Background
Vitamin C, an essential micronutrient abundant in fruits and vegetables, is essential for many physiological processes in humans [
1,
2]. Due to its beneficial effects on redox imbalance, epigenetic reprogramming, oxygen sensing regulation, host immunity, and collagen synthesis, all of which are involved in tumor angiogenesis, treatment evasion, or metastasis, many studies have suggested anticancer potential of vitamin C [
2‐
6]. Prior studies examining the therapeutic effect of vitamin C on cancer found that vitamin C at pharmacological concentrations from intravenous dosing, but not physiological vitamin C from oral dosing, exerted clinical benefits among cancer patients [
7‐
9]. However, whether lifelong exposure to high physiological concentration of vitamin C has a protective effect on cancers is still largely unknown.
Observational studies support an inverse correlation between circulating vitamin C and cancers [
10,
11]. However, the possibility of reverse causation could not be ruled out, as cancer-induced oxidative stress and reactive oxygen species formation might increase the consumption of antioxidants including circulating vitamin C, and cancer-related symptoms such as impaired taste, dysphagia, nausea, and vomiting could also contribute to an unbalanced dietary intake of vitamin C. Many prospective cohort studies have examined the associations between dietary or supplementary intake of vitamin C and risk of various types of cancers, but the conclusions were inconsistent [
12‐
17]. In contrast to observational studies, randomized controlled trials (RCT) of vitamin C supplements could potentially help establish the causal relationship. Several RCTs on this topic showed no effect of vitamin C supplementation on the risk of cancers, but the number of incident cases of site-specific cancers was small [
13,
18‐
21]. Therefore, whether the associations between circulating vitamin C and cancers are causal, and the direction of the causal associations (if any) are still unknown. Mendelian randomization (MR) analysis, exploiting inherent properties of common genetic variation for a modifiable environmental exposure of interest, has become a widely used approach to explore the potential causal relations between environmental exposures and diseases [
22]. By applying a bi-directional MR approach, on the one hand, we can explore whether circulating vitamin C casually affects the cancer risk, and on the other hand, we can examine whether the genetic predisposition of cancer risk causally influence the circulating vitamin C levels. To date, there has been no MR analysis addressing these questions.
In the present study, we applied a bi-directional MR approach to estimate the putative causal relationships between circulating vitamin C concentrations and risk of site-specific cancers, including lung and bronchus, breast, prostate, colon, and rectal cancers, which together represent half of the overall burden of cancer in Europe [
23]. To make comparison with prospective observational or interventional studies, we also conducted a meta-analysis to comprehensively summarize the results of prospective studies assessing the effect of vitamin C intakes on the cancer outcomes.
Discussion
This bi-directional MR analysis based on large-scale genetic consortia provided no evidence to support a causal association of circulating vitamin C concentrations with risk of cancer of the lung and bronchus, prostate, breast, colon, or rectum. Moreover, the meta-analysis of prospective studies of the associations of dietary or supplemental vitamin C intakes with cancer risk did not support the use of vitamin C supplements for prevention of the five cancers.
To the best of our knowledge, this was the first MR analysis to examine the potential bi-directional relationships between circulating vitamin C concentrations and site-specific cancer risk. Previous studies have explored the causal association of circulating vitamin C concentrations instrumented by only one genetic variant (rs33972313) with several health outcomes, including hyperuricaemia, ischemic heart disease, and Alzheimer disease, but not any cancer [
79‐
81].
The association of circulating vitamin C with cancer risk has been examined in several observational studies, most of which focused on total cancer [
82‐
84]. The results of meta-analyses including 5 studies involving 45,758 participants showed that each 50 μmol/L increase in vitamin C concentration was associated with a 26% lower risk of total cancer [
85]. Focusing on site-specific cancers, another systematic review reported a significant association of higher plasma vitamin C concentration with lower risk of breast cancer based on case-control studies [
86]. One explanation was that the cancer-induced oxidative stress and ROS formation may increase the consumption of vitamin C that acted as an antioxidant; thus, the observed associations in cross-sectional studies may be because of the reverse causality [
87].
Interestingly, our reverse MR analysis based on the primary approach (i.e., IVW method) found clues that smoking initiation might causally decrease circulating vitamin C or even mediate the association of lung cancer with circulating vitamin C. However, significant heterogeneity of the causal associations among different genetic variants was observed, and the median-based estimator showed non-significant associations. The sensitivity to the inclusion of invalid IVs may help explain the discrepancies. Specifically, the IVW estimate assumed that all genetic variants are valid instrumental variables, while the weighted median estimate can provide a consistent estimated of the causal effect when up to (not including) 50% of genetic variants are invalid [
45]. Therefore, the weighted median estimator is more conservative than IVW approach, especially when heterogeneity between the SNP-specific estimates presents. Nevertheless, other sensitivity analysis methods which are also robust to some violation of the instrumental variable assumptions (e.g., MR-PRESSO, MR-Robust, and MR-RAPS) still yield positive results for inferring causal effects of smoking initiation on circulating vitamin C. Notably, in the present study, smoking initiation served as a positive control for our reverse MR analysis, as previous observational studies showed smoking was associated with decreased vitamin C concentrations (39, 40). The facts that many of our analysis approaches successfully found clues of the association between smoking and circulating vitamin C may validate that our reverse MR approaches would detect signals of causal effects of cancers on plasma vitamin C, if the effect sizes were comparable with that of smoking.
As the circulating vitamin C was rarely measured in prospective cohort studies, most observational studies examined the preventative effects of vitamin C supplementation or dietary vitamin C against cancers [
88]. In a meta-analysis of 21 case-control and cohort studies, including 8938 lung cancer cases, the risk of lung cancer decreased by 7% for every 100 mg/day increase in vitamin C intake among men [
89]. However, another pooled analysis of women from five prospective studies in the UK Dietary Cohort Consortium did not find evidence of a significant association between vitamin C intakes and breast cancer incidence [
90]. Observational studies also yielded controversial results for colorectal cancer. A pooled analysis of prospective cohort studies found that high (>600 mg/day) versus low (≤100mg/day) vitamin C intake was associated with a 19% lower risk of colon cancer [
91], but no significant association was observed between vitamin C supplement use and colon cancer risk in a meta-analysis based on three studies conducted in Europe and the USA [
92].
In the present study, we performed a more comprehensive and up-to-date meta-analysis of prospective cohort studies and RCTs, involving up to 1,992,894 participants to summarize the potential effect of vitamin C intake on several common site-specific cancers. Our findings also support the abovementioned beneficial association of dietary vitamin C with lung cancer and null findings for breast cancer or colon cancer. However, only dietary vitamin C but not supplemental vitamin C intake exhibit potential protective association with lung cancer. Thus, compared with cross-sectional observational studies, prospective studies, and RCT studies tend to yield more consistent results with our MR findings. Given the null associations discovered in our MR analysis, the abovementioned controversial results based on observational studies raised concern about confounding, as the main sources of dietary vitamin C are fruits and vegetables which are also rich in polyphenols and fibers. Thus, circulating vitamin C might be just a biomarker of fruit and vegetable consumption [
24,
93]. Moreover, participants consuming high amounts of fruits and vegetables might be more health conscious.
Over the past decades, although a lot of studies supported the role of vitamin C in cancer prevention, the direction and magnitude of the association are uncertain and contradictory across observational studies [
88]. At present, in the context where pharmacological high dose of intravenous vitamin C alone or in combinations with clinically used drugs showed promising efficacy on treating several types of cancers, it is of great public health importance to clarify whether keeping high physiological circulating vitamin C levels through vitamin C intake has a beneficial effect on cancer prevention. The current study did not support a causal association of circulating vitamin C at physiological level with risk of five most common cancers in Europe. As circulating vitamin C cannot be synthesized by humans, and has to be obtained from diet [
93], our findings also imply that vitamin C supplementation is unlikely to be helpful for the prevention of the five most common cancers. Of note, our findings do not rule out the potential beneficial effects of fruits and vegetables, which besides vitamin C are rich in numerous phytochemicals and dietary fibers.
Our study has several strengths. First, in addition to the [SLC23A1]-rs33972313, which had long been used as the genetic instrument of circulating vitamin C, we further included another 9 up-to-date genetic variants identified in European populations to construct the genetic instrument. Second, our study is the first MR analysis on the relationship between circulating vitamin C and site-specific cancers, based on various large-scale cancer consortium data and the UK Biobank in European populations. The large sample size provides us with enough power to estimate the causal relationship between circulating vitamin C and site-specific cancers. Third, we summarized evidence from published prospective studies for vitamin C intake and incident site-specific cancers, which provides a comprehensive comparison with our MR findings.
This study has several limitations. First, due to limited available datasets for colorectal cancer and other different cancer subtypes, we cannot independently replicate the UK Biobank-derived findings on the colorectal cancer or explore the bi-directional relationships between circulating vitamin C and subtypes of different site-specific cancers, while different cancer subtypes may imply different etiology and pathogenesis. Second, this study can only investigate the potential effects of circulating vitamin C at physiological level on cancer prevention, but not the vitamin C exposure at a pharmaceutical level. Third, despite including data from very large genetic epidemiology networks, our study is not powered to detect very small effects. Lastly, our results are mainly based on participants of European ancestry and may not be generalizable to other ethnic populations.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.