Table 3.
Main and Sensitivity Analysis of the Association between Thyroid Stimulating Hormone within the Reference Range and the Risk of Atrial Fibrillation
TSH level (mlU/l) | 0.45–0.99 | 1.00–1.49 | 1.50–2.49 | 2.50–3.49 | 3.50–4.49 | |||||
---|---|---|---|---|---|---|---|---|---|---|
Events / Persons |
HR (95% CI) |
Events / Persons |
HR (95% CI) |
Events / Persons |
HR (95% CI) |
Events / Persons |
HR (95% CI) |
Events / Persons |
HR (95% CI) |
|
All available cohorts | ||||||||||
Age- and sex-adjusted | 372/5665 | 1.10 (0.92–1.31) | 492/6275 | 1.03 (0.87–1.22) | 893/9990 | 1.04 (0.89–1.22) | 412/4391 | 0.94 (0.79–1.12) | 190/1806 | ref. |
Multivariable adjusted analysis* | 370/5611 | 1.07 (0.89–1.28) | 487/6199 | 0.99 (0.84–1.18) | 882/9841 | 1.02 (0.87–1.19) | 410/4336 | 0.92 (0.78–1.10) | 187/1777 | ref. |
Further adjustments of multivariable models* | ||||||||||
Plus antihypertensive and lipid lowering medication | 357/4924 | 1.06 (0.88–1.27) | 469/5699 | 0.98 (0.83–1.17) | 862/9297 | 1.01 (0.86–1.19) | 405/4138 | 0.94 (0.78–1.12) | 180/1703 | ref. |
Plus BMI | 356/4911 | 1.07 (0.89–1.29) | 466/5682 | 0.99 (0.84–1.18) | 855/9257 | 1.02 (0.87–1.20) | 403/4120 | 0.94 (0.79–1.13) | 179/1696 | ref. |
Medication affecting thyroid function | ||||||||||
Including all regardless of thyroid medication use † | 399/5920 | 1.10 (0.92–1.31) | 516/6437 | 1.05 (0.89–1.24) | 925/10226 | 1.06 (0.91–1.23) | 432/4533 | 0.96 (0.81–1.14) | 200/1914 | ref. |
Excluding thyroid medication use at BL and/or FUP ‡ | 370/5552 | 1.10 (0.91–1.32) | 483/6217 | 1.01 (0.85–1.21) | 881/9902 | 1.04 (0.88–1.22) | 393/4305 | 0.93 (0.78–1.11) | 173/1722 | ref. |
Excluding users of amiodarone § | 367/5639 | 1.09 (0.91–1.30) | 490/6254 | 1.03 (0.87–1.21) | 889/9958 | 1.04 (0.89–1.22) | 408/4380 | 0.94 (0.78–1.11) | 190/1795 | ref. |
Excluding users of other drugs affecting thyroid function ‖ | 241/4536 | 1.21 (0.96–1.53) | 302/4815 | 1.12 (0.89–1.40) | 531/7364 | 1.13 (0.91–1.40) | 222/3059 | 0.99 (0.78–1.25) | 103/1273 | ref. |
Thyroid function | ||||||||||
Excluding participants with missing values of fT4 # | 288/4808 | 1.10 (0.89–1.34) | 365/4983 | 1.01 (0.83–1.22) | 650/7561 | 1.01 (0.85–1.21) | 289/3097 | 0.90 (0.74–1.10) | 147/1314 | ref. |
Including only participants with persistent thyroid function state ** | 62/905 | 0.96 (0.61–1.51) | 102/1304 | 1.04 (0.68–1.60) | 214/2237 | 1.20 (0.81–1.80) | 78/985 | 0.91 (0.59–1.41) | 27/314 | ref. |
Excluding studies | ||||||||||
Excluding studies with AF diagnosis without ECG review †† | 282/3912 | 1.13 (0.92–1.39) | 356/3651 | 1.05 (0.86–1.27) | 683/5517 | 1.11 (0.93–1.33) | 334/2673 | 1.00 (0.82–1.21) | 147/1094 | ref. |
Excluding studies with >5% lost to follow-up ‡‡ | 236/3100 | 1.05 (0.83–1.33) | 324/4515 | 1.00 (0.80–1.25) | 574/7851 | 1.03 (0.83–1.28) | 233/3407 | 0.89 (0.71–1.13) | 99/1354 | ref. |
Excluding MrOS Study §§ | 364/5595 | 1.10 (0.92–1.32) | 477/6144 | 1.03 (0.87–1.23) | 874/9752 | 1.06 (0.90–1.25) | 402/4254 | 0.96 (0.81–1.14) | 182/1749 | ref. |
Excluding SHIP Study ‖‖ | 343/4046 | 1.10 (0.92–1.32) | 483/5775 | 1.02 (0.86–1.21) | 891/9820 | 1.04 (0.89–1.22) | 412/4359 | 0.94 (0.79–1.12) | 190/1800 | ref. |
Abbreviations: AF, Atrial Fibrillation; BMI, Body Mass Index; BL, Baseline; CI, Confidence Interval; E, events; ECG, Electrocardiogram; fT4, free Thyroxin; FUP, Follow-up; HR, Hazard Ratio; MrOS, Osteoporotic Fractures in Men Study; P, participants; ref., reference; SHIP, Study of Health in Pomerania; TSH, Thyroid Stimulating Hormone
Adjusted for age, sex, systolic blood pressure, current and former smoking, diabetes, total cholesterol, and prevalent cardiovascular disease at baseline.
For this sensitivity analysis, we added a total of 1177 thyroid medication users (thyroxin or anti-thyroid drugs) to the overall sample: 253 participants in the Cardiovascular Health Study; 207 in the Health ABC Study; 43 in the Osteoporotic Fractures in Men Study; 15 in the Bari Study; 12 in the Leiden 85+ Study; 107 in the Study of Health in Pomerania; 21 in the Invecchiare in Chianti Study; 26 in the Rotterdam Study; 188 in the PROSPER Study; 301 in the EPIC-Norfolk Study; and, 4 in the Busselton Health Study.
The number of thyroid medication users (thyroxin or anti-thyroid drugs) during follow-up are indicated in Table 1
A total of 123 participants who took amiodarone were excluded for this sensitivity analysis of the association between TSH and AF: 2 participants in the Cardiovascular Health Study; 3 in the Health ABC Study; 1 in the Osteoporotic Fractures in Men Study; 79 in the Bari Study; 1 in the Leiden 85+ Study; 1 in the Study of Health in Pomerania; 6 in the Invecchiare in Chianti Study; 6 in the Rotterdam Study; 23 in the PROSPER Study; 1 in the EPIC-Norfolk Study. Information on amiodarone use was not available in the Busselton Health Study.
A total of 7,786 participants who took other medications that could alter thyroid levels (corticosteroids, amiodarone, iodine, lithium, aspirin, furosemide) were excluded for this sensitivity analysis: 1,634 participants in the Cardiovascular Health Study; 881 in the Health ABC Study; 245 in the Osteoporotic Fractures in Men Study; 251 in the Bari Study; 56 in the Leiden 85+ Study; 299 in the Study of Health in Pomerania; 122 in the Invecchiare in Chianti Study; 151 in the Rotterdam Study; 3,199 in the PROSPER Study; 948 in the EPIC-Norfolk Study.
A total of 311 participants were excluded for this analysis: 22 participants in the Cardiovascular Health Study; 5 in the Leiden 85+ Study; 1 in the InChianti Study; 282 in the Rotterdam Study; and 1 in the Busselton Health Study had missing measurements for fT4. In participants in the Health ABC study, fT4 was measured only in participants with TSH ≥7.0 mIU/L (therefore, no fT4 measurement in 2270 participants) and, in the PROSPER Study, fT4 was measured only in participants with TSH <0.45mIU/l or TSH ≥4.5mIU/l (therefore, no fT4 measurement in 4220 participants)
Persistent thyroid function state was defined as persistent category (euthyroidism, subclinical hypothyroidism) from baseline to follow-up thyroid function test. Follow-up thyroid function was measured in the Cardiovascular Health Study, Leiden 85-plus Study, PROSPER Study, Study of Health in Pomerania, and the Busselton Health Study.
The Bari and EPIC-Norfolk studies were excluded.
The Cardiovascular Health Study, MrOS Study, SHIP Study, InChianti Study, and Busselton Health Study were excluded.
In the MrOS Study, thyroid hormones were measured an average of 3.4 years before other baseline characteristics were assessed.
SHIP includes participants from Pomerania, where an iodine supplementation began in the mid-1990s, which shifted the distribution of TSH values towards the left during the first years, and lowered TSH values in the population in the SHIP Study during the baseline examinations in 1997–2001.