Serum 25-Hydroxyvitamin D and Risks of Colon and Rectal Cancer in Finnish Men

Stephanie J. Weinstein; Kai Yu; Ronald L. Horst; Jason Ashby; Jarmo Virtamo; Demetrius Albanes


Am J Epidemiol. 2011;173(5):499-508. 

In This Article

Materials and Methods

Study Population

The ATBC Study was a randomized, double-blind, placebo-controlled, primary prevention trial involving daily supplementation with α-tocopherol (50 mg/day), β-carotene (20 mg/day), both, or placebo.[17] From 1985 to 1988, the ATBC investigators recruited 29,133 men aged 50–69 years who smoked at least 5 cigarettes per day from southwestern Finland (latitude of study area, 60°N–64°N). Study supplementation continued for 5–8 years (median, 6.1 years) until death or trial closure (April 30, 1993). The study was approved by the institutional review boards of the US National Cancer Institute and the National Public Health Institute of Finland. Written informed consent was obtained from each participant prior to randomization.

Case Identification and Control Selection

Cases (n = 428) were defined as incident colonic or rectal adenocarcinoma (International Classification of Diseases, Ninth Revision, codes 153 and 154, respectively), excluding cancers of the appendix and anus and cancers of unknown histology, identified through the Finnish Cancer Registry, with follow-up through April 30, 2005. Cases from a prior analysis[7] (diagnosed through November 1993) were excluded; therefore, all but 8 cases were diagnosed after November 1993. Three subjects who were diagnosed with both colon and rectal cancer on the same day were included in analyses of both sites, resulting in a total of 239 colon cancers and 192 rectal cancers. Cases of distal colon cancer (International Classification of Diseases, Ninth Revision, codes 153.2 and 153.3) and rectal cancer were also combined (n = 292). For cases diagnosed through April 1999 (116 colon and 89 rectum), 2 study physicians reviewed medical records for diagnostic confirmation and staging, and 1 pathologist reviewed histolopathologic or cytologic specimens where available. Information on colorectal cancer cases diagnosed since May 1999 (123 colon and 103 rectum) was derived solely from the Finnish Cancer Registry, which provides almost 100% case ascertainment.[18] Controls were alive and cancer-free at the time of case diagnosis and were matched to cases (1:1) on age at randomization (±1 year) and date of baseline serum collection (±30 days).

Serum 25(OH)D Determination

Fasting baseline serum samples were stored at –70°C. 25(OH)D was measured in early 2008 at Heartland Assays, Inc. (Ames, Iowa) using the DiaSorin Liaison 25(OH)D TOTAL assay platform by means of a direct, competitive chemiluminescence immunoassay.[19,20]

Each batch contained matched case/control sets and 4 or 6 blinded quality control samples, consisting of either standard reference material provided by the US National Institute of Standards and Technology[21] or an ATBC serum pool. Interbatch and intrabatch coefficients of variation, calculated using a nested components-of-variance analysis,[22] were 12.7%–13.6% and 9.3%–11.0%, respectively, for the standard reference material and 12.3% and 10.5%, respectively, for the ATBC serum pool.

Statistical Analysis

Wilcoxon rank-sum and chi-square tests (for continuous and categorical variables, respectively) were used to compare characteristics of cases and controls. Conditional logistic regression was used to determine odds ratios and 95% confidence intervals. The 25(OH)D concentrations were analyzed using several different approaches. First, cutpoints were defined on the basis of clinical definitions in the literature[15,16,23,24] as <25, 25–<37.5, 37.5–<50, 50–<75, and ≥75 nmol/L. The referent category was defined as 50–<75 nmol/L because this includes the mean 25(OH)D concentration of the US population[24] and is the referent category used in several other analyses of 25(OH)D and cancer (12, 25–30). Although the top category had few subjects, it was retained because of interest in higher vitamin D concentrations. Results from both the overall models and those stratified on season of blood collection are presented in this manner. The second approach used season-specific quartiles of 25(OH)D, calculated on the basis of the distribution among the colorectal controls split by season and entered into the models as indicator variables. The third approach used season-standardized 25(OH)D quartiles (described below). Results from the latter 2 approaches are presented with the lowest quartile as the referent category. Tests for linear trend were conducted by assigning to each category an ordinal value (1–5 or 1–4 as necessary) and then treating this parameter as a continuous variable.

The vitamin D assays for this study were conducted in conjunction with ATBC studies of other cancer sites, which included a total of 1,221 controls. Based on monthly median 25(OH)D concentrations in this larger set of controls, season was defined as the "darker months" (November–April) versus the "sunnier months" (May–October). The season-standardized 25(OH)D values were calculated by regressing log-transformed 25(OH)D concentration against calendar week of blood collection in the 1,221 controls, using a locally weighted polynomial regression method, and creating quartiles of the residuals.[20,31] This method was also used to create a smoothed plot of predicted 25(OH)D values by calendar week of blood collection in the 1,221 controls.

Variables assessed for confounding included those shown in Table 1, as well as coastal residence, urban residence, marital status, education, vocational training, tooth loss, and intakes of butter, poultry, alcohol, iron, folate, and pyridoxine. Potential confounders were defined as covariables that were associated with either colon or rectal cancer or produced a >10% change in the 25(OH)D coefficients when added to the univariate models, and which were not likely to be direct determinants of 25(OH)D (e.g., vitamin D intake or supplement use). All potential confounders were then considered jointly and removed from the models if that produced a <10% change in the 25(OH)D coefficients (analyses were conducted separately for each outcome and each season). The resulting confounders for colon cancer were serum α-tocopherol, β-carotene, and retinol, and the confounders for rectal cancer were serum α-tocopherol, β-carotene, and years of smoking. Given the similarity between the colon and rectal cancer confounders and because the 25(OH)D risk estimates were not affected when all 4 covariables were included, these 4 identified confounders were retained in all models.

Stratum-specific subgroup models were fitted using the 25(OH)D season-specific quartile variable, with the lowest quartile serving as the referent category. Models stratified into separate subgroups based on age, body mass index, number of cigarettes smoked per day, calcium intake, serum α-tocopherol, β-carotene, and retinol concentrations (split at the median), physical activity, and α-tocopherol and β-carotene intervention groups were fitted using unconditional multivariate logistic regression, with additional adjustment for the matching factors. Analyses stratified on date of the case diagnosis (split at the median diagnosis date of November 15, 1999, for both colon and rectal cancer cases) were conducted conditionally. We evaluated effect modification statistically by comparing models with and without a cross-product term of 25(OH)D (categorical) and the effect modifier, using the log-likelihood test. Statistical analyses were performed using SAS software, version 9.1.3 (SAS Institute, Inc., Cary, North Carolina), and all P values were 2-sided.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.