Abstract
Background:
Although some studies reported on the association of serum thyroid-stimulating hormone (TSH) concentration and cognition, only one population-based study investigated the association of TSH concentration and mild cognitive impairment (MCI).
Objective:
To investigate the gender-specific association of low- and high-normal TSH concentrations with MCI in euthyroid participants.
Methods:
Analysis sample 1 included 2,563 euthyroid participants (aged 50–80 years) from the second examination of the population-based Heinz Nixdorf Recall study. Gender-specific TSH quintiles (Q1 low, Q2-Q4 middle, Q5 high TSH concentration) were determined and group comparisons of age- and education-adjusted mean scores were performed for all cognitive subtests. Analysis sample 2 included 378 participants with MCI and 931 cognitively normal participants. MCI was diagnosed according to previously published MCI criteria. Multivariate logistic regression models were performed using TSH quintiles (Q2-Q4 as reference) to assess the association of low- and high-normal TSH concentration with MCI. Models were performed unadjusted and adjusted for sociodemographic and cardiovascular risk factors.
Results:
Group comparisons showed significant differences only in the immediate recall of the verbal memory task in women. Only women showed a strong association of high-normal TSH concentration with MCI (unadjusted: odds ratio 2.09, 95% confidence interval 1.29–3.37, full adjusted: 1.86, 1.06–3.27). There was no association with low-normal TSH concentration in women and no association of either low- or high-normal TSH concentration with MCI in men.
Conclusions:
These results suggest that women with high-normal TSH concentration might be at higher risk of cognitive decline. This needs to be confirmed in the longitudinal analysis.
Keywords
INTRODUCTION
While it is widely accepted that thyroid dysfunction can be a reversible cause of cognitive impairment, this relationship remains unclear regarding non-reversible cognitive deficits. In particular, studies focusing on thyroid function within the reference range reported inconsistent findings. For example, Castellano et al. reported no association of thyroid stimulating hormone (TSH) concentration with cognitive performance in euthyroid older persons in a case-control study [1]. In a population-based study, Moon et al. described an independent association of lower normal serum TSH concentrations with mild cognitive impairment (MCI) or dementia [2]. MCI is characterized by a measurable cognitive impairment that exceeds the normal age-associated cognitive decline. This impairment is not as pronounced as in demented persons. Furthermore, persons with MCI have normal activities of daily living and complex instrumental functions are at most slightly impaired [3]. Persons with MCI have an elevated risk of developing dementia and Alzheimer’s disease (AD) [4]. But persons with MCI can also remain stable for many years or even revert to a cognitively normal state. This modifiable characteristic makes the concept of MCI a promising target in the prevention of dementia.
The variability of study results can most likely be explained by differences in study design, study populations (including different age groups and gender), and cognitive test batteries as well as differences in diagnostic criteria of thyroid dysfunction including different reference values for TSH concentration. Additionally, the geographical area and its associated iodine status have a huge impact on thyroid function and, consequently, on the prevalence of thyroid dysfunctions. These facts emphasize the need for well-designed studies that carefully consider aspects like age, gender, and geographical area. By using data from the well-described population-based Heinz Nixdorf Recall study, we aimed to further examine the association of low- and high-normal TSH concentrations with MCI in euthyroid men and women aged 50 to 80 years.
MATERIALS AND METHODS
Study population
In the population-based Heinz Nixdorf Recall (Risk Factors, Evaluation of Coronary Calcium and Lifestyle) study, participants were randomly sampled in the Ruhr area in Germany. The study design has previously been described in detail [5]. Briefly, 4,814 participants aged 45 to 75 years underwent a baseline examination in 2000–2003. Five years later, participants were invited to the first follow-up examination with a response rate of 90.2% or 4,157 participants. In the first follow-up examination, a standardized cognitive performance assessment was implemented. The cognitive data of 71 participants were incomplete or missing and were excluded from the analysis. Twenty-two participants were excluded due to dementia, defined as a previous physician’s diagnosis, meeting the DSM-IV (Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition [6]) dementia diagnosis criteria or intake of cholinesterase inhibitors (anatomic-therapeutic-chemical classification issued by the World Health Organization (WHO) [7], code: N06DA) or other anti-dementia drugs (N06DX). Participants who did not fulfill the criteria for euthyroid status were excluded (n = 1,501, see below, Assessment of thyroid function). Participants who did not meet the MCI criteria or the criteria for ‘cognitively normal’ were excluded as well (n = 1,254, see below, Cognitive assessment). Thus, the final analysis sample consisted of 1,309 participants (Fig. 1). The study was approved by the University of Essen institutional review board and followed established guidelines of good epidemiological practice. All participants provided written informed consent.
Assessment of thyroid function
Serum blood samples were collected centrifuged and serum TSH concentrations were determined using the Roche Modular E170 electro-chemiluminescence immunoassay (Roche Diagnostics, Mannheim, Germany). The analytical sensitivity of the assay was 0.005 mIU/l. The recommended manufacturer reference range is 0.27–4.2 mU/l. TSH reference range has been intensely discussed in the literature, especially with focus on the upper cut-off score. Recommended scores vary between 2.12 mIU/l and 5.30 mIU/l [8, 9]. Interestingly, reference ranges in Germany differ depending on the specific area: Vözke et al. reported a reference range of 0.25–2.12 mIU/l for northeast Germany [8], while Burkhardt et al. reported a reference range of 0.52–3.60 mIU/l for a South German population [10]. But also other reference ranges have been recommended in Germany, e.g., 0.3–3.6 mIU/l [11]. Recently, Ittermann et al. reported formulas to define age and gender specific reference ranges [12]. Unfortunately, we were not able to define age-specific reference ranges due to power problems. Based on the recommended reference ranges with additional consideration of these reported formula, we decided to modify the recommended manufacturer range and defined a conservative reference range of 0.3–3.0 mIU/l. Participants with a TSH concentration out of this reference range (n = 2,16) were excluded as well as participants with missing data regarding their TSH concentration (n = 207). Participants who reported a positive history of thyroid diseases (n = 723) and/or use of thyroid medication (n = 355) were also excluded. Overall, 1,501 participants were excluded due to thyroid dysfunction, leaving 2,563 for the analysis of cognitive raw data (“analysis sample 1”, see Fig. 1 and Supplementary Table 1 for exclusions).
Cognitive assessment
The cognitive performance assessment comprised five subtests that tested the following cognitive domains: immediate and delayed verbal memory (eight word list [13]), speed of processing/problem solving (labyrinth test [13]), verbal fluency (animals [14]), and visuo-spatial ability (clock drawing test [15]). A detailed description of the cognitive assessment has been published previously [16]. When the MCI diagnosis was validated in a previous study, the short cognitive performance assessment showed a good accuracy compared to a detailed neuropsychological and neurological examination (area under the curve = 0.82, 95% CI = 0.78–0.85) [16].
Besides raw data, age- and education-adjusted data were used by determining three age groups (50–59 years, 60–69 years, and 70–80 years) as well as three education categories (10 years or less, 11–13 years, and 14 years or more). Education was defined as total years of formal education, combining school and vocational training (International Standard Classification of Education [17]). All raw data were z transformed for each subtest according to age and education groups.
MCI diagnosis was based on meeting all of the following published MCI criteria [18]: (1) cognitive impairment (defined as a performance of one SD below the age- and education-adjusted mean or a score of ≥3 in the clock-drawing test [15] in at least one cognitive domain); (2) subjective cognitive complaint (assessed with the question “In comparison to two years ago would you rate your memory function as better, same or worse?” if the answer was “worse”); (3) normal functional abilities and daily activities; (4) no dementia diagnosis. Participants who showed neither subjective cognitive complaint nor objective impairment were defined as cognitively normal. Participants who reported either subjective cognitive complaint (without objective impairment, n = 333, Fig. 1) or showed objective impairment (without subjective cognitive complaint, n = 912) were excluded from the analysis as well as participants with missing information regarding subjective complaint (n = 9). Overall, 1,254 participants were excluded due to their cognitive status, leaving 1,309 participants for the regression analysis (“analysis sample 2”, see Fig. 1 and Supplementary Table 2 for exclusions).
Assessment of covariates
We performed computer-assisted interviews to gain information about socioeconomic status, medical history, and coronary heart disease. Participants underwent three seated blood pressure measurements with an automated oscillometric blood pressure device (Omron, HEM-705CP). The mean of the second and third measurement was used for classifying the blood pressure according to the JNC-7 (Joint-National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure) definitions into four groups (systolic blood pressure and/or diastolic blood pressure): normal (<120 mmHg and <80 mmHg), pre-hypertension (120–139 mmHg or 80–89 mmHg), stage 1 (140–159 mmHg or 90–99 mmHg), and stage 2 (≥160 mmHg or ≥100 mmHg or intake of antihypertensive medication) [19, 20]. Height and weight were measured to determine the body mass index (BMI (kg/m2)). Diabetes mellitus was defined as fasting blood glucose of ≥7 mmol/l, random blood glucose of ≥11 mmol/l, previous physician’s diagnosis or intake of anti-diabetic medication. A history of smoking was defined as: “current smoker” (history of smoking during the past year), “former smoker” (history of smoking longer than one year ago) and “never smoker” (no history of smoking, data not available for n = 3 participants). Elevated depressive symptoms were assessed using the German version of the Center for Epidemiologic Studies Depression scale (CES-D) short form comprising 15 items with a cut-off score of ≥18 (data not available for n = 111 participants) [21, 22]. We used Cardio-Metabochip BeadArrays to differentiate the APOE alleles ɛ2, ɛ3, and ɛ4 by identifying the two single nucleotide polymorphisms (SNPs, rs7412 and rs429358, n = 107 missing).
Statistical analysis
In order to compare the demographic and cardiovascular characteristic of MCI participants with cognitively normal participants, we performed Mann-Whitney U tests or Chi-square-tests, as appropriate. All of the following calculations have been done for men and women separately. In analysis sample 1, performance in all cognitive subtests was compared for low- (quintile 1), middle- (quintiles 2–4), and high-normal (quintile 5) TSH concentrations using one-way analyses of variance (ANOVA) or Chi-square-test, as appropriate. Moreover, age- andeducation-adjusted means were calculated for all continuous cognitive test scores. Group comparisons of age- and education-adjusted means were performed using analyses of covariance (ANCOVA). In analysis sample 2, we performed the following logistic regression models to estimate odds ratios (ORs) and their corresponding 95% confidence interval (CI) for the association of low- and high-normal TSH concentration (low: quintile 1, high: quintile 5, reference: quintiles 2–4) with MCI: unadjusted; model 1: adjusted for age and education; model 2: additionally adjusted for BMI, hypertension (JNC 7 classes), diabetes mellitus, coronary heart disease, stroke, APOE ɛ4 genotype, and elevated depressive symptoms (for ORs and CIs of all covariates see Tables 3.1 and 3.2). The adjustment set was selected in advance based on the most recent findings from the literature [23, 24]. Analyses were conducted using IBM SPSS Statistics 22.0.
RESULTS
Looking at the socio-demographic and cardiovascular characteristic, participants with MCI were significantly older and more often had a positive history of stroke and depression (Table 1). Additionally, male participants with MCI had a higher prevalence of diabetes mellitus type 2 and more often a positive history of coronary heart disease compared to cognitively normal participants (Table 1). We performed two steps of exclusion analyses: step 1 was performed after exclusion of participants due to their thyroid function. Excluded women had significantly higher BMIs. Excluded men were significantly older and had higher scores on the depression scale. There were no other significant differences regarding socio-demographic or cardiovascular risk factors for excluded men and women. Step 2 was performed after exclusion of participants who did not meet the criteria for MCI or cognitively normal (see above, Cognitive measures). The only significant difference was a higher score on the depression scale in excluded men. All other analyzed factors showed no significantdifference.
Comparing the gender stratified cognitive performance in every subtest for low- and high-normal TSH concentrations in the analysis sample 1, we did not see significant differences in any subtest (Table 2). Looking at the age- and education-adjusted mean scores for participants with different TSH concentrations, we saw significant differences only in the immediate recall of the verbal memory task. This difference was only present in women with the worst performance associated with high-normal TSH concentration (Table 2). There were no other significant differences in any other cognitive subtest.
In the analysis sample 2, a strong association of high-normal TSH concentration with MCI was shown in women in the unadjusted logistic regression model (OR = 2.09, 1.29–3.37, Table 3.2). This association remained significant even after full adjustment (OR = 1.86, 1.06–3.27). There was no association with low-normal TSH concentration in women. In male participants, we found no association of either low- or high-normal TSH concentration with MCI(Table 3.1).
DISCUSSION
The main finding of this population-based study is a strong association of high-normal TSH concentrations with MCI in euthyroid women. There was no association observable regarding low-normal TSH concentration and MCI. Men did not show any association of low- or high-normal TSH concentrations and MCI. Additionally, the analysis of cognitive raw data showed the only difference between TSH concentrations for immediate recall of verbal memory in women.
So far, only one study investigated the association of TSH concentrations and MCI in euthyroid participants in a population-based study. In contrast to our study, Moon et al. showed an association of low-normal TSH concentration with MCI and dementia in 313 participants after 5 years [2]. There are several socio-demographic and geographic differences to our study that could explain the divergent results: Age (65 years and older compared to 50–80 years in our study population) is an important risk factor for thyroid diseases as well as cognitive decline. While Korea is considered one of the most iodine-rich areas [25], Germany was iodine deficient until iodine prophylaxis was established in 1993, after which iodine level remained normal but low [26]. Laurberg et al. reported that even small changes in the level of iodine intake had a severe impact on thyroid abnormalities [27]. Since our elderly study population was exposed to iodine deficiency for a long time and furthermore underwent this change in iodine supply, the diverging results might be due to differences in iodine status. Besides the longitudinal study design, Moon et al. investigated the total sample and did not report gender-stratified results [2]. TSH concentration is known to be influenced by gender [28]. A gender-stratified analysis might have revealed different results.
There is no other population-based study available that investigated TSH concentration and MCI in the reference range but there are several population-based studies that examined the association of TSH concentration and dementia. Tan et al. observed an elevated risk for incident AD with low and high TSH concentrations (including outside the reference range) in women after a mean follow up of 12.7 years [29]. Limiting the analysis sample to participants without thyroid medication use and TSH concentrations within the reference range, the risk for AD in women with high-normal TSH concentration was elevated but failed to reach significance (hazard ratio: 1.56, CI 0.91–2.69). There was no association with AD in men. These results are at least partially in line with the results of our study since our participants with MCI also have a higher risk of developing dementia. Although almost 38% of MCI cases are known to revert to a normal cognitive state, 65% of these converters develop MCI again or even develop dementia over a median time follow-up of 5.1 years [30]. This emphasizes the value of an MCI diagnosis regarding the prognosis even without considering biomarkers. Our planned longitudinal analysis will allow us to verify this hypothesis. The study by Tan et al. emphasizes that it is necessary to investigate the relationship of thyroid function and cognition in a gender-stratified manner. The mechanisms that might explain the reported gender-specific association are not clear yet. One possible explanation could be the effects of female gonadal hormones on the thyroid function. It is known that the thyroid gland has several estrogen receptors and therefore estrogen might have direct effects on thyroid function. Furthermore, animal studies reported that estrogen influences the iodide uptake as well as thyroid peroxidase uptake [31]. Both mechanisms are important for the thyroid hormone synthesis and might have an increasing effect.
Proposed mechanisms that might explain the association of TSH concentration with AD include a direct effect of thyroid hormones on the gene expression of the amyloid precursor protein that has been shown in animal studies [32]. AD is characterized by pathological changes that start decades before the onset of clinical symptoms and amyloid pathology is known to be the first detectable change [33]. Thus, the postulated influence of thyroid function on amyloid deposition might also explain the association of TSH and MCI. More studies are needed to further investigate this, especially with focus on low and high TSH concentrations within the reference range.
There are a few limitations, namely the cross-sectional study design. It is not possible to conclude that a high TSH concentration in the reference range actually forecasts the risk of MCI but future analysis of the longitudinal data will allow us to specify this relationship. Another limitation is that TSH measurement was carried out only once. TSH concentration can be affected by different reasons, for example by the general health condition or intake of drugs. Nevertheless, blood TSH concentration is still the best marker for thyroid function.
Our study has several strengths. Participants were randomly selected from mandatory registries which reduced the selection or volunteer bias. We excluded all participants with either only objective impairment or only subjective cognitive complaint resulting in a reference group of cognitively healthy participants. Several studies have reported an association between midlife cardiovascular risk factors and the development of cognitive impairment [34, 35]. But most population-based studies that investigated MCI and dementia are characterized by a relatively old study population. This is due to the fact that the risk of dementia is highly elevated after the age of 65. In our study, the age range from 50 to 80 years gives us the opportunity to investigate the association of cardiovascular risk factors not only in old age but also in middle age participants.
In conclusion, our study showed a strong association of high-normal TSH concentration to be associated with MCI in women, but not in men. This association persisted even after adjusting for age, education, APOE genotype, depression, and several cardiovascular risk factors. Our results suggest a need to monitor high-normal TSH concentrations in women with regard to a potentially elevated risk of cognitive decline but require further investigation in our planned longitudinal analysis.
Footnotes
ACKNOWLEDGMENTS
We acknowledge the support of the Sarstedt AG & Co. (Nümbrecht, Germany) concerning laboratory equipment. We thank Prof. K. Lauterbach (Adjunct Professor, Harvard School of Public Health, Boston, USA) for his valuable contributions in an earlier phase of the study. We are indebted to all study participants and to the dedicated personnel of both the study center of the Heinz Nixdorf Recall Study and the EBT-scanner facilities as well as to the investigative group, in particular to U Slomiany, EM Beck, A Öffner, S Münkel, M Bauer, S Schrader, R Peter, and H Hirche.
We thank the Heinz Nixdorf Foundation (Chairman: Martin Nixdorf; Past Chairman: Dr Jur Gerhard Schmidt [deceased]) for their generous support of this study. This study is also supported by the German Ministry of Education and Science (BMBF), and the German Aero-space Center (Deutsches Zentrum für Luft- und Raumfahrt (DLR)), Bonn, Germany. The German Research Council supported the study (DFG project: ER 155/6-2 and in the framework of SPP1629 THYROID TRANS ACT FU356/7-1) funded the assessment of psychosocial factors and neighborhood level information (DFG project SI 236/8-1 and SI 236/9-1) and funded the cognitive screening (DFG; SI236/10-1). For TSH assays, we received support of Roche Diagnostics (Mannheim, Germany).
