Nephrol Dial Transplant (2005) 20: 2453–2457

doi:10.1093/ndt/gfi027
Advance Access publication 2 August 2005

Original Article

Haemoglobin level and vascular access survival

José M. Garrancho1, Judith Kirchgessner2, Mariana Arranz1, Gerdi Klinkner2, Ramón Rentero1,
Juan A. Ayala1 and Daniele Marcelli2

1
Fresenius Medical Care Spain, Madrid, Spain, 2Fresenius Medical Care Dentschland Gubh, Bad Homburg, Germany

Abstract diabetes, age >65 years and iPTH >400 pg/ml were
Background. A full correction of anaemia in haemo- negative predictive factors for VA survival.
dialysis (HD) patients may lead to an increased risk of
vascular access (VA) failure. We studied the relation- Keywords: ACE inhibitors; anemia correction;
ship between haemoglobin (Hb) level and VA survival. Diabetes Mellitus; haemoglobin; iPTH;
Methods. Incident patients between January 2000 and vascular access
December 2002 with <1 month on HD were con-
sidered. The relative risk (RR) of access failure
was evaluated in four different groups of patients
divided according to their Hb level (<10, 10–12, 12–13 Introduction
and >13 g/dl). Other factors possibly influencing VA
survival were also considered: age, gender, diabetes, Anaemia is a characteristic and important clinical
vascular disease, intact parathyroid hormone (iPTH) manifestation of progressive renal disease. It usually
and treatment with an angiotensin-converting enzyme worsens with the development of renal failure and it
(ACE) inhibitor, angiotensin receptor blocker (ARB) can be corrected with recombinant human erythro-
or recombinant human erythropoeitin therapy. poietin (rHuEPO).
Results. We studied 1254 patients (1057 with auto- Although rHuEPO has been used in dialysis patients
logous fistulae, 75 grafts and 122 permanent catheters since 1989, there is not yet a global consensus regarding
at admission). Based on Cox analysis, we found the the optimum haematocrit target in this patient popula-
next statistically significant RR of VA failure to be tion. Indeed, the appropriate target haematocrit for
2.3 times higher with grafts than with arterio-venous dialysis patients has throughout the last decade been
fistulae (AVFs) and 1.8 times higher in AVFs with Hb one of the most debated issues in nephrology.
<10 g/dl than in AVFs of the next Hb group. There Guidelines have suggested a target haemoglobin
was no statistically significant difference in the RR of (Hb) >11 g/dl, with an average value of 12–12.5 g/dl
VA failure between patients with Hb 10–12 g/dl and (European Best Practice Guidelines) [1] and a
those with Hb 12–13 g/dl or >13 g/dl. Diabetes (RR: haematocrit of 33–36% (Hb 11–12 g/dl; NKF-DOQI
1.41, P ¼ 0.06), age >65 years (RR: 1.32; P ¼ 0.11) Guidelines) [2]. However, many studies have shown
and iPTH (RR: 1.56; P ¼ 0.01) were identified as that by maintaining the Hb of dialysis patients at nearly
predictive factors for VA failure; ACE inhibitors or normal levels, great advantages in terms of quality
ARB (RR: 0.69; P ¼ 0.03) were found to be protective of life [3], cardiac function [4], brain function [5],
factors. hospitalization and cost [6] can be achieved without
Conclusions. In the studied population, the correction significant adverse effects. Nevertheless, there has been
of Hb level to >12 g/dl was not associated with a some concern that full correction of anaemia may
higher incidence of VA thrombosis than in patients increase the risk of adverse effects, such as vascular
with Hb between 10 and 12 g/dl. ACE inhibitors access (VA) thrombosis, in dialysis patients [7]. The aim
or ARBs were found to be protective factors, and of this study was to determine if there is any relation-
ship between Hb levels and VA survival. As secondary
aims, other variables were examined to analyse their
Correspondence and offprint requests to: José M. Garrancho,
influence on VA survival: diabetes, age, gender, vas-
Fresenius Medical Care Spain, Madrid, Spain. cular disease (cerebrovascular disease, cardiovascular
Email: jmgarrancho@senefro.org or jmgarranchol@hotmail.com disease and limb amputation) and intact parathyroid
ß The Author [2005]. Published by Oxford University Press on behalf of ERA-EDTA. All rights reserved.
For Permissions, please email: journals.permissions@oxfordjournals.org
2454 J. M. Garrancho et al.
hormone (iPTH) at admission, and treatment with an they received kidney grafts, they died or had a functioning
angiotensin-converting enzyme (ACE) inhibitor or an VA on the final observation date (December 31, 2002). Cox’s
angiotensin receptor blocker (ARB) and rHuEPO proportional hazards regression was used to model time to
during follow-up. event as a function of Hb level.
The potential effect of the type of VA (fistula, graft or
permanent catheter) was evaluated both testing the effect of
Patients and methods this covariate and stratifying the models. Risk factors related
to patient demographics and co-morbid conditions were also
Subjects considered. All variables used in the equations were chosen
a priori and retained in the models if there was biological
Data for this study were collected in 40 dialysis units of the plausibility or if univariate analyses suggested that they might
Fresenius Medical Care Provider, Spain, by the European be associated with the event or might confound the relation-
Clinical Database (EuCliD) [8]. This database prospectively ship between the covariate of interest and the event. The
gathers demographic and clinical information on all end-stage proportional hazards assumption was checked for each model
renal disease (ESRD) patients admitted to haemodialysis by inspection of the complementary log-minus-log plots.
(HD), and it is continuously updated. It contains detailed A stepwise method was employed to obtain the best multi-
information on VA with respect to VA surgery date, type of variate model. The –2 log likelihood ratio (–2 Log L) statistics
access, access site, date of failure and all access-related were used for goodness of fit comparisons [8,9]. Estimated
complications. RRs and the hazards ratios (HR) for time to event analyses
We selected 1254 incident patients who had been on and their 95% confidence intervals (95% CIs) were calculated
renal replacement therapy for <1 month who started HD using estimated regression coefficients and their standard
with a permanent VA between January 2000 and December error. The contribution of covariates to explain the dependent
2002. Patients with either autologous fistulae (AVFs) or variable was assessed by means of a two-tailed Wald test, with
artificial grafts were considered. Patients with catheters were P<0.05 considered significant. The P-value for variable
not included in all analyses, because catheters are also removal within the multivariate analyses was set to 0.10.
occasionally used as temporary VA. All statistical analyses were performed using SPSS software,
version 11 (SPSS Inc., Chicago, IL).
Study variables and outcome definitions
The patients were categorized into four groups based on Results
their mean Hb levels during follow-up:<10 g/dl; 10–12 g/dl;
12–13 g/dl and >13 g/dl, and they were examined for their
VA survival and relative risk (RR) of access failure.
We studied 1254 patients, of whom 1057 had auto-
To test the predictive role of Hb level, survival analyses logous fistulae (84.6%) at admission, 75 (5.9%) had
were conducted using the date of first venipuncture for the grafts and 122 (9.5%) had permanent catheters. The
initiation of HD as time zero. For this reason, only patients population characteristics of the patients with AVFs
starting dialysis in the participating dialysis centres were and grafts, respectively, are shown in Table 1.
selected; and those starting dialysis with a temporary catheter A significantly higher proportion of females and
were excluded. Vascular access survival was defined as diabetics were found among the patients with grafts. In
the length of the intervention-free period to first failure. the same group, significantly more patients turned out
Permanent catheter failure was considered when it was to have high iPTH levels and to be on rHuEPO therapy.
removed for any reason. We found no significant differences between patients
Covariates selected a priori as possible risk factors for with AVF and those with grafts for the proportion of
VA failure included age at the beginning of HD, gender, elderly patients, patients on ACE inhibitors or ARBs
diabetic status and vascular disease (cerebrovascular disease, as well as for patients with vascular disease, even when
cardiovascular disease and limb amputation). These condi- considering cardiovascular disease, cerebrovascular
tions were defined based on the International Classification disease and limb amputation separately.
of Diseases, Tenth Revision. For example, patients were Median follow-up time was 9.34 months for all
considered to have ischaemic heart disease or cerebrovascular patients, 9.57 months for patients with AV fistulae,
disease if they had an associated diagnosis falling under
I20–I25 or under I63–I69, respectively. Diabetes, type 1 or 2,
was confirmed likewise, irrespective of the underlying renal Table 1. Population characteristics of patients with fistulas or grafts
disease.
The presence of increased baseline levels of iPTH Fistula (%) Graft (%) P-value
(>400 pg/ml), rHuEPO therapy or anti-hypertensive therapy
based on ACE inhibitors or ARBs were considered as possible Age >65 years 49.9 51.3 NS
risk factors for access failure. Females 33.3 61 0.001
Diabetes 24.5 33.8 0.037
Vascular disease 16.7 10.4 NS
Statistical analyses Cardiovascular disease 11.2 6.5 0.135
Cerebrovascular disease 4.5 2.6 0.332
Survival functions were described using the Kaplan–Meier Peripheral vascular disease 2.2 1.3 0.500
technique. The log-rank test was used for univariate iPTH >400 pg/ml 27.1 41.8 0.012
comparisons. Patients were censored when changed to ACE inhibitors or ARBs 33.2 32.5 NS
rHuEPO therapy 75.0 81.3 0.008
peritoneal dialysis, transferred to another dialysis unit, or if
Hb level and vascular access survival in HD patients 2455
10.34 months for patientes with grafts and 5.67 months (stratified by type of VA, Figure 2), the following major
for patients with permanent catheters. factors influencing VA survival were found: (i) treat-
The Kaplan–Meier analysis revealed a cumulative ment with ACE inhibitors or ARB was a protective
survival for fistulae of 87 and 80% after 12 and factor for VA survival; it reduced the RR of VA
24 months, respectively. The results for graft survival failure by 31% (95% CI 0.49–0.98; P ¼ 0.03); and
were 74 and 56% after 12 and 24 months, respectively (ii) older age, diabetes and high iPTH at admission
(Figure 1). were predictive of VA failure. The corresponding RR
Median cumulative survival was
5 months longer increased significantly by 32% (0.94–1.86; P ¼ 0.11),
for AVFs than for grafts [25.38 months (24.68–26.09) vs 41% (0.98–2.02; P ¼ 0.06) and 56% (1.10–2.21;
20.54 months (17.44–23.63)]. P ¼ 0.01), respectively.
The RR of VA failure based on Cox analysis was Gender, vascular disease and rHuEPO therapy were
2.3 (95% CI: 1.37–3.86) times higher with grafts than not significantly related to VA survival.
with fistulae (P ¼ 0.002).
The distribution of patients with grafts and fistulae
in the four Hb groups is summarized in Table 2. For Discussion
both types of VA, most patients fell in the group with
Hb 10–12 g/dl. This Hb group was chosen as a reference Any of the previously described advantages for dialysis
for RR analysis. patients of the complete correction of anaemia has to
The shortest VA cumulative survival was found in be balanced with the potential adverse effects of this
patients with Hb <10 g/dl (AVF 23.41 months; graft practice. In the context of our study, we focused parti-
survival 14.3 months). The RR of AVF failure in those cularly on the possibility of an increased incidence of
severely anaemic patients was 1.8 (2.86–1.13) times VA thrombosis. It is well known that complications
higher than in the patients with Hb between 10 and
12 g/dl (P ¼ 0.013). Patients with Hb between 12 and
13 g/dl had an RR of AVF failure of 1.32 (0.84–2.08;
Table 2. Haemoglobin level during follow-up in patients with
P ¼ 0.229), but patients with Hb >13 g/dl showed a fistulae or grafts
lower RR of fistula failure, 0.77 (0.31–1.91; P ¼ 0.585).
The RR of graft failure in severely anaemic patients No. of patients (%) No. of patients (%) P-value
(Hb <10 g/dl) was 1.61 (0.56–4.63) higher than in with a fistula with a graft
patients with Hb between 10 and 12 g/dl (P ¼ 0.37).
Patients with Hb between 12 and 13 g/dl had an RR Hb <10 g/dl 180 (17.0%) 18 (24%) NS
of graft failure of 0.98 (0.28–1.68; P ¼ 0.98; reference Hb 10–12 g/dl 647 (61.2%) 45 (60%) NS
Hb, 10–12 g/dl). Hb 12–13 g/dl 170 (16.1%) 10 (13.3%) NS
With regards to the risk factors for VA failure apart Hb >13 g/dl 60 (5.7%) 2 (2.6%) NS
Total 1057 (100%) 75 (100%) NS
from the Hb level, applying the Cox regression analysis

Fig. 1. Kaplan–Meier survival curve of different vascular accesses: fistulae, grafts and permanent catheters.
2456 J. M. Garrancho et al.
2.5

2
RR 95% CI

1.5

0.5
1.32 1.41 1.56 0.69
0
>65 years Diabetes iPTH > 400 ACE inh (or ARB)
Reference <65 years absence <400 pg/dL absence
P 0,114 0,06 0,01 0,03

Fig. 2. The relative risk of access failure associated with different parameters: age >65 years, diabetes, iPTH >400 pg/ml at admission, and
treatment with an ACE inhibitor or ARB (angiotensin receptor blocker). Cox regression analysis stratified by type of vascular access.

of VA are the primary causes of morbidity (hospita- Such results conflicting with ours might be explained
lizations) in HD patients. The matter is of utmost by the fact that most of the patients in the Besarab
importance in clinical practice; and any practices that study had grafts, whereas in our study most patients
increase the incidence of these complications must be had AVFs.
avoided. The potential association between VA throm- Another relevant difference between the Besarab
bosis, higher Hb levels and rHuEPO therapy, however, study and ours is the design of the study. The former
remains controversial. is an interventional trial, where patients are randomly
In our study, which evaluated an appropriately forced to reach a certain Hb level. Our study, however,
sized cohort of incident patients (1254), no significant is an epidemiological evaluation, where the target
differences were found in autologous fistula survival Hb is established by the European Best Practice
in patients with Hb levels between 10 and 12 g/dl Guidelines [1], but the physicians are not forced
and those with Hb between 12 and 13 g/dl or Hb to normalize Hb in patients resistant to correction.
>13 g/dl; therefore, we assume that normalization of On the other hand, our finding that severe anaemia
Hb levels was not an RR factor for fistula survival was associated with a higher risk of VA failure
in our population. Moreover, it was shown that the can be explained by the association of the anaemic
severely anaemic patients (Hb <10 g/dl) were the group state with inflammatory states and the presence of
with a statistically significant shorter fistula survival; co-morbidities in many HD patients. This hypothesis is
therefore, anaemia was a significant RR factor in keeping with the findings of a study by Miller et al.
(P ¼ 0.013) for fistula failure in our population. [10], who found a shorter graft survival in patients
Only 75 patients (5.9% of all our patients) had grafts with hypoalbuminaemia compared with patients with
as VA. Therefore, the statistical evaluations in this normal serum albumin levels. Patients with inflamma-
small group have less capacity to allow reliable con- tory states would have EPO resistance and, thus,
clusions. Nevertheless, again the severely anaemic lower Hb levels. Moreover, their inflammatory state
patients (Hb <10 g/dl) showed a higher RR of graft would increase the likelihood of access failure. These
failure than those with an Hb between 10 and 12 g/dl two opposing effects would be obscured in a random-
or >12 g/dl. ized study, given that no correlation between patient
Therefore, in our population, VA survival in non- Hb and the likelihood of access thrombosis was
anaemic patients was longer than in anaemic patients. observed. Accordingly, in another randomized con-
This result is in conflict with a previous large sample, trolled trial investigating the effect of Hb levels on
randomized clinical trial published by Besarab et al. left ventricular hypertrophy, Foley et al. [11] found
in 1998 [7]. They concluded by recommending, in their no difference in the incidence of arteriovenous access
‘Normal Hematocrit Cardiac Trial’ (NHCT), against thrombosis in patients with low Hb (9.5–10.5 g/dl)
the normalization of Hb levels in HD patients. In this compared with individuals with normal levels
well-controlled clinical study, a stratified cohort of 1265 (13.0–14.0 g/dl).
HD patients with clinically evident cardiac disease were Autologous fistulae are recommended as the first-
randomized to achieve a maintenance haematocrit of choice VA for HD patients [2], in preference to pros-
either 42±3 or 30±3%. Increased access thrombosis thetic grafts, because the former have better patency
was found in the cohort of patients randomized to rates and less need for corrective interventions, and are,
achieve the higher haematocrit of 42% (39 vs 29%; therefore, associated with significantly lower morbidity
P ¼ 0.001). However, VA thrombosis was not asso- and costs. Our results also strongly support the advan-
ciated with either the achieved haematocrit level or the tages of autologous fistulae vs grafts as VA for HD: our
dose of rHuEPO, making interpretation of these results study showed that the RR of graft failure was 2.3 times
difficult. higher (P ¼ 0.002) than autologous fistula failure,
Hb level and vascular access survival in HD patients 2457
and fistula survival was significantly longer than graft in our study. ACE inhibitors, already recommended
survival. for their protective effects on the heart, should also
Apart from anaemia correction, we found other be considered for their potential beneficial impact on
factors that could be influencing VA survival: diabetes VA survival, especially in older diabetic patients
and hyperparathyroidism were associated with a higher with high iPTH levels, because of their risk of VA
risk of VA failure; treatment with an ACE inhibitor failure. Moreover, in those patients, the VA has to be
or ARB was associated with longer VA survival. monitored carefully.
In addition to the classic atherogenic risk factors
(diabetes and age), iPTH has recently been suggested
as a potential cause of vascular disease. Grandaliano Conflict of interest statement. None declared.
et al. [12] investigated the relationship between fistula
dysfunction and mean plasma iPTH in 36 patients.
They found that patients with fistula failure had References
significantly higher mean plasma iPTH. The increased 1. European Best Practice Guidelines for the management of
RR of access failure that we found in patients with anaemia in patients with chronic renal failure. Nephrol Dial
baseline iPTH >400 pg/ml might be related to uncon- Transplant 1999; 14 [Suppl 5]: 1–50
trolled hyperparathyroidism during the pre-dialysis 2. NKF-K/DOQI Clinical Practice Guidelines for Vascular
period and the consequent vascular damage. Access: update 2000. Am J Kidney Dis 2001; 37 [1 Suppl 1]:
The protective effect of ACE inhibitors has already S137–S181
3. Furuland H, Linde T, Ahlmen J et al. A randomised controlled
been described in the literature. Saran et al. [13] found trial of haemoglobin normalization with epoetin alfa in pre-
a significant reduction of the risk of fistula failure dialysis and dialysis patients. Nephrol Dial Transplant 2003; 18:
with the use of ACE inhibitors (RR: 0.56; P ¼ 0.010) 353–361
when they investigated 900 fistulae in a prospective, 4. Harnett JD, Kent GM, Foley RN et al. Cardiac function and
observational study (DOPPS: Dialysis Outcomes and hematocrit level. Am J Kidney Dis 1995; 25 [4 Suppl 1]: S3–S7
Practice Pattern Study). In a series of 121 grafts, 5. Pickett JL, Theberge DC, Brown WS et al. Normalizing
hematocrit in dialysis patients improves brain function. Am J
Gradzki et al. [14] reported a substantially lower risk Kidney Dis 1999; 33: 1122–1130
of graft failure in patients who used ACE inhibitors, 6. Collins AJ, Li S, St Peter W et al. Death, hospitalization, and
compared with those who did not (RR: 0.32; economic associations among incident haemodialysis patients
P ¼ 0.003). Our findings are consistent with these with hematocrit values of 36 to 39%. J Am Soc Nephrol 2001;
results; consistent also in terms of the magnitude of 12: 2465–2473
the decrease of the RR, calculated by us to be
31%. 7. Besarab A, Bolton WK, Browne JK et al. The effects of normal
Several mechanisms have been proposed for this as compared with low hematocrit values in patients with
cardiac disease who are receiving haemodialysis and epoetin.
protective effect: inhibition of intimal hyperplasia, N Engl J Med 1998; 339: 584–590
vascular smooth muscle cell proliferation and migra- 8. Marcelli D, Kirchgessner J, Amato C et al. EuCliD (European
tion, and extracellular matrix deposition. In addition, Clinical Database): a database comparing different realities.
an ACE inhibitor or ARB may improve blood flow by J Nephrol 2001; 14 [Suppl 4]: S94–S101
increasing cardiac output associated with afterload 9. Marcelli D, Moscardó V, Steil H et al. Data management and
reduction. Quality Assurance for Dialysis Network. Contrib Nephrol 2002;
137: 293–299
10. Miller PE, Carlton D, Deierhoi MH, Redden DT, Allon M.
Natural history of arteriovenous grafts in hemodialysis
Conclusions patients. Am J Kidney Dis 2000; 36: 68–74
11. Foley RN, Parfrey PS, Morgan J et al. Effect of hemoglobin
In our population, anaemia (Hb <10 g/dl) was asso- levels in hemodialysis patients with asymptomatic cardio-
myopathy. Kidney Int 2000; 58: 1325–1335
ciated with VA failure. No significant differences were
12. Grandaliano G, Teutonico A, Allegretti A et al. The role of
found in fistula and graft survival between patients hyperparathyroidism, erythropoietin therapy, and CMV infec-
with nearly normal Hb and those with Hb between 10 tion in the failure of arteriovenous fistula in haemodialysis.
and 12 g/dl. Kidney Int 2003; 64: 715–719
Our study showed that HD patients should benefit 13. Saran R, Dykstra DM, Wolfe RA et al. Association between
from anaemia correction without incurring any vascular access failure and the use of specific drugs: the Dialysis
increased risk of VA failure. Outcomes and Practice Patterns Study (DOPPS). Dialysis
Outcomes and Practice Patterns Study. Am J Kidney Dis 2002;
A significantly protective effect of ACE inhibitors 40: 1255–1263
or ARBs on VA survival was detected. In contrast, 14. Gradzki R, Dhingra RK, Port FK et al. Use of ACE inhibitors
older age (>65 years), diabetes and high baseline is associated with prolonged survival of arteriovenous grafts.
iPTH were predictive risk factors for access failure Am J Kidney Dis 2001; 38: 1240–1244

Received for publication: 20.5.05