Association of vitamin C, vitamin D, vitamin E and risk of bladder cancer: a dose-response meta-analysis

A dose-response meta-analysis was conducted to assess the association of vitamin C, D, E with risk of bladder cancer. Pertinent studies were identified in PubMed and Embase. The random-effect model was used. The relative risk (95% confidence interval) of bladder cancer was 0.99 (0.95–1.03) for every 100 IU/day increment in vitamin D from diet plus supplement and 0.95 (0.90–1.00) for every 10 nmol/L increment in circulating vitamin D. The effect for every 10 mg/day increment was 0.96 (0.90–1.02) for vitamin E from diet plus supplement, 0.83 (0.72–0.95) from diet and 0.88 (0.67–1.15) from supplement, and the effect was 0.84 (0.76–0.94) for every 1 mg/dL increment in circulating α-Tocopherol and 1.22 (1.00–1.49) for every 0.1 mg/dL increment in circulating γ-Tocopherol. The observed association for vitamin D and vitamin E was significant among smokers but not among non-smokers. No significant association was found between vitamin C and risk of bladder cancer in the dose-response analysis. Based on the dose-response analysis, the risk of bladder cancer might be inversely associated with vitamin D and E (especially α-Tocopherol), but positively associated with γ-Tocopherol.

W orldwide, 386,300 new cases and 150,200 deaths from bladder cancer were estimated to occur in 2008, and the majority of bladder cancer occurs in males 1 . The American Cancer Society presents 4 recommendations to reduce cancer risk, including maintain a healthy weight, adopt a physically active lifestyle, consume a healthy diet with an emphasis on plant foods, and limit alcoholic consumption 2 . Smoking and occupational exposures are the major risk factors of bladder cancer in Western countries, whereas chronic infection with Schistosoma hematobium in developing countries 1 . Vitamin D receptors have been detected in superficial transitional cell carcinoma of the human bladder 3 , and vitamin D inhibits proliferation and induces apoptosis in human bladder tumor cells in vitro 4 . Vitamin C and vitamin E are also supposed to have a protective effect against bladder development through their actions as an antioxidant and free radical scavenger 5 . Many epidemiology studies have been conducted to assess the association of Vitamin C 6-20 , D 7,16,21-25 and E [7][8][9][10][12][13][14][15][16][17][18][19][26][27][28][29] with risk of bladder risk, but the results are not consistent. Therefore, we performed a dose-response metaanalysis to quantitatively summarize the evidence from epidemiological studies on the association of Vitamin C, vitamin D and vitamin E with risk of bladder cancer.
Moderate heterogeneity was found. Subgroup and meta-regression analysis showed that the following covariates did not contribute to the between-study heterogeneity (all P values . 0.05): publication year, location, number of cases, assessment method and study design; however, number of covariates adjusted ($4 or ,4) explained the heterogeneity (P 5 0.01).
No association was found in dose-response analysis, and the risk of bladder cancer for every 100 mg/day increment was 0.96 (0.90-1.02), P for nonlinearity 5 0.95 for vitamin C from diet plus supplement, 0.93 (0.82-1.05), P for nonlinearity 5 0.13 from diet, and 0.94 (0.85-1.03), P for nonlinearity 5 0.51 from supplement, and the effect was 0.96 (0.91-1.02), P for nonlinearity 5 0.76 for every 10 umol/L increment of circulating vitamin C (Table 1).
Vitamin D and bladder cancer (Table 1 and Figure 2-3). Although no association was found between bladder cancer and vitamin D from diet plus supplement, an inverse association was found or indicated with circulating vitamin D overall [0.75 (0.57-0.99), I2 5 51.7%], in cohort studies [0.82 (0.61-1.11), I2 5 46.3%] and case-control studies [0.55 (0.36-0.85), n 5 1]. No publication bias was found for vitamin D from diet plus supplement (P 5 0.41) and for circulating vitamin D (P 5 0.87), and no individual study had an excessive influence in sensitivity analysis, respectively. Because only 3 articles for vitamin D from diet plus supplement and 4 articles for circulating vitamin D were included, subgroup analysis was not conducted further.
Low to Moderate heterogeneity was found for vitamin E from diet plus supplement and from diet; however, subgroup and meta-regression analysis found that the above-mentioned covariates did not contribute to the heterogeneity (all P values . 0.05).
Association by smoking status and sex (supplementary table 4). In this analysis, we combined the result from diet and/or supplement and circulating levels (a-Tocopherol for vitamin E) to increase the number of studies in each subgroup analysis. For vitamin C, no association was found in subgroup analysis by smoking status [smokers (current or former) and non-smokers] and sex (men and women). For vitamin D, an inverse association was found among smokers [0.72 (0.55-0.93), I2 5 0.00%], but not among non-smokers. For vitamin E, an inverse association was found among smokers [0.77 (0.64-0.92), I2 5 68.8%], but not among non-smokers. An inverse association of vitamin E with risk of bladder cancer was found for men [0.68 (0.58-0.80), I2 5 0.00%] but not for women.

Discussion
To our knowledge, this is the first dose-response meta-analysis to quantitatively summarize the evidence on vitamin C, vitamin D and vitamin E and risk of bladder cancer. In this meta-analysis, vitamin D and vitamin E (especially a-Tocopherol) were found inversely associated with risk of bladder in a linear dose-response manner, and the association was stronger among smokers. Vitamin C might not be associated with risk of bladder cancer. However, c-Tocopherol might be positively associated with risk of bladder cancer. Categories of the three micronutrients (vitamin C, vitamin D, and vitamin E) levels differed between the included studies, which might complicate the interpretation of the pooled results across study populations with different categories (i.e. pooling the results comparing highest vs. lowest categories 31,32 ). In this respect, a dose-response meta-analysis with restricted cubic spline functions provides a solution to this problem from which the dose-response relationship (linear or non-linear) was first assessed and then a summary risk estimate can be derived for a standardized increase of the micronutrients levels 33 . We also assessed the associations between circulating vitamin C levels, circulating a-Tocopherol levels, circulating c-Tocopherol, diet and supplement vitamin D intakes and risk of bladder cancer, which are not assessed in the available meta-analysis 31,32 . In addition, this meta-analysis found that the observed association for vitamin D and vitamin E was significant among smokers but not among non-smokers, which was also not assessed in the available meta-analysis 31,32 .
Several plausible mechanisms have been proposed by which vitamin E and vitamin C may delay various steps in carcinogenesis through its actions as an antioxidant and free radical scavenger 5,34,35 , including limiting free radical induced DNA damage and the formation of carcinogens such as N-nitroso compounds, decreasing the concentration of the bladder carcinogen 3-hydroxanthranilic acid, inhibiting the carcinogenic effect of saccharin and cancer cell growth, and stimulating immune function and apoptosis. Vitamin D is not an antioxidant, but its activity in preventing cancer development involves regulation of adherence and signaling, inhibition of proliferation, differentiation, cell cycle stabilization, promotion of apoptosis, and anti-neoangiogenesis 36 . In this meta-analysis, the observed association was more prominent in smokers. This finding is in line with the hypothesis that smokers may benefit more from vitamin D and vitamin E intake than nonsmokers, because the high antioxidant content and anticancer effects of vitamins may reduce the oxidative damage caused by cigarette smoking 37 , and current smokers have significantly lower concentrations of serum vitamin E than nonsmokers 38 . There was a 32% reduction in prostate cancer incidence and a 41% reduction in prostate cancer mortality after 6.1 years of treatment of a-Tocopherol in the Alpha-Tocopherol, Beta Carotene Cancer Prevention Trial including 29,133 male smokers 39 , although this benefit was attenuated during posttrial follow-up 40 . However, the benefit on prostate cancer with vitamin E supplementation was not observed in the Physicians' Health Study II Trial 41,42 and the Selenium and Vitamin E Cancer Prevention Trial 43,44 , and both trials involved participants of very low levels of smoking. These findings on prostate cancer could also help understand the finding that vitamin E selectively protected against bladder cancer in smokers in this metaanalysis. In this meta-analysis, c-Tocopherol might increase the risk of bladder. Although the mechanism has yet to be determined, the residual confounding by smoking cannot be ruled out, because smoking decreases some plasma antioxidants but increases ctocopherol levels 45,46 . Although no association was found for vitamin C from diet plus supplement and vitamin C from supplement, an inverse association was found for vitamin C from diet. However, the inverse association was mainly caused by the results from case-control studies, and no association was indicated in cohort studies [1.02 (0.85-1.23)], and this was also the case for vitamin E. The distinction between the results from case-control and cohort studies might be attributed to the differences in study design considering prospective cohort studies do not suffer from recall bias and are anticipated to be less likely to have selection bias relative to case-control studies and are also believed to provide better evidence for causality in which vitamins intake precedes bladder incidence compared with case-control studies. For vitamin D, although no association was found with vitamin D from diet plus supplement, an inverse association was found with circulating vitamin D. This discrepancy might arise from the fact that self-reported dietary intake of vitamin D is a less accurate measure of vitamin D status than are circulating concentrations, because vitamin D is produced endogenously in response to sun exposure. Among the 4 articles included for circulating vitamin D, an inverse association was found or indicated in 3 articles 21,23,24 , but an increase (although not significant) risk of bladder associated with higher circulating vitamin D levels was found in 1 article 22 . As explained in the article 22 , this difference may be explained by the inclusion of women and nonsmokers in the current analysis, as a modest inverse association was also found when restricting the analysis to male smokers 22 . In addition, although the departure from a non-linearity was not significant (P for nonlinearity 5 0.78) that might be caused by the relatively small number of studies, the shape of the dose-response analysis suggested that the increased risk of bladder cancer was most pronounced at levels of circulating vitamin D less than 50 nmol/L. Importantly, the distribution of circulating vitamin D levels was higher in the PLCO study 22 , and few participants (about 43%) had circulating vitamin D levels , 50 nmol/L than those in the other studies (73% 24 and 74% 23 ). Therefore, the different distributions of circulating vitamin D levels across studies may make comparison between studies difficult and this could also contribute to the between-study heterogeneity. For vitamin E, although the observed association was mainly evident in case-control studies, circulating  a-Tocopherol was also found inversely associated with risk of bladder cancer, and no heterogeneity (I2 5 0.00%) was found across studies.
Other factors that might influence the observed association should be considered. The inverse association between circulating vitamin D and risk of bladder cancer was stronger among men with lower vitamin D binding protein (DBP) [low DBP: 0.47 (0.23-1.00), high DBP: 0.83 (0.40-1.75)] 23 , and among low-FGFR3 expressers 24 . In addition, season of blood collection, physical activity, a-Tocopherol supplementation and time from blood collection to case diagnosis were also found to influence the association 22,24 . No association was found between circulating vitamin C and bladder cancer by prognostic subgroups (aggressive and nonaggressive) 6 . The association between circulating a-Tocopherol and bladder cancer risk differed by vitamin E supplement [users: 0.48 (0.28-0.83), non-users: 1.76 (0.71-4.35)]; however, the effect of total vitamin E on bladder cancer risk was 0.49 (0.22-1.07) among non-users and 0.84 (0.44-1.59) among users of a-Tocopherol and beta carotene supplement in another study 12 . The limited data precluded a more robust assessment in this meta-analysis.
Other limitations should also be of concern. First, although we extracted the RRs that reflected the greatest degree of control for potential confounders, the extent to which they were adjusted and the possibility that the observed association was due to unmeasured or residual confounding should be considered. Therefore, the observed association for a-Tocopherol, c-tocopherol, and vitamin D should be confirmed by randomized controlled trials. One randomized controlled trial 42  . Second, misclassification and inaccurate measurement of vitamins intake should be of concern in observational studies. Third, although no publication bias was found, or the results did not change with the trim and fill method, validity of publication bias test should be questioned because of small number of studies included, especially for vitamin D, a-Tocopherol and c-tocopherol.
In conclusion, vitamin D and vitamin E (especially a-Tocopherol) might be inversely associated with bladder cancer risk, while c-tocopherol For inclusion, studies must fulfill the following criteria: (1) exposure of interest was Vitamin C, vitamin D or vitamin E; (2) outcome of interest was bladder cancer; (3) relative risk (RR) or odds ratio with 95% confidence interval (CI) was provided (we presented all results with RR for simplicity), or data available to calculate them; (4) conducted in humans; (5) for dose-response analysis, the number of cases and participants or person-years for each category of vitamins must also be provided (or data available to calculate them). The most recent study was included for duplicate publications.
Data extraction. The following data were extracted: first author, study design, cohort name and follow-up duration for cohort studies, publication year, country where the study was conducted, participants' mean age, number of cases and total participants, methods for the measurement of vitamins intake, RR (95% CI) and adjusted covariates. For dose-response analysis, the number of cases and participants (or person-years) and RR (95% CI) for each category of vitamins were also extracted. We assigned the median level of vitamins for each category to each corresponding RR estimate. If the highest category of the studies was open-ended, we assumed that the boundary had the same amplitude as the adjacent category.
Statistical Analysis. We first pooled the study-specific logarithms of RR for the highest versus lowest level of vitamins levels, using a random-effects model. Between studies heterogeneity was evaluated using I2 statistic, and I2 values of 25%, 50%, and 75% represent low, moderate, and high heterogeneity 47 , respectively. To explore sources of heterogeneity among studies, we conducted meta-regression analysis. A sensitivity analysis was conducted by omitting one study at each turn and recalculating the pooled estimates for the remainder of the studies to evaluate whether the results could have been affected markedly by a single study. Publication bias was assessed by Egger test, and the trim and fill method 48 was used to incorporate theoretical missing studies when publication bias was detected.
We then conducted a two stage random effects dose response meta-analysis 49 . First, a restricted cubic spline model was estimated with a generalised least squares regression with 3 knots at percentiles 25%, 50%, and 75% of the distribution of vitamins levels, then the estimated study-specific 2 regression coefficients (3 knots minus 1) were combined in a multivariate random-effects meta-analysis. A test for a non-linear relation was calculated by making the coefficient of the second spline equal to zero. We used STATA version 12.0 (StataCorp LP, College Station, TX) to analyse the data, and P , 0.05 was considered statistically significant. www.nature.com/scientificreports