J Am Soc Nephrol 15: 3154–3165, 2004

Hemoglobin Targets for the Anemia of Chronic Kidney
Disease: A Meta-analysis of Randomized, Controlled Trials
GIOVANNI F.M. STRIPPOLI,*†‡ JONATHAN C. CRAIG,*‡ CARLO MANNO,† and
FRANCESCO P. SCHENA†
*Cochrane Renal Group, NHMRC Centre for Clinical Research Excellence in Renal Medicine, †The
Children’s Hospital at Westmead, University of Sydney, Australia; Department of Emergency and Organ
Transplantation, Section of Nephrology, University of Bari, Bari, Italy; and ‡School of Public Health,
University of Sydney, Australia

Abstract. Anemia affects almost all patients with chronic kid- with 95% confidence intervals (CI). Nineteen relevant trials
ney disease (CKD), reduces quality of life, and is a risk factor were identified. Twelve trials (638 patients) compared use of
for early death. Higher hemoglobin (Hb) targets have been erythropoietin versus no erythropoietin treatment, and seven
widely advocated because of data from observational studies trials (2058 patients) compared higher versus lower Hb targets.
showing that higher Hb is associated with improved survival Compared with Hb values of 130 g/L or more in the CKD
and quality of life, but higher Hb targets may cause access population with cardiovascular disease, Hb values of 120 g/L
thrombosis and hypertension and are costly. This study aimed were associated with lower all-cause mortality (RR, 0.84; 95%
to evaluate the benefits and harms of different Hb targets in CI ,0.71 to 1.00). Hb values of 100 g/L or less reduced the risk
CKD on the basis of randomized trial evidence. A comprehen- of hypertension (RR, 0.50; 95% CI, 0.33 to 0.76) but increased
sive search of the Cochrane Trials Registry, Medline, Embase, the risk of seizures (RR, 5.25; 95% CI, 1.13 to 24.34). From
and reference lists was performed. Two independent reviewers the available trial evidence, in CKD patients with
assessed studies for inclusion criteria and extracted data on cardiovascular disease, the benefits associated with higher Hb
all-cause mortality, cardiovascular disease, strokes, hyperten- targets (reduced seizures) are outweighed by the harms
sion, seizures, hyperkalemia, access thrombosis, and quality of (increased risk of hy- pertension and death). There is
life. Analysis was by a random-effects model, and results are insufficient data to guide de- cisions in patients without
expressed as relative risk (RR) or weighted mean difference cardiovascular disease or in the predialysis population.

Anemia is a common complication of chronic kidney disease ular dilation, and myocardial ischemia, which are risk factors
(CKD). The prevalence of anemia varies with the degree of for cardiovascular disease and death (5,6). It is plausible that
renal impairment in predialysis patients with CKD, but once reversing anemia may reduce this risk (7–10).
end-stage kidney failure occurs, all patients are eventually Current treatment options for anemia include blood transfu-
affected (1–3). Anemia develops once renal function decreases sion, recombinant human EPO ( or ) or darbepoetin . The
to 50% because of a deficiency in endogenous erythropoietin latter are growth factors for the bone marrow erythroid precur-
(EPO) production by the kidney, decreased red cell survival, sors, which stimulate production of erythrocytes (11,12).
blood losses, and increased red blood cell destruction once the Treatment of anemia improves muscle strength and function,
patient begins dialysis treatment, particularly hemodialysis (4). cardiac function, and cognitive and brain electrophysiologic
Anemia reduces physical capacity, well-being, neurocogni- function because of improved peripheral oxygen supply
tive function, and energy level and worsens quality of life both (10,13). Aspects of the patient’s quality of life such as fatigue,
in predialysis and dialysis patients (5). Anemia also induces depression, and relationships with others are better at higher
adaptive cardiovascular mechanisms to maintain tissue oxygen hemoglobin (Hb) levels (5,10,14), but higher Hb levels may
supply. This leads to left ventricular hypertrophy, left ventric- also lead to an increased risk of arteriovenous fistula throm-
bosis, hypertension, and an increased number of adverse car-
diovascular events. Also, reaching and maintaining higher Hb
Received April 14, 2004. Accepted August 31, 2004. targets implies major costs (8,15–18). Very recently, a number
Correspondence to Dr. Giovanni F.M. Strippoli, Centre for Kidney Research, of trials in cancer patients were terminated prematurely be-
NHMRC Centre for Clinical Research Excellence in renal medicine, Cochrane
cause of unexpectedly higher mortality in the higher Hb
Renal Group, Locked Bag 4001, The Children’s Hospital at Westmead and the
University of Sydney, Westmead, NSW 2145, Australia. Phone: 02-98453088; Fax: groups, mainly as a result of disease progression and throm-
02-98453038; E-mail: gfmstrippoli@katamail.com and GiovannS@chw.edu.au boembolic events (19).
1046-6673/1512-3154 There has been a consistent trend toward higher Hb levels in
Journal of the American Society of dialysis patients of all ages, genders, and race categories
Nephrology during the past decade (2). Guidelines have been developed for
Copyright © 2004 by the American Society of
Nephrology
Hb targets, but there remains considerable debate regarding
the
DOI:
10.1097/01.ASN.0000145436.09176.A7
J Am Soc Nephrol 15: 3154–3165, 2004 Hemoglobin Targets for the Anemia of Chronic Kidney Disease 3155

appropriate Hb levels as shown by the wide variation that still methodologic characteristics of the trials, and outcomes: all-cause
exists in anemia management practices between and within mortality, left ventricular hypertrophy, cardiovascular mortality (myo-
countries (Table 1). The aim of this systematic review is to cardial infarction, stroke), adverse events (patients with hypertension
summarize the benefits and harms of lower versus higher Hb requiring exclusion from study or administration of additional anti-
targets in the treatment of the anemia of CKD using existing hypertensive medication, seizures, access thrombosis, hyperkalemia),
randomized, controlled trial data. hospitalization rates and days of hospitalizations, renal function, BP,
and quality of life.
Quality of the trials was assessed using standard criteria (allocation
Materials and Methods concealment, blinding, analysis by intention to treat, and complete-
Inclusion Criteria ness of follow-up) (21). Any differences in data extraction were
We included any randomized, controlled trial that was of at least 2 resolved by discussion among authors. When data were missing or
mo duration and assessed the effects of different Hb targets in pre- or incomplete, the authors of the trial were contacted for clarification.
postdialysis patients with anemia of CKD. The higher Hb target could
be achieved by EPO ( or ) or darbepoetin or blood transfusion;
the lower target could be achieved by lower doses of the same drugs Statistical Analyses
or by a placebo or no treatment or blood transfusion. Dichotomous outcome data from individual trials were analyzed
using the relative risk (RR) measure and its 95% confidence intervals
(CI). The analysis was based on published event rates and took no
Search Strategy
account of the risk estimates provided by the trials, which in some
Electronic searches were performed in Medline (1966 through May
cases could be adjusted for having performed interim analyses and
2003) and Embase (1988 through May 2003) using optimally sensi-
other factors. When continuous outcome data were analyzed, differ-
tive search strategies for identification of randomized, controlled
ence in means and 95% CI at the end of treatment or difference in
trials developed by the Cochrane Collaboration (20). The Cochrane
mean change between baseline and end of treatment value were
Renal Group Specialized Register was also searched. The following
calculated for individual trials. Subgroup analysis was planned to
medical subject heading terms and text words were used: anemia,
explore potential sources of variability in observed treatment effect
hemoglobin, hematocrit, CKD, renal replacement therapy, renal
when possible (overt or nonovert cardiovascular disease, time on
dialysis and he- mofiltration, end stage renal disease, uremia, EPO,
dialysis, time on predialysis treatment). Heterogeneity of treatment
and darbepoetin. Trials were considered without language restriction.
effects between studies was formally tested using the Q
The results of these searches were analyzed in title and abstract form
(heterogeneity
by one of the authors according to the inclusion criteria (G.F.M.S.). 2
) and the I2 statistics. When appropriate, summary estimators of
Reference lists from all identified articles were also searched. The
treatment effects were calculated using the Der Simonian-Laird ran-
conference pro- ceedings of the American Society of Nephrology and
dom-effects model with RR and its 95% CI for dichotomous
European Di- alysis and Transplantation Association meetings of
outcomes and weighted mean difference with 95% CI for continuous
1999 to 2003 were searched for abstracts of relevant trials.
outcomes (22).
Information about unpub- lished trials were sought from experts in the
field, and pharmaceutical companies involved in the production of
EPO and darbepoetin (Janssen-Cilag, Amgen, Roche) were Results
approached for information about relevant trials. The combined search of Medline, Embase, and the specialist
registry of the Cochrane Renal Group identified 1365 articles,
Data Extraction and Quality Assessment 1322 of which were excluded. The major reasons for exclusion
Each trial was assessed by two independent reviewers (G.F.M.S. were that selected studies were not randomized or were ran-
and C.M.). From all included trials, data were extracted on charac- domized trials that evaluated other interventions (e.g., subcu-
teristics of the study sample, doses and modalities of treatment, taneous versus intravenous EPO treatment for anemia of CKD)
or because only surrogate and hemorheologic outcomes were

a
Table 1. Published guidelines on Hb targets in patients with CKD
Target Hb
Guidelines Country Year Level (g/L)

National Kidney Foundation-Dialysis Outcome Quality Initiative (NKF-DOQI) United States 2000 110–120
British Renal Association (BRA) United Kingdom 2002 100
Canadian Society of Nephrology (CSN) Canada 1999 110–120
European Best Practice Guidelines (EBPG) Europe 2004 110b
Health Care and Financing Administration (HCFA) United States 2000 103–120
Caring for Australians with Renal Impairment (CARI) Australia 2003 120c
120–140d
a
Hb, hemoglobin; CKD, chronic kidney disease.
b
Hb concentrations 120 g/L are not recommended for patients with severe cardiovascular disease unless continuing severe symptoms
dictate otherwise.
c
In patients with proven or likely significant cardiovascular disease (level I evidence).
d
In patients without cardiovascular disease (suggestion for clinical care).
3156 Journal of the American Society of Nephrology J Am Soc Nephrol 15: 3154–3165, 2004

reported (e.g., blood viscosity, hematopoietic progenitor cell that patients in this group were experiencing harmful effects.
assays; Figure 1). Full-text assessment of 43 potentially eligi- The publication of this trial resulted in a change in the methods
ble papers identified 19 eligible trials (2696 patients) reported of the Furuland et al. (42) trial, with 33 enrolled patients with
in 25 publications (13,15,16,23– 43). Authors of 12 trials were cardiovascular impairment excluded from this study, and the
contacted for clarification about study methods and/or request inclusion criteria were changed to avoid further enrollment of
for additional unpublished information; four responded. patients with cardiovascular impairment. No patients with car-
diovascular disease were enrolled in the Brandt et al. (26) and
Trial Characteristics Roger et al. (43) trials; left ventricular hypertrophy or dilation
Two groups of studies were identified. In seven trials (2058 was one of the inclusion criteria of the Foley et al. (29) trial
patients), the hypothesis of whether a higher and “normal” Hb (Table 2).
target was better than a lower Hb target was tested. Patients In the other 12 trials (638 patients), the tested hypothesis
were randomized to reach the two pre-established Hb targets. was a pharmacologic one; that is, whether EPO treatment
Generally a low EPO dose was administered to achieve a compared with no EPO treatment was associated with im-
lower Hb target and a higher dose was administered to achieve proved outcomes. Patients who were randomized to placebo
a higher Hb target. However, in two of these trials, some ended up having a lower Hb target of 95 g/L, whereas those
patients in the low Hb target group received lower doses of who were randomized to EPO treatment achieved higher Hb
EPO or no EPO at all (42,43). Overall, the achieved Hb values levels of 100 g/L (15,23,24,27,30 –37). Overall, the achieved
of the experimental arms in this group of trials were in the Hb targets of the experimental arms in this group of trials were
range of in the range of 95 to 133.3 g/L and the achieved Hb targets of
119 to 150 g/L and the achieved Hb values of the control arm the control arm were in the range of 75 to 104 g/L. The
were in the range of 90 to 120 g/L. Of the seven studies in this patients who were enrolled in this second group of trials had
group, the Besarab et al. (16), Berns et al. (26), and Conlon et no overt cardiovascular comorbidity. The Canadian
al. (28) trials enrolled patients with severe cardiovascular Erythropoietin Study Group trial had three arms: one of
disease (congestive heart failure or ischemic heart disease). placebo, one of low-dose EPO, and one of high dose EPO.
The Besarab et al. (16) trial was terminated at the third interim The Lim et al. (32) trial included four arms: one of placebo
analysis when differences in mortality between the groups and three of different EPO doses.
were recognized as sufficient to make it very unlikely that In summary, the first group contained studies in which the
continuation of the study would reveal a benefit for the higher intervention was to achieve different Hb targets, and trials
Hb target group. There was also a significantly higher rate of included individuals with clinical cardiovascular disease. The
vascular access loss in the higher Hb target group, indicating

Figure 1. Flowchart indicating the number of citations retrieved by individual searches and the final number and grouping of included trials;
reasons for exclusions are provided.
J
A
m
So
c
Ne
ph
rol
Table 2. Characteristics of patients and interventions in trials of EPO or randomized Hb targets for the anemia of CKDa 15:
Low Target Group High Target Group
31
Randomized 54
Treatment

Study ID Cointerventions
Modality Intervention Follow-up No. of Targeted Achieved No. of Targeted Achieved 31
(Months) 65,
Patients Hb Value (g/L) Hb Value (g/L) Patients Hb Value (g/L) Hb Value (g/L) 20
Berns et al. (25) Hemodialysise EPO versus EPO None indicated 12 17 100.0 10.0 101.0 3.3 14 140.0 10.0 140.0 3.6 04
Low Hb target
Besarab et al.( (16)
100 g/L) versus high hemoglobin target ( e 140
Hemodialysis g/L)
EPO versus EPO IV Fe 29 618 100.0 10.0 100.0 10.0 618 140.0 10.0 140.0 10.0
Brandt et al.b (26) Predialysis EPO versus EPO Oral (3 mg/kg per d) or IV 2 mg/kg 3/wk Fe (target ferritin 100 20 22 100.0 10.0 100.0 10.0 21 140.0 10.0 140.0 10.0
Peritoneal dialysis ng/mL/transferrin saturation 20%)
Hemodialysis
c
Conlon et al. (28) Hemodialysise EPO versus EPO None indicated 6 16 100.0 10.0 100 14.3 15 140.0 10.0 136.0 17.0
Foley et al. (29) Hemodialysise EPO versus EPO None indicated 12 73 95.0–105.0g 102.0–106.0g 73 130.0–140.0g 119.0–125.0g
Furuland et al. (42) Predialysis EPO versus EPO or no Oral or IV Fe (target transferrin saturation 20%/ferritin 250 g/L) 12–19 200 90.0–120.0d,g 113.0 13.0h 216 130.0–150.0d,g 143 11.0h
Hemodialysis treatment
Peritoneal dialysis
Roger et al. (43) Predialysis EPO versus EPO or no Oral or IV Fe (target transferrin saturation 20%/ferritin 100 g/L) 24 80 90.0–100.0g 110.0 10.0g 75 120.0–130.0g 122.0 7.0g

treatment
No EPO treatment (Hb 95 g/L) versus EPO treatment (Hb 100 g/L)
Abraham and Macres (23) Predialysis EPO versus standardf
Fe (dose/route NA) (target transferrin saturation 20%); folic acid 1 3 4 NA 96.0 10 4 133.0 123.0 6.6
mg/d He
f
Bahlmann et al. (24) Hemodialysis EPO versus standard None indicated 3 66 NA 100.0 63 100.0–116.0 100.0–116.6b mo
Canadian Erythropoietin Study Group (40) Hemodialysis EPO versus placebo Oral or IV Fe (dose NA) (target ferritin 250 g/L) 6 40 NA 90.0 38 95.0–110.0g 95.0–110.0g glo
Clyne and Jogestrand (27) Predialysis EPO versus standardf Oral (200–300 mg/d) or IV (100 mg/wk) Fe for all patients 3 10 NA 84.0–104.0b 12 100.0 106.0–117.0g bin
Kleinman et al. (30) Predialysis EPO versus placebo Fe (dose/route NA) 3 7 NA 94.0 7 126.0–133.0g 119.3 Ta
Kuriyama et al. (31) Predialysis EPO versus no treatment Fe “at investigator’s discretion” 9 31 NA 84.3 6.3 42 NA 118.3 13.3
80–93.0g
rge
Lim et al. (32) Predialysis EPO versus placebo Oral (300 mg 3/d) Fe and oral folic acid 1 mg/d “to all patients 2 3 NA 11 NA 126.6
except those with excess iron stores” ts
d
Morris et al. (33) Peritoneal dialysis EPO versus placebo None indicated 6 5 NA 100.0 6 105.0–120.0g 115.0 for
Revicki et al. (34) Predialysis EPO versus no treatment Iron 200 mg/d (route not indicated) if transferrin saturation 12 40 NA 89.3 12.0 43 116.0–120.0g 120.0 the
Sikole et al. (35) Hemodialysis EPO versus no treatment None indicated 12 19 NA 100.0 19 NA 100–116.6g An
Techan et al. (36) Predialysis EPO versus placebo None indicated 6 31 NA 100.0 86 123.0–133.0g 128.0–133.3d em
Watson et al. (37) Predialysis EPO versus placebo None indicated 3 6 NA 86.6 6.6 5 NA 116.6 6.6 ia
a
of
IV, intravenous; NA, not available. Ch
b
Trial enrolling children. ro
c
Substudy of the Besarab et al. trial. nic
d
Hb target was 135 to 150 g/L in females and 145–160 g/L in males. Ki
e
Known cardiovascular abnormality. dn
f
No treatment or blood transfusion. ey
g Di
Range. sea
h
The highest achieved targets are reported independent of specific subgroup (in this trial, target Hb values were presented separately for the three groups of predialysis, se
hemodialysis, and peritoneal dialysis patients).

31
57
3158 Journal of the American Society of Nephrology J Am Soc Nephrol 15: 3154–3165, 2004

second group compared EPO treatment with no EPO communication. In the Besarab et al. study, there was no
treatment. Overall, trials that were pooled in this second group significant difference in the risk of seizures across the two
resulted in lower Hb levels achieved in both the intensive and groups (1233 patients; RR, 1.14; 95% CI, 0.66 to 1.97). For
control arms than the Hb levels in the corresponding arms in the trials of EPO treatment versus no EPO treatment, there was
the first group of trials. On the basis of these considerations, a higher risk of seizures in the no EPO treatment (lower Hb)
the results of these two groups of studies were analyzed group (four trials, 219 patients; RR, 5.25; 95% CI, 1.13 to
separately. Of note, follow-up duration was higher and 24.34; P 0.03) compared with the group that was treated
publication date was more recent in the first group of trials. with EPO. No heterogeneity was demonstrated between trials
These trials seemed more generalizable to current practice, (heterogeneity 2 0.74, P 0.86, I2 0%; Figure 3).
whereas those done earlier seemed to be conducted mainly to Hypertension. Eight studies provided data on the number
demonstrate dose-response relationships and safety of EPO of patients who had hypertensive episodes and required addi-
treatment. tional antihypertensive medication or exclusion of patients
from the trial because of hypertension. Two of these were from
Trial Quality the group of trials that compared high versus low Hb targets,
Trial quality was variable. Allocation concealment was un- and six were from the group of trials of EPO treatment versus
clear in all 19 trials. Outcome assessors were not stated as no EPO treatment. No significant difference for hypertension
blinded in any of the trials, and only three studies were ana- was demonstrated between the two Hb targets in the first group
lyzed on an intention-to-treat basis. The dropout rate ranged of trials (two trials, 1277 patients; RR, 0.92; 95% CI, 0.59 to
from 0.0 to 57.8%. 1.45). Heterogeneity was not significant across these trials
(heterogeneity 2 1.41, P 0.24, I2 29.0%). There was
Trial Results a significantly lower risk of hypertension in the EPO treatment
All-Cause Mortality. In the high Hb target versus low Hb arm compared with no EPO treatment in the second group of
target trials, there was a lower risk of death with the lower Hb trials (six trials, 387 patients; RR, 0.50; 95% CI, 0.33 to 0.76;
target compared with the high Hb target (four trials, 1949 P 0.001). Heterogeneity in this group of studies was not
patients; RR, 0.84; 95% CI, 0.71 to 1.00; P 0.05). This significant (heterogeneity 2 2.88, P 0.72, I2 0%;
analysis was dominated by the Besarab et al. (16) study, Figure 4).
contributing 86.2% of the weight. There was no heterogeneity Quality of Life. Many problems were evident in the eval-
across these trials (heterogeneity 2 0.59, P 0.8, I2 0%; uation of these data. Five of 10 trials in which quality of life or
Figure 2). exercise capacity was analyzed did not use validated scales for
In the trials of EPO treatment versus no EPO treatment, the assessment of quality of life (i.e., codified scales for the
there was no statistically significant difference in the risk of assessment of quality of life or exercise capacity such as the
all-cause mortality between the two arms (three trials, 255 Short Form-36 or the Kidney Disease Quality of Life ques-
patients; RR, 1.83; 95% CI, 0.48 to 7.06). Heterogeneity across tionnaires, which have been validated using standard methods
these trials was also not significant (heterogeneity 2 0.45, and in relevant populations) (24,27,30,32,36). Also, dimen-
P 0.8, I2 0%). sion-specific and composite quality-of-life outcomes reported
Seizures. This outcome was reported only in the trials of were not prespecified, and only positive results were
EPO treatment versus no EPO treatment, and we obtained presented. When codified scales were used, overall scores
additional data of the Besarab et al. (16) trial by personal presented were

Figure 2. Effect of high versus low hemoglobin (Hb) targets on all-cause mortality.
J Am Soc Nephrol 15: 3154–3165, 2004 Hemoglobin Targets for the Anemia of Chronic Kidney Disease 3159

Figure 3. Effect of high versus low Hb targets on the reported number of patients with seizures.

Figure 4. Effect of high versus low Hb targets on the reported number of patients with hypertensive events that required exclusion from the
study or administration of additional antihypertensive medication.

not described in detail. Commonly, individual items of quality the outcomes (Table 4). Of note, data on the effect of EPO on
of life that had improved with the higher Hb target were serum creatinine were available in only four trials of EPO
reported as an indication of an overall positive effect of the treatment versus no EPO treatment, and there was no signifi-
higher target, or else a positive effect of EPO treatment on cant effect (four studies, 77 patients; weighted mean differ-
quality of life. It was not possible to perform a pooled analysis ence, 0.01; 95% CI, 1.21 to 1.22). Two trials of the high
given the wide variability of the assessment of these outcomes versus low Hb target group indicated a significantly better
and uncertainties regarding the validity of the instruments used creatinine clearance at the end of the trials in the lower Hb
(Table 3). target arms (two trials, 167 patients; weighted mean difference,
Summary data for all other patient-relevant outcomes (myo- 1.07; 95% CI, 2.10 to 0.05).
cardial infarction, serious cardiovascular events [including an-
gina, atrial fibrillation, and severe arrhythmia], stroke, hyper- Discussion
kalemia, access thrombosis, hospitalization rates and days of Key Findings
hospitalization, left ventricular hypertrophy, systolic BP, dia- We find that on the basis of available randomized, controlled
stolic BP, and mean arterial pressure) were also analyzed, trials, Hb targets of 120 g/L are associated with a lower risk
although minimal data were available. These analyses showed of death in the population with cardiovascular disease and
no significant difference between lower and higher Hb targets CKD compared with Hb targets of 130 g/L (RR, 0.84; 95%
or EPO treatment and no EPO treatment in relation to any of CI, 0.71 to 1.00). In absolute terms, the risk of death is 3.0%
Table 3. Quality of life and physical capacity assessment of patients enrolled in trials of EPO or randomized
3160 Journal of the American Society of Nephrology J Am Soc Nephrol 15: 3154–3165, 2004

(95% CI, 1.0 to 6.0%) lower in the group of patients with Hb

every 6 moincrease of physical function score; no significant changes in other scores of the scale with higher targets

Improvement of main physical symptoms in the normal compared with the lower Hb group.
target of 120 g/L compared with the group of patients with

Baseline and end of treatmentImprovement in level of energy, ability to do work, and overall quality of life with higher targets
Angina, dyspnea, “other” cardiovascular symptoms, and sexuality improved with higher targets
Hb target 130 g/L. For every 30 patients treated to an Hb
target of 120 g/L compared with an Hb target of 130 g/L,
approximately one death is avoided.
The present findings are applicable mainly to hemodialysis
patients with clinical cardiovascular disease because the ma-
jority of patients included in these trials had these attributes
(44 – 46). Higher Hb targets have been recommended in
other CKD populations (e.g., CKD stages 3 to 4 or individuals
Result

without overt cardiovascular disease) despite the absence of

A codified scale for the standard assessment of quality of life or exercise capacity whose validity has been tested in patients with kidney disease.
randomized trial data. If higher Hb targets really reduced
mortality, then this should be even more evident in studies that
include high-risk patients, as those enrolled in the Besarab et
al. (16) trial. However, the most favorable interpretation of the
Besarab et al. study can be only that there was no survival
benefit demonstrated. To find a survival benefit in lower risk
patients would be more difficult because a larger number of
individuals would have to be studied to have adequate power,
and empiric evidence of qualitative effect modification (i.e.,
Change in KDQ score for main physical symptoms, fatigue, depression, and frustration favored the normal compared with the lower hematocrit group.

harm in one group and benefit in another using the same
intervention) is very rare (47). The finding of higher mortality
Baseline and end of treatment

Baseline and end of treatment

Notreatment Baseline and end of treatment
Baseline and 3, 6, and 7 mo

Baseline and 12 and 24 mo
Time of Assessment

in higher Hb target groups from very recent cancer trials would
Baseline andSignificant

also make a survival benefit of higher Hb values in the CKD
Baseline and 12 mo

population very unlikely (19). The annualized mortality rate in
the 1233 patients of the Besarab et al. study was 24%, which
is very similar to the annualized mortality rate for every 1000
patients (20.4%) reported by the United States Renal Data
System (http://www.usrds.org), suggesting that results from
Validateda

Yes

Yes

Yes Yes

Yes

Yes Yes

this trial are in fact generalizable.
No

No

No

No of
Baseline and end

Lower Hb targets of 95 g/L in individuals who are not
treated with EPO are associated with a significantly increased
risk of seizures (RR, 5.25; 95% CI, 1.13 to 24.34) compared
Standardized symptom-limited exercise test and electrically braked bicycle ergometer

with treatment with EPO and Hb values of 100 g/L. In
absolute terms, the risk of seizures is 4% (95% CI, 3 to 10%)
Kidney Disease Questionnaire (KDQ) part 1 and part 2

lower with Hb values of 100 compared with the group of
patients with Hb 95 g/L. This means that for every 25
100 g/L) Short Form 36

patients treated to an Hb level of 100 g/L, one fewer will
Scale or Tool

develop seizures.
Profile Kidney Disease Quality of Life tool
Short Form 36 Kidney Disease Quality of Life tool

Finally, lower Hb levels of 95 g/L with no EPO treatment
are associated with a reduced risk of patients who present with
hypertensive episodes (RR, 0.62; 95% CI, 0.42 to 0.92). In
Renal Quality of Life Profile (Hb

absolute terms, the risk of developing hypertensive episodes is
Short Form 36

Study specific

Study specific Study specific

16% lower (95% CI, 8 to 24%) with Hb values 95 g/L
Low Hb target ( 100 g/L) versus high Hb target ( 140 g/L)

compared with Hb 100 g/L. For every seven patients treated
95 g/L) versus EPO treatment

to obtain an Hb 100 g/L, one patient will require additional
antihypertensive medication.
Study Group
Canadian Erythropoietin Sickness (40)
Impact

Comparison with Other Studies
This systematic review is influenced mainly by the finding
Clyne and Jogestrand (27)
Study ID

of Besarab et al. (16) of an association of higher mortality in
Bahlmann et al. (24)

Kleinman et al. (30)
No EPO treatment (Hb
Furuland et al. (42)
Besarab et al. (16)

Lim et al. (32) Teehan et al. (36)

individuals who have cardiovascular disease and are treated to
Roger et al. (43)
Foley et al. (29)

Hb targets 130 g/L. The other small trials are inadequately
powered to show any evidence of benefit or harm, both sepa-
rately and combined (RR, 0.98; 95% CI, 0.61 to 1.55). They
a

show no difference between the arms that treated to higher
J
A
m
So
c
Ne
ph
rol
15:
31
a 54
Table 4. Other outcomes analyzed in trials of EPO or randomized Hb targets in patients with anemia of CKD –
Results
31
No. of Patients with Events or No. of Patients with Events or Heterogeneity 65,
Meansa in Low Target Group Meansa in High Target Group Weighted Mean
2
P Value
Outcome Analyzed No. of Studies No. of Patients Relative Risk 95% CI
Difference 95% CI 20
04

Low Hb target ( 100 g/L) versus high Hb target ( 140 g/L)
myocardial infarction (all) 1 1389 42 41 1.03 0.68–1.56 — — NA
myocardial infarction (fatal) 1 1233 28 22 1.28 0.74–2.21 — — NA
myocardial infarction (nonfatal) 1 1233 14 19 0.74 0.37–1.46 — — NA
serious cardiovascular events 1 146 3 4 0.75 0.17–3.23 — — NA
stroke 1 1233 9 14 0.65 0.28–1.48 — — NA
hyperkalemia 2 1277 3 5 0.62 0.15–2.56 — — 0.89
access thrombosis 3 1795 242 296 1.05 0.63–1.75 — — 0.01
hospitalization for all causes 1 1233 425 445 0.96 0.89–1.03 — — NA
systolic BP 1 246 151.50 150.70 — — 0.80 4.78–6.38 NA
diastolic BP 1 246 75.90 77.30 — — 1.40 4.37–1.57 NA
mean arterial pressure 1 16 104.80 105.60 — — 0.80 12.39–10.79 NA
serum creatinine — — — — — — — — NA
creatinine clearance 2 167 13.00–23.00b 16.00–24.00b — — 1.07 2.10– 0.05 0.48 He
left ventricular mass index (g/m2) 1 127 102.00 105.00 — — 3.00 10.66–4.66 0.0005
mo
days of hospitalization 1 416 3.80 4.80 — — 1.00 2.75–0.75 NA
No EPO treatment (Hb 95 g/L) versus EPO treatment (Hb 100 g/L)
glo
myocardial infarction (all) 3 170 2 1 1.56 0.26–9.50 — — 0.60 bin
myocardial infarction (fatal) 2 156 2 0 1.36 0.80–2.31 — — 0.40 Ta
myocardial infarction (nonfatal) 1 14 0 1 0.71 0.37–1.39 — — 0.30 rge
serious cardiovascular events 1 245 11 5 1.56 0.49–5.03 — — 0.10 ts
stroke 2 1412 2 0 0.79 0.36–1.72 — — 0.60
hyperkalemia 1 14 0 1 0.56 0.15–2.01 — — 0.86
for
access thrombosis 2 1972 5 12 0.94 0.59–1.51 — — 0.20 the
systolic BP 3 123 139.90–151.00b 143.40–151.10b — — 0.66 7.50–6.19 0.78 An
diastolic BP 3 369 81.40–92.90b 84.50–93.90b — — 2.11 6.44–0.12 0.75 em
mean arterial pressure 4 60 67.30–112.30b 67.30–113.30b — — 0.35 5.63–6.33 0.89 ia
serum creatinine 4 77 4.00–10.50b 5.70–10.83b — — 0.01 1.21–1.22 0.16 of
creatinine clearance 2 19 10.30–20.00b 12.90–18.00b — — 2.15 6.14–1.84 0.50 Ch
left ventricular mass index (g/m2) 2 45 101.20–133.80b 87.60–171.80b — — 16.64 66.45–33.17 0.04 ro
nic
a
CI, confidence interval. Ki
b
Means are provided for continuous variables; lower and higher mean is provided for analyses including more than one trial. dn
ey
Di
sea
se

31
61
3162 Journal of the American Society of Nephrology J Am Soc Nephrol 15: 3154–3165, 2004

versus lower Hb targets, although the CI do not exclude benefit not consider seizures as an outcome, and there was no standard
or harm. quality assessment of the included trials. The most recently
Seizures are reported to be a complication of EPO in the published is a “systematic review” of the impact of epoetin
product information. However, this systematic review indicates on clinical end points in patients with CKD (50), which sug-
that treating with EPO may be protective against seizures. gested that epoetin therapy provides important clinical and
There is some uncertainty with this finding as a result of the quality-of-life benefits while substantially reducing hospital-
small number of trials reporting this outcome; a small number izations and transfusions. This study had the following limita-
of patients were present in each Hb target group, and very tions: It did not fulfill standard requirements for systematic
small number of events were reported. The physiologic ratio- reviews, as it lacked clear inclusion criteria and an explicit and
nale underlying the protective role of EPO against seizures comprehensive search strategy; it included both randomized
could be the increased peripheral cerebral oxygenation of neu- and cohort studies; and there was no quality assessment of
rons and the observation that neurons develop EPO receptors included studies. We conclude that on the basis of available
during ischemia, with a neuroprotective effect of EPO inde- data from randomized trials, it is not possible to draw evidence
pendent of its effects on Hb (48). of improvement in quality of life, hospitalizations, and
Hypertension is a widely known adverse effect of EPO transfusions.
treatment, and our review confirms this finding (9). Access
thrombosis is another recognized complication of EPO treat- Strengths and Weaknesses
ment, although the results in the available studies are conflict- Although there are many studies and reviews on this debated
ing. In our review, we found that the Besarab et al. (16) study topic, this is the most updated systematic review of
showed a significantly higher risk of access thrombosis with randomized trials reported. Our findings contrast with data
the higher Hb target, whereas the remaining trials did not from large observational studies that have shown a consistent
confirm this finding. association between higher Hb values and improved outcomes,
Reports of stroke episodes in the available trials in individ- including increased survival (56). This discrepancy between
uals with CKD are few. This aspect mandates further investi- observa- tional and trial data are well recognized.
gation, in light of available data suggesting that EPO promotes Observational studies are not the best study design to answer
thrombosis by interaction with platelet function (49). A recent intervention questions, because of bias and confounding, the
trial of EPO versus placebo in breast cancer patients was effects of which are unpredictable and may result in an
terminated prematurely because of significantly higher mortal- overestimate, an underes- timate, or true estimate of effect
ity in the group that was treated with EPO, as a result of an (57). In this setting, observa- tional studies may well be flawed
increase in the incidence of thrombotic and vascular events and as a result of residual confounding such that the higher Hb
breast cancer progression in the EPO-treated patients (50). values may reflect under- lying better survival potential as a
Another placebo-controlled trial of EPO in patients with head result of an inability to adjust completely for all known and
and neck cancer found that EPO corrects anemia, but there unknown predictors of survival. In short, healthier patients
was a significantly higher number of deaths in patients in the (with higher Hb values) live longer. Only randomized trials
EPO arm (51). A recent editorial in Lancet Oncology reported can show whether changing Hb values improves survival,
similar findings from other still unpublished studies (19). The because any baseline predictors of survival should be balanced
applicability of these results to CKD patients is unclear but a between treatment arms as a result of the randomization.
cause of concern. Recent examples of the unpredict- able biased nature of
EPO is widely reported to improve quality of life. We could observational studies include the findings of the ADEMEX and
not pool results on quality of life; nonvalidated scales were the HEMO Study (randomized trials), which contrast with the
often used, and authors reported individual domains of a qual- results of the CANUSA (observational) study (58,59).
ity-of-life scale, which were significantly different, without It is important to point out the potential reasons for hetero-
prespecification, rather than report the summary measure of geneity in the studies included in these analyses to understand
effect, suggesting outcome reporting bias (52). the limitations of the conclusions that can be derived from
Many narrative reviews and some meta-analyses have been them. These differences, which include variable sample size,
published on this debated topic. Two Cochrane systematic treatment protocols (treatment targets or EPO interventions),
reviews of randomized trials of EPO use or Hb targets for populations studied, follow-up times (16,42), deficiencies in
anemia of CKD are consistent with the present larger analysis design, and reporting of the published trials, all were made
that includes the most recently published trials (53,54). Our explicit in this study but did not result in statistically
study’s findings are also consistent with the results of an significant heterogeneity of any analysis.
Agency for Healthcare Research and Quality report on EPO
use for anemia of CKD (55). This report, which was also based Applicability to Clinical Practice
on randomized, controlled trials, concluded that there is not From the available trial evidence, the benefits associated
strong evidence showing that maintaining Hb 120 g/L is with EPO treatment achieving higher Hb targets (reduced
more beneficial to CKD patients than Hb 120 g/L. Compared seizures) are outweighed by the harms (increased risk of hy-
with ours, the Agency for Healthcare Research and Quality pertension and mortality), and data for patients with CKD and
report did not include some recently published trials and did cardiovascular disease clearly indicate that the preferred Hb
J Am Soc Nephrol 15: 3154–3165, 2004 Hemoglobin Targets for the Anemia of Chronic Kidney Disease 3163

target should be 120 g/L. Data relating to other populations Nissenson, and V. Montinaro kindly agreed to review the manuscript.
(predialysis patients with CKD with and without cardiovascu- We are indebted to Premala Sureshkumar for collaboration in some of
lar impairment) are unclear. Higher Hb targets have long been the data analysis, Friederike Bachmann for translating one of the trials
known to be associated with improvement in the quality of life, and assisting in data extraction, Narelle Willis for assistance as
but available evidence from randomized, clinical trials is prob- coordinator of the Cochrane Renal Group, Ruth Mitchell and Gail
lematic. Until additional, well-designed trials that adequately Higgins for assistance in the development of search strategies, and
Sandra Puckeridge for excellent administrative support. We also ac-
assess safety and quality of life are available, clinicians always
knowledge the anonymous reviewers who provided useful comments
need to consider the potential harms and costs whenever pre- to previous versions of this analysis.
scribing interventions of unproved efficacy.
References
Future Research 1. Hsu CY, McCulloch CE, Curhan GC: Epidemiology of anemia
Our study highlights the need for randomized, controlled associated with chronic renal insufficiency among adults in the
trials in the area of anemia management in CKD. The concern- United States: Results from the Third National Health and Nu-
ing results about higher risk of all-cause mortality in patients trition Examination Survey. J Am Soc Nephrol 13: 504 –510,
who have cardiovascular comorbidities and are treated with 2002
2. Obrador GT, Roberts T, St Peter WL, Frazier E, Pereira BJ,
higher EPO doses in the pursuit of higher Hb targets arise
Collins AJ: Trends in anemia at initiation of dialysis in the
mainly from the most recent trials that had the longest fol- United States. Kidney Int 60: 1875–1884, 2001
low-up duration and seem to be more generalizable to current 3. Hemodialysis. Australia and New Zealand Registry Annual Re-
practice than the remaining ones. Adequately powered, well- port 2001. Available: http://www.anzdata.org/au
designed and reported long-term, randomized, controlled trials 4. Caro J, Brown S, Miller O, Murray T, Erslev AJ: Erythropoietin
comparing the benefits and harms of different Hb levels are levels in uremic nephric and anephric patients. J Lab Clin Med
necessary. Given the findings of Besarab et al. (16) of in- 93: 449 – 458, 1979
creased all-cause mortality with a higher Hb target (Hb 140 5. Klang B, Bjorvell H, Clyne N: Quality of life in predialytic
g/L) and the uncertainties of trials in which an average (Hb uremic patients. Qual Life Res 5: 109 –116, 1996
100 to 110 g/L) target was tested, future studies should com- 6. Mann JF: What are the short-term and long-term consequences
of anaemia in CRF patients? Nephrol Dial Transplant 14[Suppl
pare the effect of treatment targets for Hb in the range of 120
2]: 29 –36, 1999
to 140 with lower targets, with adequate power, and in
7. Humphries JE: Anemia of renal failure. Use of erythropoietin.
different populations (e.g., predialysis CKD patients). Trials of Med Clin North Am 76: 711–725, 1992
Hb tar- gets could be done with EPO but also the newer agent 8. Valderrabano F, Jofre R, Lopez-Gomez JM: Quality of life in
darbe- poetin. Seizures should be reported, and quality of life end-stage renal disease patients. Am J Kidney Dis 38: 443– 464,
should be assessed using validated tools. The trials should be 2001
ade- quately powered to detect relevant patient outcomes, 9. Sundal E, Kaeser U: Correction of anaemia of chronic renal
including mortality and thrombovascular accidents. With a failure with recombinant human erythropoietin: Safety and effi-
power of cacy of one year’s treatment in a European multicentre study of
80%, ~1500 patients would be needed in a population with a 150 haemodialysis-dependent patients. Nephrol Dial Transplant
baseline mortality risk of 15% if we expected the intervention 4: 979 –987, 1989
10. McMahon LP, Mason K, Skinner SL, Burge CM, Grigg LE,
(high Hb target) to reduce mortality to 10%. Approximately
Becker GJ: Effects of haemoglobin normalization on quality of
2000 would be needed with a baseline mortality of 20% and an life and cardiovascular parameters in end-stage renal failure.
expected reduction to a value of 15%, a 2-yr accrual time and Nephrol Dial Transplant 15: 1425–1430, 2000
3-yr follow-up, and an expected noncompliance of ~10%. 11. Nissenson AR: Novel erythropoiesis stimulating protein for man-
Insights for the management of the anemia of CKD may also aging the anemia of chronic kidney disease. Am J Kidney Dis 38:
derive from the three ongoing randomized trials on the topic 1390 –1397, 2001
(CREATE, CHOIR, and TREAT). Their results are particu- 12. Sikole A, Stojanovic A, Polenakovic M, Petrusevska G,
larly expected to elucidate the potential impact of different Hb Sadikario S, Saso R, Jovanovski M: How erythropoietin affects
targets in predialysis patients with CKD. Additional informa- bone marrow of uremic patients. Am J Nephrol 17: 128 –136,
tion on the relationship between Hb values achieved and clin- 1997
13. Keown PA: Quality of life in end-stage renal disease patients
ical outcomes could be gleaned from an individual patient data
during recombinant human erythropoietin therapy. The Canadian
meta-analysis.
Erythropoietin Study Group. Contrib Nephrol 88: 81– 86, 1991
14. Hays RD, Kallich JD, Mapes DL, Coons SJ, Carter WB: Devel-
Acknowledgments opment of the Kidney Disease Quality of Life (KDQOL) instru-
This study was partly funded by the Cochrane Renal Group, the ment. Qual Life Res 3: 329 –338, 1994
NHMRC Centre for Clinical Research Excellence in Renal Medicine, 15. Laupacis A: A randomized double-blind study of recombinant
and a young investigator award granted to G.F.M.S. by the Italian human erythropoietin in anaemic hemodialysis patients. Cana-
Society of Nephrology. dian Erythropoietin Study Group. Transplant Proc 23: 1825–
We acknowledge collaboration of Drs. J. Berns, A. Besarab, and R. 1826, 1991
Foley and Ms. Cindy Wong, and Dr. A. Laupacis for providing data 16. Besarab A, Bolton WK, Browne JK, Egrie JC, Nissenson AR,
relating to their trials that were not reported or were unclear in the Okamoto DM, Schwab SJ, Goodkin DA: The effects of normal
publications. Drs. J. Berns, A. Besarab, J. Conlon, A. Laupacis, A.
3164 Journal of the American Society of Nephrology J Am Soc Nephrol 15: 3154–3165, 2004

as compared with low hematocrit values in patients with cardiac 33. Morris KP, Sharp J, Watson S, Coulthard MG: Non-cardiac
disease who are receiving hemodialysis and epoetin. N Engl benefits of human recombinant erythropoietin in end stage renal
J Med 339: 584 –590, 1998 failure and anaemia. Arch Dis Child 69: 580 –586, 1993
17. Muirhead N: Erythropoietin is a cause of access thrombosis. 34. Revicki DA, Brown RE, Feeny DH, Henry D, Teehan BP,
Semin Dial 6: 184 –188, 1993 Rudnick MR, Benz RL: Health-related quality of life associated
18. Tonelli M, Winkelmayer WC, Jindal KK, Owen WF, Manns BJ: with recombinant human erythropoietin therapy for predialysis
The cost-effectiveness of maintaining higher hemoglobin targets chronic renal disease patients. Am J Kidney Dis 25: 548 –554,
with erythropoietin in hemodialysis patients. Kidney Int 64: 1995
295–304, 2003 35. Sikole A, Polenakovic M, Spirovska V, Polenakovic B, Masin G:
19. Epoetin: For better or for worse? Lancet Oncol 5: 1, 2004 Analysis of heart morphology and function following erythro-
20. Crowther CA, Henderson-Smart DJ: Phenobarbital prior to pre- poietin treatment of anemic dialysis patients. Artif Organs 17:
term birth for preventing neonatal periventricular haemorrhage. 977–984, 1993
Cochrane Database Syst Rev (3): CD000164, 2003 36. Teehan BP, Benz RL, Sigler MH, Brown JM: Early intervention
21. Schulz KF, Chalmers I, Hayes RJ, Altman DG: Empirical evi- with recombinant human erythropoietin therapy. Semin Nephrol
dence of bias. Dimensions of methodological quality associated 10: 28 –34, 1990
with estimates of treatment effects in controlled trials. JAMA 37. Watson AJ, Gimenez LF, Cotton S, Walser M, Spivak JL:
Treatment of the anemia of chronic renal failure with subcuta-
273: 408 – 412,
1995 neous recombinant human erythropoietin. Am J Med 89: 432–
435, 1990
22. DerSimonian R, Laird N: Meta-analysis in clinical trials. Control
38. Effect of recombinant human erythropoietin therapy on blood
Clin Trials 7: 177–188, 1986
pressure in hemodialysis patients. Canadian Erythropoietin
23. Abraham PA, Macres MG: Blood pressure in hemodialysis pa-
Study Group. Am J Nephrol 11: 23–26, 1991
tients during amelioration of anemia with erythropoietin. J Am
39. Conlon PJ, Kovalik E, Schumm D, Minda S, Schwab SJ: Nor-
Soc Nephrol 2: 927–936, 1991
malization of hematocrit in hemodialysis patients does not affect
24. Bahlmann J, Schoter KH, Scigalla P, Gurland HJ, Hilfenhaus M, silent ischemia. Ren Fail 22: 205–211, 2000
Koch KM, Muthny FA, Neumayer HH, Pommer W, Quelhorst E: 40. Association between recombinant human erythropoietin and
Morbidity and mortality in hemodialysis patients with and with- quality of life and exercise capacity of patients receiving hae-
out erythropoietin treatment: A controlled study. Contrib Neph- modialysis. Canadian Erythropoietin Study Group. BMJ 300:
rol 88: 90 –106, 1991 573–578, 1990
25. Berns JS, Rudnick MR, Cohen RM, Bower JD, Wood BC: 41. Roth D, Smith RD, Schulman G, Steinman TI, Hatch FE, Rud-
Effects of normal hematocrit on ambulatory blood pressure in nick MR, Sloand JA, Freedman BI, Williams WW Jr, Shadur
epoetin-treated hemodialysis patients with cardiac disease. Kid- CA: Effects of recombinant human erythropoietin on renal func-
ney Int 56: 253–260, 1999 tion in chronic renal failure predialysis patients. Am J Kidney
26. Brandt JR, Avner ED, Hickman RO, Watkins SL: Safety and Dis
efficacy of erythropoietin in children with chronic renal failure. 24: 777–784, 1994
Pediatr Nephrol 13: 143–147, 1999 42. Furuland H, Linde T, Ahlmen J, Christensson A, Strombom U,
27. Clyne N, Jogestrand T: Effect of erythropoietin treatment on Danielson BG: A randomized controlled trial of haemoglobin
physical exercise capacity and on renal function in predialytic normalization with epoetin alfa in pre-dialysis and dialysis pa-
uremic patients. Nephron 60: 390 –396, 1992 tients. Nephrol Dial Transplant 18: 353–361, 2003
28. Conlon PJ, Kovalik E, Schumm D, Minda S, Schwab SJ: Nor- 43. Roger SD, McMahon LP, Clarkson A, Disney A, Harris D,
malization of hematocrit in hemodialysis patients with cardiac Hawley C, Healy H, Kerr P, Lynn K, Parnham A, Pascoe R, Voss
disease does not increase blood pressure. Ren Fail 22: 435– 444, D, Walker R, Levin A: Effects of early and late intervention with
2000 epoetin alpha on left ventricular mass among patients with
29. Foley RN, Parfrey PS, Morgan J, Barre PE, Campbell P, Cartier chronic kidney disease (stage 3 or 4): Results of a randomized
P, Coyle D, Fine A, Handa P, Kingma I, Lau CY, Levin A, clinical trial. J Am Soc Nephrol 15: 148 –156, 2004
Mendelssohn D, Muirhead N, Murphy B, Plante RK, Posen G, 44. Jacobs C: Normalization of haemoglobin: Why not? Nephrol
Dial Transplant 14[Suppl 2]: 75–79, 1999
Wells GA: Effect of hemoglobin levels in hemodialysis patients
45. Foley RN: Should hemoglobin be normalized in uremic patients?
with asymptomatic cardiomyopathy. Kidney Int 58: 1325–1335,
Clin Nephrol 58[Suppl 1]: S58 –S61, 2002
2000
46. Ritz E, Schwenger V: The optimal target hemoglobin. Semin
30. Kleinman KS, Schweitzer SU, Perdue ST, Bleifer KH, Abels RI:
Nephrol 20: 382–386, 2000
The use of recombinant human erythropoietin in the correction
47. Ioannidis JP, Lau J: Heterogeneity of the baseline risk within
of anemia in predialysis patients and its effect on renal function: patient populations of clinical trials: A proposed evaluation al-
A double-blind, placebo-controlled trial. Am J Kidney Dis 14: gorithm. Am J Epidemiol 148: 1117–1126, 1998
486 – 48. Chong ZZ, Kang JQ, Maiese K: Hematopoietic factor erythro-
495, 1989 poietin fosters neuroprotection through novel signal transduction
31. Kuriyama S, Tomonari H, Yoshida H, Hashimoto T, Kawaguchi cascades. J Cereb Blood Flow Metab 22: 503–514, 2002
Y, Sakai O: Reversal of anemia by erythropoietin therapy retards 49. Taylor JE, Henderson IS, Stewart WK, Belch JJ: Erythropoietin
the progression of chronic renal failure, especially in nondiabetic and spontaneous platelet aggregation in haemodialysis patients.
patients. Nephron 77: 176 –185, 1997 Lancet 338: 1361–1362, 1991
32. Lim VS, DeGowin RL, Zavala D, Kirchner PT, Abels R, Perry 50. Leyland-Jones B, BEST Investigators and Study Group: Breast
P, Fangman J: Recombinant human erythropoietin treatment in cancer trial with erythropoietin terminated unexpectedly. Lancet
pre-dialysis patients. A double-blind placebo-controlled trial. Oncol 4: 459 – 460, 2003
Ann Intern Med 110: 108 –114, 1989
J Am Soc Nephrol 15: 3154–3165, 2004 Hemoglobin Targets for the Anemia of Chronic Kidney Disease 3165

51. Henke M, Laszig R, Rube C, Schafer U, Haase KD, Schilcher B, N9.29, Rockville, MD, Agency for Healthcare Research and
Mose S, Beer KT, Burger U, Dougherty C, Frommhold H: Quality, 2001 (Prepared by the Blue Cross and Blue Shield
Erythropoietin to treat head and neck cancer patients with anae- Association Technology Evaluation Center under contract No.
mia undergoing radiotherapy: Randomised, double-blind, place- 290-97-0015; AHRQ Publication No. 01-3016)
bo-controlled trial. Lancet 362: 1255–1260, 2003 56. Collins AJ, Ma JZ, Xia A, Ebben J: Trends in anemia treatment
52. Lohr KN, Aaronson NK, Alonso J, Burnam MA, Patrick DL, with erythropoietin usage and patient outcomes. Am J Kidney
Perrin EB, Roberts JS: Evaluating quality-of-life and health Dis
status instruments: Development of scientific review criteria. 32: S133–S141, 1998
Clin Ther 18: 979 –992, 1996 57. Sackett DL: Bias in analytic research. J Chronic Dis 32: 51– 63,
53. Cody J, Daly C, Campbell M, Donaldson C, Grant A, Khan I, 1979
Vale L, Wallace S, MacLeod A: Frequency of administration of 58. Paniagua R, Amato D, Vonesh E, Correa-Rotter R, Ramos A,
recom- binant human erythropoietin for anaemia of end-stage renal Moran J, Mujais S, Mexican Nephrology Collaborative Study
disease in dialysis patients. Cochrane Database Syst Rev Group: Effects of increased peritoneal clearances on mortality
CD003895, 2002 rates in peritoneal dialysis: ADEMEX, a prospective, random-
54. Strippoli GF, Manno C, Schena FP, Craig JC: Haemoglobin and ized, controlled trial. J Am Soc Nephrol 13: 1307–1320, 2002
haematocrit targets for the anaemia of chronic renal disease. 59. Depner T, Beck G, Daugirdas J, Kusek J, Eknoyan G: Lessons
Cochrane Database Syst Rev (4): CD003967, 2003 from the Hemodialysis (HEMO) Study: An improved measure of
55. Flamm CR, Aronson N, Bohn R: Use of Epoetin for Anemia in the actual hemodialysis dose. Am J Kidney Dis 33: 142–149,
Chronic Renal Failure. Evidence Report/Technology Assessment 1999
Access to UpToDate on-line is available for additional clinical information
at http://www.jasn.org/