The Journal of Nutrition

Nutrition and Disease

Optimal Serum Selenium Concentrations Are

Associated with Lower Depressive Symptoms
and Negative Mood among Young Adults13
Tamlin S Conner,4* Aimee C Richardson,4 and Jody C Miller5
Departments of 4Psychology and 5Human Nutrition, University of Otago, Dunedin, New Zealand

Abstract
Background: There is evidence that low, and possibly high, selenium status is associated with depressed mood. More
evidence is needed to determine whether this pattern occurs in young adults with a wide range of serum concentrations of
selenium.
Objective: The aim of this study was to determine if serum selenium concentration is associated with depressive
symptoms and daily mood states in young adults.
Methods: A total of 978 young adults (aged 1725 y) completed the Center for Epidemiological StudiesDepression scale and
reported their negative and positive mood daily for 13 d using an Internet diary. Serum selenium concentration was determined
by inductively coupled plasma mass spectrometry. ANCOVA and regression models tested the linear and curvilinear
associations between decile of serum selenium concentration and mood outcomes, controlling for age, gender, ethnicity, BMI,
and weekly alcohol intake. Smoking and childhood socioeconomic status were further controlled in a subset of participants.
Results: The mean 6 SD serum selenium concentration was 82 6 18 mg/L and ranged from 49 to 450 mg/L. Participants
with the lowest serum selenium concentration (62 6 4 mg/L; decile 1) and, to a lesser extent, those with the highest serum
selenium concentration (110 6 38 mg/L; decile 10) had significantly greater adjusted depressive symptoms than did
participants with midrange serum selenium concentrations (82 6 1 to 85 6 1 mg/L; deciles 6 and 7). Depressive
symptomatology was lowest at a selenium concentration of ;85 mg/L. Patterns for negative mood were similar but more
U-shaped. Positive mood showed an inverse U-shaped association with selenium, but this pattern was less consistent
than depressive symptoms or negative mood.
Conclusions: In young adults, an optimal range of serum selenium between ;82 and 85 mg/L was associated with reduced risk
of depressive symptomatology. This range approximates the values at which glutathione peroxidase is maximal, suggesting that
future research should investigate antioxidant pathways linking selenium to mood. This trial was registered with the Australian
New Zealand Clinical Trials Registry as ACTRN12613000773730. J Nutr 2015;145:5965.

Keywords:

selenium, micronutrients, mood, depression, young adults, daily diary method

Introduction
There is growing evidence for the role of micronutrients in
mental health, although the links between multivitamin-mineral
supplements, specific micronutrients, and mood remain inconclusive (1, 2). In a recent review, 4 of 8 trials demonstrated
decreases in depressed mood after supplementation with

1
Supported by a Health Research Council Emerging Researcher First Grant
(12/709) and a University of Otago Research Grant to TS Conner.
2
Author disclosures: TS Conner, AC Richardson, and JC Miller, no conflicts of
interest.
3
Supplemental Tables 13 are available from the Online Supporting Material
link in the online posting of the article and from the same link in the online table of
contents at http://jn.nutrition.org.
* To whom correspondence should be addressed. E-mail: tconner@psy.otago.
ac.nz.

multivitamin-mineral supplements in nonclinical populations.

Research also linked low concentrations of specific micronutrients such as vitamin B-12 and folate to increased levels of
depressive symptomatology (3, 4), although not consistently (1,
2). In this report, we investigated whether low concentrations of
selenium may also be implicated in depressive symptomatology.
Selenium is an essential nutrient required for the optimal
functioning of several selenoproteins involved in antioxidant
defenses within the brain and nervous system, including the
glutathione peroxidase antioxidant enzymes (GPx)6 (5). Several
studies found lower GPx activity among those with major
6
Abbreviations used: CES-D, Center for Epidemiological StudiesDepression;
GPx, glutathione peroxidase antioxidant enzyme; RCT, randomized controlled
trial; ROS, reactive oxygen species; SES, socioeconomic status.

2015 American Society for Nutrition.

Manuscript received June 3, 2014. Initial review completed June 20, 2014. Revision accepted October 15, 2014.
First published online November 5, 2014; doi:10.3945/jn.114.198010.

59

depression vs. controls (6), suggesting that lower selenium could

be a risk factor for depression via antioxidant pathways.
Observational research in adult and elderly populations has
shown a higher risk of depression among those with lower
selenium intake (7), less selenium groundwater exposure (8),
and lower selenium status as determined by nail samples (9).
Few studies have tested the association between selenium and
depressive symptoms in young adult populations (aged 1825 y).
The one study with the youngest age range of adults aged 20
35 y found that higher, not lower, selenium status was associated
with higher risk of depressive symptoms 3 y later (10). Randomized controlled trials (RCTs) have found improvements in
mood (11, 12) and improvement in postpartum depression (13)
with selenium supplementation in adult populations. However,
the largest RCT to date found no effect of selenium supplementation on mood in an elderly population (14). Older age could
limit the antioxidant benefits of selenium supplementation
through years of exposure to free radicals, which could explain
why supplementation shows stronger benefits in adult rather
than in elderly populations. Nevertheless, it is still unclear how
selenium status and depressive symptoms are associated in
young adults. Current observational studies focus on older ages
and are limited by the methods to determine selenium intake and
concentrations. For example, measures of selenium intake may
not be reliable because of large variations in the selenium
content of foods, which may not be accurately reflected in food
composition tables (15), and nail samples tested for selenium
reflect long-term selenium exposure; yet, studies using this
approach typically measure mood symptoms in the past week (9,
10). Past-week mood symptoms may be more closely aligned
with an acute measure of selenium, such as serum selenium,
rather than a long-term measure of selenium exposure. Moreover, mood symptoms recalled over the past week can reflect
memory biases; therefore, it would be useful to determine
whether selenium status predicts mood measured accurately by
using real-time data-capture techniques (16, 17). Our aim was to
determine whether serum selenium concentration is associated
with depressive symptoms and daily mood states in a large
nonclinical sample of young adults. We predicted that young
adults with lower selenium concentrations would report greater
depressive symptoms and negative mood. We also tested
whether higher selenium concentrations would also be associated with greater depressive symptoms and negative mood,
given recent evidence (10).

Methods
Participants and inclusion/exclusion criteria. Participants were
recruited as part of the Daily Life Study, a large cross-sectional study
of the biological and genetic markers of well-being in a population of
young adults living in Dunedin, New Zealand. Participants were
recruited through flyers, an on-campus employment agency, or human
nutrition classes and reimbursed with a small payment, or through the
University of Otago Psychology Departments experimental participation program and reimbursed with course credit. The data were collected
in the fall months of March to May in 2011, 2012, and 2013 and in the
winter months of July and August 2013.
Participants were eligible to participate in the Daily Life Study if
they were at least part-time students at the University of Otago, were
aged #25 y, and had regular access to the Internet (n = 1061). We
excluded from analysis any study participants who reported current
antidepressant medication use (n = 36) or who failed to provide
sufficient data (no blood sample, n = 17; <1 wk of daily diaries, n = 30).
No other inclusion or exclusion criteria were applied due to the broad
focus of the Daily Life Study. Participants taking selenium supplements
60

Conner et al.

or other multivitamin supplements were not excluded from the study.

The final sample consisted of 978 young adults (36.5% male) who
identified as European (80%), Asian (10.1%), Maori or Pacific Islander
(3.6%), or another ethnicity (6.3%). The study was approved by the
University of Otago Human Ethics Committee (no. 10/177), and all
participants provided written informed consent. This observational
study was registered with the Australian New Zealand Clinical Trials
Registry (ACTRN12613000773730).
Procedure. In an initial clinic visit, participants completed informed
consent and information on demographic characteristics [gender,
ethnicity, socioeconomic status (SES)], and depressive symptoms were
collected with a self-administered computerized questionnaire. Participants were informed that the study aim was to understand how
genes, hormones, and other biomarkers were related to daily health
and well-being. Before enrolling in the study, participants were told
that they would be answering questions each day about common
experiences in the lives of young adults, such as their health behaviors
and emotional states. They also received training for the daily diary
procedure that began the following day. For the next 13 d, participants completed an online daily survey between 1500 and 2000 h
every day. This daily survey included measures of negative and
positive mood and alcohol consumption, among a large range of other
health and well-being variables (daily exercise, sleep, stress, etc.). On
the final day of the study, participants attended the Department of
Human Nutrition, University of Otago, clinic where a nonfasting
blood sample was drawn into additive-free Vacutainers (Becton
Dickinson) for serum. Samples were refrigerated and centrifuged
within 2.5 h of collection. Aliquots of serum were stored in cryovials
at 280C until analysis. Height and weight measures were taken in
duplicate by using standardized procedures. On this day, participants
also completed a final questionnaire to measure their current use of
antidepressant medication and multivitamins. Participants recruited
during the 2013 phase of the study were also asked about current
smoking status.
Depressive symptoms and mood. Depressive symptoms were measured by using the Center for Epidemiological StudiesDepression (CESD) scale (18). This scale is designed for use in the general population and
contains 20 items that are self-rated on a scale of 0 (Rarely or none of
the time, <1 d) to 3 (Most or all of the time, 57 d), such as I thought

TABLE 1 Participant characteristics and descriptive statistics

for the young adult sample1
Value
n
Age, y
Gender, % men
Ethnicity, % European
BMI, kg/m2
Alcohol intake, standard drinks/wk
Current smoker,2 %
Childhood SES2,3
Multivitamin use, % currently taking
Serum selenium, mg/L
Depressive symptoms, CES-D score
Negative mood score4
Positive mood score4

978
19.6 6 1.6
36.5
80.0
23.8 6 4.1
8.9 6 10.8
7.1
4.4 6 1.9
21.5
82 6 18
13.7 6 8.5
1.68 6 0.47
3.00 6 0.49

1
Values are means 6 SDs unless otherwise indicated. CES-D, Center for Epidemiological StudiesDepression; SES, socioeconomic status.
2
Smoking and SES data available for only a subset of participants: n = 379 for smoking
and n = 352 for SES.
3
SES rating of 4.4 corresponded to an average childhood household income of
$72,500 in New Zealand dollars.
4
Average mood scores ranged from 1 (not at all) to 5 (extremely). A midpoint score of
3 indicated moderately intense negative or positive mood.

Values are means 6 SDs unless otherwise indicated. *P , 0.05, **P , 0.01. B, unstandardized regression coefficient.
Only significant patterns are reported. Blinear and Bcurve reflect the linear and curvilinear patterns in the descriptive characteristics across selenium decile groups. The coefficients were obtained by entering selenium decile (110) and selenium
decile squared (1100) simultaneously to predict each characteristic. Nonsignificant curvilinear coefficients were dropped from the final models. Categorical variables were tested by using chi-square statistics.
3
Gender x2 (9, 978) = 31.11, P , 0.001; Ethnicity x2 (9, 978) = 93.5, P , 0.001.
4
Number of standard drinks of alcohol consumed per week.
5
Smoking and socioeconomic status data were available for only a subset of participants, which yielded 2842 participants per decile.
2

0.018** (0.005, 0.032)

3
3

20.037** (20.064, 20.010)

20.122 (20.274, 0.030)

3
3
20.105* (20.195, 20.015)

0.386* (0.089, 0.684)

85 6 1
84 to ,87
19.8 6 1.7
36.1
68.0
24.0 6 3.7
9.7 6 10.9
5.3
4.8 6 1.8
70 6 1
67 to ,72
19.5 6 1.5
41.4
92.9
23.2 6 3.8
8.0 6 11.0
9.5
4.0 6 1.8
Selenium, mg/L
Range
Age, y
Gender, % men
European, %
BMI, kg/m2
Alcohol intake4
Smoker,5 %
Childhood socioeconomic status4

62 6 4
49 to ,67
19.5 6 1.5
19.6
93.8
25.6 6 6.3
7.4 6 9.4
12.5
4.3 6 1.8

73 6 1
72 to ,75
19.4 6 1.6
32.0
94.8
23.6 6 3.1
8.1 6 8.0
5.6
4.5 6 1.7

77 6 1
75 to ,78
19.4 6 1.6
30.6
89.8
23.4 6 3.4
9.5 6 11.6
10.0
5.0 6 1.7

79 6 1
78 to ,81
19.5 6 1.5
35.7
82.7
24.2 6 5.0
9.0 6 10.8
2.6
5.2 6 1.8

82 6 1
81 to ,84
19.4 6 1.6
34.3
77.8
23.5 6 3.7
9.7 6 12.1
12.5
4.3 6 1.8

89 6 1
87 to ,91
19.7 6 1.5
37.8
74.5
23.7 6 3.9
9.7 6 10.8
5.1
4.3 6 2.0

94 6 2
9197
20.0 6 1.7
42.9
68.4
23.3 6 3.5
9.5 6 11.9
4.8
4.1 6 2.2

110 6 38
98450
20.2 6 1.5
54.6
56.7
23.5 6 3.5
8.3 6 10.6
2.9
4.1 6 1.9

Bcurve2 (95% CI)

Blinear2 (95% CI)
10 (n = 98)
9 (n = 98)
8 (n = 98)
7 (n = 97)
6 (n = 99)
5 (n = 98)

Decile of serum selenium concentration

4 (n = 98)
3 (n = 97)
2 (n = 99)
1 (n = 97)

TABLE 2 Characteristics of the young adult participants (n = 978) categorized by decile of serum selenium concentration1

my life had been a failure and I felt that I could not shake off the blues
even with the help of my family or friends. Items were answered in
reference to the past week, and after reverse scoring of 4 items responses
were summed to give a total score. In community samples, a cutoff total
score of 16 is usually used to indicate significant levels of depressive
symptomatology (18).
Daily mood was measured each day for 13 d by using an 18-item
scale (19). The scale contained 9 adjectives assessing negative mood
(irritable, hostile, angry, nervous, tense, anxious, dejected, sad, unhappy)
and 9 adjectives assessing positive mood (excited, energetic, enthusiastic,
pleasant, happy, cheerful, content, calm, relaxed). Mood adjectives were
based on the affective circumplex and captured a range of high-,
medium-, and low-activation states (19). Participants were required to
rate each adjective on the basis of how they felt today on a 5-point
Likert scale: 1 (not at all), 2 (slightly), 3 (moderately), 4 (very much), 5
(extremely). Scores across the 9 adjectives were averaged each day to
obtain measures of daily negative mood (a = 0.780) and positive mood
(a = 0.820), which were averaged across days for analysis (a reliabilities
for nested data were computed by using recommended guidelines from
Nezlek; 20).
Serum selenium status. Serum selenium concentrations were measured by using an Agilent 7500ce quadrupole inductively coupled plasma
mass spectrometer in the Trace Element Centre, Chemistry Department,
University of Otago. National Institute of Standards and Technology
traceable calibration standards (High Purity Standards) and serum
samples were diluted 1:20 with a diluent containing 0.7 nmol/L
ammonia, 0.01 mmol/L EDTA, 0.07% Triton X-100, and butan-1-ol
1.5% vol:vol. The instrument was tuned according to the manufacturers
recommendations for a hydrogen reaction gas mode. Accuracy was
assessed by using an external certified reference material (UTAK
Laboratories) with a mean certified selenium value of 108 mg/L. Our
analysis gave a mean 6 SD of 112 6 3 mg/L, with a CV of 2.4% (n = 31).
In addition, multiple aliquots of pooled serum were analyzed during
each batch to determine the interassay CV, which was 1.2% (n = 82) at
72 mg/L.
Covariates. Demographic covariates were age (1725 y), gender (0 =
male, 1 = female), and self-reported ethnicity (0 = European ethnicity, 1 =
other ethnicity). Health covariates were BMI computed from measured
height and weight and weekly alcohol intake computed from a selfreported alcohol intake question included in the daily diary [number of
standard drinks of alcohol consumed the previous night, where
1 standard drink = 10 g of pure ethanol in a 330-mL can or bottle of
normal strength (4%) beer/cider, 1 very small (100 mL) glass of wine, or
1 small shot (30 mL) of liquor/spirits, straight or in a mixed drink]. Selfreported physical activity and cold symptoms were measured in the daily
diary but not included as covariates because neither was related to
selenium. The month of data collection and multivitamin use were tested
as covariates but dropped from analyses because they did not change the
results. For a subset of participants who completed the study in 2013,
smoking status (n = 379) and childhood SES (n = 352) were additionally
measured. Participants were defined as nonsmokers if they smoked
cigarettes once per month or less and as smokers if they smoked at least
once per week (nonsmokers = 0, smokers = 1). Childhood SES was
measured by self-reported household income while growing up (in New
Zealand dollars): 0 = <$15,000; 1 = $15,001$25,000; 2 = $25,001
$35,000; 3 = $35,001$50,000; 4 = $50,001$75,000; 5 = $75,001
$100,000; 6 = $100,001$150,000; and 7 = >$150,000.
Data analysis. Before analysis, serum selenium values were categorized
into deciles ranging from 1 to 10, from the lowest to highest 10% of
values. The use of deciles mitigated the influence of outliers and afforded
greater sensitivity than quintiles for detecting curvilinear patterns. An
ANCOVA general linear model was constructed to test the association
between serum selenium decile as the classification variable (predictor)
and depressive symptoms as the dependent variable, adjusting for age,
gender, ethnicity, BMI, and mean weekly alcohol intake as covariates.
This analysis yielded the adjusted least-squares mean CES-D (or mood)
scores for participants in each serum selenium decile. Assumptions of
Selenium and depressive symptoms in young adults

61

Conner et al.

Values are least-squares means (95% CIs) adjusted for age, gender, ethnicity, BMI, and alcohol intake unless otherwise indicated. ***P , 0.001. B, unstandardized regression coefficient.
Blinear and Bcurve reflect the linear and curvilinear effect of serum selenium decile on the predicted depressive symptoms or mood scores saved from ANCOVA. The coefficients were obtained by entering selenium decile (110) and selenium
decile squared (1100) simultaneously to predict the predicted depressive symptoms (Center for Epidemiological StudiesDepression) or mood scores. Least-squares means and predicted scores were nearly identical.
2

Depressive 15.9 (14.2, 17.6) 13.0 (11.3, 14.7) 15.1 (13.4, 16.8) 13.9 (12.3, 15.6) 13.3 (11.7, 15.0) 12.7 (11.0, 14.4) 12.5 (10.9, 14.2) 13.1 (11.4, 14.7) 13.0 (11.3, 14.7) 14.3 (12.6, 16.0) 21.15*** (21.29, 21.01)
0.087*** (0.075, 0.100)
symptoms
Negative
1.74 (1.64, 1.84) 1.67 (1.57, 1.76) 1.73 (1.64, 1.83) 1.68 (1.59, 1.78) 1.65 (1.55, 1.74) 1.69 (1.60, 1.78) 1.56 (1.46, 1.65) 1.67 (1.57, 1.76) 1.71 (1.62, 1.84) 1.75 (1.62, 1.81) 20.052*** (20.057, 20.047) 0.005*** (0.004, 0.005)
mood
Positive
2.93 (2.83, 3.03) 3.02 (2.92, 3.12) 2.95 (2.85, 3.05) 2.99 (2.89, 3.09) 3.07 (2.97, 3.16) 3.02 (2.92, 3.12) 2.97 (2.88, 3.07) 3.02 (2.92, 3.11) 3.01 (2.92, 3.11) 3.00 (2.90, 3.09)
0.034*** (0.023, 0.044)
20.003*** (20.004, 20.002)
mood

Bcurve2 (95% CI)

Blinear2 (95% CI)
10 (n = 98)
9 (n = 98)
8 (n = 98)
7 (n = 97)
6 (n = 99)
5 (n = 98)
4 (n = 98)
3 (n = 97)
2 (n = 99)
1 (n = 97)

Decile of serum selenium concentration

TABLE 3 Mean ANCOVA-adjusted depressive symptoms, negative mood, and positive mood for the young adult participants (n = 978) categorized by decile of serum selenium
concentration1
62

homogeneity of covariance were analyzed for each covariate and found

to be met. We tested for curvilinear patterns by saving the predicted
CES-D (or mood) scores from ANCOVA, plotting them, and then
predicting them from the decile of serum selenium (deciles 110) and
the squared decile of serum selenium concentration (1100) as
simultaneous predictors. This process was repeated substituting
negative mood and positive mood as the outcome variables. Statistical
analyses were performed with SPSS (version 22.0). P values <0.05 were
considered significant.
Several supplementary analyses were conducted including ANCOVA
models with smoking and childhood SES added as additional covariates
for the subset of participants with available data (n = 348). We also
performed hierarchical linear regression to further test for curvilinear
patterns in the relation between selenium and the mood outcomes. In the
first step, age, gender, ethnicity, BMI, and mean weekly alcohol intake
were entered to predict CES-D (or negative mood or positive mood)
scores. In the second step, the decile of serum selenium (deciles 110)
and the squared decile of serum selenium (1100) were entered as
simultaneous predictors of CES-D scores, controlling for the
covariates. Values in the text are presented as means 6 SDs unless
otherwise noted.

Results
Table 1 presents the participant characteristics and descriptive
statistics for the measured variables. Serum selenium concentrations ranged from 49 to 450 mg/L with the majority of cases
(99%) ranging from 49 to 158 mg/L. Depressive symptoms
ranged from 0.0 to 50.0, with 33.8% (n = 331) of our sample
reporting CES-D scores of $16.0, indicating significant levels of
depressive symptoms. Depressive symptoms correlated with
higher negative mood (r = 0.46, P < 0.001) and lower positive
mood (r = 20.41, P < 0.001). Negative mood and positive mood
were inversely correlated (r = 20.32, P < 0.001). Table 2
presents serum selenium concentrations and participant characteristics across the selenium deciles. Participants in the higher
selenium deciles were more likely to be older, male, nonEuropean, and of lower BMI. Participants in the middle
selenium deciles reported higher childhood SES than did
participants in the lower or higher selenium deciles. Table 3
presents the adjusted least-squares means of depressive symptoms, negative mood, and positive mood for each serum
selenium decile. There were significant linear and curvilinear
patterns across the serum selenium deciles in all 3 outcomes (all
P < 0.001).
Figure 1 shows the relations between selenium decile and
depressive symptoms (panel A), negative mood (panel B), and
positive mood (panel C) for the full sample, adjusted for the
primary covariates. For depressive symptoms, the pattern was a
reverse J-shaped curve, with the lowest depressive symptoms
among participants in selenium deciles 6 and 7 (82 6 1 and 85 6
1 mg/L serum selenium, respectively). Below and above this
range, depressive symptoms increased, particularly at the lowest
decile (decile 1; 62 6 4 mg/L serum selenium). For negative
mood, the pattern was a U-shaped curve with the lowest
negative mood among participants in selenium decile 7 (85 6
1 mg/L) and the highest negative mood among participants in
deciles 1 and 10 (62 6 4 and 110 6 38 mg/L, respectively).
However, the means for negative mood were all <2.00,
suggesting that negative mood did not increase above the
mild range. For positive mood, the pattern was an inverted
U-shaped curve with the highest positive mood among participants in selenium decile 5 (79 6 1 mg/L). Below and above this
range, positive mood decreased. This association between serum
selenium decile and adjusted positive mood was not due simply
to negative mood. Results from another ANCOVA entering

These relations between serum selenium decile and depressive symptoms, negative mood, and positive mood continued to
be significant when including smoking status and childhood SES
as additional covariates in the restricted sample of participants
(Supplemental Table 1). These relations were also significant
when including quadratic terms for both age and SES to control
for their curvilinear relations with serum selenium decile.
Hierarchical multiple regression analyses controlling for age,
gender, ethnicity, BMI, and alcohol intake confirmed these
significant negative linear and curvilinear relations between
serum selenium decile and depressive symptoms (linear B:
21.071; 95% CI: 21.898, 20.024; P = 0.011; curvilinear B:
0.082; 95% CI: 0.008, 0.155; P = 0.029) and negative mood
(linear B: 20.052; 95% CI: 20.098, 20.005; P = 0.029;
curvilinear B: 0.005; 95% CI: 0.001, 0.009; P = 0.029) but not
positive mood (linear B: 0.029; 95% CI: 20.19, 0.076; P = 0.24;
curvilinear B: 20.002; 95% CI: 20.006, 0.002; P = 0.31)
(Supplemental Table 2). Similar regression results were found
when controlling for age and SES as additional covariates
(Supplemental Table 3).

Discussion

FIGURE 1 Predicted scores for depressive symptoms (A), negative

mood (B), and positive mood (C) across serum selenium deciles for
978 young adults adjusted by ANCOVA for age, gender, ethnicity,
BMI, and alcohol intake. Decile 1 represents the lowest 10% of serum
selenium scores. Decile 10 represents the highest 10% of serum
selenium scores. n in each decile ranged from 97 to 99. CES-D, Center
for Epidemiological StudiesDepression.

negative mood as an additional covariate showed the same

pattern of adjusted means in Table 3 and the same linear and
curvilinear patterns shown in Figure 1C.

The results of this study suggest that there is a nonlinear relation

between serum selenium concentration and depressive symptoms and negative mood among young adults. Depressive
symptomatology was lowest around selenium concentrations
of 82 to 85 mg/L. Below 82 mg/L, depressive symptoms began
to increase, culminating in the highest depressive symptoms for
participants in the lowest decile of serum selenium (approximate
serum selenium concentration of 62 mg/L). Higher serum
selenium also resulted in increased depressive symptomatology,
although these increases were not as sharp. This pattern suggests
that although both lower and higher concentrations of serum
selenium were associated with increases in depressive symptomatology, lower concentrations of serum selenium were more
detrimental in this sample. Previous research also found that
either low selenium exposure (79) or high selenium exposure
(10) was associated with increased levels of depressive
symptoms among adult populations. This is the first study to
demonstrate that both low and high selenium statuses were
associated with increased depressive symptomatology within the
same sample of young adults. This is also the first study to show
that selenium concentration is related to daily negative mood in
a similar way to depressive symptoms. Negative mood was also
lowest around a selenium concentration of 85 mg/L. Selenium
concentration was less reliably related to positive mood.
The finding that higher concentrations of seleniumparticularly
$110 mg/Lwere associated with increased depressive symptoms may help to explain why the RCT conducted by Rayman
et al. (14) found that selenium supplementation had no effect
on mood or quality of life. In that RCT, there were 3 different
dosage groups (100, 200, and 300 mg/d), with those in the
100-mg supplementation group having the lowest serum
selenium concentration by the end of the 6-mo supplementation period (149 mg/L). This selenium value would have been
associated with high depressive symptoms in our study.
However, Rayman et al. (14) did not report any adverse effects
on mood with selenium supplementation or with elevated
serum selenium status.
The present findings may have important implications for
people who eat a diet either low or very high in selenium and for
people who take selenium supplements. Areas in Eastern Europe
and some parts of China have lower selenium intakes than in
Selenium and depressive symptoms in young adults

63

other areas (22). Similarly, in New Zealand, the selenium

content of the soil is low and thus, historically, New Zealanders
have had a low selenium intake. Although selenium status has
improved since the 1990s, many New Zealanders still have
serum selenium concentrations that are below what is required
for maximum GPx activity, and selenium supplementation leads
to increases in GPx activity (15). Our research suggests that low
selenium intake may be associated with greater risk of depressive
symptomatology and negative mood even among healthy young
adults. However, too much selenium may also lead to increased
risk. Previous research documented the detrimental effects of
high selenium concentrations on physical health (23, 24) and
mental health (10). Observational research demonstrated associations between higher selenium status and increased risk of
type 2 diabetes (>130 mg/L; 23) and increased risk of cancer
(>150 mg/L; 24). A U-shaped curve was also reported by
Bleys et al. (24) in relation to selenium status and all-cause
mortality (>150 mg/L). Animal research also reported a similar
U-shaped curve in relation to type 2 diabetes risk (25) and heart
function (26). Seleniums role as an antioxidant may help to
explain these findings. At serum selenium concentrations of 70
90 mg/L, plasma GPx enzymes are functioning at maximum
activity levels (27). Below this level, the enzymes may not be
performing at their maximum and thus the optimal antioxidant
protection may not be achieved. This could result in increased
vulnerability to oxidative stress and may explain why levels of
depressive symptomatology were highest in the lowest decile
with serum selenium concentrations #62 mg/L. Moreover, in
high doses, selenium has been demonstrated to generate reactive
oxygen species (ROS) (28). Although the range at which
selenium begins to produce ROS in humans is not yet known,
this mechanism could suggest that, in higher doses, selenium
may influence depressive symptoms via the generation of ROS
leading to oxidative stress. Further research is needed to examine
this oxidative stress hypothesis, particularly through RCTs that
monitor concentrations of ROS alongside changes in selenium
status.
This study is limited by its observational design. Therefore, a
causal relation between selenium status and mental health
cannot be concluded. Although it is plausible that selenium
status outside the optimal range leads to poorer mental health, it
is also possible that those with poor mental health are less likely
to eat foods high in selenium, resulting in lower selenium status,
although this does not explain why higher selenium was related
to increased depressive symptoms and negative mood. It is also
possible that the relation between selenium and depressed mood
is mediated by declines in physical health rather than any direct
effect of selenium on depressive symptoms. Clinical conditions
associated with high concentrations of selenium such as type 2
diabetes and cardiovascular disease are not prevalent in this
population, although it is possible that subclinical differences in
physical health could account for the relation between selenium
and depressed mood. However, we found no relation between
selenium and self-reported health as measured by physical
activity and cold symptoms. In addition, the relations between
selenium and depressed mood remained significant when controlling for 2 of the more relevant health factors in this
population, BMI and alcohol use.
Despite being limited by its observational design, this is the
only cross-sectional study of the association between serum
selenium concentrations and depressive symptoms in young
adults to date and, to our knowledge, the only study to have
implemented a daily diary method to assess negative and
positive mood. Thus, we can have more confidence that
64

Conner et al.

selenium is associated with experiences of negative mood that

occur in daily life. This study is also the first to report that serum
selenium within an optimal range is associated with lower levels
of both depressive symptoms and negative mood. Below this
range, symptoms of depression and negative mood were found
to increase significantly. Above this range, symptoms also
increased, although to a lesser extent. This latter finding adds
to the evidence base that a high selenium intake may be
associated with adverse health effects. Although further research
is needed, care should be taken to avoid consuming a diet too
low or too high in selenium.
Acknowledgments
We thank Hadyn Youens for web programming; Maria Polak
and Dr. Lisa A Houghton for assistance with blood collection;
and Dr. Malcolm Reid and David Barr from the University of
Otago Chemistry Department for selenium processing. TSC,
ACR, and JCM designed the research; JCM coordinated the
selenium processing; TSC and ACR conducted the research and
analyzed data; TSC, ACR, and JCM wrote the manuscript; and
TSC had primary responsibility for final content. All authors
read and approved the final manuscript.

References
1.
2.

3.

4.

5.

6.

7.

8.

9.

10.

11.
12.

13.

Kaplan BJ, Crawford SG, Field CJ, Steven J, Simpson A. Vitamins,

minerals, and mood. Psychol Bull 2007;133:74760.
Rucklidge JJ, Kaplan BJ. Broad-spectrum micronutrient formulas for
the treatment of psychiatric symptoms: a systematic review. Expert Rev
Neurother 2013;13:4973.
Gilbody S, Lightfoot T, Sheldon T. Is low folate a risk factor for
depression? A meta-analysis and exploration of heterogeneity.
J Epidemiol Community Health 2007;61:6317.
Robinson DJ, OLuanaigh C, Tehee E, OConnell H, Hamilton F, Chin
A, Coen R, Molloy AM, Scott J, Cunningham CJ, et al. Associations
between holotranscobalamin, vitamin B12, homocysteine and depressive symptoms in community-dwelling elders. Int J Geriatr Psychiatry
2011;26:30713.
Rotruck JT, Pope AL, Ganther HE, Swanson AB, Hafeman DG,
Hoekstra WG. Selenium: biochemical role as a component of glutathione peroxidase. Science 1973;179:58890.
Maes M, Galecki P, Chang YS, Berk M. A review on the oxidative and
nitrosative stress (O&NS) pathways in major depression and their
possible contribution to the (neuro)degenerative processes in the illness.
Prog Neuropsychopharmacol Biol Psychiatry 2011;35:67692.
Pasco JA, Jacka FN, Williams LJ, Evans-Cleverdon M, Brennan SL,
Kotowicz MA, Nicholson GC, Ball MJ, Berk M. Dietary selenium and
major depression: a nested case-control study. Complement Ther Med
2012;20:11923.
Johnson LA, Phillips JA, Mauer C, Edwards M, Hobson Balldin V, Hall
JR, Barber R, Conger TL, Ho EJ, OBryant SE. The impact of GPX1 on
the association of groundwater selenium and depression: a Project
FRONTIER study. BMC Psychiatry 2013;13:714.
Gao S, Jin Y, Unverzagt FW, Liang C, Hall KS, Cao J, Ma F, Murrell JR,
Cheng Y, Li P, et al. Selenium level and depressive symptoms in a rural
elderly Chinese cohort. BMC Psychiatry 2012;12:729.
Colangelo LA, He K, Whooley MA, Daviglus ML, Morris S, Liu K.
Selenium exposure and depressive symptoms: the Coronary Artery Risk
Development in Young Adults Trace Element Study. Neurotoxicology
2014;41:16774.
Benton D, Cook R. The impact of selenium supplementation on mood.
Biol Psychiatry 1991;29:10928.
Scott BH. The effect of selenium supplementation on the mood of
chronic fatigue syndrome and healthy control subjects [thesis].
Christchurch (New Zealand): University of Canterbury; 1993
Mokhber N, Namjoo M, Tara F, Boskabadi H, Rayman MP, GhayourMobarhan M, Sahebkar A, Majdi MR, Tavallaie S, Azimi-Nezhad M,
et al. Effect of supplementation with selenium on postpartum depression: A randomised double-blind placebo-controlled trial. J Matern
Fetal Neonatal Med 2011;24:1048.

14. Rayman M, Thompson A, Warren-Perry M, Galassini R, Catterick J,

Hall E, Lawrence D, Bliss J. Impact of selenium on mood and quality of
life: a randomized, controlled trial. Biol Psychiatry 2006;59:14754.
15. Thomson CD. Selenium and iodine intakes and status in New Zealand
and Australia. Br J Nutr 2004;91:66172.
16. Schwarz N. Why researchers should think real-time: a cognitive
rationale. In: Mehl MR, Conner TS, editors. Handbook of research
methods for studying daily life. New York: Guilford Press; 2012.
p. 2242.
17. Conner TS, Barrett LF. Trends in ambulatory self-report: the role of
momentary experience in psychosomatic medicine. Psychosom Med
2012;74:32737.
18. Sawyer Radloff L. The CES-D scale: a self-report depression scale for
research in the general population. Appl Psychol Meas 1977;1:385401.
19. Barrett LF, Russell JA. The structure of current affect: controversies and
emerging consensus. Curr Dir Psychol Sci 1999;8:104.
20. Nezlek JB. Multilevel modelling analyses of diary-style data. In: Mehl
MR, Conner TS. Handbook of research methods for studying daily life.
Guilford Press; 2012. p. 35783.
21. Cook RD, Weisberg S. Residuals and influence in regression. New York:
Chapman & Hall; 1982.

22. Rayman MP. Selenium and human health. Lancet 2012;379:125668.

23. Bleys J, Navas-Acien A, Guallar E. Serum selenium and diabetes in US
adults. Diabetes Care 2007;30:82934.
24. Bleys J, Navas-Acien A, Guallar E. Serum selenium levels and all-cause,
cancer, and cardiovascular mortality among US adults. Arch Intern Med
2008;168:40410.
25. Labunskyy VM, Cheon Lee B, Handy DE, Loscalzo J, Hatfield DL,
Gladyshev VN. Both maximal expression of selenoproteins and
selenoprotein deficiency can promote development of type 2 diabeteslike phenotype in mice. Antioxid Redox Signal 2011;14:232736.
26. Metes-Kosik N, Luptak I, DiBello PM, Handy DE, Tang S-S, Zhi H,
Qin F, Jacobsen DW, Loscalzo J, Joseph J. Both selenium deficiency and
modest selenium supplementation lead to myocardial fibrosis in mice
via effects on redox-methylation balance. Mol Nutr Food Res 2012;
56:181224.
27. Duffield AJ, Thomson CD, Hill KE, Williams S. An estimation of
selenium requirements for New Zealanders. Am J Clin Nutr 1999;70:
896903.
28. Lee KH, Jeong D. Bimodal actions of selenium essential for antioxidant
and toxic pro-oxidant activities: the selenium paradox [review]. Mol
Med Rep 2012;5:299304.

Selenium and depressive symptoms in young adults

65