Circulating Folate and Vitamin B12 and Risk of Prostate Cancer: A Collaborative Analysis of Individual Participant Data from Six Cohorts Including 6875 Cases and 8104 Controls

Background Folate and vitamin B12 are essential for maintaining DNA integrity and may influence prostate cancer (PCa) risk, but the association with clinically relevant, advanced stage, and high-grade disease is unclear. Objective To investigate the associations between circulating folate and vitamin B12 concentrations and risk of PCa overall and by disease stage and grade. Design, setting, and participants A study was performed with a nested case–control design based on individual participant data from six cohort studies including 6875 cases and 8104 controls; blood collection from 1981 to 2008, and an average follow-up of 8.9 yr (standard deviation 7.3). Odds ratios (ORs) of incident PCa by study-specific fifths of circulating folate and vitamin B12 were calculated using multivariable adjusted conditional logistic regression. Outcome measurements and statistical analysis Incident PCa and subtype by stage and grade. Results and limitations Higher folate and vitamin B12 concentrations were associated with a small increase in risk of PCa (ORs for the top vs bottom fifths were 1.13 [95% confidence interval (CI), 1.02–1.26], ptrend = 0.018, for folate and 1.12 [95% CI, 1.01–1.25], ptrend = 0.017, for vitamin B12), with no evidence of heterogeneity between studies. The association with folate varied by tumour grade (pheterogeneity < 0.001); higher folate concentration was associated with an elevated risk of high-grade disease (OR for the top vs bottom fifth: 2.30 [95% CI, 1.28–4.12]; ptrend = 0.001), with no association for low-grade disease. There was no evidence of heterogeneity in the association of folate with risk by stage or of vitamin B12 with risk by stage or grade of disease (pheterogeneity > 0.05). Use of single blood-sample measurements of folate and B12 concentrations is a limitation. Conclusions The association between higher folate concentration and risk of high-grade disease, not evident for low-grade disease, suggests a possible role for folate in the progression of clinically relevant PCa and warrants further investigation. Patient summary Folate, a vitamin obtained from foods and supplements, is important for maintaining cell health. In this study, however, men with higher blood folate levels were at greater risk of high-grade (more aggressive) prostate cancer compared with men with lower folate levels. Further research is needed to investigate the possible role of folate in the progression of this disease.


Introduction
Folate and vitamin B 12 are essential for maintaining healthy patterns of DNA methylation, repair, and synthesis [1], and low availability of these vitamins may influence cancer development through altered methylation patterns [2]. However, a meta-analysis of prospective studies published up to 2009 reported that higher concentrations of folate and vitamin B 12 were associated with a modest increase in prostate cancer (PCa) risk [3]. Interpretation of these findings is challenging because of the heterogeneous nature of PCa, with many tumours remaining small and asymptomatic for long durations of time [4], whereas some develop into aggressive forms and are ultimately lethal [5,6]. The use of prostate-specific antigen (PSA) testing has led to much earlier detection of localised cancers, many of which may never progress to aggressive disease [7]. Therefore, the identification of risk factors for PCa needs to take into account differences in stage and grade of the disease.
To investigate the association between circulating folate and vitamin B 12 concentrations and risk of PCa and, in particular, whether these associations differ by stage and grade of disease, individual participant data from six cohort studies participating in the Endogenous Hormones, Nutritional Biomarkers, and Prostate Cancer Collaborative Group were assessed.

2.
Materials and methods Seven eligible studies were identified [3,[8][9][10][11][12], of which six are included in this analysis; one research group declined to participate in this collaboration [11]. Studies provided data in predefined case-control sets matched on age at recruitment, date of blood collection (or followup time), and, if appropriate, other matching criteria as specified by the original study investigators (Supplementary Table 1). The Carotene and Retinol Efficacy Trial (CARET [13]; unpubl. data), the European Prospective Investigation into Cancer and Nutrition study (EPIC) [9], the Northern Sweden Health and Disease Study cohort (NSHDC) [8], and the Janus study [12] used a matched nested case-control study design within a prospective cohort. The remaining two studies were observational studies conducted within randomised controlled trials. The Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study (ATBC) [10] provided cases and controls matched on intervention group. The Prostate Testing for Cancer and Treatment Trial (ProtecT) [

Statistical analysis
To allow for possible systematic differences between studies in assay methods, storage conditions, and blood sample types, study-specific quintile cut points were used to categorise men into fifths of circulating

Results
Data were available for up to 6875 PCa cases and 8104 controls from six studies with a mean follow-up time of 8.9 yr (standard deviation: 7.3 yr). The mean age of control participants at blood collection ranged from 48.6 to 62.1 yr, and mean BMI varied from 25.7 to 28.3 kg/m 2 ( Table 1). Between 13% and 40% of control participants were current smokers, except in ATBC (100%), which recruited all smokers, and CARET (52%), which recruited current and former heavy smokers. Blood collection dates varied between 1981 and 2008. There was substantial variation across the studies in the time between blood collection and PCa diagnosis: >50% of cases were diagnosed >5 yr after blood collection in ATBC, EPIC, Janus, and NSHDC. In contrast, all cases from ProtecT and 82% of cases from CARET were diagnosed within 5 yr of blood collection ( Table 2). In all studies, !66% of cases were diagnosed in men aged !60 yr, with most cases diagnosed with localised (>50%) and low grade (>85%) disease, if known. Geometric mean folate concentrations (among controls) ranged from 5.7 nmol/l in NSHDC to 16.9 nmol/l in ProtecT. Vitamin B 12 concentration ranged from 295 pmol/l in ProtecT to 444 pmol/l in Janus. Folate concentration was higher in older age groups, in those with a higher level of education, in never smokers, and in those who had higher alcohol consumption ( Supplementary Fig. 1a). Vitamin B 12 concentration was higher in younger age groups and in those with lower education and lower alcohol consumption ( Supplementary Fig. 1b). The correlation between logtransformed folate and vitamin B 12 concentration was of a small magnitude although highly statistically significant (partial correlation adjusted for age at blood collection in five groups: r = 0.11; p < 0.0001).
Circulating concentrations of both folate and vitamin B 12 were positively associated with risk of PCa; the ORs for the top versus bottom fifths were 1.13 (95% confidence interval [CI], 1.02-1.26) for folate and 1.12 (95% CI, 1.01-1.25) for vitamin B 12 (Fig. 1), with no evidence of heterogeneity between the studies for both analytes (p heterogeneity > 0.05). Owing to the cross-sectional nature of the study design, analyses were also performed after excluding ProtecT (with 1458 cases and 1506 controls); the association of folate with PCa risk was slightly strengthened (Fig. 2a ) (Fig. 3a). , although, for these end points, the differences between the subgroups were not statistically significant (Fig. 3a). To examine in more detail the association of circulating folate concentration with risk of advanced stage, aggressive, and high-grade disease, analyses were performed across the study-specific fifths of folate concentration (Fig. 4a-4c) The association of circulating vitamin B 12 with PCa risk did not differ by stage or grade of disease (Fig. 3b) (p heterogeneity > 0.05). There was some evidence of heterogeneity in the association with vitamin B 12 by smoking status, with a higher concentration of vitamin B 12 associated with an increased risk of PCa in never smokers (OR per 80% increase: 1.37 [95% CI, 1.14-1.65]) but not in past or current smokers (p heterogeneity = 0.03). We also examined the joint effects of folate and vitamin B 12 in relation to PCa and found no evidence of interaction (p heterogeneity = 0.27) (Supplementary Table 3).

Discussion
The findings from this individual participant pooled analysis of 6875 cases and 8104 controls represent almost all of the existing observational data from cohort studies for the association of circulating concentrations of folate and vitamin B 12 with risk of PCa. Our results provide evidence of weak positive associations between circulating concentrations of both folate and vitamin B 12 and risk of PCa. Furthermore, the finding that a higher folate concentration was associated with an increased risk of high-grade disease but not low grade-disease suggests that the possible role of folate in PCa progression warrants further investigation. The effect of folic acid supplementation on cancer risk has also been investigated. An individual participant metaanalysis of randomised trials [15], with 656 incident PCa cases and an average of 5.2 yr of treatment, reported a modest (15%) but nonsignificant increased risk in the supplement versus placebo arms (95% CI, À1% to 34%); however, the median circulating folate concentration in the supplement arm was substantially higher than that observed in our study, and associations by stage or grade of disease were not assessed [15].
The role that circulating concentrations of folate might play in risk of high-grade disease and, to a lesser extent, advanced stage and aggressive disease is unclear. In vitro models using human prostate tissue have shown enhanced proliferation of tumour cells with elevated folate concentrations [16], and lower cellular proliferation has been observed in transgenic adenoma of mouse prostate (TRAMP) mice fed a folate-depleted diet compared with  those on a normal or high folate diet [17]. Furthermore, in a small sample of radical prostatectomy patients, increased cancer cell proliferation was observed in prostate samples from men with higher serum concentrations of folate [18]. There is also some evidence from genetic studies of a role for folate in the development of aggressive PCa, for example, an association of the homozygote TT genotype of the methylenetetrahydrofolate reductase gene (MTHFR C677T) with lower circulating folate concentrations and reduced risk for aggressive PCa [12,19]. There has been little research into whether folate intake in men with PCa might affect the progression of the disease, but the data available have not shown any evidence of an association of folate intake with PrCa progression [20] or survival [21]. To understand better the potential role of folate in PCa progression, more data are needed from large observational studies (or randomised trials) with long-term follow-up, circulating folate, and PCa tissue to investigate whether higher circulating concentrations are associated with quantifiable local molecular changes.
The modest 12% increased risk of PCa associated with a higher vitamin B 12 concentration is similar to that reported from a meta-analysis of five studies (fixed-effects pooled estimate per 100-pmol/l increment, OR: 1.10 [95% CI, 1.03-1.18]) [3], of which three studies were eligible for inclusion in the present analysis and contributed 45% of the data. The current analysis suggests that this modest association might be modified by smoking, with higher vitamin B 12 associated with an increased risk of PCa in never smokers but not in current or past smokers. Circulating vitamin B 12 did not differ by smoking status in our study, and it is possible that our finding may be due to chance, given the multiple statistical tests performed.
The strengths of the current analysis include the large sample size and the detailed data on participant characteristics. The prospective design, with blood collected several years before diagnosis in most studies, allowed us to evaluate the association by time from blood collection to diagnosis, the results of which do not suggest that the observed associations are due to reverse causation bias.
[ ( F i g . _ 4 ) T D $ F I G ] Fig. 4 -Odds ratios and 95% confidence intervals for (a) advanced stage, (b) aggressive, and (c) high-grade prostate cancer by study-specific fifths of circulating folate concentration. The odds ratios are conditioned on the matching variables and adjusted for exact age, marital status, education level, cigarette smoking, height, and body mass index. The odds ratio per 80% le increase in folate concentration and p value for trend were calculated by replacing the fifths of the folate with a continuous variable that was scored 0, 0.25, 0.5, 0.75, and 1 in the conditional logistic regression model. 80% le = 80-percentile; CI = confidence interval; OR = odds ratio; P tr = p value for trend.
Changing diagnostic practices over time and between settings, including increasing use of PSA testing (used to a greater extent in the United States [22] than in Europe [23]), and updates to Gleason scoring will have influenced rates of case ascertainment, lead time to diagnosis, and disease classification during the follow-up period. Unfortunately, individual participant screening histories and blood PSA levels at recruitment were not available for four of the six studies included in the analysis, so it was not possible to take these factors into account. However, our stringent definition for high-grade PCa (Gleason sum of !8 or equivalent) means that the association seen for this subgroup is unlikely to include low-or intermediate-grade cancers.
The current analysis is based on measurements of folate and B 12 concentrations taken from a single blood sample from each participant, which may not characterise usual or longer term blood concentrations, leading to some attenuation of risk estimates. Previous studies have reported that the within-person reproducibility over a 1-to 4-yr period is fair to good for folate (intraclass correlation coefficient: 0.47-0.61) and better for vitamin B 12 (intraclass correlation coefficient: 0.61-0.87) [24][25][26]. However, mandatory folic acid fortification of cereals in the United States [27] during the follow-up period of the CARET study and some voluntary fortification of foods in Europe [28] might further contribute to misclassification of usual exposure in the absence of repeated measures.
Variation in circulating folate and vitamin B 12 concentrations between the contributing studies is probably mostly real (ie, due to differences in intake [29]), although it is likely that there is also some artefactual variation due to differences in blood processing, storage, and assay methodology [30][31][32][33][34]. However, samples from all studies (with the exception of a small subcohort of the EPIC study) were processed within a short time of blood collection, and samples from the Janus study (stored at À25 8C) were analysed using a novel assay capable of recovering degraded folate [35,36].

Conclusions
Higher folate concentration was associated with an increased risk of high-grade disease that was not evident for low-grade disease. This finding suggests a possible role for folate in the progression of clinically relevant PCa and warrants further investigation.
Author contributions: Alison J. Price had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Key, Allen.
Acquisition of data: Allen, Appleby.
Analysis and interpretation of data: Appleby, Price, Travis, Key, Allen.
Administrative, technical, or material support: None.