If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Department of Clinical Biochemistry, Rigshospitalet, Copenhagen University Hospital, DenmarkThe Copenhagen General Population Study, DenmarkDepartment of Clinical Medicine, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark
The Copenhagen General Population Study, DenmarkDepartment of Clinical Biochemistry, Herlev and Gentofte Hospital, Copenhagen University Hospital, DenmarkDepartment of Clinical Medicine, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark
Corresponding author. Department of Clinical Biochemistry KB3011, Section for Molecular Genetics Rigshospitalet, Copenhagen University Hospital Blegdamsvej 9, DK-2100, Copenhagen, Denmark.
Department of Clinical Biochemistry, Rigshospitalet, Copenhagen University Hospital, DenmarkThe Copenhagen General Population Study, DenmarkDepartment of Clinical Medicine, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark
We tested the hypothesis that TSH was associated with cardiovascular disease (CVD).
•
105,224 individuals from the general population were followed for a median of 7 years.
•
Observationally, TSH below the median was associated with increased risk of CVD.
•
Genetically, Mendelian randomization suggested possible causal associations.
Abstract
Background and aims
The association between thyroid stimulating hormone (TSH) and cardiovascular disease has mainly been determined using clinical categories of disease. We tested the hypothesis that TSH on a continuous scale is associated with risk of atrial fibrillation (AF), myocardial infarction (MI), stroke, heart failure (HF), aortic valve stenosis (AVS), and major adverse cardiovascular events (MACE) and whether these associations are likely to be causal.
Methods
We first tested whether plasma TSH on a continuous scale was observationally associated with incident cardiovascular events in a prospective cohort study of 105,224 individuals from the Copenhagen General Population Study followed for a median 7 years. Next, we tested whether a genetic risk score weighted on TSH was associated with cardiovascular endpoints. Finally, using Mendelian randomization, we tested whether the observed associations were likely to be causal.
Results
Using restricted cubic splines, lower concentrations of TSH relative to the population median (=1.53 mIU/L) were associated with higher risk of AF, MI, stroke, HF, AVS, and MACE. Comparing individuals with TSH ≤5th percentile (≤0.54 mIU/L) versus >50th percentile (>1.53 mIU/L), hazard ratios (HRs) ranged from 1.12 (1.00–1.26) for stroke to 1.27 (1.11–1.46) for HF. Genetic risk estimates per standard deviation decrease in TSH were 1.28 (1.08–1.52) for AF, 1.35 (1.06–1.71) for MI, 1.06 (0.89–1.26) for stroke, 1.19 (0.94–1.52) for HF, 1.53 (1.03–2.26) for AVS, and 1.09 (0.97–1.23) for MACE.
Conclusions
In 105,224 individuals from the general population low plasma TSH was observationally and genetically associated with increased risk of AF, MI, and AVS suggesting that these observations may reflect causal pathways.
While overt hypo- and hyperthyroidism increases the risk of cardiovascular morbidity and mortality, the role of subclinical thyroid dysfunction in cardiovascular disease is more controversial and the necessity for treatment therefore debatable [
]. In subclinical hypothyroidism, plasma concentration of thyroid stimulating hormone (TSH) is > 4.0 mIU/L, and in subclinical hyperthyroidism, plasma concentration of TSH is < 0.4 mIU/L, whereas levels of free thyroxine and free triiodothyronine remain within the population reference intervals [
]. Subclinical hypothyroidism is frequent with an estimated prevalence of 5–10% in the general population, while subclinical hyperthyroidism is identified in 0.6–2.0% [
Plasma TSH concentration is by far the best single laboratory test to evaluate thyroid status, therefore rather than studying clinical categories of overt and subclinical hyper- and hypothyroidism, a more biological approach could be to determine cardiovascular risk as a function of TSH on a continuous scale in individuals in the general population. So far, only few studies have adopted this approach [
Because observational studies such as those mentioned above are prone to confounding and reverse causation, these studies can determine associations, but cannot determine causality between thyroid dysfunction, as measured by thyroid parameters, and cardiovascular risk. Mendelian randomization, which mimics randomized clinical trials but uses genetic variants as the long-term random intervention, can help to elucidate possible causal inferences between TSH and cardiovascular risk [
In the present study, we tested the hypothesis that TSH on a continuous scale is associated with risk of atrial fibrillation (AF), myocardial infarction (MI), stroke, heart failure (HF), aortic valve stenosis (AVS) and major adverse cardiovascular events (MACE) in the general population, and whether the observed associations are likely to be causal. For this purpose, we followed 105,224 individuals from the Copenhagen General Population Study (CGPS) for a median of 7 years (969,209 person years).
2. Patients and methods
The study was approved by institutional review boards and a Danish ethical committee (KF100.2039/91) and was conducted according to the principles of the Declaration of Helsinki. Written informed consent was obtained from all individuals.
2.1 Participants
We included 105,224 consecutive individuals attending the first examination of the CGPS from 2003 to 15 [
]. The CGPS is an ongoing, prospective cohort study initiated in 2003, including adults of Danish descent living in the vicinity of Copenhagen, Denmark.
2.2 Endpoints
We followed all individuals using their unique national identification number from entry into the study 2003–15 until occurrence of atrial fibrillation (AF), myocardial infarction (MI), stroke, heart failure (HF), aortic valve stenosis (AVS), major adverse cardiovascular events (MACE), death, or emigration, or March 22nd, 2017, whichever came first. We did not lose track of even a single individual. For more details on participants, endpoints, and laboratory analyses and other covariates see the Supplementary Data.
2.3 Genotyping
We genotyped three top signals in two genes identified and consistently replicated in GWAS studies of TSH [
]: rs4704397 (PDE8B, intron1), rs10917477, and rs10917469 (both upstream of CAPZB). The two variants in CAPZB were not in linkage disequilibrium (R2 = 0.14). Genotyping was by TaqMan-based assays (Applied Biosystems, Foster City, CA, USA), or by KASP genotyping technology (LGC Genomics Ltd, Hoddesdon, Herts, UK).
2.4 Statistical analyses
Data were analyzed using Stata SE 16.1 (Stata Corp, College Station, TX), and R 3.6.2 by RStudio version 1.2.5033. To compare characteristics in individuals by TSH groups, Cuzick's test for trend was used for continuous covariates, and Pearson's χ2-test was used for categorical covariates.
Cox proportional hazards regression models using age as time scale (referred to as age adjusted) and delayed entry (left truncation) and with censoring at event, emigration, death, or end of follow-up, were used to estimate hazard ratios (HRs) for AF, MI, stroke, HF, AVS, and MACE as a function of plasma concentration of TSH. Models were multifactorially adjusted for age, sex, body mass index, smoking, physical inactivity, alcohol consumption, diabetes mellitus, hypertension, menopausal status and hormone replacement therapy (women only), total cholesterol, triglycerides, and lipid-lowering therapy. Restricted cubic splines were used to study the associations between plasma TSH concentration on a continuous scale and risk of AF, MI, stroke, HF, AVS, and MACE. Three knots were chosen for MI, stroke, HF, AVS and MACE, and four knots for AF to balance best fit and overfitting [
Genotypes, weighted allele score groups (based on β-coefficients) (Supplementary Table 1), or simple allele count groups were coded 0, 1 and 2.
Finally, we used inverse ranked normalized transformed TSH values to estimate genetic risk ratios for AF, MI, stroke, HF, AVS, and MACE per one standard deviation lower TSH. For more details on the statistical analyses see the Supplementary Data.
3. Results
A total of 105,224 individuals from the CGPS were included and followed for a median of 7 years (range: <1–13 years; 969,209 person years). Baseline characteristics as a function of plasma TSH in percentile groups and the corresponding concentrations of TSH are shown in Table 1. TSH was associated with all measured parameters except triglycerides and lipoprotein(a). A density plot of TSH is shown in Supplementary Fig. 1. Median TSH was 1.53 mIU/L. Number of individuals according to clinical cut-offs are shown in Supplementary Table 3.
Table 1Baseline characteristics as a function of TSH percentile groups in the Copenhagen General Population Study.
TSH groups
p-value
Percentage TSH, mIU/L
≤5% ≤0.54
>5–10% >0.54–0.71
>10–25% >0.71–1.04
>25–50% >1.04–1.53
>50–75% >1.53–2.21
>75–90% >2.21–3.15
>90–95% >3.15–4.02
>95% >4.02
Number of individuals (%)
5378 (5)
5288 (5)
15847 (15)
26418 (25)
26113 (25)
15704 (15)
5223 (5)
5253 (5)
Age, years (IQR)
63 (53–71)
60 (51–69)
59 (48–68)
57 (47–67)
56 (47–66)
56 (47–66)
56 (47–66)
58 (49–68)
3.7 × 10−118
Women, number (%)
3388 (63)
2880 (54)
8394 (53)
13653 (52)
13835 (53)
8902 (57)
3184 (61)
3729 (71)
1.8 × 10−203
Body mass index, kg/m2 (IQR)
26 (23–29)
26 (23–28)
25 (23–28)
26 (23–28)
26 (23–28)
26 (23–29)
26 (23–29)
26 (23–29)
5.8 × 10−10
Smoking, number (%)
1289 (24)
1207 (23)
3431 (22)
4844 (19)
3855 (15)
2038 (13)
589 (11)
673 (13)
4.6 × 10−222
Low physical activity, number (%)
2404 (45)
2536 (48)
7903 (50)
13634 (52)
13746 (53)
8308 (53)
2743 (53)
2613 (50)
4.5 × 10−32
Alcohol consumption, number (%)
871 (16)
874 (17)
2798 (18)
4653 (18)
4514 (17)
2754 (18)
869 (17)
816 (16)
1.4 × 10−3
Diabetes, number (%)
266 (5)
235 (4)
657 (4)
995 (4)
1006 (4)
603 (4)
197 (4)
190 (4)
9.0 × 10−4
Hypertension, number (%)
3385 (63)
3236 (61)
9345 (59)
15525 (59)
15630 (60)
9555 (61)
3318 (64)
3318 (63)
1.6 × 10−18
Systolic BP, mmHg (IQR)
140 (127–155)
140 (126–154)
139 (125–154)
139 (125–154)
140 (126–154)
140 (127–156)
142 (128–158)
140 (127–157)
6.5 × 10−21
Diastolic BP, mmHg (IQR)
82 (75–90)
83 (75–90)
83 (76–90)
84 (76–90)
84 (77–91)
85 (78–92)
85 (78–93)
85 (78–92)
2.8 × 10−105
HRT, number (%)a
392 (14)
360 (12)
1056 (12)
1656 (12)
1674 (12)
995 (12)
340 (12)
364 (13)
6.9 × 10−2
Menopause, number (%)a
2302 (79)
2219 (74)
6068 (69)
9408 (65)
9315 (64)
5514 (64)
1956 (67)
2049 (71)
2.0 × 10−84
Total cholesterol, mmol/L (IQR)
5.5 (4.8–6.2)
5.5 (4.8–6.3)
5.5 (4.8–6.2)
5.5 (4.8–6.3)
5.5 (4.8–6.3)
5.5 (4.8–6.3)
5.6 (4.9–6.3)
5.6 (4.9–6.4)
1.9 × 10−15
LDL cholesterol, mmol/L (IQR)
3.1 (2.5–3.7)
3.1 (2.5–3.8)
3.1 (2.5–3.8)
3.2 (2.6–3.8)
3.2 (2.6–3.8)
3.2 (2.6–3.9)
3.2 (2.6–3.9)
3.3 (2.6–3.9)
1.7 × 10−29
HDL cholesterol, mmol/L (IQR)
1.6 (1.3–2.0)
1.6 (1.3–2.0)
1.6 (1.3–2.0)
1.6 (1.2–1.9)
1.5 (1.2–1.9)
1.6 (1.2–1.9)
1.6 (1.2–1.9)
1.6 (1.3–2.0)
1.6 × 10−7
Triglycerides, mmol/L (IQR)
1.4 (1.0–2.0)
1.4 (1.0–2.1)
1.4 (1.0–2.0)
1.4 (1.0–2.1)
1.4 (1.0–2.1)
1.4 (1.0–2.0)
1.4 (0.9–2.1)
1.4 (1.0–2.0)
0.11
Lipid-lowering therapy, number (%)
763 (14)
688 (13)
1952 (12)
3086 (12)
3033 (12)
1771 (11)
629 (12)
651 (12)
8.5 × 10−8
Lipoprotein (a), mg/dL (IQR)b
9.6 (4.8–28.3)
9.8 (4.7–30.2)
9.9 (4.7–29.5)
9.7 (4.7–30.0)
9.7 (4.7–29.2)
9.7 (4.8–29.7)
9.8 (4.6–30.2)
10.0 (4.8–30.5)
0.69
Values are median (interquartile range), or number (percent). Body mass index was measured weight in kilograms divided by measured height in meters squared (kg/m2). Smoking was current smokers. Low physical activity was <4 h per week of light physical activity in leisure time. Alcohol consumption was more than 14/21 units of alcohol per week in women/men, respectively (1 unit = 12 g alcohol). Diabetes was self-reported disease, use of anti-diabetic medication, and/or non-fasting plasma glucose >11.0 mmol/L (198 mg/dL). Hypertension was systolic blood pressure ≥140 mmHg, diastolic blood pressure ≥90 mmHg, and/or use of antihypertensive medication. p-values are for Cuzick's test for trend for continuous covariates or Pearson's χ2-test for categorical covariates. a Women only. b Lipoprotein(a) measurements were based on 68,163 individuals from the Copenhagen General Population Study. TSH, thyroid stimulating hormone; IQR, interquartile range; BP, blood pressure; HRT, hormonal replacement therapy; LDL, low density lipoprotein; HDL, high density lipoprotein.
3.1 Cardiovascular endpoints as a function of plasma TSH
During a median follow-up of 7 years (<1–13 years), 5,105 individuals developed incident AF, 2,148 developed MI, 4,659 developed stroke, 2,915 developed HF, 1,312 developed AVS, and 8,675 had a MACE. On a continuous scale using multifactorially adjusted restricted cubic splines, lower concentrations of TSH relative to the reference of 1.53 mIU/L (=median) were associated with increased risk of AF, MI, stroke, HF, AVS, and MACE (Fig. 1). In addition, TSH above the median was associated with a slightly increased risk of atrial fibrillation up until 5mIU/L. Because the spline curves showed that risk of all endpoints was lowest for TSH concentrations around the median (1.53 mIU/L) and gradually increased with lower TSH even within the normal range (0.4–4 mIU/L), whereas risk was almost linear for all endpoints for TSH above the median (Fig. 1), we chose the >50% percentile group as the reference in stratified analyses. In these analyses, multifactorially adjusted HRs for individuals with TSH ≤5th versus >50th percentile group (≤0.54 mIU/L versus >1.53 mIU/L) were 1.21 (95% confidence interval: 1.09–1.34) for AF, 1.25 (1.05–1.48) for MI, 1.12 (1.00–1.26) for stroke, 1.27 (1.11–1.46) for HF, 1.23 (1.00–1.51) for AVS and 1.19 (1.10–1.30) for MACE (Fig. 2). Results were similar among the 99,263 individuals without thyroid disease (excluding 5961 individuals with an ICD diagnosis of hyperthyroidism or hypothyroidism, or TSH and corrected T4 and/or T3 outside the respective reference ranges) (Supplementary Figs. 2 and 3).
Fig. 1Plasma TSH concentration on a continuous scale and risk of atrial fibrillation, myocardial infarction, stroke, heart failure, aortic valve stenosis, and major adverse cardiovascular events in 105,224 individuals from the Copenhagen General Population Study.
Hazard ratio (solid red line) and 95% confidence interval (dashed lines) from Cox regression using restricted cubic splines were multifactorially adjusted for age, sex, body mass index, diabetes mellitus, hypertension, smoking, physical inactivity, alcohol consumption, menopausal status and hormone replacement therapy (women only), total cholesterol, triglycerides and lipid-lowering therapy. The median value of plasma TSH concentration (1.53 mIU/L) was used as the reference. TSH, thyroid stimulating hormone; MACE, major adverse cardiovascular events. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)
Fig. 2Plasma TSH concentration in five percentile groups and risk of atrial fibrillation, myocardial infarction, stroke, heart failure, aortic valve stenosis, and major adverse cardiovascular events in 105,224 individuals from the Copenhagen General Population Study.
Hazard ratios from Cox regression were multifactorially adjusted for age, sex, body mass index, diabetes mellitus, hypertension, smoking, physical inactivity, alcohol consumption, menopausal status and hormone replacement therapy (women only), total cholesterol, triglycerides and lipid-lowering therapy. TSH above the 50th percentile served as the reference group. TSH, thyroid stimulating hormone; MACE, major adverse cardiovascular events.
Using clinical categories corresponding to those in Supplementary Table 3 and the euthyroid group (TSH 0.4–4 mIU/L) as the reference, risk of all cardiovascular endpoints was increased in individuals with subclinical and/or overt hyperthyroidism, whereas there were borderline increases in risk of heart failure and MI in those with subclinical hypothyroidism (Supplementary Fig. 4).
3.2 Interactions between plasma TSH and age and sex on risk of cardiovascular endpoints
There was interaction between plasma TSH and age (<65/≥65 years) on risk of MI, stroke, and MACE (p-values for interaction: 0.01, 0.0004 and 0.0001, respectively), with a higher risk in those <65 years of age; and interaction between plasma TSH and sex on risk of MI and borderline on MACE (P-values for interaction: 0.02 and 0.06, respectively) with a higher risk in women only (Supplementary Figs. 5 and 6). These interactions were due to a higher risk of MI with lower TSH in men <65 years only, and higher risks of stroke, and MACE with lower TSH in both women and men <65 years (Fig. 3).
Fig. 3Interaction between plasma TSH concentration and age on risk of myocardial infarction, stroke, and MACE in women and men from the Copenhagen General Population Study.
TSH was inversed ranked normalized transformed. Hazard ratios are shown in strata of age below or at or above 65 years in women and men separately. Hazard ratios were multifactorially adjusted for body mass index, diabetes mellitus, hypertension, smoking, physical inactivity, alcohol consumption, menopausal status and hormone replacement therapy (women only), total cholesterol, triglycerides, and lipid-lowering therapy, and are shown per 1 stand deviation lower TSH. CI, confidence interval; MACE, major adverse cardiovascular events; TSH, thyroid stimulating hormone.
3.3 Plasma TSH as a function of weighted allele score
Three genetic variants in two genes (PDE8B: rs4704397; and CAPZB: rs10917477 and rs10917469) were used to construct an allele score weighted on lower TSH (Supplementary Tables 1 and 2) and divided into three groups: >50% (=reference), 5–50%, and ≤5%. TSH as a function of the individual genetic variants is shown in Supplementary Fig. 7, and for the weighted allele score and simple allele score groups in Fig. 4 and Supplementary Fig. 8. Population density as a function of the weighted allele score in the CGPS was approximately normally distributed, and lower plasma TSH concentration was associated with increasingly lower allele score (Supplementary Figs. 9 and 10).
Fig. 4Plasma TSH concentration and risk of atrial fibrillation, myocardial infarction, stroke, heart failure, aortic valve stenosis, and major adverse cardiovascular events as a function of groups of a weighted allele score or a simple allele count in the Copenhagen General Population Study.
Cox proportional hazards regression models adjusted for age as time scale and delayed entry (left truncation) and sex were used to estimate hazard ratios for atrial fibrillation, myocardial infarction, stroke, heart failure, aortic valve stenosis, and major adverse cardiovascular events as a function of an allele score weighted on lower TSH, or a simple allele count. Genetic variants included in the scores were PDE8B rs4704397, and CAPZB rs10917477 and rs10917469. p-values are tests for trend. CI, confidence interval; MACE, major adverse cardiovascular events; N, number of individuals; TSH, thyroid stimulating hormone.
Baseline characteristics as a function of weighted allele score groups were similar, implying lack of confounding for measured characteristics in genetic as opposed to observational analyses (Supplementary Table 4, compare with Table 1).
3.4 Cardiovascular endpoints as a function of weighted allele score
The age and sex adjusted HRs as a function of allele score weighted on lower TSH for individuals in the ≤5th versus >50th percentile of the allele score were 1.15 (1.02–1.31) for AF, 1.29 (1.07–1.55) for MI, 1.05 (0.92–1.20) for stroke, 1.05 (0.88–1.24) for HF, 1.09 (0.85–1.41) for AVS, and 1.06 (0.96–1.17) for MACE (Fig. 4, top). Results were similar whether using a simple allele count (Fig. 4, bottom) or excluding individuals with thyroid disease (Supplementary Fig. 8).
Approximately nine percent of the population with the lowest gene score (score ≤5%) had a plasma TSH concentration ≤0.54 mIU/L (Supplementary Fig. 10), suggesting that these individuals had a genetically determined risk for cardiovascular endpoints comparable to the risk observed in individuals with TSH ≤5% (≤0.54 mIU/L) in the observational analyses (compare with Fig. 2).
3.5 Mendelian randomization
A one standard deviation genetically determined lower TSH was associated with HRs of 1.28 (1.08–1.52) for AF, 1.35 (1.06–1.71) for MI, 1.06 (0.89–1.26) for stroke, 1.19 (0.94–1.52) for HF, 1.53 (1.03–2.26) for AVS and 1.09 (0.97–1.23) for MACE using the weighted allele score (Fig. 5). Results were similar when using an externally weighted score [
] (Supplementary Table 1), a simple allele count, or the inverse variance weighted method (Fig. 5). Cragg-Donald Wald F-statistic was 767 for the weighted allele score, implying a robust score (an F-statistic >10 is considered sufficient). In sensitivity analyses, the p-values for intercept from MR Egger were not significant indicating lack of horizontal pleiotropy, and the results from the median weighted method did not suggest weak instrument bias (Fig. 5). Because the observational analyses (spline curves in Fig. 1) might indicate nonlinear associations we applied the nlmr method (piecewise_mr) [
Emerging Risk Factors Collaboration, E-CVDVDSC Estimating dose-response relationships for vitamin D with coronary heart disease, stroke, and all-cause mortality: observational and Mendelian randomisation analyses.
] and found no evidence of nonlinearity as indicated by the quadratic p-values for all cardiac endpoints (Supplementary Fig. 11; all p-values >0.29). In agreement, applying the nlmr method yielded similar results as the classical Mendelian randomization methods (Fig. 5).
Fig. 5Risk ratios for atrial fibrillation, myocardial infarction, stroke, heart failure, aortic valve stenosis, and major adverse cardiovascular events per 1 standard deviation genetically lower TSH.
Internal and external beta coefficients were adjusted for age and sex. MR Egger and the median weighted method are sensitivity analyses addressing pleiotropy (estimated by the p-value for the intercept) and weak instrument bias, respectively. CI, confidence interval; MACE, major adverse cardiovascular events; N, number of individuals.
This study comprising 105,224 individuals from the Danish general population has several novel findings. First, plasma TSH concentration below the population median (1.53 mIU/L) was observationally associated with increased risk of myocardial infarction (MI), stroke, heart failure (HF), aortic valve stenosis (AVS), and major adverse cardiovascular events (MACE) with atrial fibrillation (AF) as a positive control. Second, risk of MI in men and risk of stroke and MACE in both sexes was increased in those <65 years of age, but not in older individuals ≥65 years. Finally, genetically lower concentrations of plasma TSH were associated with higher risk of AF, MI, and AVS, suggesting that these observations may reflect causal pathways (Fig. 6). These findings may be clinically important because they suggest the addition of TSH even within the normal range to other preventive strategies for AF, MI, and AVS.
Fig. 6Mendelian randomization of plasma TSH concentration and risk of cardiovascular disease in the general population.
We tested the hypothesis that TSH on a continuous scale is causally associated with risk of atrial fibrillation (AF), myocardial infarction (MI), stroke, heart failure (HF), aortic valve stenosis (AVS), and major adverse cardiovascular events in the Copenhagen general Population Study (n = 105,224 individuals; median follow-up 7 years).
Top panel: We first tested whether plasma TSH was observationally associated with incident cardiovascular events (1). Next, we tested whether a genetic risk score, comprising variants in PDE8B and CAPZB and weighted on TSH (2), was associated with incident cardiovascular events (3). Finally, using Mendelian randomization, we tested whether the observed associations were likely to be causal (4).
Lower panel: A one standard deviation genetically lower plasma TSH was observationally and genetically associated with increased risk of AF, MI, and AVS suggesting that these observations may reflect causal pathways. These findings may be clinically important because they suggest the addition of TSH even within the normal range to other preventive strategies for AF, MI, and AVS. TSH, thyroid stimulating hormone; CAPZB, Capping Actin Protein of muscle Z-line subunit Beta gene; PDE8B, Phosphodiesterase 8B gene. Created with BioRender.com.
]. Subclinical hyperthyroidism has been associated with changes in risk of MI in patients who were treated with levothyroxine and had normal free T4 and T3 levels but suppressed TSH [
]. Other studies have shown that fixed suppressive TSH treatment that may lead to a state of subclinical hyperthyroidism, significantly affected several cardiac parameters including reduced exercise tolerance [
Cardiac function, physical exercise capacity, and quality of life during long-term thyrotropin-suppressive therapy with levothyroxine: effect of individual dose tailoring.
]. Additional mechanisms that may explain the present associations between low TSH levels and risk of MI include: 1) Cardiac contractility measured as pre-ejection period in milliseconds was shortest in overt hyperthyroidism and increased through subclinical hyperthyroidism, euthyroid, subclinical hypothyroidism and finally towards overt hypothyroidism [
]. Consequently, a fast contractility function in subclinical hyperthyroidism will increase the oxygen demand above the supply and may lead to an ischemic event; 2) Increased fibrinogen levels are associated with low plasma TSH levels, possibly leading to a prothrombotic state increasing risk of cardiovascular disease [
Carotid Intima-Media Thickness Score, Positive Coronary Artery Calcium Score, and Incident Coronary Heart Disease: the Multi-Ethnic Study of Atherosclerosis. vol. 6. Journal of the American Heart Association,
2017
]. Future mechanistic studies are warranted to fully understand the mechanisms behind low levels of plasma TSH and risk of MI and AVS.
In the ARIC study (n = 11,359), using restricted cubic splines both TSH below the median and extreme high levels of TSH were associated with a higher risk of MI [
]. The HUNT study (n = 26,707) reported associations between subclinical hyperthyroidism and subclinical hypothyroidism and risk of coronary death in women only [
], but no associations with first time hospitalization for MI or HF. Our results support that associations with TSH or subclinical disease may be both sex and age-specific for some endpoints. We found that risk of MI was increased in women with low TSH regardless of age, while risk in men was restricted to those under 65 years who have a lower number of competing risk factors and comorbidities than older men. Likewise, we found an increased risk of stroke and MACE in women and men <65 years, but not in those ≥65 years. In contrast to our results, no significant association between TSH levels and coronary artery disease was observed in a recent study [
Thyrotropin and free thyroxine levels and coronary artery disease: cross-sectional analysis of the Brazilian Longitudinal Study of Adult Health (ELSA-Brasil).
]. However, that study did not capture low levels of plasma TSH (TSH between 0 and 0.95 mIU/L), and lack of power was likely an issue (n = 767 individuals). Coppala et al. [
] defined subclinical hyperthyroidism as TSH between 0.1 and 0.45 mIU/L, and found no associations between subclinical hyperthyroidism and coronary heart disease. Finally, Langén et al. [
] did not detect any associations between TSH concentrations and cardiovascular endpoints using a cubic spline function. However, the total number and number of events were of moderate size (n = 5,211; 402, 595, and 327 for MACE, cardiovascular disease, and AF, respectively).
Using clinical categories, we found an increased risk of all cardiovascular endpoints in individuals with subclinical and/or overt hyperthyroidism, while there were borderline increases in risk of heart failure and MI in those with subclinical hypothyroidism. These findings agree with previous meta-analyses which showed an increased risk of coronary heart disease events and/or mortality in those with subclinical hypothyroidism, particularly in those with TSH concentrations of 10 mIU/L or higher [
Relationship between subclinical thyroid dysfunction and the risk of cardiovascular outcomes: a systematic review and meta-analysis of prospective cohort studies.
]. In our study, this latter group accounts for only 0.2% of the population and is therefore not well-powered to further explore this (Supplementary Table 3).
Finally, a few smaller case-control studies (<100 participants) have shown an association between overt or subclinical hypothyroidism with very high TSH levels (median above 11 mIU/L) and increased concentration of lipoprotein(a) [
]. In the Copenhagen General Population Study, we measured lipoprotein(a) in more than 68,000 individuals and found no association between plasma TSH and lipoprotein(a) concentration. There was also no association between clinical categories of thyroid disease and plasma concentration of lipoprotein(a) (Supplementary Table 5).
For the genetic analyses, we chose variants in biologically highly relevant genes which have repeatedly been shown to associate with TSH [
]. PDE8B (encoded by PDE8B) is a cAMP-specific phosphodiesterase that is mainly transcribed in the thyroid gland. cAMP itself plays an important role in stimulation of the thyroid by TSH and interacts with the activity of CapZ proteins (encoded by CAPZB) which in turn change the uptake of thyroglobulin [
A recent genetic study using Mendelian randomization and summary GWAS data for thyroid function and cardiovascular disease from various publicly available databases supported a causal role for decreased TSH with an increased risk of AF in keeping with our findings but not with other cardiovascular endpoints, including MI, stroke, HF, and AVS [
]. However, a shortcoming of that study was that non-linear associations with TSH using restricted cubic splines or effects of very low or very high TSH could not be examined.
A major strength of our study is the large prospective general population design with no losses to follow-up; that is, every single individual could be followed to end of follow-up, death, or emigration. To our knowledge, this is the largest prospective, single-cohort, single-lab study reporting an association between plasma levels of TSH on a continuous scale and risk of cardiovascular endpoints. Our consecutive TSH measurements performed in the same laboratory make the study less prone to suffer from regression towards the null. Another strength is that our findings remained after restricting the analyses to individuals without thyroid disease. A limitation is that we only studied white individuals; however, we are not aware of data to suggest that our findings would not apply to other ethnicities.
In conclusion, plasma concentrations of TSH below the median were observationally associated with increased risk of AF, MI, stroke, HF, AVS, and MACE. Moreover, genetically lower concentrations of TSH were associated with risk of AF, MI, and AVS, suggesting that these observational associations may reflect causal pathways.
CRediT authorship contribution statement
Nawar Dalila: Formal analysis, Data curation, Writing – original draft. Ruth Frikke-Schmidt: Writing – original draft. Anne Tybjærg-Hansen: Writing – original draft. Børge G. Nordestgaard: Interpretation of data.
Declaration of competing interest
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests. AT-H reports consultancies or talks sponsored by Akcea, Amgen, AstraZeneca, Draupnir Bio, Novartis, Regeneron, Sanofi, and Silence Therapeutics. BGN reports consultancies or talks sponsored by AstraZeneca, Sanofi, Regeneron, Akcea, Amgen, Kowa, Denka Seiken, Amarin, Novartis, Novo Nordisk, and Silence Therapeutics. RF-S reports consultancy sponsored by Novo Nordisk. ND had no conflicts of interest.
Acknowledgments
We thank the staff and participants of the Copenhagen General Population Study (CGPS) for their important contribution to the study.
Appendix A. Supplementary data
The following is the Supplementary data to this article.
Parzen E. Tanabe K. Kitagawa G. Information Theory and an Extension of the Maximum Likelihood Principle. Springer,
New York, NY1998: 199-213 (New York)
Estimating dose-response relationships for vitamin D with coronary heart disease, stroke, and all-cause mortality: observational and Mendelian randomisation analyses.
Cardiac function, physical exercise capacity, and quality of life during long-term thyrotropin-suppressive therapy with levothyroxine: effect of individual dose tailoring.
Carotid Intima-Media Thickness Score, Positive Coronary Artery Calcium Score, and Incident Coronary Heart Disease: the Multi-Ethnic Study of Atherosclerosis. vol. 6. Journal of the American Heart Association,
2017
Thyrotropin and free thyroxine levels and coronary artery disease: cross-sectional analysis of the Brazilian Longitudinal Study of Adult Health (ELSA-Brasil).
Relationship between subclinical thyroid dysfunction and the risk of cardiovascular outcomes: a systematic review and meta-analysis of prospective cohort studies.
00224-1&id=ga1.jpg){kind=link}
00224-1&id=gr1.jpg){kind=link}
00224-1&id=gr2.jpg){kind=link}
00224-1&id=gr3.jpg){kind=link}
00224-1&id=gr4.jpg){kind=link}
00224-1&id=gr5.jpg){kind=link}
00224-1&id=gr6.jpg){kind=link}