Nephrol Dial Transplant (2008) 23: 3546–3553
doi: 10.1093/ndt/gfn341
Advance Access publication 18 June 2008
Original Article
Prognostic value of cardiac biomarkers for death in a non-dialysis
chronic kidney disease population
Susan Vickery1 , Michelle C. Webb2 , Christopher P. Price3 , Robert Ian John2 , Nasir A. Abbas2 and
Edmund J. Lamb1
1
Department of Clinical Biochemistry, 2 Department of Renal Medicine, East Kent Hospitals NHS Trust, Canterbury, Kent, CT1 3NG
and 3 Department of Clinical Biochemistry, University of Oxford, Oxford, 0X3 9DU, UK
Abstract
Background. Excess mortality in patients with chronic
kidney disease (CKD) is predominantly due to cardiovascu-
lar disease. We explored the prognostic value of biomarkers
of cardiac overload [B-type natriuretic peptide (BNP) and
N-terminal pro-B-type natriuretic peptide (NT-proBNP)]
and inflammation [high-sensitivity C-reactive protein
(hsCRP)] for all-cause mortality in patients with CKD.
Methods. Plasma BNP (Siemens Medical Solutions Di-
agnostics, Frimley, Surrey, UK) and NT-proBNP (Roche
Diagnostics PLC, East Sussex, UK), and hsCRP (Siemens
Medical Solutions Diagnostics) were measured at study
entry. Echocardiograms were undertaken, and left ventric-
ular mass index (LVMI) was calculated. CKD patients
(n = 213) were followed for up to 53 months. Kaplan–
Meier survival analysis with log-rank testing and hazards
ratios (HRs) were calculated for each cardiac biomarker,
stratified by respective median values, as a predictor of
death to assess outcome.
Results. Fifty-four deaths occurred. NT-proBNP concen-
tration ≥89 pmol/L (HR 5.6, P < 0.0001), BNP concentra-
tion ≥14 pmol/L (HR 3.5, P < 0.001), NT-proBNP/BNP
ratio ≥6 pmol/pmol (HR 2.6, P < 0.01) and hsCRP con-
centration ≥4.7 mg/L (HR 2.4, P < 0.01) were unadjusted
predictors of death. Only NT-proBNP ≥89 pmol/L (HR
2.5, P < 0.05) and hsCRP ≥4.7 mg/L (HR 1.9, P < 0.05)
were independent predictors of death when the HRs were
adjusted for significant clinical variables (age, estimated
glomerular filtration rate, LVMI and vascular disease).
Conclusion. NT-proBNP and hsCRP can independently
predict all-cause mortality in a non-dialysis CKD popu-
lation and may have a useful role in risk stratification.
Keywords: B-type natriuretic peptides; chronic kidney
disease; high-sensitivity CRP
Correspondence and offprint requests to: Susan Vickery, Department of
Clinical Biochemistry, East Kent Hospitals NHS Trust, Kent and Canter-
bury Hospital, Canterbury, Kent, CT1 3NG, UK. Tel: +44-1227-766877,
Ext: 74736; Fax: +44-1227-783077; E-mail: susan.vickery@ekht.nhs.uk

C The Author [2008]. Published by Oxford University Press on behalf of ERA-EDTA. All rights reserved.
For Permissions, please e-mail: journals.permissions@oxfordjournals.org
Downloaded from http://ndt.oxfordjournals.org/ by guest on August 25, 2015
Introduction
Cardiovascular disease is prevalent and the leading cause of
death in patients with end-stage renal disease (ESRD) [1–3].
The most common manifestation of cardiovascular disease
in CKD patients is left ventricular hypertrophy (LVH), pre-
dominantly as a result of hypertension and anaemia. LVH
is a powerful independent predictor of cardiovascular dis-
ease in CKD patients. However, identifying which patients
will suffer cardiovascular events is challenging and requires
early identification and treatment. The ability to detect sig-
nificant cardiovascular dysfunction at an early stage could
facilitate more aggressive and focused treatment of those
at increased risk.
B-type natriuretic peptide (BNP) and N-terminal pro-
B-type natriuretic peptide (NT-proBNP) are co-secreted in
equimolar amounts from the heart in response to left ven-
tricular overload. Diagnostically they are used as rule-out
tests to exclude heart failure in patients presenting with
shortness of breath. The natriuretic peptides have been
shown to predict survival in patients with congestive heart
failure [4–6]. Elevated concentrations of BNP and NT-
proBNP are found in both ESRD and non-dialysis CKD
patients [7–11].
Systemic inflammation plays a major role in the develop-
ment of atherosclerosis leading to coronary heart disease.
C-reactive protein (CRP), an acute phase protein, is the hep-
atic product of cytokine activation and may act as a marker
of inflammatory mediators involved in atherogenic plaque
progression and stability. CRP concentrations, when mea-
sured using highly sensitive assays (hsCRP), may predict
future coronary events [12]. It is generally agreed that a
hsCRP concentration >3 mg/L is a strong risk factor for a
coronary event [13].
Cardiac biomarkers have been used in the risk stratifi-
cation of death for ESRD patients [5,11,14,15]. However,
there is little data on the outcome of patients with earlier
stages of CKD. We have explored the prognostic value of
biomarkers of cardiac overload (BNP and NT-proBNP) and
inflammation (hsCRP) for all-cause mortality in patients
Cardiac biomarkers in CKD
with CKD who have yet to commence renal replacement
therapy.
Methods
Patients
Two hundred and twenty-nine patients with CKD, not re-
ceiving renal replacement therapy at entry to the study,
attending nephrology out-patient clinics were recruited to
this single centre study between June 2003 and June 2004
during times when the investigators were available. Ex-
cluded from the study were patients with acute renal failure
or a functioning renal transplant, those receiving dialysis
and patients with a recent (<1 month) cardiac event. In
16 patients, insufficient plasma for analysis of BNP and NT-
proBNP was obtained. A final cohort of 213 patients was
analysed. The patients were followed for up to 53 months.
All patients gave informed consent and the study had ethical
approval from the Kent and Medway Strategic Health Au-
thority Research Ethics Committee (REC no EK050/3/03).
Information on patient age, gender, body mass index
(BMI), mean arterial blood pressure (MABP), drug history
and presence or absence of diabetes mellitus, cardiovascu-
lar disease and arteriopathic disease was collated. MABP
was calculated from mean diastolic and systolic blood pres-
sure using the formula: diastolic pressure + 1/3(systolic
pressure − diastolic pressure). Vascular disease was con-
sidered present if there was a history of myocardial in-
farction, angina, arrhythmia, valvular disease, congestive
cardiac failure or a requirement for coronary intervention
(angioplasty, coronary artery bypass graft or pacemaker),
cerebrovascular or peripheral vascular disease.
CKD staging was based on the estimated glomerular
filtration rate (eGFR) calculated using the simplified Mod-
ification of Diet in Renal Disease (MDRD) study formula
[16]. Patients were stratified into stage 3 (moderate; eGFR
30–59 mL/min/1.73 m2 ), stage 4 (severe; eGFR 15–29 mL/
min/1.73 m2 ) and stage 5 (failure; eGFR <15 mL/min/
1.73 m2 ) CKD.
Analytical methods
Blood was collected in the non-fasting state. For routine
analyses blood was allowed to clot at room temperature
and, following centrifugation, the serum concentrations
were determined within 4 h of venesection. For plasma
BNP, NT-proBNP, parathyroid hormone (PTH) and blood
haemoglobin measurement, blood was collected into plastic
tubes containing potassium EDTA (1.8 mg/mL of blood).
Aliquots of plasma and serum were frozen in plastic tubes
within 3–4 h of collection at −80◦ C (for natriuretic peptide
and hsCRP measurements, respectively) and at −20◦ C (for
PTH measurements). PTH and natriuretic peptide analyses
were undertaken within 1 month and 1 year of collection,
respectively, and hsCRP was measured within 3 years of
collection. Serum creatinine was measured using a compen-
sated rate Jaffe method on an Integra 800 analyser (Roche
Diagnostics Ltd, East Sussex, UK). PTH was measured
using an immunochemiluminometric assay on an ADVIA
3547
Centaur analyser (Siemens Medical Solutions Diagnostics,
Frimley, Surrey, UK); the reference range for this assay
is 14–72 ng/L. All patients also had blood haemoglobin,
serum albumin, cholesterol, calcium and phosphate mea-
sured using standard laboratory methods.
Plasma BNP concentration was determined using the
ADVIA Centaur
R
BNP immunoassay on the ADVIA Cen-
taur analyser (Siemens Medical Solutions Diagnostics,
Frimley, Surrey, UK). Between-day imprecision (n = 10)
was 7.8%, 6.7% and 5.3% at BNP concentrations of 14
pmol/L, 133 pmol/L and 488 pmol/L, respectively. The
manufacturer’s proposed decision threshold for excluding
heart failure, in patients without renal insufficiency, is
29 pmol/L (100 ng/L).
Plasma NT-proBNP concentration was determined using
the Roche proBNP electrochemiluminescence immunoas-
say on the Elecsys 1010 analyser (Roche Diagnostics PLC,
East Sussex, UK). Between-day imprecision (n = 11) was
2.8% and 3.7% at NT-proBNP concentrations of 23 pmol/L
and 513 pmol/L, respectively. The manufacturer’s proposed
Downloaded from http://ndt.oxfordjournals.org/ by guest on August 25, 2015
decision threshold for excluding heart failure, in patients
without renal insufficiency, is 10 pmol/L (84 ng/L) in males
and 18 pmol/L (155 ng/L) in females for patients ≤50 years
and 23 pmol/L (194 ng/L) in males and 26 pmol/L
(222 ng/L) in females for patients >50 years.
Serum hsCRP concentration was determined using
the Dade Behring CardioPhase hsCRP particle-enhanced
immunonephelometric assay on the BN ProSpec
R
anal-
yser (Siemens Medical Solutions Diagnostics). Between-
day imprecision (n = 6) was 3.6% and 1.3% at hsCRP
concentrations of 1.6 mg/L and 13.3 mg/L, respectively.
The manufacturer’s proposed thresholds for cardiovascu-
lar risk prediction are <1.0 mg/L, low risk; 1.0–3.0 mg/L,
average risk; and >3.0 mg/L, high risk.
Patients underwent two-dimensional targeted M-mode
echocardiography, and left ventricular mass index (LVMI)
(n = 201) was calculated as previously described [10].
Data analysis
Data were analysed using Analyse-ItTM (Analyse-it Soft-
ware Ltd, Yorkshire, UK), InStatTM (GraphPad.com), Stats-
Direct (Cheshire, UK) and SAS version 8.01 (SAS Insti-
tute Inc., Cary, USA). A P-value <0.05 was considered
significant.
None of the data, with the exception of haemoglobin,
followed a normal distribution; therefore, concentrations
were compared between stage 3, 4 and 5 CKD using the
Kruskal–Wallis test [non-parametric analysis of variance
(ANOVA)]. Categorical variables were compared between
stage 3, 4 and 5 CKD using the chi-squared test. Receiver–
operator curve (ROC) analysis for NT-proBNP, BNP, NT-
proBNP/BNP ratio and hsCRP as a predictor of death was
undertaken to enable comparison of areas under the curve
and to determine cut-off points at which the sum of sensi-
tivity and specificity was maximized.
Survival analyses
Patients were prospectively followed for up to 53 months
[median (IQR) = 40(13–47) months]. All cause mortality
3548 S. Vickery et al.
Table 1. Patient characteristics by CKD stage

All Stage 3 Stage 4 Stage 5 P-value

No of patients 213 55 66 92
Male:female (n) 145:77 43:13 49:21 53:43 0.02
Age (years) 68 (58–75) 68 (58–76) 71 (64–75) 65 (54–71) <0.01
BMI (kg/m2 ) 29 (24–32) (n = 207) 29 (26–33) (n = 55) 29 (25–32) (n = 64) 26 (23–32) 0.2
MABP (mmHg) 97 (92–103) 96 (90–102) 95 (91–104) 99 (94–104) 0.09
eGFR (mL/min/1.73 m2 ) 18 (10–30) 39 (34–46) 22 (18–25) 9 (8–11)
Haemoglobin (g/dL) 12.2 (11.1–13.4) 13.4 (12.6–13.9) 12.2 (11.1–13.3) 11.5 (10.6–12.3) <0.001
LVMI (g/m2 )a 142 (116–178) (n = 192) 127 (112–158) (n = 54) 136 (112–174) (n = 59) 155 (123–203) (n = 79) <0.01
Presence of left ventricular 122/192 29/54 38/59 57/79 0.06
hypertrophyb
Vascular diseasec 91 30 35 26 <0.01
Diabetes 57 11 24 22 0.1
Serum total cholesterol 4.5 (3.8–5.3) 4.8 (4.0–5.8) 4.7 (3.8–5.1) 4.3 (3.7–5.1) 0.02
(mmol/L)
Serum albumin (g/L) 42.0 (40.0–44.0) 43.0 (41.0–45.0) 41.0 (39.0–44.3) 42.0 (40.0–44.0) 0.06
Plasma PTH (ng/L) 99 (50–177) 66 (47–105) 104 (47–151) 121 (67–243) <0.01
Serum calcium × phosphate 3.1 (2.7–3.7) 2.8 (2.5–3.1) 2.9 (2.6–3.2) 3.7 (3.1–4.3) <0.001
product (mmol2 /L2 )
Plasma BNP (pmol/L) 14 (7–30) 9 (4–19) 14 (6–23) 21 (11–45) <0.001
Plasma NT-proBNP (pmol/L) 89 (31–242) 33 (14–85) 68 (31–134) 193 (72–405) <0.001

Downloaded from http://ndt.oxfordjournals.org/ by guest on August 25, 2015

Plasma NT-proBNP/BNP 6 (4–9) 4 (3–6) 6 (4–8) 8 (5–12) <0.001
ratio (pmol/pmol)
Serum hsCRP (mg/L, 4.7 (1.8–9.8) 4.5 (2.1–10.1) 5.9 (2.1–10.8) 3.7 (1.3–9.3) 0.3
n = 212)
Deaths 54 12 21 21 0.3

Values expressed as mean ± SD when normally distributed, medians (interquartile range) when not normally distributed or number. Data were available
for all patients unless indicated otherwise.
a Calculated from mean diastolic and systolic blood pressure using the formula: diastolic pressure + 1/3(systolic pressure − diastolic pressure).
b Considered present when LVMI exceeded 125 g/m2 .
c Considered present if there was a history of myocardial infarction, angina, arrhythmia, valvular disease, congestive cardiac failure or a requirement
for coronary intervention (angioplasty, coronary artery bypass graft or pacemaker), cerebrovascular or peripheral vascular disease.

was recorded. Exposure was computed from the date of Results

blood draw until the date of death with censoring for re-
nal replacement therapy (n = 49 patients) or renal trans-
plantation (n = 11 patients). Differences in survival rates
were compared between patients with BNP, NT-proBNP,
NT-proBNP/BNP ratio or hsCRP concentrations above or
below their respective median values for the cohort. Sur-
vival curves were determined using the Kaplan–Meier cal-
culation and the significance between the curves tested us-
ing the log-rank test. Unadjusted and adjusted hazard ratios
(HRs) for death were calculated using Cox’s proportional
HR method. Unadjusted HRs were calculated for age, gen-
der, BMI, MABP, eGFR, haemoglobin, parathyroid hor-
mone, calcium × phosphate product, cholesterol, albumin,
LVMI, vascular disease and diabetes mellitus. Those vari-
ables that had a significant (P < 0.05) HR were included
in the Cox’s proportional HR model, and manual backward
elimination was used to remove any variables that had be-
come non-significant (P ≥ 0.05). Each cardiac biomarker
was then separately entered into the model to provide ad-
justed HRs. The final adjusted model met the constant
proportional hazards assumption. No interactions between
eGFR, age and LVMI were found, and no multicollinear-
ity was detected in the final model. The overall fit of
the model was highly significant (chi-square likelihood =
44, P < 0.0001). Likelihood ratios were calculated to en-
able comparison of pre- and post-test probabilities of death
with the biomarker assays.
Patient characteristics for the whole study population and
by CKD stage are given in Table 1 and have been described
in more detail previously [17]. Over the follow-up period
54 deaths occurred. Survival was decreased amongst those
patients with plasma BNP (41% versus 84%, P < 0.0001),
plasma NT-proBNP (43% versus 80%, P < 0.0001), plasma
NT-proBNP/BNP ratio (55% versus 73%, P = 0.003) and
hsCRP (53% versus 76%, P = 0.002) concentrations greater
than the median for the cohort (Figure 1).
Unadjusted HRs for death were determined for a range
of clinical variables (Table 2). HRs for each of the car-
diac biomarkers were then adjusted for those four clinical
variables that remained significant in a multiple regression
model (i.e. age, eGFR, LVMI and the presence of vascu-
lar disease) (Table 3). Significant unadjusted HRs were
obtained for plasma BNP (P < 0.001), NT-proBNP (P <
0.0001), NT-proBNP/BNP ratio (P < 0.01) concentrations
greater than the median for the cohort (Table 3). When
stratified by the accepted 3 mg/L ‘high-risk’ cut-off the
unadjusted HR for hsCRP was not significant. However, a
significant unadjusted HR for hsCRP was obtained for a
concentration greater than the median (HR 2.4, P < 0.01)
for the cohort. NT-proBNP (HR 2.5, P < 0.05) and hsCRP
(HR 1.9, P < 0.05) concentrations greater than their respec-
tive median values were found to be significant independent
predictors of death when the HRs were adjusted for all four
Cardiac biomarkers in CKD 3549

(A) 1.0

0.9 BNP <14 pmol/L

0.8
Relative Survival (%)

0.7

0.6

0.5
p < 0.0001
BNP >14 pmol/L
0.4

0.3
0 5 10 15 20 25 30 35 40 45 50 55
Time (months)

Downloaded from http://ndt.oxfordjournals.org/ by guest on August 25, 2015

Number at risk
<14 pmol/L 105 100 95 87 83 78 73 68 30 4
>14 pmol/L 108 92 71 58 52 44 39 31 11 1

(B) 1.0

0.9 NT-proBNP <89 pmol/L

0.8
Relative Survival (%)

0.7

0.6

0.5
p < 0.0001

0.4
NT-proBNP >89 pmol/L

0.3
0 5 10 15 20 25 30 35 40 45 50 55
Time (months)
Number at risk
<89 pmol/L 106 102 96 91 86 82 79 74 31 4
>89 pmol/L 107 90 70 54 49 40 33 25 10 1
Fig. 1. Kaplan–Meier survival curves showing the association between survival and (A) median concentration for plasma BNP, (B) median concentration
for plasma NT-proBNP, (C) median concentration for the plasma NT-proBNP/BNP ratio, (D) median concentration for serum hsCRP. The significance
between the groups was determined using the log-rank statistical test.

significant clinical variables. BNP and NT-proBNP/BNP
ratio were not found to be significant independent predic-
tors of death.
The use of biomarkers in combination was also explored,
including the previously published cardiac troponin T
(cTnT) data from the same cohort [18]. Unadjusted and ad-
justed HRs were calculated for the following combinations:
NT-proBNP concentration ≥ the median with a hsCRP
concentration ≥ the median, NT-proBNP concentration ≥
the median with a cTnT concentration ≥ 0.03 µg/L, hsCRP
concentration ≥ the median with a cTnT concentration ≥
0.03 µg/L and finally NT-proBNP concentration ≥ the
median with a hsCRP concentration ≥ the median and
a cTnT concentration ≥ 0.03 µg/L. Non-significant
unadjusted HRs and subsequently non-significant adjusted
HRs were obtained for all combinations of markers (data
not shown) and added no further prognostic power than
that seen with the individual markers.
3550 S. Vickery et al.

(C) 1.0

0.9
NT-proBNP/BNP Ratio <6 pmol/pmol

0.8
Relative Survival (%)

0.7

0.6

0.5 NT-proBNP/BNP Ratio >6 pmol/pmol

p = 0.003

0.4

0.3
0 5 10 15 20 25 30 35 40 45 50 55
Time (months)
Number at risk

Downloaded from http://ndt.oxfordjournals.org/ by guest on August 25, 2015

<6 pmol/pmol 107 103 98 90 84 75 72 66 16 2
>6 pmol/pmol 106 89 68 56 51 47 41 33 10 1

(D) 1.0

0.9
hsCRP <4.7 mg/L

0.8
Relative Survival (%)

0.7

0.6

0.5 hsCRP >4.7 mg/L

p = 0.002

0.4

0.3
0 5 10 15 20 25 30 35 40 45 50 55
Time (months)
Number at risk
<4.7 mg/L 105 98 81 73 68 61 60 52 18 3
>4.7 mg/L 102 90 81 70 63 58 51 46 23 2
Fig. 1. (Continued)

The likelihood of death in this population was 25% (i.e. P < 0.01) and 0.692 (95% CI 0.608–0.776, P < 0.0001),
54 deaths in 213 patients). This likelihood rose to 36% respectively.
in the presence of an NT-proBNP concentration exceeding
89 pmol/L or hsCRP concentration exceeding 4.7 mg/L (the
respective negative likelihoods were 14% and 15%). Discussion
ROC analyses showed the optimal cut-off points for this
cohort as 128 pmol/L for NT-proBNP, 17.9 pmol/L for BNP, We have compared the prognostic value of several car-
8.1 pmol/pmol for NT-proBNP/BNP ratio and 8.0 mg/L for diac biomarkers in a large CKD population who have yet
hsCRP with areas under the curve of 0.691 [95% confidence to commence renal replacement therapy. We have shown
interval (CI) 0.611–0.771, P < 0.0001], 0.671 (95% CI NT-proBNP and hsCRP to be independent predictors of
0.589–0.752, P < 0.0001), 0.620 (95% CI 0.534–0.706, death in this CKD population. Our data are consistent with
Cardiac biomarkers in CKD 3551
Table 2. Unadjusted hazard ratios for death by known clinical variables (n = 213 unless otherwise stated)

Unadjusted hazard ratio (95% CI) P-value

Age (years) 1.08 (1.05–1.12) <0.001

Female:male (n) 1.07 (0.60–1.90) 0.8
MABP (mm Hg) 0.99 (0.97–1.02) 0.9
BMI (kg/m2 ) 0.95 (0.90–1.00) 0.1
eGFR (mL/min/1.73 m2 ) 0.97 (0.95–0.99) <0.01
Haemoglobin (g/dL) 0.76 (0.62–0.94) <0.01
Parathyroid hormone (ng/L) 1.00 (0.99–1.00) 0.1
Calcium × phosphate product (mmol2 /L2 ) 1.15 (0.82–1.63) 0.4
Total cholesterol (mmol/L) 0.86 (0.67–1.11) 0.2
Albumin (g/L) 0.89 (0.83–0.96) <0.01
LVMI (g/m2 , n = 192) 1.01 (1.00–1.01) <0.0001
Vascular disease 2.89 (1.63–5.14) <0.001
Diabetes mellitus (n) 1.33 (0.75–2.37) 0.3

CI, confidence interval; MABP, mean arterial blood pressure; eGFR, estimated glomerular filtration rate; LVMI,
left ventricular mass index.
All variables included in the model were continuous, with the exception of gender, vascular disease and diabetes
mellitus that were dichotomous. The significant unadjusted hazard ratios were included in a Cox’s proportional
hazards ratio model where haemoglobin and albumin did not remain significant. Hazard ratios, for each biomarker,

Downloaded from http://ndt.oxfordjournals.org/ by guest on August 25, 2015

were therefore adjusted for age, eGFR, LVMI and presence of vascular disease (Table 3).
For continuous variables, the HRs are expressed as the increased risk associated with an increase (age, MABP,
BMI, PTH, calcium × phosphate, cholesterol, LVMI) or decrease (eGFR, haemoglobin, albumin) of one unit (e.g.
1 year for age). For categorical variables, the HRs are expressed as risk if conditions present (vascular disease and
diabetes) or if female, rather than male, compared to risk in the absence of conditions.

Table 3. Unadjusted and adjusted hazard ratios for death by natriuretic peptides and hsCRP

Hazard ratio (95% CI)

Biomarker Patients at risk (n) Unadjusted P-value Adjusted (age, eGFR, LVMI and vascular disease) P-value

BNP ≥14 pmol/La 213 3.5 (1.8–6.6) <0.001 1.2 (0.6–2.5) 0.6
NT-proBNP ≥89 pmo/La 213 5.6 (2.8–11.2) <0.0001 2.5 (1.3–5.8) <0.05
NT-proBNP/BNP ≥6 pmol/pmola 213 2.6 (1.4–4.8) <0.01 1.8 (0.9–3.5) 0.1
hsCRP ≥3 mg/Lb 212 1.7 (0.9–3.4) 0.1 1.5 (0.8–3.0) 0.2
hsCRP ≥4.7 mg/La 212 2.4 (1.4–4.8) <0.01 1.9 (1.0–3.8) <0.05

BNP, B-type natriuretic peptide; CI, confidence interval; eGFR, estimated glomerular filtration rate; NT-proBNP, N-terminal pro-B-type natriuretic
peptide; hsCRP, high-sensitivity C-reactive protein.
a Respective median value for the cohort.
b High-risk cut-off for hsCRP (3 mg/L).

studies in the haemodialysis population, where it has been
shown that these same markers are predictors of mortal-
ity [15], albeit that the predictive effects we have observed
have occurred at much lower NT-proBNP concentrations
and higher hsCRP concentrations than those seen in the
dialysis patients. A relatively small (n = 83) study of pre-
dialysis CKD patients also found NT-proBNP to be an in-
dependent predictor of mortality and cardiovascular events,
which occurred in 10 patients. The effect was strongest at
lower eGFRs, and consistent with our data, pre-existing
cardiovascular disease was also an independent predictor of
death [19].
Of the cardiac biomarkers we evaluated, NT-proBNP
was the strongest significant predictor of death. This is
consistent with the findings of deFilippi et al. who found
that NT-proBNP, and not BNP, was an independent predic-
tor of death in a heart failure population that had eGFRs
<60 mL/min/1.73 m2 [20]. BNP and NT-proBNP, upon
stimulus, are released in equimolar amounts from the ven-
tricles of the heart, and both natriuretic peptides are used
clinically in the diagnosis of heart failure [6]. Why then have
we observed differences between the prognostic power of
BNP and NT-proBNP in this CKD cohort? Natriuretic pep-
tide concentration is independently increased in response
to increasing LVMI, and ultimately LVH, and we have pre-
viously shown that this is also true in the CKD popula-
tion [10]. Natriuretic peptide concentration is also indepen-
dently increased in CKD patients in response to declining
GFR [10]. Almost twice the concentration of NT-proBNP
(38%) appears to be retained in the circulation of CKD
patients compared to BNP (21%) for every 10 mL/min/
1.73 m2 reduction in GFR [10]. This observation may be
due to the differences in how BNP and NT-proBNP are me-
tabolized. BNP is removed from the circulation by binding
to natriuretic peptide receptors A and C and degradation by
neutral endopeptidases distributed throughout the endothe-
lium [21,22]. It is not known whether BNP metabolism
alters with renal disease. In contrast, NT-proBNP binds to
neither natriuretic peptide receptor A nor C and does not act
as a substrate for endopeptidases. It is believed to be solely
3552
cleared in the kidney [21], although some studies have sug-
gested equivalent clearance of both natriuretic peptides by
the kidney [23,24]. It is also becoming clear that natriuretic
peptide assays may have cross reactivity with proBNP and
other cleavage products, which may circulate at higher con-
centrations than previously thought [25]. The relative influ-
ence, if any, of these products in the current assays are not
known. Whatever the mechanism, BNP concentrations as
measured by the present assay are also affected by declining
GFR, as described above, although the effect is not as great
as that seen with NT-proBNP. In this study, having adjusted
for both renal and left ventricular function, we have still ob-
served a difference between the prognostic power of BNP
and NT-proBNP.
The NT-proBNP/BNP ratio increases with declining re-
nal function [10]. Intuitively, this has been considered
to be consistent with decreased renal clearance of NT-
proBNP. However, as discussed above, it could equally
reflect upregulated extra-renal BNP metabolism as eGFR
falls, or changes in the relative contribution of cross-
reacting natriuretic peptide fragments and precursors. The
NT-proBNP/BNP ratio was not a significant independent
predictor of death, possibly because of its relationship to
eGFR.
hsCRP was an independent predictor of death when strat-
ified by the median (4.7 mg/L), but in our hands did not
independently predict death at the widely used high-risk
concentration of 3 mg/L. We are unaware of other published
survival data for hsCRP in the non-dialysis CKD popula-
tion. Whilst a significant difference was seen in the survival
rates of those patients with hsCRP concentrations below
and above the median, no significant difference in hsCRP
concentration was found between the CKD stages. This is
consistent with the findings of Ortega et al. who also found
no association between hsCRP concentration and declining
renal function in non-dialysis patients [26]. Such observa-
tions may, however, be due to survivorship bias. In contrast,
some workers have found creatinine clearance to be an inde-
pendent predictor of hsCRP concentration in non-dialysis
CKD populations; however, these studies lacked outcome
data to support the role of hsCRP in predicting mortality
[27,28].
The outcome data we have presented are at odds with
that observed in the dialysed population, where hsCRP was
found to be a powerful cardiac biomarker for predicting
all cause death at concentrations around 1–3 mg/L [15,29],
although the clinical utility of hsCRP to predict death in
ESRD is still under debate [30]. A range of factors could
account for the differences seen between hsCRP in the pre-
dialysis and dialysis populations. Exposure to the dialy-
sis membrane, even modern biocompatible dialysis mem-
branes, and vascular access infections are both associated
with cytokine activation and may contribute to a different
microinflammatory state to that observed in pre-dialysis
patients.
There are several limitations in this study. At inclusion
into the study, the patients were at different stages of CKD
and cardiac dysfunction. This may have introduced an el-
ement of survivorship bias with CKD stage 5 patients, for
example, having survived cardiovascular disease that may
have killed other patients at an earlier stage of CKD. Nev-
S. Vickery et al.
ertheless, this study reflects a reality situation in which
patients with varying stages of disease are assessed using
similar clinical approaches. As this is one of the first studies
to explore the role of these biomarkers in the pre-dialysis
CKD population, there is a lack of established cut-off points
for each cardiac biomarker. From this population we have
derived ROC cut-off points that could be applied to a sep-
arate but comparable population. Indeed, a future direction
of this work is targeted at validation studies in an attempt
to obtain clinically useful cut-off points.
In conclusion, cardiac biomarkers may have a useful role
in risk stratification of patients with CKD. Our present work
has provided evidence that NT-proBNP and hsCRP can,
depending on the cut-off point used, independently pre-
dict all-cause mortality in a non-dialysis CKD population.
These markers remained significant independent predictors
of death even in the presence of a clinical diagnosis of vas-
cular disease. In other words, knowledge of these markers
provides additional prognostic information to that supplied
by the clinical history alone, including echocardiographic
Downloaded from http://ndt.oxfordjournals.org/ by guest on August 25, 2015
assessment. The clinical challenge is what can be done
for these patients to reduce progression of cardiovascular
disease. This should include focussing more attention on
their conventional cardiovascular risk management (e.g.
blockade of the renin–angiotensin–aldosterone system [31]
and other anti-hypertensive treatments, management of
dyslipidaemia [32] and introducing lifestyle modifications
such as smoking cessation, obesity reduction and decreased
salt intake) but could extend to more aggressive interven-
tions. There is now a need for the medical community to
apply this knowledge in testing and developing interven-
tional strategies to reduce mortality in those at greatest
risk.
Acknowledgements. This study was partly funded by East Kent Hospitals
NHS Trust. We are grateful to Siemens Medical Solutions Diagnostics for
support with reagent costs. We would like to thank the staff of the Clin-
ical Biochemistry and Renal Medicine Departments at East Kent Hospi-
tals NHS Trust for their co-operation and help. We also thank Dr Alexa
Laurence, from the Institute of Mathematics, Statistics and Actuarial Sci-
ence, University of Kent, for providing expert statistical advice to this
study.
Conflict of interest statement. Prof. Price is a former employee of Siemens
Medical Solutions Diagnotics. No other conflict of interest is declared.
References
1. Foley RN, Parfrey PS. Cardiovascular disease and mortality in ESRD.
J Nephrol 1998; 11: 239–245
2. Foley RN, Parfrey PS, Sarnak MJ. Epidemiology of cardiovascular
disease in chronic renal disease. J Am Soc Nephrol 1998; 9: S16–S23
3. Levin A, Foley RN. Cardiovascular disease in chronic renal insuffi-
ciency. Am J Kidney Dis 2000; 36: S24–S30
4. Doust JA, Pietrzak E, Dobson A et al. How well does B-type natriuretic
peptide predict death and cardiac events in patients with heart failure:
systematic review. BMJ 2005; 330: 625–627
5. Ishii J, Lu JC, Iwashima S et al. Prognostic value of combination
of cardiac troponin T and B-type natriuretic peptide in outpatients
on chronic hemodialysis: 5-year outcome analysis. Circulation 2004;
110: 2193
6. Masson S, Latini R, Anand IS et al. Direct comparison of B-type
natriuretic peptide (BNP) and amino-terminal proBNP in a large pop-
ulation of patients with chronic and symptomatic heart failure: the
Cardiac biomarkers in CKD
valsartan heart failure (Val-HeFT) data. Clin Chem 2006; 52: 1528–
1538
7. Clerico A, Caprioli R, Del Ry S et al. Cardiac natriuretic peptides
measured by means of competitive and non-competitive immunoas-
say methods in patients with renal failure on chronic hemodialysis.
J Endocrinol Invest 2001; 24: 24–30
8. Mallamaci F, Zoccali C, Tripepi G et al. Diagnostic potential of car-
diac natriuretic peptides in dialysis patients. Kidney Int 2001; 59:
1559–1566
9. Nishikimi T, Futoo Y, Tamano K et al. Plasma brain natriuretic peptide
levels in chronic hemodialysis patients: influence of coronary artery
disease. Am J Kidney Dis 2001; 37: 1201–1208
10. Vickery S, Price CP, John RI et al. B-type natriuretic peptide (BNP)
and amino-terminal proBNP in patients with CKD: relationship to
renal function and left ventricular hypertrophy. Am J Kidney Dis
2005; 46: 610–620
11. Zoccali C, Mallamaci F, Benedetto FA et al. Cardiac natriuretic pep-
tides are related to left ventricular mass and function and predict
mortality in dialysis patients. J Am Soc Nephrol 2001; 12: 1508–
1515
12. Ridker PM. Clinical application of C-reactive protein for cardiovascu-
lar disease detection and prevention. Circulation 2003; 107: 363–369
13. Pearson TA, Mensah GA, Alexander RW et al. Markers of inflam-
mation and cardiovascular disease. Application to clinical and public
health practice. A statement for healthcare professionals from the
center for disease control and prevention and the American Heart
Association. Circulation 2003; 107: 499–511
14. Apple FS, Murakami MM, Pearce LA et al. Predictive value of cardiac
troponin I and T for subsequent death in end-stage renal disease.
Circulation 2002; 106: 2941–2945
15. Apple FS, Murakami MM, Pearce LA, et al. Multi-biomarker risk
stratification of N-terminal pro-B-type natriuretic peptide, high-
sensitivity C-reactive protein, and cardiac troponin T and I in end-stage
renal disease for all-cause death. Clin Chem 2004; 50: 2279–2285
16. Levey AS, Greene T, Kusek JW et al. A simplified equation to predict
glomerular filtration rate from serum creatinine. J Am Soc Nephrol
2000; 11: A0828
17. Abbas NA, John RI, Webb MC et al. Cardiac troponins and renal
function in nondialysis patients with chronic kidney disease. Clin
Chem 2005; 51: 2059–2066
18. Lamb EJ, Kenny C, Abbas NA et al. Cardiac troponin I concentration
is commonly increased in nondialysis patients with CKD: experience
with a sensitive assay. Am J Kidney Dis 2007; 49: 507–516
3553
19. Carr SJ, Bavanandan S, Fentum B et al. Prognostic potential of brain
natriuretic peptide (BNP) in predialysis chronic kidney disease pa-
tients. Clin Sci (Lond) 2005; 109: 75–82
20. DeFilippi CR, Seliger SL, Maynard S et al. Impact of renal disease on
natriuretic peptide testing for diagnosing decompensated heart failure
and predicting mortality. Clin Chem 2007; 53: 1511–1519
21. Panteghini M, Clerico A. Understanding the clinical biochemistry
of N-terminal pro-B-type natriuretic peptide: the prerequisite for its
optimal clinical use. Clin Lab 2004; 50: 325–331
22. Almirez R, Protter AA. Clearance of human brain natriuretic peptide in
rabbits; effect of the kidney, the natriuretic peptide clearance receptor,
and peptidase activity. JPET 1999; 289: 976–980
23. Schou M, Dalsgaard MK, Clemmesen O et al. Kidneys extract BNP
and NT-proBNP in healthy young men. J Appl Physiol 2005; 99:
1676–1680
24. Goetze JP, Jensen G, Moller S et al. BNP and N-terminal proBNP are
both extracted in the normal kidney. Eur J Clin Invest 2006; 36: 8–15
25. Seferian KR, Tamm NN, Seenov AG et al. The brain natriuretic pep-
tide (BNP) precursor is the major immunoreactive form of BNP in
patients with heart failure. Clin Chem 2007; 53: 866–873
26. Ortega O, Rodriguez AD, Gallar P et al. Significance of high
C-reactive protein levels in pre-dialysis patients. Nephrol Dial Trans-
plant 2002; 17: 1105–1109
Downloaded from http://ndt.oxfordjournals.org/ by guest on August 25, 2015
27. Ates K, Yilmaz O, Kutlay S et al. Serum C-reactive protein level is
associated with renal function and it affects echocardiographic car-
diovascular disease in pre-dialysis patients. Nephron Clin Pract 2005;
101: c190–c197
28. Panichi V, Migliori M, De Pietro S et al. C-reactive protein and
interleukin-6 levels are related to renal function in predialytic chronic
renal failure. Nephron 2002; 91: 594–600
29. Mallamaci F, Tripepi G, Cutrupi S et al. Prognostic value of combined
use of biomarkers of inflammation, endothelial dysfunction, and my-
ocardiopathy in patients with ESRD. Kidney Int 2005; 67: 2330–2337
30. Van Der Sande FM, Kooman JP, Leunissen KM. The predictive value
of C-reactive protein in end-stage renal disease: is it clinically signif-
icant? Blood Purif 2006; 24: 335–341
31. Strippoli GF, Bonifati C, Craig M et al. Angiotensin converting en-
zyme inhibitors and angiotensin II receptor antagonists for preventing
the progression of diabetic kidney disease. Cochrane Database Syst
Rev 2006; 4: CD006257
32. Tonelli M, Isles C, Curhan GC et al. Effect of pravastatin on cardio-
vascular events in people with chronic kidney disease. Circulation
2004; 110: 1557–1563.
Received for publication: 30.7.07
Accepted in revised form: 22.5.08