Plasma Levels of B Vitamins and Colorectal Cancer Risk: The Multiethnic Cohort Study

The B vitamins, such as folate, vitamin B6, and vitamin B12, play an important role as coenzymes in one-carbon metabolism, which is essential for nucleotide synthesis and DNA methylation (1). Disruption of this metabolic pathway can result in aberrations in DNA methylation, imbalance in DNA precursors, and deficiency in DNA repair, each of which may contribute to carcinogenesis, especially in the colon (1). Indeed, high folate status has been associated in several studies with a reduced risk of colorectal cancer (2). This association has been shown in some studies to be more marked among individuals with the TT genotype for MTHFR C677T, a genetic polymorphism that results in a 50% reduction in enzyme activity, as well as among high consumers of alcohol, a folate antagonist (2, 3).

A smaller number of studies have also linked elevated dietary or plasma vitamin B6 levels to reduced colorectal cancer risk (4-8). This association may involve mechanisms additional to one-carbon metabolism. Low levels of vitamin B6 have been associated with elevated C-reactive protein (CRP), a marker of inflammation, and have been shown to increase oxidative stress, nitric oxide synthesis, cell proliferation, and angiogenesis (9, 10).

Because few prospective studies have examined the association of plasma levels of B vitamins with colorectal cancer risk and because these existing studies typically have not considered the correlation that exists in their levels, it remains unclear which ones, if any, have an independent effect on colorectal cancer. We conducted a comprehensive investigation into the relationships of plasma levels of folate, pyridoxal-5′-phosphate (PLP, the active form of vitamin B6), vitamin B12 (cobalamin), methylmalonic acid (elevated in cobalamin deficiency), homocysteine, and cysteine with colorectal cancer risk, accounting for suspected modifiers, using prospectively collected blood samples in the Multiethnic Cohort (MEC) study.

In 1993-1996, the MEC study recruited more than 215,000 individuals to investigate the relationships among diet, genetic variation, and cancer in Hawaii and Los Angeles, California (11). The study targeted men and women, 45 to 75 y of age in 1993 and of five racial/ethnic groups (African Americans, Native Hawaiians, Japanese Americans, Latinos, and Whites). Participants completed a 26-page baseline questionnaire that included a quantitative food frequency questionnaire with detailed information on portion sizes, as well as questions on medical history and lifestyle. From that questionnaire, information on family history of colorectal cancer, body mass index, smoking history, physical activity, multivitamin supplement use, and alcohol consumption was incorporated into the analyses for the present study. Additionally, dietary intakes of specific nutrients and food groups were estimated from single items and mixed dishes as measured by the food frequency questionnaire and from questions on supplement use. A short 4-page questionnaire was administered about 5 y after baseline to update the participants' medical and cancer screening histories.

Recruitment of cohort members for prospective biospecimen collection largely took place from 2001 to 2006. They were contacted by letter, and then by phone, to request biological specimens (blood and urine). For those who agreed, a phone interview was administered that included a short screening questionnaire and an update on a few items, including colorectal cancer screening. Blood samples were drawn at a clinical laboratory or at the subjects' homes and were processed within 4 h of collection. After centrifugation and separation, blood components were stored in 0.5-cc cryotubes in the vapor phase of liquid nitrogen. A total of 67,594 cohort members contributed to the biorepository, from which the cases and controls for this study were selected. Of the participants who provided the blood samples, 95% had fasted (≥8 h).

Incident colorectal cancer cases were identified through the Hawaii and California tumor registries of the Surveillance, Epidemiology, and End Results (SEER) Program of the National Cancer Institute. For this nested case-control study, cases were defined as individuals who had contributed blood to the biospecimen subcohort prior to their diagnosis of colon or rectal cancer, and whose diagnosis was reported in the 2006 tumor linkage.

For each case, a pool of potential controls was selected from the MEC participants who contributed blood to the biorepository and were alive and free of colorectal cancer at the age of the case's diagnosis, and who matched the case on sex, birth year (±1 y), race/ethnicity, location (Hawaii or California), date of blood draw (±6 mo), time of blood draw (±2 h), and hours of fasting prior to blood draw (0-<6, 6-<8, 8-<10, and ≥10 h). From each pool, two controls were then randomly chosen.

Of the 263 eligible colorectal cancer cases and 526 eligible controls, 26 cases and 51 controls did not have any plasma B vitamin measurement due to the unavailability of a fasting sample, and an additional 4 controls were excluded due to the loss of matched sets (cases had no data). Of the remaining 237 cases and 471 controls, 12 cases and 40 controls were excluded because they had missing data on at least one of the adjustment variables. One additional case and 20 controls were excluded due to the loss of matched data sets. A total of 224 cases and 411 controls (187 cases matched to 2 controls; 37 cases matched to 1 control) were analyzed. The vitamin B panel was expanded for the second batch of samples sent for laboratory analysis to include cysteine and methylmalonic acid. These analytes were measured on only 118 cases and 211 controls (93 cases matched to 2 controls; 25 cases matched to 1 control). MTHFR polymorphism data were available on 197 cases and 358 controls (161 cases matched to 2 controls; 36 cases matched to 1 control).

All laboratory assays were conducted blind to the case-control status of the participants. The samples from each matched case-control set were analyzed together as sets within laboratory batches and in random order within sets.

Plasma B Vitamins. Plasma folate and vitamin B12 concentrations were determined by a radioimmunoassay method using a commercially available kit from Biorad. Pooled plasma samples are used for internal laboratory quality control purposes and were analyzed as blind duplicates with the study samples. The intrabatch coefficients of variation from 20 blind duplicate pairs were 5.4% for vitamin B12 and 7.1% for folate. Plasma PLP, the active form of vitamin B6, was determined enzymatically using tyrosine decarboxylase based on the principles described by Shin-Buehring et al. (12). In this method, PLP activity in the plasma sample is determined on the basis of release of tritiated tyramine following the incubation of tyrosine decarboxylase apoenzyme with the supernatant fraction of trichloroacetic acid-precipitated serum sample and tritium-labeled tyrosine. The intrabatch coefficient of variation for this assay was 6.1% based on 27 duplicate pairs. Methylmalonic acid was quantified by a modification of the liquid chromatography mass spectrometry method described by Blom et al. (13), and the intrabatch correlation coefficient was 11.7% based on 16 duplicate pairs. Total homocysteine and cysteine were determined in plasma by a high pressure liquid chromatography method derived from that of Araki and Sako (14). The intrabatch coefficient of variation was 4.2% for homocysteine and 3.5% for cysteine based on 15 duplicate pairs.

Plasma CRP. The CRP assay was done using a Cobas MiraPlus CC clinical autoanalyzer (Roche Diagnostics) and a latex particle enhanced immunoturbidimetry based kit from Pointe Scientific with a functional sensitivity of 0.1 mg/L. The coefficient of variation for CRP was 7.5% based on 34 duplicate pairs.

DNA was purified from blood buffy coat using QIAamp DNA Blood Kits (Qiagen). The DNA was plated on 384-well plates together with duplicate quality control pairs. All assays were run on an AB 7900HT Fast Real-Time System. Genotyping of the GSTP1 missense single nucleotide polymorphism rs1801133 (MTHFR C677T) was done using 10 ng of genomic DNA and the 5′ nuclease TaqMan allelic discrimination assay from Applied Biosystems. Ethnic-specific genotype frequencies for MTHFR C677T in the controls were compared against Hardy-Weinberg equilibrium. All frequencies met the Hardy-Weinberg equilibrium criteria at P > 0.05. The genotyping call rate was 99.7% and the concordance rate across 78 blind duplicate samples was 100%.

To estimate the associations of plasma levels of specific B vitamins and the risk of colorectal cancer, we used conditional logistic regression models to compute odds ratios (OR) and 95% confidence intervals (95% CI), where the matched case-controls sets made up the strata. We adjusted for age at blood draw and hours of fasting prior to blood draw as continuous variables to account for differences within sets. Quartile and tertile cut points of the overall distribution of plasma levels were used to derive dummy variables, in which the lowest quantile served as the reference group. Median cut points were also used to derive a dichotomous variable for each B vitamin. To test for linearity in the logit of colorectal cancer risk, a trend variable was assigned the median value of the appropriate quantile. Alcohol and CRP were also dichotomized at the median and MTHFR was modeled as a binary variable (TT versus CC or CT).

In addition to the initial crude regression models, secondary models were created to further adjust the risk estimates for the following risk factors for colorectal cancer: processed meat intake (density, g/1,000 kcal/d), ethanol consumption (ever in the past year at least once a month versus never), body mass index (weight, kg/height, m²), pack-years of cigarette smoking; hours of moderate or vigorous physical activity, family history of colorectal cancer among parents and siblings, and history of colorectal cancer screening (never versus ever). Plasma folate was also adjusted for in a separate set of models. Other potential confounders, such as intake of red meat, fish and dairy products, were adjusted for in select models. The likelihood ratio test was used to examine interactions between plasma B vitamins and alcohol intake, plasma CRP, the MTHFR C677T polymorphism, or plasma folate in relation to colorectal cancer risk. The one degree of freedom test compared a main effects model, including the dichotomous variables of interest, with a model containing the interaction terms. Stratified analyses by sex, race/ethnicity, supplement use, and colon and rectal cancer were also done. Polytomous logistic regression was used to test for differences in the risk estimates for colon and rectal cancer, where sets were broken and matching variables were included as additional adjustment variables. Analyses were done in SAS, version 9.1.3.

As shown in Table 1, colorectal cancer cases were more likely to be obese (23% versus 11% had body mass index ≥30 kg/m²), more likely to consume alcohol (57% versus 52%), and generally consume more red and processed meat and less dairy than controls. Little to no difference was observed for dietary fiber intake between the cases and their matched controls. A higher proportion of cases than controls had a family history of colorectal cancer (12% versus 10%) and had ever smoked (60% versus 54%). Hours spent in moderate or vigorous physical activity were similar. The median levels for the plasma B vitamins were lower for cases than controls, except for methylmalonic acid. CRP levels were higher for cases.

Table 2 presents the crude and adjusted ORs and 95% CIs for the association between each of the plasma B vitamins and the risk of colorectal cancer. Subjects in the upper quartiles for folate levels tended to show a reduced risk of colorectal cancer (adjusted OR for Q4 versus Q1, 0.61; 95% CI, 0.33-1.13; P for trend = 0.097). A significant inverse trend was observed for PLP levels, with ORs of 1.00, 0.84 (95% CI, 0.51-1.40), 0.62 (95% CI, 0.37-1.03), and 0.49 (95% CI, 0.29-0.83), P for trend = 0.009. Because PLP and folate levels were correlated (Spearman correlation coefficient = 0.40), we further adjusted the risk estimates for plasma folate (last column in Table 2). The association between PLP and colorectal cancer remained significant. No significant trends were observed for vitamin B12, methylmalonic acid, or homocysteine. We also ran a model for subjects with low vitamin B12 status, defined as vitamin B12 <148 pmol/L or methlymalonic acid >210 nmol/L (223 cases, 407 controls), compared with all others. The adjusted OR for colorectal cancer was 1.27 (95% CI, 0.73-2.20). There was a reduced risk suggested for increasing levels of cysteine, but this trend was not statistically significant. Further adjustment for intakes of red meat, fish, and dairy did not change the risk estimates.

The main effect for MTHFR C677T was not significant (OR for TT versus CC+CT genotype, 0.95; 95% CI, 0.52-1.73). Similarly, no statistically significant association was found with plasma CRP, with multivariate-adjusted ORs for increasing quartiles of ≤0.30, 0.31-0.80, 0.81-2.20, and >2.20 mg/L at 1.00, 0.88 (95% CI, 0.52-1.49), 1.00 (95% CI, 0.61-1.62), and 1.19 (95% CI, 0.71-2.00), respectively (P for trend = 0.33); or with alcohol intake, with multivariate-adjusted ORs for non-drinkers, ≤3.3, 3.4-16.4, and >16.4 g/d at 1.00, 1.24 (95% CI, 0.79-1.96), 1.27 (95% CI, 0.78-2.06), and 1.37 (95% CI, 0.81-2.31), respectively (P for trend = 0.35) in this nested-case control study.

Table 3 shows the joint effects between B vitamins and, separately, alcohol, CRP, and the MTHFR 677TT genotype in relation to the risk of colorectal cancer. An interaction between plasma folate and alcohol intake was observed (P_interaction = 0.050), with a significantly decreased colorectal cancer risk observed with a high plasma folate (>15.3 ng/mL) only among subjects with an alcohol intake of ≤0.74 g/d (OR, 0.55; 95% CI, 0.31-0.95), compared with individuals with a low alcohol intake and a lower plasma level (Table 3). Plasma folate, comparing subjects with high versus low levels, was not associated with a reduced risk among subjects with an alcohol intake over the median (OR, 1.12; 95% CI, 0.62-2.04). This interaction with alcohol intake was also suggested for PLP (P_interaction = 0.072), where among light drinkers, the association with plasma PLP >51.5 pmol/mL, compared with a lower level (OR, 0.43; 95% CI, 0.26-0.74), was stronger than among heavier drinkers (OR, 0.84; 95% CI, 0.50-1.40). No interaction was found between CRP and PLP, using for PLP either the median value (51.5 pmol/mL; P_interaction = 0.42) or the cut point usually used for deficiency (30 pmol/mL; P_interaction = 0.66). Unexpectedly, we observed a significant interaction between plasma CRP and vitamin B12 (P_interaction = 0.027), with the suggestion of increased colorectal cancer risk with higher vitamin B12 levels among subjects with high CRP levels (OR, 1.50; 95% CI, 0.92-2.44). We also observed the suggestion of an interaction between PLP and MTHFR C677T on the risk of colorectal cancer (P_interaction = 0.08), with a stronger reduced risk among TT carriers with plasma PLP above the median, compared with those with PLP less than the median (OR, 0.25; 95% CI, 0.08-0.80), than among subjects with the CT or CC genotype (OR, 0.74; 95% CI, 0.47-1.15). A similar relationship was suggested for plasma folate, with ORs of 0.37 (95% CI, 0.11-1.21) and 0.92 (95% CI, 0.55-1.51) for high versus low folate levels among TT carriers and CT/CC carriers, respectively.

We also tested the interaction between plasma folate and PLP on the risk of colorectal cancer. The colorectal cancer OR for subjects with both PLP and folate levels above the median, compared with those with both levels less than the median, was 0.55 (95% CI, 0.33-0.93) and P_interaction = 0.93.

Stratified analyses by sex, anatomic subsites (colon, rectum), or race/ethnicity did not reveal any heterogeneity in the association of colorectal cancer with plasma PLP. A similar analysis stratified by multivitamin supplement use suggested a stronger effect among nonusers, with ORs for increasing tertiles of PLP at 1.00, 0.85 (95% CI, 0.41-1.77), and 0.34 (95% CI, 0.13-0.92; P trend = 0.03) than users at 1.00, 0.72 (95% CI, 0.41-1.26), and 0.70 (95% CI, 0.40-1.22; P trend = 0.32); however, the test for interaction was not significant (P_interaction = 0.52). Exclusion of cases diagnosed during the first year of follow-up did not materially change the risk estimates.

In this case-control study nested in the MEC study, we found an inverse association between plasma PLP levels and colorectal cancer which was not explained by an association with plasma folate, seemed stronger at low levels of alcohol intake and among individuals with the MTHFR 677TT genotype, and was independent of plasma CRP levels. An inverse association with plasma folate was also observed, which seemed to be stronger among individuals with a low level of alcohol intake and was suggested among subjects with the MTHFR 677TT genotype.

Our findings for PLP are consistent with those of Wei et al. (5) who found a 52% reduction in colorectal cancer risk comparing the highest to the lowest quartile for plasma PLP (P trend = 0.07) in a nested case-control study in women. They are also consistent with those of Figueiredo et al. (6) who found an inverse association between baseline plasma PLP levels and recurrence of adenoma in a randomized trial of aspirin and folic acid supplementation. Similarly, an inverse association was reported between serum PLP and colon cancer (but not rectal cancer) in the Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study cohort (7) and between plasma PLP and colorectal cancer in the Physicians' Health Study trial (8). These findings with biomarkers agree with the evidence from dietary studies that also found an inverse association with colorectal cancer for vitamin B6 intake from food and/or supplements (4).

Vitamin B6 can reduce colorectal cancer risk by preventing disruption in the one-carbon metabolism in which it acts as a cofactor for serine hydroxymethyltransferase that catalyzes the formation of 5,10-methylenetetrahydofolate, an important carrier of one-carbon groups for nucleotide synthesis and DNA repair. Vitamin B6 is also a cofactor for cystathionine-β-synthase, which catalyzes the irreversible conversion of homocysteine to cystathionine, a reaction that affects levels of homocysteine. Vitamin B6 may also protect against colorectal cancer through mechanisms independent of the one-carbon metabolism. A study showed that vitamin B6 supplementation reduced cell proliferation and the number of colon tumors induced by a carcinogen administered to mice (15). Vitamin B6 has also been shown to reduce oxidative stress and angiogenesis (10). Although vitamin B6 is widely distributed in foods, there is evidence that many older adults are not obtaining adequate amounts of this nutrient (16). For example, the prevalence of inadequate intakes is approximately 20% for men over age 50, and 40% for women in this age group.

Animal studies have suggested that folate may have a dual role, where a beneficial role may be observed at the early stage of carcinogenesis, before the establishment of neoplastic foci in the intestine (17). However, once preneoplastic lesions are present, folate may enhance the development and progression of colorectal tumors (17). Some recent human data have raised concerns that it may be the case in humans as well (18, 19). It has even been suggested that folate intake may be too high in the United States as a result of the grain fortification with folic acid initiated in the 1990s (20-22). However, we did not observe any increased colorectal cancer risk with plasma folate.

Consistent with several past studies (2, 3), we found an inverse association between plasma folate and colorectal cancer that was of borderline significance and seemed to be stronger among subjects with the MTHFR 677TT genotype. The association with plasma vitamin B6 also seemed stronger in this subgroup in our study. Previous studies have also reported a stronger protective effect of vitamin B6 on adenoma risk with the MTHFR 677TT genotype (23-25), although the interaction was significant in only one of these studies (23). These interactions may reflect the greater flow of one-carbon groups toward nucleotide synthesis with the TT genotype which may be protective only when the overall one-carbon pool is of sufficient size.

We also observed a stronger effect of folate and vitamin B6 at a low level of alcohol intake. Past colorectal cancer studies on plasma folate and alcohol have mostly shown no interaction (7, 8, 26, 27), whereas a stronger protective effect of dietary or plasma folate has been suggested against colon cancer and other cancers among heavy drinkers in some other studies (28, 29). A modifying effect of alcohol could be explained by its effect as a folate antagonist reducing the availability of dietary methyl groups, and may vary among populations based on methyl group supply and/or alcohol consumption level. Finally, the interaction that we observed in this study between CRP and plasma vitamin B12 levels was unexpected and may be due to chance.

There are a number of limitations to this study, besides its relatively small sample size. We only did a single measurement of plasma B vitamins, which may not reflect long-term vitamin B status, resulting in a possible attenuation of our risk estimates. However, an Australian study found an intra-individual coefficient of variation of only 8.3% and an Irish study found a reliability coefficient of 0.94 over a 1-year period (30, 31) for plasma folate, suggesting that a single measurement reflects reasonably good long-term B-vitamin status. Another limitation is that the dietary and anthropometric variables were not collected concurrently with the blood draw.

The important strengths of our study are the use of biomarkers (in contrast to questionnaire information), the prospective design in which blood samples were collected and covariates were measured before colorectal cancer diagnosis, and the direct generalizability of our results due to the broad representativeness of the MEC study.

In summary, these findings add to the evidence in support of an independent protective effect of vitamin B6 against colorectal cancer and suggest that this effect is stronger than that of folate.

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank the MEC and Progam Project Grant staff, who assembled the cohort and biorepository, as well as Dr. Adrian Franke for performing the CRP assays, and Annette Lum-Jones and Ann Seifried for carrying out the genotyping. We are also indebted to the participants for their invaluable participation.