Obesity and cancer: the role of vitamin D

A comprehensive literature review of all published meta-analyses on the association between obesity and cancer was carried out. We used computerised search databases (PubMed search followed by an Embase search) to identify full text and abstracts focused on human subjects and published in English language within the last fifteen years. Searches were conducted both with and without MeSH terms for “obesity”, “cancer” and “meta-analysis”. This search was repeated for individual cancer types: “breast”, “colorectal”, “melanoma”, “oesophageal”, “liver”, “lung”, “ovarian”, “endometrial”, “prostate”, “pancreatic” and “kidney” cancer. Although lung cancer may not be the obvious cancer to investigate in the context of obesity [17, 18], some studies [19, 20] reported a positive association while others are inconclusive or conflicting. Hence, lung cancer was also included in this literature review.

A comprehensive literature search of all meta-analyses conducted on the association between vitamin D and cancer was performed. We used computerised search databases (PubMed search followed by an Embase search) to identify full text and abstracts focused on human subjects and published in English language within the last fifteen years. Searches were conducted both with and without MeSH terms for “vitamin D”, “cancer”, “vitamin D receptor”, “polymorphism” and “meta-analysis”. This search was repeated for specific cancer types: “breast”, “colorectal”, “melanoma”, “oesophageal”, “liver”, “lung”, “ovarian”, “endometrial”, “prostate”, “pancreatic” and “kidney” cancer. Moreover, we also searched clinicaltrials.gov for clinical trials focused on “vitamin D supplements” and “cancer” or “neoplasm” [21].

Thirty-two meta-analyses were identified from our literature search on obesity and cancer (Table 1). More specifically, all seven meta-analyses on colorectal cancer showed a positive association between BMI and colorectal cancer risk [22‐28]. When looking at site-specific cancer within colorectal cancer, BMI was only significantly associated with rectal cancer in males. Also upper gastro-intestinal cancers (oesophageal, oesophageal gastric junction, gastric and gall bladder cancer) were positively associated with obesity [29‐32]. The strongest link was seen for oesophageal cancer with over a two-fold increased risk reported [29, 32]. All four meta-analyses on liver cancer reported an increased risk with increasing BMI [33‐36], whereas the lung cancer meta-analysis reported an inverse association with obesity (RR: 0.79; 95% CI: 0.73-0.85) [20]. Meta-analyses on pancreatic cancer reported a positive association with obesity [37‐40], which is parallel to the conclusions that can be drawn for kidney cancer[41, 42]. For prostate cancer[43], a protective effect of obesity was reported for localised disease, whereas obesity was positively associated with metastatic disease [44]. The meta-analysis on bladder cancer reported a positive association even when adjustment for smoking was performed [45]. Some variation was observed for breast cancer depending on menopausal status and breast cancer subtype [46, 47]. A positive association between obesity and breast cancer was more distinct among postmenopausal women [48]. The meta-analysis on ovarian cancer reported a positive association with obesity, with no difference in the histological subtypes of ovarian cancer studied [49]. As for the majority of other cancers [50], there was also a positive association found for endometrial cancer[51]. However, this meta-analysis included some studies which used waist circumference as a measure of obesity instead of BMI [51]. The meta-analysis on melanoma reported a positive association in men (RR: 1.31; 95% CI: 1.19-1.44), but not in women (RR: 0.99; 95% CI 0.83-1.18) [52].

The initial PubMed search produced a total of 356 (TS) and 352 (DC) papers. Further assessment of abstracts and papers based on the above-defined inclusion criteria (Figure 2) resulted in inclusion of 12 studies for primary data analysis (three cohorts, two case–control and seven cross-sectional studies) (Table 2).

The random effects analyses showed a pooled relative risk of 1.52 (95% CI: 1.33-1.73) for the association between obesity and low vitamin D status (Figure 3). The I² statistic suggested heterogeneity (I² = 89.4%). There was no difference between those studies looking at children and adolescents combined and those looking at an adult population (RR: 1.52; 95% CI: 1.04-2.26 and 1.53; 95% CI: 1.31-1.80, respectively).

Beggs and Eggers test was used to evaluate publication bias with the funnel plot suggesting the study by Goldner et al. to be an outlier [53] (Results not shown). We performed a sensitivity analysis by excluding this study from our analysis. The pooled estimate of RR did not change drastically, although the link was strengthened to some extent (RR: 1.34; 95% CI: 1.15-1.57).

From the literature search, we identified 21 meta-analyses on the association between circulating vitamin D levels and cancer risk (Table 3), showing different results for different types of cancer. We found 34 clinical trials investigating the effect of vitamin D supplementation on cancer (Table 4) [21]. From these, two studies were terminated, 18 are active, 13 have been completed, and one has an unknown status.

All six meta-analyses on colorectal cancer reported that circulating vitamin D levels were inversely associated with cancer risk [54‐59]. A pooled analysis from multiple cohort studies on pancreatic cancer, suggested no significant association for participants with low vitamin D levels. Those with vitamin D levels ≥100 mmol/L were at a statistically significant twofold increase in pancreatic cancer compared to those with normal vitamin D levels [60]. The pooled analysis for kidney cancer only found a statistically significant decreased cancer risk among women when vitamin D levels was ≥75 nmol/L [61]. In contrast, all three meta-analyses on prostate cancer found no evidence for an inverse association with vitamin D levels [58, 62, 63]. Results from four out of five meta-analyses showed an inverse association for breast cancer, with the highest quartile of vitamin D levels decreasing the risk of breast cancer [58, 64‐67] compared to the lowest quartile. However, it is interesting to note that case–control studies generally report an inverse association, whereas nested case control studies reported null-findings [58, 64‐67]. The meta-analysis on ovarian cancer reported a non-statistically significant inverse association with high serum vitamin D levels [68]. Finally, the meta-analysis on total cancer incidence and mortality reports a moderate inverse relationship with circulating vitamin D concentrations [69].

From the 13 completed clinical trials evaluating the effect of vitamin D supplementation on cancer risk, only two have reported results [70, 71]. One randomised trial focused on risk of colorectal cancer over a period of seven years in a double-blinded, placebo-controlled setting, where one group of postmenopausal women received calcium and vitamin D3 supplementation and the other group received placebo [70]. The study found no statistically significant effects of calcium or vitamin D3 supplementation on the incidence of colorectal cancer. The other completed trial had a similar design, but focused on risk of all cancers in postmenopausal woman receiving 1400–1500 mg supplemental calcium/d alone, supplemental calcium plus 1100 IU vitamin D3/d, or placebo during a follow-up of four years [71]. In contrast, this trial found that those women on vitamin D supplementation had a lower risk of cancer, compared to the placebo group when the analysis was confined to cancers diagnosed after the first 12 months (RR: 0.23; 95% CI: 0.09-0.60). No statistical analyses were performed for specific types of cancer [71].

The majority of meta-analyses included in our review reported positive associations between obesity and risk of cancer, showing that the strength of this association varies between cancer sites, sex, and in breast cancer, the menopausal status. The World Cancer Research Fund (WCRF) suggests that obesity is associated with increased risk of oesophageal adenocarcinoma, pancreatic, colorectal, postmenopausal breast, endometrial and renal cancer [73].

There are several molecular mechanisms suggested to explain the increased risk of cancer in obese people. The most commonly postulated being the “insulin–cancer hypothesis” [74], suggesting that obesity results in chronic hyperinsulinaemia. Prolonged hyperinsulinaemia leads to raised insulin like growth factor 1 (IGF-1) levels, which are known to produce cellular changes leading to carcinogenesis via increased mitosis and reduced apoptosis. Secondly, in hormonally-driven cancers, such as endometrial and post-menopausal breast cancer, the increased risk may be partly explained by an increase in circulating levels of sex steroid hormones. In the post-menopausal state, the majority of oestrogen is derived from adipose tissue rather than from the ovaries, potentially explaining the discrepancy between pre- and post-menopausal women. Thirdly, obesity is thought to result in a state of chronic inflammation and this has an effect on the cytokine microenvironment. These changes lead to an increase in tumour cell motility, invasion and metastasis. The change in the cytokine milieu has been suggested as a possible mechanism in several cancers including post-menopausal breast cancer [75].

The majority of the meta-analyses in our literature review included a substantial number of studies, with consistent study design. However, the meta-analysis on endometrial cancer [51] only included five studies of which some used other markers than BMI to define obesity (i.e. waist circumference). None of the studies to date included additional information on vitamin D status.

In summary, there is consistent accumulating evidence for an association between obesity and risk of certain cancer with several suggested molecular mechanisms that can potentially explain these raised risks. However, the role of vitamin D is not addressed in detail in these studies.

To our knowledge this is the largest meta-analysis to date on the association between circulating vitamin D levels and obesity. The pooled estimates suggest an inverse relationship between vitamin D and obesity.

The possible relationship between vitamin D and obesity was firstly described by Rosenstreich et al. in 1971 [76], who suggested that adipose tissue serves as a large storage site for vitamin D to protect against toxicity from vitamin overdose. The inverse association between obesity and vitamin D is thus thought to be a result of increased metabolic clearance in adipose tissue [77]. However, it has recently been suggested that this association is more complex since bariatric surgery solely has temporary effect on improving circulating vitamin D levels [78]. It is also postulated that obese individuals are less likely to engage in outdoor physical activity and dress differently than non-obese individuals, hence leading to decreased sun exposure [79, 80]. Wortsman et al. have shown that the bioavailability of cutaneously synthesised vitamin D decreases by >50% in obese people [81]. Even though exposure to sunlight is the main source of vitamin D synthesis [82, 83], its ultraviolet radiation is also known to increase risk of developing malignant melanoma of the skin [83]. In general, epidemiological studies have described that the highest incidence of melanoma is seen in fair-skinned population living closest to the equator [82, 84]. Within this population the highest risk is seen in those who report sunburn or intermittent sun exposure [85‐87]. Furthermore, Newton-Bishop et al. found that low vitamin D levels were associated with a thicker and more aggressive melanoma, with a poorer outcome [88]. Overall, vitamin D levels are known to be lower in obese individuals and several studies have observed that increased BMI is associated with an increased risk of developing melanoma [89‐91]. However, to date it has not been clarified whether the risk of melanoma in obese individuals is due to lower vitamin D levels associated with high BMI or less sun exposure.

Furthermore, certain vitamin D receptor (VDR) polymorphisms are associated with obesity [92, 93]. Upon ligation with calcitriol, the VDR couples with the retinoid X receptor (RXR) forming the VDR/RXR complex. This complex then further recruits other molecules, and finally binds to vitamin D response elements in the nucleus to activate the transcription of vitamin D target genes [92, 93]. Preclinical studies report expression of human VDR in mature mice adipocytes. This results in increased adipose mass and decreased energy expenditure [94] and expression of VDR in preadipocyte cell lines; this inhibits adipocyte differentiation [95]. A positive association between obesity and the Taq1 gene was also reported in a Greek case–control study [96].

In contrast, some suggest that low vitamin D itself promotes obesity. Kong and Li demonstrated that vitamin D levels may block the expression of downstream adipocyte components such as fatty acid synthase, which consequently suppresses adipogenesis [97]. One interventional study investigated the effects of vitamin D on weight loss and showed that those with higher baseline vitamin D experienced a greater degree of weight loss than those with lower baseline vitamin D [98].

In conclusion, our meta-analysis reports a modest inverse association between obesity and low vitamin D levels. The underlying biological mechanisms are unknown. The majority of studies point towards the hypothesis that, vitamin D stored in fat tissue increases local vitamin D concentrations causing activation of the VDR in adipocytes. This may lead to low energy usage and further promotion of obesity [94].

In this literature review only those meta-analyses focusing on colorectal cancer found a consistent inverse association between circulating vitamin D levels and cancer risk [54‐59]. In contrast, of the two completed clinical trials for which results are published to date, one showed no effect on colorectal cancer risk and one showed a protective effect for all cancer risk [70, 71].

A protective effect of vitamin D in colorectal cancer was first reported by Garland and Garland [99]. Despite the inconsistency in epidemiological findings [54‐61, 64‐68], there is preclinical evidence linking vitamin D and cancer, suggesting that vitamin D has anti-proliferative effects via mechanisms such as G0/G1 arrest, differentiation, and induction of apoptosis [100].

More specifically, it is suggested that vitamin D has anti-tumour effects through its binding with the VDR. Several animal and cell culture models showed that VDR plays a key role in the anticancer effects of circulating vitamin D [9‐11]. For instance, it has been reported that downregulation of VDR correlates with poor prognosis in colon cancer [101], suggesting that some of the discrepancy observed in epidemiological studies can be explained through gene polymorphisms [102]. VDR polymorphisms have been associated, both positively and inversely, with risk of cancer depending on the type of cancer, polymorphism, and other factors such as sun exposure or circulating vitamin D levels [8, 103]. For instance, a meta-analysis for prostate cancer found no association between the recessive genotype and the risk of prostate cancer relative to the dominant genotype of Fok1 [104]. To date, the importance of the role of VDR polymorphisms in carcinogenesis is unclear [101], but when analysed with additional factors like VDR haplotype combinations, vitamin D serum levels and other confounders, polymorphisms have been shown to play an important factor in cancer prognosis [105‐107].

Interestingly, several parts of the immune system (i.e. macrophages, neutrophils, or natural killer cells) also express the VDR, but the related effects remain to be elucidated [12]. It has for instance been suggested that vitamin D can weaken antigen presentation by dendritic cells, which results in suppression of their capacity to activate T cells. Furthermore, activation of the VDR promotes a shift towards T helper 2 responses, leading to antibody-mediated immunity and promoting a chronic state of disease [108, 109]. Hence, it is plausible that vitamin D has an immunosuppressive effect, which leaves tumour cells without the necessary immunosurveillance to stop them from proliferating. Thus, this suggests that the above-described potential anti-cancer effect of vitamin D most likely occurs through different mechanisms than the immune system. Most literature to date on vitamin D and the immune system has focused on autoimmune and infectious diseases, with scarce literature focusing on cancer.

In 2008, the International Agency for Research on Cancer concluded that evidence for an association between vitamin D and cancer was inconclusive, and highlighted the need for a clinical trial with specific focus on vitamin D and colorectal cancer [101]. The inconsistent findings from two trials for which results are published to date [70, 71] may be explained by the lower dose of vitamin D in the first study (i.e. 400 IU vs. 1100 IU). Furthermore, baseline vitamin D levels were lower in the second trial (i.e. 42 nmol/L vs. 71.8 nmol/L). Thus, despite the large amount of preclinical studies trying to establish a link between vitamin D and cancer, the contradictive findings from large epidemiological studies indicate that it is prudent to wait for more results from the 34 currently on-going trials to draw a reliable conclusion.

When assessing the three conditions required for vitamin D to be a mediator we found only partial fulfilment [110]. The literature shows consistent evidence for an association between vitamin D and obesity. However, there was lack of studies showing a consistent link between vitamin D and cancer after adjustment for obesity. To date, only two clinical trials have published their results with inconsistent findings. Furthermore, to our knowledge no study has assessed the mediation effect of vitamin D by quantifying the extent of obesity on cancer, which could be explained by a potential mediator.

Several other difficulties occur when assessing the mediation effect of vitamin D in the context of obesity and cancer. Dichotomisation of vitamin D exposure (low versus normal) could lead to misclassification in exposure levels. Those with extreme high values of vitamin D may have been included in the “normal” group. Hence, bias can occur when there is misclassification of the mediator [13]. Studies to date have used different cut-offs to define vitamin D deficiency, which can potentially be addressed with a dose–response assessment of this mediator. Unfortunately, it was not possible in this meta-analysis to use dose–response data [111] as the number of relevant studies available to date was small, and the qualitative classifications of circulating vitamin D levels varied. Furthermore, the effects of dietary supplements on circulating vitamin D levels needs to be accounted for, and very few studies took this into account [112]. The latter does not necessarily affect blood levels of vitamin D, but it may influence the biological role of vitamin D. Within-person variation may also affect the results of our meta-analysis, as only one measurement in time might not be representative of a person’s average vitamin D level. Moreover, it is important to address potential important confounders for the different associations studied [13, 72]. For instance, when evaluating the effect of the mediator (vitamin D) on the outcome (cancer), one has to consider age, sex, use of dietary supplements, ethnic variations, calcium intake and sun exposure [113], as they may be effect modifiers for the association between obesity and vitamin D. It has been argued that it is also important to address the strength of the association between these mediator-exposure confounders and both the exposure (obesity) and the outcome (cancer) [13]. With respect to the mediation effect of vitamin D, one also needs to evaluate whether there is a potential interaction affecting the link between the exposure (obesity) and the mediator (vitamin D) [13]. Effect modification may also have an effect on the link between the mediator (vitamin D) and the outcome (cancer), as is suggested by the different polymorphisms affecting the VDR [8].

Additionally, the current systematic literature reviews are prone to the heterogeneity related to observational studies. For example, for the studies focused on vitamin D and obesity the included studies recruited adults residing in a particular town [114], from medical centres [115‐117], from sample surveys [2, 118], and those undergoing bariatric surgery [53]. Children were recruited from schools [119, 120], hospitals [121, 122], and sample surveys [123]. Vitamin D levels were measured using either an immunoassay [2, 53, 114, 115, 118‐121, 123] or a high-performance liquid chromatography [116, 122]. Anthropometric data, including weight, height, waist circumference and BMI, were recorded for all participants [119, 120, 122]. Furthermore, information on dietary, physical activity and sun exposure were collected either by parental report during in-person interviews [123], and interview-administered questionnaires [114, 122]. These questionnaires may be subject to recall bias, as participants may not always give accurate data [124, 125] due to the time interval, degree of detail, personal characteristics, significance of events, social desirability or interviewing techniques [126]. Furthermore, despite proven validation, many food questionnaires have been found to be imprecise [127], due to the fact that people tend to answer these type of questions based on what their dietary routines are, more than on the real consumption. These memories are usually influenced by sex, age, and concerns about weight or body image [128].