Content uploaded by Hong Xu
Author content
All content in this area was uploaded by Hong Xu on Mar 02, 2018
Content may be subject to copyright.
Original Investigation
Thyroid Function Test Derangements and Mortality in Dialysis
Patients: A Systematic Review and Meta-analysis
Hong Xu, MD,
1,2
Nele Brusselaers, MD, PhD,
3
Bengt Lindholm, MD, PhD,
1,2
Carmine Zoccali, MD, PhD,
4
and Juan Jesu
´s Carrero, PhD
1,2,3
Background: We evaluated current evidence associating thyroid function test result derangements with risk
for mortality in patients with chronic kidney failure treated by long-term dialysis.
Study Design: Systematic review and meta-analysis of cohort studies.
Setting & Population: Dialysis patients.
Selection Criteria for Studies: We searched PubMed, Web of Science, Science Citation Index, Cochrane
Library, and Embase databases from inception through December 2015.
Predictors: Hypothyroidism (thyrotropin level greater than reference range) and low triiodothyronine (T
3
)
and thyroxine (T
4
) levels.
Outcomes: All-cause and cardiovascular mortality.
Results: 12 studies involving 14,766 participants (4,450 deaths) were identified. Of those, 6 studies pro-
vided data for cardiovascular mortality (2,772 participants with 327 cardiovascular deaths). Overall, confidence
in the available evidence was moderate. Pooled adjusted HRs for all-cause mortality associated with
hypothyroidism, low T
3
level, and low T
4
level were 1.24 (95% CI, 1.14-1.34), 1.67 (95% CI, 1.23-2.27),
and 2.40 (95% CI, 1.47-3.93), respectively. Pooled adjusted HRs for cardiovascular mortality associated
with low T
3
and T
4
levels were 1.84 (95% CI, 1.24-2.74) and 3.06 (95% CI, 1.29-7.24), respectively.
Limitations: Fewer studies reporting on T
4
and thyrotropin outcomes.
Conclusions: In patients treated with long-term dialysis, (cardiovascular) mortality is consistently higher in
the presence of thyroid function test result derangements.
Am J Kidney Dis. 68(6):923-932. ª2016 by the National Kidney Foundation, Inc.
INDEX WORDS: Thyroid disorders; hypothyroidism; triiodothyronine (T
3
); thyroxine (T
4
); hormones; all-cause
mortality; cardiovascular mortality; thyroid function test derangement; haemodialysis; peritoneal dialysis;
endocrine; end-stage renal disease (ESRD); meta-analysis.
In patients with chronic kidney disease (CKD), the
progressive loss of kidney function has a negative
impact on the synthesis, excretion, metabolism, and
degradation of thyroid hormones and their metabo-
lites.
1,2
As a consequence, thyroid function test result
derangements are common in patients with advanced
stages of CKD, particularly in those with end-stage
renal disease (ESRD).
1,3-8
The prevalence of clini-
cally overt and subclinical hypothyroidism increases
with worsening kidney function.
4-6
In addition, low
circulating levels of triiodothyronine (T
3
)
1,3,7,8
and
thyroxine (T
4
) become increasingly common.
9,10
Uremic conditions that are thought to contribute to
these alterations are multiple, including retention of
iodine and toxins (causing central thyrotropin
[thyroid-stimulating hormone] inhibition, triggering
thyrotropin clearance, and influencing T
3
levels
independently of thyroid function), ineffective protein
binding of T
4
, reduced T
4
levels in tissues, and pri-
marily, impaired conversion of T
4
into T
3
. The latter
is attributed to direct effects of systemic inflamma-
tion, elevated cortisol levels, malnutrition, mineral
deficiency (eg, selenium resulting in reduced deiodi-
nase activity), metabolic acidosis, commonly used
medications, and additionally in patients with ESRD,
effects of dialytic procedures (eg, peritoneal effluent
losses).
1,3,8,11,12
Observational studies in recent years have attemp-
ted to link these thyroid function test result de-
rangements with the cardiovascular complications and
elevated mortality risk of patients with CKD and
ESRD.
9,10,13-23
If associations are causal, this opens
perspectives for thyroid replacement therapy in this
high-risk patient population. Here, we evaluate the
consistency of reported associations between thyroid
function test result derangements and hard end points
From the Divisions of
1
Renal Medicine and
2
Baxter Novum,
Department of Clinical Science, Intervention and Technology, and
3
Department of Molecular Medicine and Surgery, Karolinska
Institutet, Stockholm, Sweden; and
4
Division of Nephrology,Dialysis
and Kidney Transplantation, CNR Hospital, Reggio Calabria, Italy.
Received February 2, 2016. Accepted in revised form June 29,
2016. Originally published online September 2, 2016.
Address correspondence to Juan Jesús Carrero, PhD, Divisions
of Renal Medicine and Baxter Novum, Karolinska University
Hospital at Huddinge M99, Karolinska Institutet, SE-14186
Stockholm, Sweden. E-mail: juan.jesus.carrero@ki.se
2016 by the National Kidney Foundation, Inc.
0272-6386
http://dx.doi.org/10.1053/j.ajkd.2016.06.023
Am J Kidney Dis. 2016;68(6):923-932 923
in patients undergoing long-term dialysis by means of
a systematic review and meta-analysis.
METHODS
Data Sources and Searches
We systematically searched PubMed and the Web of Science.
Complementary searches and backward and forward citation
tracking were performed through analyses of reference lists, the
Science Citation Index, Cochrane Library, and Embase. The
search was from inception through December 2015. We also
searched unpublished studies and gray literature in a clinical trial
register (www.ClinicalTrials.gov) and conference abstracts for the
major nephrology conferences during 2014 to 2015: American
Society of Nephrology Kidney Week, European Renal Associa-
tion/European Dialysis and Transplant Association Congress, and
International Society of Nephrology Congress. The search string
consisted of 3 parts: (1) the exposure (ie, thyroid disease, hypo-
thyroidism, thyrotropin, T
4
, and T
3
), (2) study population (ie,
CKD, ESRD, kidney failure, uremia, hemodialysis [HD], and
peritoneal dialysis [PD]), and (3) outcomes (ie, all-cause mortality,
cardiovascular mortality, survival, fatal, and death). Different
spellings were accounted for, and Medical Subject Headings
(MeSH) were incorporated in the PubMed search (Item S1,
available as online supplementary material).
Exposure, Study Population, and Outcome
The exposure was thyroid function test result derangements,
defined as the following: (1) low T
3
level (as measured by total or
free T
3
, either free T
3
level less than the assay-specific reference
range or free T
3
level less than the cutoff value), (2) low T
4
level
(by total or free T
4
, either free T
4
level lower than the reference
range or free T
4
level less than the cutoff value), and (3) hypo-
thyroidism (thyrotropin level greater than the reference range).
Patient groups with free T
3
, free T
4
, and thyrotropin levels within
the normal range or within the highest level category as reported
by each study were used as reference. The study populations
consisted of adults with CKD undergoing long-term dialysis,
either HD or PD.
24
Study outcomes were all-cause and/or car-
diovascular mortality during a minimal follow-up of the study
cohort of 3 months.
Inclusion and Exclusion Criteria
Studies were considered for inclusion in the meta-analysis if
they: (1) presented data for measured thyroid function test in
adult (aged $18 years) patients with CKD undergoing dialysis
and (2) provided data for all-cause and/or cardiovascular mor-
tality associated with these measurements. Both cohort studies
and case-control studies were eligible, whereas case reports,
case series, and review articles were excluded. We did not
consider studies addressing a combination of these exposures
(eg, hypothyroidism and low free T
3
level and low free T
4
level
or thyroid function test result derangements with a concurrent
comorbid condition or lifestyle factor). No language restriction
was applied. The languages selected a priori as eligible were
English, Chinese, Swedish, Spanish, French, Dutch, and
German. Studies were eligible only if hazard ratios (HRs) of
thyroid function tests for all-cause or cardiovascular mortality
were reported.
Study Selection
An a priori established study protocol was applied (Item S2).
The search method used to identify all relevant articles was dis-
cussed and developed by the authors and the final search string
was approved by all. The initial search was performed by 2 re-
viewers (H.X. and N.B.), who eliminated clearly irrelevant articles
based on the title and abstract as defined by the preset selection
criteria. The final selection of articles was made by mutual
consideration of all authors, based on the reporting of all necessary
data and in accordance with the predefined inclusion and exclusion
criteria.
Data Extraction and Quality Assessment
For each article identified, we extracted information for study
and participant characteristics, thyroid function test description,
and analysis strategy (statistical models and adjustment for
covariates). For each study, crude HRs were extracted (if re-
ported), as well as HRs based on the most fully adjusted Cox
regression models. If different thyroid function test results were
reported in one study (eg, low free T
3
or low free T
4
levels or
thyrotropin level greater than the reference range), all HRs of the
different exposures were extracted. If several level groups (eg,
tertiles of free T
3
) were reported, the most extreme comparison,
that is, lowest versus highest level, was considered for the pri-
mary results. We contacted the authors for clarifications of the
protocol and provision of HRs in categorical groupings. Data
analysis used HRs based on the most adjusted (final) Cox
regression model in each study. Risk of bias was assessed using
the Newcastle-Ottawa Scale tool.
25
Assessment of quality and
generalizability was based on 3 key broad domains considered
fundamental for observational studies: selection of study partic-
ipants, comparability of cohorts on the basis of the design or
analysis, and assessment of outcomes. Study-level risk of bias
was assessed by 2 authors (H.X. and N.B.), and disagreements in
ratings were discussed until consensus. As an overall quality
check and in order to ensure transparent reporting of this sys-
tematic review and meta-analysis, the Meta-analysis of Obser-
vational Studies in Epidemiology (MOOSE) and Preferred
Reporting Items for Systematic Reviews and Meta-analyses
(PRISMA) guidelines were followed.
Statistical Analysis
DerSimonian-Laird random-effect meta-analysis and empirical
Bayes metaregression models were performed with STATA,
version 13.0 (StataCorp LP) and were based on the HRs and
standard errors. Values were reported by a forest plot, and un-
certainty about the pooled estimates was quantified by 95% con-
fidence intervals (CIs). Statistical heterogeneity was assessed by
means of Cochran Qtest and I
2
test. I
2
represents the percentage of
variation attributable to heterogeneity, which was categorized as
low (0%-50%), moderate (51%-75%), or high (.75%).
26
We could perform additional empirical Bayes metaregression
models in studies addressing low free T
3
levels as the exposure.
These included type of free T
3
level ascertainment (less than the
reference range or cutoff value), type of T
3
measurements (free or
total T
3
), type of dialysis therapy (HD or PD), mean follow-up
(12-36 or .36 months), study sample size (,500 or $500 par-
ticipants), confounders in fully adjusted models (with or without
adjustment for malnutrition, inflammation, and comorbid condi-
tions), and reported regression models. We also did a sensitivity
analysis to further explore the robustness of results and identify
any study that may have exerted a disproportionate influence on
the summary effect of low free T
3
level on mortality risk. The
presence of small study effects and publication bias was evaluated
by Begg or Egger regression asymmetry analysis.
27
RESULTS
Study Selection
We identified a total of 3,962 publications, of
which 3,479 remained after removing duplicates
(Fig 1). We excluded 3,448 publications based on the
title and abstract because they were unrelated to the
Xu et al
924 Am J Kidney Dis. 2016;68(6):923-932
study of association between thyroid function test
result derangements and mortality. Of the remaining
31 articles, we excluded 10 articles that did not meet
inclusion criteria after full-text screening (1 study was
excluded because of a mixed definition of exposure,
28
1 study included non–dialysis-dependent patients with
CKD,
15
and 4 other studies did not present estimates
for a comparison group
29-32
). We further excluded 9
studies because they investigated cardiovascular sur-
rogates
33-41
or cardiovascular disease events,
37
but
not mortality risk (detailed in Table S1). Twelve
studies met eligibility criteria and were considered for
meta-analysis.
Study Characteristics
The 12 studies selected for analysis enrolled a total
of 14,766 participants (Table 1), of whom 4,450 died.
Six studies provided data for cardiovascular mortality,
including 2,772 participants and 327 cardiovascular-
related deaths. The studies were from Sweden
(n 53),
9,14,20
South Korea (n 52),
10,21
Italy
(n 52),
13,18
the United States (n 52),
16,17
Turkey
(n 51),
19
Germany (n 51),
22
and Greece (n 51).
23
None of them showed an overlap in geographical area
or time.
Mean age ranged from 51 to 66 years, and the
proportion of men ranged from 51% to 74%. Mean
duration of follow-up varied between 12 and 55
months. Two studies analyzed several exposures: one
addressed both low T
3
and low T
4
levels
9
and the
other addressed low T
3
levels and hypothyroidism.
22
Nine studies reported HRs for death associated with
low free T
3
levels, 2 in relation to low free T
4
levels,
and 3 in relation to hypothyroidism. Among the 9
studies that studied low free T
3
levels, 4 defined it by
the reference range,
19,21-23
and 5, by tertiles of dis-
tribution,
9,13,18
median,
20
or receiver operating char-
acteristic–derived
14
cutoffs; 6 studies used
measurements of free T
3
,
13,18-20,22,23
and 3 used
measurements of total T
3
.
9,14,21
Seven studies
included HD patients, 3 studies included PD patients,
and 2 studies included both HD and PD patients.
14,16
Quality Assessment
All studies were population-based cohort studies
and had appropriate methods for thyroid function test
Records idenƟfied via PubMed
(n=2190) and Web of Science (n=1613)
Total (n=3803)
AddiƟonal records idenƟfied through
Chinese Journal database
(n = 159)
Records aŌer duplicates removed
(n =3479)
Records screened
(n = 3479)
Records excluded
(n =3448)
Full-text arƟcles assessed for
eligibility
(n = 31)
Full-text arƟcles excluded
(n = 10)
No comparison group (n=4)
Mixed definiƟon of exposure:
(n=1)
Non-dialysis paƟents (n=1)
Review (n=4)
Studies included in
qualitaƟve synthesis
(n = 21)
Studies included in
quanƟtaƟve synthesis (meta-
analysis)
(n = 12)
Not addressing mortality
(n=9)
Figure 1. Flow chart for study inclusion; adapted from PRISMA (Preferred Reporting of Systematic Reviews and Meta-analyses).
Am J Kidney Dis. 2016;68(6):923-932 925
Thyroid Function and Mortality in Dialysis Patients
Table 1. Description and Characteristics of 12 Observational Studies Reporting on the Association Between Thyroid Function Test Derangements and Risk for Mortality
Study Country Cohort N
Mean
Age, y
Male
Sex, %
Exposure
Mortality Events
Mean
F/U, moLow (f) T
3
Level
Low (f)
T
4
Level Hypothyroidism
a
Zoccali
18
(2006) IT HD 200 61 53 ,33rd percentile NA NA All-cause 102 42
Enia
13
(2007) IT PD 41 66 63 ,33rd percentile NA NA All-cause 27 34
Carrero
14
(2007) SE HD1PD;
euthyroid pts
187 55 63 ,Cutoffs derived
from receiver
operating
characteristics
NA NA All-cause
and CVD
66 (34 CVD) 20
Ozen
19
(2011) TR HD;
euthyroid pts
669 54 56 ,Reference range NA NA All-cause
and CVD
165 (94 CVD) 34
Meuwese
9
(2012) SE HD 210 62 55 ,66th percentile ,66th
percentile
NA All-cause
and CVD
103 (40 CVD) 38
Meuwese
20
(2013) SE PD 84 64 68 ,Median NA NA All-cause 24 32
Koo
21
(2013) KR HD 471 57 57 ,Reference range NA NA All-cause
and CVD
49 (22 CVD) 24
Rhee
16
(2013) US HD1PD 2,715 63 60 NA NA .Reference range All-cause 917 20
Drechsler
22
(2014) DE HD 1,000 66 53 ,Reference range
in euthyroid pts
NA .Reference range,
with normal (f)
T
3
and (f) T
4
All-cause
and CVD
477 (131 CVD) 48
Jung
10
(2014) KR PD 235 51 56 NA ,Median NA All-cause
and CVD
31 (6 CVD) 24
Fragidis
23
(2015) GR HD;
euthyroid pts
114 62 74 ,Reference range NA NA All-cause 69 55
Rhee
17
(2015) US HD 8,840 65 51 NA NA .Reference range,
with normal (f) T
4
All-cause 2,420 12
Note: Euthyroid patients are defined as having both thyrotropin and T
4
levels within the reference ranges.
Abbreviations: CVD, cardiovascular disease; DE, Germany; (f) T
3
, (free) triiodothyronine; (f) T
4
, (free) thyroxine; F/U, follow-up; GR, Greece; HD, hemodialysis; IT, Italy; KR, Republic of
Korea; NA, not applicable; PD, peritoneal dialysis; pts, patients; SE, Sweden; TR, Turkey; TSH, thyrotropin; US, United States.
a
Thyrotropin level greater than reference value.
926 Am J Kidney Dis. 2016;68(6):923-932
Xu et al
measurements. Six studies defined exposure cutoffs
with assay reference ranges,
16,17,19,21-23
and the others
used cohort-specific cutoffs (eg, tertiles or median).
Four studies provided a comparison of baseline pa-
tient characteristics according to the analyzed expo-
sures.
9,14,16,20
All studies reported mortality follow-up
with cause of death ascertainment from medical re-
cords, describing crude and adjusted HRs (Figs S1
and S2). The covariates used in multivariable
adjustment are detailed in Table S2. Seven studies
considered multivariable adjustment for systemic
inflammation biomarkers, which are presumably on
the causal pathway of study exposure and
outcome
9,10,13,14,18,21,23
; and 2 studies further
adjusted for nutritional status and comorbid
conditions.
9,21
Thyroid Function Test Result Derangements and
Death
The pooled adjusted HR for all-cause mortality
associated with low free T
3
level was 1.67 (95%
CI, 1.23-2.27), with moderate heterogeneity
(I
2
552.1%). The adjusted HR for low free T
4
level
was 2.40 (95% CI, 1.47-3.93; I
2
50%), and for
hypothyroidism, 1.24 (95% CI, 1.14-1.34; I
2
50%).
These estimates presented low heterogeneity (Fig 2).
Pooled adjusted HRs for cardiovascular mortality
associated with low free T
3
(HR, 1.84; 95% CI,
1.24-2.74; I
2
528.8%) and low free T
4
levels (HR,
3.06; 95% CI, 1.29-7.24; I
2
50%) showed similar
but stronger HRs, with low heterogeneity (Fig 3).
Metaregression models showed similar results but
with broader CIs: the HR for all-cause mortality
associated with low free T
3
levels was 1.70 (95%
CI, 1.16-2.50), and with low free T
4
levels, 2.40
(95% CI, 0.09-59); the HR for cardiovascular mor-
tality associated with low free T
3
levels was 1.84
(95% CI, 1.05-3.21), and with low free T
4
levels,
3.06 (95% CI, 0.01-820). Analysis of publication
bias through funnel plots with Begg or Egger tests
could not be performed because of statistical
heterogeneity.
42
NOTE: Weights are from random effects analysis
.
.
.
Low (f)T3 vs high or normal range
Zoccali
Enia
Carrero
Ozen
Meuwese
Meuwese
Koo
Drechsler
Fragidis
Subtotal (I-squared = 52.1%, p = 0.034)
Low (f)T4 vs high
Meuwese
Jung
Subtotal (I-squared = 0.0%, p = 0.668)
High TSH vs normal range
Rhee
Drechsler
Rhee
Subtotal (I-squared = 0.0%, p = 0.707)
Study
2006
2007
2007
2011
2012
2013
2013
2014
2015
2012
2014
2013
2014
2015
Year
HD
PD
HD+PD
HD
HD
PD
HD
HD
HD
HD
PD
HD+PD
HD
HD
Population
200
41
187
669
210
84
471
1000
114
210
235
2715
1000
8840
Size
2.68 (1.49, 4.84)
7.85 (1.61, 38.38)
1.90 (1.10, 3.40)
1.08 (0.73, 1.61)
1.60 (1.00, 2.60)
2.40 (0.70, 8.60)
4.54 (0.87, 30.94)
1.04 (0.70, 1.54)
1.61 (0.88, 2.92)
1.67 (1.23, 2.27)
2.20 (1.20, 4.30)
2.74 (1.25, 5.90)
2.40 (1.47, 3.93)
1.27 (1.06, 1.52)
1.55 (0.85, 2.87)
1.22 (1.11, 1.34)
1.24 (1.14, 1.34)
HR (95% CI)
12.89
3.25
13.42
17.45
15.40
4.81
2.63
17.48
12.67
100.00
59.65
40.35
100.00
21.05
1.85
77.11
100.00
Weight
%
All-cause mortality
1.5 1 2 5 10 20 30 40
Figure 2. Forest plot depicts the meta-association between various forms of thyroid function test result derangements and risk for
all-cause mortality, using the Dersimonian and Laird random-effects model. All hazard ratios (HRs) are based on the most fully
adjusted reported model. Abbreviations: CI, confidence interval; (f) T
3
, (free) triiodothyronine; (f) T
4
, (free) thyroxine; HD, hemodialysis;
PD, peritoneal dialysis; TSH, thyrotropin.
Am J Kidney Dis. 2016;68(6):923-932 927
Thyroid Function and Mortality in Dialysis Patients
Metaregression and Sensitivity Analyses of Low T
3
Exclusion of single studies from the analysis did
not alter the main findings (Table S3). Metaregression
analyses suggested that studies defining low free T
3
level by cohort-specific cutoffs (as compared with
studies using reference ranges) and studies using total
T
3
measurements (as compared with studies
measuring free T
3
) tended to have stronger associa-
tions with mortality (Table 2). HD patients with
longer follow-up and larger sample size had lower
HRs as compared with their counterparts. Studies that
considered multivariable adjustment for malnutrition,
inflammation, and comorbid conditions showed
higher HRs (Table 2). Due to an insufficient number
of studies, no subgroup analysis for patients by low
T
4
levels and hypothyroidism could be performed.
DISCUSSION
In this meta-analysis, risk for all-cause mortality
and cardiovascular-related mortality was consistently
higher in patients undergoing dialysis with thyroid
function test result derangements. This association
persisted throughout a number of sensitivity and
stratified analyses.
Meta-analysis can be limited by the comprehen-
siveness of searches, the methodological rigor of
included studies, and publication bias, especially
when the meta-analysis includes, as in the current
study, observational studies rather than randomized
controlled trials. We consider the extensive literature
evaluation as a strength of the analysis, but
acknowledge that the number of retrieved articles was
relatively small, reflecting the scarcity of literature on
this topic. The pooled HRs are dependent on certain
traits of the published studies—availability, quality,
and methods—and these might be hampered by sta-
tistical heterogeneity and publication bias. We
acknowledge a number of limitations that need to be
considered when interpreting our findings. First, by
excluding studies that did not report death outcomes,
we cannot rule out the possibility of selection bias.
Second, our analysis plan selected the most adjusted
HR presented in the studies, which despite presenting
the most conservative risk estimation, may result in
outcome reporting bias. Because of statistical het-
erogeneity, funnel plots for detecting publication bias
with the Begg or Egger test were considered not
feasible.
42
We attempted to mitigate these biases by
in-depth metaregression analyses, observing alto-
gether a general coherence with the main metafind-
ings. Because we based our search on English
language2dominated sources, language bias cannot
be excluded. Finally, and regarding the study
NOTE: Weights are from random effects analysis
.
.
Low (f)T3 vs high or normal range
Carrero
Ozen
Meuwese
Koo
Drechsler
Subtotal (I-squared = 28.8%, p = 0.229)
Low (f)T4 vs high
Meuwese
Jung
Subtotal (I-squared = 0.0%, p = 0.325)
Study
2007
2011
2012
2013
2014
2012
2014
Year
HD+PD
HD
HD
HD
HD
HD
PD
Population
187
669
210
471
1000
210
235
Size
3.10 (1.40, 7.10)
1.46 (0.89, 2.37)
2.70 (1.20, 6.30)
2.74 (0.92, 11.40)
1.12 (0.59, 2.30)
1.84 (1.24, 2.74)
2.50 (1.00, 6.70)
7.78 (1.00, 60.40)
3.06 (1.29, 7.24)
HR (95% CI)
17.71
33.71
17.16
8.65
22.77
100.00
82.30
17.70
100.00
Weight
%
Cardiovascular mortality
1.5 1 5 10 30 50 80
Figure 3. Forest plot depicts the meta-association between various forms of thyroid function test result derangements and risk for
cardiovascular mortality, using the Dersimonian and Laird random-effects model. All hazard ratios (HRs) are based on the most fully
adjusted reported model. Abbreviations: CI, confidence interval; (f) T
3
, (free) triiodothyronine; (f) T
4
, (free) thyroxine; HD, hemodialysis;
PD, peritoneal dialysis; TSH, thyrotropin.
928 Am J Kidney Dis. 2016;68(6):923-932
Xu et al
exposure, it has been postulated that commonly used
free T
4
assays may be inaccurate in ESRD given the
described alterations in T
3
and T
4
levels and the
metabolism of thyrotropin.
We found a consistent association between low T
3
level and increased risk for death in long-term dialysis
patients. Being based on observational studies, our
data cannot prove causality in the associations.
However, experimental studies show that low T
3
level
impairs cardiac tissue oxygen consumption, increases
vascular resistance, and decreases cardiac output.
43,44
Observational studies in patients with CKD and those
who progressed to ESRD suggest that low T
3
levels
are linked to adverse intermediate surrogates, such as
atherosclerosis,
33
vascular calcification,
20,41
arterial
stiffness,
33,34
impaired flow-mediated vasodila-
tion,
35,40
intravascular volume deficits and abnormal
ventricular conduction,
36,37
and impaired cardiac
function,
38
which could also explain the associations
reported here. We found overall moderate heteroge-
neity in our estimates. Heterogeneity may be attrib-
uted in part to the use of different T
3
cutoffs and
different laboratory methods and measurements of T
3
(free vs total). Other potential explanations related to
differences in participant characteristics (eg, study
population, varying follow-up time, sample size, and
adjustment for confounding factors). However, strat-
ified analyses yielded consistent estimates. Compared
with studies that used cohort-based cutoffs (eg, ter-
tiles), those using assay reference range appeared to
have lower mortality risk; this may not be surprising
if CKD (with or without ESRD) per se renders low T
3
values and thus cutoffs derived from healthy in-
dividuals may not correctly identify patients at risk.
The mortality risk estimate associated with low T
3
levels was higher in patients with shorter follow-up,
with smaller sample size, and undergoing PD treat-
ment. This collectively may indicate a risk of publi-
cation bias and the scarcity of literature available. We
also report consistency in the associations between
low thyrotropin and low T
4
levels, although fewer
studies examined these exposures. In our inclusion
criteria, we considered only baseline thyroid hormone
(T
3
and T
4
) assessments. However, 2 additional re-
ports address longitudinal thyroid hormonal states and
found that persistently low T
3
and T
4
levels were
associated with 2- to 4-fold higher risk for death in
patients with ESRD,
9,10
perhaps offering further
support to our hypothesis.
Our observations are in line with the evidence from
general population studies suggesting that low thyroid
hormone levels, even in subclinical forms, may
negatively affect cardiovascular health and increase
the risk for death.
45,46
This evidence includes various,
but not all,
47
meta-analyses reporting an overall
increased mortality risk in individuals without CKD
with subclinical thyroid functional disorders, partic-
ularly among those with younger age,
48
heart fail-
ure,
49
high comorbid condition burden,
50
and higher
thyrotropin levels.
51
Although the need to treat these
subclinical disorders is recommended in some
guidelines and consensus papers as a strategy to
reduce cardiovascular risk,
52,53
there is a paucity of
interventional data in patients with CKD and those
who progressed to ESRD. An early interventional
study showed that intake of physiologic doses of T
3
(50 mg/d) decreased thyrotropin levels and resulted in
a borderline negative nitrogen balance (increased
Table 2. Metaregression Analyses on Association Between Low Free T
3
Level and Mortality Risk
Comparison of Low (f) T
3
No. of
Studies
Empirical Bayes Metaregression
Pooled HR (95% CI) PI
2
,%
a
All-cause mortality
Low (f) T
3
, defined as ,cohort-specific cutoffs
vs ,reference range
9 1.75 (0.97-3.14) 0.06 17.8
Total T
3
vs (f) T
3
measurements 9 1.16 (0.46-2.92) 0.7 54.1
In PD vs HD patients 8 2.57 (0.63-10.60) 0.2 49.9
Studies with follow-up .36 vs 12-36 mo 9 0.83 (0.35-1.95) 0.6 58.0
Sample size $500 vs ,500 patients 9 0.53 (0.33-0.84) 0.02 0
Adjusted for vs not adjusted for malnutrition,
inflammation, and comorbid conditions
9 1.15 (0.37-3.51) 0.8 56.9
Cardiovascular mortality
Low (f) T
3
, defined as ,cohort-specific cutoffs
vs ,reference range
5 2.04 (0.66-6.27) 0.1 0
Studies with follow up .36 vs 12-36 mo 5 0.55 (0.14-2.07) 0.3 13.0
Adjusted for vs not adjusted for malnutrition,
inflammation, and comorbid conditions
5 1.69 (0.42-6.86) 0.3 19.6
Abbreviations: CI, confidence interval; (f) T
3
, free triiodothyronine; HD, hemodialysis; HR, hazard ratio; PD, peritoneal dialysis.
a
I
2
represents the percentage of variation attributable to heterogeneity, typically categorized as low (0%-50%), moderate
(51%-75%), or high (.75%).
Am J Kidney Dis. 2016;68(6):923-932 929
Thyroid Function and Mortality in Dialysis Patients
protein catabolism) in patients with ESRD.
54
This
may be the natural consequence of restoring thyroid
function and in our opinion may be easily counter-
acted by increasing protein intake. Outside
nephrology, short-term T
3
replacement therapy
greatly improved the neuroendocrine profile and
ventricular performance in patients with heart failure
with low T
3
syndrome.
55
Before trials are conducted,
other indirect approaches that may serve as proofs of
concept include correcting acidosis,
56,57
oxidative
stress,
58
or selenium deficiency.
59
In a placebo-
controlled study of 30 euthyroid patients undergoing
HD, exogenous T
4
administration over 3 months
reduced lipoprotein(a) and total and low-density li-
poprotein cholesterol levels, without evidence of
thyrotoxicosis.
60
However, in a recent large obser-
vational study of patients with ESRD, hypothyroid
patients receiving exogenous thyroid hormones were
at the same risk for death compared with those
without medication.
16
In summary, all-cause and cardiovascular mortality
was found to be consistently higher for long-term
dialysis patients with thyroid function test result de-
rangements. These derangements may represent an
under-recognized risk factor, with a biologically
plausible link to the poor clinical outcomes of this
population. The observed associations of this meta-
analysis raise the question of whether it would be
cost-effective to screen for thyroid function among
patients with ESRD and whether patients with sub-
clinical signs of hypothyroidism would benefit from
corrections of thyroid hormone deficiencies to the
normal range.
ACKNOWLEDGEMENTS
Support: This work was supported by the Stockholm County
Council and the Heart and Lung Foundation. Dr Xu is partially
supported by Karolinska Institutet faculty for funding of post-
graduate (KID). Baxter Novum is the result of a grant from Baxter
Healthcare to the Karolinska Institutet. The funders of this study
had no any role in study design; collection, analysis, and inter-
pretation of data; writing the report; or the decision to submit the
report for publication.
Financial Disclosure: Dr Lindholm is employed by Baxter
Healthcare. The other authors declare that they have no other
relevant financial interests.
Contributions: Research idea and study design: HX, JJC; data
acquisition: HX, NB; data analysis/interpretation: HX, NB, BL,
CZ, JJC; statistical analysis: HX, NB; supervision or mentorship:
BL, JJC. Each author contributed important intellectual content
during manuscript drafting or revision and accepts accountability
for the overall work by ensuring that questions pertaining to the
accuracy or integrity of any portion of the work are appropriately
investigated and resolved. HX and JJC take responsibility that this
study has been reported honestly, accurately, and transparently;
that no important aspects of the study have been omitted; and that
any discrepancies from the study as planned have been explained.
Peer Review: Evaluated by 2 external peer reviewers, a Statis-
tical Editor, a Co-Editor, and the Editor-in-Chief.
SUPPLEMENTARY MATERIAL
Table S1: Description and characteristics of excluded observa-
tional studies reporting association between thyroid function test
derangements and CV surrogates.
Table S2: Description of covariates used in fully adjusted
mortality HRs selected for meta-analysis.
Table S3: Sensitivity meta-analysis on association between low
free T
3
and mortality risk: omission of single studies.
Figure S1: Quality assessment of included studies.
Figure S2: Individual quality assessment of included studies.
Item S1: Electronic search strategy.
Item S2: Study protocol.
Note: The supplementary material accompanying this article
(http://dx.doi.org/10.1053/j.ajkd.2016.06.023) is available at
www.ajkd.org
REFERENCES
1. Meuwese CL, Dekkers OM, Stenvinkel P, Dekker FW,
Carrero JJ. Nonthyroidal illness and the cardiorenal syndrome. Nat
Rev Nephrol. 2013;9(10):599-609.
2. Abdel-Rahman EM, Mansour W, Holley JL. Thyroid hor-
mone abnormalities and frailty in elderly patients with chronic
kidney disease: a hypothesis. Semin Dial. 2010;23(3):317-323.
3. Rhee CM. Low-T3 syndrome in peritoneal dialysis: meta-
bolic adaptation, marker of illness, or mortality mediator? Clin J
Am Soc Nephrol. 2015;10(6):917-919.
4. Chonchol M, Lippi G, Salvagno G, Zoppini G, Muggeo M,
Targher G. Prevalence of subclinical hypothyroidism in patients
with chronic kidney disease. Clin J Am Soc Nephrol. 2008;3(5):
1296-1300.
5. Lo JC, Chertow GM, Go AS, Hsu CY. Increased prevalence
of subclinical and clinical hypothyroidism in persons with chronic
kidney disease. Kidney Int. 2005;67(3):1047-1052.
6. Rhee CM, Kalantar-Zadeh K, Streja E, et al. The relationship
between thyroid function and estimated glomerular filtration rate
in patients with chronic kidney disease. Nephrol Dial Transplant.
2015;30(2):282-287.
7. Song SH, Kwak IS, Lee DW, Kang YH, Seong EY, Park JS.
The prevalence of low triiodothyronine according to the stage of
chronic kidney disease in subjects with a normal thyroid-
stimulating hormone. Nephrol Dial Transplant. 2009;24(5):
1534-1538.
8. Rhee CM, Brent GA, Kovesdy CP, et al. Thyroid functional
disease: an under-recognized cardiovascular risk factor in kidney
disease patients. Nephrol Dial Transplant. 2015;30(5):724-737.
9. Meuwese CL, Dekker FW, Lindholm B, et al. Baseline
levels and trimestral variation of triiodothyronine and thyroxine
and their association with mortality in maintenance hemodialysis
patients. Clin J Am Soc Nephrol. 2012;7(1):131-138.
10. Jung HY, Cho JH, Jang HM, et al. Free thyroxine level as
an independent predictor of infection-related mortality in patients
on peritoneal dialysis: a prospective multicenter cohort study.
PLoS One. 2014;9(12):e112760.
11. Zoccali C, Mallamaci F. Thyroid function and clinical out-
comes in kidney failure. Clin J Am Soc Nephrol. 2012;7(1):12-14.
12. Reinhardt W, Dolff S, Benson S, et al. Chronic kidney
disease distinctly affects relationship between selenoprotein P
status and serum thyroid hormone parameters. Thyroid.
2015;25(10):1091-1096.
13. Enia G, Panuccio V, Cutrupi S, et al. Subclinical hypo-
thyroidism is linked to micro-inflammation and predicts death in
continuous ambulatory peritoneal dialysis. Nephrol Dial Trans-
plant. 2007;22(2):538-544.
930 Am J Kidney Dis. 2016;68(6):923-932
Xu et al
14. Carrero JJ, Qureshi AR, Axelsson J, et al. Clinical and
biochemical implications of low thyroid hormone levels (total and
free forms) in euthyroid patients with chronic kidney disease.
J Intern Med. 2007;262(6):690-701.
15. Yang JW, Han ST, Song SH, et al. Serum T3 level can
predict cardiovascular events and all-cause mortality rates in CKD
patients with proteinuria. Ren Fail. 2012;34(3):364-372.
16. Rhee CM, Alexander EK, Bhan I, Brunelli SM. Hypothy-
roidism and mortality among dialysis patients. Clin J Am Soc
Nephrol. 2013;8(4):593-601.
17. Rhee CM, Kim S, Gillen DL, et al. Association of thyroid
functional disease with mortality in a national cohort of incident
hemodialysis patients. J Clin Endocrinol Metab. 2015;100(4):
1386-1395.
18. Zoccali C, Mallamaci F, Tripepi G, Cutrupi S, Pizzini P.
Low triiodothyronine and survival in end-stage renal disease.
Kidney Int. 2006;70(3):523-528.
19. Ozen KP, Asci G, Gungor O, et al. Nutritional state alters
the association between free triiodothyronine levels and mortality
in hemodialysis patients. Am J Nephrol. 2011;33(4):305-312.
20. Meuwese CL, Carrero JJ, Cabezas-Rodriguez I, et al.
Nonthyroidal illness: a risk factor for coronary calcification and
arterial stiffness in patients undergoing peritoneal dialysis?
J Intern Med. 2013;274(6):584-593.
21. Koo HM, Kim CH, Doh FM, et al. The impact of low
triiodothyronine levels on mortality is mediated by malnutrition
and cardiac dysfunction in incident hemodialysis patients. Eur J
Endocrinol. 2013;169(4):409-419.
22. Drechsler C, Schneider A, Gutjahr-Lengsfeld L, et al.
Thyroid function, cardiovascular events, and mortality in diabetic
hemodialysis patients. Am J Kidney Dis. 2014;63(6):988-996.
23. Fragidis S, Sombolos K, Thodis E, et al. Low T3 syndrome
and long-term mortality in chronic hemodialysis patients. World J
Nephrol. 2015;4(3):415-422.
24. National Kidney Foundation: K/DOQI clinical practice
guidelines for chronic kidney disease: evaluation, classification,
and stratification. Am J Kidney Dis. 2002;39(2)(suppl 1):S1-S266.
25. Wells G, Shea B, O’Connell D, et al. The Newcastle-
Ottawa Scale (NOS) for assessing the quality of nonrandomised
studies in meta-analyses. http://www.ohri.ca/programs/clinical_
epidemiology/oxford.asp. Accessed December 15, 2015.
26. Higgins JP, Thompson SG, Deeks JJ, Altman DG.
Measuring inconsistency in meta-analyses. BMJ. 2003;327(7414):
557-560.
27. Egger M, Davey Smith G, Schneider M, Minder C. Bias in
meta-analysis detected by a simple, graphical test. BMJ.
1997;315(7109):629-634.
28. Lin YC, Chen TW, Yang WC, Lin CC. Abnormal thyroid
function predicts mortality in patients receiving long-term perito-
neal dialysis: a case-controlled longitudinal study. J Chin Med
Assoc. 2012;75(2):54-59.
29. Fernandez-Reyes MJ, Diez JJ, Collado A, et al. Are low
concentrations of serum triiodothyronine a good marker for long-
term mortality in hemodialysis patients? Clin Nephrol. 2010;73(3):
238-240.
30. Horacek J, Dusilova Sulkova S, Kubisova M, et al. Thyroid
hormone abnormalities in hemodialyzed patients: low triiodothy-
ronine as well as high reverse triiodothyronine are associated with
increased mortality. Physiol Res. 2012;61(5):495-501.
31. Perez SR, Diez JJ, Bajo MA, et al. Thyrotropin and free
thyroxine concentrations do not affect cardiovascular disease and
mortality in euthyroid peritoneal dialysis patients. Perit Dial Int.
2013;33(6):697-700.
32. Chang TI, Nam JY, Shin SK, Kang EW. Low triiodothy-
ronine syndrome and long-term cardiovascular outcome in inci-
dent peritoneal dialysis patients. Clin J Am Soc Nephrol.
2015;10(6):975-982.
33. Tatar E, Kircelli F, Asci G, et al. Associations of triiodo-
thyronine levels with carotid atherosclerosis and arterial stiffness
in hemodialysis patients. Clin J Am Soc Nephrol. 2011;6(9):2240-
2246.
34. Tatar E, Demirci MS, Kircelli F, et al. The association
between thyroid hormones and arterial stiffness in peritoneal
dialysis patients. Int Urol Nephrol. 2012;44(2):601-606.
35. Yilmaz MI, Sonmez A, Karaman M, et al. Low triiodo-
thyronine alters flow-mediated vasodilatation in advanced nondi-
abetic kidney disease. Am J Nephrol. 2011;33(1):25-32.
36. Zoccali C, Benedetto F, Mallamaci F, et al. Low triiodo-
thyronine and cardiomyopathy in patients with end-stage renal
disease. J Hypertens. 2006;24(10):2039-2046.
37. Xu G, Yan Y, Liu Y. The cardiovascular disease risks of
nonthyroidal illness syndrome and inflammatory responses on
patients with chronic kidney disease: from the association to
clinical prognosis. Cardiovasc Ther. 2014;32(6):257-263.
38. Kang EW, Nam JY, Yoo TH, et al. Clinical implications of
subclinical hypothyroidism in continuous ambulatory peritoneal
dialysis patients. Am J Nephrol. 2008;28(6):908-913.
39. Jaroszynski AJ, Glowniak A, Chrapko B, et al. Low-T3
syndrome and signal-averaged ECG in haemodialysed patients.
Physiol Res. 2005;54(5):521-526.
40. Kocyigit I, Gungor O, Unal A, et al. A low serum free
triiodothyronine level is associated with epicardial adipose tissue
in peritoneal dialysis patients. J Atheroscler Thromb. 2014;21(10):
1066-1074.
41. Meuwese CL, Olauson H, Qureshi AR, et al. Associations
between thyroid hormones, calcification inhibitor levels and
vascular calcification in end-stage renal disease. PLoS One.
2015;10(7):e0132353.
42. Lau J, Ioannidis JP, Terrin N, Schmid CH, Olkin I. The
case of the misleading funnel plot. BMJ. 2006;333(7568):597-600.
43. Klein I, Ojamaa K. Thyroid hormone and the cardiovas-
cular system. N Engl J Med. 2001;344(7):501-509.
44. Iervasi G, Nicolini G. Thyroid hormone and cardiovascular
system: from basic concepts to clinical application. Intern Emerg
Med. 2013;8(suppl 1):S71-S74.
45. Hennessey JV, Espaillat R. Reversible morbidity markers
in subclinical hypothyroidism. Postgrad Med. 2015;127(1):78-91.
46. Thvilum M, Brandt F, Brix TH, Hegedüs L. A review of
the evidence for and against increased mortality in hypothyroid-
ism. Nat Rev Endocrinol. 2012;8(7):417-424.
47. Åsvold BO, Vatten LJ, Bjøro T, et al. Thyroid function
within the normal range and risk of coronary heart disease: an
individual participant data analysis of 14 cohorts. JAMA Intern
Med. 2015;175(6):1037-1047.
48. Razvi S, Shakoor A, Vanderpump M, Weaver JU,
Pearce SH. The influence of age on the relationship between
subclinical hypothyroidism and ischemic heart disease: a meta-
analysis. J Clin Endocrinol Metab. 2008;93(8):2998-3007.
49. Ning N, Gao D, Triggiani V, et al. Prognostic role of hy-
pothyroidism in heart failure: a meta-analysis. Medicine (Balti-
more). 2015;94(30):e1159.
50. Haentjens P, Van Meerhaeghe A, Poppe K, Velkeniers B.
Subclinical thyroid dysfunction and mortality: an estimate of
relative and absolute excess all-cause mortality based on time-to-
event data from cohort studies. Eur J Endocrinol. 2008;159(3):
329-341.
Am J Kidney Dis. 2016;68(6):923-932 931
Thyroid Function and Mortality in Dialysis Patients
51. Rodondi N, den Elzen WP, Bauer DC, et al. Subclinical
hypothyroidism and the risk of coronary heart disease and mor-
tality. JAMA. 2010;304(12):1365-1374.
52. Peleg RK, Efrati S, Benbassat C, Fygenzo M, Golik A. The
effect of levothyroxine on arterial stiffness and lipid profile in
patients with subclinical hypothyroidism. Thyroid. 2008;18(8):
825-830.
53. Okosieme O, Gilbert J, Abraham P, et al. Management of
primary hypothyroidism: statement by the British Thyroid Asso-
ciation Executive Committee. Clin Endocrinol (Oxf). 2016;84(6):
799-808.
54. Lim VS, Tsalikian E, Flanigan MJ. Augmentation of pro-
tein degradation by L-triiodothyronine in uremia. Metabolism.
1989;38(12):1210-1215.
55. Pingitore A, Galli E, Barison A, et al. Acute effects of
triiodothyronine (T3) replacement therapy in patients with chronic
heart failure and low-T3 syndrome: a randomized, placebo-
controlled study. J Clin Endocrinol Metab. 2008;93(4):1351-1358.
56. Wiederkehr MR, Kalogiros J, Krapf R. Correction of
metabolic acidosis improves thyroid and growth hormone axes in
haemodialysis patients. Nephrol Dial Transplant. 2004;19(5):
1190-1197.
57. Disthabanchong S, Treeruttanawanich A. Oral sodium bi-
carbonate improves thyroid function in predialysis chronic kidney
disease. Am J Nephrol. 2010;32(6):549-556.
58. Vidart J, Wajner SM, Leite RS, et al. N-Acetylcysteine
administration prevents nonthyroidal illness syndrome in patients
with acute myocardial infarction: a randomized clinical trial. J Clin
Endocrinol Metab. 2014;99(12):4537-4545.
59. Gärtner R, Gasnier BC, Dietrich JW, Krebs B,
Angstwurm MW. Selenium supplementation in patients with
autoimmune thyroiditis decreases thyroid peroxidase antibodies
concentrations. J Clin Endocrinol Metab. 2002;87(4):1687-1691.
60. Bommer C, Werle E, Walter-Sack I, et al. D-Thyroxine
reduces lipoprotein(a) serum concentration in dialysis patients.
J Am Soc Nephrol. 1998;9(1):90-96.
932 Am J Kidney Dis. 2016;68(6):923-932
Xu et al