Plasma B-vitamin and one-carbon metabolites and risk of breast cancer before and after folic acid fortification in the United States

Abstract

Prior epidemiologic findings for plasma folate and B-vitamins and breast cancer risk are inconsistent and have not assessed the influence of folic acid fortification. Therefore, we examined the associations of plasma folate, B₁₂, pyridoxal 5′-phosphate (PLP), homocysteine, cysteine and cysteinylglycine with breast cancer risk, before and after fortification. We conducted a nested case–control study within the prospective Nurses’ Health Study. In 1989–1990 (pre-fortification), 32,826 women donated a blood sample and 18,743 donated an additional blood sample in 2000–2001 (post-fortification). Between the first blood collection and 2006, 1874 incident breast cancer cases with at least one blood sample and 367 with two were 1:1 matched to controls. Conditional logistic regression was used to estimate relative risks (RR) and 95% confidence intervals (CI) adjusting for breast cancer risk factors. Overall, higher plasma folate, B₁₂, PLP, homocysteine, cysteine and cysteinylglycine levels were not associated with breast cancer risk. Associations did not vary by in situ/invasive, hormone receptor status, or tumor molecular subtype. Additionally, associations were null before and after fortification. For example, the RR (95% CI) for the highest versus lowest tertile of 1990 (pre-fortification) plasma folate with 1990–2000 follow-up was 0.93 (0.75–1.16) and for the 2000 plasma folate (post-fortification) with 2000–2006 follow-up the RR (95% CI) was 1.17 (0.79–1.74). Plasma folate, B₁₂, PLP, homocysteine, cysteine and cysteinylglycine were not significantly associated with breast cancer overall, before and after fortification, or with specific tumor molecular subtypes. However, long term associations (>8 years) after the implementation of fortification could not be examined.

Introduction

Folate (B₉), cobalamin (B₁₂) and B₆ are essential, water-soluble vitamins found naturally in green leafy vegetables, legumes and organ meats like liver.^1,2 The term folate also includes the synthetic form, folic acid, found in supplements and fortified foods. The United States has mandated that enriched grain products be fortified with folic acid since 1998 to prevent neural tube defects in newborns.³ This has increased average total intakes of folate by approximately 200 to 400 μg/day and led to a 50–100% increase in plasma levels.¹ While fortification has reduced neural tube defects and has had other beneficial outcomes (e.g., reduced preterm birth, stroke),² the potential for adverse effects, such as higher cancer risk, has led to concern.^1,4–6

Folate is involved in one-carbon metabolism, an important metabolic cycle that regulates DNA methylation and synthesis.^1,7–9 Folic acid is reduced to tetrahydrofolate [THF], converted to 5,10-methylene-THF by a B₆-dependent enzyme and then reduced to 5-methyl-THF.^9,10 An enzyme dependent on B₁₂ transfers the methyl group to homocysteine, which is converted through a series of steps to S-adenosyl methionine (SAM).^8,9 5,10-methylene-THF is important for nucleic acid synthesis, and SAM is the primary methyl group donor for DNA methylation.⁹ Thus, excess folate might promote neoplastic lesions through DNA synthesis and methylation and inactivation of tumor suppressor genes.¹

However, folate and other B-vitamins deficiencies may stimulate carcinogenesis through this same cycle. Folate deficiency results in high homocysteine levels and may lead to cancer through reduced synthesis of SAM resulting in DNA hypomethylation¹¹ and/or increased DNA strand breaks and mutations due to impaired DNA repair systems.^1,11–14 B₁₂ deficiency reduces conversion of 5-methyl-THF and therefore lower SAM.^7–9 B₆ deficiency reduces levels of 5,10-methylene-THF, resulting in increased DNA strand breaks and reduced levels of 5-methyl-THF resulting in lower SAM.¹⁵

Several aberrant methylation patterns have been noted in breast tumors including widespread DNA hypomethylation and hypermethylation of tumor suppressor genes suggesting that one-carbon metabolism may play a role in breast cancer carcinogenesis.¹⁰ With concern for carcinogenesis at both low and high folate levels, folic acid fortification has been controversial.^1,6,16,17 To our knowledge, no studies have assessed plasma levels of folate and other B-vitamins both before and after fortification in relation to breast cancer risk. Additionally, few studies have assessed folate levels in relation to molecular subtypes of breast cancer, with most assessing only estrogen receptor (ER), progesterone receptor (PR) and human epidermal growth factor 2 (HER2) separately.^18,19 Therefore, we examined the associations of plasma folate, B₁₂, pyridoxal 5′-phosphate (PLP; the principal active form of B₆), homocysteine, cysteine and cysteinylglycine with breast cancer, including an assessment of nutrient levels before and after fortification in relation to breast cancer subtypes within a nested case–control in the prospective Nurses’ Health Study (NHS).

Materials and Methods

Study population

The NHS is a prospective cohort of 121, 700 female registered nurses aged 30–55 years enrolled in 1976 by completing a mailed baseline questionnaire.^20–22 Biennially, participants complete mailed questionnaires updating health information, including most known breast cancer risk factors and diagnoses. In 1989–1990 (pre-fortification), 32,826 women donated a blood sample.^23,24 The women arranged to have their blood drawn and shipped back overnight with an ice pack, where it was then processed, aliquoted into plasma, red blood cell and white blood cell components, and stored in continuously monitored liquid nitrogen freezers (<−130 °C). A second blood sample from a subset of these women (n = 18,743) was donated in 2000–2001 (post-fortification) using similar protocols.²⁵ Of the 32,826 women who provided at least one blood sample, the follow-up rate in 2006 was 99%. The study was approved by the Committee on the Use of Human Subjects in Research at the Brigham and Women’s Hospital, Boston, Massachusetts; the completion of the self-administered questionnaire and blood collection was considered to imply informed consent.

Case and control selection

Incident cases of breast cancer were self-reported on the biennial questionnaires and confirmed by medical record review. From first blood collection through 2006, there were 1931 incident breast cancer cases (357 carcinoma in situ, 1,544 invasive) among the women without a history of cancer (except non-melanoma skin) that provided at least one blood sample. We excluded 57 case–control pairs with missing information on plasma measures in one member of the pair, resulting in 1874 case–control pairs included in this analysis, of which 367 case–control pairs had two blood samples. The average time from the 1990 blood collection to breast cancer diagnosis was 97 months (range = 1–202) and 28 months (range = 1–73) from the 2000 blood collection to breast cancer diagnosis among those that donated a second blood. Each breast cancer case was matched to one control without a history of cancer, by year of birth (±1 year), time of blood draw (±2 h), fasting status (≥10 h vs. <10 h), month of blood draw (±1 month), use of postmenopausal hormones within 3 months before blood collection, and menopausal status at blood collection and diagnosis.

Laboratory assays

Plasma concentrations of folate, B₁₂, PLP, total homocysteine, total cysteine and cysteinylglycine were assayed at the Jean Mayer United States Department of Agriculture Human Nutrition Research Center on Aging at Tufts University. Folate and vitamin B₁₂ were measured using a radioassay kit (Bio-Rad, Richmond, CA) or microbial assay for folate after 2000.²⁶ For microbial assays, folate and B₁₂ values above the upper limit of quantification, >63 ng/mL for folate (n = 9 for 1st collection; n = 37 for 2nd collection) and > 2014 pg/mL for B₁₂ (n = 19 for 1st collection; n = 21 for 2nd collection), were assigned the values 63 and 2014 respectively. Plasma levels of the active form of B₆, pyridoxal 5′-phosphate (PLP), were determined by an enzymatic procedure using radioactive tyrosine and the apo-enzyme tyrosine.²⁷ Total homocysteine, cysteine and cysteinylglycine were measured using high performance liquid chromatography with fluorescence.²⁸ Cases and controls within matched pairs were assayed in the same analytical run. The laboratory was blinded to case and control status and the order of the specimens in each pair was randomized. Additionally, for case–control pairs diagnosed after 2000 with two blood samples, both blood samples were assayed together in the same analytical run. Assays were conducted in four batches, and quality control samples (~10% of samples) were randomly interspersed to assess laboratory precision. Average within-batch coefficients of variation were less than 10%, except one batch each for folate (12.6%), PLP (13.9%), B₁₂ (12.6%) and homocysteine (13.2%). Outliers were detected and removed using the extreme Studentized deviate many-outlier procedure²⁹ (range of n outliers excluded across plasma measures: n = 2–12 for 1st collection; n = 1–3 for 2nd collection).

Plasma values for folate and B₁₂ differed between 1990 and 2000 blood samples for case–control pairs with two blood samples consistent with fortification and increasing use of supplements. Among the 1990 blood samples, there was evidence of some batch variation. Therefore, for folate and B₁₂ we recalibrated all 1990 blood samples, but not the 2000 blood samples, to an average standard batch for each plasma measure.³⁰ For PLP, homocysteine and cysteinylglycine, since there was no change in the plasma values between 1990 and 2000 blood collection, all batches from 1990 and 2000 blood samples were recalibrated to an average standard batch. Lastly, since the distribution of the fourth batch of cysteine (includes both 1990 and 2000 blood samples) had a lower and narrower range of values than earlier batches, but did not differ between 1990 and 2000 blood samples we recalibrated the three earlier batches to an average standard batch. Using the recalibrated data, we defined quantiles based on the control values for each plasma measure. For cysteine, we created batch specific quantiles for the average of the three earlier batches and for batch four. Secondary analyses excluding the fourth batch of cysteine data did not change conclusions.

Covariate data

Dietary intakes of B-vitamins and alcohol were obtained from food frequency questionnaires (FFQ) administered every four years,³¹ using the 1990 FFQ for 1990 blood collection and the 2002 FFQ for 2000 blood collection. Information on multivitamin and supplement use was obtained from the 1990 and 2000 biennial questionnaires. Information on breast cancer risk factors were collected from the biennial questionnaires and the questionnaires administered at the time of blood collection; age at menarche, height and body mass index (BMI) at age 18 were collected on the baseline questionnaire in 1976, parity and age at first birth, family history of breast cancer and history of benign breast disease were updated to 1990 or 2000, and current weight and menopausal status were from the 1990 or 2000 questionnaires. Missing values were filled in using the prior questionnaire or by missing indicators (i.e., parity/age at first birth, weight change from age 18) or using the median of all participants (all other covariates). In multivariate models we adjusted for age at menarche, parity/age at first birth, age at menopause if postmenopausal at blood collection, family history of breast cancer in mother or a sister, history of benign breast disease, height, BMI at age 18, weight change since age 18, and alcohol intake at blood collection. Results were essentially unchanged if a menopause status and BMI interaction term was added to our models and hence models with only main effects for these two covariates are presented here.

Tumor/outcome data

Information on tumor invasiveness, grade and hormone receptor status were available from pathology reports. Described previously,^32,33 archival formalin-fixed paraffin-embedded breast cancer tissue blocks were obtained for approximately 70% of incident primary breast cancer cases through 2006 and used to construct tissue microarrays. Immunohistochemical staining was performed for estrogen receptor (ER), progesterone receptor (PR), human epidermal growth factor 2 (HER2), cytokeratin 5/6 (CK5/6) and epidermal growth factor (EGFR) markers. Luminal A cases were defined as ER-positive and/or PR-positive and HER2-negative and histologic grade 1 or 2. Cases were defined as luminal B if they were either 1) ER-positive and/or PR-positive and HER2-positive or 2) ER-positive and/or PR-positive and HER2-negative and grade 3. HER2-type cases were ER-negative, PR-negative and HER2-positive and basal like cases were ER-negative, PR-negative, HER2-negative and either CK5/6-positive or EGFR-positive.

Statistical analyses

We calculated intraclass correlation coefficients (ICC) and Spearman correlations between 1990 and 2000 blood samples for women with both blood samples. We compared characteristics between cases and controls at both blood collections using the Mantel–Haenszel tests for categorical and paired t-tests for continuous characteristics. Additionally, we examined the correlation between the plasma measures and dietary intakes of the B-vitamins for each blood sample using Spearman correlations among the controls with two blood samples.

For the main analysis using the 1990 blood and follow-up from 1990 to 2006, we calculated relative risks (RR) of incident breast cancer and 95% confidence intervals (CI) for each quintile of plasma measure compared to the lowest quintile using multivariable conditional logistic regression. To test for trend across quintiles, we modeled the median quintile values as a continuous variable and report the Wald P-value. As prior studies have suggested non-linear relationships between B-vitamins and breast cancer, we additionally tested for nonlinearity using cubic splines.

We also assessed in situ, invasive and hormone receptor subtype using quintiles and conditional logistic regression. Due to smaller numbers, we assessed molecular subtype and examined relationship by time point using tertiles and unconditional logistic regression adjusting for matching factors. Tests for subtype heterogeneity were calculated using the methods of Wang et al.³⁴ We also examined associations between plasma measures and breast cancer using plasma measures as a continuous variable [per standard deviation (SD) of natural log—transformed plasma measure].

To assess if associations varied before and after fortification, plasma levels were evaluated as 1) 1990 blood levels with cases diagnosed between 1990 and 2000, 2) 1990 blood levels with cases diagnosed between 2000 and 2006, 3) 2000 blood levels with cases diagnosed between 2000 and 2006. Additionally, to evaluate the 1990 and 2000 plasma measures in the same model, we conducted two analyses. First, we averaged the natural log transformed 1990 and 2000 plasma measures among those with both blood donations. Using the averaged values, in tertile categories and as a 1SD increase continuous value, we used unconditional logistic regression adjusting for the matching factors and breast cancer risk factors at both blood collections to estimate RR and 95% CIs. Second, we conducted a repeated measures logistic regression using PROC GENMOD in SAS, where the 1990 plasma measure was used for the 1990–2000 time period and the 2000 plasma measure was used for the 2000–2006 time period, updating covariates when relevant. Controls from the 2000–2006 case–control pairs were allowed to contribute to both the 1990–2000 and 2000–2006 time periods, but with their 1990 plasma measure used for the 1990–2000 time period and their 2000 plasma measure used for the 2000–2006 time period.

To assess if associations varied across other exposures, we conducted stratified analysis using unconditional logistic regression and to obtain a test of significance we used the multiplicative interaction term of the stratifying variable and continuous plasma measure (per SD of natural log–transformed plasma measure) in the model, and tested via a likelihood ratio test. Secondarily, we also restricted analyses to nonmultivitamin users.

All P-values were based on two-sided tests and were considered statistically significant if p < 0.05. Analyses were conducted using SAS 9.3 (SAS Institute Inc., Cary, NC, USA) for UNIX.

Results

At the first blood collection in 1990, the mean age of the participants was 56.8 years (range = 42.5–70.2) and in 2000, the mean age of the participants was 67.0 years (range = 53.9–80.3). Characteristics by breast cancer case and control status at both blood collections are shown in Table 1. Cases were more likely to have a family history of breast cancer and a history of benign breast disease at the 1990 blood collection. Mean plasma levels at the 2000 blood collection (post-fortification) showed substantial increases in the folate concentrations and smaller increases in B₁₂ and PLP concentrations. Multivitamin use also increased between the two bloods (39% in 1990 and 59% in 2000 among controls).

Table 1.

Characteristics, plasma B vitamins and thiol amino acids of breast cancer cases and their matched controls at two blood collections in the Nurses’ Health Study

Characteristics	At the 1990 blood collection			At the 2000 blood collection
	Breast cancer cases (n = 1874)	Controls (n = 1874)	p Value1	Breast cancer cases (n = 367)	Controls (n = 367)	p Value1
		Mean (SD)			Mean (SD)
Age at blood collection, y	56.7 (7.0)	56.8 (7.0)	Matched	66.9 (6.9)	67.1 (6.9)	Matched
Age at menarche, y	12.5 (1.4)	12.6 (1.4)	0.07	12.6 (1.4)	12.7 (1.4)	0.27
Age at first birth among parous, y	25.3 (3.4)	25.0 (3.1)	0.03	25.0 (3.2)	24.8 (3.2)	0.69
Age at menopause, y2	49.6 (4.3)	49.1 (4.7)	0.001	50.2 (4.1)	49.3 (4.6)	0.002
Parity among parous	3.1 (1.5)	3.2 (1.5)	0.16	3.1 (1.4)	3.1 (1.6)	0.63
BMI at age 18, kg/m2	21.1 (2.7)	21.3 (2.8)	0.03	21.0 (2.8)	21.2 (2.7)	0.36
BMI at blood collection, kg/m2	25.5 (4.5)	25.4 (4.6)	0.33	26.9 (4.9)	26.4 (4.8)	0.18
Alcohol intake at blood collection, g/d3	5.5 (9.7)	5.0 (8.8)	0.09	7.3 (12.7)	6.2 (10.5)	0.19
	%			%
Family history of breast cancer in 1st degree relatives, %	15	10	<0.0001	22	15	0.02
History of benign breast disease, %	54	46	<0.0001	64	61	0.39
Current use of multivitamins, %4	39	39	0.84	65	59	0.09
Postmenopausal women, %	67	67	Matched	98	98	Matched
Recent use of postmenopausal hormones, %2	53	53	Matched	69	69	Matched
Plasma Concentrations	Mean (SD)			Mean (SD)
Folate (ng/mL)	10.7 (8.9)	10.8 (8.8)	0.84	31.3 (17.1)	29.7 (15.9)	0.17
Vitamin B12 (pg/mL)	489 (229)	483 (217)	0.39	624 (358)	634 (351)	0.71
Pyridoxal 5′-phosphate (pmol/mL)	65.4 (78.5)	66.7 (79.8)	0.61	79.0 (92.3)	75.8 (85.2)	0.63
Homocysteine (nmol/mL)	11.5 (3.7)	11.3 (3.4)	0.09	11.0 (3.1)	10.8 (3.0)	0.31
Cysteine (nmol/mL)	275.5 (60.9)	276.7 (61.8)	0.23	239.2 (13.7)	238.2 (13.3)	0.22
Cysteinylglycine (nmol/mL)	193.1 (55.1)	193.0 (53.4)	0.99	192.4 (53.1)	193.0 (48.7)	0.97

Abbreviations: SD = standard deviation; BMI = body mass index; Matched = values were similar because of matching.

¹p Values are from the paired t-test for continuous variable or from the Mantel–Haenszel test for categorical variables. All p values were two-sided.

²Among postmenopausal women only.

³From the 1990 FFQ for the 1990 blood collection or from the 1998 FFQ for the 2000 blood collection.

⁴At the 1990 blood collection (from the 1990 FFQ) or at the 2000 blood collection (from the 2000 biennial questionnaire).

Plasma measures of folate, B₁₂ and PLP were modestly correlated with each other (r = 0.3–0.5), homocysteine was weakly inversely correlated with folate, B₁₂ and PLP, and cysteine was only correlated with homocysteine and cysteinylglycine (Supporting Information Table S1). On average, significant correlations between B-vitamin plasma measures and dietary intakes were 0.1 to 0.5, with the strongest correlations being plasma folate with total intake of folate and plasma PLP with total intake of B₆.

The mean time between the two blood collections was 11.1 years (range = 9.4–13). Plasma measures had variable reproducibility over the approximately 11-year period. The highest ICC values were observed with vitamin B₁₂ (0.51, 95% CI = 0.46–0.56) and cysteinylglycine (0.55, 95% CI = 0.50–0.60), with slightly lower values for PLP (0.47, 95% CI = 0.41–0.52), homocysteine (0.34, 95% CI = 0.28–0.41), and cysteine (0.44, 95% CI = 0.38–0.50). The ICC for folate was not calculated due to the three-fold increase in folate levels between the two blood collections. Spearman correlations over the 11-year interval were higher than ICC’s for all plasma measures: 0.35 for folate, 0.57 for B₁₂, 0.49 for PLP, 0.36 for homocysteine, 0.51 for cysteine and 0.56 for cysteinylglycine.

No significant associations were seen for plasma folate, B₁₂, PLP, homocysteine or cysteine with overall breast cancer risk (Table 2). The RR for the highest versus the lowest quintile ranged from 0.91 (95% CI = 0.73–1.13) for PLP to 1.17 (95% CI = 0.93–1.46) for homocysteine. None of the associations were significant when potential non-linearity was evaluated (not shown). Estimates were similar when we mutually adjusted for the other plasma measures and when we restricted to only those that had fasted ≥8 h prior to blood donation. No significant associations were seen when restricted to in situ or invasive breast cancer cases. None of the plasma B-vitamins or metabolites were significantly associated with breast cancer by ER/PR status. There were no significant associations for plasma folate, B₁₂, PLP, cysteine and cysteinylglycine observed when assessed by molecular subtypes of breast cancer (Table 3). Homocysteine was not significantly associated when assessed by molecular subtypes when modeled in tertiles or in overall heterogeneity tests (P-heterogeneity >0.05). However, when modeled continuously per SD increase in natural log transformed concentration homocysteine was significantly associated with basal-like breast cancer (RR = 1.63, 95% CI = 1.23–2.17).

Table 2.

Relative risks and 95% confidence intervals (CI) of breast cancer by quintiles of 1990 plasma measure for 1990–2006 follow-up

	# Cases	Quintile of plasma measure1					p for trend	p for heterogeneity
		1 (lowest)	2	3	4	5
	Folate, ng/mL		<4.7	4.7 to 6.6	6.7 to 9.7	9.8 to 15.5			≥15.6
Overall	1872	1.00 (ref.)	0.94 (0.77–1.15)	0.81 (0.65–1.00)	0.85 (0.69–1.05)	0.95 (0.77–1.17)	0.81
Insitu	346	1.00 (ref.)	0.72 (0.45–1.13)	0.73 (0.44–1.22)	1.00 (0.61–1.63)	0.83 (0.51–1.33)	0.96	0.37
Invasive	1,502	1.00 (ref.)	1.00 (0.79–1.27)	0.82 (0.64–1.04)	0.81 (0.64–1.03)	0.97 (0.77–1.23)	0.76
ER+/PR+	913	1.00 (ref.)	1.16 (0.85–1.57)	0.81 (0.60–1.11)	0.79 (0.57–1.08)	1.01 (0.74–1.38)	0.69	0.93
ER+/PR−	199	1.00 (ref.)	0.96 (0.48–1.90)	0.76 (0.38–1.52)	0.80 (0.40–1.59)	1.08 (0.55–2.14)	0.68
ER−/PR−	230	1.00 (ref.)	1.03 (0.55–1.92)	1.16 (0.60–2.22)	1.08 (0.57–2.05)	0.97 (0.52–1.81)	0.90
Vitamin B12, pg/mL		<311	311 to 403	404 to 497	498 to 618	≥619
Overall	1862	1.00 (ref.)	1.09 (0.88–1.35)	1.11 (0.91–1.37)	0.92 (0.75–1.15)	1.06 (0.85–1.31)	0.88
Insitu	344	1.00 (ref.)	1.00 (0.61–1.64)	0.85 (0.52–1.39)	0.90 (0.54–1.51)	0.85 (0.51–1.42)	0.49	0.54
Invasive	1,494	1.00 (ref.)	1.08 (0.85–1.38)	1.22 (0.96–1.54)	0.93 (0.73–1.18)	1.11 (0.87–1.41)	0.81
ER+/PR+	909	1.00 (ref.)	1.15 (0.84–1.56)	1.42 (1.05–1.91)	1.06 (0.78–1.45)	1.14 (0.83–1.56)	0.74	0.34
ER+/PR−	199	1.00 (ref.)	0.72 (0.34–1.52)	1.01 (0.50–2.03)	0.94 (0.45–1.96)	0.83 (0.40–1.71)	0.90
ER−/PR−	227	1.00 (ref.)	0.96 (0.49–1.87)	0.57 (0.30–1.11)	0.60 (0.33–1.11)	0.95 (0.51–1.76)	0.63
Pyridoxal 5′-phosphate, pmol/mL		<24.4	24.4 to 35.0	35.1 to 50.0	50.1 to 82.6	≥82.7
Overall	1872	1.00 (ref.)	0.91 (0.74–1.11)	0.93 (0.75–1.15)	0.81 (0.66–1.01)	0.91 (0.73–1.13)	0.56
Insitu	347	1.00 (ref.)	0.86 (0.52–1.41)	0.96 (0.58–1.59)	0.82 (0.47–1.41)	0.93 (0.55–1.57)	0.90	0.99
Invasive	1,501	1.00 (ref.)	0.90 (0.71–1.13)	0.89 (0.70–1.14)	0.81 (0.63–1.03)	0.86 (0.67–1.11)	0.37
ER+/PR+	912	1.00 (ref.)	0.89 (0.66–1.21)	0.91 (0.66–1.25)	0.85 (0.62–1.18)	0.87 (0.62–1.21)	0.56	0.96
ER+/PR−	200	1.00 (ref.)	1.02 (0.49–2.12)	1.20 (0.60–2.42)	0.77 (0.38–1.58)	0.72 (0.35–1.46)	0.18
ER−/PR−	229	1.00 (ref.)	0.89 (0.51–1.55)	0.92 (0.49–1.74)	0.87 (0.47–1.59)	0.86 (0.44–1.69)	0.75
Homocysteine, nmol/mL		<8.6	8.6 to 9.8	9.9 to 11.3	11.4 to 13.3	≥13.4
Overall	1863	1.00 (ref.)	1.11 (0.89–1.38)	0.98 (0.79–1.22)	1.12 (0.90–1.40)	1.17 (0.93–1.46)	0.20
Insitu	345	1.00 (ref.)	0.96 (0.57–1.61)	0.89 (0.55–1.44)	0.96 (0.58–1.57)	0.92 (0.54–1.58)	0.79	0.97
Invasive	1,495	1.00 (ref.)	1.13 (0.89–1.45)	0.98 (0.76–1.26)	1.12 (0.87–1.45)	1.19 (0.92–1.54)	0.22
ER+/PR+	909	1.00 (ref.)	0.92 (0.68–1.25)	0.99 (0.72–1.35)	1.05 (0.76–1.45)	1.23 (0.88–1.71)	0.10	0.45
ER+/PR−	198	1.00 (ref.)	1.46 (0.73–2.92)	0.78 (0.38–1.60)	1.25 (0.58–2.70)	1.37 (0.63–2.98)	0.56
ER−/PR−	230	1.00 (ref.)	1.61 (0.78–3.35)	1.54 (0.73–3.27)	1.24 (0.59–2.61)	1.24 (0.60–2.58)	0.81
Cysteine, nmol/mL2
Overall	1872	1.00 (ref.)	1.18 (0.95–1.47)	1.12 (0.90–1.41)	1.01 (0.79–1.30)	0.97 (0.73–1.30)	0.93
Insitu	347	1.00 (ref.)	1.02 (0.61–1.69)	1.18 (0.69–2.00)	0.95 (0.52–1.75)	1.14 (0.55–2.37)	0.54	0.78
Invasive	1,501	1.00 (ref.)	1.21 (0.95–1.55)	1.09 (0.85–1.41)	0.97 (0.74–1.28)	0.91 (0.66–1.26)	0.65
ER+/PR+	912	1.00 (ref.)	1.42 (1.04–1.94)	1.27 (0.92–1.76)	1.00 (0.69–1.44)	1.03 (0.68–1.56)	0.83	0.03
ER+/PR−	200	1.00 (ref.)	1.67 (0.80–3.49)	0.95 (0.46–1.98)	1.51 (0.68–3.36)	0.80 (0.30–2.13)	0.44
ER−/PR−	229	1.00 (ref.)	0.36 (0.18–0.75)	0.57 (0.27–1.22)	0.65 (0.33–1.29)	0.62 (0.25–1.56)	0.41
Cysteinylglycine, nmol/mL		<174.8	174.8 to 198.9	199.0 to 232.6	232.7 to 515.8	≥515.9
Overall	1846	1.00 (ref.)	0.95 (0.77–1.18)	0.80 (0.64–1.01)	0.93 (0.75–1.17)	0.98 (0.78–1.23)	0.96
Insitu	344	1.00 (ref.)	1.03 (0.62–1.72)	0.80 (0.46–1.40)	1.26 (0.71–2.25)	0.79 (0.45–1.41)	0.57	0.24
Invasive	1,478	1.00 (ref.)	0.93 (0.73–1.18)	0.82 (0.63–1.05)	0.86 (0.67–1.11)	1.03 (0.79–1.33)	0.81
ER+/PR+	902	1.00 (ref.)	0.99 (0.72–1.36)	0.78 (0.56–1.08)	0.87 (0.63–1.21)	0.98 (0.70–1.37)	>0.89	0.89
ER+/PR−	198	1.00 (ref.)	0.73 (0.38–1.39)	0.86 (0.44–1.71)	0.94 (0.47–1.90)	1.05 (0.53–2.08)	0.69
ER−/PR−	227	1.00 (ref.)	0.65 (0.33–1.27)	0.91 (0.44–1.89)	0.75 (0.37–1.51)	1.14 (0.54–2.43)	0.42

¹Adjustments for age at menarche (continuous variable), parity/age at first birth (nulliparous, 1–2 children/first birth <25 years, 1–2 children/first birth 25+ years, ≥3 children/first birth <25 years, ≥3 children/first birth 25+ years), age at menopause (<50, ≥50 to <55, or ≥ 55 years), family history of breast cancer in mother or a sister (yes or no), history of benign breast disease (yes or no), height (continuous), body mass index at 18 (<21, 21 to <23, 23+), weight change from age 18 (continuous) and alcohol intake at blood collection (continuous) in 1990.

²Quintile cut points: <246.9, 246.9 to 273.4, 273.5 to 300.2, 300.3 to 345.0, ≥345.1 Batch 1–3; <222.8, 222.8 to 230.8, 230.9 to 238.3, 238.4 to 245.4, ≥245.5 Batch 4.

Table 3.

Relative risks and 95% confidence intervals (CI) of breast cancer subtypes by tertiles of 1990 plasma measure for 1990–2006 follow-up among those with 1990 blood

	# Cases	Tertile of plasma measure1			P for trend	P for heterogeneity	Per 1-SD increase in natural log-transformed concentrations1
		1 (lowest)	2	3
		Folate, ng/mL		<6.0				6.0 to 11.2	≥11.3
Luminal A	510	1.00 (ref.)	0.87 (0.68–1.12)	1.05 (0.82–1.34)	0.47	0.41	1.01 (0.91–1.12)
Luminal B	182	1.00 (ref.)	0.71 (0.48–1.03)	0.79 (0.54–1.16)	0.37	0.93 (0.79–1.10)
HER 2 Type	50	1.00 (ref.)	0.82 (0.40–1.68)	0.95 (0.47–1.93)	0.99		0.89 (0.66–1.20)
Basal Like	42	1.00 (ref.)	1.47 (0.68–3.17)	1.14 (0.49–2.63)	0.94	0.94 (0.68–1.30)
Vitamin B12, pg/mL		<375	375 to 530	≥531
Luminal A	508	1.00 (ref.)	1.11 (0.87–1.42)	1.04 (0.81–1.33)	0.83	0.81	1.04 (0.94–1.16)
Luminal B	182	1.00 (ref.)	0.94 (0.64–1.36)	0.93 (0.64–1.36)	0.72		1.05 (0.90–1.22)
HER 2 Type	50	1.00 (ref.)	0.83 (0.40–1.75)	1.20 (0.61–2.38)	0.52		1.12 (0.84–1.49)
Basal Like	42	1.00 (ref.)	0.80 (0.37–1.70)	0.72 (0.33–1.54)	0.40		0.86 (0.63–1.17)
Pyridoxal 5′-phosphate, pmol/mL		<31.1	31.1 to 57.7	≥57.8
Luminal A	510	1.00 (ref.)	0.86 (0.68–1.10)	0.88 (0.68–1.12)	0.40	0.45	0.98 (0.88–1.08)
Luminal B	182	1.00 (ref.)	0.90 (0.62–1.30)	0.82 (0.56–1.21)	0.34	0.92 (0.79–1.08)
HER 2 Type	50	1.00 (ref.)	0.60 (0.30–1.20)	0.57 (0.28–1.16)	0.18		0.76 (0.55–1.05)
Basal Like	41	1.00 (ref.)	1.72 (0.77–3.87)	1.66 (0.73–3.78)	0.34	1.02 (0.74–1.40)
Homocysteine, nmol/mL		<9.5	9.5 to 11.8	≥11.9
Luminal A	509	1.00 (ref.)	0.96 (0.75–1.22)	1.02 (0.80–1.31)	0.81	0.29	0.96 (0.86–1.06)
Luminal B	182	1.00 (ref.)	1.08 (0.73–1.60)	1.22 (0.83–1.79)	0.31		1.11 (0.95–1.30)
HER 2 Type	50	1.00 (ref.)	0.51 (0.25–1.05)	0.60 (0.30–1.20)	0.16	0.85 (0.63–1.15)
Basal Like	42	1.00 (ref.)	0.85 (0.36–2.04)	1.56 (0.73–3.33)	0.18		1.63 (1.23–2.17)
Cysteine, nmol/mL2
Luminal A	510	1.00 (ref.)	1.00 (0.79–1.28)	0.86 (0.66–1.11)	0.21	0.57	0.99 (0.89–1.10)
Luminal B	182	1.00 (ref.)	1.23 (0.85–1.80)	0.88 (0.58–1.35)	0.43	1.09 (0.93–1.29)
HER 2 Type	50	1.00 (ref.)	0.84 (0.42–1.68)	0.66 (0.30–1.41)	0.28		0.86 (0.63–1.18)
Basal Like	42	1.00 (ref.)	1.57 (0.67–3.69)	1.84 (0.77–4.38)	0.20	1.24 (0.89–1.72)
Cysteinylglycine, nmol/mL2		<167.3	167.3 to 208.7	≥208.8
Luminal A	505	1.00 (ref.)	0.78 (0.61–1.00)	0.87 (0.68–1.11)	0.27	0.85	0.97 (0.88–1.08)
Luminal B	181	1.00 (ref.)	0.84 (0.57–1.24)	1.06 (0.73–1.54)	0.71		1.02 (0.87–1.19)
HER 2 Type	50	1.00 (ref.)	1.22 (0.60–2.47)	1.13 (0.55–2.34)	0.76	1.04 (0.77–1.41)
Basal Like	42	1.00 (ref.)	0.91 (0.41–2.04)	1.13 (0.53–2.42)	0.73		1.17 (0.84–1.62)

¹Adjustments for matching factors, age at menarche (continuous variable), parity/age at first birth (nulliparous, 1–2 children/first birth <25 years, 1–2 children/first birth 25+ years, ≥3 children/first birth <25 years, ≥3 children/first birth 25+ years), age at menopause (<50, ≥50 to <55, or ≥ 55 years), family history of breast cancer in mother or a sister (yes or no), history of benign breast disease (yes or no), height (continuous), body mass index at 18 (<21, 21 to <23, 23+), weight change from age 18 (continuous) and alcohol intake at blood collection (continuous) in 1990.

²Tertile cut points: <264.4, 264.4 to 313.0, ≥313.1 Batch 1–3; <228.9, 228.9 to 240.9, ≥241.0 Batch 4.

There were no significant associations for plasma folate, B₁₂, homocysteine, cysteine, or cysteinylglycine at 1990 blood collection with cases diagnosed within 10 years of blood collection (proximate), 1990 blood collection with cases diagnosed >10 years of blood collection (distant), or the 2000 blood sample (proximate cases, blood sample post fortification) with invasive breast cancer (Table 4). For plasma PLP, there was a significant inverse association observed for the proximate 1990 blood sample and invasive breast cancer (RR top vs. bottom tertile = 0.79, 95% CI = 0.64–0.97, p-trend = 0.04). However, this inverse association was not seen after fortification (2000 blood collection) or with distant invasive breast cancer cases (1990 blood collection).

Table 4.

Relative risks and 95% confidence intervals (CI) of invasive breast cancer by tertiles of repeated plasma measures

	# Cases	Tertile of plasma measure1	p for trend	Per 1-SD increase in natural log-transformed concentrations1
		1 (lowest)			2	3
		Folate, ng/mL
		<6.0	6.0 to 11.2	≥11.3
1990 blood: 1990–2000 follow-up	1,041	1.00 (ref.)	0.95 (0.77–1.16)	0.93 (0.75–1.16)	0.56	0.92 (0.83–1.01)
2000–2006 follow-up	450	1.00 (ref.)	0.79 (0.57–1.10)	0.97 (0.72–1.31)	0.92	1.03 (0.92–1.15)
		<21.9	21.9 to 31.9	≥32.0
2000 blood: 2000–2006 follow-up	282	1.00 (ref.)	1.14 (0.76–1.71)	1.17 (0.79–1.74)	0.47	1.14 (0.97–1.34)
Vitamin B12, pg/mL
		<375	375 to 530	≥531
1990 blood: 1990–2000 follow-up	1,039	1.00 (ref.)	0.93 (0.75–1.14)	1.01 (0.81–1.24)	0.87	1.05 (0.96–1.16)
2000–2006 follow-up	449	1.00 (ref.)	1.24 (0.90–1.71)	1.01 (0.75–1.35)	0.98	1.03 (0.92–1.15)
		<425	425 to 736	≥737
2000 blood: 2000–2006 follow-up	282	1.00 (ref.)	1.26 (0.86–1.85)	0.87 (0.58–1.32)	0.40	1.02 (0.87–1.20)
Pyridoxal 5′-phosphate, pmol/mL
		<31.1	31.1 to 57.7	≥57.8
1990 blood: 1990–2000 follow-up	1,041	1.00 (ref.)	0.86 (0.70–1.06)	0.79 (0.64–0.97)	0.04	0.92 (0.85–1.00)
2000–2006 follow-up	449	1.00 (ref.)	0.98 (0.72–1.32)	1.15 (0.85–1.57)	0.33	1.08 (0.95–1.22)
2000 blood: 2000–2006 follow-up	282	1.00 (ref.)	0.85 (0.56–1.30)	1.08 (0.72–1.61)	0.46	1.04 (0.88–1.23)
Homocysteine, nmol/mL
		<9.5	9.5 to 11.8	≥11.9
1990 blood: 1990–2000 follow-up	1,037	1.00 (ref.)	0.90 (0.73–1.12)	1.03 (0.84–1.26)	0.67	1.03 (0.95–1.12)
2000–2006 follow-up	449	1.00 (ref.)	0.88 (0.65–1.20)	1.21 (0.88–1.67)	0.20	1.09 (0.95–1.26)
2000 blood: 2000–2006 follow-up	281	1.00 (ref.)	1.08 (0.74–1.59)	1.12 (0.74–1.69)	0.59	1.00 (0.84–1.19)
Cysteine, nmol/mL
		<264.4	264.4 to 313.0	≥313.1
1990 blood: 1990–2000 follow-up	1,042	1.00 (ref.)	0.99 (0.80–1.22)	0.85 (0.68–1.06)	0.12	0.95 (0.86–1.04)
		<228.9	228.9–240.9	≥241.0
2000–2006 follow-up	450	1.00 (ref.)	1.12 (0.83–1.51)	0.95 (0.68–1.34)	0.73	1.08 (0.94–1.23)
2000 blood: 2000–2006 follow-up	282	1.00 (ref.)	1.30 (0.83–2.02)	1.09 (0.70–1.69)	0.85	1.13 (0.95–1.35)
Cysteinylglycine, nmol/mL
		<167.3	167.3 to 208.7	≥208.8
1990 blood: 1990–2000 follow-up	1,018	1.00 (ref.)	0.91 (0.74–1.11)	1.00 (0.81–1.23)	0.98	1.00 (0.91–1.09)
2000–2006 follow-up	450	1.00 (ref.)	0.75 (0.55–1.04)	1.00 (0.74–1.35)	0.92	1.02 (0.90–1.15)
2000 blood: 2000–2006 follow-up	281	1.00 (ref.)	0.74 (0.50–1.09)	0.88 (0.59–1.32)	0.57	0.97 (0.83–1.15)

¹Adjustments for matching factors, age at menarche (continuous variable), parity/age at first birth (nulliparous, 1–2 children/first birth <25 years, 1–2 children/first birth 25+ years, ≥3 children/first birth <25 years, ≥3 children/first birth 25+ years), age at menopause (<50, ≥50 to <55, or ≥ 55 years), family history of breast cancer in mother or a sister (yes or no), history of benign breast disease (yes or no), height (continuous), body mass index at 18 (<21, 21 to <23, 23+), weight change from age 18 (continuous) and alcohol intake at blood collection (continuous) in 1990, or in 2000 dependent on models.

No significant associations were observed for the average of the 1990 and 2000 plasma measures with breast cancer risk; though a borderline positive association was noted for the continuous folate average and breast cancer (Table 5). In updated plasma measure analyses, no association was seen for plasma folate or other plasma measures, with the exception of continuous (per1 SD increase in natural log transformed plasma measure) plasma homocysteine, where an 8% increase in breast cancer risk was seen.

Table 5.

Relative risks and 95% confidence intervals (CI) of invasive breast cancer for models that incorporate both the 1990 and 2000 plasma measures

	Tertile of plasma measure			p for trend	Per 1-SD increase in natural log-transformed concentrations
	1 (lowest)	2	3
	Average of 1990 and 2000 plasma measures1
Folate, ng/mL	<11.6	11.6 to 18.4	≥18.5
	1.00 (ref)	0.79 (0.53, 1.17)	1.38 (0.94, 2.03)	0.03	1.20 (1.00, 1.44)
Vitamin B12, pg/mL	<415	415 to 612	≥613
	1.00 (ref)	1.02 (0.70, 1.49)	0.97 (0.66, 1.41)	0.83	1.03 (0.88, 1.22)
Pyridoxal 5′-phosphate, pg/mL	<33.7	33.7 to 57.7	≥57.8
	1.00 (ref)	0.79 (0.54, 1.15)	1.06 (0.72, 1.54)	0.59	1.06 (0.88, 1.27)
Homocysteine, nmol/mL	<9.5	9.5 to 11.2	≥11.3
	1.00 (ref)	0.80 (0.54, 1.19)	1.17 (0.80, 1.73)	0.34	1.09 (0.88, 1.33)
Cysteine, nmol/mL	<229.6	229.6 to 240.6	≥240.7
	1.00 (ref)	1.36 (0.92, 2.00)	1.15 (0.75, 1.77)	0.53	1.19 (0.97, 1.46)
Cysteinylglycine, nmol/mL	<171.0	171.0 to 210.3	≥210.4
	1.00 (ref)	0.78 (0.54, 1.15)	0.84 (0.57, 1.22)	0.37	0.93 (0.78, 1.11)
Updated plasma measures2
Folate, ng/mL	1.00 (ref)	1.04 (0.88, 1.22)	0.93 (0.79, 1.10)	0.52	0.97 (0.90, 1.04)
Vitamin B12, pg/mL	1.00 (ref)	1.11 (0.94, 1.30	0.95 (0.80, 1.12)	0.49	1.02 (0.94, 1.09)
Pyridoxal 5′-phosphate, pg/mL	1.00 (ref)	0.95 (0.81, 1.12)	0.93 (0.79, 1.10)	0.42	0.98 (0.92, 1.05)
Homocysteine, nmol/mL	1.00 (ref)	0.91 (0.77, 1.08)	1.16 (0.98, 1.37)	0.08	1.08 (1.01, 1.15)
Cysteine, nmol/mL	1.00 (ref)	1.07 (0.91, 1.27)	0.99 (0.83, 1.18)	0.15	0.98 (0.92, 1.06)
Cysteinylglycine, nmol/mL	1.00 (ref)	0.84 (0.71, 0.98)	0.89 (0.76, 1.06)	0.37	0.95 (0.88, 1.02)

¹Unconditional logistic regression using the average of the 1990 and 2000 plasma measures among only those with two blood collections (2000–2006 follow-up; Cases/Controls:367/367) with adjustments for matching factors, age at menarche (continuous variable), parity/age at first birth (nulliparous, 1–2 children/first birth <25 years, 1–2 children/first birth 25+ years, ≥3 children/first birth <25 years, ≥3 children/first birth 25+ years), age at menopause (<50, ≥50 to <55, or ≥ 55 years), family history of breast cancer in mother or a sister (yes or no), history of benign breast disease (yes or no), height (continuous), body mass index at 18 (<21, 21 to <23, 23+), weight change from age 18 (continuous) and alcohol intake at blood collection (continuous) in both 1990 and 2000.

²Repeated measures logistic regression, where 1990 plasma measure used for 1990–2000 time period and the 2000 plasma measure for the 2000–2006 time period (Cases/Controls:1626/1855). Controls from the 2000–2006 case–control pairs were allowed to contribute to both the 1990–2000 and 2000–2006 time periods. Models were adjusted for matching factors, age at menarche (continuous variable), parity/age at first birth (nulliparous, 1–2 children/first birth <25 years, 1–2 children/first birth 25+ years, ≥3 children/first birth <25 years, ≥3 children/first birth 25+ years), age at menopause (<50, ≥50 to <55, or ≥ 55 years), family history of breast cancer in mother or a sister (yes or no), history of benign breast disease (yes or no), height (continuous), body mass index at 18 (<21, 21 to <23, 23+), weight change from age 18 (continuous) and alcohol intake (continuous), updated at time of blood collection when relevant. See Table 4 for blood collection specific tertile cut points.

Associations of folate and other plasma measures (per SD of natural log–transformed plasma measure) did not vary by age at blood donation (Table 6). For plasma folate, B₁₂ and homocysteine, associations varied by smoking status; though no strata reached statistical significance. The B₁₂, PLP and cysteine associations varied significantly by BMI, although only B₁₂ reached significance in stratified analyses. For example, for B₁₂, among those with BMI <25 the relative risk was 0.97 (95% CI = 0.88–1.06) and among those with BMI ≥25 the relative risk was 1.17 (95% CI = 1.05–1.30, p-interaction = <0.0001). For all plasma measures, associations did not vary by alcohol use (p-interaction >0.05 for all). Lastly, associations did not significantly vary when restricted to nonmultivitamin users (data not shown).

Table 6.

Relative risks and 95% confidence intervals (CI) of invasive breast cancer per 1 SD increase in natural log transformed 1990 plasma measure for 1990–2006 follow-up among those with 1990 blood stratified by potential effect modifiers at time of blood

Plasma measure	Cases/ Controls	Per 1-SD increase in natural log-transformed concentrations RR (95% CI)1	Cases/Controls	Per 1-SD increase in natural log-transformed concentrations RR (95% CI)1	p for interaction
	Age at 1990 blood <57 years		Age at 1990 blood ≥57 years
Folate, ng/mL	704/935	0.98 (0.88,1.08)	796/935	0.92 (0.83,1.01)	0.10
Vitamin B12, pg/mL	703/933	0.99 (0.89,1.09)	793/932	1.10 (1.00,1.21)	0.07
Pyridoxal 5′-phosphate, pmol/mL	705/938	0.99 (0.89,1.10)	796/935	0.93 (0.84,1.02)	0.06
Homocysteine, nmol/mL	703/930	1.07 (0.96,1.19)	787/932	1.03 (0.94,1.14)	0.70
Cysteine, nmol/mL	705/936	0.99 (0.89,1.11)	796/935	0.99 (0.89,1.10)	0.62
Cysteinylglycine, nmol/mL	693/920	0.98 (0.89,1.09)	773/919	1.02 (0.93,1.13)	0.28
	Never/Past Smoker at 1990 Blood		Current Smoker at 1990 blood
Folate, ng/mL	1300/1657	0.99 (0.91,1.07)	199/207	0.81 (0.66,1.00)	0.04
Vitamin B12, pg/mL	1296/1652	1.08 (1.00,1.16)	199/207	0.84 (0.69,1.04)	0.003
Pyridoxal 5’-phosphate, pmol/mL	1301/1660	0.98 (0.91,1.06)	199/207	0.93 (0.75,1.14)	0.59
Homocysteine, nmol/mL	1292/1650	1.01 (0.94,1.10)	197/206	1.17 (0.97,1.43)	0.03
Cysteine, nmol/mL	1301/1658	1.00 (0.92,1.08)	199/207	0.93 (0.74,1.18)	0.76
Cysteinylglycine, nmol/mL	1272/1629	0.99 (0.91,1.06)	193/204	1.09 (0.88,1.33)	0.17
	BMI <25 kg/m2		BMI ≥25 kg/m2
Folate, ng/mL	789/1046	0.94 (0.85,1.03)	711/824	0.97 (0.87,1.09)	0.18
Vitamin B12, pg/mL	786/1042	0.97 (0.88,1.06)	710/823	1.17 (1.05,1.30)	<0.0001
Pyridoxal 5’-phosphate, pmol/mL	790/1046	0.91 (0.82,1.00)	711/827	1.05 (0.94,1.17)	0.002
Homocysteine, nmol/mL	781/1042	1.04 (0.94,1.14)	709/820	1.07 (0.96,1.19)	0.56
Cysteine, nmol/mL	790/1046	0.92 (0.84,1.02)	711/825	1.07 (0.95,1.19)	0.0009
Cysteinylglycine, nmol/mL	771/1028	1.04 (0.95,1.15)	695/811	0.96 (0.87,1.07)	0.13
	Alcohol <10 g/day		Alcohol ≥10 g/day
Folate, ng/mL	1212/1535	0.96 (0.88,1.03)	288/335	0.97 (0.81,1.15)	0.81
Vitamin B12, pg/mL	1208/1531	1.03 (0.95,1.11)	288/334	1.12 (0.95,1.33)	0.16
Pyridoxal 5’-phosphate, pmol/mL	1212/1538	0.97 (0.90,1.05)	289/335	0.95 (0.80,1.13)	0.76
Homocysteine, nmol/mL	1206/1529	1.04 (0.96,1.13)	284/333	1.10 (0.93,1.30)	0.30
Cysteine, nmol/mL	1212/1536	0.98 (0.91,1.07)	289/335	1.05 (0.87,1.27)	0.29
Cysteinylglycine, nmol/mL	1188/1509	1.00 (0.93,1.08)	278/330	0.99 (0.83,1.18)	0.99

¹Adjustments for matching factors, age at menarche (continuous variable), parity/age at first birth (nulliparous, 1–2 children/first birth <25 years, 1–2 children/first birth 25+ years, ≥3 children/first birth <25 years, ≥3 children/first birth 25+ years), age at menopause (<50, ≥50 to <55, or ≥ 55 years), family history of breast cancer in mother or a sister (yes or no), history of benign breast disease (yes or no), height (continuous), body mass index at 18 (<21, 21 to <23, 23+), weight change from age 18 (continuous) and alcohol intake at blood collection (continuous) in 1990.

Discussion

Overall, we found that folate, B₁₂, PLP, homocysteine, or cysteine were not associated with breast cancer risk. Additionally, there were no consistent associations when exposure was assessed either before or after fortification, with proximal, distal, or the long term average exposures, or when cases were defined by specific tumor molecular subtypes. There was some suggestion that higher plasma levels of folate and B₁₂, which in turn would lower homocysteine levels, may be beneficial among smokers.

Excluding the prior two studies conducted in the Nurses’ Health Study with follow-up to 1996, our findings are in line with the majority of other prospective studies finding no association of plasma B-vitamins and risk of breast cancer overall. Of seven studies that examined plasma folate, six found no association,^{18,19,35–38} and one found a positive overall association.³⁹ Kim (2016) was a high risk population of BRCA1/2 carriers which could potentially explain the difference.³⁹ Five studies examined plasma B₁₂, all of which found no overall association.^{18,19,35,37,39} PLP was evaluated in four prior studies, of which three found no association,^18,35,39 and one found an inverse overall association.³⁷ Two examined homocysteine and found no association.^35,40 One examined cysteine, finding a positive association with breast cancer.⁴⁰ Additionally, to date, gene variants associated with vitamin B-related nutrient levels^41,42 have not been associated with breast cancer in large GWAS studies.⁴³

The two studies published previously in the NHS have cases that overlap with this analysis.^44,45 Briefly, we previously observed inverse associations for cysteine,⁴⁵ borderline significant inverse associations for folate, B₁₂ and PLP, and no association for homocysteine with breast cancer.⁴⁴ When stratified by menopausal status, inverse associations with PLP were stronger in postmenopausal women, whereas B₁₂ was inversely associated with risk among premenopausal women.⁴⁴ When stratified by alcohol intake, women who consumed at least 15 g alcohol per day had significant inverse associations with plasma folate (RR top vs. bottom = 0.11, 95% CI = 0.02–0.59).⁴⁴ This current analysis expands on the follow-up period with an additional 10 years of follow-up (1990–2006) and an additional 1,162 cases, and we found no association overall for any of the five measures. We did observe inverse associations for PLP for our proximal 1990 blood sample analysis consistent with the prior analysis.

Prior literature on interactions is limited and inconsistent. In several studies, associations did not vary by alcohol intake^18,40 similar to our results; conversely, one found B₁₂ associations were positive among non-drinkers³⁷ and one noting inverse for PLP among those consuming > 7 g/day.³⁷ We observed that among current smokers, higher plasma folate was inversely associated and conversely, homocysteine was suggestive of a positive association with breast cancer risk. B₁₂ had a significant interaction term, but the inverse association in stratified analysis among current smokers was not significant. In the two relatively small studies that assessed smoking,^18,40 no differences in associations were observed by smoking status. Smokers have been shown to have lower systemic levels of folate,^46–49 B₁₂^46,47 and B₆,^47,48 and higher homocysteine⁵⁰ levels than non-smokers. The lower folate and B₆ status persists after accounting for diet^47–49 suggesting that smoking alters the metabolism of folate. Without B₁₂, folate remains in the 5-methyl THF form unable to donate methyl groups or return to the folate cycle.² One study found that smokers had different proportions of folate forms,⁴⁷ with the majority of folate as 5-methyl THF, consistent with the “methyl trap” mechanism. Therefore, it is possible that among smokers having higher folate levels may partially mitigate this compared to the non-smokers.

The main strengths of the study are our ability to examine plasma B-vitamin levels before and after fortification due to the two blood collections and the ability to examine molecular subtypes with the unique tumor tissue resources. Other strengths of the study include its prospective design with overall large sample size, long-term follow-up, high follow-up rates, and accounting for known breast cancer risk factors.

Several of the stratified analyses were limited by small case numbers—particularly the low number of current smokers, and the small number of HER2 and basal breast cancer subtypes. There is also the potential for misclassification as a single blood sample was used to characterize a woman’s long-term levels. However, reported ICCs for women over 1–5 years ranged from 0.50 to 0.65,^51,52 suggesting only moderate misclassification. Additionally, we had two blood samples about 10 years apart for a subset of the women. Lastly, folate was measured using two different methods – cases up until 2000 measured via radioassay and cases after 2000 measured via microbial assay. Microbial assays have a tendency to recover more folate than the radioimmunoassay and are considered the gold standard now; however, the two assays are highly correlated (R² = 0.91).⁵³ As case–control pairs after 2000 had both blood samples measured together, the difference between the 1990 and 2000 blood samples would not be due to the difference in assay methods. Additionally, the average batch recalibration should “adjust” for the difference in assay methods.

As our follow-up after the implementation of folic acid fortification was relatively short, (8 years after fortification implementation) we cannot exclude the possibility of an association between post-fortification plasma folate and breast cancers diagnosed 10–15+ years after fortification was implemented. Further, the number of cases diagnosed after fortification with a post-fortification blood sample may have limited our power to detect an association. Lastly, the metabolites measured were only from the folate cycle and the transulfuration pathway. Metabolites in the methionine cycle (i.e., SAM, S-adenosylhomocysteine [SAH]), which are directly involved in methylation and reflect the methylation potential (i.e., SAM/SAH)⁵⁴ hypothesized to be the link between one-carbon metabolism and carcinogenesis, were not measured.

In summary, folate and other B-vitamins were not associated with breast cancer risk overall. There was some suggestion that higher plasma folate may be beneficial particularly to smokers. However, long term effects of folate fortification cannot be excluded, and additional large prospective studies are needed to confirm subgroup findings.

What’s new?

With concerns of increased cancer risk, there has been discussion as to whether countries should implement folic acid fortification programs. However, few prospective studies have examined plasma folate and other B-vitamins levels and breast cancer risk among populations whose food is fortified with folic acid. This study examines the association of plasma folate and other B-vitamins on breast cancer risk, prior to and after mandatory folic acid fortification of grain products in the U.S. Plasma folate, B₁₂, PLP, homocysteine, cysteine, and cysteinylglycine were not significantly associated with breast cancer overall, before and after fortification, or with specific tumor molecular subtypes.

Abbreviations:

BMI

body mass index

CI

confidence intervals

CK5/6

cytokeratin 5/6

EGFR

epidermal growth factor

ER

estrogen receptor

FFQ

food frequency questionnaires

HER2

human epidermal growth factor 2

ICC

intraclass correlation coefficients

NHS

Nurses’ Health Study

PR

progesterone receptor

RR

relative risks

SAM

S-adenosyl methionine

SD

standard deviation

THF

tetrahydrofolate