Nephrol Dial Transplant (2005) 20: 1048–1056

doi:10.1093/ndt/gfh813
Advance Access publication 5 April 2005

Special Feature

Chronic kidney disease as cause of cardiovascular

morbidity and mortality

R. Vanholder1, Z. Massy2, A. Argiles3, G. Spasovski4, F. Verbeke1 and N. Lameire1 for the European
Uremic Toxin Work Group (EUTox)

1
Nephrology Section, Department of Internal Medicine, University Hospital, Gent, Belgium, 2Divisions of Clinical
Pharmacology and Nephrology, University of Picardie and Amiens, University Hospital, Amiens, France,
3
Laboratory of Functional Genomics, UPR 2580 of CNRS, Montpellier, France and 4Department of Nephrology,
Clinical Center, University Hospital, Skopje, Macedonia

Downloaded from http://ndt.oxfordjournals.org/ by guest on June 19, 2015

Abstract replacement therapy are exposed to cardiovascular
To make an evidence-based evaluation of the relation- problems and suffer from accelerated and severe
ship between kidney failure and cardiovascular risk, atheromatosis. More recently, the extent of the
we reviewed the literature obtained from a PubMed problem was re-emphasized by Foley et al. [2].
search using pre-defined keywords related to both According to these authors, already a 5-fold increase
conditions and covering 18 years (1986 until end 2003). in cardiovascular mortality risk is present in dialysis
Eighty-five publications, covering 552 258 subjects, patients older than 75 years, but, more importantly,
are summarized. All but three studies support a link cardiovascular mortality is increased approximately
between kidney dysfunction and cardiovascular risk. 375 times in patients aged between 25 and 35 years [2].
More importantly, the association is observed very In young adults dialysed for several years during
early during the evolution of renal failure: an childhood or adolescence, an exceptionally high prev-
accelerated cardiovascular risk appears at varying alence of coronary calcifications is found for their age
glomerular filtration rate (GFR) cut-off values, which range [3]. In stage 5 chronic kidney disease (CKD) (end-
were
60 ml/min in at least 20 studies. Many studies stage), the overall risk for stroke increases 6-fold [4]. In
lacked a clear definition of cardiovascular disease patients with chronic kidney dysfunction, cardiovascu-
and/or used a single determination of serum creatinine lar disease (CVD) is twice as common as in the general
or GFR as an index of kidney function, which is population and it advances at twice the rate [5]. In an
not necessarily corresponding to well-defined chronic essentially non-dialysed population from the UK with
kidney disease. In six studies, however, chronic kidney a serum creatinine (SCrea)
1.7 mg/dl (
150 mmol/l),
dysfunction and cardiovascular disease were well standardized mortality was increased 2-fold vs the
defined and the results of these confirm the impact general population in those aged >65 years, 12-fold
of kidney dysfunction. It is concluded that there is between the ages of 50 and 64 and even 36-fold between
an undeniable link between kidney dysfunction and the ages of 16 and 49 [6].
cardiovascular risk and that the presence of even Cardiac events in the dialysis population are
subtle kidney dysfunction should be considered as one characterized by a substantial mortality [7–9]. The
of the conditions necessitating intensive prevention of expected remaining life time is shortened by
50% [10].
this cardiovascular risk. Life expectancy at the start of dialysis is comparable to,
if not worse than, that of many cancers at the time of
their diagnosis [11]. Even after successful revascu-
larization for peripheral vascular disease, local recur-
Introduction rence rate is high and early [12].
These observations have an important socio-
In a landmark publication in 1974, Lindner et al. [1] economic impact. In a multivariate analysis in dialysis
first pointed out that patients treated by chronic renal patients, angina and peripheral vascular disease were
among the most prominent reasons for hospitaliza-
Correspondence and offprint requests to: Raymond Vanholder, MD,
tion [13]. By the end of 2001, the dialysis population
PhD, Nephrology Section, 0K12, University Hospital, De Pintelaan worldwide consisted of close to 1 500 000 patients [14],
185, B-9000 Gent, Belgium. Email: raymond.vanholder@ugent.be while it increases by
7% each year [14]. End-stage
ß The Author [2005]. Published by Oxford University Press on behalf of ERA-EDTA. All rights reserved.
For Permissions, please email: journals.permissions@oupjournals.org
CKD as cause of cardiovascular morbidity and mortality 1049
renal disease (ESRD) treatment is expected to cover, The retrieval was limited to studies containing at least
worldwide, 2 million patients in 2005 and 2.5 million in one population with renal failure which was not yet on
2010. Even if these are seemingly small patient groups, dialysis.
the costs of dialysis, especially if indirect (hidden) In order to avoid selection bias, we intentionally opted to
costs related to the cardiovascular complications are include all retrieved studies, such as (I) studies using renal
included, make the socio-economic burden dramatic. function as a continuous variable; (II) studies using sCrea as
More importantly, however, the problem starts long an index of renal function; (III) studies not adjusting for other
before dialysis is needed and this extends the popula- cardiovascular risk factors; (IV) studies with overall mortality
tion at risk
100-fold [15] (see below). rather than cardiovascular mortality as an endpoint. All these
Our present study showed that (i) in the presence characteristics are specified in Tables 1–3.
of an increased SCrea or a decreased glomerular Likewise, no selection of cardiovascular endpoints was
filtration rate (GFR), morbidity and mortality because pursued and studies were included based on all possible
endpoints, such as stroke, transient ischaemic attacks,
of CVD are markedly increased and vascular degrada-
myocardial infarction; coronary heart disease; unstable
tion is accelerated; (ii) the process of cardiovascular
angina; coronary intervention (both surgical or percuta-
damage starts very early during progression in well- neous); admission to the coronary unit; vascular stenosis;
defined CKD, long before the dialysis stage is reached; atheroembolic disease; and cardiac death.
(iii) the link between kidney dysfunction and CVD is an Studies were classified according to whether they had been
important global epidemiological entity with an extent conducted in patients with proven CVD, hypertension or
comparable to the link observed between CVD and chronic heart failure. Studies conducted in the elderly (>65
diabetes mellitus; (iv) apart from traditional cardiovas- years old) were classified separately if patients were selected

Downloaded from http://ndt.oxfordjournals.org/ by guest on June 19, 2015

cular risk factors, non-traditional factors specifically because of this age rather than because of one of the above-
related to kidney failure per se, are very likely to play mentioned other risk factors together with age. A last separate
a causative role; (v) adequate preventive measures set of studies was defined as having been undertaken in
against CVD should be started early during the natural entirely unselected populations. These were still subdivided
history of kidney dysfunction, but their application in two separate subgroups, one in known chronic renal
is hampered by the incorrect interpretation of renal failure patients and one based on one single renal function
function parameters, the delayed identification of determination.
kidney failure and the (too) late referral of patients The primary endpoints of the studies contained in this
with kidney dysfunction. survey are cardiovascular morbidity and mortality as well as
The primary aim of the present study was to offer overall mortality; in studies evaluating overall and cardio-
a broad review of available studies referring to an vascular mortality together, both showed remarkably parallel
enhanced cardiovascular risk in chronic renal failure trends [47,55,59–65]. Those studies specifically focusing on
patients. The second aim was to define at which stage of cardiovascular morbidity or mortality are highlighted in bold
renal dysfunction cardiovascular problems were at first in the corresponding tables (Tables 1–3).
perceived. Studies with as endpoints non-cardiovascular events
or events only partially or indirectly related to CVD [66–79]
and studies in transplant recipients were excluded [80–82],
Subjects and methods although also these studies convincingly pointed to a rela-
tion with kidney dysfunction (17 studies covering 366 720
To appreciate the relationship between changes in parameters subjects). The following were considered as only partially
of kidney function and cardiovascular risk, a literature search or indirectly related to CVD: ventilation time, length of stay
(PubMed) was undertaken with as keywords: ‘cardio-vascular in the coronary care unit, post-operative stay, need for blood
or cardiovascular or vascular or atheromatosis or athero- transfusion, bleeding, overall adverse events, cardiac arrhyth-
sclerosis or atherogenesis or arteriosclerosis or stroke or mia or immediate in-hospital mortality following vascular
outcome or survival or mortality’ on one hand and ‘renal or surgery or intervention, acute coronary syndrome, general or
kidney or uremia or uraemia’ on the other. This search was heart valve surgery or admission to the coronary care unit.
repeated at regular intervals (every 3 months) to update the Studies on the relationship of proteinuria/albuminuria with
database, with 31 December 2003 as final publication date CVD [83] were also considered to be beyond the scope of this
for inclusion. In addition, the reference list of the retrieved survey, although such a relationship is increasingly being
publications was screened in order to add further missing recognized [16].
publications to the database. This approach finally emanated For those studies determining relative mortality risk (RR)
in the retrieval of 85 publications, covering the period in function of GFR, those two factors were correlated with
1986–2003 (18 years in total). each other in a separate analysis. A separated regression
The issue of kidney dysfunction as a cardiovascular risk analysis was undertaken for those studies in which RR was
marker was underscored recently in a comprehensive state- close to 1 and in those with an overt increased mortality. The
ment [16]. Another recent publication was restricted to confluence point of the two regression lines was plotted. Data
reports published in 2003 [17]. Two seminal reviews reported referring to clearance analysis by Cockroft and Gault or by
on, respectively, 19 and 16 publications that are included in the modification of diet in renal disease (MDRD) formula
the present survey [18,19]. The present review extends the and data from studies in which adjustments were made for
previous work by an almost complete retrieval of publications other cardiovascular risk factors were identified by specific
resulting in the identification of 40 additional references symbols and analysed separately. Data from 24 studies were
[6,20–58]. submitted to this analysis.
1050 R. Vanholder et al.
Table 1. Studies in populations selected on the presence of proven CVD: study characteristics

Author Year Selection n Agea F-Ub Cut-off Cut-off Adjuste

criterion clearc SCread

Friedman [32] 1991 STR 492 >59 1.5 120.0 (1.36) Yes
Matts [61] 1993 MI 417 51 >7.0 80.0 (0.90) Yes
Nygard [33] 1997 CHD 587 62 4.6 120.0 (1.36) Yes
Gottlieb [34]* 1998 MI 201752 73 2.0 124.0 (1.40) Yes
Anderson [84] 1999 CABG 3902 64 0.1 132.6 (1.50) No
McCullough [85]** 2000 CCU 9544 63 6.7 81.5f,g Yes
Brooks [101] 2000 CI 3610 62 5.0 NDh NDh Yes
Rubenstein [86] 2000 PCI 3334 68 4.6 132.6 (1.50) No
Beattie [87]** 2001 MI [1724] 63 8.3 63.1f,g Yes
Hemmelgarn [88] 2001 PCA 16989 2.0 203.3 (2.30) Yes
Schlipak [89] 2001 CHD 2763 67 4.1 60.0f 106.1 (1.20) Yes
Szczech [90] 2001 CI 59576 65 3.0 221.0 (2.50) Yes
Ting [35] 2001 PCI 2626 67 2.0 265.2 (3.00) Yes
Asinger [36] 2001 PCI 154 60 2.0 177.0 (2.00) No
Al Suwaidi [60] 2002 CHD 37925 72 0.5 70.0f Yes
Szczech [59] 2002 CHD 3608 63 7.0 132.6 (1.50) Yes
Best [91] 2002 PCI 5327 69 3.0 70.0f Yes
Beddhu [92]*** 2002 PCA 8600 63 8.0 57.0i Yes
Gruberg [93] 2002 PCI 5084 67 1.0 Contf,j 132.6 (1.50)k Yes

Downloaded from http://ndt.oxfordjournals.org/ by guest on June 19, 2015

124.0 (1.40)l
Januzzi [37] 2002 CCU 1537 67 Tranm 75.0f No
Wright [94] 2002 MI 3106 73 5.0 75.0f Yes
Walsh [95] 2002 MI 483 71 1.0 Contf,g,j 132.6 (1.50) Yes
McCullough [53] 2002 CP 817 64 0.1 47.0f,g Yes
Sorensen [99] 2002 MI 6252 69 6.0 55.0f Yes
Shlipak [100]* 2002 MI [130099] 77 1.0 55.0f 132.6 (1.50) Yes
Reis [102] 2002 CHD 784 60 Tranm 80.0f 106.1 (1.20) Yes
Kaplan [57] 2002 MI 2677 64 3.4 120.0 (1.36) Yes
Gruberg [96] 2003 CHD 1265 71 1.0 50.0f Yes
Reinecke [97] 2003 PCI 1049 63 3.2 115.0 (1.30) Yes
Wison [98] 2003 CHD 2503 63 Tranm 76.8f,n Yes
Zebrack [38]*** 2003 PCA [1484] 65 3.0 60.0i Yes
Reddan [39] 2003 CHD 4584 63 5.0 90.0f No
Keeley [40] 2003 CHD 4758 64 7.0 60.0i Yes
Conlon [41] 2003 RAS 1235 4.0 Contj Yes
Naidu [42] 2003 PCI 4602 64 1.0 Increasedo Yes
HPSCGp [43] 2003 OAD 20536 63 5.0 130.0 (1.47)k No
110.0 (1.25)l
Sadeghi [52] 2003 PCI 1933 65 1.0 60.0f Yes
Langston [54] 2003 MI 508 74 1.0 60.0f Yes
Scolari [55] 2003 AERD 95 71 7.0 50.0i Yes
Hillege [56] 2003 MI 298 59 1.0 81.0f Yes
a
Age: mean value, mean of means or medians of different groups, or mean of ranges; lower limit if preceded by ‘>’; bF-U: mean/median
follow-up in years; shortest reported follow-up if preceded by ‘>’; ccut-off clear: highest clearance value [creatinine clearance or equi-
valent of GFR (ml/min)] for which a deterioration in cardiovascular status was observed vs the control (‘normal kidney function’) group;
d
cut-off SCrea: lowest sCrea value [mmol/l (mg/dl)] for which a deterioration in cardiovascular status was observed vs the control (‘normal
kidney function’) group; eadjust: adjustment for comorbid factors; fCockroft–Gault; gnormalized per 72 kg body weight; hND: renal failure
not defined; iMDRD formula; jkidney function parameter(s) used as continuous variable; kmen; lwomen; mtrans-sectional study;
n
normalized per 1.73 m2; oincreased: the cut-off for kidney failure is an ‘increased’ sCrea without further specification; pHeart Protection
Study Collaborative Group [Occlusive Arterial Disease without known diabetes (14 573 subjects) or diabetes mellitus (5963 subjects)].
STR, stroke; MI, myocardial infarction; CHD, coronary heart disease, unstable angina; CCU, admission to the coronary unit; CI,
coronary intervention (both CABG and PCI); CP, chest pain; RAS, renal artery stenosis; OAD, occlusive arterial disease; AERD,
atheroembolic renal disease.
*–***Both studies were undertaken in the same population or database (*Cooperative Cardiovascular Project; **Henry Ford Hospital
Cardiac Intensive Care Unit Database; ***Intermountain Heart Study). Only the study with the highest number of patients is taken into
account for calculating the total number of analysed subjects; the omitted numbers are cited in square brackets.
Studies in bold have a cardiovascular endpoint.

It should be realized that there are a number of limitations studies rather than aiming at an in-depth pathophysiological
to our present approach, such as the unlimited introduction of interpretation of the reported data. Furthermore, the
publications into the database without selection and the approach did not follow the rigid conditions requested for a
descriptive aim pursuing an encyclopaedic listing of related meta-analysis.
CKD as cause of cardiovascular morbidity and mortality 1051

Table 2. Studies in populations selected on the presence of at least one cardiovascular risk factor or of chronic heart failure, or on age:
study characteristics

Author Year Selection n Agea F-Ub Cut-off Cut-off Adjuste

criterion clearc SCread

A. Cardiovascular risk factor

Bulpitt [20] 1986 HT 5451 51 4.3 Contj Yes
Schulman [63] 1989 HT 10940 50 5.0 150.4 (1.70) Yes
Flack [64] 1993 HT 4830 46 6.0 Contj Yes
Rossing [24] 1996 DM 939 39 10.0 Contj Yes
Pahor [25] 1998 HT 4336 72 4.0 101.7 (1.15) No
Mann [105] 2001 DMq 9287 67 4.5 65.0f 124.0 (1.40) Yes
Ruilope [62] 2001 HT 18591 63 3.8 60.0f 132.6 (1.50) Yes
Schillaci [106] 2001 HT 1829 51 4.0 83.0 (0.94)k Yes
70.0 (0.79)l
Wang [31] 2001 HT 1880 67 3.0 80.0 (0.90) Yes
De Leeuw [108] 2002 HT 4688 70 4.1 90.0 (1.02) Yes
B. Chronic heart failure
Packer [21] 1991 CHF 1088 64 1.5 115.0 (1.30) No
Madsen [22] 1994 CHF 190 66 2.1 121.0 (1.37) No
Spinar [23] 1996 CHF 300 58 1.0 Contj Yes
Feenstra [26] 1998 CHF 2246 77 0.4 115.0 (1.30) Yes

Downloaded from http://ndt.oxfordjournals.org/ by guest on June 19, 2015

McAlister [27] 1999 CHF 566 66 1.4 130.0 (1.47) Yes
Opasich [28] 2000 CHF 3327 64 1.0 221.0 (2.50) Yes
Dries [103]* 2000 LVD [5834] 63 2.8 60.0f Yes
Cowie [29] 2000 CHF 220 76 1.3 Contj Yes
Hillege [104] 2000 CHF 1906 65 3.3 76.0f Yes
Al Ahmad [30]* 2001 LVD 6635 60 2.8 75.0i,n 221.0 (2.50) Yes
Mahon [107] 2002 CHF 585 65 2.6 63.9f Yes
Kearney [109] 2002 CHF 553 63 5.0 111.0 (1.26) Yes
McClellan [110] 2002 CHF 646 76 1.0 132.6 (1.50)k Yes
124.0 (1.40)l
C. Aged
Damsgaard [44] 1990 Age 216 67 6.0 Contj Yes
Manolio [111]** 1996 Age [5201] 74 3.3 132.6 (1.50) Yes
Alcorn [45]** 1996 Age [4575] >65 5.0 80.0 (0.90) Yes
Fried [112]** 1998 Age [5201] 73 5.0 106.1 (1.20) Yes
Bursztyn [46] 1999 Age 448 70 6.5 Contj Yes
Schlipak [113]** 2002 Age 5808 75 Tranm 60.0f 132.6 (1.50)k Yes
115.0 (1.30)l
Manjunath [65]** 2003 Age [4893] 73 4.3 90.0i,n Yes
Fried [47]** 2003 Age [5808] 74 7.3 97.3 (1.10) Yes
a–e, i–n
For explanation refer to Table 1; qdiabetes mellitus and/or CVD.
HT, hypertension; DM, diabetes mellitus; CHF, chronic heart failure; LVD, left ventricular dysfunction.
*Both studies were undertaken in the same population or database (SOLVD). **All six studies were undertaken in the same population or
database (Cardiovascular Health Study). Only the study with the highest number of patients is taken into account for calculating the total
number of analysed subjects; the omitted numbers are cited in square brackets.
Studies in bold have a cardiovascular endpoint.

Results studies (20.0%), post-coronary angiography (PCA) in

three studies (7.5%), admission in the coronary care
Studies in populations selected on the unit and coronary intervention [both PCI and coronary
presence of proven CVD artery bypass graft (CABG)], each in two studies
(5.0%), and renal artery stenosis, chest pain, occlusive
Data are presented in Table 1. arterial disease, atheroembolic renal disease, left
In this section, 40 studies are included, representing CABG and stroke, each in one study (2.5%).
425 312 patients [32–43,52–57,59–61,84–102]. Of these Considering these studies together, median age was
studies, 19/40 (47.5%) had overall mortality as an 64 years (range: 51–77 years) and median follow-up
endpoint, compared with 21/40 (52.5%) studies with time was 3 years (range: 0.1–8.3 years). The median
cardiovascular morbidity or mortality as an endpoint. cut-off value of renal function linked to an increased
The cardiovascular conditions used to define this cardiovascular risk was 60.0 ml/min for GFR or one of
population were coronary heart disease in nine studies its surrogates (calculations according to Cockroft and
(22.5%), myocardial infarction in 10 studies (25.0%), Gault or one of the MDRD formulae) (highest value:
percutaneous coronary intervention (PCI) in eight 90.0 ml/min) [39] and 1.47 mg/dl (131.3 mmol/l) for
1052 R. Vanholder et al.
Table 3. Studies in entirely unselected populations: study characteristics

Author Year Selection n Agea F-Ub Cut-off Cut-off Adjuste

criterion clearc SCread

A. Studies in known chronic renal failure

Jungers [48]r 1997 – 147 65 5.4 50.0f,n Yes
Jungers [49]r 1998 – 980 63 Nors 50.0f,n No
Levin [50]r 2001 – 313 56 Tranm 75.0f No
Landray [120]r 2001 – 369 62 Tranm 44.0f 130.0 (1.47) No
Sarnak [121]r 2002 – 1342 50 Tranm Contj,t No
Drey [6] 2003 – 1076 5.5 300.9 (3.40) Yes
B. Studies based on a single determination of renal function
Wannamethee [114] 1997 – 7690 50 14.7 116.0 (1.31) Yes
Culleton [115] 1999 – 2837 54 11.0 132.6 (1.50)u Yes
Garg [116] 2002 – 2352 16.1 122.0 (1.38)k Yesv
104.0 (1.18)l
Muntner [117] 2002 – 6354 55 14.0 90.0i Yes
Henry [118] 2002 – 631 64 8.7 72.2i,n Yes
Manjunath [119]* 2003 – 15350 56 6.2 90.0i,n Yes
Abramson [51]* 2003 – [13716] 54 Nors 60.0f Yes
Jurkovitz [58]* 2003 – [13329] 54 Nors 132.6 (1.50)k Yes
106.1 (1.20)l

Downloaded from http://ndt.oxfordjournals.org/ by guest on June 19, 2015

a–e, i–n
For explanation refer to Table 1; rthese studies contained no real control group, but incidence of CVD was strikingly higher than in
the general population; srisk rates normalized per number of observation years; tGFR; usignificant relationship with kidney failure only in
men; vsignificant relation disappeared after adjustment for traditional risk factors.
*All three studies were undertaken in the same population or database (ARIC). Only the study with the highest number of patients is taken
into account for calculating the total number of analysed subjects; the omitted numbers are cited in square brackets.
Studies in bold have a cardiovascular endpoint.

SCrea [lowest value: 0.90 mg/dl (80 mmol/l); values Studies in populations selected on the presence of
in men and women combined] [61]. The median of at least one traditional cardiovascular risk factor,
the values linked to an increased cardiovascular risk including old age, or of chronic heart failure
for the MDRD formula (57 ml/min) was slightly
Data are presented in Table 2.
lower than for the Cockroft and Gault formula
(65.5 ml/min), reflecting the divergent clinical and In this section, 31 studies are included, representing
mathematical bases for both formulae. The results of 87 505 patients [20–31,44–47,62–65,103–113].
34 studies (85.0%) were adjusted for other cardiovas- Table 2 is subdivided into three sections: (A) patients
cular risk factors. selected because of the presence of at least one
The first publication for this specific subgroup was cardiovascular risk factor (either hypertension or
published in 1991, when Friedman et al. [32] described diabetes mellitus) (10 studies in 62 771 subjects;
that survival was more compromised with a higher respectively, two and eight studies with diabetes and
SCrea, in 492 elderly subjects post-stroke. hypertension as determinant); (B) patients selected
The study by Shlipak et al. [89] published in 2001 and because of the presence of chronic heart failure or left
the one by Reis et al. [102] were restricted to women ventricular dysfunction (13 studies in 18 262 subjects);
and showed the same trend as studies covering both and (C) patients selected because of old age (eight
genders. studies in 6472 subjects). Aged patients were confined
The study by Szczech et al. [59] as well as the study to this part of the survey (corresponding to section C
conducted by the Heart Protection Study Collabora- of Table 2) if no additional selection criterion was
tive Group [43] evaluated patients with and without present. In Table 1 and sections A and B of Table 2,
diabetes mellitus separately, but showed parameters of patient groups of similar old age are considered, but
kidney failure to be a discriminating factor in both in these studies, patients were selected in addition
arms. because of the presence of at least one vessel lesion, one
This section contains five study populations cardiovascular risk factor or chronic heart failure
including more than 20 000 subjects, all convincingly [22,25–27,29,31,34,35,37,54,55,60,86,89,91,93–96,99,
underscoring a link of signs of enhanced cardio- 100,105,108,110].
vascular damage with parameters corresponding to Of the 31 studies contained in this section, 16/31
a decrease of renal function [34,43,60,90,100]. Two (51.6%) had overall mortality as an endpoint, com-
of these studies contained more than 100 000 subjects, pared with 15/31 (48.4%) studies with cardiovascular
but were conducted in the same study population morbidity or mortality as an endpoint. Considering
[34,100]. these studies together, median age was 66.0 years
CKD as cause of cardiovascular morbidity and mortality 1053
(range: 39–77 years) and median follow-up time was 3.9 gender-related lack of significance might, in part,
years (range: 0.4–10.0 years). The median cut-off value be attributed to the small number of patients with
of renal function for showing increased cardiovascular renal dysfunction in the gender strata (n ¼ 270 for
risk was 63.9 ml/min for GFR or one of its sur- women).
rogates (highest value: 90.0 ml/min) [65] and 1.3 mg/dl Garg et al. [116] undertook an analysis of the
(115 mmol/l) for SCrea [lowest value overall: 0.9 mg/dl data from the first National Health and Nutrition
(80 mmol/l); in women: 0.79 mg/dl (70 mmol/l)] Examination Study (NHANES I) in 2352 adults. The
[31,45,106]. The results of 28 studies (90.3%) were hazard ratio related to moderate indices of kidney
adjusted for other cardiovascular risk factors. failure lost significance after correction for traditional
To the best of our knowledge, the very first study to risk factors. This is the only study included in the
describe a relationship between parameters of kidney present survey, where a loss of significance after
function and CVD is included in this section and was adjustment is observed.
published by Bulpitt et al. in 1986 [20]. When in more recent studies GFR was used as a
The largest population in this section was reported renal function index, a relation of cardiovascular risk
by Ruilope et al. (n ¼ 18 790) [62] in a reanalysis of the with kidney failure parameters was corroborated each
Hypertension Optimal Treatment (HOT) Study. time [117–119].
In the study by Opasich et al. [28], patients Also in the study by Abramson et al. [51], GFR-
with chronic congestive heart failure and a SCrea defined chronic renal failure resulted in a higher
>2.5 mg/dl (>220 mmol/l) had a markedly higher incidence of stroke, but most strikingly if combined
mortality. This is the only study in this section where with anaemia. Similar results were published, based on

Downloaded from http://ndt.oxfordjournals.org/ by guest on June 19, 2015

the difference in mortality was not significant, possibly the same population, for risk of coronary events [58].
because of the low number of patients with parameters Of note, in at least four other studies, however,
corresponding to kidney failure (2.8% of the popula- adjustment for anaemia did not neutralize the statistical
tion; n ¼ 93). weight of renal failure [47,54,65,85].

Studies in unselected populations

Overall summary of the data
Data are presented in Table 3.
In this section, 14 studies are included, representing a Overall, 552 258 subjects were considered in 85 studies.
total of 39 441 patients [6,48–51,58,114–121]. Of note, All studies based on larger population sizes (>5000
in all studies reported in Tables 1 and 2 and in seven of subjects) [20,30,34,43,47,51,58,60,62,63,85,88,90–93,
the 14 studies in Table 3, the enrolment in different 99, 100,103,105,111–114,117,119] and/or based on ade-
strata is based on a single determination of SCrea, GFR quate parameters of kidney function (GFR) [20,23,
or surrogate of GFR. These levels do not necessarily 24,29,30,37–41,44,46,48–56,60,62,64,65,85,87,91,92,94,
correspond to different degrees of severity of chronic 96,98–100,103,104,107,113,117–119,121] showed a sig-
kidney dysfunction, as they might have been influenced nificant relationship between the markers of renal
by instant factors, such as (de)hydratation, drug intake dysfunction and overall mortality and/or cardiovas-
(non-steroidal anti-inflammatory drugs or angiotensin- cular morbidity or mortality.
converting enzyme inhibitors), recent radio-contrast One of the most relevant findings of this analysis is
investigations, etc. The seven studies of this type in that the increase in cardiovascular risk occurs very
Table 3 are classified under ‘kidney dysfunction not early during the evolution of chronic renal failure
defined’ (35 214 subjects). Six studies covering 4227 (Figure 1), irrespective of the method of determination
subjects are based on defined CKD and are considered of renal function.
separately [6,48–50,120,121]. In 40 studies (Table 1), subjects were selected because
Considering these studies together, the median age of the presence of at least one cardiovascular abnormal-
of the patients was 55.5 years (range: 50–65 years) ity, whereas in 31 other studies (Table 2), subjects were
(i.e. markedly younger than in the previous sections selected because of the presence of at least one
A and B) and the median follow-up time was 9.85 years traditional risk factor, including old age, or chronic
(range: 5.4–16.1 years) (i.e. markedly longer than in heart failure. One might argue that this type of selection
the previous sections). The median cut-off value of does not relate to kidney dysfunction per se. Of note,
renal function linked to increased cardiovascular risk however, most of the patients attending nephrology
was 72.2 ml/min for GFR or one of its surrogates clinics have similar comorbidities. In addition, results
(highest value: 90.0 ml/min) [117,119] and 1.42 mg/dl were similar in 14 studies in unselected populations
(126.0 mmol/l) for SCrea [lowest value overall: (Table 3). In these populations, the hidden presence
1.31 mg/dl (116 mmol/l); in women: 1.18 mg/dl of the above-mentioned risk factors is not excluded,
(104 mmol/l)] [114,116]. The results of 10 studies however. The results in these low cardiovascular
(71.4%) were adjusted for other cardiovascular risk risk populations with well-defined CKD point
factors. in the same direction of enhanced risk related to
In the study by Culleton et al. [115], an impact kidney dysfunction as the studies in the remaining
of mild changes in renal function parameters on populations. However, the link is somewhat less
overall mortality was found only in men. This partial straightforward.
1054 R. Vanholder et al.
15 15
overall, data from 74 studies were available. Among
those, 26% of subjects in total conformed with preset
parameters corresponding to kidney dysfunction,
defined as the stratum or the sum of strata character-
RR_mortality (fold increase)

10 10
ized by a higher sCrea and/or a lower GFR than the
reference (control) group. If we accept that the vast
majority of these study populations are recruited
5 5
among subjects with at least one known cardiovascular
problem [Tables 1 and 2 (section B)] or cardiovascular
risk factor (Table 2, sections A and C), it might be
hypothesized from the above that approximately one-
0 0 quarter of these populations additionally suffer from a
0 2 50 75 10 125 15 certain degree of renal dysfunction which only adds to
GFR (ml/minute) their cardiovascular risk.
Fig. 1. Relative risk of mortality (RR_mortality) vs GFR. Twenty- Of note, 71 of the 85 studies under consideration
four studies were included in this analysis. The points plotted in were adjusted for other cardiovascular risk factors. In
the graph correspond to the values available in the publications only one study was the significant weight of renal
included in the present review. The majority of studies used
Cockroft and Gault formulae to estimate clearance (open triangles). failure lost [116].
Studies using the MDRD formula are indicated by the closed
squares and unadjusted studies by the crosses. Two different parts

Downloaded from http://ndt.oxfordjournals.org/ by guest on June 19, 2015

of the graph can be identified clearly: one constituted of those
points related to an increased mortality (which are associated with Discussion
a lower renal function) and those with a RR of mortality close to 1
(which are associated with a nearly normal renal function). Both
clusters follow a linear distribution. The linear regression has been Our present study showed that (i) in the presence of an
calculated separately for these two areas of the graph and the increased SCrea or a decreased GFR, morbidity and
regression lines plotted [respectively: y ¼ (0.1262x) þ 10.77, mortality because of CVD are markedly increased and
r ¼ 0.645, P < 0.001; y ¼ (0.1018x) þ 2.727, r ¼ 0.574, P < 0.004].
The confluence point of the obtained regression lines represents the vascular degradation is accelerated; (ii) the process
level at which a change in the slope is observed: to a given decrease of cardiovascular damage starts very early during
in GFR corresponds a higher increase in RR mortality. This progression in well-defined CKD, long before the
threshold point corresponds to a GFR of 75 ml/min, supporting the dialysis stage is reached; (iii) the link between kidney
results reported in the publications analysed and indicating that an dysfunction and CVD is an important global epide-
intensive follow-up and prevention of CVD should be aimed at very
early when renal function decays. Recalculation of the threshold miological entity with an extent comparable to the link
points with only studies based on Cockroft and Gault, or only with observed between CVD and diabetes mellitus; (iv) apart
adjusted studies, also resulted in values between 74 and 76 ml/min. from traditional cardiovascular risk factors, non-
traditional factors specifically related to kidney failure
per se, are very likely to play a causative role; and
(v) adequate preventive measures against CVD should
It is of note that studies from all over the world came be started early during the natural history of kidney
to the same conclusion. In addition to Europe and dysfunction, but their application is hampered by the
Northern America, which are the most common incorrect interpretation of renal function parameters,
geographical areas where this type of study is the delayed identification of kidney failure and the (too)
conducted, reports from China [31], Israel [46] and late referral of patients with kidney dysfunction.
New Zealand [32] are also included. Hence, the relation
between kidney dysfunction and cardiovascular risk
seems to be universal.
Some of the enclosed studies were not undertaken Methodological considerations regarding
primarily to evaluate the connection between kidney the analysed literature
dysfunction and cardiovascular problems, but were
e.g. evaluations of drug effects, which amongst A substantial proportion of the cited studies are post-
other elements contained a comparison between hoc analyses of interventional trials or observational
patients with and without kidney dysfunction [20–22, studies. In this section, methodological aspects related
26,28,33,34,37,43]. Also, these studies came to the same to study design and choice of outcome variables are
conclusions as those specifically considering the rela- discussed. These more likely obscured or weakened
tionship of kidney dysfunction and cardiovascular risk. rather than amplified the association of impaired renal
It might be argued that even if in total more than function with CVD.
550 000 subjects were considered in the 85 publications Retrospective analyses of the effect of a variable
reported, the fraction of patients with kidney dysfunc- (in this case a renal function parameter) on the
tion contained in these studies might be much smaller, endpoints are generally underpowered, because these
if not irrelevant. For that reason, the specific number studies were not designed to detect such effects. Since
of subjects with parameters corresponding to kidney renal failure per se enhances certain traditional risk
dysfunction was also considered. For this subanalysis, factors, adjusted risks or hazard ratios may overcorrect
CKD as cause of cardiovascular morbidity and mortality 1055
for these factors and equally will underestimate a true corroborates the impact of this pathophysiological
relationship. association.
In most studies, a single determination of a surrogate
marker for GFR is studied. A well-known problem
with single determinations is regression dilution bias,
a statistical phenomenon whereby the extreme values of Practical epidemiological considerations
high and low GFR (or surrogates) are more likely to
over- or underestimate the real values; thus, once more After it had been established that dialysis patients are
attenuating the relationship of increasing cardiovas- prone to accelerated atheromatosis [1], it has been
cular risk with decreasing renal function. The frequently suggested more recently that the cardiovascular system
used surrogate markers for GFR, such as sCrea, may becomes affected in renal failure well before the stage
be increased by multiple instant factors, whereas an of dialysis is reached [122]. This suggestion is strongly
artefactual decrease is rare and in essence only related corroborated by this present survey. The smallest
to spurious cases of fluid overload. This unequal impairments in renal function associated with a
distribution of errors may increase the scatter in the significant increase in cardiovascular risk are a GFR
relationship with cardiovascular risk. Attempts to of 90 ml/min/1.73 m2 [39,65,117,119] and a sCrea
overcome this problem by calculating GFR causes the of 0.90 mg/dl (80 mmol/l) (Figure 1) [31,45,61]. The
difficulty that known cardiovascular risk factors, such median cut-off GFR in each of Tables 1–3 was situated
as age, weight and gender, are included a second time at
60 ml/min/1.73 m2, which corresponds to a loss of
as predictor variables, so that the use of classical kidney function by a maximum 50%.

Downloaded from http://ndt.oxfordjournals.org/ by guest on June 19, 2015

multivariate methods probably overcorrects for these This problem of cardiovascular risk associated with
factors. Categorization (renal failure or not) of a kidney dysfunction, hence, affects a larger section of
continuous variable (GFR) generally further decreases the general population than until recently suspected
the power to detect a relationship with the outcome (Table 4). Based on recent data collected in the United
variable. Additionally, most studies a priori assume a States, 4.7% of the overall population suffers from
linear relationship between GFR and CVD risk. The a deterioration of GFR by
50% [123]. The vast
lack of exploration of non-linear relationships and majority of those (92%) has a GFR between 60 and
interactions between chronic renal failure and other 30 ml/min (CKD stage 3), of whom >90% are non-
factors may result in a larger scatter (and weaker diabetics [123]. One should be careful with extrapolat-
relationship) of the data when fitting the statistical ing these data from the USA to the rest of the world,
model. but, unfortunately, to the best of our knowledge,
It is known that creatinine values vary across clinical no data from outside the USA are available as yet.
laboratories and that all estimates of GFR from However, even if the number of patients at risk would
creatinine equations (Cockroft and Gault or MDRD be substantially lower outside the USA, which is rather
formula) result in biased estimates if the sCrea assay unlikely, the group size of potentially affected individ-
used in a specific study is calibrated differently from uals seems to be not much different from those
that in the study which validated the method for attributed to diabetes mellitus [124,125]. In spite of
estimating GFR. Systematic bias is relatively unim- this, the cardiovascular risk associated with kidney
portant in investigations of the relationship between dysfunction is not always recognized as such by society
CVD and renal function within a study, but hampers and even not by the nephrological community.
comparison across studies and limits the interpretation Extending this problem to a population which is
of specific GFR cut-off points. much larger than the group currently treated by dialysis
In addition, it should be realized that, apart from also extends the socio-economic implications (Table 4)
these drawbacks, sCrea is only a crude index of renal and underscores the urgent need for intervention to
function and that GFR values, even if calculated, are detect kidney dysfunction and to intensify the manage-
more valid. For that reason, the method of renal ment of CVD when it is present.
function estimation was clearly specified in the tables. Most patients with kidney dysfunction also bear
Studies based on sCrea were not excluded, however, to one or more traditional cardiovascular risk factors and
avoid selection bias (see ‘Subjects and methods’). a major pathophysiological role might be attributed
Studies considering mortality instead of cardiovas- to these factors [126,127]. They undeniably have a
cular morbidity as the endpoint may also lose power negative clinical impact and should, if possible, be
to detect a relationship, since mortality rates are much treated appropriately. In several correlation analyses,
lower than cardiovascular incidence rates. Moreover, however, the statistical weight of traditional risk factors
when all-cause mortality is considered rather than was weak to absent [50,105,128,129]. While submitting
cardiovascular mortality, the relationship may become the uraemia population to a Framingham risk analysis,
blurred by non-cardiovascular deaths. the real observed risk is far greater than the calculated
Altogether, multiple aspects related to methodologi- one based on the traditional risk factors
cal design may influence the results in a way that they [121,128,130,131], which hardly differs from the risk
tend to weaken rather than strengthen the relationship in the general population [121,128,131]. Statistical
between renal dysfunction and CVD. That such a adjustment for a host of comorbid factors in 71 studies
strong relationship is nevertheless found only further does not neutralize the relation of cardiovascular
1056 R. Vanholder et al.
a
Table 4. Potential subjects at risk CVD was confirmed. Confirmation was obtained from
85 studies covering more than 550 000 subjects, clearly
Worldwide USA Europe showing this relationship, independent of geographic or
ethnic factors. Differences in cardiovascular risk are
Overall population 6 000 000 000 389 000 000 800 000 000 already present at the very early phase of renal
GFR<60 ml/min (4.7%)b 282 000 000 18 283 000 37 600 000 impairment; those can only be detected by sensitive
c
Stage 5 (ESRD ) 1 750 000 335 000 350 000
methods of renal function assessment (GFR), which are
a
Approximative figures; bbased on epidemiological data collected in therefore necessary and fully justified.
the USA [123]; cbased on worldwide data for 2001 [14] – ESRD The epidemiological features observed bring this
corresponds to the need for renal replacement therapy (according to problem to the highest level of importance, along
the current classifications: stage 5).
with other major conditions, such as diabetes mellitus
and hypertension.
Much remains to be done, however, to improve the
risk with renal failure [6,20,23,24,26–35,38,40–42, prevention of CVD associated with renal failure. The
44–48,51–65,85,87–115,117–119]. main aspects, with a predictable positive impact, that
The risk factors which were most frequently require our particular attention include:
adjusted were age (69/71 studies), gender (56), diabetes
mellitus (55), vascular disease (52), blood pressure (49), a. Sensibilization regarding the problem and improve-
left ventricular dysfunction (49), medication (39), ment of the awareness about its existence.
smoking (36) and lipid status (32). Therefore, it is b. Inclusion in the follow-up of patients with cardio-
vascular risk of sensitive parameters assessing renal

Downloaded from http://ndt.oxfordjournals.org/ by guest on June 19, 2015

likely that non-traditional risk factors (as yet identified
as well as non-identified) are at play as well. function and in patients with kidney dysfunction of
Prominent candidates in such a role are uraemic parameters of cardiovascular risk.
retention solutes [128], which accumulate because the c. Improvement of the interpretation of parameters of
kidneys fail to eliminate them. Their accumulation in kidney (dys)function.
kidney failure might impose a mechanistically similar d. Sensibilization of the general population and
damaging effect as the one provoked by glycation general practitioners and improvement of collabora-
products in diabetes [132]. Interestingly, several sup- tion between all healthcare providers to facilitate a
portive studies suggest a relation between uraemic timely referral of these patients to the nephrologist.
retention solutes and CVD: oxidative stress products e. Application of the rules of secondary prevention in
[133–137]; pro-coagulant factors [48,135,138]; inter- the presence of even moderate renal dysfunction.
leukin-6 [139]; phosphate and calcium–phosphate
product [140–143]; asymmetric dimethylarginine
[144,145]; parathyroid hormone [146,147]; neuro- Acknowledgements. The European Uremic Toxin Work Group
peptide Y [148]; homocysteine [48,149–154]. Also, (EUTox) is a group of European researchers involved in studies
sympathetic overactivity plays a prominent patho- and reviews related to uraemic toxicity. The group was created
under the auspices of the European Society for Artificial Organs
physiological role [155]. (ESAO) and its members are A. Argiles, Montpellier, France;
However, at present, the evaluation of the factors J. Beige, Leipzig, Germany; P. Brunet, Marseille, France; G. Cohen,
responsible for increased CVD is far from complete Vienna, Austria; P.P. De Deyn, Antwerp, Belgium; B. Descamps-
and has been concentrated on those factors that had Latscha, Paris, France; S. Herget-Rosenthal, Essen, Germany;
already been incriminated in the general population. W. Hörl, Vienna, Austria; J. Jankowski, Berlin, Germany; A. Jörres,
Berlin, Germany; Z.A. Massy, Amiens, France; M. Rodriguez,
It seems very much appropriate to extend our studies Cordoba, Spain; G. Spasovski, Skopje, Macedonia; B. Stegmayr,
to those factors specific for renal failure. Recently Umeå, Sweden; P. Stenvinkel, Stockholm, Sweden; R. Vanholder,
identified but until then unsuspected compounds, such Gent, Belgium; C. Wanner, Würzburg, Germany; E. Weissinger,
as phenylacetic acid, indoxyl sulphate, guanidine, Hannover, Germany; A. Wiecek, Katowice, Poland; W. Zidek,
Berlin, Germany. The actions of this group are endorsed by the
methylguanidine, guanidinosuccinic acid, guanidino- following companies: Amgen, Baxter Healthcare, Fresenius Medical
acetic acid and p-cresol [156–158], underscore that Care, Gambro, Genzyme, Membrana and Roche.
many responsible compounds might have, as yet, gone
unrecognized. Therefore, a more systematic evaluation Conflict of interest statement. None declared.
and classification of known uraemic solutes [159]
should be undertaken, to allow the development of
more directed therapeutic approaches.
References
This review was conducted to make an objective
evaluation in how many available studies in the The references to this paper can be found as supplementary data at
literature the link between kidney dysfunction and NDT online.

Received for publication: 6.9.04

Accepted in revised form: 9.3.05