Background: In a recent randomized intervention trial, the risk of prostate cancer for men receiving a daily supplement of 200 µg selenium was one third of that for men receiving placebo. By use of a nested case-control design within a prospective study, i.e., the Health Professionals Follow-Up Study, we investigated the association between risk of prostate cancer and prediagnostic level of selenium in toenails, a measure of long-term selenium intake. Methods: In 1986, 51 529 male health professionals aged 40–75 years responded to a mailed questionnaire to form the prospective study. In 1987, 33 737 cohort members provided toenail clippings. In 1988, 1990, 1992, and 1994, follow-up questionnaires were mailed. From 1989 through 1994, 181 new cases of advanced prostate cancer were reported. Case and control subjects were matched by age, smoking status, and month of toenail return. Selenium levels were determined by neutron activation. All P values are twosided. Results: The selenium level in toenails varied substantially among men, with quintile medians ranging from 0.66 to 1.14 µg/g for control subjects. When matched case-control data were analyzed, higher selenium levels were associated with a reduced risk of advanced prostate cancer (odds ratio [OR] for comparison of highest to lowest quintile = 0.49; 95% confidence interval [CI] = 0.25–0.96; P for trend = .11). After additionally controlling for family history of prostate cancer, body mass index, calcium intake, lycopene intake, saturated fat intake, vasectomy, and geographical region, the OR was 0.35 (95% CI = 0.16–0.78; P for trend = .03). Conclusions: Our results support earlier findings that higher selenium intakes may reduce the risk of prostate cancer. Further prospective studies and randomized trials of this relationship should be conducted. [J Natl Cancer Inst 1998;90:1219–24]
Selenium is an essential trace nutrient and is critical for the activity of glutathione peroxidase, which may protect DNA and other cellular molecules against oxidative damage (1–3). At relatively high levels, selenium protects against the action of certain carcinogens in various animal models (4,5). In the United States, lower age-specific death rates from some types of cancer have been observed in states with higher soil selenium levels (6). Most studies (7) based on prediagnostic serum selenium levels tend to support an association between low levels of selenium and the risk of various cancers. In a recent double-blind, placebocontrolled cancer prevention trial (8) in which 200 mg selenium was given daily to patients with histories of basal and squamous carcinoma, selenium supplementation did not protect against the development of recurrent skin cancers as a primary end point but was inversely associated with the incidence of and mortality from total prostate, lung, and colorectal cancers.
The data are surprisingly sparse regarding a relationship between selenium and prostate cancer, a disease that causes more than 40 000 deaths annually in the United States. Combined prospective data based on serum selenium levels from three studies (9–11) sum to a total of only 75 cases of prostate cancer, with the two smallest studies (9,10) suggesting a statistically nonsignificant inverse association. The recent trial by Clark et al. (8), based on a total of 48 cases of prostate cancer, found men who were randomly assigned to receive 200 mg selenium daily to be at only one third the risk for prostate cancer (P = .001) compared with those receiving a placebo.
The provocative results from this trial support further study of the relationship between selenium and prostate cancer. This trial indicates that selenium supplementation at doses relatively high compared with the recommended daily allowance for men (70 mg) may have an anticancer effect, but few studies address the relation between the selenium acquired from normal dietary intake and the risk of prostate cancer. Toenail selenium levels reflect selenium intake integrated over many months and, thus, are useful indicators of past selenium intake in epidemiologic studies (12–15). We therefore assessed the association between prediagnostic levels of selenium in toenails and subsequent risk of advanced prostate cancer by use of a nested case-control design within the Health Professionals Follow-Up Study (16,17).
Subjects and Methods
The Health Professionals Follow-Up Study is an ongoing prospective cohort study of diet and coronary disease and cancer among 51 529 men in the United States from all 50 states who were aged 40– 75 years in 1986 (16,17). The study began in 1986 when all cohort members completed a mailed questionnaire that assessed dietary intake, risk factors for cancer and heart disease, and medical history. Among these men, 1595 (3%) whose reported daily energy intake from the 1986 semiquantitative food frequency questionnaire was less than 800 or greater than 4200 kcal/day or who left 70 or more questions on the questionnaire blank were excluded from the baseline population of 51 529 men. Every 2 years, follow-up questionnaires were sent to update information on newly diagnosed prostate cancers. We used the National Death Index to ascertain the status of nonrespondents. In 1987, 33 737 sets of toenails were collected from the cohort members and were stored for subsequent analysis. This study was approved by the Human Research Committee at the Harvard School of Public Health.
On the 1988, 1990, 1992, and 1994 questionnaires, each participant was asked if he had had a diagnosis of prostate cancer during the prior 2-year period. A letter was sent to all men who reported an incident prostate cancer on the follow-up questionnaires to confirm the report and to ask for permission to review medical records. After repeated mailings, we received questionnaires from or had confirmation of deaths for more than 94% of the eligible participants through 1994.
Prostate cancer was staged by physicians conducting the cohort study according to information received from medical reports. In these analyses, we considered only extraprostatic (advanced) cases of adenocarcinoma of the prostate (stage C or D). Stage C cancers are those cancers that have perforated the capsule but remain localized to the periprostatic area; stage D cancers are those that have metastasized to a remote site (17). We included only case subjects diagnosed at least 2 years after collection of the toenail specimen to minimize the potential impact of undiagnosed cancers on selenium levels. All prostate cancers for this analysis were confirmed through a review of histopathologic reports from medical records.
Each case subject with advanced prostate cancer was matched to a Health Professionals Follow-Up Study participant without a diagnosis of prostate cancer on the basis of age within 1 calendar year, smoking status (current, past [years since smoking was stopped], and never), and date of toenail return within 1 month (the dates of toenail return were matched to reduce differences in levels due to seasonal variation). The control subjects had to be alive and cancer free (other than nonmelanoma skin cancer) at the time that the corresponding case subject was diagnosed with cancer.
Assessment of the Level of Selenium in Toenails
Before the analysis, the toenail clippings from all toes were washed with deionized water by use of a sonicator. Case and control specimens were analyzed together, but in random order, with the case status unknown to the laboratory personnel. Selenium was analyzed by instrumental neutron activation analysis at the University of Missouri Research Reactor, Columbia, MO (18). 77mSe is produced from 76Se by neutron capture during reactor irradiation. The isomeric transition of 77mSe emits a 161.9-keV photon that is measured by high-resolution gammaray spectroscopy. The selenium concentration is quantified by standard comparison as described (14,19). The coefficient of variation for duplicate selenium neutron activation analysis for nails was less than 2%.
Assessment of Diet and Other Risk Factors
Nutrient intakes were calculated from the 1986 dietary questionnaire, which included 131 food items with specified portion sizes. The cohort member reported the average frequency at which he had consumed each item during the previous year. The average daily intake of nutrients was calculated by multiplying the frequency of consumption of each food item by the nutrient content, calculated from composition values from U.S. Department of Agriculture sources (20) supplemented with other data from the Harvard University Food Composition Database (November 1993), and summing the nutrient intake from all the food items. Specific brands of multivitamin and selenium supplements were used for the computation of selenium supplement intake (17). The nutrients were adjusted for total energy intake by using the residual method (21).
The food-frequency questionnaire was evaluated in this cohort by use of two 1-week dietary records from a random sample of 127 men living in the Boston area (22). The correlation coefficients, adjusted for total energy intake and week-to-week variation in diet records between the food frequency questionnaire and the average of two 1-week diet records, averaged 0.65 for nutrients (22) and 0.63 for specific foods (23).
On the 1990 questionnaire, participants were asked about a history of prostate cancer in their father and in any brothers. In each follow-up questionnaire, we also inquired whether participants had received a digital rectal examination during the previous 2 years. The history of screening by measurement of prostatic-specific antigen (PSA) was assessed in 1994.
Because the distribution of toenail selenium levels was skewed toward higher levels, the values were log transformed to improve normality. The mean and distribution of sample weights did not vary by case-control status, but the Pearson correlation coefficient between the log-transformed selenium level and the sample weight was -0.34 (P<.0001). Therefore, to reduce extraneous variation, we adjusted for the toenail sample weight by regression analysis of the log-transformed selenium values on the specimen weight. A constant, the log-transformed predicted selenium level for the mean of the sample weight, was added back to the residuals obtained from the linear regression model, and the values were inverse log transformed.
Differences in selenium levels between case and control groups were tested by using nonparametric two-sample tests. The association between the selenium level in toenails and the risk of advanced prostate cancer was expressed as an odds ratio (OR), with a 95% confidence interval (CI). Selenium values were categorized into quintiles based on their distribution among the control subjects. Logistic regression analysis was used to control for known and potential risk factors. We tested for linear trend by using the median value of each quintile of selenium as a continuous variable in a multiple logistic regression model with covariates. All P values reported are two-sided. In addition, we evaluated whether the selenium level was inversely linearly associated with the risk of advanced prostate cancer by using restricted cubic splines (24). Nutrients and body mass index (BMI) were also grouped into quintiles. Smoking status was a matching variable; in multivariate models, we additionally controlled for category of cigarettes per day for current smokers. We tested for an interaction between selenium status and vitamin E intake because these nutrients may interact as antioxidants (25).
Conditional and unconditional logistic regression gave similar results; therefore, in this report, only the results by the former method were given for the overall analyses. However, we used unconditional logistic regression analysis when we stratified the analyses across levels of vitamin E intake.
The association between the selenium level in toenails and the geographic distribution of selenium in soil, based on state of residence in 1986, was assessed by using published selenium values (6,26,27). The high-level selenium states had soil selenium levels in the range of greater than or equal to 0.1 µg/g (18 states), medium-level states had soil selenium levels in the range of 0.05–0.09 µg/g (11 states), and low-level states had soil selenium levels between 0.02 and 0.05 µg/g (19 states).
Table 1 shows the baseline characteristics of control subjects according to selenium levels in toenails. The intake of lycopene, calcium, and saturated fat did not vary appreciably across the levels of selenium in toenails. The percentage of men using supplements that contained selenium (including multivitamins) was lowest among those in the lowest quintile of toenail selenium level. The percentage of men living in states with high soil selenium content was 14% in the low quintile of toenail selenium and 27% in the highest quintile of toenail selenium.
Baseline characteristics of the case and control subjects are shown in Table 2. The median ages and proportions of current, never, and past smokers at the baseline in 1986 were the same in the case and control subjects due to the way in which they were matched. The median BMI in 1986 was the same in the two groups. Reported routine digital rectal examinations were slightly lower on the 1988 questionnaire for the case subjects. The proportion of men with a family history of prostate cancer was higher in the case subjects. A history of diabetes was more common among the control subjects; intake of lycopene was lower among the case subjects, whereas calcium and saturated fat intakes were higher in the case subjects. These associations are similar to those previously reported in the entire cohort (16,17,28,29).
The range of toenail selenium levels among the baseline population (control subjects) was between 0.53 and 7.09 µg/g. The mean toenail selenium level was higher in the control subjects (0.96 µg/g) than in the case subjects (0.82 µg/g). This difference was statistically significant (nonparametric paired test, P = .05).
Table 3 shows the ORs of prostate cancer according to toenail selenium levels. After adjustment for age, smoking, and date when toenail samples were returned by the matching case and control subjects, the OR was 0.49 (95% CI 4 0.25–0.96), comparing the highest with the lowest quintiles. The P value for trend using conditional logistic regression analysis on all of the data was .11. After additionally controlling for potential prostate cancer risk factors, including a family history of prostate cancer; BMI; vasectomy; and intakes of lycopene, calcium, and saturated fat, the OR was 0.39 (95% CI 4 0.18– 0.84); when all of the data were used, the P value for trend was .04. Additional adjustment for region (high, medium, and low levels of soil selenium content) did not change the risk appreciably. The cubic spline analysis indicated that the log incidence rate of advanced prostate cancer risk decreased throughout the entire range of selenium levels.
We examined the relationship between the selenium level in toenails and the risk of advanced prostate cancer across strata of vitamin E intake by use of unconditional logistic regression analysis. Among men (n 4 233) with daily intakes of vitamin E less than or equal to 30 IU, the ORs across quintiles 1–5 were 1.0 (reference), 0.63 (95% CI 4 0.26–1.53), 0.41 (95% CI 4 0.16–1.07), 0.75 (95% CI 4 0.30–1.89), and 0.29 (95% CI 4 0.11– 0.77), respectively; the P value for trend was .025. Among 125 men who obtained substantial amounts of vitamin E from supplements (>30 IU/day), those in the lowest quintile of selenium were at highest risk, but there was no trend (P = .68) across quintiles 2–5 (ORs across quintiles 1–5 were 1.0 [reference], 0.29 [95% CI = 0.07–1.18], 0.27 [95% CI 4 0.06–1.17], 0.32 [95% CI 4 0.07–1.53], and 0.41 [95% CI = 0.09–1.83], respectively). We tested for multiplicative interaction between selenium level and intake of vitamin E by including a variable for the product of selenium level and vitamin E in a logistic regression model, along with selenium and vitamin E individually, and testing for significance of this interaction term. No interaction between selenium and vitamin E was evident (P = .30).
We evaluated potential modification of the selenium effect by the level of smoking, which is known to reduce selenium levels (14). The inverse association between selenium and prostate cancer did not differ by smoking status (P value for multiplicative interaction >.5). When current smokers were excluded, the multivariate OR did not change appreciably (between high and low quintiles OR 4 0.42; 95% CI =0.20–0.91; P for trend4 .03). Also, exclusion of current smokers or men who had quit within the past 5 years yielded similar results (OR 4 0.44; 95% CI 4 0.20–0.97, controlling for the number of cigarettes among those who smoked more than 5 years in the past; P for trend 4 .05).
We found a strong inverse association between the prediagnostic selenium level assessed in toenail clippings and the risk of advanced prostate cancer. This inverse association was not a result of any confounding effect of various factors, including age, other dietary factors, smoking, BMI, geographic region, family history of prostate cancer, or vasectomy. Although the technologies for diagnosing prostate cancer changed during the course of the study, the probability of being diagnosed conditional on cancer was likely to be independent of selenium level, and the vast majority of advanced prostate cancers would be clinically diagnosed in this population. Moreover, the percent of men who had reported a recent negative digital rectal examination at baseline did not differ substantially across levels of selenium. Over the course of the study, screening by PSA measurement became widely used, and rates of PSA testing were similar by level of baseline selenium (across quintiles 1–5, respectively, 96%, 84%, 93%, 85%, and 84% of the control subjects had had a PSA test by 1994).
We minimized the possibility that undiagnosed cancers artifactually lowered selenium levels by excluding the first 2 years of follow-up. Also, the selenium level in 1987 was identical in case subjects diagnosed across the follow-up period (0.82 µg/g for cases of cancer diagnosed in the 2-year periods from 1989 through 1990, from 1991 through 1992, and from 1993 through 1994); if undiagnosed cancers had reduced selenium levels, the levels in case subjects should have been lower in the early years of followup. The mean selenium level was the same (0.82 µg/g) for case subjects who had reported a recent digital rectal examination at baseline and those who had not. Because smoking can lower selenium levels, we matched subjects for cigarette use to exclude this as a confounder and found that the inverse association between the selenium level in toenails and the risk of prostate cancer persisted among never smokers and past smokers who had not smoked for at least 5 years.
In this study, the association between selenium level and the risk of prostate cancer at stages A and B was not evaluated. On the basis of our results, selenium could plausibly either influence early stages of prostate carcinogenesis, ultimately leading to a decrease in overall prostate cancer incidence, or affect the progression from organ confined to extraprostatic disease. Alternately, cancers at stage A or B may be inherently different from cancers at stage C or D. In either case, the advanced stage cases are most important clinically because these lead to symptoms and death, and they are less prone to detection bias.
Toenail clippings provide useful measures of long-term average intake of selenium. Longnecker et al. (13) showed that toenail selenium levels reflect intake integrated over 26–52 weeks, and the Pearson correlations with selenium intake were .78 for whole blood and .67 for toenails. Furthermore, toenail selenium levels indicate long-term status; among women in the Nurses' Health Study, the Spearman correlation was .48 for specimens obtained 6 years apart from the same person (15). Single measures of various parameters (e.g., blood pressure, cholesterol, or glucose), which have similar correlations over time, are strong predictors of disease. If misclassification is random and nondifferential by case- control status, the magnitude of any true association would tend to be underestimated.
The level of selenium in toenails reflects intake from dietary and supplementary sources. The selenium content of foods is largely dependent on the selenium content of the soil where the food or animal feed was grown. In our study, the regional selenium content of soil and the use of selenium supplements predicted selenium levels in toenail clippings, albeit moderately, supporting the use of toenails as an indicator of selenium intake. Other unmeasurable sources of variation, such as dietary pattern and absorption and excretion, contribute to produce a wide range in levels, from a median of 0.66 parts per million (wt/wt) in the bottom quintile to 1.14 parts per million (wt/wt) in the top quintile. The difference in these levels is substantial and is comparable to the range found in a sample of people living in the Boston area (0.74 µg/g), a region with a relatively low soil selenium content, and another sample of people living in South Dakota (1.17 µg/g), an area with a very high soil selenium level (30). With the use of formulas derived from direct measurements of selenium in the diet based on multiple, 1-day, duplicate-plate food composites and from toenail specimens, the estimated daily median selenium intake was found to be 86 mg among men in the low quintile of selenium level in toenails and 159 mg among the men in the high quintile (13). These should be considered rough estimates because of differences, such as in the chemical form of selenium, in the diet of our population and that from which the formulas were derived (South Dakota and Wyoming).
Prospective data on the relationship between selenium status and prostate cancer are sparse. Two small studies of 11 (9) and 13 (10) case subjects found nonsignificant inverse associations between the level of selenium in plasma and the risk of prostate cancer, but a larger study of 51 case subjects did not (11). However, in a recent randomized selenium supplementation trial (8) conducted in areas with low soil selenium content, men who received 200 mg of selenium daily had a relative risk of 0.35 (95% CI 4 0.18– 0.65) for prostate cancer compared with those who received a placebo (P = .001). These results, based on 48 cases of prostate cancer, were remarkably similar to our relative risk based on a comparison of high and low quintiles of toenail selenium levels. In the trial, the reduction in risk among those randomly assigned to receive selenium was observed for individuals with cancers at stage A or B as well as for individuals with cancers at more advanced stages (31).
The results from the trial alone do not allow us to distinguish whether a benefit occurs only at relatively high supplemental doses or whether a relationship within normal dietary ranges exists; the men received 200 mg daily in addition to their normal dietary intake (the recommended daily allowance for men is 70 mg). Because supplementary intake in our population was low, our results suggest that the recommended and actual dietary intakes of selenium may be suboptimal for many men in the United States with regard to prostate cancer risk. Our analyses indicated an inverse association that was linear throughout the entire range of selenium levels.
Vitamin E is known to interact with selenium in preventing the formation of oxidative products of polyunsaturated fatty acids, some of which may be carcinogenic (25). Also, the Alpha-Tocopherol, Beta-Carotene Trial indicated a benefit of a-tocopherol on prostate cancer (32). The inverse association between selenium intake and advanced prostate cancer risk was slightly stronger after excluding men with an intake of vitamin E that exceeded 30 IU/day that was mostly from supplementary sources (OR 4 0.29). The interaction between selenium and vitamin E was not statistically significant, but our sample size to test for interaction was limited.
The biologic mechanism whereby selenium affects prostate cancer is unknown, but several mechanisms have been hypothesized. Selenium was suggested as an anticarcinogenic agent by Griffin (33) because of its role as an essential component of selenium-dependent glutathione peroxidase. This hypothesis is supported by the mutagenicity and carcinogenicity of malonyldialdehyde, which is thought to be formed with selenium deficiency. Regulation of apoptosis appears to be an important determinant of cancer risk, and antitumorigenic activities of selenium compounds have been related to apoptotic responses (34,35).
In conclusion, the findings of our study provide further support for the hypothesis that a higher intake of selenium may reduce the risk of advanced prostate cancer. Our results support further prospective studies and randomized trials of this relationship to confirm or to refute the findings and to determine optimum dose levels.