [PubMed] [Google Scholar]
- 27.Thomson CA, Giuliano A, Rock CL, Ritenbaugh CK, Flatt SW, Faerber S, Newman V, Caan B, Graver E, Hartz V, et al. Measuring dietary change in a diet intervention trial: comparing food frequency questionnaire and dietary recalls. Am J Epidemiol 2003;157:754–62 [DOI] [PubMed] [Google Scholar]
- 28.Phillips KA, Milne RL, Buys S, Friedlander ML, Ward JH, McCredie MRE, Giles GG, Hopper JL. Agreement between self-reported breast cancer treatment and medical records in a population-based breast cancer family registry. J Clin Oncol 2005;23:4679–86 [DOI] [PubMed] [Google Scholar]
- 29.Pierce JP, Natarajan L, Caan BJ, Parker BA, Greenberg ER, Flatt SW, Rock CL, Kealey S, Al-Delaimy WK, Bardwell WA, et al. Influence of a diet very high in vegetables, fruit, and fiber and low in fat on prognosis following treatment for breast cancer: the Women's Healthy Eating and Living (WHEL) randomized trial. JAMA 2007;298:289–98 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986;7:177–88 [DOI] [PubMed] [Google Scholar]
- 31.Smith-Warner SA, Spiegelman D, Ritz J, Albanes D, Beeson WL, Bernstein L, Berrino F, van den Brandt PA, Buring JE, Cho E, et al. Methods for pooling results of epidemiologic studies: the Pooling Project of Prospective Studies of Diet and Cancer. Am J Epidemiol 2006;163:1053–64 [DOI] [PubMed] [Google Scholar]
- 32.Ainsworth BE, Haskell WL, Leon AS, Jacobs DR, Jr, Montoye HJ, Sallis JF, Paffenbarger RS., Jr Compendium of physical activities: classification of energy costs of human physical activities. Med Sci Sports Exerc 1993;25:71–80 [DOI] [PubMed] [Google Scholar]
- 33.WHO Expert Consultation. Appropriate body-mass index for Asian populations and its implications for policy and intervention strategies. Lancet 2004;363:157–63 (Published erratum appears in Lancet 2004;363:902.) [DOI] [PubMed] [Google Scholar]
- 34.Messina M, Nagata C, Wu AH. Estimated Asian adult soy protein and isoflavone intakes. Nutr Cancer 2006;55:1–12 [DOI] [PubMed] [Google Scholar]
- 35.Chun OK, Chung SJ, Song WO. Urinary isoflavones and their metabolites validate the dietary isoflavone intakes in US adults. J Am Diet Assoc 2009;109:245–54 [DOI] [PubMed] [Google Scholar]
- 36.Dent R, Trudeau M, Pritchard KI, Hanna WM, Kahn HK, Sawka CA, Lickley LA, Rawlinson E, Sun P, Narod SA. Triple-negative breast cancer: clinical features and patterns of recurrence. Clin Cancer Res 2007;13:4429–34 [DOI] [PubMed] [Google Scholar]
- 37.Dawson SJ, Provenzano E, Caldas C. Triple negative breast cancers: clinical and prognostic implications. Eur J Cancer 2009;45(suppl 1):27–40 [DOI] [PubMed] [Google Scholar]
- 38.Allison P. Survival analysis using SAS: a practical guide. Cary, NC: SAS Institute, 1995 [Google Scholar]
- 39.Mauriac L. Aromatase inhibitors: effective endocrine therapy in the early adjuvant setting for postmenopausal women with hormone-responsive breast cancer. Best Pract Res Clin Endocrinol Metab 2006;20:S15–29 [Google Scholar]
- 40.Messina M, Messina V. Increasing use of soyfoods and their potential role in cancer prevention. J Am Diet Assoc 1991;91:836–40 [PubMed] [Google Scholar]