Thyroid Status and Brain Circulation: The Rotterdam Study.

CONTEXT: Whether thyroid dysfunction is related to altered brain circulation in the general population remains unknown.
OBJECTIVE: We determined the association of thyroid hormones with different markers of brain circulation within community-dwelling elderly people.
METHODS: This was a population-based study of 3 subcohorts of the Rotterdam Study, starting in 1989, 2000, and 2006. A total of 5142 participants (mean age, 63.8 years; 55.4% women), underwent venipuncture to measure serum thyroid-stimulating hormone (TSH) and free thyroxine (FT4). Between 2005 and 2015, all participants underwent phase-contrast brain magnetic resonance imaging to assess global brain perfusion (mL of blood flow/100 mL of brain/minute). Arteriolar retinal calibers were assessed using digitized images of stereoscopic fundus color transparencies in 3105 participants as markers of microcirculation. We investigated associations of TSH, FT4 with brain circulation measures using (non)linear regression models.
RESULTS: FT4 (in pmol/L) levels had an inverse U-shaped association with global brain perfusion, such that high and low levels of FT4 were associated with lower global brain perfusion than middle levels of FT4. The difference in global brain perfusion between high FT4 levels (25 pmol/L) and middle FT4 levels (FT4 = 15 pmol/L; P nonlinearity = .002) was up to -2.44 mL (95% CI -4.31; -0.56). Higher and lower levels of FT4, compared with middle FT4 levels, were associated with arteriolar retinal vessels (mean difference up to -2.46 µm, 95% CI -4.98; 0.05 for lower FT4).
CONCLUSION: These results suggest that thyroid dysfunction could lead to brain diseases such as stroke or dementia through suboptimal brain circulation that is potentially modifiable.

Thyroid dysfunction is common in the general population, yet underdiagnosed, and associated with an increased risk of stroke and dementia (1-3). Although some previous studies addressed these associations, the mechanism through which thyroid dysfunction causes dementia or stroke remains unclear. Low brain perfusion is implicated in dementia (4) and in the pathophysiology of acute transient ischemic attack (TIA) and stroke (5). There is growing evidence that persons with thyroid dysfunction have an altered brain circulation, which in turn could lead to an increased risk of cerebrovascular and neurodegenerative disease (6). For example, several previous studies have shown that in patients with Alzheimer’s disease, higher serum levels of thyroid-stimulating hormone (TSH) were correlated with lower brain perfusion in the temporal regions (7-9). In addition to hypothyroidism where reduced perfusion of the brain may occur through lower cardiac output (10), relative hypoperfusion could also occur with hyperthyroidism through increased metabolism which modifies brain oxygen utilization (11). When the vasodilatory response is exhausted, the brain becomes hypoperfused. Thus far, the relation between thyroid dysfunction and circulation of the brain has only been shown in specific patient populations, such as in patients with mood disorders and thyrotoxicosis (6, 10, 12-15). The role of thyroid status, even within the normal range, in brain circulation of individuals from the general population remains unknown. Changes in thyroid status might be important for the pathogenesis of stroke and dementia, and these diseases may thus be modifiable by altering thyroid function status (eg, treating thyroid dysfunction). Therefore, we examined the association between TSH, free thyroxine (FT4), and thyroid peroxidase antibodies (TPOs) and different indices of brain circulation. Specifically, the outcome measures included global brain perfusion as assessed by magnetic resonance imaging (MRI), and alterations in retinal vessel diameter, as a measure of brain microcirculation. In addition, we investigated the effect of thyroid hormones on cerebrovascular CO2 reactivity (CVR) as assessed by transcranial Doppler, using data from the population-based Rotterdam Study.

Materials and Methods

Study Population

This study was performed within the Rotterdam Study, a large population-based cohort initially comprising 7983 persons living in the well-defined Ommoord district in the city of Rotterdam in The Netherlands. There were no prespecified exclusion criteria, meaning that all persons older than 55 years of age living in the area were invited to participate. The study started with a pilot phase in the second half of 1989. From January 1990 onwards participants were recruited for the Rotterdam Study. In 2000, the study was extended with a second cohort of 3011 participants from the study district who had become 55 years since the start of the study and had not been invited to participate previously or those of 55 years or over who had migrated into the study district. A third cohort of 3932 participants, all aged 45 years and over living in the study district who had not been examined already (ie, mainly comprising those aged 45-60 years) was added in 2006. The study included in total 14 926 participants by the end of 2008 (response rate of 72%). For more information see Ikram et al. (16).

Thyroid function measurements were collected between 1997 and 2007 in 11 740 participants from these 3 independent Rotterdam Study cohorts (once per participant): the third visit of the first cohort, the first visit of the second cohort, and the first visit of the third cohort (17). Among the sample with complete thyroid function measurements, 5142 participants underwent 2D phase-contrast brain magnetic resonance imaging (MRI) between 2005 and 2015 to assess global brain perfusion. At first, MRI scans in the Rotterdam Study were only performed in a random sample from the second cohort, and then the invitation to participate in MRI data collection was extended to all eligible participants from the 3 cohorts (18). Of these participants, 3105 underwent arteriolar and venular retinal caliber assessment (19, 20). This assessment was performed in participants from the first, second, and third cohorts. Finally, a smaller sample of 901 participants underwent CVR measurements between 1997 and 2001 during the third visit of the first cohort and the first visit of the second cohort. CVR measurements were only offered to a random smaller sample of participants because of lack of technical support and personnel (21) (Fig. 1).

Figure 1.

Flow diagram.

The Rotterdam Study was approved by the Medical Ethics Committee of Erasmus University with registration number MEC 02.1015, and by the Ministry of Health, Welfare, and Sport of The Netherlands (Population Screening Act WBO, license number 1071272-159521-PG), implementing the Population Study Act Rotterdam Study. The Rotterdam Study was entered into The Netherlands National Trial Register (www.trialregister.nl) and into the World Health Organization International Clinical Trials Registry Platform (https://apps.who.int/trialsearch/) under shared catalogue number NTR6831. In accordance with the Declaration of Helsinki, all participants in this analysis provided written informed consent to participate in the study and to have their information obtained from treating physicians.

Assessment of Thyroid Function

TSH, FT4, and TPO were measured on serum samples stored at –80°C (electrochemiluminescence immunoassay “ECLIA”, Roche). The normal range values were defined for TSH as 0.4 to 4.0 mIU/L and for FT4 as 11 to 25 pmol/L (22) according to guidelines and previous research (23). TPO levels greater than 35 kU/mL were considered as positive according to manufacturer recommendations.

Assessment of Global Brain Perfusion

Global brain perfusion was assessed using a 1.5-Tesla MRI scanner (4, 18). Specifically for the flow measurements, we performed a 2D phase-contrast sequence as described before (18), a sequence that allows measuring blood flow velocities by phase. Blood flow was calculated using interactive data language-based custom software (Cinetool version 4, General Electric Healthcare, Milwaukee, WI, USA) from the 2D phase-contrast sequence images. Calculation of cerebral blood flow (CBF, in mL/minute) was made by summing the flow rates for the carotid arteries and the basilar artery (mL/second) and multiplying this by 60 seconds/minute (24). Global brain perfusion (mL/minute blood per 100 mL brain) was computed as the division of CBF by each individual’s brain volume (mL) and by multiplying the result by 100 (4).

Retinal Vessel Measurements

Retinal vessels were measured using simultaneous stereoscopic fundus color photographs centered on the optic disk (pharmacological mydriasis, 20°; Topcon Optical Co., Tokyo, Japan; digitized with a high-resolution scanner model LS-4000; Nikon Corp., Tokyo, Japan) (25). For each participant, the image of 1 eye with the best quality was studied with the Retinal Vessel Measurement System (Retinal Analysis, Optimate, WI; Department of Ophthalmology and Visual Science, University of Wisconsin Madison) (25). One summary value was calculated for the arteriolar diameters (in µm) and 1 for the venular diameters of the blood column for each participant (26, 27). Because of different magnification in case of refractive changes due to corneal curvature, lens, and axial length differences, we adjusted this summary vessel measure with Littmann’s formula to approximate absolute measures (27, 28).

Transcranial Doppler Assessment

Transcranial Doppler (TCD) monitoring (Multi-Dop X-4; DWL, Sipplingen, Germany) was used to measure the CBF velocity in the middle cerebral artery on both sides as described before (21). End diastolic, peak systolic, and mean CBF velocities (cm/second) were recorded automatically (29). End-tidal CO2 pressure (kPa) was recorded continuously with a CO2 analyzer (Multinex; Datascope, Hoevelaken, The Netherlands) (21). CVR was assessed by continuous measurement of flow velocity in the middle cerebral artery when participants breathed room air followed by 5% CO2 inspiration through an anesthetic mask for 2 minutes (21). We defined CVR as the percentage increase in flow velocity during inspiration of 5% CO2, divided by the absolute increase in end-tidal CO2 in the same period (21). We used the mean of right and left circulatory parameters for the analyses. In case of 1-sided window absence, the contralateral parameters were used for analyses (21).

Covariates

Smoking was assessed with a questionnaire and categorized as never or past vs current smoking. Alcohol consumption was measured with a questionnaire assessing the number of glasses of alcohol per day. Diet was self-reported using the food-frequency questionnaire, and a diet quality score was calculated as previously described (30). Measures of height and weight were collected at the research center and body mass index was defined as weight (kg)/height (m)². Blood pressure was measured twice using a sphygmomanometer and averaged. Blood pressure–lowering medication was assessed through a home interview and pharmacy records. Diabetes mellitus was defined as use of antidiabetic medication, fasting serum glucose ≥7.1 mmol/L (≥127.9 mg/dL), or random nonfasting serum glucose ≥11.1 mmol/L (≥200.0 mg/dL) (31). Physical activity was assessed using the validated LASA Physical Activity Questionnaire and Zutphen Physical Activity Questionnaire (32, 33). Creatinine was measured using blood samples with an enzymatic assay method.

Statistical Analyses

We analyzed the cross-sectional association of thyroid hormones (ie, TSH, FT4, and TPO) with measures of brain circulation, including global brain perfusion, arteriolar and venular retinal diameters, and CVR. Covariables were selected based on previous literature (1) (Figure S1 (34)). The first model was adjusted by sex, age, and cohort. For the analyses involving global brain perfusion, we also adjusted for time between thyroid assessment and MRI scan. We did not adjust for time difference when we explored the association between thyroid function and retinal vessels because the time difference was very small (median time difference: 0.03 months). Similarly, the association between thyroid function and CVR was not adjusted for time difference because these measures were collected during the same research visits. The second model was additionally adjusted for smoking, alcohol consumption, and diet quality score. In the third model, we additionally adjusted for covariables that could act as confounders but also as mediators, namely body mass index, blood pressure–lowering medication, diabetes mellitus, physical activity, and glomerular filtration rate. In a fourth model, we additionally adjusted for systolic and diastolic blood pressure as potential mediators. Models are presented in Figure S1 (34). TSH was studied using natural logarithmic transformation due to its right skewed distribution. TPO was studied continuously and dichotomously based on the cut-off (35 kU/mL). The possibility of a nonlinear association between the exposures (TSH, FT4, and TPO), the covariables, and our outcomes was tested by using restricted cubic splines with 3 knots. We further quantified nonlinear associations between thyroid hormones and measures of brain circulation by plotting these associations and visually assessing the thyroid hormone value that was associated with the highest point for brain circulation and comparing that with other thyroid hormone values and their corresponding brain circulation outcomes.

We performed the following sensitivity analyses. First, we restricted our analyses to euthyroid participants. Euthyroidism was defined based on TSH values between 0.4 and 4.0 mIU/L, and without thyroid medication and antithyroid medication drugs (eg, methimazole). Second, we performed our analyses in participants who had their thyroid and MRI assessment during the same visit scheme. The reason for this analysis was the long time between blood measurement and time of MRI scan in some participants. We did the same for retinal vessel diameter as outcome. We did not perform this sensitivity analysis for CVR as thyroid hormones and CVR were measured in the same visit. Third, we studied our association between thyroid function and global brain perfusion separately for those without cortical infarcts, without carotid stenosis (>50% in the left or right carotid artery) and additionally adjusting for the other thyroid hormones (eg, we included TSH in the analyses of FT4 and vice versa). Regarding antithrombotics, we have added antithrombotic agent use (ATC-code b01) to our main model (II).

We explored effect modification by sex, age, smoking, time between thyroid measurement and MRI scan, and finally each of the other brain circulation measures by including the product of the interacting factor and thyroid hormones and planned a stratified analysis if the P for interaction was <.1.

Multiple imputation by chained equations was performed for covariates with missing data. Statistical analyses were performed using R statistical software (mice, rms packages, R project, Institute for Statistics and Mathematics, R Core Team, version 3.2.2).

Results

We included a total of 5142 subjects with their characteristics presented in Table 1. The mean age was 63.8 years, and 55.4% were women. Median TSH was 2.0 mIU/L (interquartile range [IQR] 1.4, 2.8), mean FT4 was 15.6 pmol/L with SD 2.2, and 684 participants (13.3%) were TPO positive. Median global brain perfusion was 55.3 mL of blood flowing per 100 mL of brain/minute (IQR 49.7, 61.8). The mean arteriolar diameter was 156.7 µm (SD 15.8), venular diameter 238.2 µm (SD 22.6), and the median CVR was 39.2%/kPa (IQR 25.0, 56.0) (Table 1). The median time difference between thyroid function and MRI data collection was 2.7 months (IQR 0.7, 116.0) and the median time difference between thyroid function and retinal vessels data collection was of 0.03 months (IQR 0.00, 1.3). There was no time difference between thyroid function and CVR data collection because these measures were assessed at the same research visit.

Table 1.

Study population characteristics

Characteristic	N = 5142
Women	2849 (55.4%)
Age, years	63.8 ± 10.2
Smoking
Current	1059 (20.6%)
Former or never	4083 (79.4%)
Alcohol, glasses per day	1.2 ± 2.0
Diet, sum score adherence dietary guidelines (0-14)	6.9 ± 1.9
Diabetes mellitus type 2	261 (5.1%)
Body mass index, kg/m2	27.5 ± 4.1
Systolic blood pressure, mmHg	138.0 ± 20.1
Diastolic blood pressure, mmHg	81.3 ± 10.7
Blood pressure lowering medication	1585 (30.8%)
Total physical activity, MET·h·week, standardized	0.10 ± 1.0
Glomerular filtration rate, mL/min/1.73 m2	83.9 ± 14.4
Free thyroxine, pmol/L	15.6 ± 2.2
Thyroid-stimulating hormone, mIU/L	2.0 (1.4-2.8)
Thyroid peroxidase antibodies positivity(cutoff, 35 kU/mL)	684 (13.3%)
Outcomes
Global brain perfusion, mL/min/100 mL of brain, median (IQR)	55.3 (49.7-61. 8)
Arteriolar diameter, µm (N = 3105)	156.7 ± 15.8
Venular diameter, µm (N = 3105)	238.2 ± 22.6
CO2 vasoreactivity, %/kPa (N = 901), median (IQR)	39.2 (25.0-56.0)

Abbreviations: N indicates number of participants included in study, IQR indicates interquartile range. Data presented as mean ± standard deviation for continuous variables and number (%) for categorical variables, unless otherwise specified. Number of missing values: 43 (0.8%) for smoking, 144 (2.8%) for alcohol, 1129 (22.0%) for diet, 12 (0.2%) for diabetes mellitus type 2, 5 for body mass index (0.1%), 64 for systolic and diastolic blood pressure (0.8%), 69 for blood pressure lowering medication (1.3%), 829 for total physical activity (16.1%), 81 for glomerular filtration rate (1.6%).

Thyroid Function and Global Brain Perfusion

There was no statistically significant association between TSH and global brain perfusion (mean difference in mL perfusion per natural log increase in TSH = –0.22, 95% CI –0.55; 0.10; Fig. 2 and Table S1 (34)). On the other hand, we found that higher and lower FT4 levels (in pmol/L) were associated with lower global brain perfusion (inverted U-shape) between FT4 and global brain perfusion (P for nonlinearity = .002, P value for association = .024; Fig. 2), with on average 0.80 mL (95% CI –1.95; 0.35) less brain perfusion in persons with a lower FT4 serum level of 10 pmol/L than in persons with a middle FT4 of 15 pmol/L and a difference of –2.44 mL (95% CI –4.31; –0.56) when comparing persons with a higher FT4 of 25 pmol/L with FT4 of 15 pmol/L (Table S2 (34)). Finally, there was no statistically significant association between TPO positivity and global brain perfusion (mean difference in mL perfusion in TPO positives vs negatives = –0.25, 95% CI –1.00; 0.48).

Figure 2.

A visual representation of the associations between thyroid markers and brain perfusion measures. Models are corrected for age, sex, study cohort, smoking, alcohol, and diet. For global brain perfusion, we additionally adjusted for time between thyroid measurement and MRI scan. The associations of free thyroxine and global brain perfusion and arteriolar retinal vessel diameter were performed using restricted cubic splines (3 knots). P denotes P value of association. The continuous line represents the estimate of the association and the gray shadow represents the 95% CI.

Thyroid Function and Retinal Vessels

We found on average 0.92 µm narrower arteriolar retinal vessels (95% CI –1.64; –0.20) per natural log increase in TSH (Fig. 2 and Table S1 (34)). This association did not change materially after additional adjustment by systolic and diastolic blood pressure (mean difference = –0.79, 95% CI –1.48; –0.09, Table S1 (34)). There was an inverted U-shaped association between FT4 and arteriolar retinal vessels (P < .0001 for nonlinearity, P for association = .084; Fig. 2). Participants positive for TPO antibodies had on average 1.78 µm narrower arterioles (95% CI –3.33; –0.22) compared with persons negative for TPO antibodies. This association was attenuated after adjustment by systolic and diastolic blood pressure and no longer statistically significant (mean difference = –1.35, 95% CI –2.85; 0.14, Table S1 (34)). There was no association between thyroid function and venular retinal vessels: mean difference in µm venular vessel diameter per natural log increase in TSH = –0.91 µm (95% CI –1.96; 0.13), mean difference in µm vessel diameter per pmol/L increase in FT4 = 0.18 µm (95% CI –0.16; 0.53), and mean difference in µm diameter in TPO positives vs negatives = –1.26 (95% CI –3.53; 1.01).

Thyroid Function and Cerebrovascular CO2 Reactivity

We did not find any association between thyroid function and CVR: mean difference in CVR per natural log increase in TSH was 0.03 (95% CI –0.03; 0.10), mean difference per pmol/L increase in FT4 –0.001 (95% CI –0.024; 0.023), and mean difference for TPO positive vs negative –0.02, (95% CI –0.17; 0.12).

Sensitivity and Additional Analyses

After we restricted our analyses to euthyroid participants, the nonsignificant association remained the same: TSH as well as TPO positive vs negative and global brain perfusion (mean difference in mL per natural log increase in TSH = –0.16, 95% CI –0.51; 0.66 and mean difference in TPO positive vs negative: –0.41, 95% CI –1.38; 0.55). The inverse U-shaped association between FT4 and global brain perfusion persisted; Figure S2 (34). The associations between TSH and TPO positive vs negative and arteriolar retinal vessels were slightly attenuated. When restricting the analyses to participants that had their thyroid and MRI assessment as well as their retinal vessel assessment at the same time, the inverted U-shaped associations between FT4 and global brain perfusion and arteriolar retinal vessel diameter remained (Figure S3 (34)). Furthermore, when excluding participants with cortical infarcts and carotid stenosis results were similar. Additional adjustment for the 2 other thyroid measures when assessing the relation between thyroid status and brain circulation did not alter results significantly. Additionally adjusting the analyses for antithrombotic agent use did not change the results significantly (data not shown). Finally, when exploring effect modification for sex, age, smoking and time between thyroid measurement and MRI scan, the P value for all interaction terms was >.1.

Discussion

In this study, we found that lower as well as higher FT4 levels were associated with lower global brain perfusion in middle-aged and elderly persons within the general population. Persons included in these analyses represented the full range of thyroid function. Furthermore, higher TSH and TPO levels associated with narrower arteriolar retinal vessel diameters. These associations remained in euthyroid participants and were robust for sensitivity analyses.

Existing literature on the association between thyroid status and brain circulation is conflicting. Some studies find no association (7, 8, 35), others report an association of low (10, 15, 36-39) and 1 study reports an association of high (13) thyroid function with brain circulation. However, most of these studies were performed in specific patient populations, rather than in the general population, and none investigated nonlinearity.

Our observation of an inverted U-shaped association between FT4 and global brain perfusion is novel. Although a previous study found an association between higher FT4 and increased brain perfusion (13), there are 2 important distinctions between that study and ours. First, the former assessed short-term effects of hyperthyroidism, while we investigated the full range of thyroid function. Second, the previous study investigated regional brain perfusion while we assessed global brain perfusion. According to the authors, it can be argued that the increased heart rate as a result of the hyperthyroid state may have led to the increased cerebral perfusion. However, it could not explain why the perfusion is selectively increased (ie, regional brain perfusion change) because the increased heart rate should increase whole brain perfusion (13). Although higher FT4 indeed increases cardiac output, studies have only shown that acutely increasing cardiac output increases brain perfusion (40), while a nonacutely altered cardiac output was not associated with brain perfusion (41). One explanation is that vasoconstriction of other vascular beds shunts the flow from the periphery to the brain, resulting in a lesser extent of brain perfusion reduction than what would be expected based on both the cardiac output and the peripheral blood flow (41). Another explanation is that the extent to which brain perfusion is changed is determined by the integrated effect of all brain perfusion–regulating mechanisms, including other powerful mechanisms unrelated to cardiac output (40), such as cerebral perfusion pressure via cerebral autoregulation (42) and cerebral metabolic activity via neurovascular coupling (43).

It is plausible that a chronic exposure to slightly elevated FT4 could be detrimental for brain circulation for several reasons. First, higher FT4 could lead to tachycardia and subsequently atrial fibrillation, even within the normal range of FT4 (44). Other potential mechanisms are elevated blood pressure (45), atherosclerosis (46), and increased coagulation (47). Also, both hyperthyroidism and hypothyroidism can lead to heart failure (48), which has been linked to reduced CBF (40). Indeed, additional adjustment for potential confounders that could also act as mediators in our third model attenuated our results. However, formal mediation analyses should be conducted to evaluate and quantify the role of these variables. After adjustment for blood pressure the association between TPO positivity and narrower arteriolar retinal vessels was not significant anymore, suggesting that elevated blood pressure could be a mechanism that explains the thyroid status effect on brain circulation.

The fact that the inverse U-shaped association between FT4 is not only present for global brain perfusion, but also for arteriolar retinal vessel diameter in our study further supports our findings, since arteriolar retinal vessel diameter is another proxy for brain circulation. In fact, brain perfusion is regulated by the resistance of vessels, in other words myogenic autoregulation, (42) and this phenomenon is widely distributed in vascular structures, like the retina and brain (49). Autoregulation of the retinal vessels may thus reflect autoregulation in the small brain vessels.

Interestingly, we did not find an association between thyroid status and CVR. Arterial blood CO2 reactivity is only 1 of the 4 known physiologic processes that regulate the caliber of the cerebral resistance vessels (50). It could be that thyroid hormones act through 1 or more other regulatory processes. For instance, the mechanisms for a high-normal FT4 lowering brain perfusion are likely to be arterial blood pressure and cerebral perfusion pressure via cerebral autoregulation (42), and cerebral metabolic activity via neurovascular coupling (43). However, it remains unclear how thyroid function affects these pathways.

Our findings are important, considering the clinical consequences of having a suboptimal brain perfusion. A previous study also measuring global brain perfusion by phase-contrast MRI in the Rotterdam Study found that lower brain perfusion was associated with higher risk of dementia (4). A similar study showed that lower brain perfusion was associated with a higher risk of TIA (5). Translating these findings to our study reveals that a middle-aged or elderly person with an FT4 of 25, has a decreased brain perfusion of 2.44 mL/minute/100 mL compared with another person with an FT4 of 15 and this might indicate a higher risk of 7.5% for developing either dementia or a TIA (over a 7-year period) (4, 5).

It is important to acknowledge the limitations of our study. First, due to the cross-sectional nature of the study we cannot exclude reverse causation, which means that brain circulation could influence thyroid function and not vice versa, as hypothesized in the current study. Second, by using our phase-contrast method, we obtained a measure of brain perfusion in the major brain arteries. This measure is assumed to represent homogeneous flow in the entire brain under normal conditions. Thus, this directly highlights that potentially interesting inhomogeneities of blood flow among different brain regions, i.e. that some local brain areas show increased perfusion, while others show decreased perfusion, cannot be taken into account. Third, the time between thyroid and MRI measurements (median 2.66 months) was long for some participants potentially introducing survival bias. However, sensitivity analyses in participants who had both measurements at the same time yielded similar results. Fourth, we did not have triiodothyronine measurements in our cohort, which could especially be interesting in this context due to the heart’s vulnerability to reductions in biologically active triiodothyronine (48), which could affect brain perfusion. Furthermore, since this study was performed within the general population, participants with thyroid diseases were included in the study. It remains unclear if specific thyroid diseases could be responsible for changed brain perfusion and arteriolar retinal calibers. However, after we restricted our analyses to euthyroid participants, the inverse U-shaped association between FT4 and global brain perfusion persisted. The associations between TSH and TPO positive vs negative and arteriolar retinal vessels were slightly attenuated. Fifth, the vast majority of our population is of European ancestry, limiting generalizability to other ethnicities. Finally, we did not assess cardiac output in relation to brain perfusion in this study. Future studies could evaluate this hemodynamic factor to further explore underlying mechanisms.

In summary, our findings show that lower as well as higher FT4 levels are related to lower global brain perfusion in middle-aged and elderly persons from the general population. Furthermore, higher TSH and TPO levels were associated with narrower arteriolar retinal vessel diameters. These results suggest that thyroid dysfunction could lead to brain diseases such as stroke or dementia through a suboptimal brain circulation. Prospective cohort studies and clinical trials are needed to assess whether thyroid function has a causal relationship with brain perfusion and whether interventions on thyroid function could reduce the incidence of relevant clinical outcomes, such as dementia and TIA.