Clinical review: The thyroid in mind: cognitive function and low thyrotropin in older people.

CONTEXT: Several studies have reported an association between low serum TSH, or subclinical hyperthyroidism (SH), and dementia, but little emphasis has been placed on this field because not all studies have demonstrated the same association. We performed a detailed systematic review to assess the evidence available to support the association between these two conditions.
METHODS: We performed a systematic search through the PubMed, Embase (1996 to 2012 wk 4), Cochrane Library, and Medline (1996 to January wk 4, 2012) electronic databases using key search terms encompassing subclinical hyperthyroidism, TSH, dementia, and cognitive impairment.
RESULTS: This review examines the 23 studies that provide information about the association between SH or lower serum TSH within the reference range and cognition. Fourteen of these studies, including several well-designed and well-powered cross-sectional and longitudinal analyses, have shown a consistent finding of an association between SH with cognitive impairment or dementia.
CONCLUSION: There is a substantial body of evidence to support the association between SH and cognitive impairment, but there is no clear mechanistic explanation for these associations. Nor is there an indication that antithyroid treatment might ameliorate dementia. Larger and more detailed prospective longitudinal or randomized controlled trials are required to inform these important questions.

With ready access to sensitive hormone assays, the last few decades have witnessed a dramatic increase in serum thyroid function testing. This has raised many issues about the interpretation of minor deviations in thyroid function test results, particularly in individuals with little or no conventional clinical evidence of thyroid disease. The elderly are overrepresented among individuals with minor abnormalities in serum TSH and thyroid hormone concentration, and the clinical significance of these biochemical abnormalities in older people is the least clear. Overt hypothyroidism is well-established as a reversible cause of cognitive impairment, which may sometimes be profound. However, overt hyperthyroidism is also well known to be associated with impairment of concentration, mood changes, and alterations in perception. A more difficult, but nonetheless important question is whether there is evidence to support a relationship between more subtle alterations in thyroid function, in particular subclinical hyperthyroidism (SH) or low serum TSH concentration and cognitive impairment. However, the literature on this subject is heterogeneous with regard to study design and conclusions. In this review article, we aim to address the question of whether there is an association between low serum TSH and cognitive impairment and to explore possible underlying mechanisms.

Background to SH and the Aging Thyroid Axis

SH is defined as a serum TSH concentration below the reference range, with normal free T4 (FT4) and free T3 (FT3) levels. It can be divided into two groups; patients with a mildly reduced TSH (0.1–0.4 mU/liter) can be classified as having grade I SH, whereas those with a serum TSH of less than 0.1 mU/liter, including a completely suppressed TSH level (<0.01 mU/liter) are classified as having grade II SH (1). The prevalence of SH has been reported to range from 0.63 to 2.1% in epidemiological studies (2–4).

Most people with SH have no symptoms of hyperthyroidism, and relatively few develop specific complications attributable to SH: atrial fibrillation or osteoporosis (5). Prognostically, many people with SH have only a transient abnormality of thyroid function, with resolution of biochemical abnormalities found in 25–75% of those with grade I SH on repeat testing (6, 7). Thus, the majority of individuals with grade I SH do not have intrinsic thyroid disease, and the low TSH reflects nonthyroidal illness or drug effects. Nevertheless, approximately 1–2% of patients over the age of 60 with SH progress to overt hyperthyroidism, with progression most likely in those with grade II SH. This reflects the indolent nature of multinodular goiter as the dominant cause in this age group, with mild thyroid autonomy developing slowly. A small number of individuals with SH have early Graves' disease, and progression to overt hyperthyroidism is more probable in these cases (7). Conversely, 1–2% of elderly individuals have a persistent and stably low serum TSH that remains unexplained by any primary thyroid pathology. Thus, these alterations in thyroid function may reflect either altered physiology or pathology associated with advanced age.

The study of the physiological changes in the thyroid axis during aging is complicated by the presence of confounders such as medication, chronic illness, and increased risk of thyroid disease with age. Nevertheless, many observational studies in healthy older individuals, which took into account common confounders, have revealed an age-dependent decline in serum TSH and FT3 and an age-dependent increase in rT3 with maintenance of stable serum FT4 levels (8, 9). A further study demonstrated that thyroid function was well preserved until the eighth decade of life, with a decreased level of serum FT3 only observed in centenarians (10). However, the upper limit of serum TSH has also been shown to increase with age (11), leading to a wider spread of the TSH distribution with age, expanding both above and below the reference intervals for younger individuals. The mechanisms responsible for this age-related decline in serum TSH have not been fully examined. One hypothesis suggests an age-related reduction in the secretion of pituitary TSH and/or hypothalamic TRH, due to an increased pituitary thyrotrope sensitivity to peripheral T3/T4 negative feedback (12). Interestingly, the pattern of TSH pulsatility is also changed in the healthy elderly, with a blunted nocturnal TSH peak being observed, followed by a 1- to 1.5-h backward shift in the circadian rhythm of TSH secretion, leading to an earlier peak (13, 14). Equally compelling, a primary defect in thyroid hormone inactivation and disposal might explain the observation of unchanged serum T4 levels despite reduced tropic drive from lower pituitary TSH secretion. Therefore, reduced T4 and T3 degradation and clearance may lead to reduced throughput of hormone and a lower tone for the axis (15).

Brain and Thyroid Function

The population prevalence of dementia is about 7%, and this rises to more than 30% in those aged 85 yr and above (16, 17). Dementia can be classified as primary or secondary depending on its etiology. Alzheimer's disease (AD), frontal temporal lobe dementia, and Lewy body dementia are caused by primary degeneration of the brain, comprising about 70% of cases of primary dementia (18). Vascular dementia accounts for a further 15% of primary cases (19). The “cholinergic hypothesis” of AD pathogenesis is well accepted, with characteristic depletion of acetylcholine and presynaptic cholinergic markers in AD cortex and hippocampus (20, 21). Studies using single photon emission computed tomography or magnetic resonance spectroscopy have found similar characteristics in AD and in cognitively impaired patients with hyperthyroidism (22–24). Correspondingly, reduced levels of choline-related compounds in the brain have been demonstrated in hyperthyroidism, with a gradual normalization after antithyroid medication (22). Paradoxically, T4 administration led to an enhancement of rats' learning ability, along with increased cholinergic activity in the frontal cortex and hippocampus (25). Interestingly, TRH has also been shown to exert strong stimulant action on the cholinergic pathways of the cortex and hippocampus (26). Intraseptal injection and intracerebroventricular administration of TRH increased the turnover of acetylcholine in rat hippocampus and parietal cortex, respectively (27, 28). Hence, reduced TRH secretion could contribute to acetylcholine depletion, which is associated with the cognitive changes in AD.

The β amyloid plaque is the pathological hallmark of AD, and the potential pathways of β amyloid-mediated neurotoxicity include inhibition of acetylcholine activity in the cortex and hippocampus, oxidative stress from free radical damage, mitochondrial dysfunction, and ultimately apoptotic cell death (29–33). At the same time, thyroid function has been shown to influence systemic oxidative stress. Experimental and clinical studies have demonstrated increased reactive oxygen species and lipid peroxidases in the hyperthyroid state, resulting in diminishing antioxidative enzymes (34, 35). Furthermore, T3 has been shown to affect splicing of certain β-amyloid precursor protein isoforms, which are preferentially expressed in the AD brain (36). Therefore, thyroid hormones could also have a role in modulating the intracellular and extracellular contents of β-amyloid precursor protein isoforms and directly influence the pathogenesis of AD (36, 37).

Literature triage for systematic review.

Methods

Search strategy and analysis (Fig. 1)

We performed a systematic search through the PubMed, Embase (1996 to 2012 wk 4), Cochrane Library, and Medline (1996 to January wk 4, 2012) electronic databases. The key search terms were “subclinical hyperthyroidism,” “subclinical thyroid disorders or dysfunction,” “cognitive decline or *impairment,” “dementia,” and “Alzheimer's disease.” The fields evaluated were “treatment, epidemiology, complication and mortality.” We included all cross-sectional or longitudinal studies published or translated into the English language. Meta-analysis has not been performed in light of the heterogeneity in study design, statistical analysis methods, and outcome measures among the studies.

Results

Systematic review of the relationship between SH and cognitive function

Eighty-four studies were identified that investigated the relationship between thyroid dysfunction or serum thyroid parameters and cognitive impairment. We excluded investigations that involved participants with overt thyroid disorders and only included published papers that examined the relationship between SH, or variations of thyroid function within reference intervals, with cognitive function. Twenty-three studies published from 1996 to January 2012, evaluating a total of 31,482 patients, fulfilled the inclusion criteria. Fifteen studies indicated the correlation between thyroid function and dementia as the primary study objective. Eight studies included other entities such as cardiovascular or osteoporosis risk or imaging changes as the primary outcome, with the correlation of thyroid function and cognition as a secondary objective.

Of the 23 studies, 13 were case-control or cross-sectional single-phase trials, and the remaining 10 studies had a prospective longitudinal population-based design. To date, there have been no randomized interventional studies examining the effects of antithyroid treatment on cognition in SH.

Cross-sectional/case-control studies (38–50)

Categorical analysis of subclinical thyroid disease vs. euthyroidism; or quantiles of serum TSH concentration (38–43)

Six studies made a categorical analysis, either of SH vs. euthyroidism or of quantiles of serum TSH and thyroid hormone concentration in relation to cognitive function (Table 1). The Sao Paulo Ageing and Health Study, a large cross-sectional study comprising 1119 community-dwelling participants aged 65 yr and over, found an association between SH and dementia after multivariate adjustment (38). The effect was strongest for vascular dementia [odds ratio (OR) for all types of dementia, 4.9; 95% confidence interval (CI), 1.5–15.7; vascular dementia OR, 5.8; 95% CI, 1.4–33.1]. This was confirmed using an analysis of quintiles of TSH; those with the lowest serum TSH quintile showed more than a 3-fold increase in risk for all types of dementia [age-adjusted OR for dementia, 3.6 (95% CI, 1.4–8.9); vascular dementia, 9.3 (95% CI, 1.1–75.5)]. Similarly, the InCHIANTI Study, a population-based study of 916 Italians aged 65 yr and older, demonstrated a significantly lower Mini-Mental State Examination (MMSE) score among those with SH compared with the euthyroid group (22.61 ± 6.88 vs. 24.72 ± 4.52; P < 0.03) (39). Furthermore, SH was associated with more than a 2-fold risk of scoring less than 24 out of 30 on the MMSE; a standard threshold for significant dementia [hazard ratio (HR) = 2.26 (95% CI, 1.32–3.91); P = 0.003]. These two large studies were well designed with more than a 90% participant response rate, and both carefully excluded patients with overt thyroid dysfunction as well as those taking thyroid medication. Two smaller studies [van Osch et al. (40) and Dobert et al. (41)], involving 469 and 119 participants, respectively, supported these findings. van Osch et al. (40) investigated a cognitively intact group and an AD cohort aged greater than 52 yr. Interestingly, the euthyroid participants with a TSH level in the lowest tertile (TSH, 0.5–1.3 mU/liter) had more than a 2-fold increase in AD prevalence compared with those with serum TSH in the highest tertile (2.1–6.0 mU/liter) [OR, 2.04 (95% CI, 1.18–3.53); P = 0.01]. Similarly, Dobert et al. (41) demonstrated a significantly lower TSH level in participants with both AD and vascular dementia (P < 0.01) compared with controls without cognitive impairment.

Table 1.

Cross-sectional and case control studies on the relationship between SH and cognitive decline (1996–2011)

First author, year (Ref.)	Study size (n)	Mean age (range)	Thyroid status	Thyroid function indicators (normal range)	Objective cognitive measures	Study outcomes	Covariates	Exclusion criteria
Categorical analysis of subclinical thyroid disease vs. euthyroidism; or quantiles of serum T4 concentration
Benseñor, 2010 (38) (Sao Paulo study)	1119	Patients with dementia: 78.5 (8)Non-dementia, 71.9 (6.1)	SH and euthyroid	TSH, FT4; TSH (0.4–4.0 mIU/liter)FT4 (0.8–1.9 mg/liter)	CSI-D (Community Screening Instrument for Dementia), Geriatric Mental State (GMS), HAS-DDS (History and Aetiology Schedule Dementia Diagnosis and Subtype)	SH increased the risk of developing dementia, especially vascular dementiaAll type of dementia (OR, 4.9; 95% CI, 1.5–15.7)Vascular dementia (OR, 5.8; 95% CI, 1.4–33.1)AD (OR, 2.5; 95% CI, 0.3–20.8; P, NS)Age-adjusted OR for dementia, 3.6 (95% CI, 1.4–8.9); vascular dementia OR, 9.3 (95% CI, 1.1–75.5)	Age, gender, BMI, education, smoking, history of alcohol abuse, hypertension	Overt thyroid dysfunctionPatients on thyroid medications
Ceresini, 2009 (39) (InCHIANTI study)	916	>65	SH and euthyroid	TSH, FT3, FT4TSH (0.46–4.68 mU/liter)FT4 (0.77–2.19 ng/dl)	MMSE	MMSE was significantly lower in SH group compared to euthyroid group (22.61 ± 6.88 vs. 24.72 ± 4.52; P < 0.03)HR, 2.26 (95% CI, 1.32–3.91); P = 0.003	Age, sex, smoking, chronic heart failure, DM, hypertension, Parkinson's disease, BMI, physical activity	Participants on thyroid medications, amiodarone, lithium; dementia
Van Osch, 2004 (40)	469	>52	Euthyroid only	TSH (0.5–6 mU/liter)	CAMCOG (Cambridge Examinations for Mental Disorders of the Elderly), MMSE	Lowest tertile of TSH was associated with a more than 2-fold increased risk of AD, compared to the highest tertile (OR, 2.04; 95% CI, 1.18–3.53; P = 0.01)	Smoking, hypertension, stroke, DM, CVD, alcohol. APOEϵ4 genotype, total homocysteine level, depression	TSH outside normal range; patients on thyroid medications
Dobert, 2003 (41)	119	69.8 ± 14	SH and euthyroid	TSH, FT4, FT3TSH (0.5–4.0 mU/liter)FT4 (11–24 pmol/liter)	MMSE, CERAD, Short Cognitive Performance test, MRI, PDG-PET	Patients with dementia showed 3-fold increased probability of having decreased or borderline TSH values (29%) vs. control (10%)	Age, sex	History of overt thyroid disease; patients on thyroid medications or contrast medium 4 wk before the laboratory testing
Roberts, 2006 (42)	5868	(65–98)	SH and euthyroid	TSH (0.4–5.5 mU/liter), FT4 (9–20 pmol/liter)FT3 (3.5–6.5 pmol/liter)	MMSE, MEAMS (Middlesex Elderly Assessment of Mental State)	No association between subclinical thyroid dysfunction and cognition or mood	Dementia, CVD. RA, PVD, psychiatry diseases, osteoporosis, nonspecific thyroid diseases, DM, pulmonary or gastrointestinal diseases, medications (amiodarone, lithium, β-blocker, antiepileptics, antidepressants, kelp, tranquilizers, steroid, morphine)	Overt thyroid diseases or taking thyroid medications
Van der Cammen, 2003 (43)	829	78.2	All thyroid status (TSH)	TSH (no normal range provided)	Diagnostic and Statistical Manual of Mental Disorders to diagnose AD	No differences in TSH level between AD patients and those without dementia	No information available	No exclusion criteria (patients with overt thyroid disease were included)
Multivariate analysis with thyroid function markers or cognitive performance as continuous variables
Wahlin, 1998 (44)	200	(75–96)	Euthyroid only	TSH, FT4TSH (0.4–5 mU/liter)FT4 (12–25 pmol/liter)	Two-letter fluency tasks; Block design test with WAIS-R; Trail Making Test; Episodic memory tests	Positive association between low normal TSH and worse episodic memory (P < 0.01)	Age, education, mood symptoms	Overt thyroid dysfunction, patients on neuroleptic or antithyroid medications, TFT outside the normal range, psychiatric illness
Stuerenburg, 2006 (45)	227	71.6	All thyroid status (mean TSH 17.3 ± 3.2)	FT4, TSH, FT3, TT4, TT3Lowest FT4 quartile <15.1Highest FT4 quartile >19.0	MMSE	Significant inverse correlation between plasma FT4 and MMSE score (Spearman rank correlation = −0.14; P = 0.04)	Smoking, hypertension, LDL, HDL, APOEϵ4 allele, depression	No exclusion criteria
De Jongh, 2011 (46)	1219 (34-SH)	75.5 (68.9–82.1)	Euthyroid, SH and subclinical hypothyroidism	TSH, FT4, FT3TSH (0.3–4.5 mU/liter)	MMSE, the Raven'S colored progressive matrices (RCPM), the coding task and the audiotry verbal learning test	Subclinical hyper- or hypothyroidism was not associated with impaired global cognitive function	Age, sex, alcohol use, smoking, educational level, mean arterial pressure, BMI, heart rate, total cholesterol, and physical activity	Antithyroid or T4 medications
Patterson, 2010 (47)	409	76.9 (52–94)	Euthyroid only	TSH, FT4No normal range provided	NART (National Adult Reading Test); MMSE; HVLT (Hopkins Verbal Learning test)	No relationship between cognitive function and thyroid hormones.Higher FT4 was associated with worse functional independence (P < 0.001)	Age, sex, mood	No exclusion criteria
van Boxtel, 2004 (48)	120 (healthy volunteers)	>45	All thyroid status	TSHNo normal range provided	MAAS test battery (memory, sensorimotor speed, information processing, cognitive flexibility	Higher TSH was associated with poorer memory (P < 0.05), but this association disappeared after correction for depression score	Depression, education, overt thyroid diseases	Dementia, PD, CVD, epilepsy, chronic psychotropic drug usage, CNS tumor
Quinlan, 2010 (49)	69	60.9–66.8	All thyroid status:	TSH, FT4, TT4, TT3Only TT3 level given (1.4–1.6 nmol/liter)	Trail Making Test; RAVLT delayed recall; Block design; Token test; Boston Naming test, Stroop test etc	MCI group with higher TT3 showed more cognitive impairment in episodic memory (P = 0.0080), language (P = 0.001), and executive function (P = 0.012)	Age, sex, BMI, total cholesterol, HDL, LDL, BP, use of T4, β-blocker and estrogen	Psychiatric disorders, depression, systemic illness, diabetes, cerebral tumor, CNS infection, chronic alcoholism, patients on steroid treatments, patients with dementia
Prinz, 1999 (50)	44	72	All thyroid status	TT3, TT4, TSH; no normal range provided	MMSE; CATMEAN (category fluency), FASMEAN (verbal fluency) etc	Higher TT4 was significantly associated with better WAIS score (P < 0.05) and global cognitive performance (P < 0.01)	Age, education	Overt thyroid diseases, DM, dementia (MMSE score <27) depression, MI, BMI >18 or <33 kg/m2, hypertension, CNS medication used (including 2 wk prior to testing), head/trauma or infection, other systemic illness, neurological, alcohol, sleep disorder

NS, Nonsignificant; MMSE, Mini-Mental State Examination; CERAD, Consortium to Establish a Registry for Alzheimer's disease; BMI, body mass index; DM, diabetes mellitus; CVD, cardiovascular disease; RA, rheumatoid arthritis; PVD, peripheral vascular disease; LDL, low-density lipoprotein; HDL, high-density lipoprotein; TFT, thyroid function test; APOEε4, apolipoprotein Eε4; BP, blood pressure; PD, Parkinson's disease; MI, myocardial infarction.

In contrast, the largest cross-sectional population-based study, involving 5868 participants, showed no association between SH and cognition (42). However, this study invited participants by mail, and the response rate was poor at 38%. Although the responders were representative of the regional population with respect to demographic characteristics, the average MMSE score was above 27 in all subgroups, suggesting that the responder population was skewed toward a cognitively intact group. Another cross-sectional observational study, involving 829 consecutive unselected referrals to a hospital geriatric clinic, did not demonstrate any difference in TSH level between the group with AD and the cohort without dementia (43). However, comorbidities and medication use were not described, so the results may be confounded by nonthyroidal illness, thyroid medication, or overt thyroid dysfunction.

Serum thyroid function markers (TSH/FT4/TT3) or cognitive performance as continuous variables (44–50)

An additional seven studies performed multivariate analyses to explore the relationship between cognition and thyroid function. Wahlin et al. (44) were the first to suggest an association between poorer cognitive performance and lower serum TSH concentrations within the reference range. A significant positive correlation was found between a lower episodic memory score and serum TSH concentration (P < 0.01) in 200 healthy, euthyroid, community-dwelling participants aged over 75 yr. In keeping with these findings, Stuerenburg et al. (45) demonstrated an inverse correlation (P = 0.04) between serum FT4 level and MMSE score in 227 patients aged 49–91 yr with mild to moderate AD.

In contrast, the remaining five studies of this design failed to demonstrate any association between cognition and low serum TSH or high serum FT4 level (46–50). In a cross-sectional study involving 1219 community-dwelling participants, global cognitive impairment was not increased in those with SH (n = 34) (46). In a hospital-based study involving 409 euthyroid patients aged 52–94 yr diagnosed with probable AD, no association was found between serum TSH or FT4 and dementia (47). Two smaller studies involving 120 healthy volunteers and 69 hospital-based participants, respectively, showed similar findings (48, 49). Paradoxically, a study of 44 healthy, community-dwelling male volunteers with a mean age of 72 yr observed a significant association between higher total T4 (TT4) levels and better cognition [Verbal performance in Wechsler Adult Intelligence Scale (P < 0.05) and global cognitive performance (P < 0.01)] but failed to reproduce this trend with TT3 and FT4 (50). It is worth noting that almost all of these studies have relatively small sample sizes, that each made a single measurement of thyroid function, and they used different cognitive performance tests, which limits the generalizability of these results.

Prospective longitudinal studies (51–60)

Categorical analysis of subclinical thyroid disease vs. euthyroidism; or quantiles of serum TSH concentration (51–56)

Six studies performed categorical analyses, either of SH vs. euthyroidism or quantiles of serum thyroid hormone concentration in relation to cognitive function (Table 2). The Rotterdam study was the first longitudinal study to examine the relationship between SH and dementia in 1893 population-based subjects (mean age, 69 yr) (51). Over a 2-yr follow-up, a 3-fold increased risk of dementia from all causes [relative risk (RR), 3.5; 95% CI, 1.2–10] and AD (RR, 3.5; 95% CI, 1.1–11) was found among those with a low baseline TSH level (<0.4 mU/liter). This study was limited by the small number of the SH group with dementia (n = 25) and a short follow-up period. Nevertheless, the positive findings in this study are supported by two well-conducted, large population-based studies with a longer follow-up period. The first study, a community-based observational study, was carried out over a period of 12 yr by Tan et al. (52) in 1864 patients who were free of dementia for 3 yr at recruitment. The patients were then divided into three tertiles according to baseline TSH: T1, less than 1.0 mU/liter; T2, 1.0–2.1 mU/liter; and T3, more than 2.1 mU/liter. A positive correlation in women with a serum TSH level in the lowest (T1) tertile (HR, 2.26; 95% CI, 1.36–3.77; P = 0.002) and the highest (T3) tertile (HR, 1.84; 95% CI, 1.10–3.08; P = 0.003) with AD risk was found. This relationship was not found in men, but other indices of thyroid function were not studied. However, the two aforementioned large prospective studies involved only a single measurement of thyroid function. This limitation has recently been addressed by the largest observational study involving 12,115 participants (2,004 with SH; 10,111 euthyroid individuals) with a mean age of 66.5 yr and a median follow-up period of 5.6 yr (53). This well-designed, population-based study had a carefully selected SH cohort, including only individuals with two confirmatory measurements of serum TSH at least 4 months apart. Patients whose TSH normalized or who developed overt thyroid disease during the course of follow- up were excluded. A positive association between SH and dementia was observed (adjusted HR, 1.64; 95% CI, 1.20–2.25). Interestingly, when patients were divided into two groups according to the TSH level (0.1–0.4 vs. <0.1 mU/liter), no relationship could be established between suppressed TSH (<0.1 mU/liter) and dementia, but a significant association remained for the group with TSH at 0.1–0.4 mU/liter.

Table 2.

Longitudinal populational-based studies on the relationship between SH and cognitive decline (1996–2011)

Author and year of publication	Study size (n) and setting	Mean age (range)	Follow-up interval (yr)	Participants' thyroid status	Thyroid function indicators (normal range)	Objective cognitive measures	Study outcomes	Covariates	Exclusion criteria
Categorical analysis of subclinical thyroid disease vs. euthyroidism; or quantiles of serum T4 concentration
Kalmijn, 2000 (51)	1893 community	68.8 (54–94)	2–4	SH and euthyroid	TSH, FT4, TT4TSH (0.4–4.0 mU/liter)FT4 (11–25 pmol/liter)	MMSEGMS-ACAMDEX	SH increased the risk of dementia and AD 3-fold after a 2-yr follow-up (RR, 3.5; 95% CI, 1.2–10)	Age, sex, education, smoking status, atrial fibrillation, depression	Dementia at baseline, patients on amiodarone, β-blocker or thyroid medications
Tan, 2008 (52)	1864 community	71	12.7	SH and euthyroid	TSH (0.5–5.0 mU/liter)	MMSE	Positive association between women with serum TSH in the lowest (<1.0 mIU/liter) and highest (>2.1 mIU/liter) and increased risk of AD	Age, plasma homocysteine levels, BMI, education, APOEϵ4 allele, stroke, atrial fibrillation	Dementia at baselineRisk of AD for: lowest TSH tertile (HR, 2.26; 95% CI, 1.36–3.77); highest TSH tertile (HR, 1.84; 95% CI, 1.10–3.08)
Vadiveloo, 2011 (53)	12,115 community; SH- 2004; euthyroid, 10,111	66.5 ± 15.9	Median, 5.6 yr	SH and euthyroid	TSH (0.4–4.0 mU/liter); FT4 (10–25 pmol/liter); FT3 (0.9–2.6 nmol/liter) (at least 2 measurements of TSH, minimally 4 months apart)	Diagnosis of dementia, ICD9 and 10	Positive association between SH and dementia. Adjusted HR 1.64 (95% CI, 1.20–2.25)No relationship established between TSH concentration (0.1–0.4) vs. <0.1 mU/liter and dementia (limited by small number of patients with TSH <0.1)	Age, gender, history of dementia and psychiatry disease	Age <18 yr, patients treated with antithyroid medications/RAI/ thyroidectomy before and during the first year after the first abnormal TFT, patient on amiodarone, T4 replacement during the study period, pregnancy. Patient on long-term steroid replacement/ pituitary disease
de Jong, 2009 (54)	615 community	77.3–78.6	5	SH and euthyroid	TSH, FT4, TT4. TSH (0.4–4.3 mIU/liter); FT4 (0.85–1.94 ng/dl)	CASI (100 point Cognitive Ability Screening Instruments)	Higher TT4 and FT4 were associated with dementia (HR 1.21; 95% CI, 1.04–1.40); AD (HR, 1.31; 95% CI, 1.14–1.51) and neuropathology	Age, education level, depression, albumin level, BMI, cholesterol , DM, hypertension, smoking, patients on T4, β-blocker or other cardiac antiarrhythmic drugs	T4 level out of normal range
de Jong, 2006 (55)	1,025 community	72.3 (60–90)	5.5	Euthyroid only	TSH, FT4, FT3. TSH (0.4–4.3 mU/liter); FT4 (0.85–1.94 ng/dl)	MMSE, GMS-A, CAMDEX	No association between increased risk of dementia or AD with TSH or thyroid hormoneHigher FT4 associated with greater atrophy at hippocampus and amygdala on MRI	Sex, educational level, smoking, depression, medication use (amiodarone, β-blocker, steroids), BMI, cholesterol, homocysteine, smoking, creatinine APOEϵ4 genotype, diabetes, atrial fibrillation	Blindness, dementia, contraindication to MRI, patients on thyroid medications
Volpato, 2002 (56)	464 community	77.5	3	Euthyroid only	TSH, FT4. TSH (0.3–5.0 mU/liter); FT4 (4.5–12.5 ng/dl)	MMSE	Low T4 within the normal range was associated with cognitive impairment over a 3-yr period (RR, 1.97; 95% CI, 1.10–3.5)	Age, race, educational level, coronary heart disease, hypertension, stroke, diabetes, PVD, depression, cancer	Nil
Multivariate analysis with thyroid function markers or cognitive performance as continuous variables
Hogervorst, 2008 (57)	1,047 community	(64–94)	2	Euthyroid only	TSH, FT4. TSH (0.3–4.8 mU/liter); FT4 (13–23 pmol/liter)	MMSE; AGECAT (Automated Geriatric Examination and Computer Assisted Taxonomy)	High normal FT4 had a negative association with baseline MMSE and accelerated cognitive decline after 2 yr (P = 0.03)	Age, sex, education, MMSE at baseline, mood, vascular risk factor (smoking, hypertension, heart disease, DM, stroke)	MMSE <18
Gussekloo, 2004 (58)	558 community	85	3.7	All thyroid status	TSH, FT4, FT3. TSH (0.3–4.8 mU/liter); FT4 (13–23 pmol/liter)	MMSE; Stroop test; Letter Digit Coding test; Word Learning test	Increased level of TSH was associated with better memory on follow-up (P = 0.03), when participants on T4 were excluded	Age, education	No exclusion criteria
Wahlin, 2005 (59)	200 community	75–96	3, then 6 yr	Euthyroid only	TSH, FT4. TSH (0.4–5 mU/liter); FT4 (12–25 pmol/liter)	Two-letter fluency tasks; Block design test with WAIS-R; Trail Making Test; Episodic memory tests	Positive association between decreased level of TSH with increased episodic recall deficits at 6-yr follow-up (β 0.290; P < 0.05)No association found in 3-yr follow-up	Age, education, mood symptoms	Over thyroid diseases, thyroid medications, psychiatric illness (but dementia is not excluded due to small sample size)
Annerbo, 2006 (60)	93 hospital-based	Men, 64.7; women, 65.4	5	All thyroid status	TSH (no normal range provided)	MMSE	Low TSH predicted risk of developing AD after controlled for other risk factors (OR for square root of TSH, 0.287; 95% CI, 0.088–0.931)	Stroke, cardiovascular disease, T4 treatment	No exclusion criteria

GMS-A, Geriatric Mental State schedule; CAMDEX, Cambridge examination for disorders of the elderly; BMI, body mass index; APOEε4, apolipoprotein Eε4; DM, diabetes mellitus; PVD, peripheral vascular disease; RAI, radioiodine therapy; TFT, thyroid function test.

In addition to these three large population-based studies, de Jong et al. (54) carried out a smaller prospective study among 615 Japanese-American men in the Honolulu Heart Program. The cohort had a mean age of 77.5 yr, and the mean duration of follow-up was 5 yr. This study observed a 20 and 30% increased risk for dementia and AD, respectively, for each sd increase in serum FT4 (HR for dementia, 1.21; 95% CI, 1.04–1.40; and HR for AD, 1.31; 95% CI, 1.14–1.51). In addition, neuropathological examinations from the autopsy program instituted as part of this study demonstrated a higher neocortical neurofibrillary tangle count per sd increase in TT4 (0.25; 95% CI, 0.05–0.46).

In contrast with the studies discussed so far, the Rotterdam Scan study, another prospective community-based study involving 1025 subjects aged 60–90 yr with a mean follow-up interval of 5.5 yr, showed no association between dementia and serum TSH or thyroid hormone levels (55). Nevertheless, they found a positive association between a higher FT4 and rT3 and greater atrophy at the hippocampus and amygdala on magnetic resonance imaging (MRI), in keeping with the finding in the Honolulu aging study. Although the Rotterdam Scan study had greater power than the original Rotterdam study, due to a larger-sized dementia cohort (n = 60 vs. 25) and a longer follow-up period, it is limited by single measurements of thyroid function and a small number of dementia cases with low TSH (n = 7).

Finally, a community-based study involving 464 euthyroid women aged 65 yr or older found an association between lower FT4 concentrations within the reference range and cognitive impairment over a 3-yr period (RR, 1.97; 95% CI, 1.10–3.5) (56). However, the same pattern was not exhibited in those with a frankly low TSH or an elevated FT4 level. There was a 14% dropout rate during follow-up, largely due to missing MMSE scores, which might have introduced a significant bias.

Serum thyroid function markers (TSH/FT4/TT3) or cognitive performance as continuous variables (57–60)

Four studies made multivariate analyses using thyroid function parameters or cognitive function (MMSE) as continuous variables. Hogervorst et al. (57) conducted a large prospective study involving 1047 community-dwelling patients aged 64–94 yr with MMSE scores of at least 25 in the United Kingdom and Wales. Higher serum FT4 concentrations within the reference range were associated with lower baseline MMSE scores and accelerated cognitive decline after 2 yr [OR, 1.13 (95% CI, 1.03–1.23); P = 0.006]. This study was limited by a single measurement of thyroid function. Prospective follow-up of 558 individuals aged 85 living in Leiden, The Netherlands, demonstrated an association between higher serum TSH and better memory function (β, 0.13; P = 0.03) (58). Nevertheless, this study had a small sample size and a 41.5% loss of follow-up due to death or refusal to continue participation.

After the cross-sectional study in 1998 (44), Wahlin et al. (59) conducted a follow-up survey on the same cohort and found an association between low serum TSH levels within the reference range and decreased verbal memory or episodic recall deficits at the 6-yr follow-up (β, 0.290; P < 0.05). Similarly, in a small hospital-based study, Annerbo et al. (60) demonstrated that for each unit decrease in serum TSH, the OR of developing dementia increased by 3.5.

Summary of observational studies

Twenty-three studies that met our criteria have examined the association between SH and cognition. Fourteen of these studies, including several well-designed and well-powered cross-sectional and longitudinal analyses, have shown a consistent finding of an association of SH, or low serum TSH within the reference interval, with cognitive impairment or dementia. In particular, this association was seen in more than three fourths of the prospective longitudinal studies, providing reliable evidence from robust studies. Several of the studies that did not confirm this result are potentially compromised by small samples sizes, recruitment from hospital environments, or the challenging nature of working with participants with cognitive problems (e.g. failure to return questionnaires).

Discussion

Given the weight of information suggesting an association of SH and lower serum TSH concentrations with cognitive impairment summarized above, we need to consider several explanations that could explain these findings (Table 3). A conventional explanation (explanation A) might be that mild endogenous SH reflects true thyroid overactivity causing excessive thyroid hormone action on the central nervous system. “Toxic” effects of thyroid hormones on the brain could be mediated by increased brain oxidative stress caused by the mild hyperthyroidism, which promotes reactive oxygen species production (35). Alternatively, thyroid hormone effects on the heart could mediate vascular dementia via thromboembolism from a combination of atrial fibrillation, myocardial and endothelial dysfunction, and hypercoagulability (5). However, we believe this explanation to be unlikely for several reasons. First, only about 20% of individuals with SH have a suppressed TSH (grade II SH), and so most don't have true endogenous hyperthyroidism. Clearly, neither do those people with serum TSH concentrations in the lower centiles of the healthy reference range. Indeed, the low but not suppressed TSH in the 0.1–0.4 mU/liter range is most commonly seen as a manifestation of nonthyroidal illness and is associated with a lowering of circulating FT3, rather than thyroid hormone excess. Second, for this explanation to be true, we would expect to see a biological gradient with cognitive defects being worse in individuals with the lowest TSH. In fact, this was not observed in the study of Vadiveloo et al. (53) where, if anything, the cognitive effects were most marked in the grade I SH group.

Table 3.

Possible mechanism for association of cognitive impairment with SH or low serum TSH

A. Excess circulating thyroid hormone resulting in neuronal loss.

B. Primary neurodegeneration causes reduced central nervous system TRH secretion, hence lower TSH.

C. SH and low TSH are biomarkers for age, and so are associated with other diseases of advanced age including dementias.

D. Subjects with cognitive impairment have a high burden of comorbidity, and association is due to nonthyroidal illness and drug effects on serum TSH.

An alternative explanation (explanation B) is that the organic brain diseases causing cognitive impairment also reduce TRH secretion from the hypothalamus and other brain areas. This would have a “knock-on” effect leading to lower TSH secretion and hence reduced thyroid hormone turnover (production and excretion)—in effect, a state of mild central hypothyroidism. There is little evidence to either support or refute this contention. It is known that there are projections from many areas of the brain to the paraventricular nucleus containing the TRH-secreting neurons, such as the C1–3 adrenergic neurons of the brainstem, the hypothalamic arcuate nucleus, and the neurons of the hypothalamic dorsomedial nucleus, which exert different effects on the hypophysiotropic TRH neurons (61–66). Furthermore, TRH is not only released from the paraventricular nucleus of hypothalamus but is also localized in neurons of the septal nuclei, preoptic area, raphe nuclei of the medulla oblongata, and spinal cord (67–69). Thus, TRH has a more generalized role as a central nervous system (CNS) neurotransmitter, and the brain involution of the dementia process may lead to a widespread perturbation of neurotransmitters, including TRH. If we accept this explanation as being plausible, then paradoxically, a study of thyroid hormone supplementation in dementia might be warranted.

A third explanation refers back to our understanding of thyroid hormone metabolism in older age. With this mechanism, one has to consider that low serum TSH (and indeed higher FT4) may be a marker of biological age, reflecting reduced hepatic clearance of thyroid hormones (including reduced type 1 deiodinase activity) and a consequent reduction in thyroid axis turnover. The well-established observation that older people with hypothyroidism require less levothyroxine replacement is the clinical correlate of this phenomenon (15, 70). In effect, the epidemiological studies reviewed above identify a group of individuals within a cohort who have more advanced biological age, using TSH as the biomarker (1). These individuals therefore have an excess of the degenerative disorders of advanced age, which includes cognitive impairment and dementia. Compellingly, SH is also associated with excess rates of atrial fibrillation, vascular events, low bone mineral density, fracture, and reduced muscle strength (71–74). Thus, dementia can be viewed as another noncausal association of low TSH/SH (1).

Individuals with cognitive impairment and dementias have a high burden of comorbidity and associated medication use. This may include many episodic intercurrent illnesses, such as transient infection, as well as more chronic degenerative conditions. All of these comorbidities can lead to reduced serum TSH, consequent to the well-described changes in serum thyroid hormones associated with nonthyroidal illness, known as “euthyroid sick syndrome.” Similarly, a wide variety of medications for conditions other than thyroid diseases including glucocorticoids, opiates, L-DOPA, amiodarone, and metformin are known to reduce serum TSH concentrations (1, 75). Thus, the combination of excess comorbidity and the consequent additional medications that may be prescribed for this may also explain a noncausal association of SH/low TSH with cognitive impairment. The fact that most of the studies performed in this field have relied upon a single TSH measurement makes these data particularly vulnerable to this interpretation. Nevertheless, although it is possible that this explanation contributes to some of the association between SH and dementia, it is unlikely to be the dominant factor. The large study of Vadiveloo et al. (53) involving 12,115 patients showed that this association persists after stringent ascertainment of a cohort categorized as having SH after repeated serum TSH measurement.

At the current time, we believe that there are insufficient data to allow discrimination between the various mechanistic possibilities outlined above (Table 3). Indeed, it is likely that there may be contributions from several of the above factors to the observed association between SH or low serum TSH and cognitive impairment. Until such time as more information becomes available, this remains an open question.

Conclusion

We have performed the first detailed review of a large and heterogeneous literature relating to the association between low serum TSH and cognitive impairment in older people. Overall, taking into account the largest and most robustly designed studies, there is a strong body of evidence to support the association between SH and cognitive impairment. However, there is no clear mechanistic explanation for these associations, nor is there any evidence to support the use of antithyroid measures to prevent or improve cognitive decline in this patient group. Larger and more detailed prospective longitudinal or randomized controlled trials are required to inform these important questions.