Selenium supplementation and colorectal adenomas: An analysis of the nutritional prevention of cancer trial
Article first published online: 10 OCT 2005
Copyright © 2005 Wiley-Liss, Inc.
International Journal of Cancer
Volume 118, Issue 7, pages 1777–1781, 1 April 2006
How to Cite
Reid, M. E., Duffield-Lillico, A. J., Sunga, A., Fakih, M., Alberts, D. S. and Marshall, J. R. (2006), Selenium supplementation and colorectal adenomas: An analysis of the nutritional prevention of cancer trial. Int. J. Cancer, 118: 1777–1781. doi: 10.1002/ijc.21529
- Issue published online: 20 JAN 2006
- Article first published online: 10 OCT 2005
- Manuscript Accepted: 28 JUL 2005
- Manuscript Received: 22 MAR 2005
- National Cancer Institute. Grant Number: RO1 CA49764
- colorectal cancer;
- colorectal adenomas
Selenium status has been inversely associated with colorectal cancers (CRC) and adenomas. This investigation evaluates the association between selenium supplementation and prevalent and incident colorectal adenomas and CRC detected during the Nutritional Prevention of Cancer trial follow-up. Of the 1,312 randomized to 200 mcg of selenized yeast of matching placebo, 598 participants underwent endoscopic screening (flexible sigmoidoscopy or colonoscopy) for CRC sometime during the follow-up period, which ended in February 1, 1996. There was no colorectal screening performed at baseline. Of those screened, 77% were male (with a mean age of 62.8 years), 42% were former and 25% were current smokers. Adenomas were classified as prevalent (identified at the first endoscopic examination postrandomization during the follow-up period) or incident (identified at the second or subsequent examination). Ninety-nine prevalent and 61 incident adenomas were ascertained. Logistic regression odds ratios (OR) and 95% confidence intervals (CI) were calculated, adjusting for age, gender and smoking status. For prevalent adenomas, there was a suggestive but nonsignificant decrease in risk associated with selenium treatment (OR = 0.67, 95% CI = 0.43–1.05). Subjects in the lowest tertile of baseline selenium (OR = 0.27, 95% CI = 0.09–0.77) and current smokers (OR = 0.27, 95% CI = 0.11–0.66) had significant reductions in risk. The OR for incident adenomas was 0.98 (95% CI = 0.57–1.68). In addition to being associated with a reduced risk of incident CRC, selenium supplementation was associated with a significantly reduced risk of prevalent adenomas, but only among subjects with either a low baseline selenium level or among current smokers. © 2005 Wiley-Liss, Inc.
The Nutritional Prevention of Cancer (NPC) study1 was a placebo-controlled, randomized clinical trial designed to prevent the recurrence of nonmelanoma skin cancer (NMSC) in men and women living in the low soil selenium areas of the Eastern Coastal plain of the US. The results indicated that selenium led to a significant increase in the recurrence of squamous cell carcinoma of the skin.2 The secondary endpoints of the trial (total cancer mortality, total cancer incidence, prostate, colon and lung cancer incidence), defined in 1993, were found to be inversely associated with selenium supplementation, with the greatest reduction in risk seen among subjects in the lower third of the distribution of baseline plasma selenium levels.1, 3, 4
Since the trial began, patients were interviewed at each clinical visit for new illnesses, tests and new medications. These reports included cancer screening procedures and newly diagnosed colorectal polyps, cancers and other gastrointestinal conditions. Starting in 1989, specific questions were added to the clinical interview on general cancer screening practices and exam results, focusing on breast, cervical, colorectal and prostate cancers. Patients were asked repeatedly about their personal history of cancer screening, prior to randomization and throughout the follow-up period. The additional colorectal cancer (CRC) screening queries specifically asked about (i) history of colorectal polyps or cancers (CRC) and (ii) results from fecal occult blood tests (FOBT), flexible sigmoidoscopies (FS) and colonoscopies (CS). All subjects in the trial who were over 50 years of age were encouraged with personal letters and educational materials to be screened for CRC. In addition, subjects were given FOBT kits with instructions to use the kit and then mail them back to the study coordinating office. After kit evaluation, regardless of the result, subjects were contacted by letter and encouraged to have routine CRC screening.
The analysis of the association between selenium supplementation and the prevalence and incidence of colorectal adenomatous polyps in the NPC trial is based on an extended follow-up of NPC participants through the end of the blinded clinical trial (February 1, 1996), when all the subjects were informed of their treatment group status. In addition to evaluating the main effect of selenium supplementation, this investigation evaluated the role of baseline selenium status and other characteristics (age, gender and smoking) in modifying the effects of supplementation on adenoma prevalence and incidence.
Material and methods
The methods for subject recruitment and follow-up for the NPC trial have been described previously.1, 2, 3 Briefly, 1,312 subjects were recruited from 7 dermatology practices located in the Eastern United States, between the years of 1981–1989. Subjects had confirmed recent histories of NMSC, were less than 80 years old and were under treatment by a participating dermatologist. All subjects had an estimated life expectancy of at least 5 years and had had no reported internal cancers within the previous 5 years. Subjects were randomized to treatment or an identical placebo and were followed for an average of 7.9 years until the trial was unblinded on February 1, 1996. The treatment was a 200 μg/day of selenium formulated in a high-selenium baker's yeast tablet provided by Nutrition 21 (La Jolla, CA) through 1995 and by Cypress Systems (Fresno, CA) for the remainder of the trial. Prior to randomization, all subjects signed informed consent forms approved by the University of Arizona IRB committee, Tucson, AZ.
CRC screening was reported by a total of 598 participants at some point during the follow-up period. These procedures included both flexible SC and CS. Prevalent adenomas were defined as adenomas detected from the first CRC screening procedure (either FS or CS) postrandomization. Incident adenomas were restricted to those adenomas found on subsequent examinations, after determination that the colon was clean of adenomas or existing adenomas had been removed. There was no way to determine the proportion of incident adenomas that were missed in previous examinations. All medical records, including pathology and operative reports pertaining to CRC screening, were obtained throughout the course of the blinded trial. These records were reviewed by registered nurses before the information was entered into the database. CRC cases were confirmed by a board-certified medical oncologist.
Plasma selenium concentration was measured using electrothermal atomic absorption spectrophotometry (Perkin Elmer 3030, Perkin–Elmer Corp., Norwalk, CT). All selenium assays were performed at Cornell University by Dr. Gerald Combs. Baseline selenium levels had to be drawn within 4 days (±) from the date of randomization (n = 1,250), as described in previous publications.3, 4
These analyses were based on the subset of subjects (n = 598) who reported CRC screening (FS or CS) during the course of the trial. The statistical plan was structured as an intention-to-treat analysis so that regardless of time on treatment, all subjects were grouped according to original treatment group assignment. t-Tests and χ2 analyses were used to determine differences in the distributions of baseline variables. Logistic regression odds ratios (OR) and 95% confidence intervals (CI) were adjusted for age (continuous), gender and smoking status (never, former and current). Only subjects with 2 or more CRC screening endoscopies were included in the models for incident adenomas (n = 442). Selenium status at baseline was categorized by median and tertiles. The interaction between strata of baseline characteristics (including baseline selenium) and treatment group was evaluated in separate OR models. All statistical analyses were completed using STATA 7.0.5
Table I summarizes the baseline characteristics of the NPC sample with valid baseline selenium levels (n = 1,250) and the subjects that were screened for CRC (n = 598) combined, by treatment group and by the presence or absence of prevalent or incident polyps. Overall, the sample screened for CRC had a mean age of 62.8 years, 77% were male, and 33% were never smokers. They had a mean baseline selenium level of 114.5 ng/ml. Men were more likely to have adenomas, either prevalent or incident, and current smokers were more likely than never smokers to have prevalent adenomas (32.2% vs. 23.0%, χ2p = 0.04). Among CRC screened subjects, ∼78% of both treatment groups had at least 1 FOBT test, and similar proportions had positive results (17–19%) (data not shown). Positive FOBT (28% vs. 16%) and screening with CS (25% vs. 14%) were more common in screened patients with prevalent polyps when compared with screened patients without prevalent polyps (data not shown). Overall, however, FOBT screening did not significantly contribute to the detection of either prevalent or incident adenomas. Patients with prevalent adenomas were followed for an average of 3.6 years (SD = 2.12), while participants with incident adenomas were followed for an average of 5.74 years (SD = 2.07).
|Sample(n = 1,250)||Screened (n = 598)||Selenium3 (n = 298)||Placebo4 (n = 300)||−Adenomas (n = 499)||+Adenomas (n = 99)||−Adenomas (n = 537)||+Adenomas (n = 61)|
|Mean (SD)||63.2 (10.0)||62.8 (9.1)||62.8 (9.1)||62.9 (9.1)||62.3 (9.3)||65.4 (7.4)||62.6 (8.9)||64.0 (8.7)|
|Mean (SD)||25.6 (4.0)||25.5 (3.8)||25.7 (3.8)||25.4 (3.9)||25.5 (3.9)||25.6 (3.8)||25.3 (3.9)||26.0 (4.2)|
|Smoking status (n, %)|
|Never||398 (31.8)||199 (33.2)||101 (33.9)||98 (32.6)||176 (35.2)||23 (23.2)8||134 (21.7)||20 (32.8)|
|Former||497 (39.8)||253 (42.2)||122 (40.9)||131 (43.5)||209 (41.8)||44 (44.4)||165 (43.2)||23 (37.7)|
|Current||355 (28.4)||147 (24.5)||75 (25.2)||72 (23.9)||115 (23.0)||32 (32.3)||83 (21.7)||18 (29.5)|
|Baseline plasma selenium (ng/ml)|
|Mean (SD)||114.2 (22.0)||114.5 (21.8)||115.2 (22.1)||113.9 (21.5)||114.5 (22.1)||114.9 (20.2)||115.3 (22.2)||117.4 (20.7)|
|Number of CRC procedures reported (mean (SD))|
|Sigmoidoscopy||–||1.53 (1.21)||1.61 (1.26)||1.45 (1.11)||1.63 (1.15)||1.02 (0.81)9||1.79 (1.17)||1.68 (1.34)|
|Colonoscopy||–||1.40 (1.50)||1.41 (1.60)||1.39 (1.60)||1.09 (1.38)||2.98 (1.68)9||1.81 (1.65)||3.95 (1.94)10|
Table II summarizes the OR estimates for prevalent and incident adenomas, adjusting for age, gender and smoking status at baseline. For prevalent adenomas, the adjusted OR is 0.67, (95% CI = 0.43–1.05), again with a borderline significance (p = 0.06). Incident adenomas showed a nonsignificant OR = 0.98 (95% CI = 0.57–1.68, p = 0.93).
|Prevalent adenomas||41||58||0.67 (0.43–1.05)||0.08|
|Incident adenomas||29||32||0.98 (0.57–1.68)||0.93|
Table III shows the OR estimates for prevalent adenomas by treatment group and strata of baseline characteristics (age, gender, smoking status and baseline plasma selenium levels). Nonsignificant decreases in risk were observed within the age (≤65 versus >65 years) and gender strata. Among never smokers, there was a nonsignificant increase in risk associated with selenium supplementation (OR = 1.62, 95% CI = 0.66–4.00, p = 0.29), while there was a nonsignificant reduction in risk for former smokers (OR = 0.72, 95% CI = 0.37–1.42). In contrast, among current smokers, the risk was significantly reduced (OR = 0.27, 95% CI = 0.11–0.66, p = 0.004) and the interaction between smoking status and selenium treatment was significant at the p = 0.006 level. Baseline plasma selenium was stratified by median and tertiles plasma levels. Although the OR estimates for values above and below the median showed a reduction of 40–21% respectively, these reductions were not significant. However, a significant 73% reduction in risk with selenium supplementation was seen for subjects in the lowest tertile (<105.5 ng/ml) of baseline plasma selenium (OR = 0.27, 95% CI = 0.09–0.77, p = 0.01). The highest tertile (≥122.0 ng/ml) showed a nonsignificant increase in risk of prevalent adenomas with treatment, which is consistent with other stratified analyses of the NPC trial.3, 6 Additional models that adjusted for time from randomization to first screening procedure did not significantly predict the presence of prevalent adenomas (data not shown). A comparable analysis of incident adenomas found that time on supplementation was not a significant predictor. In addition, a separate stratified analysis of incident adenomas by baseline selenium levels (tertiles and median) showed no significant reduction change in risk (OR for lowest tertile = 1.04, p = 0.88; Below the median OR = 1.23, p = 0.24).
|≤114.2 ng/ml||20||32||0.60 (0.32–1.11)||0.10|
|>114.2 ng/ml||21||26||0.79 (0.42–1.50)||0.47|
|<105.5 ng/ml||5||19||0.27 (0.09–0.77)||0.01||0.39||0.05|
|105.6–122.0 ng/ml||20||24||0.62 (0.31–1.24)||0.18|
|≥122.0 ng/ml||16||15||1.14 (0.52–2.48)||0.75|
The NPC trial is currently the only completed clinical trial among selenium replete subjects, testing the effects of selenium supplementation on the incidence of cancer. As part of this important trial, CRC screening practices and results were ascertained on all participants. Approximately 46% of the randomized subjects were screened by endoscopy at least once during the course of active follow-up. These results suggest that selenium supplementation reduced prevalent adenomas in subjects screened for CRC during the course of the trial. This effect was greatest among current smokers and those whose baseline selenium levels were in the lowest tertile. These effects were not seen for incident adenomas. In addition to the clear issue with sample size, these results might be explained by the relatively short time period to observe an effect of selenium treatment on adenomas and that these incident adenomas may represent a subgroup of adenomas that are insensitive to selenium and grew, despite the selenium-rich environment. Because no adenoma tissue samples were collected from these participants, pursuit of this hypothesis must be left to other investigations.
Colon cancer mortality has been shown to be inversely correlated with per capita selenium intake.7, 8 Multiple cross-sectional and case-control studies have also demonstrated an inverse association between plasma or toenail selenium levels and the risk of colon cancer and colorectal adenomatous polyps, the precursor lesion for colon cancer.9, 10, 11, 12, 13, 14, 15, 16, 17, 18 Clark et al.19 showed that among men undergoing CS for the first time, higher selenium status was inversely associated with a lower colorectal adenoma prevalence. In addition, subjects with a selenium status below a median of 128 ng/ml were more likely to have more adenomas and more likely to have adenomas located in the proximal colon. In a recent study that pooled data from 3 large adenoma recurrence prevention trials, subjects in the highest versus lowest quartile of blood selenium at baseline had a significantly decreased risk of adenoma recurrence (OR = 0.66, 95% CI = 0.50–0.87).20 Other studies have failed to duplicate these findings.21, 22, 23, 24
Several biological models have been explored to explain the mechanism(s) by which selenium might exert a chemopreventive effect on colorectal adenomas and cancers. For example, depression of carcinogen bioactivation, cell proliferation and cell cycling and increased apoptosis have been proposed.25 Preclinical studies show a chemopreventive effect of selenium in colon carcinogenesis models,26, 27, 28 including a reduction of the incidence of colonic aberrant crypts in Fisher F344 rats29 and the prevention of intestinal tumor formation in both APCmin mice and azoxymethane-treated rats.30 Citing evidence of differences in the expression of GI glutathione peroxidase (an isoform of GSH-Px found in the GI tract) and selenoprotein-P31 in adenomas and adjacent normal colonic mucosa, Early et al.23 suggested that expression of selenoproteins in the target tissue can be altered without a corresponding difference in overall selenium status.
Several potential effect modifiers were evaluated in this study, including age, gender, smoking status and baseline selenium status. Current smokers received the greatest benefit from supplementation; the interaction between smoking status and treatment group was significant. Smoking has been established as a risk factor for colorectal adenomas,32, 33, 34, 35 that may influence both the initiation and progression phases of colon carcinogenesis pathway.36 The interaction of treatment and smoking on adenoma prevalence suggests that selenium, like smoking, may also act in the early phases of adenoma development.
Previously reported results from the NPC trial have shown that for total cancers, prostate and lung cancers, the protective effect of selenium supplementation was greatest among individuals in the lowest tertile of baseline plasma selenium.1, 3, 4, 6 The effect of baseline plasma selenium status on cancer has been shown in other studies for esophageal and gastric cancers37, 38 and adenoma recurrence. These results were contradicted in a recent analysis of baseline selenium and adenoma recurrence, nested in an adenoma prevention trial.24 In a sample of 276 cases and 276 matched controls comparing the highest quintile of baseline selenium to the lowest, no significant reduction in adenoma recurrence was observed. Early et al.23 measured total selenium and the concentrations of separate selenoproteins, glutathione peroxidase and selenoprotein-P and found no differences in selenium status between normal, adenoma and colon cancer patients.
A major limitation of this study is that these results are from a secondary analysis, for which this study was not powered. However, these results are consistent with other secondary analyses from this trial, including a decreased incidence of CRC related to treatment, and while not definitive, offer potentially important information on the timing of the chemopreventive activity of selenium in the colon carcinogenesis pathway. The collection of CRC screening information and results initially relied heavily on participant self-reporting. However, this information was collected blinded to treatment group and the same questions were asked repeatedly throughout the follow-up period. In addition, all efforts were made to document these procedures and results with medical record evidence. While all subjects were approached in a uniform way, it is difficult to eliminate unknown influences that may have affected screening and adenoma detection. Family history of CRC and adenomatous polyps was not collected as part of the clinical interview. While this is an important risk factor for CRC and adenomas, randomization should have distributed these risk factors equally between treatment groups. However, the distribution of this factor in the subgroup analysis is not known and may have had an effect on the results. Again, the power to detect clear differences in the risk estimates is limited because of the small numbers of prevalent adenomas (n = 99) and incident adenomas (n = 61) observed in this study.
While the effect of baseline selenium status on cancers of the lung, prostate and colon has been consistent throughout the analyses from the NPC trial, the mechanisms for these effects remain unclear. These subjects were intentionally drawn from geographic areas naturally low in selenium. Environmental factors associated with low soil selenium or genetic variation in selenium metabolism may have rendered these subjects more susceptible to the effects of selenium supplementation. Since these factors may have implications for the success of follow-up intervention studies using selenium, more research into the determinants of selenium status and response in humans is needed. These data show that selenium supplementation is associated with a decrease in prevalent adenomatous polyps, particularly in current smokers and subjects with a low baseline plasma selenium level (<105.5 ng/ml). The chemopreventive effects of selenium supplementation appear impact both the formation of colorectal tumors and the premalignant lesions for CRC. Prospective studies specifically designed to test the chemopreventive efficacy of selenium supplementation on CRC and adenoma incidence are needed to confirm these results.
- 1Effects of selenium supplementation for cancer prevention in patients with carcinoma of the skin. A randomized controlled trial.Nutritional Prevention of Cancer Study Group. JAMA 1996; 276: 1957–63. [published erratum appears in JAMA 1997; 277: 1520]., , , , , , , , , , , , et al.
- 5StataCorp. Stata Statistical Software: Release 6.0, ed. 7.0 College Station, TX: Stata Corporation, 2003.
- 23Selenoprotein levels in patients with colorectal adenomas and cancer. Am J Gastroenterol 2002; 97: 745–8., , , .Direct Link: