1

Subclinical Hypothyroidism and the Risk of Coronary Disease and Mortality:

An Individual Participant Data Analysis from Nine Prospective Cohorts

N. Rodondi, W. P. J. den Elzen, D. C. Bauer, A. R. Cappola, S. Razvi, J. P. Walsh, B. O. Åsvold, G. Iervasi, M. Imaizumi, A. Bremner, P. Maisonneuve, M. Vanderpump, A. B. Newman, J. Cornuz, J. A. Franklyn, R.G.J. Westendorp, E. Vittinghoff, J. Gussekloo for the Thyroid Studies Collaboration

Switzerland, The Netherlands, United States, United Kingdom, Western Australia, Norway, Italy, Japan

Conflict of interest: none

Minneapolis, SGIM, April 2010

Background (1)

• Subclinical hypothyroidism =– elevated thyroid-stimulating hormone (TSH)– normal levels of free thyroxine (T4)

• Prevalence: – US adult population: 4.3% (NHANES III) – increases with age: ~ 10% in women > 60 years

• Controversy about screening and treatment of subclinical hypothyroidism

• Current evidence about the risks is limited 1,2

21 USPTSF 2004, Helfand M. Ann Intern Med 2004, 2 Surks M, JAMA 2004

Background (2)

• Data on cardiovascular outcomes are conflicting among several prospective cohorts1,2.

• 3 recent study-level meta-analyses 3,4,5:– Modestly increased risks for CHD and mortality– Limitations: clinical heterogeneity, with different

TSH cutoffs, confounding factors for adjustment and varying CHD definitions3-5

3

1 Cappola AR. JAMA 2006, 2 Walsh JP. Arch Int Med 2005, 3 Rodondi N. Ann Intern Med 2008, 4 Razvi S. J Clin Endocrinol Metab 2008. 5 Volzke H. J Clin Endocrinol Metab 2007

Objectives

• To perform an analysis of individual participant data (IPD) from large cohort studies to define the influence of age, TSH levels, and preexisting CVD on the association between subclinical hypothyroidism and:

- CHD events, CHD mortality, Total mortality

• IPD analysis- Gold standard for synthesizing evidence across

several studies- Subgroup analyses: not subject to potential bias from

study level meta-analyses (ecological fallacy)

4

• Cardiovascular Health Study

• Health, Aging and Body Composition Study Pisa cohort

- Leiden 85+ Study

• Birmingham Study• Whickham Survey

HUNT Study

Nagasaki Adult Health Study

Busselton Health Study

Thyroid Studies Collaboration

5

Standardized Definitions

• Difference with study-level meta-analyses• Thyroid function:

- Euthyroidism: • TSH 0.50-4.49 mU/L

- Subclinical hypothyroidism: • TSH 4.5 mU/L & TSH <20 mU/L• Normal free T4 (site and study specific)

• CHD mortality• CHD events:

- nonfatal myocardial infarction- CHD death- hospitalization for angina or coronary revascularization

6

Statistical Analyses

• Summary estimates and 95%CI- Two stage method:

• Cox proportional hazard models for each cohort separately (SAS Version 9.2)

• Combine estimates (generic reverse variance, random effects model, RevMan 5)

• Heterogeneity- I2 statistic (% of total variation across trials is

attributable to heterogeneity rather than chance)1

71 Higgins et al., BMJ, 2003

Risks Associated with Subclinical Hypothyroidism (n=41’685)

N events / Participants

Adjusted for age and gender

HR (95% CI)

Multivariate model*

HR (95% CI)I2

CHD events 2791 / 13355 1.25 (0.98, 1.59) 1.23 (0.97, 1.56) 67%

CHD mortality 1715 / 41676 1.14 (0.98, 1.34) 1.14 (0.96, 1.34) 0%

Total mortality 7770 / 41685 1.09 (0.94, 1.25) 1.12 (0.95, 1.31) 67%

* Adjusted for gender, age, systolic blood pressure, current and former smoking, total cholesterol, and prevalent diabetes at baseline

8

Study sample:

-41,685 adults comprising 2,621 (6.3%) with subclinical hypothyroidism

Number of outcomes:

-2791 CHD events, 1715 CHD deaths and 14’449 total deaths

Hazard Ratios for CHD Events, CHD Mortality and Total Mortality

Panel A: Elevated TSH Categories vs. Euthyroid

CHD even ts †

TSH 4.5 -6.9 mU/L

TSH 7.0 -9.9 mU/LTSH 10 -20 mU/L

Hazard Ratio (95% CI) *

TSH 4.5 -6.9 mU/L

TSH 7.0 -9.9 mU/LTSH 10 -20 mU/L

CHD mo rtality ‡

To tal mo rtality §

TSH 4.5 -6.9 mU/L

TSH 7.0 -9.9 mU/LTSH 10 -20 mU/L

Panel A: TSH

1.07 (0.84, 1.35)

1.12 (0.88, 1.44)2.00 (1.25, 3.20)Ptrend=0.004

1.11 (0.91, 1.34)

1.40 (0.96, 2.04)1.64 (1.11, 2.42)Ptrend=0.007

1.06 (0.97, 1.17)

1.04 (0.82, 1.32)1.13 (0.69, 1.86)Ptrend=0.65

559 / 1873 6978 / 39064

145 / 496 6978 / 3906488 / 252 6978 / 39064

114 / 1873 1536 / 39056

41 / 496 1536 / 3905624 / 251 1536 / 39056

202 / 854 2465 / 12063

69 / 285 2465 / 1206355 / 153 2465 / 12063

Euth yro idismEven ts / Participants

Subclin ical h ypothyroidismEven ts / Participants

HR adjusted for age and gender

Sizes of data markers are proportional to the inverse of the variance of the hazard ratios.9

Age 18-49 yearsAge 50-64 yearsAge 65-79 yearsAge ≥80 years

CHD events

CHD mortality Age 18-49 yearsAge 50-64 yearsAge 65-79 yearsAge ≥80 years

Total mortalityAge 18-49 yearsAge 50-64 yearsAge 65-79 yearsAge ≥80 years

0.2 0.5 1 2 5

Panel B: Age ||

1.95 (1.02, 3.70)1.45 (0.97, 2.19)1.21 (0.90, 1.61)1.30 (0.93, 1.82)Ptrend=0.22

2.45 (0.79, 7.60)1.61 (0.95, 2.73)1.33 (1.04, 1.70)1.01 (0.62, 1.63)Ptrend=0.14

1.50 (0.85, 2.67)1.13 (0.84, 1.51)1.18 (0.98, 1.42)0.96 (0.81, 1.14)Ptrend=0.16

13 / 334 268 / 1084170 / 705 984 / 12857

539 / 1358 4134 / 13048170 / 224 1592 / 2318

2 / 334 44 / 1084113 / 705 213 / 12855

135 / 1357 979 / 1304729 / 224 300 / 2313

10 / 111 197 / 268725 / 150 359 / 2220

249 / 907 1649 / 614842 / 124 260 / 1008

Hazard Ratios for Coronary Heart Disease (CHD) Events, CHD Mortality and Total Mortality

Panel B: Subclinical Hypothyroidism vs. Euthyroid Stratified by Age

Risks did not significantly differ by gender or preexisting CVD

HR adjusted for age and gender as a continuous variable to avoid residual confounding within age strata, Sizes of data markers are proportional to the inverse of the variance of the hazard ratios.

10

Sensitivity Analysis on the Risks of CHD Events and CHD Mortality

CHD Events TSH 10-20 mU/L

CHD Mortality TSH 10-20 mU/L

All eligible studies :

Random-effects 2.00 (1.25, 3.20) 1.64 (1.11, 2.42)

Excluding those treated by thyroid medication at baseline

1.84 (1.07, 3.16) 1.58 (1.04, 2.39)

Excluding those treated by thyroid medication at baseline and during follow-up

2.26 (1.39, 3.68) 2.02 (1.26, 3.26)

Excluding “soft” CHD outcomes * 1.88 (1.00, 3.53) NA

4 Studies with formal adjudication procedures1-4 2.05 (1.14, 3.68) 1.77 (1.08, 2.89)

Further adjustment for lipid lowering and anti-hypertensive medications in addition to cardiovascular risk factors

1.96 (1.14, 3.35) 1.60 (1.06, 2.42)

11

* Possible in 4 studies 1-4

1 Cappola AR et al, JAMA 2006; 2 Rodondi N et al. Arch Intern Med 20053 Iervasi G, et al. Arch Intern Med 2007, 4 Gussekloo J, et al. JAMA 2004

Limitations

• Our IPD analysis included predominantly white populations, except for a study in Japan1

• Thyroid function testing performed only at baseline: - limitation of all published large cohorts

• Commencement of thyroid medication during follow-up (by 0-12.6%) might have attenuated any true effects of subclinical hypothyroidism:- higher estimators in the sensitivity analysis excluding

such participants

121 Imaizumi M, et al. J Clin Endocrinol Metab 2004

Conclusions

• Subclinical hypothyroidism is associated with an increased risk of CHD in those with higher TSH levels among 41,685 participants.

• Our results might help refine a TSH threshold at which larger benefits of thyroxine replacement would be expected:

- Many adults with minimal TSH elevation currently treated1 despite no significant increased risk of CHD (or other risks)

• An appropriately powered RCT is needed to examine the efficacy of screening for and treating subclinical hypothyroidism.

131 Fatourechi V et al. Mayo Clin Proc 2003

14

Thank you for your attention