Nephrol Dial Transplant (2003) 18: 1167–1173
DOI: 10.1093/ndt/gfg066
Original Article
White blood cells as a novel mortality predictor in haemodialysis
patients
Donal N. Reddan1,2, Preston S. Klassen1,2, Lynda A. Szczech1,2, Joseph A. Coladonato1,2,
Susan O’Shea3, William F. Owen Jr2,5 and Edmund G. Lowrie2,4
1
Division of Nephrology, Department of Medicine, Duke University Medical Center, Durham, NC, 2Duke Institute for
Renal Outcomes Research and Health Policy, Durham, NC, 3Division of HematologyuOncology, Duke University Medical
Center, Durham, NC, 4Fresenius Medical Care, North America, Lexington, MA and 5Renal Division, Baxter Healthcare,
Gurnee, IL, USA
Abstract
Background. Many conventional cardiovascular risk
factors in the general population are not as predictive
in end-stage renal disease (ESRD). As absolute
neutrophil count and total white blood cell (WBC)
count are associated with adverse cardiovascular
outcomes and all-cause mortality, this analysis was
undertaken to explore the associations of WBC
variables with mortality risk in ESRD.
Methods. Of a total study population of 44 114 ESRD
patients receiving haemodialysis during 1998 at facilities operated by Fresenius Medical Care, North
America, 25 661 patients who underwent differential
white cell count and had complete follow-up were
included. Information on case mix (age, gender, race),
clinical (diabetes, body mass index), and laboratory
variables (haematocrit, albumin, creatinine, potassium,
calcium, phosphorus, bicarbonate, ferritin, transferrin
saturation and differential WBC count) was obtained.
Associations between lymphocyte count, neutrophil
count and demographic and clinical variables were
examined using linear regression. Associations between
WBC variables and survival were estimated using Cox
proportional hazard regression.
Results. A higher lymphocyte count was associated
with higher serum albumin and creatinine, lower age
and black race. High neutrophil count was associated
with lower serum albumin and creatinine, younger age
and white race (all Ps -0.0001). Cox proportional
hazard regression showed an increased lymphocyte
count was associated with reduced mortality risk
[HR 0.86 (0.83–0.89) per 500uml increase in lymphocyte count] and an increased neutrophil count was
associated with increased mortality risk [HR 1.08
(1.06–1.09) per 1000uml increase in neutrophil count].
Correspondence and offprint requests to: Dr Donal Reddan, Duke
Institute of Renal Outcomes Research and Health Policy, Box 3646,
Duke University Medical Center, Durham, NC 27710, USA. Email:
[email protected]
Conclusions. An increased neutrophil count is strongly
associated with, and reduced lymphocyte count associated less strongly with, many surrogates of both
malnutrition and inflammation. An increased neutrophil count and reduced lymphocyte count are independent predictors of increased mortality risk in
haemodialysis patients.
Keywords: end-stage renal disease; haemodialysis;
lymphocytes; mortality; neutrophils
Introduction
Mortality among maintenance dialysis patients is
high, and cardiovascular disease is the major cause of
mortality. Mortality at 1 year for incident end-stage
renal disease (ESRD) patients was 19.5% in 2000, and
;50% of deaths were reported to be due to cardiac
disease [1]. ESRD patients also experience significantly
worse outcomes following myocardial infarction and
following cardiac interventions [2,3]. Many conventional cardiovascular risk factors such as elevated
cholesterol [4] and hypertension [5], which are robust
predictors of adverse cardiovascular outcomes in the
non-ESRD population, do not follow conventional
intuitive mortality relationships in maintenance dialysis patients. Approximately one third of the general
population who have suffered a myocardial infarction
have no conventional risk factors for atherosclerotic
disease [6] and so, in the general population, increased
scrutiny has been placed on non-conventional risk
factors such as prothrombotic lipids, like lipoprotein
(a) or endothelial inflammation, reflected by biomarkers like C-reactive protein (CRP) [7]. Similarly, some
of these non-conventional risk factors have been found
to individually add to the risk profile for cardiovascular events or for overall mortality in the ESRD
1168
population. Examples include CRP [8,9], lipoprotein
(a) [10], homocysteine [11] and oxidative stress [12]. In
the general population, an increased total white blood
cell (WBC) count has been found to correlate with
increased cardiovascular mortality in elderly men [13]
and with increased mortality following myocardial
infarction in general [14]. Increased total WBC counts
[15] and neutrophil count [16] have also been
implicated as a biomarker of atherosclerosis. Alternatively, a reduced lymphocyte count has been
associated with increased mortality in patients with
congestive heart failure [17], and has also been
identified as a bad prognostic sign in patients with
coronary artery disease [18]. An association between
WBC counts and mortality in ESRD has also been
suggested in the past [19]. Because of the limited
correlation between conventional risk factors and
cardiovascular mortality in dialysis patients, and the
similar profile of non-conventional risk factors and
death risks for the general population and dialysis
patients, we proposed that routine haematological
measures like the total WBC count and selected
components of the differential WBC count would
predict death risk for the dialysis population.
Methods
Patient population
Of 44 114 ESRD patients receiving haemodialysis on
1 January 1998 at facilities operated by Fresenius
Medical Care, North America (FMCNA), 25 661
patients who underwent differential WBC count and
had complete follow-up for the year of observation
were included in the analysis. Demographic, clinical
and laboratory data were collected on all patients
during the months of October, November and
December 1997. All laboratories were performed predialysis. For patients who had more than one
determination of the differential WBC count between
October and December 1997, the results of all
determinations during this interval were averaged
for each patient. The number of differential WBC
determinations made from monthly blood samples
ranged between one and three per patient. All the
biochemical analyses were performed by a single
clinical laboratory, LifeChem Clinical Laboratory
(Rockleigh, NJ).
Patients were followed until 31 December 1998.
Patients who left FMCNA facilities or received a renal
transplant were censored from the analyses. Date of
death or censoring was recorded for time-to-event
analysis of all-cause mortality.
D. N. Reddan et al.
count and lymphocyte count. Univariate and multivariable analyses were conducted to examine associations of lymphocyte count and neutrophil count with
demographic and laboratory variables. Differences
among groups in continuous variables were tested by
a two-tailed Student’s t-test. Mean values were presented as mean"standard deviation (SD). Differences
among groups in categorical variables were tested by
chi-square analysis. Stepwise linear regression was
performed to examine the associations of lymphocyte
count and neutrophil count with clinical, demographic
and laboratory variables. The correlation between
neutrophil count with lymphocyte count was also
assessed using a Pearson correlation coefficient. Final
models were generated using the variables noted to be
significant in the stepwise analysis and other variables
thought to be clinically relevant.
Survival was estimated using the Kaplan–Meier
method for patients with and without WBC data and
compared using the log-rank test. Kaplan–Meier
survival curves among patients across quartiles of
lymphocyte count and neutrophil count were compared similarly. The associations of neutrophil count
and lymphocyte count with mortality were subsequently assessed in adjusted and unadjusted models
using Cox proportional hazards regression. Variables
entered into the models included all variables found
to be predictive of lymphocyte count and neutrophil
count in the linear regression models and other
variables thought to be clinically relevant. The regression models were built using both forward addition
and backward elimination stepwise methods (entry
threshold: P-0.05; retention threshold: P-0.1) and
subsequent models were also generated with imputed
values for missing lymphocyte and neutrophil values.
Finally, patients were stratified into a matrix of 81
groups (nine groups of neutrophil count stratified by
nine groups of lymphocyte count), and using Cox
proportional hazards modelling, unadjusted hazard
ratios (HRs) were derived for each subgroup. The
selection of nine subgroups for each variable was for
ease of data presentation. The analysis was then
repeated adjusting for demographic and clinical variables and adjusted HRs derived for each sub-group.
The interaction between neutrophil count and lymphocyte count was tested in both the models with
lymphocyte count and neutrophil count as linear
variables and the model where neutrophil count and
lymphocyte count were categorized into a matrix of
81 groups.
All P-values reported are two-sided, and all confidence intervals (CIs) reported are 95% intervals.
Analyses were performed using SAS software version
8.1 (SAS Institute Inc., Cary, NC).
Measurements and data analysis
Absolute neutrophil count and absolute lymphocyte
count were calculated by multiplying total WBC
quantitated as 109ul of blood by the percentages of
neutrophils and lymphocytes, respectively. Patients
were described overall and by quartile of neutrophil
Results
Of the 25 661 patients included in the analysis, 4087
died by the end of 1998, and 4006 patients were
White cells and mortality in ESRD
1169
A difference in mortality was confirmed using
Kaplan–Meier analysis P-0.001 (data not shown).
The characteristics of the final study population are
described in Table 1. This population had a slightly
higher proportion of black patients (45%) when
compared to the total US ESRD population [1]. The
proportion of males (51%), mean haematocrit (0.33)
and albumin concentrations (38 gul) were all consistent
with those of the US ESRD population [1]. Table 2
portrays differences in patient characteristics across
quartiles of lymphocyte count. Patients with higher
lymphocyte count were more often black, female,
diabetic and younger, and had a greater body mass
index (BMI). Patients with a higher lymphocyte count
had higher serum creatinine and serum albumin
concentrations. Total WBC and neutrophil counts
increased monotonically in parallel with higher lymphocyte count. Table 3 portrays differences in patient
characteristics across quartiles of neutrophil count.
Patients with higher neutrophil count were more often
white, female, diabetic and older. Patients with higher
neutrophil count had lower serum creatinine and
serum albumin concentrations. Higher neutrophil
count was accompanied by a monotonic increase in
ferritin concentration and WBC and lymphocyte
counts.
The results of the stepwise linear regression analyses,
which describe independent predictors of both neutrophil count and lymphocyte count, are presented in
Tables 4 and 5, respectively. Significant independent
predictors of a higher lymphocyte count included
younger age, female gender, presence of diabetes,
increased BMI, higher serum creatinine, lower serum
potassium and higher transferrin saturation (TSAT)
and haematocrit. The total F-value for the linear model
predicting lymphocyte count was 32.7. Significant
independent predictors of higher neutrophil count
included younger age, white race, presence of diabetes,
higher BMI, lower serum creatinine and albumin
Table 1. Characteristics of the final patient population
No. of patients
White (%)
Black (%)
Other (%)
Male (%)
Diabetes mellitus (%)
Age (years)
BMI (kgum2)
Creatinine (mmolul)
Potassium (mmolul)
Phosphorus (mmolul)
Albumin (gul)
Haematocrit (%)
Ferritin (pmolul)
TSAT (%)
WBC ( 3 109ul)
Lymphocyte count ( 3 109ul)
Neutrophil count ( 3 109ul)
25 661
48
45
7
51
46
59.9"15.2
26.7"7.9
849"292
4.9"0.7
1.87"0.52
38"4
0.332"0.33
1191"1148
29.3"12.3
7.0"2.5
138"87
466"202
Values are mean"SD where indicated.
censored from the analysis due to transplantation,
loss to follow-up, or because they moved to another
dialysis unit. In comparison to the 18 453 excluded
because of missing WBC data, the final analytic group
of 25 661 patients were slightly younger (mean age 59.9
and 60.4 years, respectively), had higher serum albumin concentrations (mean albumin 39 and 38 gul,
respectively), and higher serum ferritin concentrations
(mean 1191 and 1045 pmoluml). The final analytical
group of patients was also disproportionately of black
race (45 and 39%, for included and excluded patients,
respectively). There were no other clinically significant differences noted between included and excluded
patients with respect to baseline characteristics. Mortality was significantly higher among patients included
than in those excluded. Over the year of follow-up,
18.9% of uncensored patients included in the analysis
died, in comparison to 15.4% of the uncensored
patients that were excluded because of missing data.
Table 2. Patient characteristics by quartile of lymphocyte count (value
White (%)
Black (%)
Other (%)
Male (%)
DM (%)
Age (years)
BMI (kgum2)
Creatinine (mmolul)
Potassium (mmolul)
Phosphorus (mmolul)
Albumin (gul)
Hct (%)
Ferritin (pmolul)
TSAT (%)
WBC ( 3 109ul)
Neutrophil count ( 3 109ul)
Values are mean"SD where indicated.
a
P-0.0001.
b
Not significant.
9
3 10
ul)
O99
99–129
129–165
P165
56
38
6
55
44
62"14
25"7
786.8"282.9
4.9"0.7
1.84"0.55
37"4
33"4
1234"1135
28.4"12.5
6.0"2.3
445"211
50
43
7
53
47
61"15
26"7.3
839.8"282.9
4.9"0.7
1.87"0.52
38"4
33"3
1193"1112
28.9"11.8
6.6"2.1
456"189
45
48
7
50
47
59"15
27"8.3
875.2"291.7
4.9"0.6
1.87"0.52
39"4
33"3
1200"1070
29.7"12.2
7.1"2.1
469"193
41a
52
7
48a
47a
56"16a
28"8.6a
892.8"300.6a
4.9"0.6a
1.87"0.52a
39"4a
34"3a
1195"1270b
30.1"12.8a
8.2"2.9a
495"211a
1170
D. N. Reddan et al.
9
Table 3. Characteristics by quartile of neutrophil count ( 3 10 ul)
White (%)
Black (%)
Other (%)
Male (%)
Diabetic (%)
Age (years)
BMI (kgum2)
Creatinine (mmolul)
Potassium (mmolul)
Phosphorus (mmolul)
Albumin (gul)
Hct (%)
Ferritin (pmolul)
TSAT (%)
WBC ( 3 109ul)
Lymphocyte count ( 3 109ul)
-331
331–434
435–560
P561
31
63
6
55
35
57"16
26"7
928"318
4.8"0.6
1.84"0.52
39"4
33.4"3.4
1121"1009
32.6"13.3
4.6"1.1
129"65
44
49
7
53
45
60"15
27"8
866"292
4.9"0.6
1.87"0.52
39"4
33.4"3.3
1139"1002
30.0"12.0
6.1"0.8
136"61
54
38
8
49
50
61"15
27"8
822"274
4.9"0.6
1.87"0.52
38"4
33.2"3.2
1193"1029
28.2"11.3
7.3"1.43
142"112
63a
30
7
48a
56a
61"14a
27"9a
769"256a
4.9"0.7a
1.84"0.55b
37"4a
32.9"3.4a
1368"1476a
26.3"11.8a
10.0"2.4a
147"99a
Values are mean"SD where indicated.
a
P-0.0001.
b
P-0.05.
Table 4. Factors predictive of lymphocyte count in linear model
Parameter
Age
Female
Diabetes
BMI (kgum2)
Creatinine (mmolul)
Potassium (mEqul)
PO4 (mmolul)
TSAT (%)
Haematocrit (%)
0.35
8.6
3.48
0.69
0.01
2.574
5.38
0.13
1.56
F-valuea
P-value
63.7
48.0
7.65
82.9
24.2
7.7
16.4
6.25
71.7
-0.0001
-0.0001
0.0057
-0.0001
-0.0001
0.0080
-0.0001
0.0130
-0.0001
Variables tested, but not found to be significant in the model,
included race, serum calcium, albumin, ferritin and bicarbonate
concentrations.
a
F-value Type III sums of squares.
Table 5. Factors predictive of neutrophil count in linear model
Parameter
Age
White
Diabetes
BMI
Creatinine (mmolul)
Albumin (gul)
Potassium (mmolul)
Calcium (mmolul)
Bicarbonate (mEqul)
TSAT (%)
Ferritin (pmolul)
0.30
74.49
30.53
1.12
0.057
7.388
6.71
9.99
9.64
3.51
0.07
F-valuea
P-value
10.9
749.9
136.4
50.7
89.1
440.5
11
37.9
463.5
1063.0
781.0
0.001
-0.0001
-0.0001
-0.0001
-0.0001
-0.0001
0.0009
-0.0001
-0.0001
-0.0001
-0.0001
Variables tested but not found to be significant in the model included
gender, PO4 and haematocrit.
a
F-value Type III sums of squares.
concentrations, higher potassium and calcium concentrations, and lower bicarbonate and transferrin saturation and higher ferritin. For the linear model predictive
of neutrophil count, the total F-value was 327.3. The
Table 6. Final Cox proportional hazards model
Lymphocyte count
(per 50 3 109ul increase)
Neutrophil count
(per 100 3 109ul increase)
Age (per year)
Female
Diabetes
Black
BMI
Creatinine (100 mmolul)
Albumin (10 gul)
Potassium (mEqul)
Calcium (mmolul)
PO4 (mmolul)
Bicarbonate (mEqul)
TSAT (%)
Ferritin (100 pmolul)
HR
CI
P-value
0.859
0.830, 0.888
-0.001
1.078
1.063, 1.093
-0.001
1.023
0.766
1.110
0.892
0.984
0.881
0.404
1.090
2.161
1.494
0.979
0.995
1.011
1.020,
0.715,
1.035,
0.828,
0.979,
0.864,
0.369,
1.034,
1.805,
1.387,
0.967,
0.992,
1.008,
-0.001
-0.001
0.0033
0.0024
-0.001
-0.001
-0.001
0.0014
-0.001
-0.001
0.008
0.0014
-0.001
1.026
0.821
1.191
0.960
0.989
0.897
0.441
1.150
2.585
1.611
0.991
0.998
1.013
All variables selected for inclusion were significant after the stepwise
process.
Pearson correlation coefficient for the correlation of
neutrophil count with lymphocyte count was 0.074.
Table 6 displays the results of the Cox proportional
hazards modelling. All variables tested in the stepwise
process were found to be significant except for race,
and this variable was found to be significant when
added back to the final model. The variables found to
be predictive of mortality included lower lymphocyte
count, higher neutrophil count, older age, male gender,
diabetes, white race, lower BMI, lower serum creatinine and albumin concentrations, higher serum potassium, calcium, and phosphate concentrations, lower
serum bicarbonate concentration, and lower transferrin saturation. Each 50 3 109ul decrement in lymphocyte
count was associated with a 14% increase in mortality,
and each 100 3 109ul increase in neutrophil count was
associated with an 8% increase in mortality. Mortality
White cells and mortality in ESRD
risk was highest for those with a high neutrophil count
and a low lymphocyte count. When the interaction
between neutrophil count and lymphocyte count was
added to the multivariable survival model, it was not
found to be significant (chi-squares1.30, Ps0.25).
The interaction was also not significant when further
tested in the survival models incorporating nine
categories of lymphocyte count and nine categories
of neutrophil count (Ps0.3). When survival models
were generated that also incorporated imputed data
for missing neutrophil and lymphocyte counts, results
were remarkably similar to those reported here (data
not shown).
Figures 1 and 2 describe the relationships between
neutrophil count and lymphocyte count and mortality
in adjusted and unadjusted analyses. Figure 1 illustrates an increase in hazard of death with increasing
Fig. 1. Unadjusted HRs by lymphocyte count and neutrophil count.
Fig. 2. Adjusted HRs by lymphocyte count and neutrophil count.
1171
neutrophil count and decreasing lymphocyte count
in an unadjusted analysis. Figure 2, which is an
adjusted model, illustrates similar relationships, even
after adjustments for the aforementioned mortality
predictors.
Discussion
This analysis demonstrates a novel finding of significant mortality risk associated with selected white cell
components in ESRD patients receiving maintenance
haemodialysis. Higher neutrophil count and lower
lymphocyte count are each independently associated
with increased risk of death. These relationships are
similar to findings in patients with cardiovascular
disease [14,17], and suggest that examination of the
1172
differential WBC count may be useful in assessing
mortality risk profiles in ESRD patients.
A limitation of this data set is that it does not
contain information about the cause of mortality. The
fact that cardiovascular disease is the major cause of
mortality in ESRD suggests that the increased mortality risk is a consequence of cardiovascular disease,
but future evaluation of these relationships among
other data sets are needed to validate or refute this
supposition. This study has a number of additional
limitations germane to observational database
research. The absence of follow-up beyond 1 year is
a limitation. There is a possible bias in that those
patients who had a differential WBC evaluation
performed were demographically different than those
who did not. Although suggested, it does not mean
that their comorbid conditions were different. Moreover, WBC counts are often performed as part of a
routine laboratory panel in dialysis units, and so it is
unclear how much bias by indication has occurred in
this data set. The patients who had WBC counts
performed had a higher mortality than those patients
who did not, which suggests that the strength of the
relationships may be affected in a broader patient
base. However, the finding that the other parameter
estimates of death were in the same direction and
magnitude as had been seen previously suggests that
the leukocyteumortality relationship is likely genuine,
too. Recent results from the Dialysis Outcomes and
Practice Patterns Study (DOPPS) are also consistent
with the observation of higher neutrophil count and
lower lymphocyte count predicting mortality [20]. The
clinical predictive importance of an elevated WBC
count in an individual patient may not be readily
derived from these data and the relevance of such
individual findings are best interpreted at a clinical
level.
The similarity of these findings to those in the
general population with cardiovascular disease is not
surprising in view of the weak associations of cardiovascular disease in ESRD patients with conventional
cardiovascular risk factors. Cardiovascular disease
remains the most important contributor toward
ESRD mortality [1] and just as in the non-ESRD
population a correlation between inflammation and
cardiovascular disease in ESRD has been suggested [9].
The pathobiology of this correlation may involve the
interaction of soluble mediators such as interleukin-6
[21,22], advanced glycation end products (AGEs) and
lipoxidation products [23,24], and oxidative stress [9].
The leukocytes examined herein are likely affected by
an inflammatory process, and so may serve as a
biomarker for it. Data was not captured on other
inflammatory biomarkers like CRP or fibrinogen. Nor
were blood cytokine levels or their antagonists measured directly or from isolated stimulated leukocytes.
So, it is unclear if the strength of these relationships
is affected in models in which other biomarkers or
provocateurs are included. However, these are not
routinely performed laboratory tests, and are less
available as clinical tools to identify patients at risk.
D. N. Reddan et al.
Leukocyte counts are also affected by the patient’s
nutritional status. For example, lymphopaenia is well
described in malnourished patients [25]. So it is
unsurprising that a higher lymphocyte count was
associated with greater anthropometric attributes
and higher serum creatinine values. The relationship
between protein-calorie malnutrition, inflammation,
and mortality is noteworthily complex, especially for
scrutiny by a cross-sectional study design and the
finding that neutrophil count and lymphocyte count
display opposing relationships with mortality suggests
that they may indicate different pathobiologic relationships. For example, we posit that a low lymphocyte
count may reflect protein-calorie malnutrition and
increased death risk secondary to this co-morbid
condition; an increased death risk associated with an
increased neutrophil count may reflect subclinical
infection anduor inflammation and be a consequence
of these processes. The identified predictors of an
increased neutrophil count from the linear regression
model are consistent with its proposed association
with inflammation and poor nutrition in that higher
neutrophil count is associated with lower serum
creatinine, lower serum albumin, lower transferrin
saturation and higher serum ferritin. The predictors of
lymphocyte count selected from an identical variable
pool were strikingly less robust in terms of the overall
model and in terms of the individual predictors
identified. The fact that lower lymphocyte count was
a strong mortality predictor but was not well predicted
by the variables in the models elucidated suggests that
the pathobiology underlining the relationship between
lymphocyte count and mortality remains somewhat
unclear. Although there was a correlation the observation that neutrophil count and lymphocyte count did
not correlate highly with each other also suggests that
they are associated with different disease processes.
Based on the data offered herein, we suggest that
higher neutrophil count and lower lymphocyte count
are independent predictors of mortality risk in
haemodialysis patients, and simple measures to be
followed.
Acknowledgements. D.N.R. was supported by a grant from the
National Kidney Foundation, Inc., with matching funds from the
National Kidney Foundation of North Carolina. W.F.O. Jr
received support for this work through an unrestricted educational
grant from Amgen, Inc. (Thousand Oaks, CA). L.A.S. is
supported by grant DK02724-01A1 from the NIH. P.S.K. received
support from an American Kidney Fund Clinical Scientist in
Nephrology Fellowship.
References
1. U.S. Renal Data System. USRDS 2001 Annual Data Report
National Institutes of Health, National Institute of Diabetes and
Digestive and Kidney Diseases, Bethesda, MD: 2001
2. Herzog CA, Ma JZ, Collins AJ. Poor long-term survival after
acute myocardial infarction among patients on long-term
dialysis. N Engl J Med 1998; 339: 799–805
3. Szczech LA, Reddan DN, Owen WF et al. Differential survival
after coronary revascularization procedures among patients with
renal insufficiency. Kidney Int 2001; 60: 292–299
White cells and mortality in ESRD
4. Pedro-Botet J, Senti M, Rubies-Prat J, Pelegri A, Romero R.
When to treat dyslipidaemia of patients with chronic renal
failure on haemodialysis? A need to define specific guidelines.
Nephrol Dial Transplant 1996; 11: 308–313
5. Zager PG, Nikolic J, Brown RH et al. ‘U’ curve association
of blood pressure and mortality in hemodialysis patients.
Medical Directors of Dialysis Clinic, Inc. Kidney Int 1998;
54: 561–569
6. Steering Committee of the Physicians’ Health Study Research
Group. Final report on the aspirin component of the ongoing
Physicians’ Health Study. N Engl J Med 1989; 321: 129–135
7. Ridker PM, Stampfer MJ, Rifai N. Novel risk factors for
systemic atherosclerosis: a comparison of C-reactive protein,
fibrinogen, homocysteine, lipoprotein(a), and standard cholesterol screening as predictors of peripheral arterial disease. J Am
Med Assoc 2001; 285: 2481–2485
8. Owen WF, Lowrie EG. C-reactive protein as an outcome
predictor for maintenance hemodialysis patients. Kidney Int
1998; 54: 627–636
9. Stenvinkel P, Heimburger O, Paultre F et al. Strong association
between malnutrition, inflammation, and atherosclerosis in
chronic renal failure. Kidney Int 1999; 55: 1899–1911
10. Koda Y, Nishi S, Suzuki M, Hirasawa Y. Lipoprotein(a) is a
predictor for cardiovascular mortality of hemodialysis patients.
Kidney Int 1999; 71 [Suppl]: S251–S253
11. Bostom AG, Lathrop L. Hyperhomocysteinemia in end-stage
renal disease: prevalence, etiology, and potential relationship to
arteriosclerotic outcomes. Kidney Int 1997; 52: 10–20
12. Himmelfarb J, Stenvinkel P, Ikizler TA, Hakim RM. The
elephant in uremia: oxidant stress as a unifying concept of
cardiovascular disease in uremia. Kidney Int 2002; 62: 1524–1538
13. Weijenberg MP, Feskens EJ, Kromhout D. White blood cell
count and the risk of coronary heart disease and all-cause
mortality in elderly men. Arterioscler Thromb Vasc Biol 1996;
16: 499–503
14. Furman MI, Becker RC, Yarzebski J, Savegeau J, Gore JM,
Goldberg RJ. Effect of elevated leukocyte count on in-hospital
mortality following acute myocardial infarction. Am J Cardiol
1996; 78: 945–948
1173
15. Fuster V, Lewis A. Conner Memorial Lecture. Mechanisms
leading to myocardial infarction: insights from studies of
vascular biology. Circulation 1994; 90: 2126–2146
16. Ernst E, Hammerschmidt DE, Bagge U, Matrai A, Dormandy
JA. Leukocytes and the risk of ischemic diseases. J Am Med
Assoc 1987; 257: 2318–2324
17. Ommen SR, Hodge DO, Rodeheffer RJ, McGregor CG,
Thomson SP, Gibbons RJ. Predictive power of the relative
lymphocyte concentration in patients with advanced heart
failure. Circulation 1998; 97: 19–22
18. Thomson SP, Gibbons RJ, Smars PA et al. Incremental value of
the leukocyte differential and the rapid creatine kinase-MB
isoenzyme for the early diagnosis of myocardial infarction. Ann
Intern Med 1995; 122: 335–341
19. Friedman EA. Death on Dialysis: Preventable or Inevitable?
Kluwer Academic Publishers, DordrechtuBostonuLondon: 1994
20. Pifer TB, McCullough KP, Port FK et al. Mortality risk in
hemodialysis patients and changes in nutritional indicators:
DOPPS. Kidney Int 2002; 62: 2238–2245
21. Stenvinkel P, Heimburger O, Jogestrand T. Elevated interleukin6 predicts progressive carotid artery atherosclerosis in dialysis
patients: association with Chlamydia pneumoniae seropositivity.
Am J Kidney Dis 2002; 39: 274–282
22. Pecoits-Filho R, Barany P, Lindholm B, Heimburger O,
Stenvinkel P. Interleukin-6 is an independent predictor of
mortality in patients starting dialysis treatment. Nephrol Dial
Transplant 2002; 17: 1684–1688
23. Miyata T, Maeda K, Kurokawa K, van Ypersele de Strihou C.
Oxidation conspires with glycation to generate noxious advanced
glycation end products in renal failure. Nephrol Dial Transplant
1997; 12: 255–258
24. Sugiyama S, Miyata T, Horie K et al. Advanced glycation endproducts in diabetic nephropathy. Nephrol Dial Transplant 1996;
11 [Suppl 5]: 91–94
25. Allende LM, Corell A, Manzanares J et al. Immunodeficiency
associated with anorexia nervosa is secondary and improves
after refeeding. Immunology 1998; 94: 543–551
Received for publication: 13.5.02
Accepted in revised form: 20.12.02