Psychoneuroendocrinology (2015) 58, 114—119

Available online at www.sciencedirect.com

ScienceDirect

journal homepage: www.elsevier.com/locate/psyneuen

SHORT COMMUNICATION

Relationship between thyroid-stimulating

hormone levels and risk of depression
among the general population with normal
free T4 levels
Eun Young Kim a,1, Se Hyun Kim b,1, Sang Jin Rhee a, Iksoo Huh c,
Kyooseob Ha d,e,f, Jayoun Kim g, Jae Seung Chang h,
Dae Hyun Yoon i, Taesung Park c,∗∗, Yong Min Ahn a,e,∗

a
Mental Health Clinic, National Cancer Center, Goyang, Republic of Korea
b
Department of Neuropsychiatry, Dongguk University Medical School, Dongguk University International
Hospital, Goyang, Republic of Korea
c
Department of Statistics, Seoul National University, Seoul, Republic of Korea
d
Department of Psychiatry, Seoul National University Bundang Hospital, Seongnam, Republic of Korea
e
Institute of Human Behavioral Medicine, Seoul National University College of Medicine, Seoul,
Republic of Korea
f
Seoul National Hospital, Seoul, Republic of Korea
g
Biomedical Research Institute, Seoul National University Bundang Hospital, Seongnam, Republic of Korea
h
Department of Psychiatry and Institute of Clinical Psychopharmacology, Dongguk University College of
Medicine, Goyang, Republic of Korea
i
Department of Psychiatry, Seoul National University Hospital Healthcare System Gangnam Center, Seoul,
Republic of Korea

Received 30 December 2014; received in revised form 21 April 2015; accepted 22 April 2015

KEYWORDS Summary
Thyroid-stimulating Objective: This study assessed the relationship between thyroid-stimulating hormone (TSH)
hormone; level and risk of depressive symptom in a population with no clinical or laboratory evidence of
Subclinical thyroid thyroid dysfunction.
function; Methods: This retrospective cohort study included 13,017 subjects (7913 males and 5104
Depression females), 17—84 years of age, who underwent health examinations at the hospital. Subjects

∗ Corresponding author at: Department of Psychiatry and Behavioral Science, Institute of Human Behavioral Medicine, Seoul National

University College of Medicine, 28 Yongon-Dong, Chongno-Gu, Seoul 110-744, Republic of Korea. Tel.: +82 2 2072 0710; fax: +82 2 744 7241.
∗∗ Corresponding author at: Department of Statistics, Seoul National University, gwanak-gu, Seoul 151-747, Republic of Korea.

Tel.: +82 2 880 8924; fax: +82 2 888 6693.

E-mail addresses: tspark@stats.snu.ac.kr (T. Park), aym@snu.ac.kr (Y.M. Ahn).
1 These authors contributed equally to this study.

http://dx.doi.org/10.1016/j.psyneuen.2015.04.016
0306-4530/© 2015 Elsevier Ltd. All rights reserved.
Suboptimal thyroid function and risk of depression among general population 115

had a Beck Depression Inventory (BDI) total score of ≤9 and fell within the normal range of free
T4 levels at baseline. The association between gender-specific serum TSH tertile at baseline and
the development of clinically significant depressive symptom (i.e., ≥19 BDI total score) on the
follow-up visit was evaluated using the Cox proportional hazards model, with adjustment for
demographic and life style factors.
Results: The risk of depressive symptom was increased among subjects with the highest tertile
TSH level (adjusted hazard ratio [HR], 2.236; 95% confidence interval [CI], 1.443—3.466; p < 0.001)
as compared with subjects with the lowest tertile in females, but not in males. Even among
patients with normal TSH levels, females in the lowest-normal TSH tertile had a higher risk of
depressive symptoms (adjusted HR, 2.279; 95% CI, 1.456—3.567; p < 0.001) than did those in the
highest tertile. The TSH level as a continuous variable significantly predicted the depressive
symptoms in females (adjusted HR, 1.402; 95% CI, 1.002—1.812; p = 0.027).
Conclusions: Our finding suggests that suboptimal thyroid function increases vulnerability to
the occurrence of depressive symptom and represents a modifiable risk factor for depression in
females.
© 2015 Elsevier Ltd. All rights reserved.

1. Introduction questions remain about whether subclinical thyroid dysfunc-

tion is a risk factor for depressive symptoms and, this being
There has been a long history of interest in the associ- the case, the question of which population is vulnerable to
ation between the hypothalamic—pituitary—thyroid (HPT) depression remains largely unanswered.
axis and mood disorders. Substantial literatures have doc- We conducted a large retrospective cohort study to assess
umented the neurobiological effects of the thyroid axis in the association between thyroid function and the devel-
mood modulation and the role of thyroid function in the opment of clinically significant depressive symptoms after
pathophysiology or treatment of mood disorders (Bauer and controlling for potential confounding factors in the general
Whybrow, 2001). The development of assays for thyroid- population who had no clinical or laboratory evidence of thy-
stimulating hormone (TSH) in the early 1970s allowed for the roid dysfunction (free T4 values within the normal range).
assessment of more subtle thyroid dysfunction (Cole et al., We present our results separately according to gender.
2002), and many studies have documented that depres-
sion can be accompanied by subtle forms of hypothyroidism 2. Methods
(Haggerty et al., 1993; Joffe and Levitt, 1992). Frank thyroid
disorder is rare in depressive patients (Maes et al., 1993),
2.1. Participants
whereas subclinical hypothyroidism (defined as normal T4
with raised TSH) is highly prevalent in depressed populations
This retrospective cohort study included data from 13,017
(Gold et al., 1981). Patients with subclinical hypothyroidism
subjects, between 17 and 84 years of age, who under-
have a greater lifetime prevalence of depression (Haggerty
went health examinations at the Seoul National University
et al., 1993), and subtle thyroid dysfunction has been
Hospital Healthcare System, Gangnam Center, Seoul, South
suggested as a negative prognostic factor for depression
Korea, between October 2004 and July 2012. All subjects
(Fountoulakis et al., 2006).
received anthropometric measurements, completed self-
Although suboptimal thyroid function is considered a
administered questionnaires including the Beck Depression
reversible cause of depression, most of prior researches
Inventory (BDI) and underwent physical examinations and
were cross-sectional and few prospective studies have
laboratory tests including thyroid function tests. Our study
investigated the association between the development of
included 7913 males and 5104 females, who completed the
depression and suboptimal thyroid function using large sam-
BDI on ≥ two occasions during the observation period and
ples. A recent study showed that subclinical thyroid disease
had a BDI total score <10 on their initial visit. We excluded
was not associated with incident depression in 3932 elderly
subjects who had abnormal free T4 levels (normal reference
males free of overt thyroid disease (Almeida et al., 2011).
range, 0.7 ng/dl ≤ free T4 ≤ 1.8 ng/dl) or diagnosis of other
A study of 2269 middle-aged males found a weak positive
thyroid disease, or who were currently taking thyroid medi-
effect of total T4 on incident and chronic psychiatric morbid-
cation. This study was approved by the Institutional Review
ity; however, this was consistent with chance after adjusting
Board of the Seoul National University Hospital.
for alcohol, smoking and social status (Williams et al., 2009).
Gender-based relationships between thyroid function and
depressive symptoms may constitute an important factor, 2.2. Measures
due to the difference in serum TSH levels according to gen-
der in the general population and in depressed patients The 21-item BDI, a validated instrument for measuring the
(Forman-Hoffman and Philibert, 2006), as well as differ- severity of depressive symptoms, was used to determine
ences in the prevalence of thyroid abnormalities (Flynn the development of clinically significant depressive symp-
et al., 2004). To date, no large study has thoroughly investi- toms. Each BDI question was scored from 0—3 with a total
gated this issue with a focus on adult females. Fundamental score ranging from 0 to 63, with higher scores indicating
116 E.Y. Kim et al.

greater depressive symptom severity. Scores of 0—9, 10—18, 3. Results

19—29, or over 30 indicated the severity of symptoms as min-
imal, mild, moderate, or severe depression, respectively. 3.1. Subject characteristics
This study included subjects with a total score of 0—9 on BDI
to ensure absent or minimal depressive symptoms at base-
Table 1 summarizes demographic and clinical characteristics
line, and a BDI cutoff score of 19 indicated the development
of study participants in groups according to the gender-
of clinically significant depressive symptoms as a primary
specific tertiles of TSH levels at baseline (males: 0.05—1.34,
outcome measure.
1.35—1.89, and 1.90—31.4 mU/L; females: 0.05—1.44,
Demographic characteristics (age, gender, and mari-
1.45—2.30, and 2.31—22.0 mU/L). The mean (±SD) age of
tal status), educational level, lifestyle factors, and past
the study cohort was 46.9 ± 9.9 years at baseline. Subjects
medical history were collected from self-administered ques-
in the highest tertile group were older than subjects in the
tionnaires on the initial visit. Medical history was assessed
other groups. Subjects with higher TSH levels at baseline
for the presence or absence of hypertension, diabetes melli-
were less likely to be current smokers. Participants in all
tus, stroke, cardiovascular disease, and any type of cancer.
groups had similar rates of somatic illness. Pearson corre-
lation coefficient between TSH levels and free T4 levels
was significant (r = −0.142, p < 0.001 in males and r = −0.157,
p < 0.001 in females). The mean (± SD) follow-up period was
2.3. Thyroid function
3.2 (±1.5) years, ranging from 0.5 to 7.6 years.
Over the period covered by this study, thyroid function assay
methodologies changed several times at Seoul National Uni- 3.2. Risk of clinically significant depressive
versity Hospital. Serum TSH levels were measured using symptom and thyroid stimulating hormone levels
a commercially available kit (Daiichi Radioisotope Labs,
Tokyo, Japan), beginning in August 1998, and the kit was Fig. 1 shows the results from Cox proportional hazards
replaced by the Liaison TSH kit (Diasorin, Saluggia, Italy) in regression analyses. Clinically significant depressive symp-
February 2007. Free T4 levels were measured using a dif- toms developed in 40 male and 144 female subjects. There
ferent commercial kit (RIA-gnostFT4; CIS Bio International, was no association between risk of developing depressive
Cedex, France) beginning in March 2003. The RIA kit was symptoms and TSH tertile in male subjects. Conversely, in
eventually replaced by the Riakey TSH IRMA Kit and Riakey female subjects, when compared with subjects in the low-
free T4 RIA Kit (Shin Jin Medics, Seoul, Korea) after April est tertile group, the risk of depressive symptoms increased
2012. Regardless of the assay kit used, the reference range among subjects in the 2nd tertile group (adjusted hazard
in Seoul National University Hospital was consistent during ratio (HR), 1.795; 95% CI, 1.138—2.832; p = 0.012) and among
the study period (0.4—4.1 mU/L for TSH, and 0.7—1.8 ng/dl subjects in the highest tertile group (adjusted HR, 2.236;
(8.9—23.2 pmol/l) for free T4)). 95% CI, 1.443—3.466; p < 0.001). The TSH level as a continu-
ous variable also significantly predicted the development of
depressive symptoms only in females (adjusted HR, 1.402;
95% CI, 1.002—1.812; p = 0.027). Even among patients with
2.4. Statistical analyses normal TSH levels, females in the lowest-normal TSH ter-
tile had a higher risk of depressive symptoms (adjusted HR,
We categorized patients into three groups based on the 2.279; 95% CI, 1.456—3.567; p < 0.001) than did those in the
gender-specific tertiles of TSH levels. Baseline character- highest tertile, compared with persons in the highest ter-
istics between the groups were compared using an analysis
of variance for continuous variables and chi-square tests for
categorical variables. We applied Cox proportional hazards
regression model to estimate the hazard of clinically sig-
nificant depressive symptom according to gender-specific
TSH tertiles as well as TSH levels as a continuous vari-
able. In each model, the following possible confounding
factors were used: age, BMI, marital status, educational
level, smoking status, alcohol consumption, exercise fre-
quency, and somatic illness including hypertension, diabetes
mellitus, cardiovascular disease, and stroke. We also per-
formed similar analyses after additional adjustment for
menopause status and hormonal treatment only in female
subjects to determine whether these factors would influ-
ence the associations. Finally, we performed the same
analysis using samples with normal TSH levels. All p-values Fig. 1 Cox proportional hazards regression analysis for the
were two-tailed, and a p-value of 0.05 was considered statis- development of clinically significant depressive symptoms in
tically significant. Statistical analyses were performed using groups according to the gender-specific tertiles of TSH lev-
the Statistical Package for Social Science version 19.0 for els. a .a Ranges for TSH tertiles were as follows (mU/L):
Windows (SPSS, Inc., Chicago, IL, USA) and R Project for men: 0.05—1.34, 1.35—1.89, 1.90—31.4; women: 0.05—1.44,
Statistical Computing software version 2.11.0. 1.45—2.30, 2.31—22.0
Suboptimal thyroid function and risk of depression among general population 117

Table 1 Baseline demographic and clinical characteristics of study participants in groups according to the gender-specific
tertile of TSH levels.a

1st tertile (N = 4372) 2nd tertile (N = 4332) 3rd tertile (N = 4313)

Mean (SD) Mean (SD) Mean (SD)

Age (years)*** 46.2 (9.9) 46.9 (9.7) 47.7 (10.1)

BMI (kg/m2 ) 23.4 (3.0) 23.5 (3.0) 23.5 (3.0)
N (%) N (%) N (%)
Female 1711 (39.1) 1702 (39.3) 1691 (39. 2)
Currently married 3933 (90.0) 3949 (91.2) 3933 (91.2)
Education
<High school graduation 118 (2.7) 73 (1.7) 87 (2.0)
<College graduation 512 (11.7) 501 (11.6) 499 (11.6)
College graduation 2326 (53.2) 2365 (54.6) 2336 (54.2)
Post-graduate studies 1416 (32.4) 1393 (32.2) 1391 (32.3)
Cigarette smoking***
Non-smoker 2110 (48.3) 2168 (50.0) 2215 (51.4)
Ex-smoker 1219 (27.9) 1256 (29.0) 1437 (33.3)
Current smoker 1043 (23.9) 908 (21.0) 661 (15.3)
Alcohol use
≤1/month 1644 (37.6) 1678 (38.7) 1699 (39.4)
2/month∼2/week 2097 (48.0) 2008 (46.4) 1994 (46.2)
3∼4/week 515 (11.8) 519 (12.0) 495 (11.5)
≥5/week 116 (2.7) 127 (2.9) 125 (2.9)
Exercise
<1/week 1415 (32.4) 1311 (30.3) 1322 (30.7)
1∼2/week 1394 (31.9) 1416 (32.7) 1341 (31.1)
3∼4/week 1117 (25.5) 1133 (26.2) 1154 (26.8)
≥5/week 446 (10.2) 472 (10.9) 496 (11.5)
Somatic illness
Hypertension 687 (15.7) 676 (15.6) 718 (16.6)
Diabetes mellitus 181 (4.1) 183 (4.2) 180 (4.2)
Stroke 10 (0.2) 12 (0.3) 14 (0.3)
Cardiovascular disease 67 (1.5) 71 (1.6) 65 (1.5)
Menopause 481 (28.1) 469 (27.6) 494 (29.2)
Hormonal treatmentb 74 (4.3) 70 (4.1) 78 (4.6)
a Ranges for TSH tertiles were as follows (mU/L): men: 0.05—1.34, 1.35—1.89, 1.90—31.4; women: 0.05—1.44, 1.45—2.30, 2.31—22.0.
b Hormonal treatment includes use of oral contraceptive.
*** p < 0.001.

tile. In females, this significant relationship remained after
more adjustment for menopause and hormonal treatment.
4. Discussion
This large retrospective cohort study suggests that subop-
timal HPT axis function, even within the normal range of
thyroid hormone, is a major neuroendocrine condition pre-
disposing females to develop depressive symptoms.
The adjusted hazard ratio increased to 2.236 in the high-
est as compared with the lowest tertile of TSH levels only
in females but not in men. Interestingly, in a cross-sectional
analysis of general population, lower TSH and higher T4 lev-
els were associated with current depression in young adults
and elderly people (Forman-Hoffman and Philibert, 2006;
Medici et al., 2014), although no association in adults aged
50 to 70 years has been reported. In several clinical studies,
hyperactivity of the HPT axis was observed in depressive
patients, and thyroxine levels decreased with treatment
(Maes et al., 1993). Meanwhile, low normal thyroid function
was associated with poorer treatment response and with
lower rate of remission in depressive patients (Abulseoud
et al., 2013; Cole et al., 2002) Collectively, previous findings
are consistent with the hypothesis that hyperactivity of the
HPT axis in a depressive state may represent a compensatory
homeostatic response of the central nervous system to the
depressive state (Whybrow and Prange, 1981) and that fail-
ure of such response, indicated by high TSH or low free T4
levels, is related to depressive recurrence as well as neg-
ative prognosis. Accordingly, our findings show that subtle
thyroid dysfunction, even within the normal range, indicates
a decreased capacity of compensatory thyroid response to
the development of depressive symptoms.
In our study, the relationship between suboptimal thy-
roid function and development of depressive symptoms was
118
significant only in females. Overt and subclinical hypothy-
roidism is more common in females (Flynn et al., 2004), and
the risk of developing overt hypothyroidism from subclini-
cal thyroid dysfunction is also higher in females than males
(Cooper and Biondi, 2012). Moreover, the thyroid reserve
in response to depression can be mobilized more effec-
tively in males than in females (Abulseoud et al., 2007).
In this regard, baseline suboptimal thyroid function may
increase vulnerability to the development of clinically signif-
icant depressive symptoms, especially in females. However,
this association was not significant in males. Almeida et al.
(2011) investigated the association between thyroid dys-
function and incident depression in 3,932 males, 69 to 87
years of age. They found that serum TSH and free T4 lev-
els did not affect the hazard of incident depression during
the follow-up period over 5.5 years. Williams et al. (2009)
found no significant association between total T4 levels and
minor psychiatric morbidity in 2269 middle-aged males over
12.3 years. Our results were generally consistent in showing
no association between thyroid function and risk of depres-
sion in males. On the other hand, a recent study reported
that elderly people in the lowest normal-range TSH tertile
had a higher risk of incident depressive syndromes with no
sex-specific effect (Medici et al., 2014). However, this is not
directly comparable with our study because of differences
in age ranges, the depressive symptom assessment tool, and
covariates.
Our findings implicate the clinical utility of TSH levels to
identify patients who are vulnerable to depressive disorders
in female. Diagnosis of subclinical hypothyroidism is cur-
rently based on laboratory tests of serum TSH determination
(defined as an elevation in serum TSH above the upper limit
of the reference range with normal free T4 levels) and is
inevitably arbitrary in nature (Col et al., 2004). No clear evi-
dence shows that this ‘endocrinological’ reference range of
serum TSH levels can be applied to determine the relation-
ship between thyroid function and depression and identify
a high-risk population for development of depression. Sim-
ilar to our study, a previous study reported that serum
TSH concentrations in the highest quartile of the normal
range was positively associated with recurrent depression
(Berlin et al., 1999), questioning the appropriateness of the
‘endocrinological’ reference range of thyroid function in
depressive patients.
Our study is limited by the retrospective cohort design
and the fact that past medical history was obtained through
self-reported questionnaire. This may result in underesti-
mation of residual confounding factors including history of
thyroidectomy, use of radioactive iodine, and any type of
thyroid disease or medications such as T4 supplementation
which could affect TSH levels. Second, lack of structured
diagnostic interviews for depression may limit the con-
clusions. Although we used a BDI cut-off point of 19 to
determine clinically significant depressive symptoms, shown
as valid criteria in previous studies, under- or overestima-
tion of depression cannot be entirely ruled out. Finally, our
results may not be generalizable to other ethnic groups.
Nevertheless, this is the first study to show that inci-
dent depression is associated with higher serum TSH levels
in females with normal range free T4 levels using a large
study sample. Our findings highlight the need to establish
practical standard values of thyroid index to identify the
E.Y. Kim et al.
risk of depression. A controlled prospective study with a
longer follow-up period is necessary to establish a clear ref-
erence range to detect the population at risk and to assess
the potential treatment benefits such as thyroid hormone
supplementation, while considering sex differences.
Role of the funding source
The funding source had no role in study design; in the col-
lection, analysis, and interpretation of data; in the writing
of the report; or in the decision to submit the paper for
publication.
Conflict of interest
Prof. Yong Min Ahn receives a research grant and served as
a lecturer for Janssen, Lilly, Lundbeck and Otsuka. All other
authors report no financial relationships with commercial
interests.
Acknowledgments
This work was supported by grant A121987 from the Korea
Healthcare Technology R & D Project, Ministry of Health &
Welfare, Republic of Korea and by the National Research
Foundation of Korea (NRF) grant funded by the Korea gov-
ernment (MSIP) (No. 2012R1A3A2026438).
References
Abulseoud, O., Sane, N., Cozzolino, A., et al., 2007. Free T4 index
and clinical outcome in patients with depression. J. Affect. Dis-
ord. 100, 271—277.
Abulseoud, O.A., Gitlin, M., Altshuler, L., et al., 2013. Baseline
thyroid indices and the subsequent response to citalopram treat-
ment, a pilot study. Brain Behav. 3, 89—94.
Almeida, O.P., Alfonso, H., Flicker, L., et al., 2011. Thyroid hor-
mones and depression: the health in men study. Am. J. Geriatr.
Psychiatry 19, 763—770.
Bauer, M., Whybrow, P.C., 2001. Thyroid hormone, neural tissue and
mood modulation. World J. Biol. Psychiatry 2, 59—69.
Berlin, I., Payan, C., Corruble, E., et al., 1999. Serum thyroid-
stimulating-hormone concentration as an index of severity of
major depression. Int. J. Neuropsychopharmacol. 2, 105—110.
Col, N.F., Surks, M.I., Daniels, G.H., 2004. Subclinical thyroid dis-
ease: clinical applications. JAMA 291, 239—243.
Cole, D.P., Thase, M.E., Mallinger, A.G., et al., 2002. Slower
treatment response in bipolar depression predicted by lower
pretreatment thyroid function. Am. J. Psychiatry 159, 116—121.
Cooper, D.S., Biondi, B., 2012. Subclinical thyroid disease. Lancet
379, 1142—1154.
Flynn, R.W., MacDonald, T.M., Morris, A.D., et al., 2004. The
thyroid epidemiology, audit, and research study: thyroid dys-
function in the general population. J. Clin. Endocrinol. Metab.
89, 3879—3884.
Forman-Hoffman, V., Philibert, R.A., 2006. Lower TSH and higher T4
levels are associated with current depressive syndrome in young
adults. Acta Psychiatr. Scand. 114, 132—139.
Fountoulakis, K.N., Kantartzis, S., Siamouli, M., et al., 2006. Periph-
eral thyroid dysfunction in depression. World J. Biol. Psychiatry
7, 131—137.
Suboptimal thyroid function and risk of depression among general population
Gold, M.S., Pottash, A.L., Extein, I., 1981. Hypothyroidism and
depression evidence from complete thyroid function evaluation.
JAMA 245, 1919—1922.
Haggerty Jr., J.J., Stern, R.A., Mason, G.A., et al., 1993. Subclinical
hypothyroidism: a modifiable risk factor for depression? Am. J.
Psychiatry 150, 508—510.
Joffe, R.T., Levitt, A.J., 1992. Major depression and subclini-
cal (grade 2) hypothyroidism. Psychoneuroendocrinology 17,
215—221.
Maes, M., Meltzer, H.Y., Cosyns, P., et al., 1993. An evalua-
tion of basal hypothalamic—pituitary—thyroid axis function in
depression: results of a large-scaled and controlled study. Psy-
choneuroendocrinology 18, 607—620.
119
Medici, M., Korevaar, T.I., Schalekamp-Timmermans, S., et al.,
2014. Maternal early-pregnancy thyroid function is associ-
ated with subsequent hypertensive disorders of pregnancy:
the generation R study. J. Clin. Endocrinol. Metab. 99,
2591—2598.
Whybrow, P.C., Prange Jr., A.J., 1981. A hypothesis of thyroid-
catecholamine-receptor interaction: its relevance to affective
illness. Arch. Gen. Psychiatry 38, 106—113.
Williams, M.D., Harris, R., Dayan, C.M., et al., 2009. Thyroid func-
tion and the natural history of depression: findings from the
Caerphilly Prospective Study (CaPS) and a meta-analysis. Clin.
Endocrinol. (Oxf) 70, 484—492.