deFilippi C, Wasserman S, Rosanio S, Tiblier E, Sperger H, Tocchi M, Christenson R, Uretsky B, Smiley M, Gold J, Muniz H, Badalamenti J, Herzog C, Henrich W. Cardiac Troponin T and C-Reactive Protein for Predicting Prognosis, Coronary Atherosclerosis, and Cardiomyopathy in Patients Undergoing Long-term Hemodialysis. JAMA. 2003;290(3):353-359. doi:10.1001/jama.290.3.353
Author Affiliation: Departments of Medicine (Drs deFilippi, Wasserman, and Henrich) and Pathology (Dr Christenson), University of Maryland School of Medicine, Baltimore; Department of Cardiology, San Raffaele University Hospital, Milan, Italy (Drs Rosanio and Tocchi); Department of Medicine, University of Texas Medical Branch, Galveston (Drs Tiblier, Uretsky, Smiley, and Badalamenti, and Ms Sperger); Renal Care Group, Angleton, Tex (Dr Gold); Renal Specialists, Houston, Tex (Dr Muniz); and Department of Medicine, Hennepin County Medical Center, University of Minnesota School of Medicine, Minneapolis (Dr Herzog).
Context Cardiac troponin T (cTnT) and C-reactive protein (CRP) are prognostic
markers in acute coronary syndromes. However, for patients with end-stage
renal disease (ESRD) the ability of combinations of these markers to predict
outcomes, and their association with cardiac pathology, are unclear.
Objective To investigate the association between levels of cTnT and CRP and long-term
risk of cardiac pathology and death in patients with ESRD.
Design, Setting, and Participants A prospective cohort study initiated February through June 1998 and
enrolling 224 patients with ESRD from 5 hemodialysis centers in the Houston-Galveston
region of Texas. Levels of cTnT and CRP were analyzed at study entry in patients
without ischemic symptoms.
Main Outcome Measures All-cause mortality during a mean follow-up of 827 (range, 29-1327)
days. Secondary outcomes in predefined substudies were coronary artery disease
(CAD), decreased (≤40%) left ventricular ejection fraction (LVEF), and
presence of left ventricular hypertrophy (LVH).
Results One hundred seventeen (52%) patients died during follow-up. For levels
of cTnT and CRP, progressively higher levels predicted increased risk of death
compared with the lowest quartile (for cTnT quartile 2: unadjusted hazard
ratio [HR], 2.2; 95% confidence interval [CI], 1.2-4.1; quartile 3: HR, 2.7;
95% CI, 1.5-4.9; quartile 4: HR, 3.0; 95% CI, 1.6-5.3. For CRP quartile 2:
HR, 0.9; 95% CI, 0.5-1.6; quartile 3: HR, 1.8; 95% CI, 1.1-3.1; quartile 4:
HR, 1.8; 95% CI, 1.1-3.2). Both cTnT and CRP remained independent predictors
of death after adjusting for a number of potential confounders. The combination
of cTnT and CRP results provided prognostic information when patients were
divided into groups at or above and below the biomarker medians (high cTnT/high
CRP levels vs low cTnT/low CRP levels for risk of death: HR, 2.5; 95% CI,
1.5-4.0). Elevated levels of cTnT, but not CRP, were strongly associated with
diffuse CAD (n = 67; 0%, 25%, 50%, and 62% prevalence of multivessel CAD across
progressive cTnT quartiles, P<.001). An LVEF of
40% or less was identified in 4 (9%), 3 (8%), 10 (27%), and 7 (19%) of patients
across cTnT quartiles (P = .07). No trend for cTnT
levels was found among patients with LVH (P = .45);
similarly, no trend for CRP was found among patients with LVH (P = .65) or an LVEF of 40% or less (P = .75).
Conclusions Among stable patients with ESRD, increasing levels of cTnT and CRP are
associated with increased risk of death. Furthermore, higher levels of cTnT
may identify patients with severe angiographic coronary disease.
Mortality among patients with end-stage renal disease (ESRD) remains
as high as 23% per year despite advances in dialysis, with cardiac causes
accounting for 40% to 45% of all deaths.1 Because
this high burden of disease results from coronary artery disease (CAD) or
cardiomyopathy, cardiac risk stratification is a key issue in the clinical
management of these patients.
Inflammation, critical to the pathogenesis of atherosclerosis,2 can be demonstrated by elevation of serum C-reactive
protein (CRP) levels in more than 70% of patients receiving hemodialysis.3- 7 Furthermore,
levels of cardiac troponin T (cTnT), a serum marker of myocardial infarction,8 can be elevated in 30% to 75% of these patients.7,9- 14 When
elevated, these biomarkers may be associated with increased all-cause mortality.3,4,6,7,10- 14 However,
little is known regarding the potential complementary roles of measuring levels
of cTnT and CRP for predicting all-cause mortality, and the relationship between
these biomarker levels and underlying cardiac pathology in stable patients
To address these issues, we undertook a prospective multicenter study
of patients receiving long-term hemodialysis without symptoms of myocardial
ischemia in whom follow-up outcomes, CAD, and left ventricular function and
mass were assessed. Our goals were to compare the predictive values of cTnT
and CRP levels for long-term risk of all-cause mortality and to investigate
if these biomarkers are indicators of the severity of either coronary atherosclerosis
demonstrated by angiography or of cardiomyopathy as manifested by left ventricular
systolic dysfunction or hypertrophy determined by echocardiography.
The protocol was approved by the institutional review board at the University
of Texas Medical Branch at Galveston, and written informed consent was obtained
from each patient. The study was conducted in the Houston-Galveston region
of Texas, and from February through June 1998 224 of 334 screened dialysis
patients were prospectively enrolled at 5 centers, including 3 community-based
dialysis sites located in suburban and semirural areas, and 2 university-based
sites located in urban settings (range of enrollment per center was 29-70
patients). To be enrolled, patients had to have been undergoing hemodialysis
for more than 30 days, be 18 years of age or older, and be free from any acute
coronary event for more than 4 weeks. Of the 110 screened patients not included,
89 declined to participate and 21 consented but did not have blood drawn secondary
to transfer (n = 7), transplant (n = 1), withdrawal of consent (n = 7), poor
compliance with dialysis (n = 1), or death (n = 5). The initial clinical evaluation
included a history, physical examination, and a review of medical records
for prior cardiac events. The adequacy of dialysis was estimated at study
entry by a dialysis dose (Kt/V) greater than 1.2 calculated by the single-pool
method.15 Levels of serum biomarkers were obtained
immediately prior to the midweek dialysis at study entry.
Outcomes through October 2001 were determined by linking to data in
the United States Renal Data System (USRDS) for deaths and kidney transplantation.
The primary end point was all-cause mortality. Cause of death was determined
from the USRDS, with cardiac etiologies defined as deaths related to myocardial
infarction, congestive heart failure, cardiomyopathy, arrhythmia, and cardiac
arrest.1 Patients were censored from further
follow-up if they underwent kidney transplantation (n = 25; mean [SD] days
enrolled in the study, 604 ).
Blood samples were collected without an anticoagulant and allowed to
clot for at least 30 minutes, then centrifuged at 1000g for 12 minutes. The resulting serum was aliquoted, frozen, and maintained
at –70°C. Samples were thawed within 6 months for measurements of
cTnT levels using a third-generation assay (Troponin T STAT immunoassay, ElecSys
2010 system, Roche Diagnostics, Indianapolis, Ind) and then refrozen at –70°C
within 24 hours. Rethawed samples were used to measure high-sensitivity CRP
(BN II analyzer, Dade Behring, Glasgow, Del). Routine clinical chemistry variables
(including albumin and creatine kinase-MB) were analyzed by standardized methods.
An angiographic substudy was undertaken to determine the prevalence
and severity of CAD according to biomarker concentrations. Volunteers were
recruited for angiography in the absence of coronary angiography within 2
years of study enrollment or vascular disease precluding access to the femoral
arteries. The study paid the technical fees; professional fees were deferred.
All patients without contraindications were provided an opportunity to volunteer
during the first 10 months of the study.
Angiography was performed on off-dialysis days using a low-osmolality
nonionic contrast medium (iodixanol) and were interpreted independently by
2 experienced angiographers who were blinded to the levels of biomarkers.
Differences were resolved by consensus. High interobserver and intraobserver
correlation for these readers has been shown.16 Coronary
artery disease was defined as a 50% or greater lumen narrowing of a major
epicardial artery or its branches. A left main stenosis of 50% or greater
was regarded as equivalent to 2-vessel disease. In addition, we applied a
CAD prognostic index that accounts for both the severity and location of coronary
stenosis. For the analysis we identified patients with a score greater than
48; this score has been previously associated with a poor prognosis in patients
with ischemic cardiomyopathy.17 For this index
analysis we assumed a stenosis of 70% or greater was equivalent to 75% or
Echocardiograms were prospectively obtained at 1 of 3 points during
the first 18 months of the study: just prior to dialysis, at the time of angiography
for patients volunteering for the procedure, or as part of a clinically indicated
study. Echocardiograms were interpreted without knowledge of biomarker concentrations.
Left ventricular volumes were quantified using the Simpson biplane formula.
Left ventricular mass was calculated using the area-length method18 and indexed for body surface area. Left ventricular
hypertrophy (LVH) was defined as left ventricular mass index greater than
Continuous clinical variables are reported as medians and interquartile
ranges or as means and SDs. Kruskal-Wallis tests and Fisher exact tests were
used to assess between-group differences for continuous and categorical data,
respectively. For the relationships of biomarker quartiles with multivessel
CAD, CAD prognostic index, and echocardiographic parameters, the existence
of trends was explored using the Cochran-Armitage test for categorical variables
and Spearman rank correlation analysis for continuous variables. Logistic
regression was used to test if levels of cTnT were an independent predictor
of multivessel CAD. The stability of this estimate was tested using the method
of Hosmer and Lemeshow.20 Prevalence ratios
were calculated from the odds ratios using the method of Zhang and Yu.21
Kaplan-Meier estimates were generated to describe outcome time-course
for quartiles of biomarker levels and cTnT/CRP combinations, with each biomarker
dichotomized at its median. The log-rank test was used to compare differences.
Cox regression models for all-cause mortality were used to calculate
the hazard ratio (HR) of progressively higher quartiles of biomarker levels,
cTnT/CRP combinations, and to determine the effects of baseline cTnT and CRP
levels after adjustment for clinical variables. Various transformations (ie,
none, inverse, square root, square, quadratic, quartiles) for each biomarker
were explored for best goodness-of-fit in individual Cox models, adjusting
each for age. The transformations selected in this way were: for cTnT, transformation
to quartiles; for CRP, no transformation. In addition to biomarker levels,
the following clinical variables were considered in Cox regression models:
age; white race; sex; length of time receiving dialysis; history of smoking,
coronary disease, or diabetes; levels of albumin; Kt/V; and body surface area.
Interactions between levels of cTnT and CRP and each potential confounder
were explored in Cox models that included 1 biomarker and the interaction
of the confounder being tested with that biomarker. Stepwise Cox models were
created that forced cTnT and CRP into the model and selected among the confounders
to create a final model that was identical for P-to-enter
values between .05 and .25 and for P-to-remove values
of .05 or .10. The assumption of proportionality was tested using the method
described by SAS.22 The test for trend in the
HR involved the inclusion of a time-dependent interaction term. The stability
of point estimates of HRs was verified using 1000 bootstrap resampling analyses.
The final Cox model was run for each sample, and the lower 2.5% confidence
limits on the HRs for all independent variables were calculated.
Two-sided P values less than .05 were considered
statistically significant. Analyses were performed using SAS version 8.02
(SAS Institute, Cary, NC).
Baseline clinical and biochemical characteristics of the study population
are shown in Table 1. No patients
were lost to follow-up. After a mean follow-up period of 827days (SD, 430
days; range, 29-1327 days), 117 patients (52%) died (62 [53%] cardiac, 33
[28%] noncardiac, and 22 [19%] of uncertain etiology). Etiologies of noncardiac
deaths included infection/sepsis (43%), cerebrovascular accident (21%), malignancy
(14%), and other causes (22%).
The distribution of all-cause mortality over time by quartiles of cTnT
and CRP levels is shown in Kaplan-Meier plots in Figure 1. Separation of the biomarker quartiles curves occurred
initially within the first months of follow-up and typically continued to
widen during the next 3.5 years for both biomarkers. Unadjusted Cox regression
analyses are summarized in Table 2.
For levels of cTnT and CRP, progressively higher levels predicted increased
risk of death compared with the lowest quartile.
To test if combinations of the 2 biomarker levels were complementary
for predicting death, patients were divided into 4 groups on the basis of
whether their biomarker levels fell at or above (high) or below (low) the
medians for cTnT and CRP. The results of the analyses are shown in Table 2 and Figure 2. There was a 2.5-fold (95% CI, 1.5-4.0) increased risk
of death over the study period in patients having high levels of both cTnT
and CRP compared with those having low levels of both. For patients with high
levels of only 1 marker, the risk of death was between that for those with
low or high levels of both biomarkers, although it was not significantly different
from that for patients with low levels of both markers. Among patients with
low levels of CRP, low and high levels of cTnT did not differentiate a risk
of a cardiac cause of death (63% vs 42% respectively; age-adjusted HR, 0.53;
95% CI, 0.23-1.25; P = .22). Among patients with
high levels of CRP the presence of low or high levels of cTnT trended toward
discriminating those at risk for cardiac death (33% vs 65% respectively; age-adjusted
HR, 2.08; 95% confidence interval [CI], 0.90-4.84; P =
.03). Nearly half (49%) of all cardiac deaths occurred in patients with high
levels of both cTnT and CRP.
Risk of death determined by each biomarker level was stratified separately
by presence or absence of a history of CAD and by white race. No interaction
was detected between these 2 factors and the biomarkers for prediction of
A multivariable model was developed to identify if levels of cTnT and
CRP remained independent predictors of all-cause mortality (Table 3). Levels of CRP considered as a continuous variable and
high levels of cTnT (fourth quartile) both remained independent predictors
of all-cause mortality along with age, white race, history of diabetes, and
body surface area. No important interaction (P<.01)
was detected between any of the clinical risk factors and either biomarker.
Sixty-seven patients volunteered for coronary angiography. No significant
clinical differences were found between patients who underwent angiography
and the remaining study patients, other than a lower age and a modestly lower
Kt/V in the former group (mean [SD], 57  vs 61  years, P = .03; 1.68 [0.75] vs 1.72 [0.57], P = .04,
respectively). There were no differences in levels of cTnT or CRP between
volunteers for angiography and the remaining patients. Coronary artery disease
was found in 28 patients (42%). Multivessel disease was present in 21 patients
(31%) (2-vessel disease in 10, 3-vessel disease in 11).
Multivessel disease and a CAD index greater than 48 (indicative of high-risk
multivessel CAD) were more prevalent across progressively higher quartiles
of cTnT (as defined by the entire study population), but not quartiles of
CRP level (0%, 25%, 50%, and 62% prevalence of multivessel CAD across progressive
cTnT quartiles) (Table 4). A high
level of cTnT (>median) vs a low level of cTnT remained an independent predictor
of multivessel disease (prevalence ratio [PR], 3.7; 95% CI, 1.2 –13.0)
after adjustment for age (PR, 1.05 per year; 95% CI, 1.01-1.10) and history
of clinical CAD (PR, 7.1; 95% CI, 2.0-26.3).
A total of 173 patients (77%) underwent echocardiography, with 155 studies
suitable for quantitative analysis. There were 51 echocardiograms not obtained
because of death (31 patients), transfer (15 patients), or scheduling problems
(5 patients). The prevalence of LVH and systolic dysfunction across cTnT or
CRP quartiles was assessed (Table 4).
No trend for LVH by level of cTnT and CRP (P = .45)
was found, but depressed systolic function (left ventricular ejection fraction
[LVEF] ≤40%) was approximately twice as frequent in the highest cTnT quartiles
(4 [9%], 3 [8%], 10 [27%], and 7 [19%] of patients across cTnT quartiles, P = .07). Patients with high cTnT levels had a higher prevalence
of depressed left ventricular function vs patients with low levels (23% vs
9%, P = .03). Interestingly, only 1 of 35 patients without CAD (determined
by angiography) who had had an echocardiogram performed had an LVEF of 40%
or less. In these patients without CAD, the prevalence of LVH was also not
different based on high or low cTnT values (65% vs 73%; P = .99). No trend for CRP levels was found among patients with LVH
(P = .65) or an LVEF of 40% or less (P = .75).
This study provides insight into the association between elevated levels
of cTnT with coronary athersclerosis and the long-term prognostic importance
for biomarkers of myocardial damage and inflammation in patients receiving
long-term hemodialysis. An angiographic substudy found that an elevated level
of cTnT is a marker of extensive CAD, as manifested by a 3.7-fold higher PR
of multivessel CAD in patients with high levels of cTnT vs patients with low
levels. This finding provides pathophysiological support to our main finding
that elevated levels of cTnT were associated with an increased long-term risk
of all-cause mortality.
Our results also indicate that measurement of inflammation in a population
of patients receiving long-term hemodialysis, as determined by level of CRP,
was an independent predictor of death. Moreover, the presence of a high level
of CRP along with a high level of cTnT conveyed the highest risk of death.
Proposed mechanisms of cTnT level elevation in patients receiving hemodialysis
have included silent ischemic injury and an apoptotic process.23 We
found that even small elevations of cTnT concentration, at levels lower than
those traditionally used for the diagnosis of acute coronary syndromes,24 are associated with an increased likelihood of multivessel
CAD in stable patients with ESRD. The fact that elevated levels
of cTnT retain their predictive value for death in patients with and without
known CAD, even after adjustment for other cardiac risk factors including
white race, suggests the common nature of coronary atherosclerosis in the
population of patients with ESRD. In this setting where the timing of troponin
release is unknown, elevation of cTnT level, as a marker of diffuse multivessel
CAD, fits with current thought that many acute coronary events, with associated
plaque rupture and microembolization, are clinically silent.25
It has been proposed that elevation of levels of cTnT in patients receiving
dialysis may result from nonischemic cardiomyopathy or microvascular disease
in the setting of LVH.26,27 Our
echocardiographic findings demonstrated no correlation between left ventricular
mass and elevated cTnT level, possibly because LVH is present in the majority.28 In addition, cTnT did not appear to correlate with
either depressed LVEF or hypertrophy in patients without angiographically
validated CAD. These findings highlight the need for further investigation
using more sensitive techniques in patients without angiographic CAD to detect
coronary atherosclerosis and myocardial ischemia.
Despite the near ubiquitous presence of elevated levels, CRP retained
an independent association with all-cause mortality consistent with several,3,4,6 but not all,5,7 previous findings in patients with
ESRD. Studies of patients receiving dialysis often show striking elevations
of this marker of systemic inflammation.3- 7 Our
use of the high-sensitivity assay for CRP detected values above the 80th percentile
of the general population (>4.19 mg/L, highest risk group) in more than 70%
of our study patients.29 Unlike the general
population, in which low levels of CRP elevation may reflect chronic vascular
inflammation and increased risk of vascular events, patients receiving dialysis
are subject to multiple nonvascular inflammatory stimuli, including chronic
infections and the dialysis process itself.30 Therefore,
given the multiple etiologies of CRP level elevation in these patients, it
should not be surprising that prediction of risk of cardiac death is complemented
by simultaneous biochemical evidence of myocardial injury. This risk appears
graded, with the spectrum of risk continuing to increase at CRP levels exceeding
those found to predict risk in the general population.31 One
hypothesis for this finding is that, while many patients with ESRD and a large
burden of coronary atherosclerosis are identified by elevated levels of cTnT
resulting from clinically silent plague ruptures, ultimately it is the systemic
inflammatory milieu that raises the risk of triggering a fatal cardiac event.
The absence of an association between elevated levels of CRP and angiographic
evidence of multivessel CAD in patients receiving hemodialysis is consistent
with the poor correlation that exists between angiographic findings and results
of electron-beam computed tomography with levels of CRP in clinically stable
patients in the general population.32,33 Based
on the frequent presence of elevated levels of cTnT and their association
with diffuse CAD, stratification of risk might best compare hemodialysis patients
without ischemic syndromes with acute coronary syndrome patients without renal
disease. In such patients, elevated levels of CRP also provide additional
long-term risk prediction independent of cTnT.34
Selection of patients with ESRD but without ischemic symptoms to undergo
coronary angiography would ideally be randomized. However, such a strategy
could have introduced a selection bias into this prospective outcomes study
by intentionally enrolling only patients who were also willing to undergo
angiography. Reliance on volunteers potentially improves the generalizability
of the findings, particularly compared with series of dialysis patients referred
for clinical indications.
Echocardiography comprised a second complementary substudy for evaluating
the pathophysiology of biomarker elevation. The diverse and distant sites
of this multicenter study limited the rapid acquisition of high-quality echocardiograms.
Despite the large number of echocardiograms analyzed, bias due to missing
data and temporal changes in left ventricular function or mass cannot be excluded.
Determination of cause of death using death notifications completed
by clinicians (as used by the USRDS) may be inaccurate.35 However,
when the USRDS-determined cause of death was compared with the adjudicated
end points of a large prospective clinical trial, the classification of death
as cardiac compared favorably whereas the type of cardiac death correlated
poorly between the two.36 Therefore, we did
not comment on the type of cardiac death identified by high levels of biomarkers.
For patients with ESRD but without ischemic symptoms undergoing hemodialysis,
randomly assessed levels of cTnT and CRP independently identify patients at
risk of death, and the combination of the 2 levels identify patients at particularly
high risk. Furthermore, small elevations of cTnT level predict a markedly
increased risk of multivessel coronary atherosclerosis. Taken together, these
findings identify a potential role for these markers to be incorporated into
future diagnostic and therapeutic strategies aimed at the earlier detection
and management of clinically silent, but high-risk, diffuse CAD.