Unmetabolized Folic Acid, Tetrahydrofolate, and Colorectal Adenoma Risk

Extensive epidemiological and preclinical data have suggested that folates, a family of water-soluble vitamins that serve as cofactors for DNA synthesis and methylation reactions, exert substantial anti-neoplastic effects (1). Motivated by these findings, we conducted a randomized controlled trial to test the effect of 1 g/day folic acid for the prevention of colorectal adenoma recurrence (2). Although we found no effect during the first three years of treatment, we saw unexpected indications of increased risk during subsequent follow-up and treatment: the relative risk for the occurrence of advanced adenomas was 1.67 (95% CI, 1.00–2.80) and for 3 or more adenomas, 2.32 (95% CI, 1.23–4.35; ref. 2).

Although our finding of increased adenoma risk from folic acid supplementation may have been due to chance, there are other indications of increased cancer-related risks of folic acid, depending on dose, timing, and folate status before supplementation. Whereas moderate doses may reduce risk, carcinogenesis may be enhanced by larger doses, or by supplementation after a neoplastic lesion has developed (3–6). The mechanisms for this adverse effect have not been fully elucidated but may include the provision of nucleotides for carcinogenesis (7).

Folic acid is a stable, oxidized molecule rarely found in nature, though widely used for food fortification and in supplements. Folic acid is the molecular backbone of folates, but it is inactive until it is metabolized into the natural reduced forms, including 5-methyltetrahydrofolate (5m-THF), the prevailing circulating folate species (8). Because folic acid enters the folate metabolic cycles differently than 5m-THF and has a distinct metabolism, it may have biological effects that differ from those of the naturally occurring folates. For example, some studies suggest that UFA reduces natural killer cell cytotoxicity in vitro (9, 10). Thus, the increased risks associated with folic acid supplementation in our study could have been the result of folic acid itself, not the 5m-THF generated by it.

For both 5m-THF and folic acid, genetic variants in folate metabolism may also be pertinent. For example, the homozygous C677T variant of methylenetetrahydrofolate reductase (MTHFR), which metabolizes methylenetetrahydrofolate to mTHF, may modify the association of folate status with risk of colorectal carcinoma and adenoma (11–13). MTRR, methionine synthase reductase, maintains methionine synthase in its active form, and in this state maintains intracellular folate levels (14).

It is widely recognized that colorectal cancer is a heterogeneous disease, with two distinct precursor pathways: the conventional pathway (accounting for about 60% of colorectal cancers) and the serrated pathway, which leads to CpG island methylated colorectal cancers (accounting for the remaining ∼40%; refs. 15, 16). Because these involve distinct genetic and epigenetic processes, it is possible that folates affect these pathways in different ways.

In a secondary, observational analysis, we previously found no consistent association between folate status and adenoma risk (17). However, during early follow-up, total folate intake and plasma total folate were inversely associated with adenoma risk among subjects randomized to folic acid placebo, while in later follow-up there were inconsistent suggestions of increased risks for dietary folate intake and red blood cell folate. In this analysis, we extend our analyses using separate measures of plasma UFA and methylated folate (mF), and separate assessment of the conventional adenoma pathway and the serrated pathway. We hypothesized that long-term risk after folic acid supplementation is independently associated with plasma UFA levels.

Briefly, we conducted a double-blinded, placebo-controlled, randomized trial (NCT00272324) in a 3 × 2 factorial design to compare the effects of oral aspirin (81 mg/day, 325 mg/day, or placebo) with placebo or folic acid 1 g/day, in the prevention of colorectal adenomas in persons with a history of adenomas (2, 18). The trial's primary outcome was the occurrence of ≥1 colorectal adenomas. Secondary outcomes were the occurrence of advanced adenomas (tubulovillous or villous adenomas, or adenomas that were large (≥1 cm) or had invasive cancer or high-grade dysplasia); high-risk findings (≥1 advanced adenomas or ≥3 adenomas); and serrated lesions (comprising hyperplastic polyps, sessile serrated adenomas/polyps, and traditional serrated adenomas). The recruitment period (1994–98) coincided with the beginning of folic acid food fortification in the United States and Canada that became mandatory in 1998 (19, 20). All participants provided written informed consent; the trial was approved by each participating center's institutional review board.

Participants aged 21–80 with a recent adenoma and no known remaining adenomas after a complete colonoscopy were identified at nine clinical centers, as described previously (18). Exclusion criteria included vitamin B₁₂ deficiency, contraindication to aspirin treatment, or ongoing treatment with aspirin or related compounds (2, 18).

After a run-in phase of approximately 3 months, 1,021 participants were randomized to aspirin (81 mg, 325 mg or placebo) and folic acid (1 mg or placebo); 100 early participants were randomized only to aspirin/placebo. The initially planned study interventions continued until a follow-up colonoscopy anticipated approximately 3 years after the qualifying examination. This demarcated the end of the first follow-up interval and the beginning of the second interval, which lasted until 6 years later, one year longer than a 5-year colonoscopic surveillance cycle (Fig. 1).

At the end of the first follow-up interval, participants were asked to continue placebo/folic acid treatment until their next surveillance colonoscopy, anticipated 3 to 5 years later; 729/1,084 (67%) agreed to do so (Fig. 1). Of the 355 who chose to discontinue randomized treatment, 278 (78%) agreed to observational follow-up. Participants still on study treatment were asked to stop taking study tablets by October 1, 2004, because folic acid supplementation had been associated with increased adenoma risk.

Blood samples were collected at enrollment (“year 0,” reflecting prerandomization folate status), and shortly before the end of the first treatment interval (“year 3”), reflecting folate status on treatment. The blood draws were not timed to intake of food or study tablets. This study examined associations with circulating concentrations of folic acid and naturally occurring folates shortly before the year 3 colonoscopy.

Plasma levels of 5m-THF, 4-α-hydroxy 5-methyl-THF (hm-THF), and unmetabolized folic acid (UFA) were analyzed at Bevital (21), using LC MS/MS (22). hm-THF is formed upon degradation of 5m-THF during storage so we pooled these measures for each sample and refer to the sum as mF (22). The levels of detection were 0.27 nmol/L for UFA, 0.20 nmol/L for hm-THF, and 0.07 nmol/L for 5m-THF (22). The intra-run CVs ranged from 5.0% to 9.9%, and the inter-run CVs from 3.3% to 9.5% (22). We identified the 677C>T, V222A (rs1801133) polymorphism of the methylenetetrahydrofolate reductase (MTHFR) gene and the 66A>G, I22M (rs1801394) polymorphism of methionine synthase reductase (MTRR) as previously described (23).

Participants taking study treatments were contacted every 4 months for information on adherence, use of personal nutritional supplements including folic acid, large bowel endoscopies and major medical events. During observational follow-up, they were contacted annually. For all tissue removed from the bowel, slides were obtained if possible and sent to a single, blinded pathologist for classification.

We used conventional descriptive statistics to summarize characteristics of the study population. Correlations were assessed using Kendall's Tau b. We used the circulating folate measures at year 3 to characterize the folic acid and folate status of subjects during the first follow-up interval. In the analysis for the second follow-up interval, we included participants whose study folic acid treatment remained unchanged from that in the first interval: those randomized to folic acid who agreed to continue study treatment, those allocated folic acid placebo whether or not they agreed to continue study pills, and 100 participants randomized only to aspirin or placebo (i.e., no randomization to folic acid/placebo). Thus, apart from changes in diet, and non-compliance with the study treatment regimen, the year 3 blood specimen represented the UFA and mF status during the second follow-up interval as well as the first.

We fit binomial regression models with a log link to describe the associations of circulating folate measures at year 3 with the occurrence of adenomas and serrated polyps during each of the surveillance intervals. The endpoints were ≥1 adenomas; ≥1 advanced adenomas; high-risk findings; ≥1 serrated lesions. All analyses were adjusted for age and sex, but not for treatment group (which would be reflected in the year 3 plasma measurements). Clinical center did not materially affect the analyses, and was excluded from all models. Personal folic acid supplement use during the second follow-up interval did not contribute usefully to the models for any of the measured outcomes, either as a covariate or as an effect modifier. It was therefore omitted from the analysis, and the same model was used for both intervals (i.e., age, sex, mF, and UFA). Estimates for mF and UFA were mutually adjusted for each other so that the risk ratios reflected the independent associations of each analyte. mF was modeled in quartiles; UFA was modeled in four categories, undetectable (set to zero) plus tertiles of the remaining values. We assessed patterns of associations with a general test of heterogeneity (Wald test) based on the categorical variables. We assessed tests for trend using the linear and quadratic functions of the continuous measurements. This approach allowed testing of both monotonic trends and the possibility that modest increases in mF or UFA levels might be beneficial, while higher levels detrimental. Results are given as relative risks of outcome by mF or UFA category. The possibility of effect modification by randomized aspirin assignment (dichotomous) or by MTHFR or MTRR polymorphisms was assessed using likelihood ratio tests. All analyses were performed with SAS v9.4.

Year 3 plasma measurements and colonoscopy data were available for 924 participants in the first interval, 680 of whom completed a further 3 to 6 years of colonoscopic follow-up (Fig. 1; Table 1). At enrollment, the mean concentration of mF was 37.7 nmol/L (SD 22.4); at year 3 this had increased to 46.3 nmol/L (SD 21.7) in the placebo group and 84.8 nmol/L (SD 29.5) in the folic acid treatment group. At enrollment, UFA was detected in 19.3% of participants, with a mean concentration of 4.2 (6.5) nmol/L for the 170 with detectable levels. Just before the 3-year surveillance colonoscopy, detectable UFA was found in 23.2% of the placebo group and 73.0% of those in the folic acid group, with mean concentrations of 3.4 (8.5) and 21.7 (24.9) nmol/L for those with detectable levels, respectively. At year 3, mean plasma mF increased from 48.0 to 97.0 nmol/L within increasing categories of UFA (Supplementary Fig. S1). The correlation between mF and UFA at year 3 was r = 0.46 (P < 0.0001).

There were no significant associations between circulating UFA or mF levels and overall adenoma risk during either follow-up interval (Table 2). However, during the second follow-up interval, there were indications of an association of mF with increasing risk of advanced adenomas and high risk findings (advanced or ≥3 adenomas; Table 2). For high-risk findings, there was a linear trend of increasing risk with increasing levels, with a 58% increased risk for participants in the upper quartile compared with the lowest (Table 2). There was a weaker, borderline-significant trend for advanced adenomas (25% increased risk for upper vs. lowest quartile). During the second interval, UFA showed an erratic association with high-risk findings, with a borderline-significant inverse linear trend (P = 0.049). Analyses without mutual adjustment for mF and UFA showed broadly similar findings, though with nonsignificant trends (Supplementary Data; Table 1).

There was no evidence for any differences between men and women in the association of mF or UFA with any of these adenoma endpoints (all P_interaction > 0.15; Supplementary Data; Table 2). There was also little evidence for a quadratic relationship between mF or UFA and these outcomes during early or later follow-up (all P > 0.15; not shown). No significant interactions were seen for any outcome between randomized aspirin use (dichotomous) and either UFA or mF.

Associations for serrated pathway lesions differed markedly from those for adenomas. During the second surveillance interval, the risk of any serrated lesion was significantly inversely associated with plasma mF, with a 39% lower risk for upper vs. lower quartile participants (Table 2). There was no association with plasma UFA (Table 2).

As reported previously for the entire study population (24), there were no significant associations between adenoma risk and polymorphisms of MTHFR or MTRR in either follow-up interval (Supplementary Table S3). There were also no indications of interactions between plasma levels of mF or UFA and genetic variants in either gene with regard to risk of all adenomas (Supplementary Table S3). However, in the corresponding analysis of high risk-findings, there was an interaction of MTHFR with mF (P = 0.03) during the first follow-up interval, with higher mF-associated risk in CC participants than in those with the CT/TT genotype. (Supplementary Table S4). In the second interval, there was a broadly similar pattern, but the interaction did not reach statistical significance (P = 0.08).

In our clinical trial, we previously found that participants randomized to 1 mg/day folic acid had an increased risk of advanced and multiple adenomas during longer follow-up (2). In this post hoc analysis, high circulating levels of mF were positively associated with the composite high-risk endpoint of advanced and/or multiple adenomas during this same period, while there was lower risk in the highest category of UFA, with a borderline significant trend. These findings suggest that the increased risks seen after folic acid supplementation (2) were due to the high mF levels, rather than due to a direct effect of UFA itself (Fig. 2). In contrast, our findings for serrated lesions suggested a protective effect of higher levels of mF during later follow-up.

We found an interaction between MTHFR 677C→T and mF levels such that in the first follow-up interval there was a particularly elevated risk of high-risk findings associated with high mF only among individuals homozygous for the CC allele. This is consistent with our finding from the second follow-up interval that mF is associated with increased risk, as the enzyme variant associated with the T allele generates 5 methyl-tetrahydrofolate from 5, 10-methylenetetrahydrofolate less efficiently than the variant associated with the C allele. However, during the second follow-up interval, the interaction was less marked.

Cho and colleagues identified no overall association between circulating UFA and colorectal cancer, but with significant heterogeneity by sex, with higher risks among men and lower risks among women, with increasing UFA (25). In the second follow-up interval of our study, the directions of effects were similar to those seen by Cho and colleagues, but the interactions were not statistically significant in our smaller study population. Cho and colleagues also reported a significant interaction between UFA and MTHFR C677T, with a colorectal cancer OR of 2.20 (95% CI, 1.22–3.94) for the highest versus lowest levels of UFA among those with CT or TT genotype. Increasing risks associated with the TT or CT genotype have been found previously, either as main effects, or with increasing total folate levels (11, 26). We also found only slightly and nonsignificantly higher risk ratios in those with these genotypes at highest UFA levels.

Several laboratory studies indicate that folate or folic acid may enhance carcinogenesis in some settings. For example, in mouse models of colorectal cancer, low doses of dietary folic acid were protective if given before tumors developed, but promoted further tumor growth if given after tumors had emerged (27), and in a rat model of breast cancer, folate deficiency was protective (28, 29). Other in vitro studies have shown that folate repletion after deficiency has the potential to lead to carcinogenesis (5). Some argue that, because folate antagonists are effective anticancer drugs, the rationale for folate supplementation to prevent cancer may be flawed; indeed, the use of antifolates in leukemia was first prompted by the observation that large doses of folic acid made patients with leukemia substantially worse (30).

Although the inverse association between high-risk findings and UFA may be a chance finding, there are reasons to speculate that synthetic folic acid might have different biological effects than naturally occurring folates (31, 32), some of them potentially harmful (33). UFA has been associated with neurological and cognitive effects in individuals with lower B₁₂ status (34, 35) and with reduced natural killer cell cytotoxicity in some studies in vivo (9, 10), although not in vitro (36). There are differences in the carrier systems through which these natural folates and UFA enter the cell, the points at which they enter the folate cycle, and the bioavailability of folates from diet and folic acid supplements (31, 37). The assimilation of folic acid into tetrahydrofolate is catalyzed by dihydrofolate reductase (DHFR), whose activity varies considerably between individuals and is inhibited by UFA (38). In addition, UFA may inhibit MTHFR, the enzyme that produces methyltetrahydrofolate (39). Thus, it is possible that UFA has tissue effects that decrease the bioavailability of intracellular mF levels following the supplementation. Further, after administration of radiolabeled folic acid to 10 volunteers, 40% of the increase in plasma total folate was due to mobilization of endogenous (unlabeled) folate rather than conversion of the ingested folic acid (40). The physiological differences between UFA and naturally occurring folate species are incompletely understood and require further study. The irregular risk ratios seen in the second interval of our study may also merit further investigation in future studies, to assess the possibility of a threshold effect.

After folic acid supplementation, 73.0% of our participants had detectable UFA, compared with 23.2% of controls; these figures are consistent with the wide range of detectable UFA seen in other studies (41–44). UFA has been found in the serum following boluses of dietary folic acid as low as 200 mcg in folate-replete women (34), suggesting that it should be expected in the serum of individuals taking 1 mg folate/day (45). A variety of factors influence the concentration of mF and UFA, including the time since ingestion of the supplement (46–48), variability in dietary intake including food fortification and DHFR polymorphisms. We did not control the timing of blood tests relative to pill-taking or meals in our study; variable timing would increase the variability of plasma measures and reduce our ability to identify associations if any exist. This would be particularly true for UFA, which has a short half-life (∼1 hour; ref. 49) as compared with the more deep folate compartments that have previously been detected after chronic supplementation with deuterium labeled folic acid (50). A future study should time blood draws consistently in relation to intake of folic acid from foods and supplements.

Our approach had the advantage of adjustment for serum mF while analyzing the effect of UFA and vice versa. By summing the measured concentrations of hm-THF and 5m-THF, we minimized the measurement error associated with 5m-THF breakdown to hm-THF in stored samples, (22). We also used a single blood sample during intervention (near the end of the first follow-up interval) to represent UFA or mF levels during two observation cycles; this is more reflective of in-trial levels than the presupplementation samples used in our previous report (17). However, a single measurement is an imperfect correlate of long-term exposure and may cause regression dilution bias. During a multiyear study, changes in adherence and in diet would likely reduce the reliability of the single blood measure in representing long-term folate status, and this would tend to reduce our ability to identify associations (51). A further limitation is that our findings with respect to adenoma and especially serrated polyps apply only indirectly to the study of colorectal cancer. Finally, in exploring trends in these data, we performed many significance tests, with the possibility of type 1 errors, particularly for the quadratic trend tests that were conducted; one of 20 tests was statistically significant with P = 0.04.

Our results indicate that the increased risk of advanced/multiple adenomas in later follow-up of our trial was likely related to higher levels of plasma mF, and not due to UFA. In contrast, findings regarding serrated lesions provide some suggestions of protection from mF in the serrated pathway. While our data do not indicate any significant harm associated with UFA itself, the associated high levels of mF seem to be detrimental with respect to adenoma risk. The widespread use of folic supplementation and food fortification makes associations between the levels of folate species and disease an important avenue for further research.

No potential conflicts of interest were disclosed.

Conception and design: J.R. Rees, J.A. Baron

Development of methodology: J.R. Rees, P.M. Ueland

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G.E. McKeown-Eyssen, J.C. Figueiredo, J.A. Baron

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.R. Rees, C.B. Morris, J.L. Peacock, G.E. McKeown-Eyssen, J.C. Figueiredo, J.A. Baron

Writing, review, and/or revision of the manuscript: J.R. Rees, C.B. Morris, J.L. Peacock, P.M. Ueland, E.L. Barry, G.E. McKeown-Eyssen, J.C. Figueiredo, D.C. Snover, J.A. Baron

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.C. Snover

Study supervision: J.A. Baron

The authors wish to acknowledge the participants and staff of the Polyp Prevention Study.

This work was supported in part by grants 5 R01CA059005 and U54 CA100971 from the NIH. Study tablets were provided by Wyeth Consumer Health Care, Madison, NJ. J.L. Peacock is supported by the NIHR Biomedical Research Centre based at Guy's and St Thomas' NHS Foundation Trust and King′s College London, UK.

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.