X Tang, Z-H Song et al. Spectrum of thyroid 10:4 410–421 dysfunction and dementia RESEARCH Spectrum of thyroid dysfunction and dementia: a dose–response meta-analysis of 344,248 individuals from cohort studies Xingyao Tang1,*, Zhi-Hui Song2,*, Dawei Wang3, Jinkui Yang4, Marly Augusto Cardoso5, Jian-Bo Zhou 4 and Rafael Simó6,7 1 Beijing Tongren Hospital, Capital Medical University, Beijing, China 2 Department of Pharmacy, Beijing Tongren Hospital, Capital Medical University, Beijing, China 3 General Practice Department, Beijing Tongren Hospital, Capital Medical University, Beijing, China 4 Department of Endocrinology, Beijing Tongren Hospital, Capital Medical University, Beijing, China 5 Department of Nutrition, School of Public Health, University of Sao Paulo, Sao Paulo, Brazil 6 Endocrinology and Nutrition Department, Hospital Universitari Vall d’Hebron, Diabetes and Metabolism Research Unit, Vall d’Hebron Institut de Recerca (VHIR), Universitat Autònoma de Barcelona, Barcelona, Spain 7 Centro de Investigación Biomédica en Red de Diabetes y Enfermedades Metabólicas Asociadas (CIBERDEM), Instituto de Salud Carlos III (ISCIII), Madrid, Spain Correspondence should be addressed to J-B Zhou: jbzhou@ccmu.edu.cn *(X Tang and Z-H Song contributed equally to this work) Abstract Thyroid hormone, as a modifiable risk factor for dementia, promotes neurocognitive Key Words function and regulates metabolic processes. Various studies have defined different thyroid- ff thyroid dysfunction stimulating hormone cutoffs, but the safest thyroid-stimulating hormone concentration ff dementia was absent. A dose–response meta-analysis describing the overall functional relation ff dose–response and identifying exposure intervals associated with a higher or lower disease risk is thus ff meta-analysis desirable. Therefore, our current analysis was conducted to understand the influence of thyroid dysfunction on dementia risk. We searched PubMed, Embase, and Web of Science before May 1, 2020 for human studies published in English. Studies were considered for inclusion if they used a cohort study design to measure the risk of dementia in different thyroid function status groups, diagnosed thyroid functional status and all-cause dementia, included participants aged >18 years, and provided quantitative measures of data. The analysis contained 17 articles with 344,248 individuals with a 7.8-year mean follow-up. Ten studies with 329,287 participants indicated that only subclinical hyperthyroidism was associated with an increased risk of dementia. In contrast, subclinical hypothyroidism, clinical hyperthyroidism, and clinical hypothyroidism did not affect dementia. In the dose– response meta-analysis with 46,417 samples from 11 studies, the association of thyroid- stimulating hormone with the risk of dementia exhibited a U-shaped curve. Our study indicated that subclinical hyperthyroidism was associated with the risk of dementia and the thyroid-stimulating hormone concentration at around 1.55–1.60 mU/L as the optimum Endocrine Connections range for the risk of dementia. (2021) 10, 410–421 https://ec.bioscientifica.com © 2021 The authors This work is licensed under a Creative Commons https://doi.org/10.1530/EC-21-0047 Published by Bioscientifica Ltd Attribution-NonCommercial 4.0 International License. X Tang, Z-H Song et al. Spectrum of thyroid 10:4 411 dysfunction and dementia Introduction Method Dementia is a progressive disease with declining Study inclusion cognitive function and affects daily activities (1). The For our analysis, we searched the PubMed and Embase global prevalence of dementia is 1.2–7.2%, and the databases and Web of Science for relevant studies number of dementia patients is estimated to increase published until May 1, 2020. All types of abnormal thyroid to 131.5 million by 2050 (2). Prevention of dementia function and dementia were searched, including clinical is crucial owing to the lack of disease-modifying hypothyroidism, subclinical hypothyroidism, subclinical medication. Recent reports have declared that risk hyperthyroidism, clinical hyperthyroidism, abnormal factor modifications could prevent 30–35% of dementia thyrotropin (TSH), dementia, Alzheimer’s disease, and cases (3). vascular dementia. Reference lists of eligible articles were The thyroid hormone, as a modifiable risk factor searched for further pertinent articles. The screening was for dementia, promotes neurocognitive function and completed independently by two reviewers (X Y T, J B Z), regulation metabolic processes (4). Thyroid dysfunction, and discrepancies, if any, were resolved with a discussion. including clinical hyperthyroidism, subclinical This report was conducted according to the Meta-analysis hyperthyroidism, subclinical hypothyroidism, and of Observational Studies in Epidemiology (16) and the clinical hypothyroidism, has been associated with an Preferred Reporting Items for Systematic Reviews and increased risk of cognitive impairment in many studies Meta-Analysis (17). (5). A recent meta-analysis indicated a positive association Subclinical hypothyroidism was defined as high between subclinical hyperthyroidism and an increased TSH concentrations but normal blood concentrations risk of dementia among prospective cohorts, while no of thyroid hormones. Subclinical hyperthyroidism such relationship existed in subclinical hypothyroidism was defined as low TSH concentrations but normal (5). New evidence has emerged recently (6, 7, 8, 9), blood concentrations of thyroid hormones. Clinical involving the impact of clinical hyperthyroidism and hypothyroidism was defined as high TSH concentrations clinical hypothyroidism on the risk of dementia in a large and low blood concentrations of thyroid hormones. sample size (7, 8). Clinical hyperthyroidism was defined as low TSH For the diagnosis of thyroid dysfunction, the thyroid- concentrations and high blood concentrations of stimulating hormone (TSH) (also known as thyrotropin) thyroid hormones. concentration was also measured as a risk factor for dementia (10, 11, 12, 13). Another meta-analysis combined 11 studies and found that TSH concentrations Inclusion and exclusion criteria below the normal range had an increased risk of dementia (14). However, other studies have defined different Studies were considered for inclusion if they (1) were an TSH cutoffs. Moreover, a lower, but average serum TSH original article published in English, (2) used a cohort concentration, also pertained to negative impacts on the study design to measure the risk of dementia in different development and progression of cognitive function (15). thyroid function status groups, (3) included participants Thus, a detailed analysis of the impact of specific TSH aged > 18 years, and (4) provided quantitative measures of concentrations on dementia is desirable. A dose–response data. Exclusion criteria followed were (1) the publication meta-analysis – aimed at describing the overall functional was a review, case report, animal study, or letter to the relation and identifying exposure intervals associated editor, (2) the reviews did not clearly define clinical with a higher or lower disease risk – is also desirable. outcomes, (3) the authors could not provide valid data Various studies have defined different thyroid- after being contacted, and (4) duplicated data. stimulating hormone cutoffs, but the safest thyroid- For the meta-analysis, studies were included if they stimulating hormone concentration was absent. It is measured the risk of dementia between euthyroid and desirable to confirm the relationship between thyroid clinical hypothyroidism or subclinical hypothyroidism or dysfunction and risk of dementia in prospective cohort subclinical hyperthyroidism or clinical hyperthyroidism. studies as well as to conduct a dose–response meta- In the dose–response meta-analysis, studies with at least analysis using published findings to understand the three categories of TSH concentrations and the number influence of TSH concentration change, even within the of dementia cases presented, according to the different normal range, on the risk of dementia. categories of TSH concentrations, were retained. https://ec.bioscientifica.com © 2021 The authors This work is licensed under a Creative Commons https://doi.org/10.1530/EC-21-0047 Published by Bioscientifica Ltd Attribution-NonCommercial 4.0 International License. X Tang, Z-H Song et al. Spectrum of thyroid 10:4 412 dysfunction and dementia Data extraction and quality assessment Sensitivity analysis was used to identify more potential resource of heterogeneity. Two investigators (X Y T, M A C) independently extracted In the dose–response meta-analysis, we changed the the following data from the 17 enrolled studies using the reference group to the lowest dose group. The median same standardized protocol: concerning study design, concentration of TSH was chosen as the TSH category, and population characteristics, follow-up years, TSH cutoff the number of dementia cases and the total participants concentrations as well as outcomes. Two investigators were extracted. For the studies that did not present the (J B Z, M A C) independently utilized the Newcastle– median or mean doses of TSH, the midpoint of each Ottawa Quality Assessment Scale criteria (NOS) (18) to category was chosen. The value for the open-ended upper assess the risk of bias for the enrolled studies. We rated the interval was arbitrarily assigned as 1.5 times that of the quality of the studies by awarding stars in each domain lower end of the interval. The aggregate generalized following the guidelines of NOS. Discrepancies, if any, least squares for trend (glst) method by Greenland and were resolved with a discussion between the investigators Longnecker was used for assessing the dose–response and other authors. relationship between TSH and the risk of dementia (19). We used restricted cubic splines with three knots to explore a potential non-linear association between TSH Statistical analysis concentrations and the risk of dementia. We used STATA (version 12.0, Stata Corporation, College Potential publication bias was evaluated using Egger’s Station, TX, USA) to conduct the current analysis. asymmetry test and with a visual inspection of a funnel Heterogeneity between studies was evaluated using plot (20). P-values were two-tailed, and P < 0.05 was the I2 metric and the variance between studies using considered significant. Tau2. Random-effects models were performed if I2 > 50% to determine the association between different thyroid function status and the risk of dementia. Random-effects models provide more weight to smaller studies. They Results typically have wider CIs because the total effect is the Literature search outcomes and average value of the real effect of each study, which not validity assessment only focuses on the study of a large sample but also pays attention to all included studies to balance the effect The search strategy identified 2957 potentially relevant of each study. Fixed-effects models were performed if records, of which 1016 were excluded as duplicates. I2 ≤ 50%. The remaining 1941 manuscripts were subjected to title Risk ratios (RR) as a measure of association across all and abstract screening. Further, 1765 publications were studies were pooled. If studies used a hazard ratio (HR), we excluded, as they were reviews, letters or conference regarded it as RR. RRs with a 95% CI were used to assess abstracts, and independent studies. Therefore, 176 articles time-to-event outcomes in the cohort studies, from the were eligible for full-text review and data assessment. data obtained directly from the articles. If studies had Finally, 159 articles were excluded owing to inadequate both unadjusted and covariate-adjusted odds ratios, we information or lack of a complete publication. A total of chose the one that adjusted for the maximum number 17 cohort studies (6, 7, 8, 9, 10, 11, 12, 13, 15, 21, 22, 23, of covariates. We calculated RRs and 95% CIs using 24, 25, 26, 27, 28) published from 2000 to 2020 met the data from some studies that provided only the number inclusion criteria (Fig. 1). Table 1 provides an overview of dementia cases in both abnormal thyroid status and of the eligible studies. All 17 studies used a longitudinal control groups. We performed a sensitivity analysis to cohort design. The total number of participants was examine the robustness of our findings by excluding one 344,248, with a mean follow-up duration of 7.8 years study at a time and recalculating the pooled estimates. (2.1–21.9 years). The mean age of the participants was Further sub-group analysis was conducted owing to 65.1 years, and 50.1% were women. Ten studies (6, 7, the different types of thyroid dysfunction, including 8, 10, 13, 21, 24, 25, 27, 28) were enrolled in the meta- subclinical hyperthyroidism, subclinical hypothyroidism, analysis and eleven studies (6, 7, 10, 11, 13, 21, 24, 25, 26, clinical hypothyroidism, and clinical hyperthyroidism. 27, 28) were enrolled in the dose–response meta-analysis. Further analysis was performed only for those who had a Nine of them are present in both the analyses (6, 7, 10, diagnosis of Alzheimer’s disease, a sub-type of dementia. 13, 21, 24, 25, 27, 28). Five other studies (9, 12, 15, 22, 23) https://ec.bioscientifica.com © 2021 The authors This work is licensed under a Creative Commons https://doi.org/10.1530/EC-21-0047 Published by Bioscientifica Ltd Attribution-NonCommercial 4.0 International License. X Tang, Z-H Song et al. Spectrum of thyroid 10:4 413 dysfunction and dementia Figure 1 PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) study. were not included in the meta-analysis or dose–response eight (6, 7, 10, 13, 21, 24, 27, 28) mentioned subclinical meta-analysis owing to irrelevant data. hypothyroidism, while two (7, 8) mentioned clinical hyperthyroidism, and one (7) mentioned clinical hypothyroidism. Supplementary Table 2 (see section on Quality assessment supplementary materials given at the end of this article) Quality assessment of the studies revealed scores of 5–9 on provides the meta-analysis data. NOS. One study (28) was scored 5, another (9) was scored The pooled effect value of the relationship between 6, four (10, 22, 23, 27) were scored 7, four (13, 15, 21, thyroid status and the risk of dementia was not significant 24) were scored 8, and seven (6, 7, 8, 11, 12, 25, 26) were (RR = 1.14, 95% CI: 0.97–1.35). Further sub-group analysis scored 9. High-quality studies were included (Table 1). exhibited that subclinical hyperthyroidism increased the risk of dementia (RR = 1.39, 95% CI: 1.08–1.79), while subclinical hypothyroidism, clinical hyperthyroidism, Meta-analysis of the association between and clinical hypothyroidism had no effect (RR = 0.96, 95% abnormal thyroid function status and the risk CI: 0.75–1.22, RR = 1.06, 95% CI: 0.63–1.79 and RR = 0.96, of dementia 95% CI: 0.73–1.26, respectively) (Fig. 2). Ten studies (6, 7, 8, 10, 13, 21, 24, 25, 27, 28) with a Three studies (10, 13, 24), with 3410 individuals, sample size of 329,187 individuals provided effective further mentioned subclinical thyroid dysfunction data and were included in the current meta-analysis affecting the risk of Alzheimer’s disease. The results showed (I2 = 66.1%, P = 0.000). Four different abnormal thyroid no association between subclinical thyroid dysfunction, function statuses were assessed: clinical hypothyroidism, including subclinical hyperthyroidism (RR = 1.52, 95% CI: subclinical hypothyroidism, subclinical hyperthyroidism, 0.69–3.32) and subclinical hypothyroidism (RR = 1.15, and clinical hyperthyroidism. Nine studies (6, 7, 8, 13, 21, 0.68–1.94), and the risk of Alzheimer’s disease (RR = 1.25, 24, 25, 27, 28) mentioned subclinical hyperthyroidism, 95% CI: 0.81–1.94) (Fig. 1). https://ec.bioscientifica.com © 2021 The authors This work is licensed under a Creative Commons https://doi.org/10.1530/EC-21-0047 Published by Bioscientifica Ltd Attribution-NonCommercial 4.0 International License. Table 1 Characteristics of studies included. Abnormal thyroid status TSH cutoff Study Year Country Sample size Women, % Mean age (s.d.) Follow-up (years) definition level (mU/L) Results NOS Kalmijin 2000 Netherlands 1843 61.90 68.8 (7.5) 2.1 Subclinical hypothyroidism: 0.4–4.0 Subclinical hyperthyroidism 8 TSH > 4 mU/L, FT4 in the elderly increases 11–25 pmol/L; the risk of dementia and subclinical hyperthyroidism: Alzheimer’s disease. https://ec.bioscientifica.com TSH < 0.4 mU/L, FT4 11–25 pmol/L https://doi.org/10.1530/EC-21-0047 Annerbo 2006 Sweden 93 51.60 64.7 (9.2) 6 / 0.2–4.0 Lower TSH levels were 7 found to be predictive factors of Alzheimer’s disease. de Jong 2006 Netherlands 1077 51.20 72.3 (7.4) 5.5 Subclinical hypothyroidism: 0.4–4.3 TSH and thyroid hormones 8 TSH > 4.3 mU/L, FT4 were not associated with 0.85–1.94 ng/dL, T3 the risk of dementia or 92.8–162.9 ng/dL; Alzheimer’s disease. subclinical hyperthyroidism: TSH < 0.4 mU/L, FT4 0.85–1.94 ng/dL, T3 92.8–162.9 ng/dL X Tang, Z-H Song et al. Tan 2008 USA 1864 59.00 71 (7) 12.7 / 0.5–5.0 Low and high thyrotropin 9 levels were associated © 2021 The authors Published by Bioscientifica Ltd with an increased risk of incident Alzheimer’s disease in women but not in men. Annerbo 2009 Sweden 200 79.50 81.0 (4.6) 6.7 / 0.2–4.0 TSH is not involved in the 7 development of Alzheimer’s disease. de Jong 2009 Hawaii 665 0 78.1 (5.7) 4.7 Subclinical hypothyroidism: 0.4–4.3 No associations were 7 TSH > 4.3 mU/L, FT4 found for TSH and Spectrum of thyroid 0.85–1.94 ng/dL; dementia or Alzheimer’s subclinical hyperthyroidism: disease. TSH < 0.4 mU/L, FT4 dysfunction and dementia 0.85–1.94 ng/dL Vadivelo 2011 UK 12115 77.40 66.2 (16.4) 5.6 Subclinical 0.4–4.0 Subclinical hyperthyroidism 9 hyperthyroidism: TSH ≤ patients were associated 0.4 mU/L, FT4 with an increased risk of 10–25 pmol/L or total T4 dementia. 65–155 nmol/L and total T3 0.9–2.6 nmol/L 10:4 Forti 2012 Italy 660 52.90 73.3 (6.0) 3.8 Subclinical hypothyroidism: 0.45–4.5 Baseline TSH was not 7 TSH > 4.50 mU/L, FT4 related to the risk of 10.3–25.7 pmol/L developing Alzheimer’s disease, but high TSH was associated with an This work is licensed under a Creative Commons increased risk of vascular 414 dementia. Attribution-NonCommercial 4.0 International License. (Continued) Table 1 Continued. Abnormal thyroid status TSH cutoff Study Year Country Sample size Women, % Mean age (s.d.) Follow-up (years) definition level (mU/L) Results NOS Yeap 2012 Australia 3401 0 76.8 (3.5) 5.9 Subclinical hypothyroidism: 0.4–4.0 There was no association 8 TSH > 4.0 mU/L, between TSH quartiles FT4 ≥ 10 pmol/L and incident dementia. Formiga 2014 Spain 307 54.60 85.0 (0.0) 3 Subclinical hypothyroidism: 0.25–5.0 Subclinical hypo- and 5 https://ec.bioscientifica.com TSH > 5.0 mU/L, FT4 hyperthyroidism were not 0.85–1.94 ng/dL; associated with an https://doi.org/10.1530/EC-21-0047 subclinical increased risk of hyperthyroidism: dementia. TSH < 0.25 mU/L, FT4 0.85–1.94 ng/dL Moon 2014 Korea 313 50 72.5 (6.9) 5 / 0.4–4.1 Lower serum TSH level 8 within the reference range was independently associated with the risk of dementia in elderly subjects. Cappola 2015 USA 2843 56.20 74.5 (5.1) 17 / 0.45–4.5 Individuals with TSH in the 9 fourth quartile had a 9.6 X Tang, Z-H Song et al. per 1000 person-year lower incidence of Published by Bioscientifica Ltd © 2021 The authors dementia. Chaker 2016 Netherlands 9446 56.70 64.9 (9.7) 8 / 0.4–4.0 High and high-normal 9 thyroid function is associated with increased dementia risk. Aubert 2017 USA 2558 51.80 75.1 (2.8) 9 Subclinical hypothyroidism: 0.45–4.49 Among older adults, 9 TSH > 4.50 mU/L, subclinical FT4 10.3–23.2 pmol/L; hyperthyroidism with a subclinical TSH < 0.10 mU/L was hyperthyroidism: associated with a higher Spectrum of thyroid TSH < 0.45 mU/L, risk of dementia. FT4 10.3–23.2 pmol/L dysfunction and dementia George 2019 USA 12481 56 57 (5.7) 21.9 Clinical hypothyroidism: 0.56–5.1 Subclinical hypothyroidism 9 TSH > 5.1 mU/L, was associated with a FT4 < 0.85 ng/dL; reduced risk of dementia, subclinical hypothyroidism: whereas overt TSH > 5.1 mU/L, hyperthyroidism, FT4 0.85–1.4 ng/dL; particularly very elevated subclinical FT4, was associated with 10:4 hyperthyroidism: an increased risk of TSH < 0.56 mU/L, dementia. FT4 0.85–1.4 ng/dL; clinical hyperthyroidism: TSH < 0.56 mU/L, FT4 > 1.4 ng/dL This work is licensed under a Creative Commons 415 Attribution-NonCommercial 4.0 International License. X Tang, Z-H Song et al. Spectrum of thyroid 10:4 416 dysfunction and dementia Sensitivity analysis 6 9 Sensitivity analyses demonstrated similar results, TSH did not associate with indicating that the overall results were not significantly compared to individuals influenced by any single study, with RRs and 95% CIs decreased TSH was associated with an dementia by 16%, ranging from 1.11 (0.94–1.31) to 1.23 (0.99–1.52) for the with normal TSH. Every 6 months of increased risk of risk of dementia. We limited the analysis to individuals aged >65 dementia. years (6, 10, 13, 21, 24, 25, 27, 28) and found similar results. Subclinical hyperthyroidism was significantly linked to the risk of dementia (RR = 1.51, 95% CI: 1.21–1.89), while subclinical hypothyroidism did 0.3–4.2 0.3–4.0 not affect dementia (RR = 1.03, 95% CI: 0.84–1.25). Furthermore, we limited the ranalysis to studies with NOS scores ≥ 7 (6, 7, 8, 10, 13, 21, 24, 25, 27). 135.0 nmol/L and/or T3 > Subclinical hyperthyroidism increased the risk of register cohort: nmol/L, T3 < 2.2 nmol/L; hyperthyroidism: TSH < clinical hyperthyroidism: dementia (RR = 1.40, 95% CI: 1.08–1.83), while TSH < 0.3 mU/L, T4 > 0.3 mU/L, T4 < 135.0 subclinical hypothyroidism was not a risk factor for dementia (RR = 0.91, 95% CI: 0.71–1.44). The effect of clinical thyroid dysfunction on dementia 2.2 nmol/L Subclinical FT3, free triiodothyronine; FT4, free thyroxine; NOS, score of Newcastle–Ottawa Scale; TSH, thyroid-stimulating hormone. remains unchanged. / DNPR cohort: Dose–response meta-analysis of the association OPENTHYRO between TSH concentrations and the risk 2.8 of dementia 7.2 7.3 As most primary studies analyzed reported different DNPR cohort: cutoffs of TSH, we further performed a dose–response OPENTHYRO 65.0 (7.8) 65 (54–75) 62 (49–74) meta-analysis. A total of 11 studies (6, 7, 10, 11, 13, 21, register cohort: 24, 25, 26, 27, 28) containing 46,417 individuals with more than three categories of TSH concentrations were analyzed. Supplementary Table 3 provides data of the dose–response meta-analysis. 54 / A non-linear relationship was observed between the TSH concentrations and the risk of dementia. The graph OPENTHYRO showed that both low and high TSH concentrations had adverse effects, which indicated that not only 302 National register registry 280640 cohort: cohort: (DNPR) Patient Danish 13440 hyperthyroidism but also hypothyroidism might lead to dementia. Using a restricted cubic spline model, we found that 1.55–1.60 mU/L was the safest concentration of TSH for preventing dementia. Lower or higher TSH 2020 Denmark 2019 Sweden concentrations increased the risk of dementia, even TSH concentrations were in the upper and lower limits of normal range. The most and second harmful TSH concentration for the risk of dementia was measured as lower than 0.05 (RR = 1.46, 95% CI: 1.31–1.64) Folkestad mU/L and higher than 5.39 mU/L (RR = 1.44, 95% CI: Quinlan 1.22–1.69), respectively. We added age and follow-up years as covariates, and the results showed no change (Fig. 3). https://ec.bioscientifica.com © 2021 The authors This work is licensed under a Creative Commons https://doi.org/10.1530/EC-21-0047 Published by Bioscientifica Ltd Attribution-NonCommercial 4.0 International License. X Tang, Z-H Song et al. Spectrum of thyroid 10:4 417 dysfunction and dementia Figure 2 The association between thyroid dysfunction and dementia. RR, ratio risk. Where I2 is the variation in effect estimates attributable to heterogeneity, overall is the pooled random effect estimate of all studies. Subtotal is the pooled random-effects estimate of sub-group analysis studies. Weights are from the random-effects analysis. %Weight is the weight assigned to each study, based on the inverse of the within- and between-study variance. The size of the gray boxes around the point estimates reflect the weight assigned to each study. Publication bias In contrast, subclinical hypothyroidism, clinical hyperthyroidism, and clinical hypothyroidism did not According to the Cochrane Handbook version (29), as a affect the risk of dementia. Furthermore, the optimum rule of thumb, tests for funnel plot asymmetry should concentration of TSH was 1.55–1.60 mU/L is. Both lower be used only when an adequate number of studies are and higher concentrations of TSH were associated with included in the meta-analysis, as with very few studies, the increased risk of dementia, even TSH concentrations the power of the tests is extremely low to distinguish were in the upper and lower limits of normal range. chance from real asymmetry. In this study, the P-value of Egger’s test was > 0.05 (P = 0.225), for the relationship between thyroid function and dementia, indicating no Comparison with other studies significant bias among them. The funnel figure of these studies showed asymmetrical inverted distribution, which The current analysis contained entirely of cohort is consistent with the results of Egger’s test (Fig. 2). studies and demonstrated subclinical hyperthyroidism as the sole risk factor for dementia. Two cross-sectional studies with 2072 (30) and 119 (31) samples indicated a similar association between subclinical hyperthyroidism Discussion and dementia. Our findings were in line with a recent The current analysis comprised 17 cohort studies, cross-sectional study, with 918 participants, which including 4 different thyroid dysfunctions (clinical demonstrated that subclinical hypothyroidism did not hypothyroidism, subclinical hypothyroidism, subclinical affect dementia (32). Our results verified the findings of a hyperthyroidism, and clinical hyperthyroidism) and meta-analysis conducted in 2016 after adding three new different TSH concentration range. In the meta-analysis, reports (6, 7, 8). Compared with the former meta-analysis, 10 studies with 329,287 thirty participants indicated that we further analyzed the relationship between clinical only subclinical hyperthyroidism was associated with hypothyroidism and clinical hyperthyroidism with increased risk of dementia (RR = 1.39, 95% CI: 1.08–1.79). dementia and conducted a dose–response analysis aimed https://ec.bioscientifica.com © 2021 The authors This work is licensed under a Creative Commons https://doi.org/10.1530/EC-21-0047 Published by Bioscientifica Ltd Attribution-NonCommercial 4.0 International License. X Tang, Z-H Song et al. Spectrum of thyroid 10:4 418 dysfunction and dementia at determining the effects of different TSH concentrations even within a normal range. A recent meta-analysis on the risk of dementia. declared TSH concentrations below the normal range Although in our meta-analysis, clinical thyroid as a risk factor for dementia, while TSH concentrations dysfunction had no effect on dementia, the evidence was within the normal range and above the normal range had limited and needed more reliable reports for confirmation. no significant influence on dementia (14). Moreover, the The difference between the effect of clinical and subclinical linear model showed a significant association between hyperthyroidism on dementia may be explained by the per s.d. increment of TSH and dementia (RR = 0.91, the lack of treatment and extended duration of thyroid 95% CI: 0.84–0.99). Some other studies paid attention to dysfunction in patients with subclinical hyperthyroidism the influence of TSH on dementia only with respect to (33). Individuals with subclinical hyperthyroidism are a linear relationship. One report (15) declared per unit less likely to receive treatment than those with clinical decrease of TSH within the normal range (0.4–4.1 mU/L) hyperthyroidism, which may influence dementia. In increased the risk of dementia (HR = 0.583, 95% CI: the current meta-analysis, one study (7) showed the 0.403–0.843), while another (12) found no link of per relationship between clinical hypothyroidism and clinical TSH unit increase with dementia (HR = 0.88, 95% CI: hyperthyroidism with dementia, and we chose the model 0.76–1.01). Our analysis further clarified the effect of that adjusted for the baseline thyroid medication use. different TSH concentrations on dementia in the non- Another study (8) contained the relationship between linear model and found a U-shaped relationship. The clinical hyperthyroidism and dementia, but the treatment normal range of TSH is widely considered as 0.5–4.5 mU/L, methods were not mentioned. Thus, further analysis and the dose range with the lowest risk of dementia was could not be carried out. As an important laboratory 1.55–1.60 mU/L in the current analysis, which is closer indicator of clinical hyperthyroidism, free thyroxine to the lower limit of the normal range. However, the (FT4) is another important hormone related to dementia, different effects of subclinical hypothyroidism and higher and a former meta-analysis found that per standard TSH concentrations on dementia are unclear and need deviation (s.d.) increment of FT4 was associated with the additional evidence to elucidate. increased risk of dementia (RR = 1.08, 95% CI: 1.00–1.17) (14). Owing to the limited evidence of the relationship Potential mechanisms between dementia and clinical hyperthyroidism, we did not include FT4 concentrations for further analysis. Thyroid dysfunction and dementia may be linked Because of the different definitions of TSH cutoff through potential pathways. The thyroid hormone concentrations, we performed a dose–response meta- has been suggested to promote neuroprotection and analysis of the relationship between TSH concentrations regulate neurogenesis (34). However, long-term exposure and dementia. The results suggested the optimal cutoff of of the thyroid hormone can lead to cardiovascular and TSH to be 1.55–1.60 mU/L. Compared with 1.55–1.60 mU/L cerebrovascular diseases such as fibrillation and systolic of TSH, both lower than 1.55 mU/L TSH and higher hypertension, which are both associated with the risk of than 1.60 mU/L TSH increased the risk of dementia, dementia (35). An alternate explanation is to elucidate the relationship between the thyroid hormone and the amount and phosphorylation of tau protein (36). High concentrations of the thyroid hormone may increase reactive oxygen species and reduce antioxidant enzymes, which may cause cellular damage and make the brain more vulnerable to amyloid toxicity (37). Finally, as a result of genetic susceptibility, excess thyroid hormone may alter gene expression in relevant pathways, such as neurogenesis (34). Public health impact Dementia is a growing clinical and socio-economic Figure 3 The dose–response association between TSH concentration and problem of aging societies. Recent estimates suggested dementia. TSH, thyroid-stimulating hormone; RR, ratio risk. more than 47 million individuals worldwide are affected https://ec.bioscientifica.com © 2021 The authors This work is licensed under a Creative Commons https://doi.org/10.1530/EC-21-0047 Published by Bioscientifica Ltd Attribution-NonCommercial 4.0 International License. X Tang, Z-H Song et al. Spectrum of thyroid 10:4 419 dysfunction and dementia by dementia (1). Thyroid dysfunction had an adverse studies is desirable. It may be valuable for further impact on many aspects, including diabetes, elevated studies to pay more attention to the safest TSH and FT4 blood pressure, metabolic syndrome, heart failure, and concentrations. cardiovascular disease (38). Regulation of thyroid function is beneficial to health. Our analyses demonstrated an Supplementary materials increased risk of lower and higher TSH concentrations This is linked to the online version of the paper at https://doi.org/10.1530/ on dementia, even within a normal range. The optimum EC-21-0047. concentration of TSH was suggested as 1.55–1.60 mU/L. With the increasing prevalence of thyroid dysfunction, it may be valuable for public health to focus on monitoring Declarations of interest The authors declare that there is no conflict of interest that could be TSH concentrations. perceived as prejudicing the impartiality of the research reported. Strengths and limitations of the study Funding This work was supported by the National Natural Science Foundation of Our meta-analysis and dose–response meta-analysis China (No. 82070851, 81870556, 81930019, 81670738), Beijing Municipal produced valuable information regarding associations Administration of Hospital’s Youth Program (QML20170204), Excellent between thyroid dysfunction and TSH concentrations Talents in Dongcheng District of Beijing. with the risk of dementia. Some desirable strengths and limitations must be outlined. The strengths of our analysis Acknowledgements included choosing data that adjusted for the maximum The authors are solely responsible for the design and conduct of this study, number of covariates, developing a comprehensive search all study analyses, the drafting and editing of the manuscript, and its final strategy, and conducting a comprehensive search. contents. There were also a few limitations. (1) Studies focused on the relationship of clinical hyperthyroidism and clinical hypothyroidism with the risk of dementia are References few. Additional reliable evidence is needed to clarify the 1 Livingston G, Sommerlad A, Orgeta V, Costafreda SG, Huntley J, relationships. (2) In the current study, the lack of data Ames D, Ballard C, Banerjee S, Burns A, Cohen-Mansfield J, et al. Dementia prevention, intervention, and care. Lancet 2017 390 on treatment and the effects of treatment on thyroid 2673–2734. (https://doi.org/10.1016/S0140-6736(17)31363-6) dysfunction limited a more rigorous analysis. Some reports 2 Vargas DM, De Bastiani MA, Zimmer ER & Klamt F. Alzheimer’s declared the impacts of antithyroid drugs on dementia disease master regulators analysis: search for potential molecular targets and drug repositioning candidates. Alzheimers Research (39). The influence of antithyroid drugs on dementia Therapy 2018 10 59. (https://doi.org/10.1186/s13195-018-0394-7) needs additional research. (3) Some other factors may 3 Norton S, Matthews FE, Barnes DE, Yaffe K & Brayne C. Potential for influence the concentration of TSH, such as pituitary- primary prevention of Alzheimer’s disease: an analysis of population- based data. Lancet: Neurology 2014 13 788–794. (https://doi. related disease (40). Further analysis was not performed org/10.1016/S1474-4422(14)70136-X) owing to the limited information. (4) The heterogeneity 4 Cheng SY, Leonard JL & Davis PJ. Molecular aspects of thyroid of results may be due to differences in race, method of TSH hormone actions. Endocrine Reviews 2010 31 139–170. (https://doi. org/10.1210/er.2009-0007) measurement, and the definition of thyroid dysfunction. 5 Rieben C, Segna D, da Costa BR, Collet TH, Chaker L, Aubert CE, Owing to inadequate information, it was difficult to Baumgartner C, Almeida OP, Hogervorst E, Trompet S, et al. analyze the heterogeneity further. Subclinical thyroid dysfunction and the risk of cognitive decline: a meta-analysis of prospective cohort studies. Journal of Clinical Endocrinology and Metabolism 2016 101 4945–4954. (https://doi. org/10.1210/jc.2016-2129) 6 Aubert CE, Bauer DC, da Costa BR, Feller M, Rieben C, Simonsick EM, Conclusion Yaffe K, Rodondi N & Health ABC Study. The association between subclinical thyroid dysfunction and dementia: the Health, Aging and Our study indicated that subclinical hyperthyroidism Body Composition (Health ABC) Study. Clinical Endocrinology 2017 is associated with the risk of dementia, and the TSH 87 617–626. (https://doi.org/10.1111/cen.13458) 7 George KM, Lutsey PL, Selvin E, Palta P, Windham BG & Folsom AR. concentration at around 1.55–1.60 mU/L was the peak Association between thyroid dysfunction and incident dementia range for the risk of dementia. Thus, it is desirable for in the atherosclerosis risk in communities neurocognitive study. public health to manage TSH concentrations, especially in Journal of Endocrinology and Metabolism 2019 9 82–89. (https://doi. org/10.14740/jem588) those with a higher risk of dementia. However, validating 8 Folkestad L, Brandt F, Lillevang-Johansen M, Brix TH & Hegedüs L. and replicating these findings in more rigorously designed Graves’ disease and toxic nodular goiter, aggravated by duration https://ec.bioscientifica.com © 2021 The authors This work is licensed under a Creative Commons https://doi.org/10.1530/EC-21-0047 Published by Bioscientifica Ltd Attribution-NonCommercial 4.0 International License. X Tang, Z-H Song et al. Spectrum of thyroid 10:4 420 dysfunction and dementia of hyperthyroidism, are associated with Alzheimer’s and vascular 23 Annerbo S, Wahlund LO & Lökk J. The significance of thyroid- dementia: a registry-based long-term follow-up of two large cohorts. stimulating hormone and homocysteine in the development of Thyroid 2020 30 672–680. (https://doi.org/10.1089/thy.2019.0672) Alzheimer’s disease in mild cognitive impairment: a 6-year follow-up 9 Quinlan P, Horvath A, Wallin A & Svensson J. Low serum study. American Journal of Alzheimer’s Disease and Other Dementias concentration of free triiodothyronine (FT3) is associated with 2006 21 182–188. (https://doi.org/10.1177/1533317506289282) increased risk of Alzheimer’s disease. Psychoneuroendocrinology 2019 24 Kalmijn S, Mehta KM, Pols HA, Hofman A, Drexhage HA & 99 112–119. (https://doi.org/10.1016/j.psyneuen.2018.09.002) Breteler MM. Subclinical hyperthyroidism and the risk of dementia. 10 Forti P, Olivelli V, Rietti E, Maltoni B, Pirazzoli G, Gatti R, Gioia MG The Rotterdam study. Clinical Endocrinology 2000 53 733–737. & Ravaglia G. Serum thyroid-stimulating hormone as a predictor (https://doi.org/10.1046/j.1365-2265.2000.01146.x) of cognitive impairment in an elderly cohort. Gerontology 2012 58 25 Vadiveloo T, Donnan PT, Cochrane L & Leese GP. The Thyroid 41–49. (https://doi.org/10.1159/000324522) Epidemiology, Audit, and Research Study (TEARS): morbidity in 11 Tan ZS, Beiser A, Vasan RS, Au R, Auerbach S, Kiel DP, Wolf PA & patients with endogenous subclinical hyperthyroidism. Journal of Seshadri S. Thyroid function and the risk of Alzheimer disease: the Clinical Endocrinology and Metabolism 2011 96 1344–1351. (https:// Framingham Study. Archives of Internal Medicine 2008 168 doi.org/10.1210/jc.2010-2693) 1514–1520. (https://doi.org/10.1001/archinte.168.14.1514) 26 Chaker L, Wolters FJ, Bos D, Korevaar TI, Hofman A, van der 12 Cappola AR, Arnold AM, Wulczyn K, Carlson M, Robbins J & Psaty BM. Lugt A, Koudstaal PJ, Franco OH, Dehghan A, Vernooij MW, Thyroid function in the euthyroid range and adverse outcomes in et al. Thyroid function and the risk of dementia: the Rotterdam older adults. Journal of Clinical Endocrinology and Metabolism 2015 100 Study. Neurology 2016 87 1688–1695. (https://doi.org/10.1212/ 1088–1096. (https://doi.org/10.1210/jc.2014-3586) WNL.0000000000003227) 13 de Jong FJ, den Heijer T, Visser TJ, de Rijke YB, Drexhage HA, 27 de Jong FJ, Masaki K, Chen H, Remaley AT, Breteler MM, Hofman A & Breteler MM. Thyroid hormones, dementia, and Petrovitch H, White LR & Launer LJ. Thyroid function, the risk atrophy of the medial temporal lobe. Journal of Clinical Endocrinology of dementia and neuropathologic changes: the Honolulu-Asia and Metabolism 2006 91 2569–2573. (https://doi.org/10.1210/ aging study. Neurobiology of Aging 2009 30 600–606. (https://doi. jc.2006-0449) org/10.1016/j.neurobiolaging.2007.07.019) 14 Wu Y, Pei Y, Wang F, Xu D & Cui W. Higher FT4 or TSH below the 28 Formiga F, Ferrer A, Padros G, Contra A, Corbella X, Pujol R & normal range are associated with increased risk of dementia: a meta- Octabaix Study Group. Thyroid status and functional and cognitive analysis of 11 studies. Scientific Reports 2016 6 31975. (https://doi. status at baseline and survival after 3 years of follow-up: the org/10.1038/srep31975) OCTABAIX study. European Journal of Endocrinology 2014 170 69–75. 15 Moon JH, Park YJ, Kim TH, Han JW, Choi SH, Lim S, Park DJ, (https://doi.org/10.1530/EJE-13-0722) Kim KW & Jang HC. Lower-but-normal serum TSH level is associated 29 Higgins JPT, Thomas J, Chandler J, Cumpston M & Welch VA. with the development or progression of cognitive impairment in Cochrane Handbook for Systematic Reviews of Interventions. elderly: Korean Longitudinal Study on Health and Aging (KLoSHA). 30 Benseñor IM, Lotufo PA, Menezes PR & Scazufca M. Subclinical Journal of Clinical Endocrinology and Metabolism 2014 99 424–432. hyperthyroidism and dementia: the Sao Paulo Ageing and Health (https://doi.org/10.1210/jc.2013-3385) Study (SPAH). BMC Public Health 2010 10 298. (https://doi. 16 Stroup DF, Berlin JA, Morton SC, Olkin I, Williamson GD, Rennie D, org/10.1186/1471-2458-10-298) Moher D, Becker BJ, Sipe TA & Thacker SB. Meta-analysis of 31 Döbert N, Hamscho N, Menzel C, Peters J, Frölich L, Tsolakis A, observational studies in epidemiology: a proposal for reporting. Zaplatnikov K, Kratzsch T, Diener J, Maurer K, et al. Subclinical Meta-analysis Of Observational Studies in Epidemiology (MOOSE) hyperthyroidism in dementia and correlation of the metabolic index group. JAMA 2000 283 2008–2012. (https://doi.org/10.1001/ in FDG-PET. Acta Medica Austriaca 2003 30 130–133. jama.283.15.2008) 32 Park YJ, Lee EJ, Lee YJ, Choi SH, Park JH, Lee SB, Lim S, Lee WW, 17 Moher D, Liberati A, Tetzlaff J, Altman DG & PRISMA Group. Jang HC, Cho BY, et al. Subclinical hypothyroidism (SCH) is not Preferred reporting items for systematic reviews and meta-analyses: associated with metabolic derangement, cognitive impairment, the PRISMA statement. PLoS Medicine 2009 6 e1000097. (https://doi. depression or poor quality of life (QoL) in elderly subjects. org/10.1371/journal.pmed.1000097) Archives of Gerontology and Geriatrics 2010 50 e68–e73. (https://doi. 18 Stang A. Critical evaluation of the Newcastle-Ottawa scale for the org/10.1016/j.archger.2009.05.015) assessment of the quality of nonrandomized studies in meta- 33 Lillevang-Johansen M, Abrahamsen B, Jørgensen HL, Brix TH analyses. European Journal of Epidemiology 2010 25 603–605. (https:// & Hegedüs L. Excess mortality in treated and untreated doi.org/10.1007/s10654-010-9491-z) hyperthyroidism is related to cumulative periods of low serum TSH. 19 Greenland S & Longnecker MP. Methods for trend estimation from Journal of Clinical Endocrinology and Metabolism 2017 102 2301–2309. summarized dose-response data, with applications to meta-analysis. (https://doi.org/10.1210/jc.2017-00166) American Journal of Epidemiology 1992 135 1301–1309. (https://doi. 34 Kapoor R, Fanibunda SE, Desouza LA, Guha SK & Vaidya VA. org/10.1093/oxfordjournals.aje.a116237) Perspectives on thyroid hormone action in adult neurogenesis. 20 Egger M, Davey Smith G, Schneider M & Minder C. Bias in meta- Journal of Neurochemistry 2015 133 599–616. (https://doi. analysis detected by a simple, graphical test. BMJ 1997 315 629–634. org/10.1111/jnc.13093) (https://doi.org/10.1136/bmj.315.7109.629) 35 Reddy V, Taha W, Kundumadam S & Khan M. Atrial fibrillation and 21 Yeap BB, Alfonso H, Chubb SA, Puri G, Hankey GJ, Flicker L & hyperthyroidism: a literature review. Indian Heart Journal 2017 69 Almeida OP. Higher free thyroxine levels predict increased incidence 545–550. (https://doi.org/10.1016/j.ihj.2017.07.004) of dementia in older men: the Health in Men Study. Journal of 36 Oyanagi K, Negishi T & Tashiro T. Action of thyroxine on the Clinical Endocrinology and Metabolism 2012 97 E2230–E2237. (https:// survival and neurite maintenance of cerebellar granule neurons in doi.org/10.1210/jc.2012-2108) culture. Journal of Neuroscience Research 2015 93 592–603. (https:// 22 Annerbo S, Kivipelto M & Lokk J. A prospective study on the doi.org/10.1002/jnr.23519) development of Alzheimer’s disease with regard to thyroid- 37 Aslan M, Cosar N, Celik H, Aksoy N, Dulger AC, Begenik H, stimulating hormone and homocysteine. Dementia and Soyoral YU, Kucukoglu ME & Selek S. Evaluation of oxidative status Geriatric Cognitive Disorders 2009 28 275–280. (https://doi. in patients with hyperthyroidism. Endocrine 2011 40 285–289. org/10.1159/000242439) (https://doi.org/10.1007/s12020-011-9472-3) https://ec.bioscientifica.com © 2021 The authors This work is licensed under a Creative Commons https://doi.org/10.1530/EC-21-0047 Published by Bioscientifica Ltd Attribution-NonCommercial 4.0 International License. X Tang, Z-H Song et al. Spectrum of thyroid 10:4 421 dysfunction and dementia 38 Kannan L, Shaw PA, Morley MP, Brandimarto J, Fang JC, Sweitzer NK, fibrillation. Aging Clinical and Experimental Research 2020 32 433–439. Cappola TP & Cappola AR. Thyroid dysfunction in heart failure and (https://doi.org/10.1007/s40520-019-01217-3) cardiovascular outcomes. Circulation: Heart Failure 2018 11 e005266. 40 Yamamoto M, Iguchi G, Bando H, Kanie K, Hidaka-Takeno R, (https://doi.org/10.1161/CIRCHEARTFAILURE.118.005266) Fukuoka H & Takahashi Y. Autoimmune pituitary disease: new 39 Wändell P, Carlsson AC, Li X, Sundquist J & Sundquist K. concepts with clinical implications. Endocrine Reviews 2020 41 Levothyroxine treatment and incident dementia in adults with atrial bnz003. (https://doi.org/10.1210/endrev/bnz003) Received in final form 1 March 2021 Accepted 19 March 2021 Accepted Manuscript published online 19 March 2021 https://ec.bioscientifica.com © 2021 The authors This work is licensed under a Creative Commons https://doi.org/10.1530/EC-21-0047 Published by Bioscientifica Ltd Attribution-NonCommercial 4.0 International License.