Vitamin D and Prevention of Colorectal Adenoma: A Meta-analysis

Colorectal cancer is one of the leading causes of cancer mortality in the United States and in many western countries. Adenomas are precursors to the majority of colorectal cancers. Because adenomas can be detected decades before development of cancer, they can serve as a predictive indicator for cancer (1, 2). The malignancy transformation rate for adenomas ranges from 5% for small adenomas to 50% for villous adenomas over 2 cm in diameter (3, 4). For some exposures, such as cigarette smoking, the latency period for colorectal cancer can be as long as 35 to 40 years, but a much shorter latency can be found for adenomas (2).

A role for vitamin D in prevention of colorectal cancer was first hypothesized in 1980 based on ecologic patterns of colorectal cancer mortality by latitude as an indication of regional variation in solar radiation, which is required for vitamin D synthesis (5). It is now known that the vitamin D receptor and the enzyme 1α-hydroxylase, which converts 25-hydroxyvitamin D [25(OH)D] to the active 1,25(OH)₂D, are expressed in the colon, rectum, and nearly all tissues (6, 7). When activated by 1,25(OH)₂D, the vitamin D receptor is a transcription factor that has been shown to decrease epithelial cell proliferation and to induce differentiation and apoptosis in colorectal neoplasia (8-10).

An inverse linear dose-response relation has been shown for both dietary intake of vitamin D and circulating levels of 25(OH)D with colorectal cancer risk (11). Recent studies including a meta-analysis of vitamin D exposure and colorectal cancer risk estimate a 50% decrease in colorectal cancer incidence for individuals with ≥1,000 IU/d vitamin D intake or ≥33 ng/mL circulating 25(OH)D (11, 12). Here, we conducted a meta-analysis to estimate the overall summary effect of vitamin D intake and circulating (plasma or serum) 25(OH)D status on adenoma risk and recurrence.

Studies were identified by searching PubMed and MEDLINE before December 2007 for the terms vitamin D, cholecalciferol, calcifediol, calcidiol, and calcitriol in combination with colorectal adenoma or adenomatous polyp. Other forms of vitamin D were identified and searched using the Medical Subject Heading feature in PubMed. References from relevant articles were crosschecked to identify any studies that were missed in the database search.

For inclusion in this analysis, studies must have provided information in quantiles, or at minimum, contrasted high versus low vitamin D, and included 95% confidence intervals (95% CI). Individual authors were contacted for further data when these criteria were not met. Three of five authors responded; data from these authors were included in this analysis, whereas nonrespondents were recontacted twice by E-mail before being excluded. Further, studies must have adjusted for age and sex or concluded no differential effect by sex before combining results (13). When available, odds ratios (OR) adjusted for additional factors were selected.

Calcium was an important covariable given its independent effect on colorectal cancer and adenomas and that calcium absorption is strongly influenced by vitamin D status. Most studies did not adjust or stratify by calcium. Thus, the modifying effect of calcium could only be analyzed in four studies that examined calcium through clinical supplementation (14) or stratification by calcium status (15, 16, 25).

Vitamin D intake was assessed through dietary intake and total intake. Dietary intake encompassed food sources captured by self-report. Total intake included dietary as well as supplementary sources such as multivitamins.

Circulating 25(OH)D, the stored form of vitamin D, was used as a marker of vitamin D status. The active form 1,25(OH)₂D is tightly regulated by the body, and circulating 1,25(OH)₂D does not reveal vitamin D status until extreme deficiency or excess ensues. Thus, we considered 25(OH)D as a marker of vitamin D status in this analysis.

The OR and 95% CI for high versus low vitamin D exposure groups were extracted for each study. Peto's Assumption-Free Method was used to obtain the summary OR (17). If a study presented more than one OR for a result stratified by covariates, the stratified results were pooled into one summary OR before being entered into the analysis.

To establish the appropriate weighing of each study toward the summary OR, the 95% CI was used to compute the SE for each logarithm of the OR (equal to the difference between the upper and lower bounds of the 95% CI divided by 3.92). The square of the SE was the estimated variance of the log OR.

Fixed- and random-effects models were considered. The random-effects measure was used because it enabled the individual studies to estimate a different effect (18) by assuming the effect within each study varies around an overall average effect. Funnel plots were analyzed to assess for publication bias.

Homogeneity among studies was assessed using DerSimonian-Laird (19) to determine the appropriateness of combining ORs from individual studies. In the case of heterogeneity, a sensitivity analysis was done. Heterogeneity was additionally assessed through the I² statistic. I² represents the proportion of total variation across study estimates due to heterogeneity. It is a transformation of the H statistic (square root of the χ² heterogeneity statistic divided by its degrees of freedom; ref. 20).

Calculations were done using Review Manager version 4.2 (RevMan; Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2003).

Seventeen case-control, cohort, and cross-sectional studies that examined the relation between intake or circulating vitamin D and colorectal adenoma were identified and described in this analysis. A summary of included studies examining total and dietary vitamin D intake and risk of colorectal adenoma is included in Table 1 and those of circulating 25(OH)D in Table 2.

Twelve of 13 identified studies met criteria for inclusion in the analysis. One study was excluded because the results were stratified to the extent that selection of one subgroup would not be generalizable, and insufficient data were provided for the subgroups to be combined and a summary OR created (21). Correspondence was made with the first author, but the necessary stratification data for our analysis could not be obtained.

A statistically significant inverse association of high vitamin D intake versus low intake was reported in two studies (16, 22) and an additional six studies found inverse associations that were of marginal significance or nonsignificance (13, 15, 23-26). Four studies reported no association or possibly a slightly but nonsignificantly increased risk of colorectal adenoma with increased vitamin D intake (27-30).

The overall estimate using the random-effects model was a marginally significant 11% decreased risk of colorectal adenomas among individuals with high vitamin D intake compared with low vitamin D intake (OR, 0.89; 95% CI, 0.78-1.02; Fig. 1). The DerSimonian-Laird test for heterogeneity among the studies was not significant (χ² = 14.57; df = 11; P = 0.20; I² = 24.8%). Fixed-effects models yielded findings of the same magnitude (OR, 0.90; 95% CI, 0.81-1.00).

Included studies comprised seven cohort studies (16, 22, 25-28, 30) and five case-control studies (13, 15, 23, 24, 29). Results were similar for hospital- and population-based case-control studies. The overall suggestive decreased risk of colorectal adenomas among high versus low dietary vitamin D intake was shown with comparable magnitude by case-control studies (OR, 0.86; 95% CI, 0.67-1.09) and cohort studies (OR, 0.91; 95% CI, 0.76-1.08; Table 3). Both tested null for heterogeneity, although there was greater homogeneity among case-control studies (DerSimonian-Laird χ² = 2.58; df = 4; P = 0.63; I² = 0%) than cohort studies (DerSimonian-Laird χ² = 11.85; df = 6; P = 0.07; I² = 49.4%).

Of the vitamin D intake studies, six examined vitamin D from dietary sources (13, 15, 22, 25, 27, 30), four studies examined vitamin D supplements (15, 23, 25, 30), and eight studies examined dietary plus supplemental intake (total intake; refs. 15, 16, 24-26, 28-30). An inverse but not statistically significant association was observed among studies that examined dietary vitamin D intake alone (OR, 0.90; 95% CI, 0.79-1.03; Table 3). The dietary intake studies were consistent and homogenous (DerSimonian-Laird χ² = 8.07; df = 6; P = 0.23; I² = 25.7%). Supplemental intake had no relation with an OR of 1.00 (95% CI, 0.95-1.05), also with consistent and homogenous results (DerSimonian-Laird χ² = 1.94; df = 3; P = 0.58; I² = 0%). Total vitamin D intake was inversely but not significantly associated with adenomas (OR, 0.89; 95% CI, 0.79-1.01; DerSimonian-Laird χ² = 7.72; df = 8; P = 0.46; I² = 0%) at a similar magnitude of dietary vitamin D alone.

The inverse association of vitamin D intake was observed for a range of adenomatous outcomes including small adenomas (13), recurrent adenomas (23, 25, 26, 30), and advanced colorectal neoplasia (refs. 16, 22, 23, 27; Table 3). The inverse association on recurrent adenomas was highly consistent among the studies (DerSimonian-Laird χ² = 0.99; df = 3; P = 0.80; I² = 0%) but did not reach significance (OR, 0.88; 95% CI, 0.72-1.07). Advanced adenomas yielded significant inverse findings with OR of 0.77 (95% CI, 0.63-0.95).

Most of the vitamin D intake studies included men and women, whereas four studies drew from sex-specific cohorts. Kampman et al. presented separate results for men and women from two large cohorts, with the results for women updated by Oh et al. (16). Inverse relations were observed for women but not men, although neither result reached statistical significance (28). A recent analysis by Jacobs et al. stratified by sex and found a stronger inverse relation for women, although results for neither men nor women reached statistical significance (26). Kesse et al. also found a positive association in a cohort of French women, but results did not reach statistical significance (27). The study by Lieberman et al. was a predominately male (97%) cohort, and an inverse association was observed (22). Martinez et al. reported no difference in the relation when stratified by sex (30).

For the Health Professionals Follow-up Study cohort of men (28), the highest quintile did not confer an inverse association, although there was a linear inverse trend for total vitamin D intake through the fourth quintiles, suggesting the finding for quintile 5 may have been due to chance. In a sensitivity analysis in which we excluded Health Professionals Follow-up Study from the analysis, a significant inverse association was observed: OR of 0.87 (95% CI, 0.77-0.98) that was homogenous (DerSimonian-Laird χ² = 11.19; df = 10; P = 0.34; I² = 10.6%).

Eight studies that examined the relationship between circulating 25(OH)D and colorectal adenomas were identified, including four case-control studies (15, 29, 31, 32), one nested case-control study (33), two cohort studies (14, 26), and one cross-sectional study (34). Seven of the eight studies identified were included in this analysis (Table 2). Data provided by Sieg et al. (30) did not include ORs and were excluded from this analysis.

Blood samples were drawn from individuals in multiple sites across the United States. Circulating 25(OH)D levels were presented as means, medians, or ranges. The highest value of 25(OH)D reported was 115 ng/mL in Levine et al. (29). Six studies accounted for season of blood draw: one study matched controls by date of blood draw (33), and five studies (14, 15, 26, 31, 34) adjusted for month of blood draw. One study did not adjust for seasonality (29). Studies that presented more than one OR for results stratified by calcium intake or sex were pooled into one summary OR for each study.

The Peto OR of primary colorectal adenoma risk for the overall combined studies using a random-effects model was 0.70 (95% CI, 0.56-0.87) for the highest versus lowest circulating 25(OH)D levels (Fig. 2). Fixed-effects models yielded similar findings (OR, 0.74; 95% CI, 0.65-0.84). The DerSimonian-Laird test suggested moderate heterogeneity among the pooled studies (χ² = 13.06; df = 6; P = 0.04; I² = 54.1%). Following a sensitivity analysis that excluded studies with the strongest inverse findings (34), the heterogeneity was no longer significant; both studies excluded χ² = 1.72; df = 4; P = 0.79; I² = 0%. The overall effect of the pooled studies remained inverse (OR, 0.81; 95% CI, 0.70-0.93). Results were stratified by sex in the Peters et al. study, and an inverse association between circulating vitamin D and colorectal adenomas was observed for women but not men (31).

The effect of including further adjusted models over more parsimonious models was examined within studies. The most adjusted models were selected when available. For the three studies where several multivariate models could have been selected, the effect of using the most adjusted model weakened the vitamin D and adenoma inverse relation by 8% (33) to 22% (35) or did not change the effect (29).

Four studies examined the vitamin D and adenoma relation in the context of calcium supplementation (14) or stratified by calcium levels (15, 29, 34). In the study by Grau et al. (14), subjects were randomized to 3 g calcium carbonate (1,200 mg elemental calcium) or placebo. In this study, an inverse association with circulating 25(OH)D was present for adenoma recurrence only among subjects receiving calcium supplements: relative risk = 0.88 (95% CI, 0.77-0.99) per 12 ng/mL increase of 25(OH)D, P for interaction = 0.006, but no effect of vitamin D was observed in the placebo group (14). In the three observational studies, subjects were stratified by calcium status above or below the median or mean total calcium intake (712, 711, and 744 mg/d, respectively). In the Peters et al. study, there was a 28% greater decreased risk of adenomas for each 10 ng/mL increase in 25(OH)D among subjects with calcium intake above the median (average 768 mg/d for cases and controls; ref. 15). Miller et al. also showed a 49% greater inverse relation of 25(OH)D among those in the high calcium group (34). These studies suggest that the risk of colorectal adenoma is lowest when both calcium intake and vitamin D status are optimized. Contrary to these results, Levine et al. found that individuals with high vitamin D but low calcium had a greater decreased risk of adenomas (OR, 0.40; 95% CI, 0.22-0.76) compared with those with calcium above the mean (OR, 1.17; 95% CI, 0.69-1.99; ref. 27). The overall compilation of the 25(OH)D studies stratifying for calcium yielded an inverse association for vitamin D and colorectal adenomas for both high and low calcium, with a stronger effect among those with high calcium intake (OR, 0.67; 95% CI, 0.46-0.97) compared with those with low calcium intake (OR, 0.78; 95% CI, 0.54-1.12). Of note, the Grau et al. study was a randomized trial of calcium while the others were observational (Table 3).

Two studies showed an overall 36% decreased risk of advanced adenoma for high versus low 25(OH)D (Table 3). Grau et al. reported a statistically significant 42% reduction in advanced adenomas (OR, 0.58; 95% CI, 0.36-0.94) among individuals with 25(OH)D levels above versus below the study median of 29.1 ng/mL and randomized to receive calcium supplementation (14). For subjects in the control group, there was a nonsignificant reduction in advanced adenomas among subjects in the highest quartile of circulating vitamin D (OR, 0.83; 95% CI, 0.39-1.79), no association in the second quartile (OR, 1.06; 95% CI, 0.52-2.17), and a marginally significant increased association in the third quartile (OR, 1.98; 95% CI, 1.01-3.88).⁷

Combining results from 12 studies of vitamin D intake and 7 studies of circulating 25(OH)D, high compared with low dietary vitamin D intake was associated with an 11% marginally decreased risk of colorectal adenomas and high versus low circulating 25(OH)D with a statistically significant 30% decreased risk.

The inverse association with colorectal adenomas was similar for dietary intake compared with total vitamin D intake (OR, 0.90 versus 0.89), whereas no association with supplemental vitamin D intake was found (OR, 1.0). That total intake did not have a stronger association compared with dietary sources alone was unexpected given that dietary sources of vitamin D are uncommon, and in some populations, supplements are an important source of total vitamin D intake. Furthermore, supplemental sources of vitamin D may contain folate and other micronutrients that could enhance an inverse relation between supplemental vitamin D and adenomas (36, 37).

The lack of association with supplemental vitamin D and absence of a stronger association between total versus dietary vitamin D suggest that supplemental sources of vitamin D may not be of maximum effectiveness. Of note, vitamin D present as ergocalciferol or vitamin D₂ may be less efficient than cholecalciferol or vitamin D₃ in raising 25(OH)D levels. Many multivitamins have used ergocalciferol, and supplemental vitamin D predicted 25(OH)D levels much less strongly than did dietary vitamin D in one study (38). Bioactivity may be further decreased by the presence of high levels of vitamin A in the form of retinol in multivitamins, which has been shown to bind and antagonize the vitamin D receptor by reducing the availability of the retinoid X receptor, which is required for the vitamin D receptor function (39). In the study by Oh et al., there was an inverse association of colorectal adenoma with higher total vitamin D intake, but this association was attenuated with increasing levels of retinol intake in a dose-response relation for increased tertiles of total vitamin D intake (16). No other studies of colorectal adenoma (or cancer) have examined this potential interaction with retinol, but high retinol intake appears to antagonize the positive influence of vitamin D on bone mass density (40) by inducing bone resorption (41) and decreasing the intestinal calcium absorption response to vitamin D in humans (42). Further, high retinol intake is associated with reduced bone mineral density (43) and increased risk of hip fractures (44, 45), although one study did not show any increase in hip or other fractures (46).

Dietary vitamin D may be an effective indicator of vitamin D status given its ability to reflect long-term exposure through multiple assessments of intake over several years. The food frequency questionnaires (47) used in the Oh et al. study (16), one of only two studies that showed a significant inverse association with vitamin D intake, captured almost two decades of intake.

However, vitamin D intake does not account for vitamin D produced by sun exposure or other potential confounding factors. Optimal daily UV-B sunlight exposure translates to an oral dose of 20,000 IU vitamin D and may be the main source (∼90%) of vitamin D for most individuals. Physical activity has been used as a surrogate for outdoor sunlight exposure, with correlation between physical activity and 25(OH) D level corresponding to r = 0.14 (39). Where sunlight exposure is limited, at least 1,000 IU/d has been recommended to achieve 25(OH)D levels of 30 to 40 ng/mL (78-100 nmol/L) for optimal bone health (48). However, vitamin D is rare in food sources, mainly fish, fortified dairy products, and multivitamins and supplements.

Whether one should adjust for calcium intake when assessing the association between vitamin D and adenoma risk is questionable, especially given the suggestion of interaction between calcium intake and vitamin D status. In studies of dietary vitamin D, the correlation between calcium and vitamin D may be high given that fortified milk is an important source for both. Thus, a potential additional explanation for the stronger associations for dietary vitamin D than for supplemental vitamin D may be that most individuals high in dietary vitamin D will likely also have high dietary calcium intake, whereas the same assumption cannot be made for supplemental or solar sources of vitamin D. Two dietary studies that adjusted for calcium (16, 30) showed a weaker inverse association of dietary vitamin D among models further adjusted for calcium and other variables, although one cannot attribute this effect to calcium alone.

Potential effect modification by calcium for circulating 25(OH)D and adenoma incidence relation was suggested in three studies (14, 15, 34) but not in another (29). Effect modification by calcium was also shown for vitamin D intake in the study by Oh et al., which found that women with high vitamin D but low calcium did not have a decreased risk of adenoma (relative risk, 1.01; 95% CI, 0.79-1.30), but women with high calcium and vitamin D intakes had a decreased risk (relative risk, 0.75; 95% CI, 0.61-0.92) compared with women with both low vitamin D and low calcium intakes (16).

Circulating 25(OH)D is presumably the gold standard for assessing vitamin D status, although 25(OH)D has a half-life of only about 2 to 3 weeks and its levels tend to fluctuate throughout the year due to variation in solar radiation. Some heterogeneity in the summary OR was apparent for circulating 25(OH)D. Six of the seven studies reported a significant or suggestive inverse association, but for two of the studies, the inverse association appeared to be substantially stronger than for the others (15, 32). Following sensitivity analyses that excluded both of these two studies, the heterogeneity was no longer significant, whereas the overall effect of the pooled studies remained inverse for high versus low circulating 25(OH)D (OR, 0.81; 95% CI, 0.70-0.93). Thus, the heterogeneity was caused more by variation in the magnitude rather than in the direction of the association. The circulating 25(OH)D finding for adenomas (overall and for advanced adenomas) was larger in magnitude than the finding for intake but smaller in magnitude than the 50% reduction in incidence reported for colorectal cancer risk among individuals with high (33 ng/mL) compared with low circulating 25(OH)D (11, 12). One explanation for the weaker results for adenomas than for cancer is that vitamin D could have a role in preventing the formation of adenomas as well as their progression to cancer. Of note, although based on only a small fraction of the studies, an inverse association with vitamin D appeared to be stronger for advanced adenomas compared with the overall association.

Because the studies were observational, and the associations were moderate, residual confounding cannot be excluded. Potential sources for confounding include skin pigmentation, diet, age, and adiposity. Melanin in dark-pigmented individuals blocks effective UV-B radiation, and up to 10 times greater UV-B radiation exposure is required to produce vitamin D levels comparable with that in light-skinned individuals (39, 49). Hypovitaminosis D (≤15 ng/mL) is more prevalent among African Americans compared with White individuals (50). Elderly individuals have also been observed to be vitamin D insufficient year-round regardless of residency in sunny regions (51). Finally, obese individuals have decreased 25(OH)D and 1,25(OH)₂D possibly due to down-regulation of the vitamin D receptor and >50% decreased bioavailability of serum 25(OH)D secondary to deposition in adipose tissue (52). As indicated in Tables 1 and 2, cofactors were variably controlled across studies.

A limitation of this meta-analysis is that it examined only high versus low circulating 25(OH)D and vitamin D intake. The ability of this analysis to analyze characteristics of the adenomas such as size, location, and recurrence is also limited due to a small number of studies to quantify. Detection bias was reduced by using endoscopy screened controls or noncases.

In addition, all studies of circulating 25(OH)D had only a single measurement of 25(OH)D. Stronger associations between 25(OH)D and colorectal cancer and adenomas were found when measurements were made during winter months versus summer months in case-control and cohort studies (32, 53). Although adjustment or matching decreases some extraneous variability and potential measurement error, these do not entirely offset the limitations of using a single measurement to estimate long-term status.

All studies took place between 1980 and 2003, with two which were conducted in the 1980s (13, 28). The potential for publication bias was explored, and although funnel plot analysis indicated no substantial publication bias, this result was based on a limited number of studies and potential for bias cannot necessarily be excluded (54). The I² statistic was included with primary study results and sensitivity analyses were done when heterogeneity was suggested.

Subgroup analysis for sex was done, although a limited number of studies were stratified for these variables. Some of the studies suggested a more consistent inverse association in women, but overall more data would be necessary to ascertain a sex-specific role for vitamin D and colorectal adenomas.

The overall results of this meta-analysis indicate that vitamin D status, assessed through intake and circulating 25(OH)D, is associated with a decreased risk of colorectal adenoma, including advanced adenoma and recurrent adenoma. Some studies also suggest that the inverse association between vitamin D status and adenoma risk may be strongest when calcium intake is also high. These data add to the growing body of evidence supporting a potential benefit of improved vitamin D status and risk of colorectal neoplasia.

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.