Skip to main content
PMC home page
The Journal of Clinical Endocrinology and Metabolism logoLink to The Journal of Clinical Endocrinology and Metabolism
. 2016 Sep 7;101(11):4270–4282. doi: 10.1210/jc.2016-2255

Thyroid Function Within the Reference Range and the Risk of Stroke: An Individual Participant Data Analysis

Layal Chaker 1, Christine Baumgartner 1, Wendy P J den Elzen 1, Tinh-Hai Collet 1, M Arfan Ikram 1, Manuel R Blum 1, Abbas Dehghan 1, Christiane Drechsler 1, Robert N Luben 1, Marileen L P Portegies 1, Giorgio Iervasi 1, Marco Medici 1, David J Stott 1, Robin P Dullaart 1, Ian Ford 1, Alexandra Bremner 1, Anne B Newman 1, Christoph Wanner 1, José A Sgarbi 1, Marcus Dörr 1, W T Longstreth Jr 1, Bruce M Psaty 1, Luigi Ferrucci 1, Rui M B Maciel 1, Rudi G Westendorp 1, J Wouter Jukema 1, Graziano Ceresini 1, Misa Imaizumi 1, Albert Hofman 1, Stephan J L Bakker 1, Jayne A Franklyn 1, Kay-Tee Khaw 1, Douglas C Bauer 1, John P Walsh 1, Salman Razvi 1, Jacobijn Gussekloo 1, Henry Völzke 1, Oscar H Franco 1, Anne R Cappola 1, Nicolas Rodondi 1, Robin P Peeters 1,✉; for the Thyroid Studies Collaboration1
PMCID: PMC5095234  PMID: 27603906

Abstract

Context:

The currently applied reference ranges for thyroid function are under debate. Despite evidence that thyroid function within the reference range is related with several cardiovascular disorders, its association with the risk of stroke has not been evaluated previously.

Design and Setting:

We identified studies through a systematic literature search and the Thyroid Studies Collaboration, a collaboration of prospective cohort studies. Studies measuring baseline TSH, free T4, and stroke outcomes were included, and we collected individual participant data from each study, including thyroid function measurements and incident all stroke (combined fatal and nonfatal) and fatal stroke. The applied reference range for TSH levels was between 0.45 and 4.49 mIU/L.

Results:

We collected individual participant data on 43 598 adults with TSH within the reference range from 17 cohorts, with a median follow-up of 11.6 years (interquartile range 5.1–13.9), including 449 908 person-years. Age- and sex-adjusted pooled hazard ratio for TSH was 0.78 (95% confidence interval [CI] 0.65–0.95 across the reference range of TSH) for all stroke and 0.83 (95% CI 0.62–1.09) for fatal stroke. For the free T4 analyses, the hazard ratio was 1.08 (95% CI 0.99–1.15 per SD increase) for all stroke and 1.10 (95% CI 1.04–1.19) for fatal stroke. This was independent of cardiovascular risk factors including systolic blood pressure, total cholesterol, smoking, and prevalent diabetes.

Conclusion:

Higher levels of TSH within the reference range may decrease the risk of stroke, highlighting the need for further research focusing on the clinical consequences associated with differences within the reference range of thyroid function.

In an individual participant data analysis of 43,598 adults, there was an increased risk of stroke with lower TSH levels and higher FT4 levels within the reference range of thyroid function.

Subclinical hypothyroidism is associated with hypertension, hyperlipidemia, atherosclerosis, and an increased risk of coronary artery disease, whereas subclinical hyperthyroidism is associated with an increased risk of atrial fibrillation (AF) and coronary artery disease (1–4). Subclinical thyroid dysfunction is defined by a TSH level outside the reference range with a free T4 (FT4) within the reference range.

However, the currently applied reference ranges for thyroid function are under debate (5, 6) because thyroid function within these reference ranges is also associated with several adverse health outcomes (7–9). A previous systematic review found that lower TSH values and higher FT4 values within the reference range are associated with reduced bone mineral density and AF and an increased risk of fractures (8). Furthermore, higher levels of TSH and lower levels of FT4 within the reference range are associated with cardiovascular events and an unfavorable metabolic profile (8). On the other hand, a previous individual participant data (IPD) analysis provided no evidence for a higher risk of coronary heart disease (CHD) within the reference range as it is currently defined (10).

A considerable amount of data exists on the association of thyroid function within the reference range and cardiovascular risk factors such as AF, hypercholesterolemia, and hypertension (8). Whereas these risk factors related to differences within the reference range are also associated with cardiovascular disease, few data are available on clinical outcomes and no data are available on the risk of stroke, the second major vascular cause of morbidity and mortality worldwide (11). A previous study-level meta-analysis on the association of subclinical thyroid dysfunction and stroke risk included only a small number of studies and did not include any analyses on TSH within the reference range (12). Assessing the consequences of differences within the reference range of thyroid function on clinical outcomes is important for understanding the definition of the reference range and to improve care and preventive measures. Furthermore, it can help identify clinical outcomes that need to be addressed in future randomized controlled trials assessing the benefits and risks of thyroid treatment in subclinical thyroid dysfunction (13).

Therefore, we aimed to investigate the association between TSH and FT4 differences within the reference range and the risk of stroke (fatal and nonfatal) in an IPD analysis. An IPD analysis provides the opportunity to standardize definitions of thyroid function and statistical analyses, include unpublished data, and pool results from several cohorts. Also, an IPD can provide the opportunity to conduct subgroup analyses due to the large number of events included.

Materials and Methods

Data sources and study selection

Studies were identified through the Thyroid Studies Collaboration (TSC). The TSC is a consortium of cohorts with thyroid function measurements at baseline and prospective follow-up of cardiovascular outcomes (1, 4, 10, 14–16). Its primary purpose is to examine the association of subclinical thyroid dysfunction and cardiovascular disease. Eligible cohorts were originally identified through systematic literature reviews (1), and this has been described in detail previously (12). From the 19 cohorts identified by these two literature searches, 17 cohorts had information available on baseline thyroid function and follow-up stroke incidence, agreed to participate, and were therefore eligible for the current study. No additional inclusion criteria were applied. None of the cohorts has previously published on the risk of stroke within the reference range of thyroid function, and five cohorts (17–21) previously published on the association of subclinical thyroid dysfunction and the risk of stroke (Table 1). Investigators from the 17 eligible studies were invited to join the IPD analysis. The local medical ethics committees of each included study approved the distinct original study protocols, and informed consent was obtained from all study participants by the original cohort studies.

Table 1.

Baseline Characteristics of Individuals in the Included Studies (n = 43 598)

Study, Year (Reference) Description of Study Sample n Median Age (Range), ya Women, n, % Thyroid Medication at Baseline, n, %b Thyroid Medication at Follow-Up n, %c TSH, Median (IQR) FT4 Mean (SD)d Follow-Up Median (IQR) Person-Years
4D Study, 1998 (18) Trial of atorvastatin in type 2 diabetes and hemodialysis patients, Germany 841 66 (30–83) 368 (43.8) 0 11 (1.3) 1.10 (0.77–1.60) 13.90 pmol/L (2.92) 1.5 (0.2–3.6) 1666
Birmingham Study, 1988 (20) CDAs aged ≥60 y from primary care practice in Birmingham, England 1015 69 (60–94) 550 (54.2) 0 NA 1.60 (1.10–1.20) NA 10.2 (5.7–10.6) 8301
Brazilian Thyroid Study, 1999 (41) Adults from Japanese descent living in São Paulo, Brazil 890 56 (30–92) 459 (51.6) 0 NA 1.40 (0.90–2.20) 1.07 ng/dL (0.18) 7.3 (7.1–7.5) 6274
Busselton Health Study, 1981 (25) Adults in Busselton, Western Australia 1902 50 (18–90) 912 (47.9) 0 11 (0.6) 1.42 (1.00–1.96) 16.35 pmol/L (2.89) 20.0 (19.9–20.0) 33 825
Cardiovascular Health Study, 1989 (17) CDAs with Medicare eligibility in four US communities 2526 71 (64–100) 1488 (58.9) 0 52 (2.1) 2.05 (1.45–2.89) NA 14.1 (8.6–16.4) 31 099
EPIC-Norfolk Study, 1995 (24) Adults living in Norfolk, England 11 986 58 (40–78) 6365 (53.1) 0 NA 1.70 (1.20–2.30) 12.58 pmol/L (3.17) 13.4 (12.6–14.3) 153 766
Health ABC Study, 1997 (21) CDAs with Medicare eligibility in two US communities 2170 74 (69–81) 1033 (47.6) 0 37 (1.7) 2.00 (1.37–2.72) NA 11.8 (7.5–12.2) 21 057
InCHIANTI Study, 1998 (35) Adults aged 20–102 y living in Chianti geographic area, Italy 1049 71 (21–102) 575 (54.8) 11 (1.0) NA 1.38 (0.96–1.98) 1.42 ng/dL (0.29) 9.1 (8.2–9.2) 8435
Leiden 85-plus Study, 1997 (36) Adults aged 85 y living in Leiden, The Netherlands 452 85 (NA) 290 (60.4) 0 6 (1.3) 1.65 (1.15–2.31) 14.5 pmol/L (2.26) 5.2 (2.5–8.5) 2555
MrOS Study, 2000 (34) Community-dwelling US men aged ≥65 y 1410 73 (65–99) 0 83 (5.9) NA 1.97 (1.36–2.72) 0.99 ng/dL (0.15) 12.0 (8.5–12.7) 14 541
Nagasaki Adult Health Study, 1984 (19) Atomic bomb survivors in Nagasaki, Japan 2342 57 (38–92) 1419 (60.6) 27 (1.2) NA 2.60 (2.00–3.40) 1.45 ng/dL (0.46) 13.0 (12.3–13.7) 28 574
Pisa cohort, 2000 (38) Patients admitted to cardiology department in Pisa, Italye 2695 63 (19–92) 840 (31.2) 0 0 1.53 (1.02–2.30) 1.19 ng/dL (0.24) 2.6 (1.6–3.8) 7326
PREVEND Study, 1997 (37) Adults living in Groningen, The Netherlands 2493 46 (28–75) 1255 (50.3) 0 4 (0.2) 1.37 (0.99–1.90) 12.81 pmol/L (2.25) 10.9 (10.6–11.1) 24 621
PROSPER trial, 1997 (28) Trial on the benefits of pravastatin vs placebo in adults 4953 75 (69–83) 2403 (48.5) 0 28 (0.6) 1.80 (1.26–2.51) NA 3.3 (3.0–3.5) 15 937
Rotterdam Study, 1989 (40) Adults ≥55 y living in Rotterdam, The Netherlands 1577 68 (55–93) 934 (59.2) 0 NA 1.54 (1.06–2.26) 16.29 pmol/L (2.93) 17.0 (11.2–18.9) 23 217
SHIP Study, 1997 (39) Adults in West Pomerania, northeast of Germany 2977 47 (20–81) 1476 (49.6) 0 90 (3.0) 0.79 (0.61–1.07) 12.67 pmol/L (3.42) 11.3 (10.6–11.8) 32 238
Whickham Survey, 1974 (27)f Adults living in and near Newcastle upon Tyne, England 2320 46 (18–92) 1213 (52.3) 92 (4.0) 54 (2.3) 2.10 (1.20–3.00) 8.41 pmol/L (1.95) 19.0 (15.8–20.0) 37 252
Overall 43 598 64.9 (18–102) 21 580 (49.6) 213 (0.5) 293 (1.4) 1.65 (1.10–2.40) 13.6 pmol/L (2.6) 11.6 (5.1–13.9) 450 684
Open in a new tab

Abbreviations: CDA, community-dwelling adult; 4D, Die Deutsche Diabetes Dialyze; IQR, interquartile range (25th-75th percentile); MrOS, Osteoporotic Fractures in Men; NA, not applicable; SHIP, Study of Health in Pomerania.

a

Participants younger than 18 years of age were not included.

b

Participants with missing information on thyroid medication at baseline: Health ABC Study, seven; MrOs Study, 59; Rotterdam Study, 463; and Whickham Survey, three.

c

Participants with missing information on thyroid medication at follow-up: Whickham Survey 1430.

d

Measure of 1 pmol/L is 0.0777 ng/dL.

e

Excluded patients with acute coronary syndrome or severe illness.

f

The Whickham Survey used a first-generation assay for the measurement of TSH and did not measure FT4 but total T4.

Data extraction

We requested individual participant characteristics related to prior cardiovascular risk factors and disease, including systolic blood pressure (BP), serum total cholesterol, a history of diabetes, smoking, previous cardiovascular disease, and previous stroke. We also collected available information on demographic information (age, sex, race), anthropometric measurements (height and weight), medication use (thyroid hormone replacement, lipid lowering, and antihypertensive therapy), and the outcome. Individual participant information from all cohorts were collected and analyzed in one center (Rotterdam, The Netherlands). The primary outcome measures were all stroke (combined fatal and nonfatal) and fatal stroke. Stroke was defined according to World Health Organization criteria as a syndrome of rapidly developing clinical signs of focal (or global) disturbance of cerebral function, with symptoms lasting 24 hours or longer or leading to death, with no apparent cause other than of vascular origin, including ischemic or hemorrhagic strokes.

Thyroid function testing definition

We used a common definition of the reference range of thyroid function (ie, euthyroidism) to increase comparability among the different studies and in concordance with previous analyses (1, 4, 16), expert reviews (22, 23), and several large cohorts (17, 24, 25). Euthyroidism was defined as TSH level between 0.45 and 4.49 mIU/L (1). Most studies used a third-generation TSH RIA, but the Whickham Survey used a first-generation assay that reports higher measured TSH values than current assays (26), for which we adjusted the range to 0.5–6.0 mIU/L to define euthyroidism, as previously described (1, 15, 27). In addition, the Whickham Survey was the only study to perform total T4 assays (27); the remainder of the cohorts performed FT4 assays.

For FT4 values, we excluded studies that measured FT4 in TSH values only outside the reference range for these analyses (17, 20, 21, 28). In studies that measured FT4 independent of TSH values, we used all FT4 levels with individuals with TSH in the reference range, not limited by the FT4 reference range.

Data synthesis and statistical analysis

We performed a Cox proportional hazards model in each cohort separately to assess the association of TSH or FT4 continuously with all stroke and fatal stroke (IBM SPSS Statistics for Windows, version 21.0.; IBM Corp). We investigated the linearity assumption using cubic restricted splines (rms package, R-project; Institute for Statistics and Mathematics, R Core Team (2013), version 3.0.2). Due to the departure from linearity for the TSH analysis in the Die Deutsche Diabetes Dialyze (4D) cohort (P for nonlinearity = .03), TSH was log transformed for all continuous analyses (natural logarithm). We found no departure from nonlinearity in the transformed TSH or any of the FT4 analyses and no threshold effect was therefore detected. The analyses are presented as hazard ratios (HR) across the reference range of TSH (0.45–4.49 mIU/L). This corresponds to the HR when comparing participants with a TSH in the upper limit of the reference range (4.49 mIU/L) with participants with a TSH in the lower limit of the reference range (0.45 mIU/L). The FT4 analyses were performed in a standardized manner (per SD) as well as per 1 ng/dL increase, for which the Whickham study (27) was excluded. We assessed the proportional hazard assumption in each cohort for each outcome by Schoenfeld residual plots and the Schoenfeld test. All studies met the proportional hazard assumption except for the Birmingham study and Pravastatin in Elderly Individuals at Risk of Vascular Disease (PROSPER) trial for the analyses with TSH, for which we performed a sensitivity analysis excluding these two cohorts. There was no interaction between FT4 and TSH levels for the all stroke events or stroke mortality analyses (P = .099 and P = .28, respectively), as assessed by introducing an interaction term between FT4 (nanograms per deciliter) and TSH values.

We used a random-effects model according to DerSimonian and Laird (29) to pool outcomes estimates (two step approach). Pooled estimates were summarized in forest plots using the metafor package for R (R-project; Institute for Statistics and Mathematics, R Core Team (2013), version 3.0.2, Vienna, Austria). Heterogeneity across studies was measured using the I2 statistic and 95% confidence interval (CI) (30).

The primary analyses were adjusted for age and sex. We also conducted multivariable analyses, additionally adjusting for systolic BP, smoking, total cholesterol, and diabetes. These covariates were available in all cohorts except for the Birmingham cohort, in which none was available (20). We conducted multiple imputation of covariates in cohorts when there was 5% or more of missing data for the smoking, total cholesterol, systolic BP, or prevalent diabetes covariates, which was the case for one study (19). We considered the age- and sex-adjusted analysis the primary analysis because the covariates used in the multivariable analyses could also be considered as mediators and it includes all studies, whereas the multivariable analysis does not include the Birmingham cohort.

To evaluate the robustness of our findings and identify possible sources of heterogeneity and populations at risk, we conducted predefined subgroup and sensitivity analyses. We performed stratified analyses by age, sex, history of stroke, subtype of stroke (including only classified strokes), and race, in concordance with previous reports (1, 4). If the parameter estimates were infinite due to a small number of events in a stratified study-specific analysis, we used Firth's penalized maximum likelihood bias reduction method for the Cox proportional hazards model (31, 32) to estimate HRs and 95% CIs.

For the continuous TSH analyses, we conducted the following sensitivity analyses: 1) excluding participants who had thyroid hormone replacement at baseline and during follow-up, 2) excluding studies that included transient ischemic attack as a stroke event, 3) excluding studies with self-reported stroke data, 4) excluding studies that did not meet the proportional hazard assumption, 5) excluding cohorts with potential comorbidities (eg, diabetes patients), and 6) excluding studies without formal adjudication procedures. We also conducted additional multivariable analyses including prevalent AF, prevalent cardiovascular disease, body mass index or lipid lowering, and antihypertensive therapy at baseline to the previous multivariable model. Furthermore, we performed the following methodological sensitivity analyses: 1) performed the meta-analysis in a two-step approach using the restricted maximum-likelihood estimator also using the metafor package and 2) calculated the risk estimates using a one-step frailty Cox proportional hazards model (coxme package, R-project; Institute for Statistics and Mathematics, R Core Team (2013), version 3.0.2.) We assessed age- and sex-adjusted funnel plots and conducted Egger tests (33) to evaluate potential publication bias statistically. There was no specific funding for this study.

Results

We identified 17 cohorts from the United States (17, 21, 34), Europe (18, 20, 24, 27, 28, 35–40), Australia (25), Brazil (41), and Japan (19) that assessed stroke outcomes prospectively and agreed to share IPD (Table 1). The included studies provided information on a total of 43 598 participants with thyroid function within the reference range and a follow-up from 1972 to 2014, a median follow-up ranging between 1.5 and 20 years, and a total follow-up of 450 684 person-years. All studies, except one (34), included both female (49.6%) and male participants. All cohorts reported fatal stroke and 12 studies reported both fatal and nonfatal stroke, contributing to the all stroke analyses among 34 853 participants. During follow-up, 2271 participants had a stroke, with an incidence rate of 6.3 per 1000 person-years and 907 a fatal stroke with 2.0 per 1000 person-years. The FT4 analyses included 24 888 participants for all stroke and 32 580 for fatal stroke. Two studies (25, 39) used variations of the World Health Organization criteria to define all stroke and fatal stroke (Supplemental Table 1) and four studies included information on type of stroke (hemorrhagic vs ischemic) (17, 21, 28, 40). One study (39) used questionnaires for the assessment of nonfatal stroke. Formal adjudication, defined as having clear criteria for the outcomes that were reviewed by experts for each potential case, was used for all stroke in six studies (17, 21, 28, 36, 42, 43) and for fatal stroke in two additional studies (34, 38).

All but three cohorts had information on participants' race (18, 24, 25). For the additional multivariate analyses, information on AF at baseline was available for eight studies (17, 18, 21, 25, 35, 36, 39, 40, 42). Data on lipid-lowering and hypertensive medications were not available in all but two studies (19, 24). Data on history of cardiovascular disease were not available for two studies (34, 35).

All studies provided information on the proportion of participants taking thyroid hormone medication at baseline. In all but four cohorts, none of the participants used thyroid medication at baseline. In the cohorts in which thyroid medication was used, the proportion varied from 1% to 6%. All but six studies also provided follow-up information on thyroid hormone replacement use, with a range between 0% and 3%.

The association between TSH and the risk of stroke

The age- and sex-adjusted pooled HR for all stroke was 0.78 (95% CI 0.65–0.95, across the reference range of TSH [milliinternational units per liter]) and for fatal stroke 0.83 (95% CI 0.62–1.09) (Figure 1). This corresponds to a 1.28-fold and 1.20-fold increase in all and fatal stroke risk, respectively, for a participant with a TSH in the lower limit of the reference range (0.45 mIU/L) compared with a participant with a TSH in the upper limit of the reference range (4.49 mIU/L). We found no heterogeneity for the analyses of all stroke or fatal stroke analyses (I2 = 0%). Multivariable analyses, adjusting for sex, age, smoking, total cholesterol, systolic BP, and history of diabetes yielded similar results with a HR of 0.76 (95% CI 0.63–0.91) for all stroke and 0.78 (95% CI 0.58–1.07) for fatal stroke (Table 2). Subsequent subgroup analyses did not show a differential risk when stratifying by sex, age groups, history of stroke, or race (Table 2). The information on the type of stroke was available in a subgroup of 11 192 participants in four studies (17, 21, 28, 40). Stratifying by type of stroke showed a lower estimate in hemorrhagic fatal stroke compared with ischemic stroke (HR 0.37, 95% CI 0.12–1.12 vs HR 0.78, 95% CI 0.33–1.80) but with an insignificant P value for interaction (P = .30). Sensitivity analyses excluding specific studies or participants using thyroid hormone replacement therapy did not meaningfully affect the risk estimates (Supplemental Table 2). Additional adjustment for prevalent AF, prevalent cardiovascular disease (defined as previous CHD or stroke), body mass index, or lipid-lowering and antihypertensive therapy did not attenuate the associations. Estimates derived by the methodological sensitivity analyses were similar to the results of the two-step random-effects model according to DerSimonian and Laird (29) (Supplemental Table 3). We did not find any evidence of publication bias for the TSH analyses, either with visual assessment of age- and sex-adjusted funnel plots or with the Egger test for all stroke (P = .75) or fatal stroke (P = .29).

Figure 1.

Figure 1.

Open in a new tab

The association between TSH and risk of all stroke (A) and fatal stroke (B). HRs and their 95% CIs are represented by squares and are across the range of TSH (0.45 and 4.49 mIU/L). Sizes of data markers are proportional to the inverse of the variance of the HRs. Data for all stroke were available in 12 studies. Three hundred ninety-three participants were excluded from the analysis of all stroke due to missing follow-up data. Data for fatal stroke were available in 17 studies. Two hundred sixty-five participants were excluded from the analysis of fatal stroke due to missing cause of death.

Table 2.

Stratified Analyses for the Associations Between TSH and the Risk of All Stroke and Fatal Stroke

All Strokea
 I2 Fatal Strokeb
Events/Total Participants, n Age- and Sex-Adjusted HR (95% CI) Multivariable HR (95% CI)c Events/Total Participants, n Age- and Sex-Adjusted HR (95% CI) Multivariable HR (95% CI)c I2
Total population TSH 2271/34 853 0.78 (0.65, 0.95) 0.76 (0.63, 0.91) 0% 907/43 333 0.83 (0.62, 1.09) 0.78 (0.58, 1.07) 0%
Sexd
    Men 1091/16 723 0.80 (0.62, 1.07) 0.78 (0.60, 1.02) 0% 422/21 874 0.85 (0.50, 1.41) 0.85 (0.50, 1.35) 0%
    Women 1180/18 130 0.78 (0.58, 1.07) 0.75 (0.55, 1.02) 25% 485/21 459 0.80 (0.52, 1.25) 0.80 (0.52, 1.22) 12%
    P for interaction .90 .85 .86 .85
Age, ye
    18–49 60/8305 0.95 (0.31, 2.86) 1.45 (0.37, 4.17) 0% 12/9525 0.71 (0.07, 7.47) 1.14 (0.06, 23.85) 0%
    50–64 358/9145 0.75 (0.47, 1.19) 0.75 (0.47, 1.22) 0% 104/12 303 1.35 (0.55, 3.25) 1.22 (0.48, 3.16) 0%
    65–79 1588/15 667 0.83 (0.67, 1.05) 0.80 (0.63, 1.00) 0% 623/19 198 0.89 (0.62, 1.27) 0.95 (0.85, 1.09) 0%
    ≥80 265/1736 0.69 (0.40, 1.17) 0.63 (0.36, 1.09) 0% 168/2307 0.43 (0.22, 0.85) 0.36 (0.17, 0.78) 0%
    P for trend .66 .28 .61 .43
Stroke historyf
    No 1875/31 626 0.78 (0.63, 0.98) 0.75 (0.60, 0.93) 0% 710/36 222 0.71 (0.47, 1.07) 0.71 (0.47, 1.05) 24%
    Yes 206/1266 1.00 (0.53, 1.83) 1.14 (0.60, 2.20) 0% 92/1440 0.83 (0.26, 2.50) 1.58 (0.47, 5.27) 20%
    P for interaction .47 .23 .80 .22
Stroke typeg
    Hemorrhagic 129/11 192 0.47 (0.26, 0.83) 0.47 (0.25, 0.89) 5% 87/11 192 0.38 (0.14, 1.07) 0.37 (0.12, 1.12) 27%
    Ischemic 817/11 192 0.71 (0.50, 1.00) 0.69 (0.48,0.98) 0% 182/11 192 0.69 (0.34, 1.35) 0.78 (0.33–1.80) 0%
    P for interaction .24 .30 .34 .30
Raceh
    White 1430/19 037 0.76 (0.60, 0.95) 0.73 (0.58, 0.91) 0% 520/23 213 0.71 (0.48, 1.02) 0.67 (0.47 1.00) 0%
    Asian NA NA NA 63/3230 0.48 (0.06, 11.22) 0.62 (0.10, 3.97) 41%
    Black 150/1090 0.85 (0.26, 2.78) 0.91 (0.41, 1.99) 47% 59/1055 0.95 (0.30, 3.12) 0.89 (0.26, 2.91) 0%
    P for interaction .88 .60 .83 .91
Open in a new tab

The HRs are across the reference range of TSH in milliinternational units per liter (0.45–4.49).

a

Data were available from 12 studies, and 393 participants were excluded due to missing stroke event data.

b

265 Participants were excluded due to missing data on cause of death.

c

Adjusted for sex, age, systolic BP, total cholesterol, smoking, and prevalent diabetes at baseline. The Birmingham Study was excluded in this analysis because of lack of data on cardiovascular risk factors.

d

These analyses were not adjusted for sex.

e

These HRs were adjusted for sex and age as continuous variable to avoid residual confounding within age strata.

f

Information on history of stroke was not available for the Pisa cohort, Birmingham Study, and Busselton Health Study. Data concerning history of stroke were missing for 64 participants in total.

g

Information on type of stroke was available for the Cardiovascular Health Study, Health ABC Study, PROSPER, and the Rotterdam Study.

h

Information on race was not available for the 4D study, Birmingham study, Busselton Health Study, and EPIC-Norfolk Study. Ninety-six participants from Osteoporotic Fractures in Men Study were excluded due to no events in subgroup.

The association between FT4 and the risk of stroke

The age- and sex-adjusted pooled HRs for the per-SD increase of FT4 and stroke analyses were 1.08 (95% CI 0.99–1.15) for all stroke and 1.10 (95% CI 1.04–1.19) for fatal stroke (Table 3 and Figure 2). We found substantial heterogeneity for the analyses on all stroke (I2 = 55%) but no heterogeneity for fatal stroke (I2 = 0%). When analyzing the association per 1-ng/dL FT4 increase and risk of stroke, the age- and sex-adjusted pooled HRs were 1.40 (95% CI 0.95–2.05) for all stroke and 1.44 (95% CI 1.10–1.89) for fatal stroke (Supplemental Table 4). Multivariable analyses, adjusting for sex, age, smoking, total cholesterol, systolic BP, and history of diabetes did not change risk estimates substantially (Table 3).

Table 3.

Stratified Analyses for the Associations Between Standardized FT4 and the Risk of All Stroke and Fatal Strokea

All Strokeb
 I2 Fatal Strokec
 I2
Events/Total Participants, n Age- and Sex-Adjusted HR (95% CI) Multivariable HR (95% CI)d Events/Total Participants, n Age- and Sex-Adjusted HR (95% CI) Multivariable HR (95% CI)d
Total population FT4 per SD 1307/24 888 1.08 (0.99, 1.17) 1.06 (0.99, 1.15) 55% 598/32 580 1.10 (1.04, 1,19) 1.09 (1.02, 1.18) 0%
Sexe
    Men 639/11 848 1.02 (0.94, 1.11) 1.00 (0.92, 1.08) 0% 284/16 651 1.10 (0.99, 1.24) 1.08 (0.96, 1.21) 0%
    Women 668/13 040 1.10 (0.99, 1.22) 1.10 (1.01, 1.20) 52% 314/15 929 1.12 (1.03, 1.23) 1.12 (1.01, 1.24) 0%
    P for interaction .27 .12 .79 .65
Age, yf
    18–49 59/8289 0.81 (0.61, 1.07) 0.75 (0.55, 1.03) 0% 12/9507 1.50 (0.62, 3.67) 0.93 (0.32, 2.71) 36%
    50–64 342/9019 1.03 (0.93, 1.29) 1.03 (0.84, 1.27) 66% 99/11 929 1.09 (0.88, 1.35) 1.06 (0.84, 1.32) 0%
    65–79 759/6803 1.12 (1.05, 1.19) 1.10 (1.04, 1.17) 0% 376/9897 1.11 (1.01, 1.22) 1.09 (0.99, 1.21) 0%
    ≥80 147/777 1.15 (0.98, 1.35) 1.15 (0.96, 1.38) 0% 111/1247 1.12 (0.94, 1.33) 1.09 (0.89, 1.33) 0%
    P for trend .024 .015 .54 .76
Stroke historyg
    No 1013/22 446 1.06 (0.95, 1.18) 1.05 (0.95, 1.15) 58% 472/27 256 1.10 (1.02, 1.19) 1.09 (1.00, 1.18) 0%
    Yes 104/483 1.11 (0.95, 1.29) 1.12 (0.95, 1.32) 0% 60/668 1.07 (0.78, 1.45) 1.15 (0.77, 1.73) 26%
    P for interaction .64 .51 .87 .80
Stroke typeh
    Hemorrhagic 17/1577 1.37 (0.82–2.29) 1.15 (0.64–2.07) NA 10/1577 1.15 (0.63–2.12) 1.01 (0.49, 2.07) NA
    Ischemic 157/1577 1.30 (1.14–1.47) 1.20 (1.06–1.37) NA 39/1577 1.00 (0.71–1.41) 0.90 (0.62–1.28) NA
    P for interaction .84 .88 .70 .77
Racei
    White 617/10 208 1.12 (0.99, 1.26) 1.11 (0.99 1.23) 51% 319/14 528 1.13 (1.03, 1.23) 1.10 (1.00 1.21) 0%
    Asian NA NA NA NA 63/3228 1.27 (0.74, 2.18) 1.27 (0.74, 2.18) 58%
    Black NA NA NA NA 2/48 0.94 (0.23, 3.88) 1.00 (0.14, 7.09) NA
    P for interaction NA NA .89 .87
Open in a new tab

Abbreviation: NA, not applicable. The HRs are per one increase in SD of FT4.

a

The Whickham Survey did not measure FT4 but total T4.

b

Data were available from 12 studies, and 384 participants were excluded due to missing stroke event data.

c

Twenty-seven participants were excluded due to missing data on cause of death.

d

Adjusted for sex, age, systolic BP, total cholesterol, smoking, and prevalent diabetes at baseline.

e

These analyses were not adjusted for sex.

f

These HRs were adjusted for sex and age as continuous variable to avoid residual confounding within age strata.

g

Information on stroke history was not available for the Pisa cohort, Birmingham Study, and Busselton Health Study. Data on stroke history were missing for 64 subjects.

h

Information on type of stroke was available for the Rotterdam Study.

i

Information on race was not available for the 4D study, Busselton Health Study, and EPIC-Norfolk Study. Ninety-six participants from the Osteoporotic Fractures in Men Study were excluded due to no events in subgroup.

Figure 2.

Figure 2.

Open in a new tab

The association between standardized FT4 and risk of all stroke (A) and fatal stroke (B). HRs and their 95% CIs are represented by squares and are per one increase of 1 SD of FT4. Sizes of data markers are proportional to the inverse of the variance of the HRs. Data for all stroke were available in nine studies. Three hundred eighty-seven participants were excluded from the analysis of all stroke due to missing follow-up data. Data for fatal stroke were available in 13 studies. Twenty-seven participants were excluded from the analysis of fatal stroke due to missing cause of death.

Subsequent subgroup analyses showed a differential risk for the different age categories, in which the risk estimates went from protective to deleterious with increasing age (P for trend = .024; Table 3). When stratifying by sex, history of stroke, or race, no differential effects were detected. Stratifying for type of stroke also did not show differential risk (Table 3), but this was only possible in one study that was included in the FT4 analyses. We did not find any evidence of publication bias for the FT4 and stroke analyses, either with a visual assessment of age- and sex-adjusted funnel plots or with the Egger test for all stroke (P = .41) or for fatal stroke (P = .28).

Discussion

In the current IPD analysis of 43 598 participants from 17 prospective cohort studies, higher levels of TSH within the reference range of thyroid function were significantly associated with a lower risk of stroke in age- and sex-adjusted and in multivariable analyses. The analyses concerning the association between TSH levels and fatal stroke were qualitatively similar but did not reach statistical significance. The analyses on the association between FT4 and all stroke and fatal stroke support the finding of a higher risk of stroke with differences within the reference range of thyroid function.

Thyroid dysfunction is defined by the biochemical reference ranges for TSH and FT4. These reference ranges, defining the normal range, depend on the assay used, the distribution of thyroid measurements in the population, or both. A thyroid function within the normal range would imply that the levels of circulating thyroid hormone are not accompanied by symptoms, an increased risk of disease, or adverse events. In recent years, the applied reference ranges have been debated in the context of mainly the latter two: adverse events and diseases. Higher levels of TSH within the reference range are associated with an increase in systolic and diastolic BP (44, 45). Moreover, increased TSH levels within the reference range are linearly associated with an unfavorable serum lipid profile (46). On the other hand, lower TSH levels within the reference range are associated with an increased risk of heart failure, CHD, and AF in an elderly population (7). The arbitrary nature of the cutoffs currently used is an important factor hampering decision making on screening and treatment of thyroid dysfunction (13). In the context of defining the reference range of thyroid function, our study provides additional evidence that lower levels of TSH and higher levels of FT4 within the reference range are associated with a negative clinical outcome, namely stroke, a major cause of morbidity and mortality. In contrast to BP or cholesterol, reference ranges for thyroid function are currently based on the distribution in the population rather than the risks of major diseases. It is more challenging to establish reference ranges for thyroid function based on risk of outcomes than for cardiovascular risk factors such as BP and cholesterol, in which the increase in risk mainly occurs for values higher than the upper limit. However, both low and high thyroid function is associated with clinical disease, also within the reference range. Furthermore, a previous study from the TSC provided no evidence for a higher risk of CHD within the normal reference range as currently defined (10). Also, thyroid function is not solely associated with cardiovascular disease but also a wide variety of clinical outcomes including fracture risk and possibly cognitive function decline (7, 14). Therefore, future research should investigate whether reevaluation of the currently used reference ranges for thyroid function is meaningful and, if so, to what extent this should be done for specific populations or subgroups (eg, elderly).

Several pathways could explain the relation between thyroid function and stroke. Thyroid hormone has direct effects on the cardiovascular system and is known to decrease systemic vascular resistance (47), increase left ventricular contractile function, and alter systolic and diastolic cardiac function (48). Differences in thyroid hormone function are associated with the risk of several cardiovascular risk factors including hypertension (49), dyslipidemia (50), and atherosclerosis (51). These changes have also been reported in subjects with subclinical thyroid dysfunction (42) and also some with differences of thyroid function within the reference range (44–46). The fact that adjustment for these cardiovascular risk factors in our multivariable analyses did not substantially alter the risk estimates suggests an effect on the risk of stroke, which is independent of classical risk factors such as hypertension.

Another explanation might be that the lack of effect of a multivariable adjustment is due to residual confounding or unmeasured mediators. For example, in the current analysis, an additional adjustment for AF, a plausible biological mediator for the association between thyroid function and the risk of stroke (52), did not alter risk estimates substantially. However, detecting an effect may have been hampered by the lack of information on prevalent AF in nine studies and insufficient incidence information. There was no sufficient information available on anticoagulant medication use of participants, which did not allow for further exploration of possible mediating and confounding effects.

Various abnormalities in the hemostatic system have been reported in overt (53) and subclinical thyroid dysfunction (54). Hypercoagulability is seen in hyperthyroidism, whereas hypothyroidism has been associated with mainly hypocoagulability (55, 56). Alterations in coagulability and the fibrinolytic system have been linked to a higher risk of cardiovascular disease (57). Whether hemostasis is also affected within the reference range of thyroid function is not known but might be one of the pathways that play a role in the increased risk of stroke associated with differences in thyroid function within the reference range. Changes in coagulation patterns due to thyroid hormone could imply that thyroid function tending toward hyperthyroidism might increase the risk of ischemic stroke mainly. We only had a small subgroup of studies including information on the type of stroke (hemorrhagic vs ischemic), limiting our analysis on type of stroke. The exact mechanism explaining the association between differences in thyroid function within the reference range and the risk of stroke therefore remains to be determined.

Previous studies have reported that the association of thyroid dysfunction with the risk of cardiovascular disease is influenced by age or sex. A study on the association of thyroid disorders and stroke found a decreased risk of ischemic stroke in treated male patients with thyroid disorders, but not in females (58). A study level meta-analysis found that subclinical hypothyroidism was associated with an increased risk of ischemic heart disease and cardiovascular mortality only in younger populations (59). In line with this finding, a study in participants aged 85 years in the general population revealed no adverse effects of abnormally high levels of TSH (36). In contrast, an IPD meta-analysis of 55 287 participants did not show significant trend in the risk of CHD across different age groups (1). In our study, stratification by age, sex, and race did not reveal differential risk patterns. It should, however, be noted that no study to date has looked at the association of thyroid function within the reference range and stroke by age or sex, and this could be one of the reasons for the discrepancies found between previous literature and our study.

The association of TSH with the risk of stroke in participants without a prior history of stroke was similar to the overall analyses, whereas in participants with a prior stroke, the association was not present. The total number of participants with a history of stroke was small, and therefore, the power to detect a possible differential risk between participants with and without history of stroke could have been limited. The risk of all stroke associated with FT4 levels seemed to increase with older age. However, this finding was not replicated in the TSH or fatal stroke analyses.

Strengths of our study include the ability to perform an IPD analysis including 43 598 participants from 17 studies, based on published and unpublished data. By performing an IPD analysis, we were able to standardize the definition of reference range thyroid function and covariates within our study for the analyses. There were differences between the study populations regarding age and sex distribution, among others. Nevertheless, these was limited to no heterogeneity of the outcome estimates between the studies. This could indicate the robustness of the findings.

Despite the large number of participants, we had limited numbers of events in those with a history of stroke, and only four studies included data on the type of stroke. Information needed for certain stratification and sensitivity analyses, eg, by race or prevalent AF, was not available for some cohorts. Also, there was no information available on anticoagulant use or anticoagulant factor levels, hampering analyses concerning possible underlying pathways. Furthermore, TSH and FT4 measurements were performed only at baseline, and data on thyroid medication use during follow-up were not complete, which could change the risk over time, in almost all cohorts, and therefore, it was not possible to take changes of thyroid function over time into account. Residual confounding cannot be excluded, as is the case in all observational studies.

Conclusions

In summary, higher TSH levels within the reference range were associated with a lower risk of all stroke. The analyses for fatal stroke and FT4 were qualitatively similar. These data provide additional evidence that differences within the reference range of thyroid function, as currently defined, are associated with an increased risk of a major adverse event. Future studies should investigate whether reevaluation of the currently used reference ranges for thyroid function, which are based on fixed biochemical parameters instead of health and treatment outcomes and risk of disease and mortality, should be considered. This is pivotal information when designing randomized controlled trials sufficiently equipped to address possible risks and benefits of thyroid function treatment.

Acknowledgments

We gratefully acknowledge the contribution of all studies, study participants, the staff from the all participating studies, and the participating general practitioners, pharmacists and other health care professionals. We express our gratitude to Professor Eric Vittinghoff, PhD (Division of Biostatistics, Department of Epidemiology and Biostatics, University of California, San Francisco, San Francisco, California), for the statistical assistance and Dr Joost van Rosmalen, PhD (Department of Biostatistics, Erasmus Medical Center, Rotterdam, The Netherlands), for the provided support in the statistical analyses. We also thank Wichor M. Bramer (Medical Library, Erasmus Medical Center) for the important contribution to the literature search.

Authors contributions include the following: Drs Chaker and Peeters had full access to all data in all the studies and take responsibility for the integrity of the data and the accuracy of the data analysis. The following authors contributed the following: study concept and design, Peeters and Rodondi; acquisition of data, Chaker, Baumgartner, den Elzen, Ikram, Blum, Bakker, Dehghan, Collet, Drechsler, Luben, Hofman, Portegies, Medici, Iervasi, Stott, Ford, Bremner, Wanner, Ferrucci, Newman, Dullaart, Sgarbi, Dörr, Longstreth, Psaty, Ceresini, Maciel, Westendorp, Jukema, Imaizumi, Franklyn, Bauer, Walsh, Razvi, Khaw, Cappola, Völzke, Franco, Gussekloo, Rodondi, and Peeters; analysis and interpretation of the data, Chaker, Baumgartner, Peeters, Dehghan, Franco, Ikram, Collet, Blum, and Rodondi; drafting of the manuscript, Chaker and Peeters; critical revision of the manuscript for important intellectual content, Chaker, Baumgartner, den Elzen, Ikram, Blum, Bakker, Dehghan, Collet, Drechsler, Luben, Hofman, Portegies, Medici, Iervasi, Stott, Ford, Bremner, Wanner, Ferrucci, Newman, Dullaart, Sgarbi, Dörr, Longstreth, Psaty, Ceresini, Maciel, Westendorp, Jukema, Imaizumi, Franklyn, Bauer, Walsh, Razvi, Khaw, Cappola, Völzke, Franco, Gussekloo, Rodondi, and Peeters; statistical analysis, Chaker and Peeters; obtained funding, Peeters, Franco, and Rodondi; administrative, technical, or material support, Chaker, Blum, Portegies, Medici, den Elzen, and Rodondi; and study supervision: Peeters, Franco, Dehghan, Ikram, and Rodondi.

The views of the authors do not necessarily reflect those of the governments of Japan and the United States.

R.P.P. and L.C. are supported by an Erasmus Medical Center MRACE grant and a ZonMW TOP grant (Grant 91212044). R.P.P. has received lecture and consultancy fees from Genzyme B.V. T.-H.C.'s research is supported by Grants PBLAP3-145870 and P3SMP3-155318 from the Swiss National Science Foundation. B.M.P. serves on a DSMB for a clinical trial of a device funded by the manufacturer (Zoll LifeCor) and on the Steering Committee of the Yale Open Data Access Project funded by Johnson & Johnson. The Thyroid Studies Collaboration is supported by Grants SNSF 320030-138267 and 320030-150025 from the Swiss National Science Foundation (to N.R.).

Cohort-related funding included the Cardiovascular Health Study supported by Grants R01AG032317 and K24AG042765 from the National Institute on Aging; Contracts HHSN268201200036C, HHSN268200800007C, N01HC55222, N01HC85079, N01HC85080, N01HC85081, N01HC85082, N01HC85083, and N01HC85086 and Grant HL080295 from the National Heart, Lung, and Blood Institute, with additional contribution from the National Institute of Neurological Disorders and Stroke. Additional support was provided by Grant AG023629 from the National Institute on Aging. A full list of the principal Cardiovascular Health Study investigators and institutions can be found at chs-nhlbi.org.

The Health ABC Study was supported by National Institute on Aging Contracts N01-AG-6-2101, N01-AG-6-2103, and N01-AG-6-2106; National Institute on Aging Grant R01-AG028050; and National Institute of Nursing Research Grant R01-NR012459. This research was supported in part by the Intramural Research Program of the National Institutes of Health, National Institute on Aging.

The Study of Health in Pomerania analyses were funded by the Research Network of Community Medicine of the University Medicine Greifswald (www.community-medicine.de) and the German Research Foundation Grant DFG Vo 955/12-2.

The Radiation Effects Research Foundation (Hiroshima and Nagasaki, Japan) is a public interest foundation supported by the Japanese Ministry of Health, Labor, and Welfare and the US Department of Energy. This publication was supported by Radiation Effects Research Foundation Research Protocol Grant A3-13.

The Brazilian Thyroid Study was supported by an unrestricted grant from the Sao Paulo State Research Foundation (Fundação de Amaparo a Pesquisa do Estado de Sao Paulo) Grant 6/59737-9 (to R.M.B.M.).

The Dutch Kidney Foundation supported the infrastructure of the PREVEND program from 1997 to 2003 (Grant E.033). The University Medical Center Groningen supported the infrastructure from 2003 to 2006. Dade Behring, Ausam, Roche, and Abbott financed laboratory equipment and reagents by which various laboratory determinations could be performed. The Dutch Heart Foundation supported studies on lipid metabolism (Grant 2001-005).

The Osteoporotic Fractures in Men Study is supported by the National Institutes of Health. The following institutes provide support: the National Institute on Aging, the National Institute of Arthritis and Musculoskeletal and Skin Diseases, the National Center for Advancing Translational Sciences, and National Institutes of HealthRoadmap for Medical Research under the following grants: U01 AG027810, U01 AG042124, U01 AG042139, U01 AG042140, U01 AG042143, U01 AG042145, U01 AG042168, U01 AR066160, and UL1 TR000128.

Disclosure Summary: The authors have nothing to disclose.

Footnotes

Abbreviations:
AF
atrial fibrillation
BP
blood pressure
CHD
coronary heart disease
CI
confidence intervals
FT4
free T4
4D
Die Deutsche Diabetes Dialyze
HR
hazard ratio
IPD
individual participant data
PROSPER
Pravastatin in Elderly Individuals at Risk of Vascular Disease
TSC
Thyroid Studies Collaboration.

References

  • 1. Rodondi N, den Elzen WP, Bauer DC, et al. Subclinical hypothyroidism and the risk of coronary heart disease and mortality. JAMA. 2010;304:1365–1374. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2. Bauer DC, Ettinger B, Browner WS. Thyroid functions and serum lipids in older women: a population-based study. Am J Med. 1998;104:546–551. [DOI] [PubMed] [Google Scholar]
  • 3. Biondi B, Cooper DS. The clinical significance of subclinical thyroid dysfunction. Endocr Rev. 2008;29:76–131. [DOI] [PubMed] [Google Scholar]
  • 4. Collet TH, Gussekloo J, Bauer DC, et al. Subclinical hyperthyroidism and the risk of coronary heart disease and mortality. Arch Intern Med. 2012;172:799–809. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5. Surks MI, Boucai L. Age- and race-based serum thyrotropin reference limits. J Clin Endocrinol Metab. 2010;95:496–502. [DOI] [PubMed] [Google Scholar]
  • 6. Wartofsky L, Dickey RA. The evidence for a narrower thyrotropin reference range is compelling. J Clin Endocrinol Metab. 2005;90:5483–5488. [DOI] [PubMed] [Google Scholar]
  • 7. Cappola AR, Arnold AM, Wulczyn K, Carlson M, Robbins J, Psaty BM. Thyroid function in the euthyroid range and adverse outcomes in older adults. J Clin Endocrinol Metab. 2015;100(3):1088–1096. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8. Taylor PN, Razvi S, Pearce SH, Dayan CM. Clinical review: a review of the clinical consequences of variation in thyroid function within the reference range. J Clin Endocrinol Metab. 2013;98:3562–3571. [DOI] [PubMed] [Google Scholar]
  • 9. van Tienhoven-Wind LJ, Dullaart RP. Low-normal thyroid function and the pathogenesis of common cardio-metabolic disorders. Eur J Clin Invest. 2015;45:494–503. [DOI] [PubMed] [Google Scholar]
  • 10. Asvold BO, Vatten LJ, Bjoro T, et al. Thyroid function within the normal range and risk of coronary heart disease: an individual participant data analysis of 14 cohorts. JAMA Intern Med. 2015;175(6):1037–1047. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11. Feigin VL, Forouzanfar MH, Krishnamurthi R, et al. Global and regional burden of stroke during 1990–2010: findings from the Global Burden of Disease Study 2010. Lancet. 2014;383:245–254. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12. Chaker L, Baumgartner C, Ikram MA, et al. Subclinical thyroid dysfunction and the risk of stroke: a systematic review and meta-analysis Eur J Epidemiol. 2014;29(11):791–800. [DOI] [PubMed] [Google Scholar]
  • 13. LeFevre ML, US Preventive Services Task Force. Screening for thyroid dysfunction: US Preventive Services Task Force recommendation statement. Ann Intern Med. 2015;162:641–650. [DOI] [PubMed] [Google Scholar]
  • 14. Blum MR, Bauer DC, Collet TH, et al. Subclinical thyroid dysfunction and fracture risk: a meta-analysis. JAMA. 2015;313:2055–2065. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15. Chaker L, Baumgartner C, den Elzen WP, et al. Subclinical hypothyroidism and the risk of stroke events and fatal stroke: an individual participant data analysis. J Clin Endocrinol Metab. 2015;100:2181–2191. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16. Gencer B, Collet TH, Virgini V, et al. Subclinical thyroid dysfunction and the risk of heart failure events: an individual participant data analysis from 6 prospective cohorts. Circulation. 2012;126:1040–1049. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17. Cappola AR, Fried LP, Arnold AM, et al. Thyroid status, cardiovascular risk, and mortality in older adults. JAMA. 2006;295:1033–1041. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18. Drechsler C, Schneider A, Gutjahr-Lengsfeld L, et al. Thyroid function, cardiovascular events, and mortality in diabetic hemodialysis patients. Am J Kidney Dis. 2014;63(6):988–996. [DOI] [PubMed] [Google Scholar]
  • 19. Imaizumi M, Akahoshi M, Ichimaru S, et al. Risk for ischemic heart disease and all-cause mortality in subclinical hypothyroidism. J Clin Endocrinol Metab. 2004;89:3365–3370. [DOI] [PubMed] [Google Scholar]
  • 20. Parle JV, Maisonneuve P, Sheppard MC, Boyle P, Franklyn JA. Prediction of all-cause and cardiovascular mortality in elderly people from one low serum thyrotropin result: a 10-year cohort study. Lancet. 2001;358:861–865. [DOI] [PubMed] [Google Scholar]
  • 21. Rodondi N, Newman AB, Vittinghoff E, et al. Subclinical hypothyroidism and the risk of heart failure, other cardiovascular events, and death. Arch Intern Med. 2005;165:2460–2466. [DOI] [PubMed] [Google Scholar]
  • 22. Helfand M, US Preventive Services Task Force. Screening for subclinical thyroid dysfunction in nonpregnant adults: a summary of the evidence for the US Preventive Services Task Force. Ann Intern Med. 2004;140:128–141. [DOI] [PubMed] [Google Scholar]
  • 23. Surks MI, Ortiz E, Daniels GH, et al. Subclinical thyroid disease: scientific review and guidelines for diagnosis and management. JAMA. 2004;291:228–238. [DOI] [PubMed] [Google Scholar]
  • 24. Boekholdt SM, Titan SM, Wiersinga WM, et al. Initial thyroid status and cardiovascular risk factors: the EPIC-Norfolk prospective population study. Clin Endocrinol (Oxf). 2010;72:404–410. [DOI] [PubMed] [Google Scholar]
  • 25. Walsh JP, Bremner AP, Bulsara MK, et al. Subclinical thyroid dysfunction as a risk factor for cardiovascular disease. Arch Intern Med. 2005;165:2467–2472. [DOI] [PubMed] [Google Scholar]
  • 26. Nicoloff JT, Spencer CA. Clinical review 12: The use and misuse of the sensitive thyrotropin assays. J Clin Endocrinol Metab. 1990;71:553–558. [DOI] [PubMed] [Google Scholar]
  • 27. Razvi S, Weaver JU, Vanderpump MP, Pearce SH. The incidence of ischemic heart disease and mortality in people with subclinical hypothyroidism: reanalysis of the Whickham Survey cohort. J Clin Endocrinol Metab. 2010;95:1734–1740. [DOI] [PubMed] [Google Scholar]
  • 28. Nanchen D, Gussekloo J, Westendorp RG, et al. Subclinical thyroid dysfunction and the risk of heart failure in older persons at high cardiovascular risk. J Clin Endocrinol Metab. 2012;97:852–861. [DOI] [PubMed] [Google Scholar]
  • 29. DerSimonian R, Laird N. Meta-analysis in clinical trials. Controlled Clinical Trials. 1986;7:177–188. [DOI] [PubMed] [Google Scholar]
  • 30. Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat Med. 2002;21:1539–1558. [DOI] [PubMed] [Google Scholar]
  • 31. Firth D. Bias reduction of maximum likelihood estimates. Biometrika. 1993;80:27–38. [Google Scholar]
  • 32. Heinze G, Schemper M. A solution to the problem of monotone likelihood in Cox regression. Biometrics. 2001;57:114–119. [DOI] [PubMed] [Google Scholar]
  • 33. Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. BMJ. 1997;315:629–634. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34. Waring AC, Harrison S, Samuels MH, et al. Thyroid function and mortality in older men: a prospective study. J Clin Endocrinol Metab. 2012;97:862–870. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35. Ferrucci L, Bandinelli S, Benvenuti E, et al. Subsystems contributing to the decline in ability to walk: bridging the gap between epidemiology and geriatric practice in the InCHIANTI study. J Am Geriatr Soc. 2000;48:1618–1625. [DOI] [PubMed] [Google Scholar]
  • 36. Gussekloo J, van Exel E, de Craen AJ, Meinders AE, Frolich M, Westendorp RG. Thyroid status, disability and cognitive function, and survival in old age. JAMA. 2004;292:2591–2599. [DOI] [PubMed] [Google Scholar]
  • 37. Hillege HL, Janssen WM, Bak AA, et al. Microalbuminuria is common, also in a nondiabetic, nonhypertensive population, and an independent indicator of cardiovascular risk factors and cardiovascular morbidity. J Intern Med. 2001;249:519–526. [DOI] [PubMed] [Google Scholar]
  • 38. Iervasi G, Molinaro S, Landi P, et al. Association between increased mortality and mild thyroid dysfunction in cardiac patients. Arch Intern Med. 2007;167:1526–1532. [DOI] [PubMed] [Google Scholar]
  • 39. Volzke H, Alte D, Schmidt CO, et al. Cohort profile: the study of health in Pomerania. Int J Epidemiol. 2011;40:294–307. [DOI] [PubMed] [Google Scholar]
  • 40. Hofman A, Brusselle GG, Darwish Murad S, et al. The Rotterdam Study: 2016 objectives and design update. Eur J Epidemiol. 2015;30:661–708. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41. Sgarbi JA, Matsumura LK, Kasamatsu TS, Ferreira SR, Maciel RM. Subclinical thyroid dysfunctions are independent risk factors for mortality in a 7.5-year follow-up: the Japanese-Brazilian thyroid study. Eur J Endocrinol. 2010;162:569–577. [DOI] [PubMed] [Google Scholar]
  • 42. Hak AE, Pols HA, Visser TJ, Drexhage HA, Hofman A, Witteman JC. Subclinical hypothyroidism is an independent risk factor for atherosclerosis and myocardial infarction in elderly women: the Rotterdam Study. Ann Intern Med. 2000;132:270–278. [DOI] [PubMed] [Google Scholar]
  • 43. Schultz M, Kistorp C, Raymond I, et al. Cardiovascular events in thyroid disease: a population based, prospective study. Horm Metab Res. 2011;43:653–659. [DOI] [PubMed] [Google Scholar]
  • 44. Asvold BO, Bjoro T, Nilsen TI, Vatten LJ. Association between blood pressure and serum thyroid-stimulating hormone concentration within the reference range: a population-based study. J Clin Endocrinol Metab. 2007;92:841–845. [DOI] [PubMed] [Google Scholar]
  • 45. Iqbal A, Figenschau Y, Jorde R. Blood pressure in relation to serum thyrotropin: the Tromso study. J Hum Hypertens. 2006;20:932–936. [DOI] [PubMed] [Google Scholar]
  • 46. Asvold BO, Vatten LJ, Nilsen TI, Bjoro T. The association between TSH within the reference range and serum lipid concentrations in a population-based study. The HUNT Study. Eur J Endocrinol. 2007;156:181–186. [DOI] [PubMed] [Google Scholar]
  • 47. Klein I, Ojamaa K. Thyroid hormone: targeting the vascular smooth muscle cell. Circ Res. 2001;88:260–261. [DOI] [PubMed] [Google Scholar]
  • 48. Klein I, Ojamaa K. Thyroid hormone and the cardiovascular system. N Engl J Med. 2001;344:501–509. [DOI] [PubMed] [Google Scholar]
  • 49. Nagasaki T, Inaba M, Kumeda Y, et al. Increased pulse wave velocity in subclinical hypothyroidism. J Clin Endocrinol Metab. 2006;91:154–158. [DOI] [PubMed] [Google Scholar]
  • 50. Duntas LH. Thyroid disease and lipids. Thyroid. 2002;12:287–293. [DOI] [PubMed] [Google Scholar]
  • 51. Cappola AR, Ladenson PW. Hypothyroidism and atherosclerosis. J Clin Endocrinol Metab. 2003;88:2438–2444. [DOI] [PubMed] [Google Scholar]
  • 52. Heeringa J, Hoogendoorn EH, van der Deure WM, et al. High-normal thyroid function and risk of atrial fibrillation: the Rotterdam study. Arch Intern Med. 2008;168:2219–2224. [DOI] [PubMed] [Google Scholar]
  • 53. Erem C. Coagulation and fibrinolysis in thyroid dysfunction. Endocrine. 2009;36:110–118. [DOI] [PubMed] [Google Scholar]
  • 54. Jorde R, Figenschau Y, Hansen JB. Haemostatic function in subjects with mild subclinical hypothyroidism. The Tromso study. Thromb Haemost. 2006;95:750–751. [PubMed] [Google Scholar]
  • 55. Erem C. Thyroid disorders and hypercoagulability. Semin Thromb Hemost. 2011;37:17–26. [DOI] [PubMed] [Google Scholar]
  • 56. Squizzato A, Romualdi E, Buller HR, Gerdes VE. Clinical review: thyroid dysfunction and effects on coagulation and fibrinolysis: a systematic review. J Clin Endocrinol Metab. 2007;92:2415–2420. [DOI] [PubMed] [Google Scholar]
  • 57. Wiman B, Andersson T, Hallqvist J, Reuterwall C, Ahlbom A, deFaire U. Plasma levels of tissue plasminogen activator/plasminogen activator inhibitor-1 complex and von Willebrand factor are significant risk markers for recurrent myocardial infarction in the Stockholm Heart Epidemiology Program (SHEEP) study. Arterioscl Thromb Vasc Biol. 2000;20:2019–2023. [DOI] [PubMed] [Google Scholar]
  • 58. Six-Merker J, Meisinger C, Jourdan C, et al. Treatment of thyroid dysfunctions decreases the risk of cerebrovascular events in men but not in women: results of the MONICA/KORA cohort study. PloS One. 2016;11:e0155499. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 59. Razvi S, Shakoor A, Vanderpump M, Weaver JU, Pearce SH. The influence of age on the relationship between subclinical hypothyroidism and ischemic heart disease: a metaanalysis. J Clin Endocrinol Metab. 2008;93:2998–3007. [DOI] [PubMed] [Google Scholar]

Articles from The Journal of Clinical Endocrinology and Metabolism are provided here courtesy of The Endocrine Society

RESOURCES