Vitamin D Supplementation and Total Cancer Incidence and Mortality

A Meta-Analysis of Randomized Controlled Trials

N. Keum; D. H. Lee; D. C. Greenwood; J. E. Manson; E. Giovannucci


Ann Oncol. 2019;30(5):733-743. 

In This Article


In this meta-analysis of RCTs, we found that vitamin D supplementation was associated with 13% reduced cancer mortality over 3–10 years of follow-up, which confirms findings in the previous meta-analysis in 2014.[24] Corresponding to the 13% reduction in cancer mortality, there was a statistically significant 7% reduction in total mortality in our meta-analysis where cancer death (n = 1591) accounted for 33% of total death (n = 4872). The VITAL study was the only trial specifically powered to examine cancer mortality as a prespecified end point, yielding a contrast of nearly 30 nmol/l between intervention (105 nmol/l) and control group (73 nmol/l) with a vitamin D dose of 2000 IU/day.[28] Interestingly, in VITAL, increasing benefit over time was indicated, with RR reducing to 0.75 (95% CI, 0.59–0.96) upon excluding deaths occurring during the first 2 years of follow-up.[28] With additional restriction of analysis to cancer deaths confirmed by medical records or other adjudication beyond the National Death Index coding, RR further reduced to 0.63 (95% CI, 0.43–0.92).[28]

In contrast to the results for mortality, we found that supplementation of vitamin D did not lower cancer incidence compared to the control groups. The results remained null across diverse subgroups tested. Of note, while statistically insignificant, the direction of association slightly favored vitamin D supplementation (RR, 0.95; 95% CI, 0.83–1.09; P = 0.48) when attained 25(OH)D levels were > 100 nmol/l. Current epidemiologic evidence based on circulating 25(OH)D levels suggests a potential benefit only for colorectal and ovarian cancer incidence.[12,16] If any anticancer benefit of vitamin D is truly limited to specific cancer sites, then estimates for total cancer incidence may be weak and statistical significance lost. Alternatively, adult cancers typically develop over several decades and vitamin D could act on early stages of carcinogenesis. In this case, 3–10 years of follow-up periods of RCTs in this meta-analysis may have been insufficient to observe an effect. Specifically for colorectal carcinogenesis, vitamin D may act early given that 25(OH)D levels have been shown to be inversely associated with incidence of colorectal adenoma, a likely precursor for most colorectal cancers.[42]

Concern has been raised over the safety of 'high' circulating levels of 25(OH)D, generally considered to be those exceeding 100 nmol/l. In our meta-analysis conducted among trials whose achieved 25(OH)D in the intervention group exceeded 100 nmol/l, there was no statistically significant evidence for an increased risk of cancer incidence and mortality. In contrast, a decreased risk was suggested for cancer incidence (RR, 0.95) and cancer mortality (RR, 0.85), albeit not statistically significant. Particularly, two of the trials had baseline levels around 80 nmol/l and the intervention groups achieved average levels around 105 nmol/l with daily supplementation of 2000 IU vitamin D.[27,28] One of these, a relatively small study,[27] found a reduction in incidence (RR, 0.65; 95% CI, 0.42–1.00) and did not report on mortality. The other study (VITAL) showed a borderline nonsignificant lower incidence and an increasing reduction in mortality over time. These findings provide reassurance for the safety of attaining levels at least in the range of 100–120 nmol/l. This level is much higher than 50 nmol/l, a level currently considered sufficient by the Institute of Medicine but lower than the tolerable upper level of 125 nmol/l.[43] According to the Endocrine Society, vitamin D intoxication including hypercalcemia and kidney stones rarely occurs with levels below 375 nmol/l.[44]

Not only the attained levels of 25(OH)D but also supplementation regimen to reach the level may modify the effect of vitamin D supplementation on cancer outcomes. Six of the trials in this meta-analysis used daily dosing,[27,28,33,34,40,41] but one used weekly dosing,[39] two used monthly dosing,[29,37] and one used yearly dosing.[38] In a trial that tested a bolus of 500 000 IU of vitamin D3 per year, the median 25(OH)D level in the intervention group rose to approximately 120 nmol/l 1 month after dose, with ≥25% of the group reaching 150 nmol/l.[38] This trial was small and there were only 7 incident cancers in the intervention group versus 10 in the control group. Intermittent bolus dosing might yield nonphysiologic fluctuations in vitamin D.[45] Additionally, vitamin D intoxication only results from taking very large doses (e.g. 50 000–1 000 000 IU/day) of vitamin D for several months to years.[46] In this meta-analysis, the significant findings for cancer mortality were largely driven by trials that assigned daily intake of more modest levels of vitamin D supplements (Figure 3B).

The reason for the divergent findings for incidence and mortality of total cancer is not clear. Nonetheless, there are plausible mechanisms for vitamin D operating at multiple stages of carcinogenesis. Most relevant to the findings on cancer mortality, vitamin D may decrease tumor invasiveness and propensity to metastasize and influence immunomodulatory properties.[2,47] Although some of the RCTs had some participants already diagnosed with cancer at baseline,[29,38] the vast majority of the cancer mortality cases were those that were diagnosed and became fatal over the course of the study period. Thus, the potential benefit for vitamin D status on cancer mortality could operate during the prediagnostic stages by influencing late-stage tumor progression and metastatic seeding, during treatment by complementing or enhancing effects of therapies, or in post-diagnostic stages by improving survival. It is rational to design studies testing the survival benefit of enhancing vitamin D status after cancer diagnosis, though such designs may underestimate any benefit if some or all of the effects of vitamin D are in the prediagnostic stages.

Our study confirms the findings of the previous meta-analyses by our group[24] and Bjelakovic et al..[23] Our studies excluded trials of small size (≤10 participants with outcomes) and/or short term (≤1 year of follow-up) as explained in the Methods section. In contrast, the meta-analysis by Bjelakovic et al. included trials of any size and any duration. Despite differential inclusion/exclusion criteria, all meta-analyses consistently concluded that vitamin D supplements had no effect on cancer incidence but reduced cancer mortality by 12%–13%. Bjelakovic et al. noted that the positive finding with cancer mortality could have been due to random errors given too few participants examined (4 trials, 1192 cancer deaths, 44 492 participants) and the evidence of low quality. Our meta-analysis including recent trials of large size, testing higher vitamin D doses, and/or with cancer mortality set as a prespecified end point (5 trials, 1591 cancer deaths, 75 239 participants), provides more robust confirmation of these relationships and alleviates the earlier concerns.

In contrast, in the meta-analysis by Goulao et al., an inverse association with cancer mortality was suggested but not statistically significant (RR, 0.85; 95% CI, 0.70–1.04; 17 trials, 407 cancer deaths, 15 893 participants).[26] Unlike ours, Goulao et al. included trials of any size and testing any form of vitamin D supplementation (cholecalciferol, ergocalciferol, calcitriol, and vitamin D analog), and excluded trials if supplements (e.g. calcium) co-administered with vitamin D supplements were not provided in the control groups (e.g. the Women's Health Initiative[34]). They also included data on cancer incidence (e.g. Sanders et al.[38]) in the analysis of cancer mortality by counting cancer incidence as cancer death. Collectively, these may have affected statistical significance of the findings. Yet, in their sensitivity analysis additionally including after-study events (n = 160 deaths) from the RECORD trial, the results became marginally significant (RR, 0.85; 95% CI, 0.72–1.00).[26]

Our meta-analysis has several strengths. Although the time required to see an effect of vitamin D supplementation on cancer, if present, is not known, we were able to assess the influence on total cancer incidence and mortality over a period ranging from 3 to 10 years. At least for cancer mortality, which possibly may be influenced over a relatively short period, this length of time may be adequate. In addition, the studies varied in doses, but this allowed us to examine a potentially interesting and wide range of attained 25(OH)D levels, up to around 120–135 nmol/l of 25(OH)D. Unlike our earlier meta-analysis, we were able to separate effects of vitamin D from those of calcium and to evaluate higher doses such as 2000 IU/day of vitamin D.

Limitations of our meta-analysis also warrant considerations. Most of the trials were not initially designed to test the hypothesis that vitamin D influenced the risk of cancer incidence or mortality. Most of the study populations were composed of whites. Notably, in VITAL, which prespecified cancer end points and oversampled African-Americans, a suggestive inverse association was observed for cancer incidence in this racial subgroup (RR, 0.77; 95% CI, 0.59–1.01, 224 total cases), which is consistent with prior epidemiologic data.[48] Finally, because many trials did not provide data on site-specific cancers, we could not assess if a benefit of vitamin D supplementation was differential across cancer sites. Nevertheless, the epidemiologic data suggest potentially broad effects against many cancer types, though perhaps with stronger effects against some cancer types such as digestive tract malignancies.[49]

Despite the globally declining trend in cancer mortality over the past decades,[50] approximately 9.6 million cancer deaths were projected in 2018 worldwide.[51] In addition, a substantial percentage of the world population has vitamin D levels below 50 nmol/l, at least in winter.[52] Given the results from our meta-analysis, efforts to achieve circulating levels of 25(OH)D around 54–135 nmol/l may contribute to reducing cancer mortality. To consistently raise the level above 75 nmol/l, at least 1500–2000 IU/day intake of vitamin D may be required for adults as suggested by the Endocrine Society.[44] This requirement is higher than the recommended dietary allowance for vitamin D, which was established to avoid poor bone health: 600 IU for individuals aged 19–70 years and 800 IU for those aged over 70 years.[43] To determine the optimal level of vitamin D intake to reduce cancer mortality, future studies are warranted that examine the dose–response relationship between vitamin D supplementation and cancer death beyond the level of recommended dietary allowance.[43]