For years, the key question surrounding soy intake and breast cancer has been about safety: Does soy increase risk, especially in low-consuming Western populations? Numerous studies have shown safety, however most studies have been in an Asian population. This large, prospective study drawn from the Multiethnic Cohort (MEC) provides a clear and reassuring answer for non-Asian populations.

The fact that the overall finding for the entire cohort was a statistically non-significant null association (HR = 0.96; 95% CI 0.85, 1.08) effectively neutralizes the primary theoretical concern. When you combine five distinct racial/ethnic groups—including those with very low baseline intake (like whites) and those with very high traditional intake (like Japanese Americans)—and observe no overall increased risk, it sends a powerful message of safety.

This study’s strength lies in its applicability. It confirms that a moderate intake of isoflavones is not harmful across a wide spectrum of women living in the US. 

Clinical Bottom Line

This prospective cohort study of over 84,000 women across five ethnic groups found no statistically significant association between estimated dietary isoflavone intake (mainly from soy foods) and overall breast cancer risk in the combined cohort.This finding persisted when analyses were restricted to postmenopausal women, invasive breast cancer cases, or when stratified by estrogen receptor (ER) status .

While there was no significant interaction with ethnicity, exploratory, stratified analyses suggested a non-significant inverse (protective) association of higher intake in Japanese American, Latina, and African American women.Conversely, Native Hawaiian women in the highest intake group had a non-significant increase in risk.However, these stratified findings should be viewed with caution due to smaller sample sizes and the risk of chance findings from multiple comparisons.The study supports meta-analyses showing no protection at low intake levels, but suggests inverse associations in populations with high soy consumption .

Results in Context

Main Results

For the overall cohort, the study found no statistically significant association between the highest quartile of absolute dietary isoflavone intake (upper Q4: 20.3 – 178.7 \text{ mg/d}) compared to the lowest quartile (Q1: 0 – < 3.2 \text{ mg/d}) and overall breast cancer risk .

• Overall Breast Cancer Risk (All Cases): The Hazard Ratio (HR) for the upper Q4 versus Q1 was 0.96 (95% CI 0.85, 1.08) . The P _trend was 0.40, indicating no significant dose-response relationship across quartiles .

• Hazard Ratio (HR): In this context, an HR less than 1 indicates a lower hazard (risk) of developing breast cancer in the higher-intake group compared to the lowest-intake group over the follow-up period.An HR of 0.96 means a 4% lower risk, but the confidence interval crossing 1.0 means this difference is not statistically significant .

• Subgroup Findings (Non-Significant):

  • Japanese American Women: The inverse association (potential protection) with high intake was close to statistical significance (HR for upper Q4 vs. Q1: 0.86; 95% CI 0.70, 1.05; P _trend = 0.06) .

  • Native Hawaiian Women: This group showed a non-significant increase in breast cancer risk with the highest intake (HR for upper Q4 vs. Q1: 1.45; 95% CI 1.02, 2.07; P _trend = 0.10) . The association was positive and significant in the small subset of ER- breast cancer cases in Native Hawaiian women (HR 3.87; 95% CI 1.30, 11.54; P _trend < 0.01) .

  • Latina and African American Women: Also showed non-significant inverse associations .

Participants

The analysis included 84,450 women from the Multiethnic Cohort (MEC) who did not have breast cancer at baseline .

• Cases Identified: 4,769 incident breast cancer cases were identified during a mean follow-up of 12.5 years (median 13.7 years), including 896 in situ and 3,873 invasive cases .

• Ethnic Groups: Japanese Americans, whites, Latinos, African Americans, and Native Hawaiians .

Assertive Critical Appraisal

Limitations & Bias (STROBE Framework)

The study design, a prospective cohort study, is strong for examining exposure-outcome associations over time and minimizes the risk of recall bias common in case-control studies.However, as an observational study, it cannot prove causation, only association .

• Confounding: The authors adjusted for numerous known confounders, including age, BMI, reproductive history, menopausal status, total energy intake, and lifestyle factors.However, unmeasured confounding remains a significant threat to internal validity.Differences in reproductive and other characteristics across isoflavone intake quartiles indicate that women likely differed in unmeasured factors that could confound the relationship .

• Exposure Measurement Error (Major Flaw): Isoflavone intake was estimated using a Food Frequency Questionnaire (FFQ) at baseline.The authors acknowledge that FFQ-related measurement errors likely attenuated the association toward the null.The FFQ did not capture intake from all soy foods (e.g., soy milk, green soybeans, health bars) or from soy used broadly as a component in mixed dishes and processed foods (e.g., canned tuna, white bread) . This is a critical methodological limitation that suggests the true association might be stronger than the observed null finding.The authors themselves suggest that biomarker measures (like urinary isoflavone excretion) may be more accurate for future studies .

• Ethnicity-Specific Effects: The observed inverse association in Japanese American, Latina, and African American women, and the potential increased risk in Native Hawaiian women, suggest important ethnic differences.However, the lack of a statistically significant interaction test and the small sample sizes in the ethnic strata for certain outcomes (like ER- cases) mean these findings could be chance findings from multiple comparisons .

Reporting Quality Assessment (STROBE)

The paper generally adheres to high reporting standards for a cohort study .

• Confounding: The authors clearly describe the efforts to address confounding by including ethnicity as a stratum variable and adjusting for a comprehensive list of covariates in their Cox regression models.The finding that physical activity was not included as a covariate because it “did not materially change the risk estimates” is a transparent reporting practice .

• Data Sources: The cohort and follow-up procedures, including the use of state cancer registries, are adequately described, bolstering the quality of the outcome ascertainment .

Applicability

The study has high generalizability for multiethnic populations in the US, particularly in areas with a broad range of traditional and non-traditional soy food consumption.It includes Japanese American women who reported very high soy intakes, comparable to those in Japan, allowing for robust comparisons across a wide spectrum of exposure.The long follow-up period (mean 13 years) is also a strength.The study’s focus primarily on postmenopausal women (>80% of the cohort) limits its applicability to premenopausal women .

Study Details

Research Objective 🔬

The objective was to examine the association of dietary isoflavone intake with breast cancer risk (in situ and invasive, and by estrogen receptor status) in the Multiethnic Cohort (MEC) .

Study Design 📝

• Design: Prospective Cohort Study .

• Exposure Measurement: Dietary isoflavone intake was estimated in milligrams per day (\text{mg/d}) from a baseline Food Frequency Questionnaire (FFQ) completed between 1993 and 1996.The analysis categorized intake into quartiles (Q1-Q4), with the highest quartile (Q4) being further subdivided .

• Outcome: Incident cases of in situ or invasive breast cancer (ICD-0-3 code C50) .

• Statistical Analysis: Hazard Ratios (HR) and 95% Confidence Intervals (CI) were calculated using Cox proportional-hazards models, stratified by age and adjusted for known confounders .

Setting and Participants 🧑‍🤝‍🧑

• Setting: The Multiethnic Cohort Study (MEC) in Los Angeles County, California, and the state of Hawaii .

• Participants: 84,450 women aged 45–75 years at recruitment, selected from five racial/ethnic populations: Japanese Americans, whites, Latinos, African Americans, and Native Hawaiians .

• Follow-up: Follow-up began at the completion of the baseline questionnaire (or age 45) and continued to the earliest of breast cancer diagnosis, death, or the end of follow-up (31 December 2007) .

Bibliographic Data

• Title: Dietary isoflavone intake is not statistically significantly associated with breast cancer risk in the Multiethnic Cohort

• Authors: Yukiko Morimoto, Gertraud Maskarinec, Song-Yi Park, Reynolette Ettienne, Rayna K. Matsuno, Camonia Long, Alana D. Steffen, Brian E. Henderson, Laurence N. Kolonel, Loïc Le Marchand and Lynne R. Wilkens

• Journal: British Journal of Nutrition

• Year: 2014

• DOI: 10.1017 / \text{S} 0007114514001780