Effect of Thyroxine Therapy on Depressive Symptoms Among Women With Subclinical Hypothyroidism

Abstract

Objective:

To estimate the effect of antenatal treatment of subclinical hypothyroidism on maternal depressive symptoms.

Methods:

We conducted an ancillary study to a multicenter trial in singleton pregnant women diagnosed with subclinical hypothyroidism randomized to antenatal thyroxine therapy or placebo. Treatment was discontinued at the end of pregnancy. Women with overt thyroid disease, diabetes, autoimmune disease, and those diagnosed with depression were excluded. Participants were assessed for depressive symptoms using the Center for Epidemiological Studies-Depression (CES-D) scale prior to starting the study drug (between 11-20 weeks), between 32-38 weeks, and at one-year postpartum. The primary outcome was maternal depressive symptoms score as assessed using the CES-D scale. Secondary outcome was the percentage of women who scored ≥ 16 on the CES-D, as such a score is considered screen positive for depression.

Results:

Two hundred forty-five (36.2% of parent trial) women with subclinical hypothyroidism were allocated to thyroxine (n=124) or placebo (n=121). Median CES-D scores and the proportion of participants with positive scores were similar at baseline between the two groups. Treatment with thyroxine was not associated with differences in CES-D scores (10 [5-15] vs. 10 [5-17]; p=0.46) or in odds of screening positive in the third trimester compared with placebo, even after adjusting for baseline scores (24.3% vs. 30.1%, aOR 0.63, 95% CI 0.31-1.28, p=0.20). At one-year postpartum, the CES-D scrores were not different (6 [3-11] vs. 6 [3-12]; p=0.79), nor was the frequency of screen-positive CES-D scores in the treated compared with the placebo group (9.7% vs. 15.8%; p=0.19). Treatment with thyroxine during pregnancy was also not associated with differences in odds of screening positive at the postpartum visit compared with placebo even after adjustng for baseline scores. Sensitivity analysis including women who were diagnosed with depression by the pospartum visit did not change the results.

Conclusions:

This study did not achieve its planned sample size, thus our conclusions may be limited, but in this cohort of pregnant women with subclinical hypothyroidism, antenatal thyroxine replacement did not improve maternal depressive symptoms.

Précis:

Antenatal thyroxine replacement did not improve depressive symptoms in a cohort of pregnant women with subclinical hypothyroidism.

Introduction

Mood disorders are a leading cause of morbidity and disability and are responsible for billions of dollars of lost work productivity. ¹ Mood disorders affect women more commonly than men, and the highest risk period for women is during their reproductive years, with one out of seven women treated for depression between the year prior to pregnancy and the year after their pregnancy. ^1,2,3 However, depressive symptoms that do not meet DSM-V criteria, i.e. “subclinical”, are even more prevalent (20-38%). ^4-6 Although the vast majority of patients with depression do not have overt thyroid disease, subclinical hypothyroidism is found in 15 - 20% and it is the most common thyroid dysfunction in patients with mood disorders. ⁷ Subclinical hypothyroidism is the most common form of thyroid disorder in pregnancy with an estimated prevalence of 2-5%. ^8,9 It is diagnosed based on the presence of normal serum thyroid hormone levels in the face of elevated thyroid stimulating hormone (TSH). While it is well known that overt hypothyroidism can negatively impact mood, ^10-12 subclinical hypothyroidism has also been associated with depressive disorders, subtle somatic symptoms, cognitive deficits and mood changes. ^13-17 Moreover, treatment of subclinical hypothyroidism in women of child-bearing age has been shown to improve their emotional state based on scores from various mood disorder tests, ^16-19 and a positive correlation has been found between TSH reduction and improvement in mood scores. ¹⁸

The evidence for an association between subclinical hypothyroidism and mood state for pregnant women is limited, and studies on the effect of treatment of pregnant women on their mood symptoms are unavailable. Additionally, there are no data on progress in antenatal depressive mood throughout pregnancy and in the postpartum period in this population. Our objective was to estimate the effect of treatment of subclinical hypothyroidism during pregnancy on maternal depressive symptoms. We hypothesized that treatment of women with subclinical hypothyroidism during pregnancy lowers depressive mood symptoms scores antenatally, and this benefit is lost if treatment is discontinued after delivery.

Methods

We performed an ancillary study to the Eunice Kennedy Shriver National Institute of Child Health and Human Development Maternal-Fetal Medicine Units Network study of thyroxine therapy for subclinical hypothyroidism or hypothyroxinemia diagnosed during pregnancy (www.clinicaltrials.gov NCT00388297). ²⁰ The study consisted of two multi-center double-blinded placebo – controlled trials conducted in parallel. The ancillary study was limited to the trial in women with subclinical hypothyroidism. In the primary trial, women with singleton pregnancies who presented for prenatal care between 8 and 20 weeks of gestation were screened for thyroid disease and those diagnosed with either subclinical hypothyroidism were randomized to daily treatment with either thyroxine or placebo. The details of the study, which was conducted at 15 centers in the U.S. between October 2006 and October 2009, are described elsewhere. ²⁰ The primary outcome of the trial was the child’s IQ at 5 years of age. ²⁰

For this ancillary study we included women with singleton, non-anomalous pregnancy, who were diagnosed with subclinical hypothyroidism. Those with reported clinical diagnosis of depression, other psychiatric, overt thyroid, diabetes, or autoimmune diseases, and those receiving anti-depressant medications were excluded. The ancillary study was approved by the institutional review board at each participating site.

Patients who were enrolled in the subclinical hypothyroidism trial were approached for participation. Those who met the eligibility criteria, were interested in this ancillary study, and signed the informed consent were assessed for depressive symptoms using the CES-D on randomization, and prior to starting the study drug. Covariables, including sociodemographic and obstetric data, were collected as part of the original trial. Race and ethnicity were coded as reported by mothers. We also collected family depression history, occurrence of a life stressor (such as major illness, family member death, divorce/marital separation, natural or man-made disaster) event in the immediate postpartum period, and whether the patient was diagnosed with postpartum depression. Another administration of the CES-D was done in the third trimester between 34-38 weeks, and then at one year postpartum. CES-D scales were not scored until the completion of the primary trial.

We used the Center for Epidemiological Studies-Depression (CES-D)²¹ Scale to measure symptoms of depression in this study. The CES-D is a 20-item questionnaire that measures depressive and social interactions symptoms. It includes both positively and negatively worded items, scaled from 0 to 3. CES-D scores range from 0 to 60, with higher scores indicating greater symptoms of depression; and a score of 16 or greater suggesting the presence of (screen positive) depressive symptoms.²² Spanish speaking women took CES-D in their native language as it is validated in Spanish. In a general population of non-pregnant child bearing age women, the mean score of CES-D was 10.80 ± 8.7 (range 0 – 46),²³ and in a community-based sample of pregnant women 9.70 ± 8.7. ²⁴

Our primary outcome was the maternal depressive symptoms score as assessed in the third trimester using the CES-D scale expressed as a continuous variable. Secondary outcomes included significant depressive symptoms as defined by a CES-D score of ≥16 in the third trimester and 1 year postpartum. Such a score is considered screen positive for depression. The postpartum CES-D score was also a secondary outcome.

The required sample size was based on comparing mean CES-D scores in the third trimester for two independent groups (those who received levothyroxine vs. placebo), as well as comparing the prevalence of depressive mood symptoms i.e. the prevalence of elevated CES-D scores ≥ 16. In a study evaluating the association between depressive symptom CES-D scores in pregnancy and social status as well as birthweight, CES-D score in the third trimester was reported to be 12.2±9 and the rate of CES-D ≥ 16 as 30.6%.²⁴ The NICHD early child care research network reported a mean CES-D score in the postpartum period of 12.8 for women who reported to be sometimes depressed.²³ Moreover, the prevalence of depressive symptoms in the third trimester of pregnant women is reported to be between 30.6% and 51%.^4,6,24,25 These studies included women from the general population, therefore it is plausible that the prevalence of elevated scores in the placebo arm will be higher in this trial since it includes a group of pregnant women that are at higher risk of depressive symptomatology secondary to their thyroid disorder. For our sample size calculations, we assumed the mean CES-D score in the third trimester to be 12.2±9 as previously reported, and the prevalence of depressive symptoms to be at least 38% in women with subclinical hypothyroidism assigned to the placebo group. ^12-14,36 Based on these assumptions, a sample size of 150 patients in each arm provided 80% power to detect at least a 3 point difference in CES-D scores, with α=0.05. This sample size also had 80% power to detect at least a 40% reduction in the prevalence of depressive mood symptoms (CES-D) ≥16.

Demographic characteristics of women, and the CES-D scores and the proportion of patients with scores ≥ 16 at the three time points were compared between the two groups (those receiving Levothyroxine vs. placebo) using the Chi-square or Fisher exact test for categorical data and the Wilcoxon Rank Sum test for continuous data as appropriate. The change in proportion with CES-D ≥16 from the third trimester to postpartum was evaluated using the McNemar test. The association between Levothyroxine treatment and screen positive score (CES-D ≥16) was assessed in logistic regression models adjusting for baseline scores for the third trimester and postpartum visits. Women who were diagnosed with depression (n=15) at the one-year postpartum follow-up visit were excluded from the postpartum analysis. However, we also performed a sensitivity analysis at the postpartum visit comparing the rate of women who either had a screen positive CES-D or were diagnosed with depression. A two-sided p-value of less than 0.05 was considered to indicate statistical significance. Statistical analyses were performed using SAS statistical software (SAS Institute, Inc, Cary, NC).

Results

The primary trial started in October 2006 and the ancillary study began in July 2008. A total of 677 women were enrolled in the primary trial and 245 (36.2%) were enrolled in the ancillary study (82% of planned sample size). The baseline CES-D was administered prior to starting the study drug. Therefore, by the time that the ancillary study started, we could not enroll those who were already enrolled in the primary trial (n=372). Ancillary study participants were allocated to thyroxine (n=124) or placebo (n=121). (Figure 1)

Figure 1:

Study flowchart of participants. CES-D, Center for Epidemiological Studies-Depression.

The characteristics of women enrolled in this ancillary study and those not enrolled are summarized in Table 1, and found to be similar between the two cohorts except for higher baseline thyrotropin concentrations and higher rates of being Caucasian and preterm delivery among those enrolled in the ancillary study. Among ancillary study participants, maternal demographic, obstetrical characteristics and outcomes, rates of postpartum life stressor events, postpartum depression, and family history of depression, and baseline thyrotropin and free thyroxine levels were similar between those who received thyroxine treatment and those assigned to placebo. (Table 2)

Table 1.

Maternal and neonatal characteristics of patients participating in the ancillary study compared with those from the subclinical hypothyroidism stratum who did not.

	Ancillary study (n=245)	Not in ancillary study (n=432)	P
Maternal Age (years)	27.9±5.9	27.3±5.6	0.24
Race or ethnic group			0.03
Black	19 (7.8)	33 (7.6)
Hispanic	122 (49.8)	258 (59.7)
White	99 (40.4)	127 (29.4)
Other	5 (2.0)	14 (3.2)
Maternal prepregnancy BMI (kg/m2)	28.4±6.7	27.9±6.3	0.52
Nulliparous	88 (35.9)	170 (39.4)	0.38
Baseline thyrotropin (mU/liter)	4.7 (4.2, 5.7)	4.2 (3.5, 5.2)	<0.001
Baseline free thyroxine (μg/liter)	1.02 (0.95, 1.11)	1.01 (0.95, 1.09)	0.32
GA at randomization	16.6±3.1	16.7±2.9	0.74
GA at delivery	38.8±2.8	39.1±2.8	0.05
Preterm delivery < 37 weeks	33 (13.5)	35 (8.1)	0.03
Fetal demise	3 (1.2)	8 (1.9)	0.75
Birthweight	3308±564	3354±539	0.58
Parent study primary outcome *	96 (84, 104)	96 (85, 107)	0.62

^*The parent study primary outcome was death or IQ score at 5 years of age (or at 3 years of age if the 5-year examination was missing). Results are expressed as an age-standardized score, with an expected population mean of 100 and a standard deviation of 15. Death before 3 years of age was assigned a score of 0 and was included in the estimation of the median.

Data reported as mean±SD, median [IQR] or n (%).

Table 2.

Maternal and neonatal characteristics of patients participating in the ancillary study receiving levothyroxine vs. placebo

	Thyroxine (n=124)	Placebo (n=121)
Maternal Age (years)	27.8±5.6	28.0±6.2
Race or ethnic group
Black	10 (8.1)	9 (7.4)
Hispanic	66 (53.2)	56 (46.3)
White	45 (36.3)	54 (44.6)
Other	3 (2.4)	2 (1.7)
Maternal BMI (kg/m2)	28.0±6.1	28.9±7.3
Marital status (married)	104 (83.9)	101 (83.5)
Household status
Own house	45 (36.3)	43 (35.5)
Rent	58 (46.8)	66 (54.6)
Live with parent/other adult	21 (16.9)	12 (9.9)
Employment status
Full time	33 (26.6)	41 (33.9)
Part-time	18 (14.5)	10 (8.3)
None	73 (58.9)	70 (57.9)
Insurance status
Government-assisted insurance	61 (49.2)	57 (47.1)
Private insurance	49 (39.5)	49 (40.5)
Self-pay	14 (11.3)	15 (12.4)
Cigarette use	10 (8.1)	4 (3.3)
Alcoholic drink	10 (8.1)	9 (7.4)
Nulliparous	43 (34.7)	45 (37.2)
Life stressor event 6 weeks before pospartum visit	16 (12.9)	22 (18.3)
Postpartum depression	7 (6.4)	8 (7.3)
Family depression history	17 (15.5)	23 (21.1)
Baseline thyrotropin (mU/liter)	4.7 (4.3, 5.9)	4.8 (4.2, 5.6)
Baseline free thyroxine (μg/liter)	1.01 (0.95, 1.09)	1.03 (0.95, 1.12)
GA at randomization	16.8±3.1	16.4±3.0
GA at delivery	38.8±3.3	38.9±2.1
Preterm delivery < 37 weeks	18 (14.5)	15 (12.4)
Fetal demise	2 (1.6)	1 (0.8)
Birthweight	3282±550	3334±579
Parent study primary outcome *	97 (83, 104)	95 (84, 105)

^*The parent study primary outcome was death or IQ score at 5 years of age (or at 3 years of age if the 5-year examination was missing). Results are expressed as an age-standardized score, with an expected population mean of 100 and a standard deviation of 15. Death before 3 years of age was assigned a score of 0 and was included in the estimation of the median.

Data reported as mean±SD, median [IQR] or n (%).

Median CES-D scores and the proportion of participants with positive scores (CES-D ≥16) were similar at baseline between the two groups (Table 3). Treatment with levothyroxine was not associated with improvement in median CES-D scores in the third trimester (10 [5-15] vs. 10 [5-17], p=0.46) compared with those who received placebo. Similarly, treatment with levothyroxine did not reduce the odds of having a positive score in the third trimester compared with placebo (24.3% vs. 30.1%, OR 0.75 95% CI 0.41-1.37, p=0.35). Findings were not different even after adjusting for baseline CES-D scores (aOR 0.63, 95% CI 0.31-1.28, p=0.20).

Table 3.

Primary and secondary outcomes per treatment group.

Primary Outcome
	Thyroxine	Placebo	P- value
Baseline CES-D score (n=244)	10 [5, 16]	9 [4, 15]	0.31
Third trimester CES-D score (n=210)	10 [5, 15]	10 [5, 17]	0.46
Secondary Outcomes
One-year postpartum CES-D score (n=213)	6 [3, 11]	6 [3, 12]	0.79
Rate of CES-D screen positive
Baseline CES-D score (n=244)	32 (26.0%)	28 (23.1%)	0.60
Third trimester CES-D score (n=210)	26 (24.3%)	31 (30.1%)	0.34
One-year postpartum CES-D score (n=213)	10 (9.7%)	16 (15.8%)	0.19
Rate of CES-D screen positive or depression diagnosis at one–year postpartum (n=219)	17 (15.5)	24 (22.0)	0.21

The postpartum CES-D screen was done at 12±1 months postpartum; the screening time was not different between women who were previously assigned to levothyroxine or placebo. Thirty-eight (15.6%) women reported that they experienced a life stressor within 6 weeks of the study visit and 15 (6.8%) reported that they were diagnosed with depression since delivery with six patients in the levothyroxine and nine in the placebo groups, respectively. Excluding these 15 patients, the difference in the frequency of screen-positive CES-D scores in women who were assigned to levothyroxine during pregnancy compared with the placebo group was not statistically significant (9.7% vs. 15.8%, p=0.19). As mentioned, fifteen patients were diagnosed with depression by the postpartum follow up visit, with 6 who received treatment during pregnancy. Sensitivity analysis including those patients did not change the results with the rates of women who either had a screen positive CES-D or diagnosis of depression similar between those received levothyroxine (n=17; 15.5%) compared with those assigned to placebo (n=24; 22.0%) (P=0.21). Treatment with thyroxine during pregnancy was not assocuated with differences in odds of screening positive at the postpartum visit compared with placebo (OR 0.57, 95% CI 0.25 – 1.33; P=0.19) even after adjustng for baseline scores (aOR 0.41, 95% CI 0.16 – 1.06; P=0.07).

Overall, CES-D scores and the proportion of women who scored positive were lower in the postpartum period compared with antepartum. The CES-D scores dropped by a median of 2 points (IQR of change in score [−6, 2], p<0.0001), and the rate of positive CES-D dropped from 23.8% in the third trimester to 13.5% (CES-D screen positive; P=0.003) or 18.4% (CES-D screen positive or diagnosis depression; P=0.06) in the postpartum period. The drop in the CES-D score and rate of positive scores in the women in the levothyroxine group was not significant.

Discussion

We found that at least one quarter of pregnant women with subclinical hypothyroidism screened positive for depression and that antenatal thyroxine treatment was not associated with improved depressive symptoms during pregnancy. Currently, routine screening for thyroid disorders in pregnancy has not been widely endorsed as there is no evidence of benefit from such an approach. ^8,26 This study did not achieve its planned sample size, thus our conclusions may be limited, but we did not find a benefit from screening and treating subclinical hypothyroidism in pregnancy on women’s mood. However, this study does support screening for depression in pregnant women with subclinical hypothyroidism due to the high prevalence of depressive mood symptoms.

Subclinical hypothyroidism is a mild form of hypothyroidism, that is diagnosed based on laboratory values of normal serum thyroid hormone levels in the face of elevated TSH. While it is well known that overt hypothyroidism can negatively impact mood, ^10-12 the association between subclinical hypothyroidism and depressive disorders is less certain. In non-pregnant patients, subclinical hypothyroidism is associated with subtle somatic symptoms, cognitive deficits and mood changes compared with euthyroid patients. ^13,15-17 In a series of 43 patients who were predominately female (88%) with an average age of 40.9 years, Gulseren et al. ¹⁹ found that physical and mental scores on the Health-Related Quality of Life questionnaire were worse in the subclinical hypothyroidism group than in the control euthyroid group. In addition, treatment of subclinical hypothyroidism in women of child-bearing age has been shown to improve based on scores on various mood disorder tests. ^16-19 In a group of women with subclinical hypothyroidism (n = 36), Bono et al. ¹⁸ reported improvement in depression scores, as assessed by the Hamilton Depression scale, after thyroxine treatment. A positive correlation was found between TSH reduction (baseline 8.5 ± 5.7 to 3 ± 3.1 post treatment, p<0.0001) and improvement in mood scores (baseline 6.3 ± 3.4 to 5.3 ± 3.4 post treatment, p<0.05).

The evidence for an association between subclinical hypothyroidism and mood disorders for pregnant women is limited. However, antenatal total and free thyroxine levels in the lower range of normal in the third trimester of pregnancy are associated with a greater risk of developing postpartum depression,²⁷ and studies suggest a relation between thyroid function in late pregnancy and postpartum depressive symptoms.^27,28 In addition, there are no data on the progress of antenatal depressive mood throughout pregnancy and in the postpartum period, but it has been suggested that antenatal depressive mood worsens in the immediate postpartum period.²⁹ However, in our study, the CES-D scores and the proportion of women who screened positive were significantly lower in the postpartum period compared with antepartum. The CES-D scores dropped by a median of 2 points, and the rate of positive CES-D dropped from 23.8% in the third trimester to 13.5% in the postpartum period.

Several theories have been proposed linking thyroid hormones to mood symptoms through their influence on neurotransmitters implicated in depression including serotonin, norepinephrine and dopamine. These include the effect of thyroid hormone on postsynaptic beta-adrenergic receptor activity leading to a functional decrease in catecholamines with reduction of thyroid hormone.³⁰ Also, low thyroid hormone activity is associated with hypo-serotenergic activity centrally.³¹ Decreased levels of transthyretin (TTR), the carrier protein that transports T4 into the brain, in the CSF of depressed patients also have been reported.^32,33 With reduced CNS TTR, delivery of T4 into the brain might be affected despite normal peripheral thyroid levels. Finally, the activity of 5’ deiodinase type II might be disturbed in patients with depression, ^34,35 creating a state of relative deficiency of the metabolically active T3 levels, which may predispose to depression. ⁷ Other theories include disruption in cerebral blood flow and cerebral glucose use. ^7,36,37

The CES-D is a 20 item self-report questionnaire that was developed by the National Institute of Mental Health as a research instrument to measure mood symptoms in the general population.²¹ The items that comprise the scale assess general psychological impairment. They were selected from other measurement instruments such as the Zung Self-Rating Depression scale, the Beck Depression Inventory, the Raskin Three-Area Severity of Depression scale, and the Minnesota Multiphasic Personality Inventory Depression scale.²¹ The CES-D provides multiple advantages including being normed on a community sample rather than a psychiatric population, as well as including items regarding social interactions. Other frequently used measures, such as the Edinburgh postnatal depression scale (EPDS), include items that are worded such as they relate more to “depression”. While the latter measures may be appropriate to identify women at risk for major/minor depression; a measure that captures symptoms related to social interactions could be more sensitive to changes in mood symptoms in the current study. Similar to the other frequently used measures, the CES-D retains important features: 1) balance of positive/negative wording of items, 2) use of a 0 to 3 scale per question, and 3) ease and speed of administration (estimated at 5 min). ²¹ CES-D scores range from 0 to 60, with higher scores indicating greater symptoms. A strong correlation has been seen between scores on CES-D and other instruments such as EPDS with an r=0.8, p<0.01 but the CES-D appears to be a more sensitive measure of depressive symptoms especially antenatally.⁶

This was an ancillary study of a multicenter RCT of pregnant women with subclinical hypothyroidism disorders. The strengths of our study are that the maternal and neonatal data were collected prospectively by study personnel blinded to maternal treatment assignment, and outcomes were carefully ascertained. The CES-D scales at the three study visits were completed by the patient, but not scored until after completion of the primary trial. Eighty percent of women in the ancillary study completed all three visits, and 95% completed two visits.

Despite our relatively large study, the lack of a significant association between treatment of subclinical hypothyroidism and depressive symptoms may have been related to the smaller than planned sample size, and represent a type II error. Only 82% of planned sample size was achieved, primarily due to the two year gap between the ancillary study and the primary trial start dates. The final sample size of 245 patients had 74% power to detect the hypothesized difference in the primary outcome. Although some trends were seen, it is unlikely that statistical differences would have been detected even if the full sample size had been reached, given the observed results of the primary and secondary outcomes. In addition, patients included in the ancillary study had higher baseline thyrotropin concentrations and higher rates of being Caucasian and preterm delivery compared with those from the primary trial who did not enroll in the ancillary study. This may limit the generalizability of the results. Our study did not have the ability to elicidate the potential direct mechanisms underlying the potential for an association between subclinical hypothyroidism treatment and maternal depressive symptoms. Other study limitations included the time frame in which postnatal stressors were assessed, i.e., within 6 weeks of the postpartum study visit, so stressful life events during pregnancy or immediate postpartum period may have been missed. We also did not collect information on other important factors associated with depression such as social support, anxiety, and child health in the first year of life, which could have affected our outcomes. It is possible that women who have a complicated postpartum course or whose child have medical complications, may be more likely to have depressed mood or anxiety symptoms. Our study did include randomization of women through the primary study. We also excluded women who were on antidepressant medications in the postpartum since this may affect the CES-D scores, but there was no significant difference in percents between women in the placebo and levothyroxine groups. The overall findings of the study did not differ significantly if the analyses included women who were on anti-depressants. Most of the women did not have elevated depressive symptoms at baseline; therefore it is plausible that the effect of levothyroxine on mood scores, if there was an effect, was diluted. Future research may focus on women with elevated CES-D scores. Lastly, we did not evaluate the effect of levothyroxine treatment in pregnant women with subclinical hypothyroidism on DSM-V diagnosed depression, which may be an area for future research.

In conclusion, antenatal thyroxine treatment in a cohort of pregnant women with subclinical hypothyroidism does not improve depressive symptoms score. This study does not support screening and treatment of subclinical hypothyroidism in pregnant women to reduce depressive symptoms.