Health Status, Mood, and Cognition in Experimentally Induced Subclinical Hypothyroidism

Abstract

Objective: The objective of the study was to determine whether subclinical hypothyroidism causes decrements in health status, mood, and/or cognitive function.

Design: This was a double-blinded, randomized, crossover study of usual dose l-thyroxine (L-T4) (euthyroid arm) vs. lower dose L-T4 (subclinical hypothyroid arm) in hypothyroid subjects.

Patients: Nineteen subjects on L-T4 therapy for primary hypothyroidism participated in the study.

Measurements: Subjects underwent measurements of health status, mood, and cognition using validated instruments: Short Form 36, Profile of Mood States, and tests of declarative memory (paragraph recall, complex figure), working memory (N-back, subject ordered pointing, digit span backward), and motor learning (pursuit rotor). The same measures were repeated after 12 wk on each of the study arms.

Results: Mean TSH levels increased to 17 mU/liter on the subclinical hypothyroid arm (P < 0.0001). Mean free T₄ and free T₃ levels remained within the normal range. The Profile of Mood States fatigue subscale and Short Form 36 general health subscale were slightly worse during the subclinical hypothyroid arm. Measures of working memory (N-back, subject ordered pointing) were worse during the subclinical hypothyroid arm. These differences did not depend on mood or health status but were related to changes in free T₄ or free T₃ levels. There were no decrements in declarative memory or motor learning.

Conclusions: We found mild decrements in health status and mood in L-T4-treated hypothyroid subjects when subclinical hypothyroidism was induced in a blinded, randomized fashion. More importantly, there were independent decrements in working memory, which suggests that subclinical hypothyroidism specifically impacts brain areas responsible for working memory.

ADULT-ONSET HYPOTHYROIDISM causes deficits in specific cognitive domains, including attention and concentration, memory, perceptual function, language, executive function, and psychomotor speed (1–7). Subclinical hypothyroidism (elevated TSH, normal free T₄) may cause milder alterations in cognitive function, which may improve with l-thyroxine (L-T4) therapy, but this literature is inconclusive (4, 6, 8–19). In 2004 an expert panel concluded that there were insufficient data on the association between subclinical hypothyroidism and cognition and recommended against treating patients based on neurocognitive effects (20).

Subclinical hypothyroidism may also affect mood. Some (but not all) studies have shown increased rates of anxiety or depression in subclinical hypothyroidism, with improvement after L-T4 therapy (6, 11–14, 17, 21–23). To address these issues, we designed a novel model of subclinical hypothyroidism. In a double-blinded, random-order, crossover study, L-T4-treated subjects received their usual dose of L-T4 or a lower dose of L-T4 to induce subclinical hypothyroidism. We measured health status, mood, and cognitive function at the end of each arm. We chose to measure memory, based on studies suggesting that memory may be preferentially affected in mild hypothyroidism or in L-T4-treated subjects (10, 12, 13, 18) and because animal studies support a major role for L-T4 in brain areas that subserve memory (24–28). We studied three distinct forms of memory linked to specific brain systems: declarative memory (medial temporal lobe), working memory (prefrontal cortex), and motor learning (cerebellum and basal ganglia) (29). We included validated measures of health status and mood because these may be impaired in subjects with subclinical hypothyroidism (6, 13, 23, 30) and because they can affect cognition. We hypothesized the following: 1) subjects with induced subclinical hypothyroidism have decrements in general health status and mood, compared with the euthyroid state; 2) subjects with induced subclinical hypothyroidism have decrements in memory, which may be related to changes in health status or mood; and 3) in subjects with induced subclinical hypothyroidism, changes in health status, mood, and memory are correlated with changes in serum TSH, free T₄, and/or free T₃ levels.

Subjects and Methods

Experimental subjects

Nineteen women, aged 20–75 yr, with adult-onset primary hypothyroidism were recruited. Thirteen subjects had autoimmune hypothyroidism and six had received I-131 therapy for Graves disease. Men were not intentionally excluded, but the few male volunteers did not pass the screening tests. All subjects had elevated TSH levels before treatment or while on previous lower doses of L-T4. Subjects had to be receiving stable L-T4 doses as the sole treatment for at least 3 months, with normal TSH levels. Subjects did not have any acute or chronic illnesses and were not receiving medications that affect thyroid hormone levels, mood, or cognition. Oral contraceptive or estrogen therapy was allowed, with a stable dose for at least 3 months, and with no changes during the study. Testing was done during the first 10 d after onset of menstrual bleeding or in the first 10 d of an oral contraceptive cycle.

Experimental design

The protocol was approved by the Oregon Health and Science University Institutional Review Board, and subjects gave written informed consent.

Screening visit

Subjects were screened for general health, medicines, drug abuse, thyroid status, and mood or cognitive disorders by history, physical examination, and laboratory testing. General intelligence was measured by the Wechsler Adult Intelligence Scale Revised (WAIS-R) vocabulary subtest, which correlates with IQ (31). Subjects were excluded if they scored less than 8 (scaled). The Symptom Checklist 90-R was administered to screen subjects for psychiatric diseases (32). Subjects who scored outside the normative range were excluded. Subjects were excluded if serum cholesterol or triglyceride levels were above normal or if electrocardiograms were abnormal.

Baseline visit

Within 3 wk of screening, subjects returned for a 90- to 120-min baseline visit. Subjects refrained from taking their L-T4 doses that morning. Serum TSH, free T₄, and free T₃ levels were obtained. The Billewicz scale, a validated questionnaire for hypothyroid symptoms and signs, was completed (33). Subjects completed validated measures of health status and mood. The first was the Short Form 36 health survey (SF-36), a questionnaire about general health and well-being (31). Higher scores on the SF-36 summary scales and subscales reflect better health status and well-being. This instrument has been used in studies involving thyroid disease (30). The second was the Profile of Mood States (POMS), a questionnaire about mood (31). Higher scores on the POMS subscales reflect mood decrements, except for the vigor subscale, in which higher scores reflect improved mood.

Cognitive tests were then administered by a single research assistant in a standard fashion.

Tests of declarative memory

Paragraph recall (verbal memory)

For this subtest of the Wechsler Memory Scale-Revised, subjects were read two brief stories and then verbally recalled each of them immediately and after a 30-min interval. The outcome measure was the total number of story elements recalled at each interval. This measure requires the hippocampus and medial temporal lobe (34).

Complex figure test (visual memory)

For the Medical College of Georgia Complex Figure Test, subjects copied a standard complex figure and then drew it from memory immediately and after a 30-min interval. The outcome measure was the total number of elements correctly reproduced. This measure requires the hippocampus and medial temporal lobe (31).

Tests of working memory

N-back test

A series of letters was presented one at a time on a screen. Subjects responded when a letter appeared that they had seen on the previous screen (1-back). The task was repeated with an increase in memory load by having the subject respond when a letter appeared 3-back. The outcome measure was the total number correct. This measure requires the prefrontal cortex (35).

Subject-ordered pointing (SOP)

Subjects were presented with stacks of cards (six, eight, 10, or 12 per set). Each card showed an array of abstract drawings, in a different spatial arrangement on each card. The subject was instructed to touch one drawing on each card but not to touch the same drawing on subsequent cards in the set. Subjects erred when they touched a drawing that had been touched on a previous card. Each card set was repeated three times. The outcome measure was the total number of errors across each card set (31).

Digit span backward

The examiner read number sequences of increasing length, and after each sequence the subject was asked to repeat the sequence backward. Subjects erred when they could not successfully repeat two sequences of the same length. This measures requires the prefrontal cortex (31).

Motor learning

Pursuit rotor

Subjects held a photosensitive wand to maintain contact with a 2-cm light disk rotating on a variable speed turntable (model 30014; Lafayette Instrument Co., Lafayette, IN). An initial block of four trials was administered at 15, 30, 45, and 60 revolutions/min to establish the optimal speed for further trials. Two blocks of eight 20-sec trials were then administered, with a 20-sec rest after each trial and a 60-sec rest period after four trials. After a 30-min interval, the two blocks were repeated. The outcome measure was the mean total time the stylus remained on target during each trial. This measure requires the basal ganglia and cerebellum (36).

Subjects were then randomized to receive either their usual doses of L-T4, (euthyroid arm) or lower doses of L-T4, calculated to lead to a serum TSH level of 10–20 mU/liter (subclinical hypothyroid arm) (Table 1). For the subclinical hypothyroid arm, we targeted TSH levels of 10–20 mU/liter because studies have showed more consistent clinical effects at these TSH levels (20). Based on limited data, estimated L-T4 dose reductions were 30% below euthyroid doses (37, 38). The subjects and the physician (M.H.S.) and research assistant (P.C.) with direct subject contact were blinded, whereas the monitoring physician (K.G.S.) was unblinded to adjust L-T4 doses and monitor for side effects. L-T4 pills were placed in gel capsules by the research pharmacy to maintain blinding.

TABLE 1.

Clinical parameters and thyroid function tests at the end of each arm of the study (euthyroid and subclinical hypothyroid)

Measure	Baseline (mean ± sem)	Euthyroid (mean ± sem)	Subclinical hypothyroid (mean ± sem)	P value, paired t test
TSH (mU/liter)	3.53 ± 1.12	2.19 ± 0.35	17.37 ± 3.04	<0.0001
Free T4 (ng/dl)	1.32 ± 0.065	1.43 ± 0.03	0.99 ± 0.04	<0.0001
Free T3 (pg/dl)	236.95 ± 11.47	259 ± 9.0	231 ± 12.5	0.06
Weight (kg)	77.37 ± 2.64	78.0 ± 3.0	78.4 ± 3.1	0.41
BMI (kg/m2)	28.13 ± 1.16	28.4 ± 1.3	28.5 ± 1.3	0.43
Pulse (beats/min)	75.53 ± 2.84	74.9 ± 3.0	74.2 ± 3.1	0.77
Systolic blood pressure (mm Hg)	127.89 ± 3.38	126.7 ± 5.1	126.1 ± 3.6	0.89
Diastolic blood pressure (mm Hg)	74.74 ± 2.65	76.6 ± 2.5	75.5 ± 2.9	0.75
Billewicz scale	2.95 ± 0.57	1.74 ± 0.33	2.58 ± 0.62	0.08

P values refer to differences between the end of the euthyroid arm and the subclinical hypothyroid arm of the study.

Six-week interim visit

Six weeks later, subjects returned for a brief visit. Compliance was assessed by pill counts. Safety was assessed by questioning for symptoms and measurement of serum cholesterol and triglyceride levels to ensure that they had not increased above the normal range. Serum TSH and free T₄ levels were measured. The monitoring physician (K.G.S.) made L-T4 dose adjustments as needed to attain target TSH levels for that arm of the study and keep free T₄ levels within the normal range.

Crossover visit

Twelve weeks after the baseline visit, subjects returned for an extended visit. Serum TSH, free T₄, and free T₃ levels were measured. The Billewicz scale, SF-36, POMS, and cognitive tests were repeated in the same order as the baseline visit. Subjects were then crossed over to the second arm of the study and had their new doses of L-T4 dispensed.

Eighteen-week interim visit

This visit was identical with the 6-wk visit, with subjects now on the alternate dose of L-T4.

End-of-study visit

Twelve weeks after the crossover visit, subjects returned for an extended visit identical with the crossover visit. When testing was completed, subjects were asked which arm of the study they thought was the lower dose arm and which arm they preferred. They were then placed back on their usual doses of L-T4.

Analytic methods

TSH was measured by immunochemiluminometric assay (Nichols Institute, San Juan Capistrano, CA), with a sensitivity of 0.003 mU/liter and a normal range of 0.28–5.00 mU/liter. Intraassay coefficient of variation (CV) was 9.5% at 0.03 mU/liter and 4.7% at 11.6 mU/liter. Interassay CV was 17% at 0.02 mU/liter and 4.6% at 14 mU/liter. Free T₄ was measured by immunochemiluminometric assay (Nichols Institute), with a sensitivity of 0.08 ng/dl and a normal range of 0.7–1.8 ng/dl. Intraassay CV was 5.7% at 0.27 ng/dl and 1% at 4.6 ng/dl. Interassay CV was 6.8% at 0.3 ng/dl and 1.6% at 3.8 ng/dl. Free T₃ levels were measured by tracer dialysis (Nichols Institute) with a sensitivity of 25 pg/dl and a normal range of 210–440 pg/dl. Intraassay CV was 6% and interassay CV was 4%.

Statistical methods

Differences in outcomes between the euthyroid and subclinical hypothyroid arms were first explored by t tests. Then to control for multiple comparisons, sets of psychological, mood, and cognitive measures were analyzed together using a mixed-effects model. Age, years of education, WAIS-R, and the baseline measures were covariates to reduce residual variability. Generalized estimating equations (39) were used to estimate group differences for measures with ceiling or floor effects (e.g. role physical, social functioning, and emotional subscales in the SF-36). The sets of measures analyzed by each modeling technique are grouped together in Tables 2 and 3. Bonferroni adjustments were used for formal pairwise comparisons when applicable. P < 0.05 was considered significant.

TABLE 2.

Summary statistics and P values for health status and mood measures

Measure	Baseline (mean ± sem)	Euthyroid (mean ± sem)	Subclinical hypothyroid (mean ± sem)	P value, paired t test	Adjusted P values
SF-36
Mental component summary	52.26 ± 1.51	51.65 ± 1.45	47.46 ± 2.36	0.24	Treatment: 0.35
Physical component summary	51.50 ± 1.46	52.42 ± 1.13	53.89 ± 0.74	0.27	Treatment: 0.69
					Treatment∗test: 0.09
Bodily pain	70.68 ± 4.24	67.67 ± 4.12	72.58 ± 4.86	0.20	0.22
General health	78.89 ± 3.04	81.39 ± 3.07	75.53 ± 3.40	0.13	0.08
Mental health	81.26 ± 2.47	78.89 ± 2.68	72.84 ± 4.11	0.34	0.23
Physical functioning	89.74 ± 2.00	91.11 ± 1.96	92.37 ± 1.96	0.40	0.41
Vitality	59.74 ± 3.74	61.47 ± 3.97	53.68 ± 5.35	0.10	0.16
					Treatment: 0.78
Role physical	84.21 ± 7.46	84.72 ± 4.58	90.79 ± 3.43	0.39
Social functioning	92.11 ± 2.57	91.67 ± 2.67	89.47 ± 3.48	0.77
Role emotional	87.72 ± 6.84	88.89 ± 6.60	80.70 ± 6.41	0.56
POMS					Treatment: 0.13
Anxiety	45.26 ± 1.65	45.47 ± 2.01	47.00 ± 2.68	0.65
Confusion	41.42 ± 1.52	42.47 ± 1.60	43.95 ± 2.30	0.51
Depression	41.95 ± 1.11	43.53 ± 1.43	43.89 ± 1.85	0.84
Fatigue	43.84 ± 1.55	44.11 ± 1.74	48.95 ± 2.26	0.03
Tension	41.21 ± 1.19	42.79 ± 1.48	42.95 ± 2.07	0.94
Vigor	51.58 ± 1.70	51.79 ± 1.79	52.63 ± 2.50	0.75	Treatment: 0.83

P values for unadjusted paired t tests comparing the two arms of the study are shown in the fifth column. Adjusted P values in the repeated-measures analysis are shown in the sixth column. These P values were adjusted for age, years of education, WAIS-R, and baseline measure for that particular variable. Each test was analyzed as a group to adjust for within-subject correlation and correlations between responses on similar tests. For the adjusted analyses, treatment∗test P values were examined initially. If the treatment∗test P value was greater than 0.1, it is not shown in the table, and only the overall treatment effect is shown. If the treatment∗test P value was 0.1 or less, this value is listed in the table, and subsequent analyses were done for individual measures. P values are bold if they are significant or marginally significant and are discussed further in Results (P < 0.10).

TABLE 3.

Cognitive measures during the two treatment arms of the study

Test	Baseline (mean ± sem)	Euthyroid (mean ± sem)	Subclinical hypothyroid (mean ± sem)	P value, paired t test	Adjusted P values
Declarative memory
Paragraph recall					Treatment: 0.02
Immediate	13.32 ± 1.07	15.84 ± 0.92	17.53 ± 0.82	0.03
30-min delay	12.53 ± 1.09	15.37 ± 0.89	16.37 ± 0.98	0.30
Complex figure copy	34.61 ± 0.036	35.00 ± 0.33	34.89 ± 0.41	0.82	Treatment: 0.49
3 min	28.47 ± 1.37	29.39 ± 1.30	28.82 ± 1.75	0.62
30 min	28.29 ± 1.19	29.58 ± 1.28	27.82 ± 2.08	0.26
Working memory
N-back number correct					Treatment: 0.01
1-back	13.63 ± 0.38	15.05 ± 0.24	14.53 ± 0.36	0.13
3-back	9.05 ± 0.63	10.84 ± 0.71	9.63 ± 0.71	0.02
SOP errors					Treatment∗test: 0.02
6	0.80 ± 0.11	0.28 ± 0.09	0.32 ± 0.10	0.72	>0.99
8	1.11 ± 0.14	0.82 ± 0.12	0.68 ± 0.16	0.23	0.88
10	1.60 ± 0.13	1.12 ± 0.17	1.39 ± 0.21	0.05	0.17
12	2.02 ± 0.24	1.12 ± 0.19	1.59 ± 0.27	0.03	0.06
Digit span	7.32 ± 0.63	8.00 ± 0.60	8.89 ± 1.57	0.56	Treatment: 0.39
Motor learning
Pursuit rotor trial					Treatment: 0.80
1	22.77 ± 1.70	33.33 ± 3.03	33.85 ± 2.63	0.90
2	25.02 ± 1.83	33.91 ± 2.73	34.50 ± 2.30	0.94
3	26.90 ± 1.87	34.73 ± 2.70	35.51 ± 2.44	0.91
4	26.34 ± 1.89	33.50 ± 2.69	35.71 ± 2.84	0.54

Individual cognitive tests are grouped by the three memory subdomains (first column). P values for unadjusted paired t tests comparing the two treatment arms are shown in the fifth column. Adjusted P values in the repeated-measures analysis are shown in the sixth column. These P values were adjusted for age, years of education, WAIS-R, and baseline measure for that particular variable. Each test was analyzed as a group to adjust for within-subject correlation and correlation between responses on similar tests. For the adjusted analyses, treatment∗test P values were examined initially. If the treatment∗test P value was greater than 0.1, it is not shown in the table, and only the overall treatment effect is shown. If the treatment∗test P value was 0.1 or less, this value is listed in the table, and subsequent analyses were done for individual measures. P values are bold if they are significant or marginally significant and are discussed further in Results (P < 0.10).

To assess whether differences in cognitive measures could be explained by differences in health status or mood, models with significant cognitive differences were adjusted for changes in SF-36 and/or POMS measures. Separate models were fitted to assess effects of each individual change in SF-36 and/or POMS score that showed significant euthyroid-hypothyroid differences.

To investigate associations between changes in TSH, free T₄, or free T₃ and outcomes, regression models between the change in psychological, mood, or cognitive measure and change in hormone level were constructed. Three models were developed for each outcome, in which one thyroid hormone was added as a covariate. In a final model for each outcome, the association between a single thyroid hormone measure was adjusted for the two other thyroid hormone measures to assess for independent and/or confounding effects. The number of models was large, so we were interested in the patterns of P values across similar measures and similar hormones rather than in individual P values.

Results

Clinical parameters and thyroid function tests (Table 1)

Thyroid hormone levels were slightly improved at the end of the euthyroid arm, compared with the baseline visit, likely reflecting improved compliance during the study. Per the study design, mean TSH levels increased, whereas free T₄ and free T₃ levels decreased at the end of the subclinical hypothyroid arm. One subject had a slightly low free T₄ level at the end of the subclinical hypothyroid arm, and seven other subjects had slightly low free T₃ levels (five already had slightly low free T₃ levels with normal TSH levels at baseline). Thus, we were successful in inducing subclinical hypothyroidism, with almost all serum free T₄ and free T₃ levels remaining within the normal range and with TSH levels in our target range. Mean L-T4 doses were 45% lower during the subclinical hypothyroid arm (mean 1.50 vs. 0.83 μg/kg·d, range 0.6–2.35 vs. 0.3–1.44 μg/kg·d).

All subjects completed the study, and none dropped out due to clinical issues. There were no differences in weight, pulse, or blood pressure at the end of the two arms of the study. The Billewicz scale was marginally worse at the end of the subclinical hypothyroid arm compared with the end of the euthyroid arm (P = 0.08).

Subjects did not reliably predict which arm was the subclinical hypothyroid arm: 11 guessed correctly, seven guessed incorrectly, and one had no opinion (P = 0.25 by binomial calculation). Six preferred the actual lower (subclinical hypothyroid) dose, 12 preferred the actual higher (euthyroid) dose, and one had no preference (P = 0.12). One preferred the dose she perceived was the lower dose, 17 preferred the dose they perceived was the higher dose, and one had no preference (P < 0.0001).

Health status and mood (SF-36 and POMS) (Table 2)

SF-36 mental and physical component summary scale scores were not different between the two arms of the study. The SF-36 general health subscale was marginally worse during the subclinical hypothyroid arm (adjusted value = 0.08). The fatigue subscale of the POMS was worse during the subclinical hypothyroid arm (P = 0.03 by unadjusted t test, treatment effect on POMS by adjusted P = 0.13). There were no significant effects on other SF-36 or POMS subscales.

There were no associations between changes in the SF-36 or POMS scales and changes in TSH, free T₄, or free T₃ levels between the two study arms (P = 0.10–0.99, data not shown).

Cognitive tests (Table 3 and Fig. 1)

Fig. 1.

Measures of working memory at the end of the euthyroid arm (solid lines) and subclinical hypothyroid arm (dashed lines). The top panel shows the average incorrect for the SOP test; higher numbers indicate worse performance. SOP performance was worse during the subclinical arm of the study (P = 0.02 by mixed effects model). The bottom panel shows the mean number correct for the N-back test; higher numbers indicate better performance. N-back performance was worse during the subclinical hypothyroid arm of the study (P = 0.01 by mixed effects model).

Declarative memory

Paragraph recall (verbal memory) was slightly better at the end of the subclinical hypothyroid arm (adjusted P = 0.02). This was due to a difference of 1.7 items on immediate recall, with no difference in delayed recall. There were no differences in performance on the complex figure test (visual memory).

Working memory

N-back number correct and subject-ordered pointing were significantly worse at the end of the subclinical hypothyroid arm, compared with the end of the euthyroid arm (adjusted P = 0.01, 0.02) (Fig. 1). Digit span was not different between the two arms.

Motor learning

Pursuit rotor results did not differ between the two arms of the study.

The differences in N-back and SOP between the two arms of the study were not related to health status (SF-36) or mood (POMS) changes by mixed effects models (data not shown).

The differences in N-back and SOP were related to changes in free T₄ levels between the two arms of the study by mixed-effects models (P = 0.005–0.05). The difference in N-back was also marginally related to change in free T₃ and TSH levels between the two arms (P = 0.04–0.06) (data not shown).

There were slight differences between baseline and euthyroid arm levels for some of the cognitive tests, likely reflecting learning effects of repeated testing. We accounted for these differences by randomized order of the two intervention arms and by using baseline levels as a covariant in our mixed models.

Discussion

We investigated the effects of subclinical hypothyroidism on health status, mood, and cognition, using a unique experimental model in subjects with treated hypothyroidism. This model circumvents limitations inherent in studying endogenous subclinical hypothyroidism: subject recruitment and heterogeneity, variable natural history, and biases due to subjects’ awareness of their thyroid status. To our knowledge, there is only one report of a similar model; Walsh et al. (40) adjusted L-T4 doses in hypothyroid subjects to vary TSH levels within the normal range. We show here for the first time that subclinical hypothyroidism can be safely and effectively induced in a blinded, controlled fashion.

We found marginal decrements in the Billewicz scale in the subclinical hypothyroid arm. Despite this, our subjects did not accurately guess which arm included the lower L-T4 dose. Subjects overwhelmingly preferred the arm they perceived was the higher dose arm, confirming the impression that patients often have a bias toward higher doses of L-T4.

Confirming our first hypothesis, we found slight decrements in the SF-36 general health and POMS fatigue subscales on the subclinical hypothyroid arm. Previous studies have reported variable effects of subclinical hypothyroidism on health status and mood (6, 10, 11–14, 17, 19, 21–23, 41). Our study had the advantage of subjects being blinded to their treatment status, and we found subtle but significant effects on these outcomes.

Confirming our second hypothesis, we found decrements in two measures of working memory (N-back, SOP) on the subclinical hypothyroid arm. This was specific for working memory because we found no decrements in declarative memory or motor learning. We did not find differences in performance on the digit span backward test, but this is not as sensitive a measure. We did not find correlations between the decrements in working memory and alterations in health status or mood. Thus, these appear to be direct effects of altered thyroid status on the brain areas that subserve working memory. We do not have an explanation for the observed difference in paragraph recall unless it was an effect of multiple testing.

Previous studies of cognition in subclinical hypothyroidism have been variable in terms of patients and control subjects, cognitive domains and test measures, study designs, and outcomes (4, 6, 8–19). We did not use a battery of general cognitive measures, which might be less sensitive to subtle changes but rather validated measures targeted to specific domains likely to be affected by altered thyroid status. To our knowledge, this is the first study that investigated subdomains of memory in subclinical hypothyroidism, and our results provide targeted areas for further investigations.

A reasonable question is whether our results have clinical significance. The difference in the SF-36 general health subscale is similar to changes reported with weight loss among overweight adults (42) or conversion from atrial fibrillation to sinus rhythm (43). The difference in the POMS fatigue subscale is similar to changes reported with a very low energy diet in healthy women (44). The difference in N-back performance is similar to that reported during aging or sleep deprivation (45, 46). Thus, although the magnitude of the effects we documented is small, we believe they are clinically relevant.

We targeted TSH levels of 10–20 mU/liter for the subclinical hypothyroid arm, recognizing that our results may not pertain to more mild TSH elevations (5–10 mU/liter). Further studies are necessary to address this issue, and our results suggest that working memory measures are important outcomes.

Our final hypothesis was that the changes in health status, mood, and/or memory between the two arms would be correlated with changes in thyroid hormone levels. This was not true for the minor decrements in health status and mood. However, we found correlations between changes in free T₄ and working memory decrements. This supports the concept that subclinical hypothyroidism is part of the spectrum of thyroid dysfunction, with similar, although milder, decrements in clinical endpoints as those in overt hypothyroidism.

There are four limitations to this study: the sample size, variability in thyroid hormone levels during the subclinical hypothyroid arm, the short-term nature of the induced subclinical hypothyroidism, and the crossover design. We do not believe that the sample size is a major limitation because we found effects on two measures of working memory and accounted for multiple comparisons by statistical adjustment. Our study was a priori powered on the cognitive measures, and we had limited power to detect significant differences in the other measures. Post hoc analysis indicated that we would require 50–60 subjects for the marginal differences in the Billewicz scale and SF-36 to reach statistical significance. One subject had a low free T₄ and seven other subjects had low free T₃ levels at the end of the subclinical hypothyroid arm. Five of these seven already had low free T₃ levels at baseline, despite normal TSH levels. After excluding these eight subjects, results were similar for all measures of working memory, but the significance was attenuated due to loss of sample size. We limited the time period of subclinical hypothyroidism to 12 wk due to concerns over subject retention and clinical effects. A longer period of subclinical hypothyroidism might cause more decrements in outcome measures. However, patients might also adapt to subclinical hypothyroidism over time, such that our findings would lose significance. Published studies on subclinical hypothyroidism and cognition found positive effects of L-T4 treatment between 3 and 10 months (8–14, 17). Finally, the crossover design could lead to problems with learning effects during repeated testing and carry-over effects. We used alternative test forms during repeated testing, and randomized the order of the treatment arms, to minimize these effects.

In summary, we induced subclinical hypothyroidism in a controlled and blinded fashion in subjects with primary hypothyroidism. We found minor decrements in health status and mood and specific decrements in working memory. There were significant correlations between changes in free T₄ levels and the degree of impairment in these cognitive measures. This suggests that subclinical hypothyroidism induces specific deficits in brain areas that control working memory. Controlled studies of these outcomes in subjects with a wider range of serum TSH levels would help elucidate the spectrum of cognitive changes in mild thyroid disease.

Acknowledgments

This work was supported in part by Grants R21 DK062787 (to M.H.S.) and M01 RR00334 (to Oregon Health and Science University General Clinical Research Center).

Author Disclosure Summary: M.H.S., K.G.S., N.E.C., P.C., and J.S.J. have nothing to declare.

Abbreviations

 
• CV,

  Coefficient of variation;

 
• L-T4,

  l-thyroxine;

 
• POMS,

  Profile of Mood States;

 
• SF-36,

  Short Form 36 health survey;

 
• SOP,

  subject-ordered pointing;

 
• WAIS-R,

  Wechsler Adult Intelligence Scale Revised.