Folate intake, serum folate levels, and prostate cancer risk: a meta-analysis of prospective studies

This review was conducted and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Statement issued in 2009 (Additional file 1: Checklist S1).

We systematically searched the PubMed, Embase, and Cochrane Library electronic databases (from database inception to Jan 2014), with no language restrictions for studies in humans. We included all studies investigating an association between dietary folate intake or serum folate levels and prostate cancer incidence. Our core search included the following terms: (“folate” OR “folic acid”) AND (“prostate cancer” OR “prostate neoplasm” OR “prostate carcinoma”) AND (“cohort” OR “cohort studies” OR “nest case–control studies”). If a site-specific dataset had been published more than once, we used the most recent publication. We reviewed the reference lists of the identified reports, reviews, meta-analyses, and other relevant publications to find additional pertinent studies. The medical subject heading, methods, population, study design, exposure, and outcome variables of these articles were used to identify relevant studies.

A study was eligible for inclusion if the following criteria were met: (1) the study had a prospective observational design (prospective cohort or prospective nested case–control study); (2) the study investigated the association between dietary folate intake or serum folate levels and the risk of prostate cancer; and (3) the authors reported effect estimates (risk ratio [RR], hazard ratio [HR], or odds ratio [OR]) and 95% confidence intervals (CIs) for comparisons between high and low dietary folate intake or serum folate levels (with more than 2 categories). We excluded all retrospective case–control studies because various confounding factors could bias the results.

The literature search was independently undertaken by 2 authors (RW and YZ) using a standardised approach. Any inconsistencies were resolved by discussion with the primary author (YHZ) and a consensus was reached.

The following data elements were collected: name of the first author or study group, publication year, country, study design, assessment of folate levels, sample size, age at baseline, effect estimate, comparison categories, follow-up duration, and covariates in the fully adjusted model. We also extracted the numbers of cases per person or per person-year, effect of the different exposed categories, and 95% CIs. For studies that reported several multivariable adjusted RRs, we selected the effect estimate that was maximally adjusted for potential confounders.

The Newcastle-Ottawa Scale (NOS) was used to evaluate methodological quality [19, 20]. The NOS is a comprehensive tool that has been partially validated for evaluating the quality of observational studies in meta-analyses [19]. The NOS is based on the following 3 subscales: selection (4 items), comparability (1 item), and outcome (3 items). A “star system” (range, 0–9) has been developed for assessment (Additional file 2: Table S1). The data extraction and quality assessment were conducted independently by 2 authors (YZ and JYH). Information was examined and adjudicated independently by an additional author (JNW), who referred to the original studies.

We examined the relationship between dietary folate intake or serum folate levels and the risk of prostate cancer based on the effect estimate (OR, RR, or HR) and its 95% CI published in each study. We first used the random-effects model to calculate summary RRs and 95% CIs for the highest folate levels compared to the lowest folate levels [21, 22]. We subsequently transformed category-specific risk estimates into estimates of the RR associated with every 100 μg per day increase in dietary folate intake and every 5 nmol/L increase in serum folate levels via the generalised least squares for trend estimation. These estimates were calculated from the assumption of a linear relation between the natural logarithm of the RR and increasing folate levels [23, 24]. The value assigned to each folate level category was the mid-point for closed categories and the median for open categories (assuming a normal distribution for folate levels) [23]. We combined the RRs for each 100 μg per day increase in dietary folate intake and each 5 nmol/L increase in serum folate levels via a random-effect meta-analysis. Unless otherwise stated, we used the most adjusted risk estimate from each study as stated previously. We finally conducted a dose response random-effects meta-analysis from the correlated natural log of RRs or HRs across the folate levels categories. To derive the dose–response curve, we modelled folate using restricted cubic splines with 3 knots at fixed percentiles of 10, 50, and 90% of the distribution [23]. This method requires the effect measure with its variance estimate for at least 3 known categories of exposure. We assessed the heterogeneity between studies using the I² statistic to measure the proportion of total estimate variation that was attributable to study heterogeneity; I² values of 25%, 50%, and 75% were used as cut-off points for low, moderate, and high degrees of heterogeneity, respectively [25, 26]. Subgroup analyses were conducted for prostate cancer based on country, effect estimate, duration of follow-up, and adjustment for age, BMI, alcohol consumption or smoking status. We also performed a sensitivity analysis by removing each individual study from the meta-analysis [27]. Several methods were used to assess potential publication bias. Visual inspections of funnel plots for prostate cancer were conducted. The Egger [28] and Begg tests [29] were also used to statistically assess publication bias for prostate cancer. All reported P values were 2-sided, and P values <0.05 were considered statistically significant for all included studies. Statistical analyses were performed using STATA software (version 12.0; Stata Corporation, College Station, TX, USA).

After pooling the included studies [8‐12], the summary RR illustrated that a high dietary folate intake was not associated with prostate cancer risk (RR = 1.02; 95% CI = 0.95–1.09; P = 0.598, Figure 2A), and no evidence of heterogeneity was observed (I² = 0.0%; P = 0.959). The findings of the dose–response meta-analysis also suggested no association between the risk of prostate cancer and a 100 μg/day increment of dietary folate intake (RR = 1.01; 95% CI = 0.99–1.02; P = 0.433, [I² = 0.0%; P = 0.784], Figure 2B). As a result, a sensitivity analysis was conducted, and after each study was sequentially excluded from the pooled analysis, the conclusion was not affected by exclusion of any specific study.All studies were included in the dose–response curve between dietary folate intake and the incidence of prostate cancer. As shown in Figure 3A and illustrated by the P value for nonlinearity (P = 0.012), we found evidence of nonlinear relationships between dietary folate intake and the risk of prostate cancer. Dietary folate intake of more than 300 μg per day appeared to be associated with a non-significant reduction in the risk of prostate cancer.

A total of 5 prospective nest case control studies [7, 13‐16] reported an association between serum folate levels and the risk of prostate cancer. The pooled analysis results for prostate cancer incidence indicated that the comparison of the high versus low categories of serum folate levels was associated with a harmful effect (RR = 1.21; 95% CI = 1.05–1.39; P = 0.008, with no evidence of heterogeneity [I² = 0.0%; P = 0.724]; Figure 4A). The dose–response meta-analysis suggested that a 5 nmol/L increment of serum folate levels was associated with increased risk of prostate cancer (RR = 1.04; 95% CI = 1.00–1.07; P = 0.042, with no evidence of heterogeneity [I² = 0.0%; P = 0.418], Figure 4B). Furthermore, as shown by the P value of nonlinearity (P = 0.111), there was no evidence of a potential non-linear relationship (Figure 3B).

Heterogeneity testing for the analysis identified a P value >0.10 for prostate cancer incidence. We concluded that heterogeneity is not significant in the overall analysis, suggesting that most variation was attributable to chance alone. Subgroup analyses were conducted to evaluate the effect of folate on prostate cancer risk in a specific population. Overall, we noted that a 5 nmol/L increment of serum folate levels was associated with the increased risk of prostate cancer if the duration of the follow-up less than 15 years. No other significant differences in effects were detected between dietary folate intake or serum folate levels and the risk of prostate cancer (Table 2).

A review of funnel plots could not eliminate the potential for publication bias for prostate cancer (Figure 5). The Egger [28] and Begg test [29] results disclosed no evidence of publication bias for prostate cancer (Egger: P = 0.694 for dietary folate intake and P = 0.181 for serum folate levels; Begg: P = 0.806 for dietary folate intake and P = 1.000 for serum folate levels; Figure 5).