Association between Vitamin D Supplementation and Cancer Mortality: A Systematic Review and Meta-Analysis

Simple Summary It has been questioned whether vitamin D supplements can reduce the mortality and incidence of tumors. In this systematic review and meta-analysis of 12 randomized controlled trials with a total of 72,669 participants, vitamin D supplementation could not reduce the cancer mortality or cancer incidence. Our results suggest a reconsideration of the previous view that vitamin D supplementation could reduce overall cancer mortality is needed. Abstract Background: Vitamin D deficiency is related to increased cancer risk and deaths. However, whether vitamin D supplementation reduces cancer mortality remains unclear, and several randomized controlled trials yield inconsistent results. Methods: Medline, Embase, and the Cochrane Central Register of Controlled Trials were searched from their inception until 28 June 2022, for randomized controlled trials investigating vitamin D supplementation. Pooled relative risks (RRs) and their 95% confidence intervals (CIs) were estimated. Trials with vitamin D supplementation combined with calcium supplementation versus placebo alone and recruiting participants with cancer at baseline were excluded in the present study. Results: This study included 12 trials with a total of 72,669 participants. Vitamin D supplementation did not reduce overall cancer mortality (RR 0.96, 95% CI 0.80–1.16). However, vitamin D supplementation was associated with a reduction in lung cancer mortality (RR 0.63, 95% CI 0.45–0.90). Conclusions: Vitamin D supplementation could not reduce cancer mortality in this highly purified meta-analysis. Further RCTs that evaluate the association between vitamin D supplementation and total cancer mortality are still needed.


Introduction
In recent years, supplementation with vitamin D has been viewed as a potential strategy for preventing cancer [1][2][3]. Evidence from observational, preclinical, and clinical 2 of 12 studies strongly suggests that low 25-hydroxyvitamin D [25(OH)D] status is associated with the risk of developing colorectal cancer [4], breast cancer [5,6], bladder cancer [7,8], lung cancer [9,10], pediatric cancer [11], pancreatic cancer [12], and prostate cancer [13]. If adequate vitamin D concentrations reduce cancer risk, vitamin D supplementation may be a readily available, safe, and economical modality to reduce cancer incidence and mortality [2]. However, randomized controlled trials (RCTs) testing Vitamin D supplementation have been inconsistent, with one study finding that the incidence of cancer is reduced, while the other concluding that cancer mortality remains unchanged [14,15].
Previous systematic reviews found that vitamin D supplementation reduced cancer mortality [16][17][18][19][20]. However, these studies lacked enough detail on the associations for site-specific cancers and have not evaluated the quality of evidence using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) and an estimation of optimum sample size using trial sequential analyses (TSA). Since this study, the results of a new large randomized trial, the D-Health trial, changed the landscape of evidence, which suggested a trend of an increase in cancer mortality (hazard ratios 1.15, 95% CI 0.96 to 1.39) in a Vitamin D-replete Australian population.
Therefore, we performed a systematic review, meta-analysis, and trial sequential analyses to summarize the most recent evidence and assess the effect of vitamin D supplementation on cancer mortality.

Protocol and Guidance
We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines for reporting our systematic review [21]. This study was conducted according to the protocol registered in the PROSPERO database (CRD42019119639).

Eligibility Criteria
Studies that met the following criteria were included: (1) Population: adults (age ≥ 18) with any health condition; (2) Intervention: vitamin D supplements at any dose and for any duration. Trials of vitamin D plus calcium vs. calcium alone were considered vitamin D interventions; (3) Comparison intervention: placebo or no treatment. If other interventions were given (e.g., calcium), they had to be the same in all groups; (4) Outcome: cancer mortality or cancer incidence, with a follow-up of more than one year. The primary outcome was overall cancer mortality. Secondary outcomes were overall cancer incidence, site-specific cancer mortality, and incidence (i.e., breast, lung, prostate, colorectal). (5) Study design: randomized controlled trials (RCT), including quasi-randomized and cluster-randomized.
Studies were excluded if they were (1) case reports, case series, and observational studies, (2) trials of hydroxylated vitamin D or vitamin D analogs, (3) trials where all participants received vitamin D, (4) trials where all participants have cancer, (5) trials of pregnant or lactating women, (6) trials of critically ill patients, (7) trials with the total number of an outcome less than ten because of the small effect size and/or short follow-up time [16], (8) trials with vitamin D supplementation combined with calcium supplementation versus placebo alone because evidence showed calcium supplementation was associated with other unfavorable effects, including mortality [22], cardiovascular (e.g., myocardial infarction) [23][24][25], and breast cancer risk [26].

Data Sources and Search Strategy
An experienced research librarian (PX) developed and executed the search strategy. The electronic databases Medline, Embase, and Cochrane Central Register of Controlled Trials were searched (Table S1). We also checked the reference lists of eligible studies as well as screened scientific abstracts and relevant clinical trial registries (ClinicalTrials.gov and the World Health Organization International Clinical Trials Registry Platform). The last electronic search was performed on 28 June 2022. There were no restrictions on language.

Study Selection and Collection
Eight investigators were divided into two groups independently, and in duplicate screened the titles and abstracts of all identified studies using a priori selection criteria. They screened the full text of potentially relevant studies. Disagreements were resolved by discussion or, if needed, by consensus. Then, data were extracted from the included RCTs using a purpose-built spreadsheet containing the following information: Author names, publication years, the interventions in each arm, the number of total participants and events in each arm, baseline circulating 25(OH)D levels, primary outcome, and the follow-up time.

Assessment of Risk of Bias and Quality of Evidence
Two investigators independently performed quality assessments. The Cochrane risk of bias assessment tool was used to assess the risk of bias among the eligible trials. The quality assessment took random sequence generation, allocation concealment, blinding of participants, staff, and outcome assessors; incomplete outcome data; selective outcome reporting; and other potential biases into account. The risk of bias for each domain was graded as high, low, or unknown. The overall risk of bias for the study was reflected by the highest risk of bias for any criteria.
We used the Grading of Recommendation, Assessment, Development, and Evaluation (GRADE) approach (GRADE Pro-version 3.6 software) to generate the absolute and relative risk of the outcomes [27]. The GRADE guidance rated the quality of evidence and strength of recommendations depending on study design limitations, inconsistency, indirectness, publication bias, and imprecision in each result.

Data Synthesis and Analysis
The meta-analysis for the included studies were conducted using Review Manager (RevMan, version 5.4.1, the Nordic Cochrane Center, the Cochrane Collaboration) and the metafor package in R (version 4.0.1; R Project for Statistical Computing). All analyses were based on the intention-to-treat approach. The meta-analysis was conducted using random-effect models regardless of the level of heterogeneity. The risk ratio (RR) and 95% confidence intervals (CI) were calculated for dichotomous data. All tests of statistical inference reflect a 2-sided of p < 0.05. Statistical heterogeneity of the data was assessed by using the I 2 test [28]. We directly applied the random-effect models to our meta-analysis, considering the potential inconsistency in the included studies. If there are more than ten RCTs in a meta-analysis, publication bias was assessed by funnel plot techniques and the Egger and Begg tests.
Trial sequential analysis (TSA) was used to evaluate the statistical reliability of the pooled results and adjust for the random error risk using TSA software (version 0.9.5.10, beta) [29]. When the cumulative Z-curve entered the futility area or crossed the trial sequential monitoring boundaries, it suggested the anticipated intervention effect was sufficient and conclusive; thus, no further trials were needed. We applied TSA to keep an overall 5% risk of type I error and 80% power, assuming the intervention effect could reduce 20% relative risk.
Subgroup analyses were performed according to baseline vitamin D status (insufficiency and adequacy), type of vitamin D (vitamin D2 and vitamin D3), dose (≥2000 IU/d and <2000 IU/d), the dosing frequency of treatment (daily and intermittently), length of follow-up (≥3 years and <3 years), treatment duration (≥3 years and <3 years), and co-therapy status (without calcium and with calcium). We conducted post-hoc subgroup analyses based on the number of patients (≥2000 and <2000), number of events (≥200 and <200), mean age (≥70 years and <70 years), sex (female and both), and published year (before 2014 and in or after 2014).
Sensitivity analyses were conducted by (1) excluding trials with high or unknown risk of bias, (2) excluding trials with a high risk of bias of each domain, (3) excluding quasi-randomized or cluster-randomized trials, (4) excluding the largest trial, and (5) using fixed-effect models.

Characteristics of Included Studies
After identifying 29,776 articles, a total of 12 RCTs met the inclusion criteria ( Figure  1) [14,15,[30][31][32][33][34][35][36][37][38][39][40][41][42][43][44]. Characteristics of included studies are present in Table 1. Among these RCTs, 5 RCTs were conducted in Europe, 4 RCTs were in the United States, and 2 RCTs were in Australia, and 1 in New Zealand. Two RCTs only included female participants, while others included male and female participants. Mean circulating levels of 25(OH)D for the vitamin D supplementation group and placebo group ranged from 38 to 77 nmol/L.   Included RCTs were generally at low or unclear risk of bias. Risk-of-bias assessments are reported in Figures S1 and S2. Of the 12 included trials, 5 were low risk of bias, 6 were unclear risk, and 1 was high risk.

Cancer Mortality
Of these, 6 RCTs with a total of 61,882 participants were included in the meta-analysis for cancer mortality [15,37,38,40,41,45]. Pooled RR showed that vitamin D supplementation did not reduce cancer mortality risk (RR 0.96, 95% CI: 0.80-1.16, I 2 = 58%; Figure 2A). TSA analyses of cancer mortality showed that future trials are unlikely to change the pooled estimate ( Figure 2B) [15]. The funnel plot revealed no evidence of publication bias for the overall cancer mortality ( Figure S3).
Subgroup analyses demonstrated that only participants with daily dosing vitamin D have lower cancer mortality compared with those dosing vitamin D intermittently (RR 0.84, 95% CI 0.72-0.97, Table S2). All sensitivity analyses on cancer mortality were consistent with the main analyses, demonstrating vitamin D supplementation did not reduce cancer mortality (Table S3). For the site-specific cancer mortality, vitamin D supplementation significantly reduced lung cancer mortality (RR 0.63, 95% CI: 0.45-0.90, I 2 = 0%) while the results of other outcomes were consistent with the overall cancer mortality (Figure 3). Cancers 2022, 14, x 6 of 13 Subgroup analyses demonstrated that only participants with daily dosing vitamin D have lower cancer mortality compared with those dosing vitamin D intermittently (RR 0.84, 95% CI 0.72-0.97, Table S2). All sensitivity analyses on cancer mortality were consistent with the main analyses, demonstrating vitamin D supplementation did not reduce cancer mortality (Table S3). For the site-specific cancer mortality, vitamin D supplementation significantly reduced lung cancer mortality (RR 0.63, 95% CI: 0.45-0.90, I 2 = 0%) while the results of other outcomes were consistent with the overall cancer mortality (Figure 3).

Cancer Incidence
A total of 11 RCTs with a total of 51,369 participants were included in the meta-analysis for cancer incidence [14,30,32,34,35,37,38,[41][42][43]45,46]. No significant association of vitamin D supplementation with overall cancer incidence was found (RR 0.99, 95% CI 0.93-1.06, I 2 = 0%; Figure 4A). Similar results were also found in the analyses of site-specific cancer incidence, including lung, breast, prostate, and colorectal cancer ( Figure 4B). TSA analysis showed that the pooled sample size was sufficient and further trials are unlikely to change the result for cancer incidence ( Figure S4). The funnel plot, Egger and Begg's tests showed no evidence of publication bias for the overall cancer incidence (Eg-

Cancer Incidence
A total of 11 RCTs with a total of 51,369 participants were included in the meta-analysis for cancer incidence [14,30,32,34,35,37,38,[41][42][43]45,46]. No significant association of vitamin D supplementation with overall cancer incidence was found (RR 0.99, 95% CI 0.93-1.06, I 2 = 0%; Figure 4A). Similar results were also found in the analyses of site-specific cancer Cancers 2022, 14, 3717 7 of 12 incidence, including lung, breast, prostate, and colorectal cancer ( Figure 4B). TSA analysis showed that the pooled sample size was sufficient and further trials are unlikely to change the result for cancer incidence ( Figure S4). The funnel plot, Egger and Begg's tests showed no evidence of publication bias for the overall cancer incidence (Egger's test: p = 0.78, Begg's tests: p = 0.78, Figure S5).

Grading of Evidence
The GRADE summary findings for overall and site-specific cancer outcomes are shown in Table 2. The outcome of overall cancer mortality was found to be of moderate quality of evidence because of the inconsistency between studies, while the outcome of overall cancer incidence was deemed to be of high quality.

Grading of Evidence
The GRADE summary findings for overall and site-specific cancer outcomes are shown in Table 2. The outcome of overall cancer mortality was found to be of moderate quality of evidence because of the inconsistency between studies, while the outcome of overall cancer incidence was deemed to be of high quality.

Discussion
The findings of our meta-analysis indicate that vitamin D supplementation does not reduce cancer mortality or incidence overall. For site-specific cancer outcomes, we found that vitamin D supplementation could reduce lung cancer mortality. Furthermore, only participants with daily dosing vitamin D have lower cancer mortality compared with those dosing vitamin D intermittently (RR 0.84, 95% CI 0.72-0.97).
Compared with early meta-analyses that included trials with mixed interventions of vitamin D supplementation combined with calcium supplementation [16,19,47], we did not include these trials in the present study because evidence showed calcium supplementation was associated with other unfavorable effects, including mortality [22], cardiovascular (e.g., myocardial infarction) [23][24][25], and breast cancer risk [26]. In addition, we did not include RCTs with a follow-up time of less than one year as 25(OH)D levels need 3 to 6 months to attain homeostasis after vitamin D supplementation and cancer mortality of less than one year is mostly due to undiagnosed metastasis of cancer at the start of study [16].
Our findings on cancer morality were inconsistent with the recent meta-analyses conducted [20,48]. The most recent meta-analysis conducted by Guo et al. found vitamin D supplementation to reduce cancer mortality (RR = 0.88, 95% CI 0.8 to 0.96) while our results found a null association (RR = 0.96, 95% CI 0.80 to 1.16; p = 0.68). The main inconsistency mainly came from the results of the D-Health Trial, which was published recently [15]. In the D-Health Trial, Neale et al. found that the vitamin D supplementation arm has, although statistically insignificant, higher cancer mortality than the control group with a median 5.7 years follow-up (hazard ratios 1.15, 95%CI 0.96 to 1.39; p = 0.13) [15]. Our subgroup analyses found that participants with daily dosing vitamin D have lower cancer mortality compared with those dosing vitamin D intermittently (RR 0.84, 95% CI 0.72-0.97), which might partly explain the difference between the results of the D-Health Trial (which used monthly dosing) and other large RCTs including VITAL and RECORD trial (which used daily dosing) [37,38]. The results were consistent with Keum et al., which also found daily dosing instead of intermittent dosing of vitamin D, could reduce total cancer mortality [20]. Daily vitamin D might be a more effective way to increase 25(OH)D than intermittent dosing [49]. In addition, according to our TSA analyses, under the assumption of 20% relative risk reduction to maintain an overall 5% risk of type I error and 80% power, the D-Health Trial was an essential update to previous results of meta-analyses which had been underpowered for cancer mortality. By adding the results of the D-Health trial, our present meta-analysis results have reached the required information size.
In the site-specific cancer analysis, we observed that vitamin D supplementation was associated with lower lung cancer mortality. These results were partly consistent with previous in vitro and in vivo studies, which have shown vitamin D could inhibit tumor growth, and diet-derived vitamin D might be a direct therapeutic agent in the EGFR-mutant lung cancer [50,51]. It has also been found that calcitriol, the active form of vitamin D, could inhibit lung cancer growth, metastases, and recurrence in mouse models [52,53]. Some epidemiological evidence, including a dose-response meta-analysis of prospective cohort studies, also supported our results that higher plasma 25-hydroxyvitamin D concentrations are associated with lower lung cancer mortality [54]. However, several studies have reported the opposite results, showing higher lung cancer mortality in participants with higher circulating 25-Hydroxyvitamin D [55,56] or a lack of difference [57,58]. Thus, our findings regarding lung cancer should be interpreted with caution because of the limited number of studies and sample size. Thus, further RCTs or large observational studies may be warranted.
We conducted the present review based on a protocol published in the PROSPERO database, which used a rigorous methodological approach based on the Cochrane Handbook. The strengths of this study included a rigorous assessment of the quality of evidence of included studies and the minimum information size was satisfied according to TSA.
Limitations should also be noted. First, our meta-analysis was based on published trials that reported cancer mortality. However, most trials of vitamin D supplementation did not include cancer mortality as an outcome, which might lead to bias of selective reporting. Second, the pooled sample size was large enough to evaluate the associations of vitamin D supplementation with total cancer mortality; however, the sample size is insufficient for specific subtypes of cancer. Additionally, studies included in our metaanalysis were highly purified compared with other meta-analyses, which may introduce additional bias.

Conclusions
The results of the current meta-analysis may have significant implications for clinicians and researchers. We suggest a reconsideration of the previous view that vitamin D supplementation could reduce overall cancer mortality. Different dosing frequencies might be necessary for future studies investigating the relationships between vitamin D supplementation and cancer mortality.  Table S3: Sensitivity analyses of cancer mortality; Figure S1: Risk of bias summary: review authors' judgements about each risk of bias item for each included study. Figure S2: Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies; Figure S3. Funnel plot of cancer mortality; Figure S4. TSA analyses for cancer incidence; Figure S5. Funnel plot of cancer incidence.

Conflicts of Interest:
The authors declare no conflict of interest.