Selenium and Mortality in the Elderly: Results from the EVA Study

Abstract

Background: Inadequate plasma selenium can adversely affect the maintenance of optimal health; therefore, reported decreases in plasma selenium in an aging population are cause for concern. To further examine this hypothesis, we explored the relationships between plasma selenium and mortality in an elderly population: the EVA (Etude du Vieillissement Artériel) study.

Methods: The EVA study was a 9-year longitudinal study with 6 periods of follow-up. During the 2-year period from 1991 to 1993 (EVA0), 1389 men and women born between 1922 and 1932 were recruited. The effects of plasma selenium at baseline on mortality were determined by Cox proportional hazards regression analysis, adjusting for the following variables: sociodemographic characteristics, dietary habits, health, and cognitive factors.

Results: During the 9-year follow-up, 101 study participants died. Baseline plasma selenium was higher in individuals who were alive at the end of follow-up [mean (SD), 1.10 (0.20) μmol/L] than in those who died during the follow-up [1.01 (0.20) μmol/L; P <10⁻⁴]. Mortality rates were significantly higher in individuals with low selenium [increments = 0.2 μmol/L; relative risk (RR) = 1.56 (95% confidence interval, 1.28–1.89)]. After we controlled for various potential confounding factors, this association remained significant [RR = 1.54 (1.25–1.88)]. When the underlying causes of death were considered, we found an association with cancer-related mortality [adjusted RR = 1.79 (1.32–2.44)].

Conclusions: Even if it is premature to present selenium as a longevity indicator in an elderly population, our results are in accordance those of large, interventional, randomized trials with selenium, which suggest that this essential trace element plays a role in health maintenance in aging individuals.

Selenium is a trace element found in selenoproteins in the form of an amino acid, selenocystein, which is essential for enzymatic functioning of selenoproteins such as glutathione peroxidase, thioredoxine reductase, and iodine deiodinase (1)(2). Glutathione peroxidase acts against hydrogen peroxide and lipid peroxidation and is an important line of defense against free radicals; thioredoxine reductase is involved in nucleus redox status; and iodine deiodinase is involved in thyroid hormone metabolism, which is frequently affected in the elderly. Selenium also has an anticarcinogenic effect that is thought to be induced by the production of methylselenol, a seleno-metabolite that affects gene expression and modifies cell cycling and immune functions(3). Inadequate amounts of plasma selenium may contribute to the disease process in some cancers and cardiovascular diseases(4).

The anticarcinogenic properties of selenium at high doses suggest a possible link between selenium deficiency and cancer (3). A recent metaanalysis of 14 randomized trials that studied whether antioxidant supplements may prevent gastrointestinal cancers reported that selenium might have a beneficial effect on the incidence of gastrointestinal cancers(5); however, prospective longitudinal studies(6)(7)(8), case–control studies(9)(10)(11), and one randomized interventional trial(12) investigating the relationship between selenium and cancer produced conflicting results.

Selenium may also protect against cardiovascular diseases via counteraction of lipid oxidation by glutathione peroxidase and reduction of platelet aggregation (13). Epidemiologic studies have shown conflicting results(6)(14)(15)(16). An animal study analyzed the role of selenium in fly mortality and showed that the survival rate of flies fed a diet deficient in selenium was one half that of flies fed a diet supplemented with optimal amounts of selenium(17).

The aim of the Etude du Vieillissement Artériel (EVA)1 study was to evaluate the relationships between cognitive decline (18), vascular disease(19), and oxidative stress in elderly individuals. Our team has shown previously that EVA participants with low selenium concentrations at baseline exhibited an increased risk of cognitive decline(18). With the aging population, it is essential to investigate antioxidant status as a potential predictive health factor. Among antioxidants, selenium is of particular interest because in Europe, selenium status, determined by selenium concentrations in serum or plasma(1) or by selenium intake(20)(21), is not adequate for optimal glutathione peroxidase, selenoprotein P, and immune functions(1)(2). Inadequate selenium status may reduce human life expectancy either by accelerating the aging process or by increasing vulnerability to various diseases. To further examine this hypothesis, we explored the relationship between baseline plasma selenium concentrations and mortality.

Materials and Methods

study population

The EVA study is a 9-year longitudinal study with 6 periods of follow-up (18)(19)(22). During the first 2 years, 1991–1993 (EVA0), 1389 volunteers (men and women; age range, 59–71 years) residing in the town of Nantes (western France) were recruited from electoral rolls and, to a lesser extent, via information campaigns. All participants were community residents and were able to undergo a complete examination in the EVA study center, where they spent half a day(23). The next follow-up periods were EVA 2 (1993–1995), EVA3 (1995–1997), EVA4 (1997–1999), EVA5 (1999–2000), and EVA6 (June 2000–December 2001). The number of participants who completed a general questionnaire during each period were as follows: 1272 for EVA2, 1188 for EVA3, 1042 for EVA4, 792 for EVA5, and 705 for EVA6. The study protocol was approved by the Ethics Committee of the University Center Hospital of Kremlin-Bicêtre (Paris), and signed, informed consent was obtained from all participants at enrollment.

data collection

Vital statistics and date and cause of death were collected throughout the 9 years of follow-up. For each of the EVA steps and at the end of the 9th and last year of study, the health status of individuals for whom we had no feedback were collected from town hall civil registries. The cause of death was determined with the help of both families and physicians.

sociodemographic and other variables at baseline

The questionnaire elicited information on standard demographic variables such as sex and age. Educational achievement was divided into 2 categories: no school or primary school (primary school) and high school or university (high school or beyond).

Information about medication use, smoking status, and alcohol consumption was obtained from a standardized questionnaire at baseline. Individuals were classified as current smokers, nonsmokers, or former smokers. Alcohol consumption was determined from the estimated typical amount of alcoholic beverages ingested weekly and expressed in milliliters of alcohol per day. From this information, 3 groups were identified: abstainers, consumers of <20 mL per day, and consumers of >20 mL per day. Individuals who took 3 or more medicines per day were considered high medication users.

health variables

At baseline, participant weight and height were measured, and the body mass index was calculated; a body mass index >29 kg/m² defined obesity. Participants with plasma glucose ≥7.80 mmol/L (1.40 g/L), those taking antidiabetic medications, or those reporting a history of diabetes were considered to have diabetes. Participants were considered to have dyslipidemia if they had a total cholesterol ≥2.80 g/L, used lipid-lowering drugs, or reported a history of dyslipidemia. Two independent measurements of systolic and diastolic blood pressure were made with a digital electronic tensiometer (SP9 Spengler) after a 10-min rest; the mean values were used in the analysis. Individuals with systolic blood pressure ≥140 mmHg or diastolic blood pressure ≥90 mmHg, those using antihypertensive medications, and those reporting a medical history of hypertension were classified as hypertensive. Individuals who reported a history of myocardial infarction, angina pectoris, or stroke and those who used vascular drugs were classified as having a history of cardiovascular disease. The Mini Mental Status Examination (MMSE) was used for cognitive assessment. The MMSE includes 18 items that assess various cognitive skills, with scores ranging from 0 to 30 (24). The lowest cognitive functioning was defined by an MMSE score below the 25th percentile of MMSE score distribution, which corresponded to a MMSE cutoff score of 28.

plasma selenium evaluation

We determined serum selenium with electrothermal atomic absorption spectrometry (PerkinElmer; 5100) according to Arnaud et al. (25). A selenium electrodeless discharge lamp and a Zeeman longitudinal background correction were used. Serum diluted in a solution containing 0.1 mol/L nitric acid, 2 g/L Triton X-100, and matrix modifier was introduced onto the platform of a pyrolytic graphite furnace. Concentration was obtained with an addition calibration. Seronorm® trace element serum (Sero) was chosen as an internal quality control. The target value was 1.30 μmol/L, and the mean (SD) observed value was 1.22 (0.10) μmol/L. Imprecision (as the CV) varied from 1.4% (within-run; n = 20) to 8% (between-run; n = 22) at a concentration of 1.30 μmol/L. In addition, the laboratory participated in 2 interlaboratory comparison trials organized by the French Society for Clinical Biology (Nancy, France) and the “Centre de Toxicologie du Québec” (Sainte-Foy, Québec, Canada) with results within the range of acceptable values.

statistical analysis

Vital statistics, sex, education, smoking status, alcohol intake, medication use, obesity, and health variables were considered categorical variables. Participant age was considered a continuous variable. Plasma selenium at baseline followed a gaussian distribution. Plasma selenium was first analyzed as a continuous variable, with 0.2 μmol/L corresponding to 1 SD of selenium distribution. For the second analysis, we considered plasma selenium concentration by quartiles. The median (range) values for each quartile (Q) were as follows: 0.87 (0.18–0.95) μmol/L for Q1 (below the 25th percentile); 1.03 (0.96–1.09) μmol/L for Q2 (at or above the 25th to below the 50th percentile); 1.15 (1.10–1.21) μmol/L for Q3 (at or above the 50th to below the 75th percentile); and 1.32 (1.22–1.97) μmol/L for Q4 (75th percentile and above). Baseline characteristics of participants who died during the follow-up period and those who remained alive were compared with the Student t-test for continuous variables and the χ² test for categorical variables. Correlations between plasma selenium and the other continuous variables were calculated with the Pearson correlation coefficient. Survival was analyzed with actuarial methods, and a log-rank test was used to compare survival rates of individuals in different plasma selenium quartiles. The effects of plasma selenium on mortality were determined by the Cox proportional hazard regression analysis adjusted for potential confounding variables. We verified the proportionality assumption by adding a time-dependent variable to the model (26). The results of the Cox multivariate regressions were expressed by relative risk (RR) with the 95% confidence interval. All interactions between plasma selenium and other variables were tested. Statistical analysis was performed with SAS software, Ver. 8.2 (SAS Institute.).

Results

baseline characteristics of surviving vs nonsurviving participants

Of the 1389 study participants, 101 died during the 9-year follow-up period (Table 1¹ ). For 88.1% of the individuals who died, the cause of death was determined by the general practitioner. Cancer was the leading cause of death (n = 45; 44.5%), followed by cardiovascular diseases (n = 22; 21.8%). The main factors related to mortality were: sex (male), smoking (current and former smoker), alcohol intake, medication use, obesity, diabetes, hypertension, and cardiovascular diseases. Surprisingly, age, education, and dyslipidemia were not related to mortality. Mean concentrations of plasma selenium measured at EVA0 were significantly higher in surviving individuals [mean (SD), 1.10 (0.20) μmol/L [than in those who died [1.01 (0.20) μmol/L; P <10⁻⁴]. Survival bivariate analysis showed that the RR of mortality during the EVA follow-up increased significantly with a decrease in plasma selenium [selenium = 0.2 μmol/L; RR = 1.56 (95% confidence interval, 1.28–1.89); P <10⁻⁴]. Comparisons of survival distributions among quartiles of plasma selenium showed that mortality increased in subgroups with low plasma selenium concentrations (Fig. 1¹ ).

At baseline, plasma selenium concentrations were not significantly different between men and women (P = 0.189) and were not correlated with age in either sex (r = −0.05; P = 0.09). Plasma selenium was significantly higher in individuals who were in the highest education category [1.08 (0.20) μmol/L for those with less education vs 1.11 (0.20) μmol/L for those with more education; P = 0.003]. No link was observed between selenium and smoking status, alcohol consumption, or medication use. Concerning health status, selenium concentrations were higher in individuals with dyslipidemia [1.07 (0.19) μmol/L for normolipidemic individuals vs 1.13 (0.21) μmol/L for dyslipidemic individuals; P <0.001] and lower in obese individuals [1.10 (0.20) μmol/L in nonobese individuals vs 1.04 (0.19) μmol/L in obese individuals; P <0.001]. For diabetes, hypertension, and cardiovascular diseases, no significant association was found. As shown previously (23), mean plasma selenium concentrations were significantly lower in individuals with the lowest cognitive function [1.07 (0.21) μmol/L] than in other individuals [1.11 (0.20) μmol/L; P = 0.002].

The complete multivariate Cox models, controlling for all factors, showed that a 0.2 μmol/L decrease in plasma selenium was significantly associated with a higher mortality risk [RR = 1.54 (1.25–1.88); P <10⁻⁴]. Among all other included factors, only sex, diabetes, and a history of cardiovascular disease remained significantly associated with mortality risk (Table 2² ). When we analyzed plasma selenium by quartiles, we observed that individuals with plasma selenium concentrations in the 2 lowest quartiles (Q1 and Q2) had a significantly higher risk of mortality than individuals in the highest plasma selenium quartile [Q1 vs Q4: RR = 3.34 (1.71–6.53); P = 0.0004; Q2 vs Q4: RR = 2.49 (1.25–4.94); P = 0.009]. We found no significant association for individuals who had a plasma selenium concentration in Q3 compared with individuals in Q4 [Q3 vs Q4: RR = 1.67 (0.78–3.56); P = 0.18].

plasma selenium and causes of death

In the individuals who died from cancer (n = 45), the mean (SD) baseline plasma selenium concentration was 1.00 (0.22) μmol/L. The baseline plasma selenium concentration was 1.06 (0.20) μmol/L in individuals who died from vascular diseases (n = 22) and 0.97 (0.20) μmol/L in individuals who died from other causes.

Cox models indicated no significant association with cardiovascular diseases or with other causes of death. In contrast, bivariate Cox models showed a significant association between cancer-related mortality and low plasma selenium concentrations [increment = 0.2 μmol/L; RR = 1.61 (1.19–2.13); P = 0.001; Table 3³ ]. Analysis by quartile of plasma selenium showed that mortality risk increased significantly in individuals with plasma selenium in Q1 and Q2 compared with individuals with plasma selenium in Q4 [Q1 vs Q4: RR = 4.06 (1.51–10.92); P = 0.006; Q2 vs Q4: RR = 2.95 (1.06–8.18); P = 0.04]. Cancer-related mortality did not differ significantly between individuals who had plasma selenium concentrations in Q3 compared with those in Q4 [Q3 vs Q4: RR = 1.92 (0.63–5.86); P = 0.25]. The multivariate Cox model (Table 3³ ) confirmed the results of the bivariate Cox model, with the exception of smoking status. The RR of cancer-related mortality associated with plasma selenium decreases in increments of 0.2 μmol/L, adjusting for all factors, was 1.79 (1.32–2.44); P = 0.0002. In the medical history interview at baseline, 62 individuals self-reported a lifetime cancer history. The baseline plasma selenium in these individuals was 1.07 (0.23) μmol/L [not significantly different from other individuals (P = 0.3)]. Among the 45 individuals who died of cancer during the follow-up, 5 self-reported a history of cancer at baseline. Analysis gave similar results for all-cause mortality or cancer mortality when the 62 individuals who self-reported a lifetime cancer history were excluded.

Discussion

To our knowledge, very few studies on plasma selenium and mortality in elderly populations have been published. The EVA study included volunteers with higher educational status, higher incomes, and greater cognitive function than the average elderly French population. None of the study participants displayed evidence of poor nutrition. Plasma selenium concentrations in the EVA study population were in the same range as those in most European populations (13) but lower than the suggested optimal plasma selenium concentration for glutathione peroxidase activity (1.25 μmol/L) or for cancer protection (1.50 μmol/L)(2)(3). Our data indicate that in elderly individuals living independently, low plasma selenium concentrations were associated with higher mortality, even when we controlled for various potential confounding factors. This result is noteworthy because only a small number of individuals had baseline selenium concentrations below the cutoff of 0.75 μmol/L, which has been defined by a group of European experts(27) as a value related to selenium subdeficiency; however, as stated previously, many individuals have plasma selenium concentrations below the cutoff considered as optimal(2)(3). These results are in agreement with the low percentages of selenium deficiencies reported as a possible explanation for longevity in the nonagenarian–centenarian study(28). On the other hand, Wei et al.(6) found no association between total death and baseline selenium status in a cohort with a mean serum concentration of 0.93 μmol/L in younger individuals (mean age, 57 years).

The relationships that we observed between mortality and confounding factors such as sex (male), smoking, alcohol consumption, medication, obesity, diabetes, and vascular disease, but not age and dyslipidemia, suggest that the relationship between plasma selenium and mortality is not an artifact. Moreover, significant associations between selenium and mortality, taking into account the potential mortality risk factors, suggest an effect of selenium per se on mortality risk.

When considering the underlying causes of death, we found an association between plasma selenium concentration and cancer mortality. Our findings for selenium and cancer should be viewed with some caution, given that only 45 cancer deaths occurred and that different cancers often have different risk factors; however, our observations are in agreement with previous studies (6)(7)(10)(11)(12), despite controversy and debate over the protective effect of selenium against cancer(8). In the Health Professionals Follow-up Study, a reduction in cancer risk was associated with higher toenail selenium concentrations(7). In a large case–control study on breast cancer risk, a statistically significant preventive effect of quite high concentrations of plasma selenium was found in 278 cases and 135 controls(10). In the Linxian study, Mark et al.(11) found a significant inverse association between selenium and esophageal and gastric cardia cancer mortality [RR = 0.90 (0.83–0.97) and 0.87 (0.79–0.96), respectively]. In the Nutritional Prevention of Cancer Trials(12), the administration of selenium (200 μg/day for 4.5 years) induced a significant reduction in total cancer mortality [RR = 0.50 (0.31–0.80)], incidence of total cancer [0.63 (0.47–0.85)], and incidence of lung, colorectal, and prostate cancers.

In our study, 22 deaths related to cardiovascular disease were identified, limiting the interpretation of the nonsignificant relationship between selenium and cardiovascular mortality. No association between cardiovascular disease and selenium status has been reported in the Health Professionals Follow-up Study (15), but other studies have demonstrated that low serum selenium concentrations were significantly associated with deaths from heart disease(6)(16).

Our results suggest that plasma selenium could be an indicator of longevity in a preaging, independently living population not specifically at risk for cancer and cardiovascular diseases. Survival curves well illustrate that the relationship between plasma selenium and mortality was not restricted to the first years of follow-up but remained pertinent during the entire 9-year period. However, the mechanism of this potential relationship is still under debate. The determination of selenoproteins such as selenoprotein P, glutathione peroxidase at circulating or tissular concentrations (29), or selenium metabolites such as methylselenol(3) could help to clarify this mechanism.

In conclusion, our preliminary results reinforce the importance of adequate selenium status for health maintenance in an aging population. Even if it is premature to portray selenium as a longevity indicator, our data support those of large interventional randomized trials, which suggest that this essential trace element may play a role in health maintenance during the aging process.

This EVA study was carried out under an agreement between Institut National de la Santé et de la Recherche Médicale and the Merck, Sharp and Dohme-Chibret Laboratories (West Point, PA) and was supported by the EISAI laboratory (France). N.T.A. was supported by a grant from the French Alzheimer’s Disease Association.