Folate, Vitamin B₆, and Vitamin B₁₂ Intake and the Risk of Breast Cancer Among Mexican Women

Various dietary factors have been hypothesized to be determinants of breast cancer, but few have been unequivocally associated with the disease. In contrast with most known risk factors for breast cancer, dietary factors are potentially modifiable, making their identification essential. The incidence of breast cancer in Mexico is still relatively low as compared with Western countries (1). However, recent changes in reproductive patterns, diet, and life-style are likely to contribute to increasing breast cancer incidence in Mexico. From 1979 to 2000, age-standardized mortality almost doubled from 6.4 to 12.2 deaths per 100,000 women (2).

Low vegetable and fruit intake (97% of Mexican women consume <400 g/d or less than four servings per day) and increasing consumption of processed foods may account for the high prevalence of micronutrient deficiencies observed in Mexican women (3). According to the 1999 National Nutrition Survey, median folate intake among Mexican women was 221 μg/d and median vitamin B₁₂ intake was 1.6 μg/d, both close to half of the U.S. recommended dietary allowance (4). Based on a subsample of 461 Mexican women from the National Nutrition Survey, 35% of Mexican women had serum folate levels below the normal 3 ng/mL threshold, and 17% had vitamin B₁₂ serum levels below the reference range.³

Inadequate folate intake has been associated with several cancers and low levels of serum folate, vitamin B₆, and vitamin B₁₂ have been associated with increased breast cancer risk (5, 6). These vitamins all participate as coenzymes in the synthesis of purines and thymidylate for DNA synthesis. Altered DNA methylation, disruption of DNA integrity, and interference with DNA repair are hypothesized mechanisms by which imbalances in folate and other B vitamins may influence nucleic acid metabolism and participate in carcinogenesis (7, 8). Given the high prevalence of vitamin deficiencies in the Mexican population, we conducted a case-control study among women living in Mexico City to examine dietary intakes of folate, vitamin B₆, and vitamin B₁₂ in relation to the risk for breast cancer.

Participants in this study were enrolled in a population-based case-control study to assess the relationships of dietary and reproductive factors with breast cancer risk among residents of Mexico City (9). Cases that had been resident for at least 1 year in metropolitan Mexico City were recruited using a network of six hospitals that are part of the two major healthcare providers in Mexico City, the social security system and the Ministry of Health. These hospitals provided medical care to 80% of breast cancer cases reported to the Mexico City Tumor Registry. From 1990 to 1995, incident confirmed breast cancer cases without previous treatment were identified among women ages 20 to 75 years attending gynecologic clinics for the biopsy of a breast lump. Only women whose biopsy histologically confirmed the diagnosis of breast cancer were eligible. A total of 537 cases were identified. Of those, 88% (n = 475) agreed to participate and provided dietary information. Controls were an age-stratified random sample of metropolitan Mexico City residents. Households were first randomly selected using the National Household Sampling Frame. Study personnel ascertained whether the selected units contained a woman in the intended age group and then visited these households and determined willingness to be interviewed and to provide a blood sample. Only one eligible control was included per household. From 1,534 eligible controls, 90% (n = 1,391) agreed to participate in the study.

Interviewers administered a questionnaire on sociodemographic variables and potential risk factors for breast cancer including reproductive and lactation history and diet. Cases were interviewed at the gynecologic clinics of the study hospitals before they were aware of the diagnosis; the lag time between interview and biopsy was between 1 and 3 months. Controls were interviewed in their homes. Self-reported height and weight have proven unreliable in this population; we therefore sent participants to their health center to obtain actual measurements. These were available for only 226 cases (48% of all cases) and 699 controls (50% of all controls) that attended the health center.

We used a dietary questionnaire developed by Willett et al. (10), which was adapted to the Mexican population (11). We first identified relevant foods using a dietary survey from Mexico City residents. We then contacted a group of Mexican dietitians to include additional food items and made a pilot study in a small sample of women before conducting the validation study. Our modified questionnaire included 104 items and 10 multiple choice frequency categories of consumption: 6 or more per day, 4 to 5 per day, 2 to 3 per day, 1 per day, 5 to 6 per week, 2 to 4 per week, 1 per week, 2 to 3 per month, 1 per month, or less and never. For each food in the questionnaire, a commonly used unit or portion size (specified serving size: slice, glass, or natural unit such as one apple) was specified, and women were asked how often on average over the previous year they had consumed that amount of each food. Nutrient intakes were computed by multiplying the frequency response by the nutrient content of specified portion sizes using a program developed at the National Institute of Public Health, Mexico. The database for calculating the nutrient intakes used information from the U.S. Department of Agriculture food composition tables (12) complemented, when necessary, by a nutrient database developed by the Mexican National Institute of Nutrition (13). This questionnaire has been validated against 16 24-hour recalls among a sample of 134 women in Mexico City comparable to our present population, and has been shown to perform well. Correlations between the food frequency questionnaires and dietary records for total energy carbohydrate, protein, and total fat intakes were 0.52, 0.57, 0.32, and 0.63, respectively (11).

We used χ² tests for categorical variables and t tests or Kruskal-Wallis tests for continuous variables for comparisons between cases and controls. Folate, vitamin B₆, and vitamin B₁₂ intakes were adjusted for total energy intake using the regression-residual method (14). These variables were categorized as quartiles based on the distribution among the control group, and odds ratios (OR) for breast cancer were determined by comparison with the lowest quartile. Socioeconomic status was defined using an index for the Mexican population developed by Bronfman et al. (15). This variable was categorized as tertiles to define low, medium, and high socioeconomic status. Logistic regression was used to calculate OR and their 95% confidence intervals (CI) which were used to estimate relative risks. Crude models were age-adjusted. For multivariate analyses, we further controlled for socioeconomic status, family history of breast cancer, parity, total caloric intake, dietary fiber, carbohydrate, and polyunsaturated fat. In addition, because women for whom body mass index (BMI) was available may differ from those for whom it was not, we included an indicator variable in the models for availability of BMI [BMI = weight (kg) / height² (m)]. Dietary fiber, carbohydrates, and polyunsaturated fat were considered confounders because in a previous study they seemed to be strong predictors of breast cancer and were associated with intake of the micronutrients of interest (16). To test for linear trend, the quartile median value for dietary folate, vitamin B₁₂, and vitamin B₆ was assigned to each subject in that quartile and these values were used as a continuous variable. Spearman correlation coefficients were used to evaluate correlations between nutrients. Because BMI is a predictor of breast cancer (17), analyses were repeated in the subsample of women who had BMI data, with and without controlling for BMI in addition to the variables previously mentioned and compared with the results. Data were stratified by menopausal status because the effect of dietary intake of vitamins may be modified by hormonal status, and postmenopausal breast cancer is thought to be more susceptible to environmental exposures (18). Information on type of menopause (natural or surgical) was obtained by questionnaire. Natural menopause was defined as 12 consecutive months of amenorrhea without an obvious cause. Family history of breast cancer may modify nutrient intake and breast cancer risk (19). Similarly, because vitamins B₆ and B₁₂ serve as important cofactors in folate metabolism and their availability for metabolic processing may affect folate metabolites, we evaluated whether the observed association with folate intake differed by family history of breast cancer and by levels of vitamin B₆ and vitamin B₁₂ intake. Intakes of vitamins B₆ and B₁₂ were categorized by the median intake of controls. Log-likelihood tests were used to evaluate interaction terms. We used SAS statistical software (version 9.0, SAS Institute, Inc., Cary, NC) for data analysis. All tests of statistical significance were two-sided.

Cases were 475 women with histologically confirmed breast cancer, 189 of which occurred in premenopausal and 286 in postmenopausal women. A total of 1,391 women frequency-matched on 5-year strata served as controls. Table 1 presents the demographic, gynecologic, and nutritional characteristics of cases as compared with their controls. Cases were older, had a higher socioeconomic status, more often had a family history of breast, were more often nulliparous, had less children, and were less often premenopausal than controls. Total caloric and carbohydrate intake were higher among cases, whereas their dietary fiber intake seemed to be lower than controls. For the whole population, median folate intake was 310 μg/d (5th-95th percentile, 132-604), vitamin B₆ intake was 1.27 mg/d (5th-95th percentile, 0.65-2.15), and vitamin B₁₂ intake was 4.58 μg/d (5th-95th percentile, 1.29-9.48).

For all women, greater intakes of folate, vitamin B₆, and vitamin B₁₂ were associated with lower risk of breast cancer (Table 2). After adjusting for age, socioeconomic status, family history of breast cancer, menopausal status, parity, availability of BMI, total caloric intake, dietary fiber, animal fat intake, total carbohydrate, and polyunsaturated fat intake; the ORs for the highest quartile of intake compared with the lowest (OR_Q1-Q4) were 0.64 (95% CI, 0.45-0.90) for folate intake, 0.84 (95% CI, 0.61-1.18) for vitamin B₆ intake, and 0.32 (95% CI, 0.22-0.49) for vitamin B₁₂ intake. However, a significant inverse trend was observed only between folate intake (P, test for trend = 0.009) and vitamin B₁₂ intake (P, test for trend ≤0.0001) and breast cancer risk.

Because intakes of folate, vitamin B₆, and vitamin B₁₂ may be correlated; we conducted analyses including two of the nutrients at a time to identify the nutrient responsible for the observed inverse association. When we included both folate and vitamin B₁₂ (r = 0.39) in the multivariate model, the inverse associations previously observed remained present for both nutrients. No significant association was observed with vitamin B₆ after inclusion of the other nutrients in the model one at a time.

The apparent protective effect of folate intake was slightly strengthened after further adjustment one at a time for β-carotene, lutein, total vitamin A, vitamin E, and vitamin D intake and the OR_Q1-Q4 ranged between 0.44 and 0.61. After inclusion of all these nutrients simultaneously in the model, the OR somewhat shifted away from the null (OR_Q1-Q4 = 0.44; 95% CI, 0.29-0.65). For vitamin B₁₂, the OR for extreme quartiles did not appreciably change after adjustment for β-carotene, lutein, total vitamin A, vitamin E, and vitamin D. Therefore, these nutrients are unlikely to explain the association we observed with folate and vitamin B₁₂.

Based on observations made by others, we examined whether the association between these nutrients and breast cancer differed by menopausal status and family history of breast cancer (18-23). Among premenopausal women, intakes of folate and vitamin B₆ were not appreciably associated with breast cancer risk (Table 3). For vitamin B₁₂, the multivariate OR_Q1-Q4 was 0.54 (95% CI, 0.29-0.99; P, test for trend = 0.01) Among postmenopausal women, folate and vitamin B₁₂ intake were associated with a lower risk of breast cancer and this effect seemed to be stronger than among all women combined. For folate intake, the multivariate OR_Q1-Q4 was 0.55 (95% CI, 0.35-0.86; P, test for trend = 0.008; P, for interaction with menopausal status = 0.16). After multivariable adjustment, postmenopausal women in the lowest quartile of vitamin B₁₂ intake had a 4.8 times higher breast cancer risk when compared with women in the highest quartile (OR_Q1-Q4 = 0.21; 95% CI, 0.12-0.37; P, test for trend <0.0001). No association was observed with vitamin B₆ intake. We found no evidence of effect modification after stratifying by family history of breast cancer (folate; P, for interaction = 0.24).

Because vitamin B₆ and vitamin B₁₂ serve as important cofactors in folate metabolism, we evaluated whether the observed association between folate intake differed by levels of vitamin B₁₂ and vitamin B₆ intake (Table 4), as previously reported by Shrubsole et al. (22). Low vitamin B₁₂ intake seemed to reduce the apparent protection in the risk for breast cancer conferred by folate. Among women in the two lowest quartiles of vitamin B₁₂ intake, folate intake was not related to the risk for breast cancer. In contrast, among women in the highest quartile of vitamin B₁₂ intake, we observed a significant inverse trend (P = 0.006) in the risk for breast cancer with increasing folate intake (P, for interaction <0.0001). The joint effect of folate and vitamin B₁₂ comparing women in the highest quartiles of both folate and vitamin B₁₂ intake with those in the lowest quartiles was an OR of 0.16 (95% CI, 0.07-0.41). There was no evidence that vitamin B₆ intake modified the observed association between folate and breast cancer risk.

Because BMI could confound the association between these nutrients and the risk of breast cancer, we repeated these analyses in the subsample of our study population for which information on height and weight was available. Including BMI in the models did not substantially change the apparent protective effect of folate and vitamin B₁₂ intake. Thus, in this population, BMI did not seem to confound the association between vitamin intake and the risk of breast cancer.

In this population-based case-control study in Mexican women, we observed inverse associations between intakes of folate and vitamin B₁₂ and the risk of breast cancer. The observed associations were stronger among postmenopausal women and the inverse relation of folate to breast cancer risk was strengthened among participants with high intakes of vitamin B₁₂. We observed no association with vitamin B₆ intake.

The relation of folate intake to breast cancer risk has been investigated in five large prospective cohorts, most of which have not found an overall association of folate intake and breast cancer risk (20, 21, 24-26). However, in three of these studies, low folate intake seemed to increase the risk of breast cancer among women with high alcohol intake (20, 21, 24). In addition, two case-control studies did not find the association (27, 28). Our study is consistent with various case-control studies (22, 23, 29-32) that have reported a protective relationship and a large nested case-control study using prospectively collected blood that reported a strong inverse association between circulating levels of folate and risk for breast cancer (5). This apparent contradiction may be explained by differences in folate intake between populations. Four of the studies that reported a protective effect were conducted in non-American populations in which vitamin supplementation is unusual (22, 23, 29, 32). In our study, 72% of individuals reported inadequate folate intake (<400 μg/d) and close to 20% had a vitamin B₁₂ intake of less than the recommended dietary allowance of 2.4 μg/d.

Few epidemiologic studies have evaluated vitamin B₆ and vitamin B₁₂ and the risk for breast cancer. The reports are inconsistent (5, 6, 22, 23). In our study, the strongest inverse association was observed with vitamin B₁₂ intake. This association confirms previous observations of vitamin B₁₂ circulating levels (5, 6). We cannot exclude the possibility that the lack of association with vitamin B₆ which we observed might be due to reasons other than biological factors.

The main contributors to folate intake were spinach, maize porridge with milk, orange juice, beans, papaya, and mango. For vitamin B₁₂, the main contributors were maize porridge with milk, liver, red meat, milk, chicken, and sardines. We included these foods in multivariate models to confirm our results and observed that the apparent protective effects between intake of beans, liver, and sardines and breast cancer risk were particularly strong. Strong correlations between micronutrients may be an obstacle in the isolation of effects of specific nutrients, observations made on folate intake may reflect a combined effect with vitamin B₆.

The association of micronutrient intake and breast cancer seems to differ by menopausal status and levels of other micronutrients (18, 22). We observed stronger associations of folate and vitamin B₁₂ with breast cancer among postmenopausal women, which are consistent with other reports on fruit and vegetable intake, micronutrient intake and circulating levels (18, 21-23). Our most striking result is the confirmation that the protection conferred by high folate intake was modified by vitamin B₁₂ intake. A large population-based case-control study in Shanghai first reported a stronger inverse association of folate and breast cancer among women who consumed a high level of vitamin B₁₂ (22). Our study also parallels the Chinese study in the levels of micronutrient intake; median intakes of folate, vitamin B₆, and vitamin B₁₂ are very close to mean intakes reported in that study, whereas they substantially differ from those reported in American populations (5, 25). Hence, a similar variability of exposure in these two populations may explain our ability to detect these associations. Given the strength of the association observed for vitamin B₁₂ in our population, and the paucity of information on this micronutrient and breast cancer risk; these results should be taken with caution.

In the fourth quartile of vitamin B₁₂ intake, an increase in folate intake (from third to fourth quartiles) did not seem to further lower the risk of breast cancer substantially. This suggests the possibility of a threshold effect. Folate absorption in the jejunum, cellular uptake, as well as the metabolic pathway itself are saturable systems such that micronutrient beyond a certain level of intake may have little added effect.

Substantial experimental data have implicated folate in carcinogenesis and several mechanisms have been suggested (7). Folate is a precursor of S-adenosylmethionine, a methyl donor used both for common methylation reactions and for de novo synthesis of thymidilate. This substrate is necessary for both DNA replication and repair. In addition, DNA methylation is thought to play a role in the regulation of gene expression and gene integrity (33-35). DNA demethylation may seem to be affected by folate status: thymidilate is synthesized from deoxyuridilate and requires 5,10-methylene tetrahydrofolate as a coenzyme. Folate depletion induces an imbalance in the availability of nucleotides, uracil misincorporation into DNA, and subsequent abnormal replication and strand breaks (37, 38). Vitamin B₁₂ participates in folate metabolism as a cofactor in the methyl transfer of methyltetrahydrofolate to homocysteine to form methionine. Similar to what is observed with folate depletion, low levels of vitamin B₁₂ reduce methyl group exchange, thus affecting DNA methylation. Synthesis and repair are also affected because folate is “trapped” as methyltetrahydrofolate, preventing tetrahydrofolate from participating in thymidilate synthesis (39).

Bias and confounding must be considered as possible explanations for the observed results. The documentation of established breast cancer risk factors in this study argues against serious bias. Recall bias is always a concern in case-control studies, but lack of awareness among women in this population of possible links between micronutrient intake and the risk for breast cancer should minimize this problem. Recall of the diet before the disease onset could be biased toward current dietary intake, which may change due to the disease (40, 41). We aimed to limit this bias by recruiting incident cases before they knew their diagnosis and at an early state of their disease, thus reducing the likelihood of dietary changes resulting from the diagnosis of cancer. Alcohol intake has been associated with breast cancer risk (42), affects serum vitamin B₁₂ status, and modifies the association of folate with breast cancer risk (5, 19, 24, 43). However, in our study population, alcohol intake was very low, only three individuals consumed >15 g/d; therefore, we could not fully explore its effect on the risk of breast cancer. Oral contraceptive use and hormone replacement therapy were also very low in our population (<3%) and similar among cases and controls. Most (95%) of the women in our study were involved only in housekeeping as physical activity. These variables are therefore unlikely to bias our results.

The strengths of this study include a range of micronutrient intake different from most Western populations. Although the traditional diet of Mexico is likely to provide adequate folate because of regular consumption of beans, an excellent source of this vitamin (44), these traditions are being lost rapidly in urban populations. At the same time, consumption of animal products in Mexico is low so that vitamin B₁₂ intake may be marginal. In addition, food fortification with folate, zinc, iron, vitamins B₁, B₂, and B₁₂ started in 1997 in Mexico, well after data collection, and Mexican women rarely consume vitamin supplements. This facilitates micronutrient intake assessment and limits misclassification of exposure. In a validation study of our questionnaire, conducted in a population similar to that of this study, we observed correlations between the intakes of folate estimated by the food frequency questionnaires and 24-hour recalls (r = 0.22), and vitamin B₆ (r = 0.31) and vitamin B₁₂ (r = 0.29; ref. 10). Even though measurement error remains, it is likely to be nondifferential attenuating the observed associations. A source of bias that is difficult to exclude is that the cases may not be fully representative of the population from which they are derived. The minimal effect on the ORs due to control for socioeconomic status provides some reassurance that selection bias is not substantial. However, these results would require confirmation in a prospective study in Mexican women.

Vitamin deficiencies are common in the Mexican population even with national fortification programs in place (3). Abandonment of traditional foods, increasing intake of processed foods, and low intake of fruits and vegetables will probably worsen this situation. In addition, BMI has dramatically increased in Mexico so that 54% of women in reproductive age are overweight or obese (BMI ≥ 25 kg/m²), a known risk factor for breast cancer (3). Also, the reproductive pattern that was protective against breast cancer is rapidly changing in Mexico as women are having later first pregnancies, fewer children and shorter lactation periods (2). Furthermore, Mexico has the highest reported prevalence of homozygous TT genotype for the 677C → T transition in the methylenetetrahydrofolate reductase (MTHFR) gene (32%; ref. 45). Among individuals with this genotype, low folate intake has been associated with a more substantial increased risk of breast cancer than those with other genotypes (46). All these factors are expected to increase the incidence of breast cancer in Mexican women. However, vitamin deficiency is a potentially modifiable risk factor which can be addressed by health education and expansion of fortification programs. The relation between vitamin intake and breast cancer among Mexican women deserves further evaluation, in particular, the interaction of vitamin intake and genetic polymorphisms affecting their metabolism.

We are indebted to Luisa Sánchez-Zamorano for her support in database management.