Abstract
Despite the substantial epidemiologic evidence on the inverse association between circulating 25-hydroxyvitamin D [25(OH)D] and colorectal cancer, it remains controversial whether this relationship is causal or due to confounding by inflammation. We reevaluated the association between plasma 25(OH)D and colorectal cancer risk by additionally accounting for inflammatory markers in a prospective case–control study nested within the Nurses' Health Study and Health Professionals Follow-up Study (615 cases and 1,209 matched controls). Conditional logistic regression was used to estimate relative risk (RR) and 95% confidence interval (CI) of colorectal cancer in relation to quartiles of plasma 25(OH)D. Results were compared before and after adjusting for inflammatory markers in the multivariable model. Plasma 25(OH)D was associated with reduced risk of colorectal cancer (multivariable RR comparing extreme quartiles = 0.71; 95% CI, 0.52–0.97; Ptrend = 0.01). Additional adjustment for C-reactive protein, IL6, soluble tumor necrosis factor receptor 2, or a composite inflammatory score did not change the results [multivariable (including inflammatory score) RR = 0.72; 95% CI, 0.53–0.98; Ptrend = 0.02). Our findings suggest that confounding by inflammation, as reflected by circulating inflammatory markers, does not appear to account for the inverse association between plasma 25(OH)D and colorectal cancer. Cancer Epidemiol Biomarkers Prev; 23(10); 2175–80. ©2014 AACR.
This article is featured in Highlights of This Issue, p. 1957
Introduction
The effect of vitamin D on a wide variety of nonskeletal health outcomes, including colorectal cancer, has been the subject of considerable debate (1). High vitamin D has been associated with reduced risk of colorectal cancer in observational studies evaluating dietary intake (2, 3), circulating 25-hydroxyvitamin D [25(OH)D] (2, 4, 5), and predicted 25(OH)D (6). A recent meta-analysis of observational data reported a 26% lower risk of colorectal cancer [95% confidence interval (CI), 11%–37%) per 10 ng/mL increment in blood 25(OH)D levels (4). Possible mechanisms include antiproliferative, proapoptotic, prodifferentiating, and anti-inflammatory actions by the physiologically most active molecular form of vitamin D, formed locally 1α,25-dihydroxyvitamin D3 [1,25(OH)2D3] (7).
Despite the preponderance of observational and mechanistic evidence, no benefit was reported for vitamin D supplementation against colorectal cancer occurrence in several randomized trials (1), including the Women's Health Initiative (WHI; ref. 8), the largest trial in this regard. However, these data are far from conclusive given critical limitations of the WHI, including poor adherence, inadequate vitamin D dose and duration, complex factorial design, and low baseline risk of the study population. Each of these factors may have contributed to the null findings (9–11).
Recently, Autier and colleagues summarized the observational and experimental findings on vitamin D and nonskeletal disorders with a special emphasis on colorectal cancer, which is the only cancer related to low 25(OH)D in observational studies but not in randomized trials in their analyses (1). Considering the potential link between inflammation, vitamin D, and colorectal cancer, they hypothesized that low 25(OH)D is just a marker of ill health, with inflammation, measured by elevated concentrations of circulating C-reactive protein (CRP), IL6, and TNFα, acting as a confounding factor in the observed inverse association of 25(OH)D and colorectal cancer (1).
To assess the hypothesis of Autier and colleagues (1) and to assess the robustness of our prior findings on the association between plasma 25(OH)D and colorectal cancer risk (5), we performed an updated analysis with simultaneous adjustment for inflammatory markers in two large ongoing cohorts, the Nurses' Health Study (NHS) and the Health Professionals Follow-up Study (HPFS).
Materials and Methods
Study population
The NHS enrolled 121,701 female nurses aged 30 to 55 years in the United States in 1976. The HPFS enrolled 51,529 male health professionals aged 40 to 75 years in the United States in 1986. Details of the two cohorts have been described elsewhere (12, 13). Briefly, in both cohorts, follow-up questionnaires were administered biennially to collect and update medical and lifestyle information; validated food frequency questionnaires were completed every 4 years to update dietary data. Both cohorts have a follow-up rate of approximately 90% in each 2-year cycle. The study protocol was approved by the institutional review board of the Brigham and Women's Hospital (Boston, MA).
Ascertainment of cases and controls
Self-reported diagnoses of colorectal cancer were obtained in biennial questionnaires, and participants who reported a diagnosis of colorectal cancer were asked for permission to acquire their medical records and pathologic reports. We identified deaths with more than 96% sensitivity through the National Death Index and next-of-kin. For all colorectal cancer deaths, we requested permission from next-of-kin to review medical records. A study physician, blinded to exposure information, reviewed records to confirm colorectal cancer diagnosis and to extract information on anatomic location, stage, and histologic type of cancer.
A blood specimen was returned from 32,826 women in the NHS between 1989 and 1990 and from 18,225 men in the HPFS between 1993 and 1995. Among these participants, we documented 360 incident colorectal cancer cases in the NHS during follow-up through October 1, 2008 and 287 colorectal cancer cases in the HPFS through January 1, 2008. Among participants who provided plasma samples, we randomly selected up to 2 controls for each colorectal cancer patient matched on age (within 2 years) and on year (same year) and month (within 1 month) of blood donation from those who were alive and free of cancer at the time of diagnosis of the colorectal cancer case. A total of 706 and 550 controls were selected in the NHS and in the HPFS, respectively.
Laboratory assays
We used the radioimmunoassay method to measure plasma 25(OH)D, a highly sensitive immunoturbidimetric assay (Denka Seiken Co) to measure CRP, an ultra-sensitive ELISA (R&D Systems) to measure IL6, and an ELISA (R&D Systems) to measure soluble TNF receptor 2 (sTNFR2, also known as TNF receptor superfamily member 1B, TNFRSF1B), as previously described (5, 14, 15). Samples from cases and their matched controls were handled together and analyzed in the same batch. Quality control samples were randomly interspersed among the case–control samples. Personnel blinded to quality control and case–control status conducted all assays. For 25(OH)D, the mean intra-assay coefficient of variation from blinded quality control samples was less than 15% for all batches. For inflammatory markers, the intra-assay coefficients of variations in the NHS and HPFS were as follows: CRP, 2.2% and 7.8%; IL6, 10.6% and 12.1%; sTNFR2, 6.7% and 10.1% (14, 15). We excluded plasma samples that failed any of the four laboratory assays from 19 cases and 28 controls in the NHS, and 13 cases and 19 controls in the HPFS.
Statistical analyses
We corrected biomarker concentrations for measurement batch using the average batch correction method (16), with adjustment for age, body mass index (BMI), physical activity and case control status. We employed principal component analysis to construct a composite inflammatory score based on log-transformed CRP, IL6, and sTNFR2 concentrations among control participants. The first principal component explaining 57.5% of variance in women and 57.6% in men was extracted and the score was calculated as a linear combination, weighted by eigenvectors, of the three markers (all standardized to have a mean of zero and SD of one). Age-adjusted Spearman correlation coefficients were calculated between biomarkers, BMI, and physical activity among control participants. We calculated quartiles of 25(OH)D based on the distribution among control participants within each cohort and used conditional logistic regression to estimate relative risk (RR) and 95% CI of colorectal cancer. In the multivariable model, we adjusted for several confounding factors that are generally considered as important risk factors for colorectal cancer (see footnote of Table 3). We used SAS version 9.3 (SAS Institute, Inc) for analyses. All statistical tests were two-sided and P < 0.05 was considered statistically significant.
Results
After excluding samples that failed laboratory assays, a total of 341 cases and 678 controls from the NHS and 274 cases and 531 controls from the HPFS were included in the present analysis (Table 1). Plasma 25(OH)D levels were lower among colorectal cancer patients than controls (P < 0.001 in women and P = 0.08 in men). In the cross-sectional assessment, plasma 25(OH)D was not significantly correlated with inflammatory markers except a weak inverse association with IL6 among women (r = −0.08; P = 0.04). High BMI and low physical activity were inversely associated with 25(OH)D and positively associated with inflammatory markers in both men and women (Table 2).
. | Women . | Men . | ||
---|---|---|---|---|
Variable . | Cases (n = 341) . | Controls (n = 678) . | Cases (n = 274) . | Controls (n = 531) . |
Age at blood draw, year | 59.0 (6.7) | 59.0 (6.7) | 65.8 (8.3) | 65.9 (8.5) |
Body mass index, kg/m2 | 25.9 (4.9) | 25.4 (4.7) | 26.3 (3.4) | 25.5 (3.0) |
Physical activity, MET-h/wk | 16.8 (19.2) | 17.1 (21.4) | 31.9 (26.9) | 30.9 (25.1) |
Pack-year of smoking before age 30 | 3.9 (5.5) | 3.7 (5.5) | 5.6 (6.7) | 4.9 (6.8) |
Current smoker, % | 13.9 | 12.7 | 4.92 | 4.75 |
Family history of colorectal cancer, %b | 23.2 | 18.1 | 22.6 | 15.3 |
History of previous endoscopy, % | 35.2 | 38.8 | 53.7 | 63.7 |
Current multivitamin use, % | 35.2 | 39.5 | 47.5 | 51.8 |
Regular aspirin use, %c | 37.8 | 45.3 | 42.0 | 48.4 |
Regular NSAID use, %d | 15.9 | 16.6 | 11.4 | 12.0 |
Postmenopausal, % | 88.0 | 87.9 | — | — |
Current hormone use, %e | 32.9 | 41.5 | — | — |
Alcohol consumption, g/day | 5.7 (9.7) | 5.6 (9.5) | 12.4 (16.2) | 12.3 (15.1) |
Folate intake, μg/day | 425 (260) | 453 (239) | 513 (257) | 544 (281) |
Calcium intake, mg/day | 987 (544) | 1073 (558) | 931 (441) | 946 (424) |
Total fiber intake, g/day | 18.6 (5.8) | 19.0 (5.7) | 22.9 (7.5) | 23.3 (7.2) |
Dietary DASH score | 23.7 (4.1) | 24.3 (4.3) | 24.2 (4.6) | 24.7 (4.5) |
Time between blood draw and colorectal cancer diagnosis, y | 9.3 (5.1) | — | 6.3 (3.6) | — |
Plasma biomarker concentrations, median (interquartile range) | ||||
25(OH)D, ng/mL | 23.9 (17.8–30.1) | 26.0 (19.6–31.6) | 27.5 (22.4–33.4) | 28.6 (22.7–34.1) |
CRP, mg/L | 1.47 (0.50–3.07) | 1.52 (0.60–3.48) | 1.36 (0.68–2.62) | 1.11 (0.59–2.17) |
IL6, pg/mL | 1.03 (0.56–1.62) | 0.94 (0.54–1.61) | 1.60 (0.99–2.65) | 1.39 (0.93–2.25) |
sTNFR2, pg/mL | 2,732 (2,335–3,216) | 2,666 (2,249–3,155) | 2,746 (2,346–3,232) | 2,723 (2,342–3,339) |
. | Women . | Men . | ||
---|---|---|---|---|
Variable . | Cases (n = 341) . | Controls (n = 678) . | Cases (n = 274) . | Controls (n = 531) . |
Age at blood draw, year | 59.0 (6.7) | 59.0 (6.7) | 65.8 (8.3) | 65.9 (8.5) |
Body mass index, kg/m2 | 25.9 (4.9) | 25.4 (4.7) | 26.3 (3.4) | 25.5 (3.0) |
Physical activity, MET-h/wk | 16.8 (19.2) | 17.1 (21.4) | 31.9 (26.9) | 30.9 (25.1) |
Pack-year of smoking before age 30 | 3.9 (5.5) | 3.7 (5.5) | 5.6 (6.7) | 4.9 (6.8) |
Current smoker, % | 13.9 | 12.7 | 4.92 | 4.75 |
Family history of colorectal cancer, %b | 23.2 | 18.1 | 22.6 | 15.3 |
History of previous endoscopy, % | 35.2 | 38.8 | 53.7 | 63.7 |
Current multivitamin use, % | 35.2 | 39.5 | 47.5 | 51.8 |
Regular aspirin use, %c | 37.8 | 45.3 | 42.0 | 48.4 |
Regular NSAID use, %d | 15.9 | 16.6 | 11.4 | 12.0 |
Postmenopausal, % | 88.0 | 87.9 | — | — |
Current hormone use, %e | 32.9 | 41.5 | — | — |
Alcohol consumption, g/day | 5.7 (9.7) | 5.6 (9.5) | 12.4 (16.2) | 12.3 (15.1) |
Folate intake, μg/day | 425 (260) | 453 (239) | 513 (257) | 544 (281) |
Calcium intake, mg/day | 987 (544) | 1073 (558) | 931 (441) | 946 (424) |
Total fiber intake, g/day | 18.6 (5.8) | 19.0 (5.7) | 22.9 (7.5) | 23.3 (7.2) |
Dietary DASH score | 23.7 (4.1) | 24.3 (4.3) | 24.2 (4.6) | 24.7 (4.5) |
Time between blood draw and colorectal cancer diagnosis, y | 9.3 (5.1) | — | 6.3 (3.6) | — |
Plasma biomarker concentrations, median (interquartile range) | ||||
25(OH)D, ng/mL | 23.9 (17.8–30.1) | 26.0 (19.6–31.6) | 27.5 (22.4–33.4) | 28.6 (22.7–34.1) |
CRP, mg/L | 1.47 (0.50–3.07) | 1.52 (0.60–3.48) | 1.36 (0.68–2.62) | 1.11 (0.59–2.17) |
IL6, pg/mL | 1.03 (0.56–1.62) | 0.94 (0.54–1.61) | 1.60 (0.99–2.65) | 1.39 (0.93–2.25) |
sTNFR2, pg/mL | 2,732 (2,335–3,216) | 2,666 (2,249–3,155) | 2,746 (2,346–3,232) | 2,723 (2,342–3,339) |
Abbreviations: DASH, Dietary Approaches to Stop Hypertension; MET, metabolic equivalent = (caloric need/kilogram body weight per hour activity)/(caloric need/kilogram body weight per hour at rest); NSAID, nonsteroidal anti-inflammatory drug.
aNumbers in parenthesis indicate SD for means unless otherwise specified.
bDefined as having a diagnosis of colorectal cancer among parents or siblings.
cA standard tablet contains 325 mg aspirin, and regular users were defined as those who used at least 2 tablets per week.
dRegular users were defined as those who used at least 2 tablets per week.
ePercentage is among postmenopausal women.
Variable . | 25(OH)D . | CRP . | IL6 . | sTNFR2 . | Inflammatory scorea . | BMI . | Physical activity (METs-h/wk) . |
---|---|---|---|---|---|---|---|
25(OH)D | — | −0.05 | −0.08b | 0.05 | −0.04 | −0.17c | 0.21c |
CRP | −0.008 | — | 0.45c | 0.28c | 0.79c | 0.38c | −0.18c |
IL6 | −0.06 | 0.48c | — | 0.35c | 0.75c | 0.22c | −0.08b |
sTNFR2 | 0.02 | 0.23c | 0.26c | — | 0.67c | 0.28c | −0.06 |
Inflammatory scorea | −0.02 | 0.78c | 0.80c | 0.60c | — | 0.38c | −0.14c |
BMI | −0.15c | 0.20c | 0.15c | 0.11b | 0.20c | — | −0.16c |
Physical activity (METs-h/wk) | 0.23c | −0.10b | −0.14d | −0.13d | −0.16c | −0.13d | — |
Variable . | 25(OH)D . | CRP . | IL6 . | sTNFR2 . | Inflammatory scorea . | BMI . | Physical activity (METs-h/wk) . |
---|---|---|---|---|---|---|---|
25(OH)D | — | −0.05 | −0.08b | 0.05 | −0.04 | −0.17c | 0.21c |
CRP | −0.008 | — | 0.45c | 0.28c | 0.79c | 0.38c | −0.18c |
IL6 | −0.06 | 0.48c | — | 0.35c | 0.75c | 0.22c | −0.08b |
sTNFR2 | 0.02 | 0.23c | 0.26c | — | 0.67c | 0.28c | −0.06 |
Inflammatory scorea | −0.02 | 0.78c | 0.80c | 0.60c | — | 0.38c | −0.14c |
BMI | −0.15c | 0.20c | 0.15c | 0.11b | 0.20c | — | −0.16c |
Physical activity (METs-h/wk) | 0.23c | −0.10b | −0.14d | −0.13d | −0.16c | −0.13d | — |
Abbreviations: MET, metabolic equivalent = (caloric need/kilogram body weight per hour activity)/(caloric need/kilogram body weight per hour at rest).
aInflammatory score was created by principal component analysis using plasma CRP, IL6, and sTNFR2 levels among controls.
bP < 0.05.
cP < 0.001.
dP < 0.01.
Table 3 shows the association between plasma 25(OH)D and colorectal cancer with and without adjusting for inflammatory markers. High 25(OH)D was associated with decreased risk of colorectal cancer in both women and men after adjusting for multiple confounding factors, although the trend test was not statistically significant in men. In the pooled cohorts, participants in the highest quartile had a 29% reduced risk of colorectal cancer compared with those in the lowest quartile with a 95% CI of 3.0%–48% (Ptrend = 0.01). After additionally adjusting for CRP, IL6, and sTNFR2 in our multivariable model, the association between 25(OH)D and colorectal cancer was minimally altered. The RR comparing the highest to the lowest quartile of 25(OH)D was 0.72 (95% CI, 0.53–0.98; Ptrend = 0.02) when inflammatory score was included in the model. For colon cancer alone, the multivariable RRs comparing extreme quartiles of 25(OH)D were 0.67 (95% CI, 0.47–0.95; Ptrend = 0.01) and 0.68 (95% CI, 0.48–0.97; Ptrend = 0.01) in the pooled cohorts before and after adjusting for inflammatory score, respectively (data not shown).
. | Quartile of plasma 25(OH)D levels . | . | |||
---|---|---|---|---|---|
. | 1 . | 2 . | 3 . | 4 . | Ptrenda . |
Women | |||||
No. of cases/controls | 118/184 | 91/174 | 72/173 | 60/147 | |
Median (interquartile range), ng/mL | 15.8 (12.8–18.3) | 23.5 (21.8–25.2) | 29.2 (27.7–30.7) | 38.7 (35.3–44.5) | |
Age-adjusted RR (95% CI)b | 1.00 (reference) | 0.81 (0.57–1.15) | 0.62 (0.43–0.90) | 0.59 (0.40–0.88) | 0.005 |
Multivariable RR (95% CI)c | 1.00 (reference) | 0.84 (0.58–1.20) | 0.67 (0.45–0.98) | 0.64 (0.42–0.98) | 0.02 |
Multivariable + CRPd | 1.00 (reference) | 0.86 (0.60–1.23) | 0.66 (0.45–0.98) | 0.65 (0.43–1.00) | 0.03 |
Multivariable + IL6d | 1.00 (reference) | 0.85 (0.59–1.23) | 0.67 (0.45–0.98) | 0.63 (0.41–0.97) | 0.02 |
Multivariable + sTNFR2d | 1.00 (reference) | 0.82 (0.57–1.18) | 0.67 (0.45–0.99) | 0.62 (0.40–0.95) | 0.02 |
Multivariable + inflammatory scoree | 1.00 (reference) | 0.82 (0.57–1.17) | 0.68 (0.46–1.00) | 0.64 (0.42–0.99) | 0.03 |
Men | |||||
No. of cases/controls | 75/131 | 80/131 | 60/135 | 59/134 | |
Median (interquartile range), ng/mL | 18.6 (14.9–20.8) | 25.7 (24.5–27.1) | 31.3 (29.9–32.9) | 38.9 (36.6–43.1) | |
Age-adjusted RR (95% CI)b | 1.00 (reference) | 1.07 (0.72–1.60) | 0.78 (0.51–1.20) | 0.74 (0.47–1.14) | 0.10 |
Multivariable RR (95% CI)c | 1.00 (reference) | 1.05 (0.69–1.61) | 0.80 (0.51–1.26) | 0.80 (0.50–1.28) | 0.22 |
Multivariable + CRPd | 1.00 (reference) | 1.03 (0.67–1.58) | 0.79 (0.50–1.25) | 0.77 (0.47–1.24) | 0.18 |
Multivariable + IL6d | 1.00 (reference) | 1.13 (0.73–1.73) | 0.86 (0.54–1.36) | 0.82 (0.51–1.32) | 0.28 |
Multivariable + sTNFR2d | 1.00 (reference) | 1.07 (0.70–1.64) | 0.79 (0.50–1.25) | 0.77 (0.48–1.24) | 0.18 |
Multivariable + inflammatory scoree | 1.00 (reference) | 1.05 (0.68–1.60) | 0.81 (0.51–1.27) | 0.81 (0.50–1.30) | 0.25 |
Pooled cohorts | |||||
Age-adjusted RR (95% CI)b | 1.00 (reference) | 0.91 (0.70–1.19) | 0.68 (0.51–0.90) | 0.65 (0.48–0.87) | 0.001 |
Multivariable RR (95% CI)c | 1.00 (reference) | 0.94 (0.71–1.23) | 0.72 (0.54–0.97) | 0.71 (0.52–0.97) | 0.01 |
Multivariable + CRPd | 1.00 (reference) | 0.93 (0.71–1.22) | 0.72 (0.53–0.96) | 0.70 (0.51–0.95) | 0.01 |
Multivariable + IL6d | 1.00 (reference) | 0.96 (0.73–1.25) | 0.74 (0.55–0.99) | 0.71 (0.52–0.97) | 0.01 |
Multivariable + sTNFR2d | 1.00 (reference) | 0.94 (0.71–1.23) | 0.72 (0.54–0.97) | 0.69 (0.51–0.95) | 0.009 |
Multivariable + inflammatory scoree | 1.00 (reference) | 0.93 (0.71–1.22) | 0.73 (0.55–0.98) | 0.72 (0.53–0.98) | 0.02 |
. | Quartile of plasma 25(OH)D levels . | . | |||
---|---|---|---|---|---|
. | 1 . | 2 . | 3 . | 4 . | Ptrenda . |
Women | |||||
No. of cases/controls | 118/184 | 91/174 | 72/173 | 60/147 | |
Median (interquartile range), ng/mL | 15.8 (12.8–18.3) | 23.5 (21.8–25.2) | 29.2 (27.7–30.7) | 38.7 (35.3–44.5) | |
Age-adjusted RR (95% CI)b | 1.00 (reference) | 0.81 (0.57–1.15) | 0.62 (0.43–0.90) | 0.59 (0.40–0.88) | 0.005 |
Multivariable RR (95% CI)c | 1.00 (reference) | 0.84 (0.58–1.20) | 0.67 (0.45–0.98) | 0.64 (0.42–0.98) | 0.02 |
Multivariable + CRPd | 1.00 (reference) | 0.86 (0.60–1.23) | 0.66 (0.45–0.98) | 0.65 (0.43–1.00) | 0.03 |
Multivariable + IL6d | 1.00 (reference) | 0.85 (0.59–1.23) | 0.67 (0.45–0.98) | 0.63 (0.41–0.97) | 0.02 |
Multivariable + sTNFR2d | 1.00 (reference) | 0.82 (0.57–1.18) | 0.67 (0.45–0.99) | 0.62 (0.40–0.95) | 0.02 |
Multivariable + inflammatory scoree | 1.00 (reference) | 0.82 (0.57–1.17) | 0.68 (0.46–1.00) | 0.64 (0.42–0.99) | 0.03 |
Men | |||||
No. of cases/controls | 75/131 | 80/131 | 60/135 | 59/134 | |
Median (interquartile range), ng/mL | 18.6 (14.9–20.8) | 25.7 (24.5–27.1) | 31.3 (29.9–32.9) | 38.9 (36.6–43.1) | |
Age-adjusted RR (95% CI)b | 1.00 (reference) | 1.07 (0.72–1.60) | 0.78 (0.51–1.20) | 0.74 (0.47–1.14) | 0.10 |
Multivariable RR (95% CI)c | 1.00 (reference) | 1.05 (0.69–1.61) | 0.80 (0.51–1.26) | 0.80 (0.50–1.28) | 0.22 |
Multivariable + CRPd | 1.00 (reference) | 1.03 (0.67–1.58) | 0.79 (0.50–1.25) | 0.77 (0.47–1.24) | 0.18 |
Multivariable + IL6d | 1.00 (reference) | 1.13 (0.73–1.73) | 0.86 (0.54–1.36) | 0.82 (0.51–1.32) | 0.28 |
Multivariable + sTNFR2d | 1.00 (reference) | 1.07 (0.70–1.64) | 0.79 (0.50–1.25) | 0.77 (0.48–1.24) | 0.18 |
Multivariable + inflammatory scoree | 1.00 (reference) | 1.05 (0.68–1.60) | 0.81 (0.51–1.27) | 0.81 (0.50–1.30) | 0.25 |
Pooled cohorts | |||||
Age-adjusted RR (95% CI)b | 1.00 (reference) | 0.91 (0.70–1.19) | 0.68 (0.51–0.90) | 0.65 (0.48–0.87) | 0.001 |
Multivariable RR (95% CI)c | 1.00 (reference) | 0.94 (0.71–1.23) | 0.72 (0.54–0.97) | 0.71 (0.52–0.97) | 0.01 |
Multivariable + CRPd | 1.00 (reference) | 0.93 (0.71–1.22) | 0.72 (0.53–0.96) | 0.70 (0.51–0.95) | 0.01 |
Multivariable + IL6d | 1.00 (reference) | 0.96 (0.73–1.25) | 0.74 (0.55–0.99) | 0.71 (0.52–0.97) | 0.01 |
Multivariable + sTNFR2d | 1.00 (reference) | 0.94 (0.71–1.23) | 0.72 (0.54–0.97) | 0.69 (0.51–0.95) | 0.009 |
Multivariable + inflammatory scoree | 1.00 (reference) | 0.93 (0.71–1.22) | 0.73 (0.55–0.98) | 0.72 (0.53–0.98) | 0.02 |
aTest for trend was performed using the median value of each quartile of plasma 25(OH)D as a continuous variable in the regression model.
bConditional logistic regression adjusted for matching factors (age [within 2 years], year [same year], and month [within 1 month] of blood collection).
cIn addition to matching factors (age and time of blood draw), multivariable conditional logistic regression adjusted for family history of colorectal cancer, history of endoscopy, pack-years of smoking before age 30, BMI, physical activity, multivitamin use, regular aspirin/NSAID use, alcohol consumption, and DASH score.
dAdditional adjustment for plasma inflammatory marker levels (in quartile).
eInflammatory score was created by principal component analysis using plasma CRP, IL6, and sTNFR2 levels among controls.
Because geographic location is an important determinant of vitamin D levels, in the sensitivity analysis, we additionally adjusted for this variable (defined according to the state of residence; categorized as South, North, Midwest, and West, as previously described; ref. 17) and the results remained essentially unchanged (comparing the extreme quartiles: multivariable RR = 0.72; 95% CI, 0.53–0.99; Ptrend = 0.02). We also considered the possibility that the inverse association may be related to reverse causation due to clinically undetected disorders at the time of blood collection, and thus conducted a stratified analysis by time interval. Plasma 25(OH)D was associated, in a similar magnitude, with colorectal cancer cases that were diagnosed within and after 6 years of blood draw with a RR of 0.61 (95% CI, 0.33–1.12) and 0.74 (95% CI, 0.52–1.07), respectively, thus arguing against the reverse causation explanation.
Because the first principal component used to create the inflammatory score only explained less than 60% of the variance of inflammatory markers, we additionally extracted the second principal component and created the new inflammatory scores based on the first two principal components, which cumulatively explained 80.0% of variance in women and 83.0% in men. We adjusted for the scores in the multivariable model and the results remained the same with the RR of colorectal cancer of 0.71 (95% CI, 0.52–0.97; Ptrend = 0.01) comparing the extreme quartiles of 25(OH)D.
Discussion
Using data from two large cohorts, we assessed the hypothesis of Autier and colleagues that confounding by inflammation, as reflected by elevated levels of circulating inflammatory markers, explains the observed inverse association between 25(OH)D and colorectal cancer in epidemiologic studies. CRP, IL6, and sTNFR2, which is a surrogate for TNFα, have been shown to mediate inflammatory response and are sensitive to dysregulation in inflammatory processes (18). Through further adjustment for these markers, we did not find any meaningful change in the association between 25(OH)D and colorectal cancer. It is possible that the inflammatory markers we measured did not fully reflect inflammatory factors that contribute to colorectal cancer pathogenesis. However, if the hypothesis of Autier and colleagues is correct, our findings would suggest that low 25(OH)D is a superior marker for inflammation than these three well-accepted inflammatory markers, which seems unlikely, especially given the little correlation between plasma 25(OH)D and inflammatory markers observed in the current study.
In addition, the hypothesis of Autier and colleagues builds on the postulations that systemic inflammation is the common underlying factor for both low 25(OH)D and increased colorectal cancer risk. However, animal and human studies have suggested a potential anti-inflammatory effect of vitamin D (19). A recent study in mice found that increased dietary vitamin D intake reduced inflammation and colon cancer incidence, possibly through suppression of MAPK and NFκB signaling (20). Human vitamin D supplementation has been found to reduce inflammatory marker levels, especially among patients with autoimmune disease or other chronic conditions (e.g., chronic kidney disease, heart disease, type 2 diabetes, and osteoporosis), who are more likely to have vitamin D deficiency (21). Regarding the association between general markers of systemic inflammation and colorectal cancer risk, the evidence remains inconclusive with a positive relationship reported in some studies (22, 23) but not in others (15, 24). Therefore, on the basis of the current evidence, it seems implausible that the inflammatory process acts as a predominant confounding factor and explains the increased colorectal cancer risk associated with low vitamin D status.
Our study has some limitations. First, we assessed 25(OH)D and inflammatory markers from plasma collected at the same time and in a single measurement that may not represent long-term concentrations. However, these inflammatory markers have been shown to be generally stable over time and any misclassification of 25(OH)D should be nondifferential, which should have biased the results towards unity. Second, given the observational nature of our study, residual confounding cannot be excluded. However, the robust adjustment for a host of colorectal cancer–related lifestyle factors makes it unlikely that residual confounding from major risk factors of colorectal cancer explains our findings.
The strength of the current study is the prospective design within the two large well-established cohorts, which have both laboratory data on biomarkers and detailed information on a variety of lifestyle factors, therefore providing us the unique opportunity to examine the relationship between plasma 25(OH)D and colorectal cancer while accounting for inflammatory markers and other potential confounders.
In conclusion, the inverse association between plasma 25(OH)D and colorectal cancer persisted after adjustment for inflammatory markers, therefore highlighting the continuing priority for further research on vitamin D and colorectal cancer prevention.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Disclaimer
Certain data used in this publication were obtained from the DPH. The authors assume full responsibility for analyses and interpretation of these data.
Authors' Contributions
Conception and design: K. Wu, C.S. Fuchs, E.L. Giovannucci
Development of methodology: C.S. Fuchs, E.L. Giovannucci
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Song, A.T. Chan, C.S. Fuchs, E.L. Giovannucci
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Song, A.T. Chan, C.S. Fuchs, E.L. Giovannucci
Writing, review, and/or revision of the manuscript: M. Song, K. Wu, A.T. Chan, C.S. Fuchs, E.L. Giovannucci
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Song, C.S. Fuchs
Study supervision: K. Wu, A.T. Chan, C.S. Fuchs, E.L. Giovannucci
Acknowledgments
The authors thank the participants and staff of the NHS and the HPFS, for their valuable contributions, as well as the following state cancer registries for their help: AL, AZ, AR, CA, CO, CT, DE, FL, GA, ID, IL, IN, IA, KY, LA, ME, MD, MA, MI, NE, NH, NJ, NY, NC, ND, OH, OK, OR, PA, RI, SC, TN, TX, VA, WA, WY. In addition, this study was approved by the Connecticut Department of Public Health (DPH) Human Investigations Committee. M.S. is a trainee of the Harvard Transdisciplinary Research Center on Energetics and Cancer (TREC), which is supported by 1U54CA155626.
Grant Support
This work was supported by the NIH (P01 CA87969 to S.E. Hankinson, M. Song, A.T. Chan, E.L. Giovannucci; R01 CA49449 to S.E. Hankinson; UM1 CA167552 to W.C. Willett; P01 CA 55075 to W.C. Willett, M. Song, A.T. Chan, E.L. Giovannucci; and 1U54CA155626 to F.B. Hu).
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.