ORIGINAL ARTICLE

Thyroid Status and Death Risk in US

Veterans With Chronic Kidney Disease
Connie M. Rhee, MD, MSc; Kamyar Kalantar-Zadeh, MD, MPH, PhD;
Vanessa Ravel, MPH; Elani Streja, MPH, PhD; Amy S. You, MS;
Steven M. Brunelli, MD, MSCE; Danh V. Nguyen, MS, PhD; Gregory A. Brent, MD;
and Csaba P. Kovesdy, MD

Abstract

Objective: Given that patients with nonedialysis-dependent chronic kidney disease (NDD-CKD) have a
disproportionately higher prevalence of hypothyroidism compared with their non-CKD counterparts, we
sought to determine the association between thyroid status, defined by serum thyrotropin (TSH) levels,
and mortality among a national cohort of patients with NDD-CKD.
Patients and Methods: Among 227,422 US veterans with stage 3 NDD-CKD with 1 or more TSH
measurements during the period October 1, 2004, to September 30, 2012, we first examined the asso-
ciation of thyroid status, defined by TSH categories of less than 0.5, 0.5 to 5.0 (euthyroidism), and more
than 5.0 mIU/L, with all-cause mortality. We then evaluated 6 granular TSH categories: less than 0.1, 0.1
to less than 0.5, 0.5 to less than 3.0, 3.0 to 5.0, more than 5.0 to 10.0, and more than 10.0 mIU/L. We
concurrently examined thyroid status, thyroid-modulating therapy, and mortality in sensitivity analyses.
Results: In expanded case-mix adjusted Cox analyses, compared with euthyroidism, baseline and time-
dependent TSH levels of more than 5.0 mIU/L were associated with higher mortality (adjusted hazard
ratios [aHRs] [95% CI], 1.19 [1.15-1.24] and 1.23 [1.19-1.28], respectively), as were baseline and time-
dependent TSH levels of less than 0.5 mIU/L (aHRs [95% CI], 1.18 [1.15-1.22] and 1.41 [1.37-1.45],
respectively). Granular examination of thyroid status showed that incrementally higher TSH levels of
3.0 mIU/L or more were associated with increasingly higher mortality in baseline and time-dependent
analyses, and TSH categories of less than 0.5 mIU/L were associated with higher mortality (reference,
0.5-<3.0 mIU/L) in baseline analyses. In time-dependent analyses, untreated and undertreated hypo-
thyroidism and untreated hyperthyroidism were associated with higher mortality (reference, spontaneous
euthyroidism), whereas hypothyroidism treated-to-target showed lower mortality.
Conclusion: Among US veterans with NDD-CKD, high-normal TSH (
3.0 mIU/L) and lower TSH
(<0.5 mIU/L) levels were associated with higher death risk. Interventional studies identifying the target
TSH range associated with the greatest survival in patients with NDD-CKD are warranted.
ª 2018 Mayo Foundation for Medical Education and Research n Mayo Clin Proc. 2018;93(5):573-585

T
hyroid dysfunction is a highly prevalent more, 60 to 89, 45 to 59, 30 to 44, and less
yet underrecognized endocrine compli- than 30 mL/min/1.73 m2, respectively.5 More
cation affecting a large proportion of recently, among a large national cohort of For editorial
patients with chronic kidney disease (CKD).1,2 461,607 US veterans with stage 3 to 5 CKD, comment, see
Several large population-based studies found for every 10 mL/min/1.73 m2 decrement in page 555
that hypothyroidism is increasingly more com- eGFR, there was an 18% higher likelihood of hy-
mon with incrementally impaired kidney func- pothyroidism.6 Although there are compara- From Harold Simmons
Center for Chronic Dis-
tion.3-7 Data from 14,623 participants in the tively fewer reports in patients with end-stage ease Research and Epide-
Third National Health and Nutrition Examina- renal disease (ESRD), a similarly high prevalence miology, University of
California Irvine School of
tion Survey have shown that the prevalence of (w22%) of hypothyroidism has been observed Medicine, Orange, CA
hypothyroidism was 5.4%, 10.9%, 20.4%, in large dialysis cohorts.8,9 (C.M.R., K.K.-Z., V.R., E.S.,
23.0%, and 23.1% among those with estimated In the general population, adverse cardio-
glomerular filtration rates (eGFRs) of 90 or vascular sequelae may result from untreated Affiliations continued at
the end of this article.

Mayo Clin Proc. n May 2018;93(5):573-585 n https://doi.org/10.1016/j.mayocp.2018.01.024 573

www.mayoclinicproceedings.org n ª 2018 Mayo Foundation for Medical Education and Research
 MAYO CLINIC PROCEEDINGS

hypothyroidism (eg, altered cardiac structure 30, 2012.23-26 Patients were included pro-
and function, endothelial dysfunction, dyslipi- vided that they underwent at least 1 TSH
demia, accelerated atherosclerosis, and electro- measure anytime during the study period;
physiologic changes10-13) as well as with had the requisite covariates needed to calculate
hyperthyroidism (eg, atrial fibrillation, heart eGFR (eg, age, race, and serum creatinine)
failure, and coronary ischemia12,14). Hence, within 1 year of study entry (ie, date of the
this high burden of thyroid dysfunction may baseline TSH); and had stage 3 CKD (eGFR,
have an important bearing on the survival of 30-<60 mL/min/1.73 m2) at study entry.
patients with kidney disease, who suffer from Patients were excluded if they were receiving
disproportionately high cardiovascular death dialysis at the time of study entry; had an
risk (w40% of deaths15,16). Indeed, a growing improbable TSH level (ie, 0 mIU/L); or had
body of evidence reports that hypothyroidism, an implausible follow-up time value. The
defined by elevated serum thyrotropin (TSH) study was approved by the institutional review
levels as the most sensitive and specific clinical committees of the Memphis and Tibor Rubin
goal standard of thyroid function assess- VA Medical Centers.
ment,17-19 is associated with higher mortality
in some,8,9,20,21 but not all,22 studies of the Exposure Ascertainment
dialysis population. Recent data have also The exposure of interest was thyroid status
corroborated a link between hyperthyroidism defined by serum TSH concentration (irrespec-
and sudden cardiac death among hemodialysis tive of treatment status). In primary analyses, we
patients from the Die Deutsche Diabetes Dialyse examined thyroid status categorized as TSH
Studie (also known as the 4D Trial),22 and with levels of more than 5.0, 0.5 to 5.0, and less
all-cause mortality in a national peritoneal than 0.5 mIU/L (based on thresholds used in
dialysis cohort.9 However, little is known the general population for ascertainment of
about the association between thyroid status hypothyroidism, euthyroidism, and hyperthy-
and mortality risk in patients with CKD who roidism, respectively).8,9,17 In secondary ana-
are nonedialysis-dependent (NDD). lyses, we examined thyroid status using more
To address this knowledge gap, we conduct- granular categorizations of TSH, defined
ed a study examining the relationship between according to the usual TSH ranges for these des-
thyroid status and mortality risk among a large ignations: overt-hypothyroid (>10.0 mIU/L),
longitudinal cohort of US veterans with stage 3 subclinical-hypothyroid (>5.0-10.0 mIU/L),
CKD and repeated measures of serum TSH high-normal (3.0-5.0 mIU/L), low-normal
over time. We hypothesized that both higher (0.5-<3.0 mIU/L), subclinical-hyperthyroid
and lower TSH levels were independently asso- (0.1-<0.5 mIU/L), and overt-hyperthyroid
ciated with higher mortality risk in this nation- (<0.1 mIU/L) ranges.8,9,20 We were specifically
ally representative NDD-CKD cohort. On the interested in the TSH threshold of 3.0 mIU/L or
basis of previous studies of thyroid status and more as the level above which higher mortality
mortality in the dialysis population,8,20 we may be observed.8,20 We also examined TSH
were also specifically interested in examining as a continuous predictor of mortality using
the TSH threshold of 3.0 mIU/L or more as restricted cubic spline analyses with knots
the level above which higher mortality is defined at the 33rd and 66th percentiles of
observed in patients with NDD-CKD. observed TSH values.
We first examined the association between
PATIENTS AND METHODS baseline thyroid status and all-cause mortality
to ascertain long-term associations of thyroid
Source Cohort status with death risk.27 Given that underlying
We conducted a historical cohort study using illness may influence serum TSH levels in the
data from the “Racial and Cardiovascular Risk absence of true thyroid functional disease while
Anomalies in Chronic Kidney Disease” study, also increasing patients’ risk of death, we con-
constructed to examine US veterans with inci- ducted 2 types of sensitivity analyses of baseline
dent CKD who underwent care within the thyroid status and mortality risk: (1) analyses
Veterans Affairs (VA) health care system over that included a 30-day lag period between the
the period of October 1, 2004, to September date of baseline TSH measurement and start
n n
574 Mayo Clin Proc. May 2018;93(5):573-585 https://doi.org/10.1016/j.mayocp.2018.01.024
www.mayoclinicproceedings.org
THYROID STATUS IN CKD

of follow-up time to minimize bias from reverse has been observed to have a sensitivity and
causation,20 and (2) analyses that examined the specificity of 98.3% and 99.8%, respectively,
association of incident hypothyroidism and in comparison to the National Death Index.30
incident hyperthyroidism (defined as those
who had a baseline TSH level that was within Sociodemographic, Comorbidity, Medication,
the reference range [0.5-5.0 mIU/L] and had a and Laboratory Covariates
second TSH level that was >5.0 mIU/L and Patients’ baseline sociodemographic informa-
<0.5 mIU/L, respectively) with mortality. tion (age, sex, race, ethnicity) was obtained
We then examined the association between from the VA Corporate Data Warehouse and
time-dependent thyroid status and all-cause mor- from Medicare through the VA-Medicare data
tality, in which thyroid status was time-updated merge project31 as previously described.23-26
with repeated TSH measures to ascertain short- Information about comorbidities was extracted
term associations of thyroid status with death from the VA Inpatient and Outpatient Medical
risk, and to account for changes in thyroid SAS data sets using International Classification
status over time.27 The median (interquartile of Diseases, Ninth Revision diagnostic and pro-
range [IQR]) and minimum-maximum number cedure codes and Current Procedural Termi-
of TSH measurements contributed by each nology codes.32 Charlson comorbidity index
patient was 4 (2-6) and 1-54, respectively. scores were estimated using the Deyo modifica-
To determine the impact of thyroid- tion for administrative data sets without
modulating therapy on the association of thy- including kidney disease.33 Body mass index
roid status with mortality, we compared death data were obtained from the VA Vital Status
risk across the following categories that concur- File. Laboratory data except serum creatinine
rently considered patients’ TSH levels and were obtained from the Decision Support
medication status: hypothyroid overtreated System National Data Extracts Laboratory
(TSH<0.5 mIU/L, on thyroid hormone replace- Results files.28 VA Corporate Data Warehouse
ment), hypothyroid untreated (TSH>5.0 mIU/ LabChem data files were used to extract serum
L, not on thyroid hormone replacement), hypothy- creatinine data.34 Using serum creatinine and
roid undertreated (TSH>5.0 mIU/L, on thyroid demographic data, eGFR was calculated using
hormone replacement), hypothyroid treated- the Chronic Kidney Disease Epidemiology
to-target (TSH¼0.5-5.0 mIU/L, on thyroid Collaboration equation.35
hormone replacement), spontaneously euthyroid
(TSH¼0.5-5.0 mIU/L, not on thyroid hormone
Statistical Analyses
replacement or antithyroid medication), hyper-
We estimated the association between thyroid
thyroid undertreated (TSH<0.5 mIU/L, on anti-
status and mortality using Cox proportional
thyroid medication), hyperthyroid untreated
hazard models with 5 hierarchical levels of co-
(TSH<0.5 mIU/L, not on antithyroid medication),
variate adjustment. In time-dependent analyses,
and hyperthyroid overtreated (TSH>5.0 mIU/L,
all laboratory data were examined as time-
on antithyroid medication). Serum TSH data were
dependent covariates:
obtained from the Decision Support System Na-
tional Data Extracts Laboratory Results file.28 1. Unadjusted model: Included serum TSH
Medication data were obtained from VA phar- level as the primary exposure of interest;
macy dispensation records.29 2. Case-mix model: Adjusted for age, sex,
race, and ethnicity;
Outcome Ascertainment 3. Expanded case-mix model: Adjusted for
Our outcome of interest was all-cause mortal- covariates in the case-mix model, as well as
ity. Patients were followed for outcomes start- diabetes, congestive heart failure (CHF), car-
ing the day after TSH measurement. Patients diovascular disease (CVD), hypertension,
were censored for kidney transplantation, hyperlipidemia, and Charlson comorbidity
loss to follow-up, or the last date of available index score;
follow-up data (July 26, 2013), whichever 4. Expanded case-mix þ laboratory model, also
occurred first. All-cause mortality data, known as adjustment for malnutrition-
censoring events, and associated dates were inflammation-cachexia syndrome covariates:
obtained from VA Vital Status File, which Adjusted for covariates in the expanded

Mayo Clin Proc. n May 2018;93(5):573-585 n https://doi.org/10.1016/j.mayocp.2018.01.024 575

www.mayoclinicproceedings.org
 MAYO CLINIC PROCEEDINGS

case-mix model, as well as eGFR, serum albu- mIU/L) across clinically relevant categories of
min, hemoglobin, cholesterol, triglycerides, sociodemographic characteristics, comorbidity
and low-density lipoprotein; status, and laboratory measures. Proportional
5. Expanded case-mix þ laboratory þ thy- hazards assumptions were confirmed by
roid medication model: Adjusted for cova- graphical analysis. Analyses and figures were
riates in the expanded case-mix þ generated using SAS version 9.4 (SAS Institute
laboratory model, as well as thyroid hor- Inc), Stata version 14 (Stata Corporation), and
mone supplementation and antithyroid SigmaPlot version 12.5 (Systat Software).
medication use.
RESULTS
We a priori defined the expanded case-mix
adjusted model as our preferred model, which Study Population
included core sociodemographic measures Among 227,426 patients who met eligibility
and other confounders of the association be- criteria (Supplemental Figure 1, available
tween thyroid status and mortality. There online at http://www.mayoclinicproceedings.
were no missing data for covariates in the org), 6.4%, 90.0%, and 3.4% of patients had
case-mix and expanded case-mix models (eg, baseline TSH levels of more than 5.0, 0.5 to
age, sex, race, ethnicity, and comorbidities), 5.0 (euthyroid), and less than 0.5 mIU/L,
whereas there were varying degrees of missing respectively. The mean  SD, median (IQR),
data for laboratory variables including albumin and minimum-maximum of observed baseline
(17.3%), hemoglobin (15.4%), triglycerides TSH values were 2.76.7, 1.9 (1.2-2.8), and
(13.7%), cholesterol (13.6%), and low-density 0.001-531.5 mIU/L, respectively. Compared
lipoprotein (18.9%; proportions shown repre- with patients whose baseline TSH levels were
sent missing baseline data). Hence, further less than 0.5 mIU/L or in the euthyroid range,
adjustments for potential confounders in those with TSH levels of more than 5.0 mIU/L
expanded case-mix þ laboratory and expanded were older, more likely to be non-Hispanic
case-mix þ laboratory þ thyroid medication white, and less likely to be non-Hispanic black;
models (which have the potential for inherent were more likely to have cardiovascular disease
selection bias due to missing data) were con- and less likely to have hypertension; were more
ducted as sensitivity analyses. In these analyses, likely to have higher total cholesterol and tri-
missing data were handled using multiple glyceride levels; and were less likely to be pre-
imputation. scribed an ACEI or ARB within 1 year of study
To address differential quality of care as a entry (Table). Compared with patients who
potential confounder of thyroid status- were euthyroid at baseline, those who were hy-
mortality associations, we conducted a series perthyroid and hypothyroid were also more
of sensitivity analyses that incrementally likely to have CHF and less likely to have hyper-
adjusted for proxies of provider attentive- lipidemia. Baseline characteristics of patients
ness/adherence with health care recommenda- categorized by finer gradations of TSH are pre-
tions and patient medical compliance, in sented in Supplemental Table 1 (available online
addition to covariates in the expanded case- at http://www.mayoclinicproceedings.org).
mix þ laboratory þ thyroid medication
model: (1) angiotensin-converting enzyme Thyroid Status and Mortality
inhibitor (ACEI), angiotensin receptor blocker Patients contributed a total of 1,134,837 patient-
(ARB), or statin prescription within 1 year of years of follow-up during which time 67,706 all-
study entry; (2) receipt of influenza vaccina- cause deaths occurred. Median (IQR) at-risk
tion within 1 year of study entry; or (3) time was 65 (38-83) months. In analyses of base-
patient’s history of noncompliance with medi- line thyroid status adjusted for expanded case-
cal treatment ascertained by International Clas- mix covariates, TSH levels of more than 5.0
sification of Diseases, Ninth Revision code and less than 0.5 mIU/L were each associated
V15.81. with higher mortality risk (reference, euthyroid-
We additionally conducted subgroup ana- ism): adjusted HRs (aHRs) (95% CI), 1.19 (1.15-
lyses of thyroid status (categorized as TSH 1.24) and 1.18 (1.15-1.22), respectively
levels of >5.0, 0.5-5.0 [euthyroid], and <0.5 (Figure 1A; Supplemental Table 2, available
n n
576 Mayo Clin Proc. May 2018;93(5):573-585 https://doi.org/10.1016/j.mayocp.2018.01.024
www.mayoclinicproceedings.org
THYROID STATUS IN CKD

TABLE. Baseline Characteristics According to Thyroid Status Defined by Thyrotropin (TSH)a

TSH (mIU/L)
Overall <0.5 0.5-5.0 >5.0
Characteristic (N¼227,426) (N¼7807) (N¼205,166) (14,453) P
Age (y), mean  SD 7110 7011 7110 7311 <.001
Sex: female (%) 3 6 3 4 <.001
Race/ethnicity (%)
Non-Hispanic white 84 78 84 88 <.001
Non-Hispanic black 12 19 12 7
Hispanic 2 2 2 3
Other 2 2 2 2
Body mass index (kg/m2) 29.35.8 28.85.7 29.45.7 29.06.0 <.001
Comorbidities
Diabetes (%) 37 36 37 36 <.001
Congestive heart failure (%) 11 13 11 14 <.001
Cardiovascular disease (%) 36 35 36 39 <.001
Cerebrovascular disease (%) 9 9 9 9 .70
Hypertension (%) 80 80 81 75 <.001
Hyperlipidemia (%) 70 64 70 65 <.001
Charlson comorbidity 1 (1-3) 2 (1-3) 1 (1-3) 2 (1-3) <.001
index (median [IQR])
Laboratory values, mean  SD or median (IQR)
eGFR (mL/min/1.73 m2) 536 527 536 527 <.001
Serum albumin (mg/dL) 4.0 (3.8-4.3) 4.0 (3.7-4.2) 4.0 (3.8-4.3) 4.0 (3.7-4.3) <.001
Hemoglobin (g/dL) 14.0 (12.9-15.1) 13.8 (12.5-14.9) 14.1 (12.9-15.1) 13.8 (12.5-14.9) <.001
Total cholesterol (mg/dL) 165 (142-193) 163 (140-190) 165 (142-193) 167 (141-197) <.001
Triglycerides (mg/dL) 128 (89-188) 121 (84-177) 128 (89-187) 134 (92-201) <.001
LDL (mg/dL) 93 (74-117) 92 (73-115) 93 (74-117) 93 (73-119) <.001
Medications (%)
Thyroid hormone supplement 6 17 4 20 <.001
Antithyroid medication <1 <1 <1 <1 <.001
ACEI and/or ARB 40 41 40 35 <.001
Statin 22 20 22 21 <.001
Quality-of-care indicators (%)
Receipt of influenza vaccine within 27 27 27 26 .003
1 y of study entry
History of noncompliance with 3 3 3 3 .77
medical treatmentb
a
ACEI ¼ angiotensin-converting enzyme inhibitor; ARB ¼ angiotensin receptor blocker; eGFR ¼ estimated glomerular filtration rate; ICD-
9 ¼ International Classification of Diseases, Ninth Revision; IQR ¼ interquartile range; LDL ¼ low-density lipoprotein.
b
Ascertained by ICD-9 code (V15.81).

online at http://www.mayoclinicproceedings. robust association of TSH levels of more than

org). These associations were mildly attenuated 5.0 and less than 0.5 mIU/L with higher mortal-
after further adjustment for expanded case- ity risk in expanded case-mix models (reference,
mix þ laboratory and expanded case-mix þ euthyroidism): aHRs (95% CI), 1.17 (1.13-
laboratory þ thyroid medication covariates, 1.22) and 1.17 (1.14-1.21), respectively
but remained statistically significant. Similarly, (Supplemental Figure 2A and Supplemental
in analyses of time-dependent thyroid status, Table 3, available online at http://www.
TSH levels of more than 5.0 and less than 0.5 mayoclinicproceedings.org). In a subcohort of
mIU/L were each associated with higher mortal- patients who had a baseline TSH in euthyroid
ity risk across all levels of adjustment (Figure 1B; range and a subsequent TSH measurement
Supplemental Table 2). (N¼161,802), we identified 6027 and 3445
In sensitivity analyses of baseline TSH that patients who had incident hypothyroidism
included a 30-day lag period, we observed a and hyperthyroidism, respectively. Sensitivity

Mayo Clin Proc. n May 2018;93(5):573-585 n https://doi.org/10.1016/j.mayocp.2018.01.024 577

www.mayoclinicproceedings.org
 MAYO CLINIC PROCEEDINGS

1.6 1.6
All-cause mortality HR (95% CI)

All-cause mortality HR (95% CI)

1.2 1.2

1 1

0.8 0.8
TSH <0.5 TSH 0.5-5.0 (Euthyroid) TSH >5.0 TSH <0.5 TSH 0.5-5.0 (Euthyroid) TSH >5.0
TSH (mIU/L) TSH (mIU/L)
Unadjusted Case-mix Expanded case-mix Unadjusted Case-mix Expanded case-mix
Expanded case-mix + laboratory Expanded case-mix + laboratory
A Expanded case-mix + laboratory + thyroid medication B Expanded case-mix + laboratory + thyroid medication

FIGURE 1. Association between baseline (A) and time-dependent (B) thyroid status and all-cause mortality. Unadjusted model
included serum TSH level. Case-mix model adjusted for covariates in the unadjusted model, as well as age, sex, race, and ethnicity.
Expanded case-mix model adjusted for covariates in the case-mix model, as well as diabetes, congestive heart failure, cardiovascular
disease, hypertension, hyperlipidemia, and Charlson comorbidity index. Expanded case-mix þ laboratory model adjusted for cova-
riates in the expanded case-mix model, as well as estimated glomerular filtration rate, serum albumin, hemoglobin, cholesterol, tri-
glycerides, and low-density lipoprotein. Expanded case-mix þ laboratory þ thyroid medication model adjusted for covariates in the
expanded case-mix þ laboratory model, as well as thyroid hormone supplementation and antithyroid medication use. TSH ¼
thyrotropin.

analyses showed that incident hypothyroidism range, and overt-hypothyroid range, respec-
and incident hyperthyroidism were each tively (Supplemental Figure 3A and
associated with higher mortality risk in Supplemental Table 5, available online at
expanded case-mix models (reference, euthyr- http://www.mayoclinicproceedings.org). These
oidism): aHRs (95% CI), 1.27 (1.20-1.35) and patterns of association persisted in analyses
1.22 (1.16-1.27), respectively (Supplemental adjusted for expanded case-mix þ laboratory
Table 4, available online at http://www. and expanded case-mix þ laboratory þ thyroid
mayoclinicproceedings.org). medication covariates. We also observed that
TSH categories of less than 0.5 mIU/L were asso-
Granular Thyrotropin Levels and Mortality ciated with higher mortality risk in expanded
A continuum of TSH increase and reduction case-mix analyses: aHRs (95% CI), 1.21 (1.16-
from the reference range is recognized, with 1.26) and 1.13 (1.02-1.25) for TSH levels in
early or subclinical disease defined by an the subclinical-hyperthyroid and overt-
abnormal TSH with a reference range serum hyperthyroid range, respectively. Associations
free thyroxine (FT4) measurement, which between subclinical-hyperthyroid range TSH
were not available for our study. In analyses levels and mortality persisted in expanded
that more granularly examined baseline TSH case-mix þ laboratory and expanded case-
levels, we found that TSH categories of 3.0 mix þ laboratory þ thyroid medication
mIU/L or more were associated with incremen- analyses. Similar associations were observed in
tally higher death risk in expanded case-mix analyses of baseline TSH that included a 30-
models (reference, low-normal TSH): aHRs day lag period (Supplemental Figure 2B and
(95% CI), 1.04 (1.02-1.06), 1.15 (1.11-1.18), Supplemental Table 3).
and 1.38 (1.31-1.46) for TSH levels in the In analyses of time-dependent TSH levels
upper-normal range, subclinical-hypothyroid adjusted for expanded case-mix covariates, we
n n
578 Mayo Clin Proc. May 2018;93(5):573-585 https://doi.org/10.1016/j.mayocp.2018.01.024
www.mayoclinicproceedings.org
THYROID STATUS IN CKD

1.4 50,000 2 200,000

1.8

All-cause mortality hazard ratio

All-cause mortality hazard ratio

1.3 40,000
150,000
1.6
1.2

Frequency
30,000

Frequency
1.4
100,000
1.1 20,000
1.2
50,000
10,000
1
1
0 0
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
A Baseline TSH (mIU/L) B Time-dependent TSH (mIU/L)

FIGURE 2. Association between continuous baseline (A) and time-dependent (B) thyrotropin (TSH) and all-cause mortality.
Figures present hazard ratios (bold solid lines) and 95% CIs (short-dashed lines) for TSH analyzed as a spline with knots at the 33rd
and 66th percentiles of observed values (TSH levels, 1.4 and 2.4 mIU/L, respectively). A histogram of observed TSH values and a
hazard reference ratio of 1 (solid line) is overlaid. Analyses shown are expanded case-mix adjusted models, which included serum
TSH level, age, sex, race, ethnicity, diabetes, congestive heart failure, cardiovascular disease, hypertension, hyperlipidemia, and Charlson
comorbidity index.

analogously observed that TSH categories of 3.0 Adjustment for Quality-of-Care Indicators
mIU/L or more were associated with increas- To account for differential quality of care as a
ingly higher mortality (reference, low-normal potential confounder of thyroid status-mortality
TSH): aHRs (95% CI), 1.05 (1.03-1.07), 1.34 associations, in sensitivity analyses we also exam-
(1.30-1.38), and 1.87 (1.76-1.99) for TSH ined several models that incrementally adjusted
levels in the upper-normal range, subclinical- for proxies of provider attentiveness/adherence
hypothyroid range, and overt-hypothyroid with health care recommendations and patient
range, respectively (Supplemental Figure 3B medical compliance, in addition to expanded
and Supplemental Table 5). Associations case-mix þ laboratory þ thyroid medication
remained significant in expanded case-mix þ covariates. Similar to expanded case-mix þ
laboratory and expanded case-mix þ laboratory þ thyroid medication models, we
laboratory þ thyroid medication adjusted ana- observed that baseline and time-dependent
lyses. We similarly observed that TSH levels in TSH levels of more than 5.0 and less than 0.5
the subclinical-hyperthyroid range and overt- mIU/L were each associated with higher mortality
hyperthyroid range were associated with higher risk in analyses that incrementally adjusted for
mortality: aHRs (95% CI), 1.27 (1.22-1.32) and ACEI/ARB or statin use, receipt of the influenza
1.13 (1.02-1.25), respectively. Associations be- vaccination, and history of noncompliance
tween subclinical-hyperthyroid range and (Supplemental Table 6, available online at
overt-hyperthyroid range TSH levels and mor- http://www.mayoclinicproceedings.org).
tality persisted in expanded case-mix þ Following incremental adjustment for these
laboratory þ thyroid medication analyses quality-of-care indicators, we also found that
(Supplemental Figure 3B and Supplemental associations of baseline TSH levels in the
Table 5). upper-normal, subclinical-hypothyroid, overt-
In sensitivity analyses of baseline and time- hypothyroid, and subclinical-hyperthyroid
dependent TSH as continuous predictors of ranges with mortality risk remained robust
mortality using restricted cubic splines (Supplemental Table 7, available online
adjusted for expanded case-mix covariates, at http://www.mayoclinicproceedings.org).
we observed a U-shaped association between Following incremental adjustment for ACEI/
TSH level and mortality, with a nadir of risk ARB or statin use, receipt of the influenza vacci-
observed at TSH levels of approximately 1.7 nation, and history of noncompliance, time-
to 2.1 mIU/L (Figure 2). dependent TSH levels in the upper-normal,

Mayo Clin Proc. n May 2018;93(5):573-585 n https://doi.org/10.1016/j.mayocp.2018.01.024 579

www.mayoclinicproceedings.org
 MAYO CLINIC PROCEEDINGS

Age <60 Age <60

P=.51 P<.001
Age 60+ Age 60+

Female Female
P=.37 P=.66
Male Male

Black Black
P=.06 P<.001
Nonblack Nonblack

Diabetes Diabetes
P=.07 P=.53
No diabetes No diabetes

CHF CHF
P=.42 P=.05
No CHF No CHF

CVD CVD
P=.02 P=.59
No CVD No CVD

BMI <25 BMI <25

P=.05 P=.04
BMI 25+ BMI 25+

0.3 0.5 1 1.5 2 3 0.3 0.5 1 1.5 2 3

All-cause mortality HR (95% CI) All-cause mortality HR (95% CI)

A Hyperthyroid Euthyroid Hypothyroid B Hyperthyroid Euthyroid Hypothyroid

FIGURE 3. Baseline (A) and time-dependent (B) thyroid status and all-cause mortality across clinically relevant subgroups. P values for
interaction tests presented. Analyses shown are expanded case-mix adjusted models, which included serum TSH level, age, sex, race,
ethnicity, diabetes, congestive heart failure, cardiovascular disease, hypertension, hyperlipidemia, and Charlson comorbidity index.
BMI ¼ body mass index; eGFR ¼ estimated glomerular filtration rate; CHF ¼ congestive heart failure; CVD ¼ cardiovascular disease;
TSH ¼ thyrotropin.

subclinical-hypothyroid, overt-hypothyroid, except those who were female, and that hyper-
and subclinical-hyperthyroid ranges, as well thyroidism was associated with higher mortality
as in the overt-hyperthyroid range, were associ- across all subgroups (Figure 3A; Supplemental
ated with higher mortality risk, similar to Table 8, available online at http://www.
expanded case-mix þ laboratory þ thyroid mayoclinicproceedings.org). Interaction tests
medication covariates. demonstrated that differences in estimates of
the thyroid function-mortality association
THYROID STATUS AND MORTALITY ACROSS across subcategories were statistically signifi-
CLINICALLY RELEVANT SUBGROUPS cant for subgroups of underlying cardiovascular
We also examined the association between thy- disease and body mass index (BMI [calculated
roid status (categorized as TSH levels of >5.0, as the weight in kilograms divided by the
0.5-5.0 [euthyroidism], and <0.5 mIU/L; refer- height in meters squared]) (P interactions¼.03
ence, euthyroidism) and mortality across clini- and .05, respectively; Figure 3A). In these ana-
cally relevant subgroups. In baseline analyses lyses, we observed that point estimates of the
adjusted for expanded case-mix covariates, we hypothyroidism-mortality association were
observed that hypothyroidism was associated stronger among those without vs with underlying
with higher mortality across all subgroups cardiovascular disease and higher (
25 kg/m2)
n n
580 Mayo Clin Proc. May 2018;93(5):573-585 https://doi.org/10.1016/j.mayocp.2018.01.024
www.mayoclinicproceedings.org
THYROID STATUS IN CKD

4.0 4.0

3.0 3.0
All-cause mortality HR (95% CI)

All-cause mortality HR (95% CI)

2.0 2.0

1.0 1.0

0.5 0.5

Euthyroid Hypothyroid Hypothyroid Hypothyroid Hypothyroid Euthyroid Hyperthyroid Hyperthyroid Hyperthyroid Hyperthyroid
at target undertreated untreated overtreated at target undertreated untreated overtreated
A Baseline analyses B Baseline analyses

4.0 4.0

3.0 3.0
All-cause mortality HR (95% CI)

All-cause mortality HR (95% CI)

2.0 2.0

1.0 1.0

0.5 0.5

Euthyroid Hypothyroid Hypothyroid Hypothyroid Hypothyroid Euthyroid Hyperthyroid Hyperthyroid Hyperthyroid Hyperthyroid
at target undertreated untreated overtreated at target undertreated untreated overtreated
C Time-dependent analyses D Time-dependent analyses

FIGURE 4. Baseline (A and B) and time-dependent (C and D) thyroid status, thyroid medications, and all-cause mortality. Analyses
shown are expanded case-mix adjusted models, which included serum TSH level, age, sex, race, ethnicity, diabetes, congestive heart
failure, cardiovascular disease, hypertension, hyperlipidemia, and Charlson comorbidity index. TSH ¼ thyrotropin.

vs lower BMI (<25 kg/m2), and that stronger point BMI (P interactions¼<.001, <.001, .05, and
estimates of the hyperthyroidism-mortality associ- .04, respectively; Figure 3B), with stronger point
ation were also observed among those with higher estimates of hypothyroidism-mortality among
vs lower BMI. those of younger (<60 years) vs older (
60
In time-dependent analyses, we similarly years) age, black vs nonblack race, with vs
observed that hypothyroidism was associated without CHF, and higher vs lower BMI, whereas
with higher mortality across all subgroups except stronger point estimates of the hyperthyroidism-
those who were female, and that hyperthyroid- mortality association were observed among those
ism was associated with higher mortality of younger vs older age, nonblack vs black race,
across all subgroups (Figure 3B; Supplemental with vs without CHF, and higher vs lower BMI.
Table 9, available online at http://www.
mayoclinicproceedings.org). Interaction tests Thyroid Status, Medication Use, and
demonstrated that differences in estimates of Mortality
the thyroid function-mortality association across In the overall cohort, 6.0% and less than 0.1% of
subcategories were statistically significant for patients were prescribed thyroid hormone sup-
subgroups of age, race, underlying CHF, and plementation or antithyroid medication,

Mayo Clin Proc. n May 2018;93(5):573-585 n https://doi.org/10.1016/j.mayocp.2018.01.024 581

www.mayoclinicproceedings.org
 MAYO CLINIC PROCEEDINGS

respectively, at the time of study entry. In baseline of studies examining the efficacy and safety of
analyses adjusted for expanded case-mix covari- thyroid hormone replacement therapy in these
ates, we observed that presumed untreated and populations. In a study of 2715 dialysis patients
undertreated hypothyroidism as well as untreated whose TSH levels and medication status were
hyperthyroidism were associated with higher concurrently examined at study entry, we found
mortality risk (reference, spontaneous euthyroid- that patients who were presumed to be hypo-
ism): aHRs (95% CI), 1.20 (1.16-1.24), 1.13 thyroid treated-to-target had similar mortality
(1.07-1.20), and 1.23 (1.18-1.28), respectively vs patients who were spontaneously euthyroid,
(Figure 4A-B; Supplemental Table 10 available whereas those with undertreated or untreated
online at http://www.mayoclinicproceedings.org). In hypothyroidism had higher mortality risk.20
time-dependent analyses, the same patterns of as- Although these limited data suggest benefit, thy-
sociation were observed. In addition, hypothy- roid hormone supplementation is a therapy that
roidism treated-to-target TSH range was carries risk, given its (1) narrow toxic-to-
associated with lower mortality risk (aHR [95% therapeutic window,37 (2) augmentation of pro-
CI], 0.88 [0.85-0.91]) (Figure 4C-D; tein catabolism and protein-energy wasting38
Supplemental Table 10). (ie, one of the strongest death predictors in pa-
tients with advanced CKD39), and (3) potential
DISCUSSION to precipitate adverse cardiovascular sequelae
To our knowledge, this is the first study that (eg, atrial fibrillation and angina) with overtreat-
has examined the association between thyroid ment.18,40 Using the national VA database with
status defined by repeated measures of serum exceptional capture of prescription dispensation
TSH with mortality in patients with NDD- data over time,29 in both baseline and time-
CKD. Among a large national cohort of US dependent analyses we found that untreated
veterans with stage 3 CKD, granular examina- and undertreated hypothyroidism and un-
tion of thyroid status showed that TSH levels treated hyperthyroidism were each associated
in the high-normal (TSH
3.0 mIU/L) and with higher mortality risk compared with spon-
hyperthyroid (TSH <0.5 mIU/L) ranges were taneous euthyroidism, whereas hypothyroidism
independently associated with higher mortal- treated-to-target (ie, TSH 0.5-5.0 mIU/L based
ity risk. These associations persisted across on general population thresholds17) had similar
multiple secondary and sensitivity analyses to slightly lower mortality risk. In time-
that considered baseline and time-dependent dependent analyses, there also appeared to be
(ie, repeated) TSH measures (as a proxy of a trend between overtreated hypothyroidism
long-term and short-term associations, respec- and lower mortality risk, although only margin-
tively); incrementally adjusted for potential ally decreased. Although mild medication-
confounders of the thyroid status-mortality induced thyrotoxicosis may be tolerated in the
associations across multivariable models, short-term in particular patient groups (ie,
including quality-of-care indicators (eg, younger patients), further studies are needed
ACEI/ARB or statin use, receipt of the influ- to determine the long-term safety and effective-
enza vaccination, and history of noncompli- ness of thyroid-modulating therapy and specific
ance); and examined multiple clinically TSH treatment targets.
relevant subgroups. In sensitivity analyses As another noteworthy finding, our cate-
that identified cases of incident thyroid dis- gorical analyses showed that incrementally
ease, we also observed that TSH levels in the higher TSH levels even in the high-normal
hypothyroid and hyperthyroid range were range (TSH
3.0 mIU/L) were associated
each associated with higher mortality risk. with increasingly higher mortality risk. Several
Our study also adds new knowledge to the other studies have similarly found that TSH
field regarding the impact of thyroid- levels exceeding 3.0 mIU/L are associated
modulating therapy on the survival of patients with higher mortality in dialysis patients.8,20
with kidney disease. Although United States Even in the general population, there remains
Renal Data System data show that levothyroxine considerable controversy regarding the optimal
is one of the most commonly prescribed medi- upper limit of the target TSH range. Some ex-
cations among NDD-CKD and ESRD Medicare perts recommend a therapeutic TSH upper
Part D beneficiaries,36 there have been a paucity limit of 2.5 to 3.0 mIU/L,19,41,42 whereas others
n n
582 Mayo Clin Proc. May 2018;93(5):573-585 https://doi.org/10.1016/j.mayocp.2018.01.024
www.mayoclinicproceedings.org
THYROID STATUS IN CKD

favor using upper TSH laboratory reference Furthermore, that the vast majority of our
range values (w5.0 mIU/L).43,44 Given the po- source population (82.3% of >3.5 million pa-
tential risks of undertreatment and overtreat- tients) underwent TSH testing provides further
ment, rigorous prospective studies are reassurance regarding the study’s external val-
warranted to determine the optimal TSH target idity. Second, the categories of thyroid status
range in the CKD and ESRD populations. are generally classified on the basis of serum
Although thyroid dysfunction was linked TSH and FT4 measurements. Given the sparsity
with higher death risk across almost all sub- of concurrent FT4 and free triiodothyronine
groups, it bears mention that stronger measurements and their unclear reliability in
hypothyroidism-mortality associations were advanced CKD (ie, peripheral conversion of
particularly observed among subgroups of T4-to-triiodothyronine is sensitive to nonthyr-
younger vs older age, black vs nonblack race, oidal illness; routinely used FT4 assays are
and those with vs without underlying CHF, dependent on protein-hormone binding, and
corroborating emerging data in the general the presence of uremic toxins that interfere
population.45 Because some experts advise with protein-hormone binding may lead to
that the “normal” TSH range may differ spurious levels),1,2,47-49 we elected to define
according to age and race/ethnicity,46 future thyroid status on the basis of TSH levels only.
studies are needed to more precisely define Although some aberrations of TSH have been
the ideal TSH thresholds among these CKD described in the context of CKD,50,51 it remains
subgroups. Indeed, the increase in the upper a more robust metric of thyroid status particu-
reference for serum TSH with age may be larly in the setting of underlying illness (ie,
especially relevant in the older NDD-CKD TSH levels typically remain normal in mild-
population, as most now recommend targeting moderate nonthyroidal illness, and become
a higher TSH for the elderly general popula- suppressed only in severe critically ill
tion on levothyroxine therapy.44 Furthermore, states1,2,52), and serum TSH has been strongly
previous studies in the general population associated with outcomes in our previous
have suggested that those with CHF may studies of dialysis patients.8,9,20,21 In addition,
have lower reserve with which to tolerate we observed persistent associations between
hemodynamic, volume, and electrophysiologic TSH alterations and mortality in sensitivity an-
changes associated with hypothyroidism. alyses that incorporated a 30-day lag period and
Hence, patients with NDD-CKD who have restricted analyses to incident cases of thyroid
structural heart disease may be particularly disease, hence excluding patients who died
vulnerable to the ill effects of hypothyroidism, immediately after TSH measurement to mini-
and future studies are needed to determine the mize risk of observing a reverse-causal associa-
specific mechanistic pathways linking thyroid tion (ie, severe illness may lead to TSH
perturbations and death in this population. alterations as well as heightened mortality
The strengths of our study include its exam- risk). Third, because our analyses included pa-
ination of a large national cohort of patients tients who were both incident (ie, treatment-
with NDD-CKD with comprehensive availabil- naive) and prevalent (ie, existing) users of thy-
ity of detailed patient-level information, roid hormone supplementation and antithyroid
including longitudinal laboratory and prescrip- medications, we cannot exclude potential bias
tion data; a stable study population to examine towards a protective effect of thyroid-
the long-term and short-term associations of modulating therapy (ie, patients receiving treat-
thyroid status, thyroid-modulating therapy, ment at study entry may have been healthier
and outcomes; and reduced confounding by than those who stopped or died using treatment
differential health care access. Several limita- before study entry because of adverse effects).53
tions of our study should be acknowledged. Fourth, given the retrospective nature of the
First, the indications for TSH testing in this study, we are not able to determine whether dif-
cohort, a requirement for inclusion in the study, ferential quality of care by providers or
are unknown. However, this inclusion criterion noncompliance with medical treatment may
was required of all patients irrespective of un- be a potential confounder of aberrant thyroid
derlying thyroid status and thus should not status and death risk. However, it should be
impair the study’s internal validity. noted that we observed robust thyroid status-

Mayo Clin Proc. n May 2018;93(5):573-585 n https://doi.org/10.1016/j.mayocp.2018.01.024 583

www.mayoclinicproceedings.org
 MAYO CLINIC PROCEEDINGS

mortality associations after incremental adjust- Grant Support: The work was supported by the National
ment for proxies of quality of care. Finally, as Institutes of Health/National Institute of Diabetes and Diges-
tive and Kidney Diseases research grants K23-DK102903
with all observational studies, we cannot
(C.M.R.), R03-DK114642 (C.M.R.), and R01-DK096920
exclude the possibility of residual confounding, (C.P.K., K.K.-Z.). The authors are also supported by the Na-
and our findings do not confirm a causal rela- tional Institutes of Health/National Institute of Diabetes and
tionship between thyroid status, thyroid- Digestive and Kidney Diseases grants K24-DK091419 (K.K.-
modulating therapy, and mortality risk. Z.), R01-DK092232 (D.V.N.), and U01-DK102163 (K.K.-Z.,
C.P.K.), as well as grants from the American Thyroid Asso-
ciation (C.M.R.) and the National Kidney Foundation
CONCLUSION
(C.M.R). The funding organizations/sponsors did not partic-
In conclusion, our study has found that higher ipate in the study design; collection, analysis, and interpreta-
TSH levels even in the high-normal range tion of the data; in the writing of the report; or in the
(
3.0 mIU/L) and lower TSH levels (<0.5 decision to submit the manuscript for publication.
mIU/L) were independently associated with
Data Previously Presented: Portions of these data have
higher mortality risk in a national cohort of been presented as an oral abstract at the 2016 American
patients with NDD-CKD. We also observed Society of Nephrology Kidney Week Meeting, November
that untreated and undertreated hypothyroidism 16-20, 2016, Chicago, IL.
and hyperthyroidism were associated with higher
Correspondence: Csaba P. Kovesdy, MD, Nephrology Sec-
mortality compared with spontaneous euthyr- tion, Memphis VA Medical Center, 1030 Jefferson Ave,
oidism, whereas hypothyroidism treated-to- Memphis, TN 38104 ([email protected]).
target was associated with similar to slightly
improved survival. At this time, rigorous REFERENCES
prospective studies are needed to determine the 1. Rhee CM. The interaction between thyroid and kidney disease:
precise optimal TSH range in patients with an overview of the evidence. Curr Opin Endocrinol Diabetes
NDD-CKD, and whether treatment to these tar- Obes. 2016;23(5):407-415.
2. Rhee CM, Brent GA, Kovesdy CP, et al. Thyroid functional dis-
gets ameliorates mortality risk in this population. ease: an under-recognized cardiovascular risk factor in kidney
disease patients. Nephrol Dial Transplant. 2015;30(5):724-737.
SUPPLEMENTAL ONLINE MATERIAL 3. Carrero JJ, Qureshi AR, Axelsson J, et al. Clinical and biochem-
ical implications of low thyroid hormone levels (total and free
Supplemental material can be found online at
forms) in euthyroid patients with chronic kidney disease. J Intern
http://www.mayoclinicproceedings.org. Sup- Med. 2007;262(6):690-701.
plemental material attached to journal articles 4. Chonchol M, Lippi G, Salvagno G, Zoppini G, Muggeo M, Targher G.
Prevalence of subclinical hypothyroidism in patients with chronic
has not been edited, and the authors take re-
kidney disease. Clin J Am Soc Nephrol. 2008;3(5):1296-1300.
sponsibility for the accuracy of all data. 5. Lo JC, Chertow GM, Go AS, Hsu CY. Increased prevalence of
subclinical and clinical hypothyroidism in persons with chronic
Abbreviations and Acronyms: ACEI = angiotensin- kidney disease. Kidney Int. 2005;67(3):1047-1052.
converting enzyme inhibitor; aHR = adjusted hazard ratio; 6. Rhee CM, Kalantar-Zadeh K, Streja E, et al. The relationship be-
ARB = angiotensin receptor blocker; BMI = body mass index; tween thyroid function and estimated glomerular filtration rate
CHF = congestive heart failure; CKD = chronic kidney disease; in patients with chronic kidney disease. Nephrol Dial Transplant.
eGFR = estimated glomerular filtration rate; ESRD = end- 2015;30(2):282-287.
7. Targher G, Chonchol M, Zoppini G, et al. Prevalence of thyroid
stage renal disease; FT4 = free thyroxine; IQR = interquartile
autoimmunity and subclinical hypothyroidism in persons with
range; NDD-CKD = nonedialysis-dependent chronic kidney chronic kidney disease not requiring chronic dialysis. Clin
disease; TSH = thyrotropin; VA = Veterans Affairs Chem Lab Med. 2009;47(11):1367-1371.
8. Rhee CM, Kim S, Gillen DL, et al. Association of thyroid functional
Affiliations (Continued from the first page of this disease with mortality in a national cohort of incident hemodial-
article.): A.S.Y.); Tibor Rubin Veterans Affairs Medical Cen- ysis patients. J Clin Endocrinol Metab. 2015;100(4):1386-1395.
ter, Long Beach, CA (K.K.-Z., E.S.); DaVita Clinical Research, 9. Rhee CM, Ravel VA, Streja E, et al. Thyroid functional disease
Minneapolis, MN (S.M.B.); Division of General Internal Med- and mortality in a national peritoneal dialysis cohort. J Clin Endo-
icine, University of California Irvine School of Medicine, Or- crinol Metab. 2016;101(11):4054-4061.
ange, CA (D.V.N.); Division of Endocrinology, Diabetes and 10. Hak AE, Pols HA, Visser TJ, Drexhage HA, Hofman A, Witteman JC.
Metabolism, David Geffen School of Medicine at UCLA, Los Subclinical hypothyroidism is an independent risk factor for athero-
sclerosis and myocardial infarction in elderly women: the Rotterdam
Angeles, CA (G.A.B.); Department of Medicine, Veterans
Study. Ann Intern Med. 2000;132(4):270-278.
Affairs Greater Los Angeles Healthcare System, Los
11. Klein I, Danzi S. Thyroid disease and the heart. Circulation. 2007;
Angeles, CA (G.A.B.); Nephrology Section, Memphis Veter- 116(15):1725-1735.
ans Affairs Medical Center, Memphis, TN (C.P.K.); and Divi- 12. Klein I, Ojamaa K. Thyroid hormone and the cardiovascular sys-
sion of Nephrology, University of Tennessee Health Science tem. N Engl J Med. 2001;344(7):501-509.
Center, Memphis, TN (C.P.K.). 13. Taddei S, Caraccio N, Virdis A, et al. Impaired endothelium-
dependent vasodilatation in subclinical hypothyroidism: benefi-
Potential Competing Interests: The authors report no cial effect of levothyroxine therapy. J Clin Endocrinol Metab.
competing interests. 2003;88(8):3731-3737.

n n
584 Mayo Clin Proc. May 2018;93(5):573-585 https://doi.org/10.1016/j.mayocp.2018.01.024
www.mayoclinicproceedings.org
THYROID STATUS IN CKD

14. Gencer B, Collet TH, Virgini V, et al; Thyroid Studies Collabo- 34. US Department of Veterans Affairs; Health Services Research
ration. Subclinical thyroid dysfunction and the risk of heart fail- and Development Service; VA Information Resource Center.
ure events: an individual participant data analysis from 6 VIReC Resource Guide: VA Corporate Data Warehouse. Hines,
prospective cohorts. Circulation. 2012;126(9):1040-1049. IL: VA Information Resource Center; 2012.
15. United States Renal Data System. 2016 USRDS Annual Data 35. Levey AS, Stevens LA, Schmid CH, et al. A new equation to
Report: Epidemiology of Kidney Disease in the United States. estimate glomerular filtration rate. Ann Intern Med. 2009;
Bethesda, MD: National Institutes of Health, National Institute 150(9):604-612. Erratum in: Ann Intern Med. 2011;155(6):408.
of Diabetes and Digestive and Kidney Diseases; 2016. 36. United States Renal Data System. 2012 USRDS Annual Data
16. Saran R, Li Y, Robinson B, et al. US Renal Data System 2015 Report: Epidemiology of Kidney Disease in the United States.
Annual Data Report: Epidemiology of Kidney Disease in the Bethesda, MD: National Institutes of Health, National Institute
United States. Am J Kidney Dis. 2016;67(3, Suppl 1). Svii, S1-S305. of Diabetes and Digestive and Kidney Diseases; 2012.
17. Ladenson PW. Diagnosis of hypothyroidism. In: Braverman LE, 37. Wartofsky L. Levothyroxine: therapeutic use and regulatory is-
Cooper DS, eds. Werner and Ingbar’s The Thyroid. 10th ed. Phil- sues related to bioequivalence. Expert Opin Pharmacother.
adelphia, PA: Lippincott Williams and Wilkins; 2013:606-611. 2002;3(6):727-732.
18. Jonklaas J. Treatment of hypothyroidism. In: Braverman LE, 38. Lim VS. Thyroid function in patients with chronic renal failure.
Cooper DS, eds. Werner and Ingbar’s The Thyroid. 10th ed. Phil- Am J Kidney Dis. 2001;38(4, Suppl 1):S80-S84.
adelphia, PA: Lippincott Williams and Wilkins; 2013:611-628. 39. Kalantar-Zadeh K, Rhee C, Sim JJ, Stenvinkel P, Anker SD,
19. Baloch Z, Carayon P, Conte-Devolx B, et al; Guidelines Commit- Kovesdy CP. Why cachexia kills: examining the causality of
tee, National Academy of Clinical Biochemistry. Laboratory med- poor outcomes in wasting conditions. J Cachexia Sarcopenia
icine practice guidelines: laboratory support for the diagnosis and Muscle. 2013;4(2):89-94.
monitoring of thyroid disease. Thyroid. 2003;13(1):3-126. 40. Sawin CT, Geller A, Wolf PA, et al. Low serum thyrotropin
20. Rhee CM, Alexander EK, Bhan I, Brunelli SM. Hypothyroidism concentrations as a risk factor for atrial fibrillation in older per-
and mortality among dialysis patients. Clin J Am Soc Nephrol. sons. N Engl J Med. 1994;331(19):1249-1252.
2013;8(4):593-601. 41. Baskin HJ, Cobin RH, Duick DS, et al. American Association of
21. Rhee CM, You AS, Nguyen DV, et al. Thyroid status and mor- Clinical Endocrinologists medical guidelines for clinical practice
tality in a prospective hemodialysis cohort. J Clin Endocrinol for the evaluation and treatment of hyperthyroidism and hypo-
Metab. 2017;102(5):1568-1577. thyroidism. Endocr Pract. 2002;8(6):457-469.
22. Drechsler C, Schneider A, Gutjahr-Lengsfeld L, et al. Thyroid 42. Wartofsky L, Dickey RA. The evidence for a narrower thyro-
function, cardiovascular events, and mortality in diabetic hemo- tropin reference range is compelling. J Clin Endocrinol Metab.
dialysis patients. Am J Kidney Dis. 2014;63(6):988-996. 2005;90(9):5483-5488.
23. Kovesdy CP, Norris KC, Boulware LE, et al. Association of race 43. Garber JR, Cobin RH, Gharib H, et al; American Association of
with mortality and cardiovascular events in a large cohort of US Clinical Endocrinologists and American Thyroid Association
veterans. Circulation. 2015;132(16):1538-1548. Taskforce on Hypothyroidism in Adults. Clinical practice guide-
24. Lu JL, Molnar MZ, Naseer A, Mikkelsen MK, Kalantar-Zadeh K, lines for hypothyroidism in adults: cosponsored by the American
Kovesdy CP. Association of age and BMI with kidney function Association of Clinical Endocrinologists and the American Thy-
and mortality: a cohort study. Lancet Diabetes Endocrinol. roid Association. Thyroid. 2012;22(12):1200-1235.
2015;3(9):704-714. 44. Jonklaas J, Bianco AC, Bauer AJ, et al; American Thyroid Asso-
25. Molnar MZ, Kalantar-Zadeh K, Lott EH, et al. Angiotensin-con- ciation Task Force on Thyroid Hormone Replacement. Guide-
verting enzyme inhibitor, angiotensin receptor blocker use, and lines for the treatment of hypothyroidism: prepared by the
mortality in patients with chronic kidney disease. J Am Coll Car- American Thyroid Association Task Force on Thyroid Hor-
diol. 2014;63(7):650-658. mone Replacement. Thyroid. 2014;24(12):1670-1751.
26. Novak M, Mucsi I, Rhee CM, et al. Increased risk of incident 45. Rhee CM, Curhan GC, Alexander EK, Bhan I, Brunelli SM. Sub-
chronic kidney disease, cardiovascular disease, and mortality clinical hypothyroidism and survival: the effects of heart failure
in patients with diabetes with comorbid depression. Diabetes and race. J Clin Endocrinol Metab. 2013;98(6):2326-2336.
Care. 2016;39(11):1940-1947. 46. Surks MI, Boucai L. Age- and race-based serum thyrotropin
27. Dekker FW, de Mutsert R, van Dijk PC, Zoccali C, Jager KJ. Sur- reference limits. J Clin Endocrinol Metab. 2010;95(2):496-502.
vival analysis: time-dependent effects and time-varying risk fac- 47. Soldin OP. Measuring serum thyroid-stimulating hormone, thy-
tors. Kidney Int. 2008;74(8):994-997. roid hormones, thyroid-directed antibodies, and transport pro-
28. US Department of Veterans Affairs; Health Services Research teins. In: Braverman LE, Cooper DS, eds. Werner and Ingbar’s
and Development Service; VA Information Resource Center. The Thyroid. 10th ed. Philadelphia, PA: Lippincott Williams and
VIReC Research User Guide: VHA Decision Support System Clinical Wilkins; 2013:279-297.
National Data Extracts. 2nd ed. Hines, IL: VA Information 48. Kaptein EM. Thyroid hormone metabolism and thyroid diseases
Resource Center; 2009. in chronic renal failure. Endocr Rev. 1996;17(1):45-63.
29. US Department of Veterans Affairs; Health Services Research 49. Langton JE, Brent GA. Nonthyroidal illness syndrome: evalua-
and Development Service; VA Information Resource Center. tion of thyroid function in sick patients. Endocrinol Metab Clin
VIReC Research User Guide: VHA Pharmacy Prescription Data. North Am. 2002;31(1):159-172.
2nd ed. Hines, IL: VA Information Resource Center; 2008. 50. Carrero JJ, Stenvinkel P, Lindholm B. Endocrine aspects of
30. Sohn MW, Arnold N, Maynard C, Hynes DM. Accuracy and chronic kidney disease. In: Taal MW, Chertow GM,
completeness of mortality data in the Department of Veterans Marsden PA, Skorecki K, Yu AS, Brenner BM, eds. Taal: Brenner
Affairs. Popul Health Metr. 2006;4:2. and Rector’s The Kidney. 9th ed. Philadelphia, PA: Elsevier Saun-
31. Stroupe KT, Tarlov E, Zhang Q, Haywood T, Owens A, ders; 2012:2122-2137.
Hynes DM. Use of Medicare and DOD data for improving 51. Kaptein EM, LoPresti JS, Kaptein MJ. Is an isolated TSH elevation
VA race data quality. J Rehabil Res Dev. 2010;47(8):781-795. in chronic nonthyroidal illness “subclinical hypothyroidism”?
32. US Department of Veterans Affairs; Health Services Research J Clin Endocrinol Metab. 2014;99(11):4015-4026.
and Development Service; VA Information Resource Center. 52. Wiersinga WM, Van den Berghe G. Nonthyroidal illness syn-
VIReC Research User Guide: VHA Medical SAS Datasets FY2006- drome. In: Braverman LE, Cooper DS, eds. Werner and Ingbar’s
2007. Hines, IL: VA Information Resource Center; 2007. The Thyroid. 10th ed. Philadelphia, PA: Lippincott Williams and
33. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbid- Wilkins; 2013:203-216.
ity index for use with ICD-9-CM administrative databases. J Clin 53. Ray WA. Evaluating medication effects outside of clinical trials:
Epidemiol. 1992;45(6):613-619. new-user designs. Am J Epidemiol. 2003;158(9):915-920.

Mayo Clin Proc. n May 2018;93(5):573-585 n https://doi.org/10.1016/j.mayocp.2018.01.024 585

www.mayoclinicproceedings.org