Thyroid Function in the Euthyroid Range and Adverse Outcomes in Older Adults

Abstract

Context:

The appropriateness of current reference ranges for thyroid function testing in older adults has been questioned.

Objective:

This study aimed to determine the relationship between thyroid function tests within the euthyroid range and adverse outcomes in older adults not taking thyroid medication.

Design, Setting, and Participants:

US community-dwelling adults years of older (n = 2843) enrolled onto the Cardiovascular Health Study with TSH, free T₄ (FT₄), and total T₃ concentrations in the euthyroid range.

Main Outcome Measures:

Incidence of atrial fibrillation, coronary heart disease, heart failure, hip fracture, dementia, and all-cause death were measured.

Results:

No departures from linearity were detected. Higher TSH was negatively associated (P = .03) and higher FT₄ was positively associated (P = .007) with mortality. Higher FT₄ was associated with atrial fibrillation (P < .001) and heart failure (P = .004). Compared with the first quartile, individuals with TSH in the fourth quartile had a 9.6 per 1000 person-year lower incidence of dementia (P < .05) and those with FT₄ in the fourth quartile had higher incidences of atrial fibrillation, coronary heart disease, heart failure, and mortality (11.0, 8.0, 7.8, and 14.3 per 1000 person-years, respectively, all P < .05). Total T₃ was not associated with any outcome.

Conclusions:

Higher TSH and lower FT₄ concentrations within the euthyroid range are associated with lower risk of multiple adverse events in older people, including mortality. This suggests tolerance for lower thyroid hormone levels in this age group. Clinical trials are needed to evaluate the risk-benefit profile of new thresholds for initiating treatment and optimal target concentrations for thyroid hormone replacement in older people.

The definition of euthyroid or “normal” thyroid function is based on the 95% confidence interval (CI) of the TSH distribution in a young adult population without thyroid disease. Individuals with TSH concentrations outside the reference range and normal free T₄ (FT₄) concentrations are labeled as having subclinical hypothyroidism if the TSH concentration is above the reference range or subclinical hyperthyroidism if the TSH concentration is below it. Numerous observational studies have examined the relationship between subclinical thyroid dysfunction and adverse outcomes, using euthyroid individuals as the reference group (1). These studies have shown increased risk of atrial fibrillation (AF), fracture, and dementia in subclinical hyperthyroidism (2–5) and coronary heart disease (CHD) and heart failure in subclinical hypothyroidism, at higher levels of TSH (6, 7).

From a physiologic standpoint, it is likely that there is a continuum of effects across the spectrum of thyroid function, including within the designated euthyroid range. These effects may be opposing and organ specific at different ends of the thyroid function spectrum. For example, higher levels of thyroid hormones may adversely stimulate the heart, increasing the risk of AF, while simultaneously improving metabolism (8). Several large cohort studies have examined, individually, the relationship between thyroid function tests within the euthyroid range and AF (9, 10), CHD (11–13), nonvertebral fracture (14–16), dementia (17–20), and mortality (13, 21–24). What is lacking, and of clinical relevance, is the simultaneous examination of cardiovascular, musculoskeletal, and cognitive outcomes in the same cohort. Furthermore, few longitudinal studies have focused on older individuals, in whom the baseline risk of adverse events is high and mild differences in thyroid function could be the tipping point.

Although TSH is the primary thyroid function test in clinical use, FT₄ concentrations have been shown in some analyses to predict adverse events better than TSH concentrations do (18, 19, 22, 25). We sought to examine the relationships, individually, between TSH, FT₄, and total T₃ concentrations and incident AF, CHD, heart failure, hip fracture, dementia, and total mortality in euthyroid men and women enrolled onto the Cardiovascular Health Study (CHS), a cohort of community-dwelling men and women age 65 years and older.

Materials and Methods

Study participants

CHS is a multicenter, population-based cohort study of risk factors for cardiovascular disease and stroke in older adults (26). The original cohort of 5201 participants was sampled from Medicare eligibility lists of men and women age 65 years and older from four US centers (Forsyth County, North Carolina; Sacramento County, California; Washington County, Maryland; and Allegheny County, Pennsylvania) in 1989–1990. In 1992–1993, 687 additional African-American participants were recruited. To be eligible individuals had to be noninstitutionalized, expecting to remain in the area for 3 years, not in active treatment for cancer, not wheelchair bound at home, not requiring a proxy respondent at entry, and capable of providing medical consent. Eligible household members of the sampled individual were recruited. Institutional review boards at all four sites and the coordinating center at the University of Washington approved the study. All participants gave written informed consent.

Annual study visits through 1999 included a detailed medical history, examination of medication bottles, physical examination, and assessment of health status. Blood was stored in −70°C freezers for future testing. Participants have been contacted semiannually (by phone or in-person at study visits) regarding hospitalizations or new occurrences of cardiovascular events since the study's inception. When a participant reports a hospitalization or cardiovascular event, the field center obtains medical records, including discharge abstracts, for endpoint assessments. Greater than 90% of participants were retained through the end of study visits in 1999, and retention remains over 85% in ongoing followup. Additional outcome data have been obtained in all participants, including those lost to followup, using data collected from Centers for Medicare and Medicaid Services (CMS). Ascertainment of mortality is 100%.

Thyroid function measurements

Serum TSH (third generation assay), FT₄, total T₃, and antithyroid peroxidase antibody concentrations (TPOAb) were measured in 3996 participants in 2010 from banked samples obtained at the 1992–1993 visit. TSH and FT₄ can be analyzed reliably in samples stored for up to 23 years (27). Assays were performed at the CHS Central Blood Analysis Laboratory at the University of Vermont using chemiluminescent immunoassays on the Elecsys 2010 analyzer (Roche Diagnostics) as previously described (25). Reference ranges were 0.45–4.5 mU/L for TSH, 9.2–21.8 pmol/L (0.7–1.7 ng/dL) for FT₄, 1.3–3.1 nmol/L (85–201 ng/dL) for T₃, and at least 37 IU/L for TPOAb positivity.

Outcomes

Atrial fibrillation was determined from annual electrocardiography through 1999; hospital discharge codes 427.3, 427.31, or 427.32; and diagnosis from outpatient or carrier claims from CMS through 2010. Coronary heart disease was defined as the occurrence of angina, myocardial infarction, coronary angioplasty, bypass surgery, or coronary heart death. Provisional CHD and heart failure diagnoses based on interview, review of medical records, and other support documents were reviewed and adjudicated by the study's morbidity and mortality subcommittee (28). A composite cardiovascular outcome was defined as CHD, heart failure, or AF. Diagnoses of hip fractures were ascertained from hospital discharge codes and CMS inpatient data. A committee of neurologists, psychiatrists, and neuropsychologists classified dementia status using data collected at study visits, as defined in the CHS Memory study (29). Information about deaths was obtained through medical records, death certificates, autopsy, and coroner reports. Atrial fibrillation events were considered through June 30, 2009, dementia events through the 1998–1999 visit, and cardiovascular events, hip fractures, and deaths through December 31, 2010.

Statistical analysis

Participants whose endogenous TSH, FT₄, and T₃ concentrations were all within reference ranges at the 1992–1993 visit were included in analyses. Participants with prevalent disease for a specific outcome were excluded from analyses of the incidence of that outcome. In addition, participants taking bisphosphonates (n = 7) were excluded from the hip fracture models. We created spline plots to explore the shape of the association of each thyroid function test with each outcome and used generalized additive models to test for statistically significant departures from linearity. Incidence rates were calculated for each outcome by quartile of each thyroid test. Cox proportional hazards models were used to assess associations of each thyroid function test with the incidence of each outcome and a composite cardiovascular outcome. Modeling was performed in stages, adjusting for age, sex, race, and initiation of thyroid hormone replacement as a time-dependent covariate in Model 1. Outcome-specific covariates, selected based on prior analyses in the CHS cohort, were added in Model 2. The staged approach was used because many of the traditional outcome-specific covariates could be mediators, rather than confounders, of the association between thyroid function and the outcome, raising concern for potential overadjustment in Model 2. TSH was modeled with and without log transformation; results did not differ, and the untransformed model is presented for greater interpretability. Proportional hazards assumptions were checked using Schoenfeld residuals and no violations were found. All models were tested for interaction with age, sex, and race by examining the statistical significance of interaction terms with a Wald test. Sensitivity analyses were performed excluding participants with positive TPOAb tests, to examine associations in those without autoimmunity. Competing risks analyses were performed for all outcomes except mortality, with death as the competing outcome. Statistical analyses were performed using STATA version 12 (StataCorp LP). Spline plots were prepared using R version 2.13.0 (http://www.r-project.org).

Results

Of 3996 participants with thyroid function testing in 1992–1993, 3264 (82%) were in the euthyroid TSH range. After excluding participants taking thyroid hormone preparations (n = 216), antithyroid medication (n = 1), or amiodarone (n = 0); missing FT₄ or T₃ measurements (n = 43); or with FT₄ or T₃ measurements outside the reference range (n = 161), 2843 participants remained for analysis. These euthyroid participants had a mean age of 74.5 years, comprised 56.2% women, and 81.4% were Caucasian race (Table 1). The mean TSH concentration was 2.15 mU/L, FT₄ was 15.6 pmol/L (1.2 ng/dL), and total T₃ was 1.81 nmol/L (117.8 ng/dL); 7.8% were TPOAb positive.

Table 1.

Characteristics of the Euthyroid Cohort

Characteristic	Summary
Age, y	74.5 (5.1)
Female	1599 (56.2)
Caucasian	2313 (81.4)
Current smoker	317 (11.2)
Any alcohol use	1288 (45.3)
BMI, kg/m2	26.8 (4.7)
Diabetes	491 (17.3)
Coronary heart disease	628 (22.1)
Heart failure	137 (4.8)
Stroke	149 (5.2)
Atrial fibrillation	170 (6.0)
TPO Ab positive	217 (7.8)
TSH, mU/L	2.15 (0.97)
FT4, pmol/L	15.6 (2.1)
T3, nmol/L	1.81 (0.29)

Abbreviation: BMI, body mass index.

n = 2843. Data are mean (sd) or N (%).

Repeat thyroid function tests 2 years later were available in 80% of participants; 95% (2166/2279) remained euthyroid, 1.1% (26/2279) became subclinically hyperthyroid, 3.4% (77/2279) became subclinically hypothyroid, 2 (<1%) became overtly hypothyroid, and eight (<1%) initiated thyroid medication. Through the end of followup, 133 participants (4.7%) initiated thyroid hormone replacement and two initiated antithyroid medication.

Figure 1 displays the risk of each outcome within the reference range of each thyroid test, adjusted for age, sex, and race. No statistically significant departures from linearity were detected. Over 17 years of followup, 812 participants developed AF. TSH and T₃ concentrations were not associated with incident AF, whereas FT₄ concentration was (Model 1: hazard ratio [HR], 1.06; 95% CI, 1.03–1.10; P < .001; Model 2: HR, 1.04; 95% CI, 1.01–1.08; P = .02) (Table 2). Incident CHD occurred in 786 participants. No association between any of the thyroid function tests and incident CHD was found. Incident heart failure occurred in 870 participants. TSH and T₃ concentrations were not associated with incident heart failure, whereas FT₄ concentration was positively associated (Model 1: HR, 1.05; 95% CI, 1.02–1.09; P = .004; Model 2: HR, 1.04; 95% CI, 1.00–1.07; P = .03). TSH concentration was negatively associated (Model 2: HR, 0.94; 95% CI, 0.88–1.00; P = .04) and FT₄ concentration positively associated (Model 1: HR, 1.04; 95% CI, 1.01–1.07; P = .008; Model 2: HR, 1.03; 95% CI, 1.00–1.06; P = .02) with a composite measure of cardiovascular disease.

Figure 1.

Spline regression graphs depicting the adjusted associations of continuous concentrations of TSH (left), FT₄ (middle), and T₃ (right) with incident AF, CHD, heart failure (HF), hip fracture (Hip Fx), dementia, and total mortality. Spline regressions are adjusted for age, sex, and race.

Table 2.

Risk of Atrial Fibrillation, Coronary Heart Disease, Heart Failure, or a Composite Cardiovascular Disease Measure for a One Unit Higher TSH, FT₄, or Total T₃

Model	No. of Events/No. at Risk	TSH, mU/L			Free T4, pmol/L			Total T3, nmol/L
		HR	95% CI	P Value	HR	95% CI	P Value	HR	95% CI	P Value
Atrial fibrillation	812/2673
Model 1		0.94	0.87–1.01	.09	1.06	1.03–1.10	<.001	0.90	0.70–1.15	.40
Model 2		0.94	0.87–1.02	.12	1.04	1.01–1.08	.02	0.86	0.67–1.11	.24
Coronary heart disease	786/2215
Model 1		0.95	0.88–1.03	.20	1.03	0.99–1.06	.13	1.23	0.97–1.56	.09
Model 2		0.95	0.88–1.02	.15	1.02	0.98–1.05	.28	1.13	0.89–1.45	.31
Heart failure	870/2706
Model 1		0.95	0.88–1.02	.17	1.05	1.02–1.09	.004	1.07	0.85–1.35	.54
Model 2		0.94	0.87–1.01	.09	1.04	1.00–1.07	.03	1.04	0.82–1.32	.77
Composite cardiovascular outcome	1128/2073
Model 1		0.94	0.88–1.00	.05	1.04	1.01–1.07	.008	1.09	0.89–1.33	.43
Model 2		0.94	0.88–1.00	.04	1.03	1.00–1.06	.02	0.97	0.79–1.19	.78

Model 1 for all outcomes is adjusted for age, sex, race and initiation of thyroid meds (time-varying).

Model 2 for atrial fibrillation additionally adjusted for BMI, hypertension, fasting glucose, current smoking, any alcohol use, valvular disease history, and use of any diuretic or β blocker.

Model 2 for coronary heart disease additionally adjusted for BMI, hypertension, diabetes, current smoking, low-density lipoprotein cholesterol, lipid lowering medication use, and C-reactive protein.

Model 2 for heart failure additionally adjusted for BMI, hypertension, diabetes, current smoking, coronary heart disease, atrial fibrillation, stroke, claudication, any alcohol use, and creatinine (natural log).

Model 2 for composite cardiovascular disease outcome additionally adjusted for BMI, hypertension, diabetes, stroke, claudication, current smoking, C-reactive protein, and creatinine.

No association was found between any of the thyroid function tests and hip fracture (340 events) or dementia (229 events) (Table 3). There were a total of 2161 deaths. In Model 1, TSH concentration was negatively associated (HR, 0.95; 95% CI, 0.91–0.99; P = .03) and FT₄ concentration was positively associated (HR, 1.03; 95% CI, 1.01–1.05; P = .007) with total mortality. These estimates were no longer statistically significant after additional adjustment (Model 2: HR, 0.96; 95% CI, 0.91–1.00; P = .05 for TSH, and HR, 1.01; 95% CI, 0.99–1.03; P = .17 for FT₄).

Table 3.

Risk of Hip Fracture, Dementia, and Total Mortality for a One Unit Higher TSH, FT₄, or Total T₃

Model	No. of Events/No. at Risk	TSH, mU/L			Free T4, pmol/L			Total T3, nmol/L
		HR	95% CI	P Value	HR	95% CI	P Value	HR	95% CI	P Value
Hip fracture	340/2803
Model 1		0.93	0.83–1.04	.20	0.99	0.94–1.05	.79	1.07	0.74–1.55	.73
Model 2		0.94	0.84–1.05	.26	0.98	0.94–1.04	.56	1.03	0.71–1.51	.86
Dementia	229/1824
Model 1		0.87	0.76–1.00	.05	1.04	0.97–1.10	.26	1.01	0.63–1.59	.98
Model 2		0.88	0.76–1.01	.07	1.02	0.96–1.09	.52	1.06	0.66–1.68	.81
Total Mortality	2161/2843
Model 1		0.95	0.91–0.99	.03	1.03	1.01–1.05	.007	1.12	0.97–1.30	.12
Model 2		0.96	0.91–1.00	.05	1.01	0.99–1.03	.17	1.12	0.96–1.30	.15

Model 1 for hip fracture and total mortality are adjusted for age, sex, race, and initiation of thyroid meds (time varying). Model 1 for dementia also includes an age² term.

Model 2 for hip fracture additionally adjusted for oral steroid use, kilocalories of exercise (ln), current smoking, alcohol use, and weight.

Model 2 for dementia additionally adjusted for body mass index (BMI), hypertension, diabetes, current smoking, any alcohol use, total cholesterol, stroke history, claudication, and apoe4 genotype.

Model 2 for mortality additionally adjusted for BMI, hypertension, diabetes, current smoking, coronary heart disease, heart failure, stroke, claudication, depression score, creatinine, and kcal of activity.

There were no interactions between any of the thyroid function tests and sex. Sex-stratified models were examined specifically for hip fracture, which showed similar, nonsignificant results. In the incident heart failure models, there was a significant interaction of 80 years of age or older with FT₄ concentration (P = .03) and of race with TSH concentration (P = .01).

For all outcomes with the exception of dementia, results were unchanged in sensitivity analyses excluding individuals who were TPOAb positive (Supplemental Table 1) and in competing risk analyses (Supplemental Table 2). The dementia HR for TSH was 0.84 (95% CI, 0.72–0.98; P = .02) after excluding TPOAb-positive individuals and 0.86 (95% CI, 0.75–0.99; P = .04) in competing risk analyses, suggesting increased risk of dementia at lower TSH concentrations.

We also examined differences in incidence rates across quartiles within each reference range (Figure 2). Compared with individuals with a FT₄ concentration in the first quartile, those with FT₄ concentrations in fourth quartile had significantly higher rates of AF (difference of 11.0 per 1000 person years; 95% CI, 5.0–16.9), CHD (difference of 8.0 per 1000 person years; 95% CI, 1.1–15.0), heart failure (difference of 7.8 per 1000 person years; 95% CI, 2.0–13.6), and mortality (difference of 14.3 per 1000 person years; 95% CI, 6.1–22.5) (all P < .05). Incidence rates of hip fracture (difference of 0.5 per 1000 person years; 95% CI, −2.8–3.8) and dementia (difference of 7.0 per 1000 person years; 95% CI, −1.4–15.5) were also higher in quartile 4 vs quartile 1, although these comparisons were not statistically significant. Compared with individuals with TSH concentrations in the first quartile, those with TSH concentrations in the fourth quartile had significantly lower absolute rates of dementia (difference of −9.6 per 1000 person years; 95% CI, −17.9 to −1.2).

Figure 2.

Incidence rates per 1000 person years (95% CI) for incident AF, CHD, heart failure (HF), hip fracture (Hip Fx), dementia, and total mortality by quartile of TSH (left), FT₄ (middle), and T₃ (right). For TSH, Q1 is 0.45–1.39 mU/L, Q2 is 1.40–2.00 mU/L, Q3 is 2.01–2.82 mU/L, and Q4 is 2.83–4.50 mU/L. For FT₄, Q1 is 9.20–14.20 pmol/l (0.70–1.10 ng/dL), Q2 is 14.3–15.6 pmol/L (1.11–1.21 ng/dL), Q3 is 15.7–19.0 pmol/L (1.22–1.48 ng/dL), and Q4 is 19.1–21.8 pmol/L (1.48–1.70 ng/dL). For T₃, Q1 is 1.31–1.60 nmol/L (85–104 ng/dL), Q2 is 1.61–1.77 nmol/L (105–115 ng/dL), Q3 is 1.78–1.98 nmol/L (116–129 ng/dL), and Q4 is 1.99–3.10 nmol/L (130–201 ng/dL).

Discussion

We examined the relationship between thyroid function testing within the euthyroid range and outcomes encompassing the cardiovascular, musculoskeletal, and cognitive systems in a cohort of community-dwelling individuals 65 years of age and older. We found increased risk of AF and heart failure at higher concentrations of FT₄. In select models, we found increased risk of dementia at lower concentrations of TSH. Lower TSH and higher FT₄ concentrations were associated with an increased risk of mortality, although these associations were attenuated after adjustment for covariates that could be in the causal pathway, including hypertension, heart failure, and physical activity. The clinical significance of the reported relative risks is difficult to interpret because they are measured in terms of one unit increments of the assays. However, the absolute risks suggest that these differences have clinical relevance, with clinically meaningful differences in rates of AF, CHD, heart failure, dementia, and total mortality at opposite ends within the reference range.

Thyroid function is generally conceived as having a U-shaped relationship with harmful outcomes, with adverse clinical consequences at hyperthyroid or hypothyroid extremes. In our study population and for levels within the reference range, no U-shaped associations were present. This suggests that there is no optimal set of thyroid function tests within current reference ranges to reflect the euthyroid ideal in this age group.

When modeled as a continuous measure, we detected associations between higher concentrations of FT₄ within the euthyroid range and AF and heart failure. In addition, for all outcomes except hip fracture, the point estimate of the effect was greater than one, and incidence rates were highest in the upper part of the reference range, suggesting a consistency of adverse FT₄ effects at higher concentrations. Based on negative feedback regulation, relationships between TSH and outcomes would be expected to be inverse to those of FT₄. TSH concentration was negatively associated with all the outcomes, although this was not statistically significant across all models. Because of the log-linear relationship between TSH and FT₄, minor alterations in FT₄ induce larger changes in TSH concentrations. Thus, it would be expected that TSH would be the more sensitive biomarker of thyroid status. However, TSH concentrations reflect pituitary effects of thyroid hormone, which may differ from end-organ effects. It is interesting to note that the TSH association was strongest for the outcome affecting the brain (dementia), whereas the FT₄ associations were strongest for outcomes directly involving the heart (AF and heart failure).

In addition, the TSH-T₄ setpoint may be altered at older ages, as evidenced by the increase in TSH with increasing age, shown both cross sectionally (30) and longitudinally (25, 31). The mortality data from our study also support a survival advantage in those with higher TSH, suggesting a superimposed survivor effect in the oldest-old. Prescription of levothyroxine has increased, particularly in older individuals with TSH concentrations less than 10 mU/L (32). However, associations between mildly elevated TSH concentrations and clinical outcomes are not detectable when studying populations of older individuals (2, 25, 33–37), leading some researchers to question the benefits of treating mildly elevated TSH concentrations and to advocate for a change in the TSH reference range. Our analyses support shifting the TSH reference range upward and the FT₄ reference range downward in older people. However, this is an observational study of endogenous thyroid function, and we did not study the relative benefits of achieving higher TSH targets or lower FT₄ targets through intervention. In addition, we did not detect a threshold of risk for any outcome, only linear effects. There is no physiologic reason to select the 95% CI to define the reference range. Basing the reference range on clinical outcomes is a more logical approach. Our data support existing recommendations to monitor older individuals with mild subclinical hypothyroidism without treatment, because the lowest-risk individuals in our analytic sample were those with the highest TSH and/or lowest FT₄ concentrations. Our findings also suggest the possibility of using higher target TSH concentrations in older people already using thyroid hormone replacement, as well as the opportunity to study the effects of antithyroid treatment to lower FT₄ and raise TSH in older people with higher FT₄ and lower TSH concentrations within the reference range.

This report of associations between TSH and FT₄ concentrations within the euthyroid range and multiple outcomes in the same cohort of participants allowed the assessment of consistency across measures. Other longitudinal studies of participants not taking thyroid medication and focusing on the euthyroid range have shown concordance with our findings with respect to AF (10), CHD (12, 13), hip fracture (14, 16), dementia (18–20), and mortality (13, 22, 23), although others have reported increased CHD death in women with higher TSH concentrations (11), increased nonvertebral fracture risk in postmenopausal women with lower TSH and higher FT₄ and FT₃ concentrations (15), and no association between TSH concentration and incident dementia (17) or mortality (21). A large Korean study with a younger study population than ours (mean age, 40 y) reported associations between lower FT₄ and FT₃ concentrations within the euthyroid range and all-cause mortality, although no relationships with TSH were seen (24).

Although T₃ is the direct mediator of thyroid function, we were unable to detect associations between total T₃ concentrations and any outcome. Possible explanations include the short half-life of T₃, measurement of total instead of FT₃ concentrations, or inadequacy of serum T₃ as a surrogate for tissue T₃ concentrations. Participants had comorbid illnesses, raising the possibility of an effect of nonthyroidal illness on T₃ concentrations. However, participants included in this analysis all had T₃ concentrations within the euthyroid range. In addition, even if a subset were affected by nonthyroidal illness, there should have been concordance between FT₄ and T₃ concentrations in the upper ranges for the AF outcome, which was not detected.

Study strengths include the large number of participants with up to 18 years of followup, full set of thyroid function tests, assessment of thyroid medication use longitudinally, wealth of data on covariates, adjudicated events, and robustness in sensitivity models and competing risk analyses. In addition, we were able to demonstrate relative stability in TSH testing, with over 95% remaining euthyroid 2 years later. Use of different reference ranges may have affected our results slightly, although a linear trend was seen throughout the reference range used. The interactions detected in the heart failure models should be interpreted in the setting of multiple testing and require replication. Because extensive testing was required to diagnose dementia, followup for this outcome was limited to the main study years of CHS, and thus had shorter followup than the other outcomes. It is possible that there were higher-risk participants who were excluded due to use of thyroid medication who would have had thyroid function testing in the upper portion of the TSH range without thyroid medication. In addition, our findings may not be generalizable outside of older populations.

Conclusions

These analyses provide physiologic insight into organ-specific events related to mild differences in thyroid function in older people and suggest that higher TSH and lower FT₄ concentrations within the euthyroid range are associated with lower risk of multiple adverse events, including mortality. Our findings have implications for the optimal TSH range in older people, which may need to be higher than the currently defined reference ranges. Our findings may also provide a basis for refinement of target thyroid function testing goals in older people who require thyroid hormone replacement.