Full text of DOI:10.1053/ajkd.2002.29886

Hemodialysis Vascular Access Survival: Upper-Arm Native
Arteriovenous Fistula
Bradley S. Dixon, MD, Lisa Novak, BS, and Jerry Fangman, BS
● Achieving Dialysis Outcomes Quality Initiative guidelines for native arteriovenous fistulae using the radioce-
phalic forearm fistula (lower-arm fistula [LAF]) is difficult. This study reports results using the upper-arm native
arteriovenous fistula (UAF). From a prospective access database (1992 to 1998), this study was based on 204
patients (322 accesses). Average patient age was 56 ؎ 1 years, 63% were men, and 47% had diabetes. A native
fistula was the first access in 73% of patients (36%, LAFs; 37%, UAFs) and accounted for 48% of subsequent
accesses (13%, LAFs; 35%, UAFs). Younger men were more likely to receive an LAF, but there was no demographic
difference between patients receiving a UAF or arteriovenous graft (AVG). Both primary unassisted and cumulative
access patencies were significantly better for UAFs than either LAFs or AVGs. For first accesses, cumulative
access patency rates at 1, 3, and 5 years were 71%, 57%, and 57% for UAFs; 54%, 46%, and 36% for LAFs; and 54%,
28%, and 0% for AVGs (P < 0.01). Despite shorter access survival, AVGs required more total access procedures
than either UAFs or LAFs (procedures per access: 2.5, 1.0, and 0.6 for AVGs, UAFs, and LAFs, respectively). When
used, catheters were required for dialysis for a longer time for UAFs (median catheter days, 36, 53, and 56 for AVGs,
LAFs, and UAFs, respectively; P < 0.05). Access flow rates were greater in UAFs (1,247 mL/min; n ؍ 48; P < 0.01)
than AVGs (851 mL/min; n ؍ 30) or LAFs (938 mL/min; n ؍ 31). There was no evidence that UAFs were banded or
ligated for steal syndromes or heart failure more often than AVGs or LAFs. These results show that the UAF is a
good alternative to an AVG for achieving Dialysis Outcomes Quality Initiative guidelines.
© 2002 by the National Kidney Foundation, Inc.
INDEX WORDS: Surgical arteriovenous shunt; hemodialysis (HD); brachiocephalic fistula; access potency; access
complications.
ASCULAR ACCESS failure is a major
problem in providing care to patients on
outcome for native arteriovenous fistulae, the
trend in the United States has been to place more
prosthetic AVGs.7,8 By the most recent estimates,
prosthetic grafts account for 65% of accesses
used in the United States.8,9 By comparison,
prosthetic grafts are created in less than 35% of
hemodialysis patients in Canada and only 10%
of patients in Europe.6,8,9
V
hemodialysis therapy.1 A recent report analyzing
US Renal Data System (USRDS) data found an
overall primary unassisted access patency rate of
only 53% at 1 year.1 For prosthetic arteriovenous
grafts (AVGs), the 1-year primary unassisted
patency rate was only 49% compared with 62%
for native arteriovenous fistulae.1 Many recent
studies have confirmed the improved patency
rate for native arteriovenous fistulae compared
with prosthetic grafts.1-6 Despite this improved
Clinical practice guidelines for vascular ac-
cesses were recently established by the National
Kidney Foundation-Dialysis Outcomes Quality
Initiative.10 These guidelines encouraged the
placement of at least 50% native arteriovenous
fistulae. The forearm radiocephalic fistula is con-
sidered the optimal first choice for a native
fistula.10 Once established, it has the advantage
of good long-term survival, a low complication
rate, and preservation of more proximal sites in
the arm for future accesses.10 However, the radio-
cephalic fistula has the disadvantage that it is
more difficult to create and often fails to mature
in older patients, women, and those with signifi-
cant underlying vascular disease, particularly dia-
betics.11-15 The upper-arm native arteriovenous
fistula (UAF) offers an alternative surgical op-
tion for creating a native arteriovenous access in
these difficult patients.2,15-20 However, less infor-
mation is available on the long-term survival of
From the Department of Medicine, Division of Nephrol-
ogy, Veterans Affairs Medical Center, University of Iowa
College of Medicine, Iowa City, IA.
Received April 16, 2001; accepted in revised form July
27, 2001.
Supported in part by a Process Improvement Grant from
the University of Iowa Hospitals and Clinics.
Presented in part at the 32nd Annual Meeting of the
American Society of Nephrology, November 7, 1999, Miami
Beach, FL, and published in part in J Am Soc Nephrol
10:205A, 1999 (abstr).
Address reprint requests to Bradley S. Dixon, MD,
Division of Nephrology, Department of Medicine, Veter-
ans Affairs Medical Center, University of Iowa College
of Medicine, Iowa City, IA 52242-1081. E-mail:
bradley-dixon@uiowa.edu
© 2002 by the National Kidney Foundation, Inc.
0272-6386/02/3901-0013$35.00/0
doi:10.1053/ajkd.2002.29886
92
American Journal of Kidney Diseases, Vol 39, No 1 (January), 2002: pp 92-101

LONG-TERM SURVIVAL OF UPPER-ARM FISTULA
93
present analysis was limited to the period before January 1,
1999.
this type of access, and there is concern regard-
ing the potential for high-output failure and dis-
tal steal syndromes, making it is unclear how this
access should fit into clinical practice guide-
lines.10,11,15,16,18,21
Since 1991, our dialysis program has kept a
prospective database of all access-related proce-
dures performed at our institution. Our program
has a surgical emphasis on placing native fistulae
and particularly has emphasized the use of the
UAF for the last 5 years. The purpose of this
study is to analyze our database and compare
long-term survival, complications, and safety of
the UAF compared with forearm fistulae and
prosthetic grafts.
Data Coding and Outcome Measures
All vascular accesses created in the upper extremity and
followed up at our dialysis unit between January 1, 1992,
and January 1, 1999, were used for the present analysis.
Hospital billing records were used to cross-check the vascu-
lar access database for accuracy and completeness. Ten
patients in the database had their first access placed before
January 1, 1992, and were excluded. This left 204 patients
who received 322 accesses (64% native fistulae) and were
followed up at our center. Hospital charts of all 204 patients
were reviewed to confirm the accuracy, uniformity, and
completeness of the database for these subjects.
Any access attempt was recorded as a procedure. Primary
unassisted patency was defined as the time from access
creation to the first access failure, manifested by access
thrombosis or the need for an access-related procedure
(surgical revision, angioplasty, or thrombolysis). Cumula-
tive (assisted) patency was defined as the total time from
access creation to the time that the access could not be used
for dialysis, regardless of the number of procedures required
to maintain access patency. Primary access failure (failure to
mature) was defined as an access that failed before starting
dialysis therapy, or if the patient started on dialysis therapy,
had a catheter in place for dialysis for all but 7 days of the
cumulative life of the access. For accesses that failed to
mature before being used for dialysis, access failure was
defined by the earliest occurrence of either loss of a palpable
thrill on follow-up examinations or need to place a different
access for dialysis. All catheter placements and catheter
survival days were recorded. Catheter days were attributed
to an access only for the time that they both overlapped. To
determine comorbid conditions that might be associated
with access survival, we used hospital computer records
based on International Classification of Diseases, Ninth
Revision codes for patient diagnoses, including diabetes,
hypertension, peripheral vascular and carotid disease com-
bined, coronary disease, hyperlipidemia, and tobacco use.
METHODS
Study Setting
The setting is a single-center university-based teaching
hospital. Six staff surgeons accounted for 80% of all vascu-
lar access placements. Twenty-three accesses (7%) were
placed outside the university setting before transferring care
to our hemodialysis unit. The study population included all
patients seen at the university hospital who required access
placement for hemodialysis.
Vascular Access Database
Starting in 1991, a prospective database of all vascular
access–related procedures performed at the university hospi-
tal was established and maintained by dialysis unit staff.
Database entries were made for every new vascular access
placement, including catheters. Every subsequent revision
or procedure performed also was recorded. When an access
could no longer be used for dialysis, the date, as well as the
reason for discontinuation, was recorded. Patient status and
dates of important events also were recorded in the database,
including whether the patient received a transplant, was
transferred to another unit, died, or withdrew from dialysis
for another reason.
Access Flow Measurements
Flow measurements were performed monthly by a saline-
infusion ultrasound dilution technique using the Transonic
Hemodialysis Monitor (HD01; Transonic Systems Inc) and
dual flow/dilution sensors. At least two measurements at
each visit were performed on all patients. If measurements
varied by more than 10%, a third measurement was per-
formed, and results were averaged to provide the monthly
flow for each patient. To determine whether access flow rate
influenced adequacy of dialysis, we recorded the machine
blood flow rate at the time of the access measurement and
calculated the single-pool Kt/V (spKt/V) for urea performed
the same month as the access flow measurement. spKt/V
was calculated using a standard formula from the plasma
urea nitrogen level obtained predialysis and at 5 minutes
postdialysis with the machine in bypass.22
Decisions on Vascular Access Management
Selection of the location and type of access to be con-
structed was at the discretion of the attending surgeon.
Preoperative Doppler mapping was routinely performed to
assess venous anatomy before access creation. The preferred
forearm native fistula (LAF) was the radiocephalic fistula.
Surgical preference for the UAF was the end-to-side brachio-
cephalic fistula (93% of UAF). The most common configura-
tion for a prosthetic graft (AVG) was the polytetrafluoroeth-
ylene forearm loop. Starting in January 1999, detection of
access dysfunction was based on direct measurement of
access flow rates (Transonic Hemodialysis Monitor; Tran-
sonic Systems Inc, Ithaca, NY). Because this might influ-
ence the primary outcome variable, access patency, the

94
DIXON, NOVAK, AND FANGMAN
to compare log-transformed access flow data between access
type. Data for continuous variables is reported as mean Ϯ
SEM. P Ͻ 0.05 is considered statistically significant.
Statistical Analysis
Data were analyzed using StatView statistical software
(version 5.0; Abacus Concepts Inc, Berkeley, CA). Differ-
ences between the frequency of nominal variables were
compared using the chi-square statistic. Continuous vari-
ables were compared by analysis of variance (ANOVA).
Logistic regression analysis was used to determine whether
patient demographic and comorbidity factors varied as a
function of access number. Access patency and patient
survival data were analyzed using the Kaplan-Meier method,
with differences compared by the log-rank statistic. Patient
survival data were calculated from the date of starting
dialysis therapy and censored if a patient transferred to
another unit or initiated another form of renal replacement
therapy. Access patency was calculated from the date of each
access placement and censored if the patient had a functional
access at the time of transfer to another unit, transfer to an
alternate form of renal replacement therapy, or death. Covari-
ates for both patient and access survival were analyzed using
a Cox proportional hazards model. In this model, all nominal
patient demographic and comorbidity data were coded as
either present or absent. The continuous variable age was
analyzed per unit decade of life. Body mass index was
analyzed in several ways, but did not significantly affect
patient or access survival for any of these analyses. Results
for body mass index as a linear variable are reported. Data
for the number of access procedures per year were highly
skewed. We used Poisson regression analysis (performed
using S-Plus-2000; MathSoft, Seattle, WA) to compare the
number of access procedures per year for different access
types, recognizing that this may somewhat overstate the
observed level of significance for the statistic because the
number of access procedures in a given patient may be
correlated. To analyze access flow rates, we averaged monthly
flow rates for each patient, and this value was used in
subsequent analyses. The distribution of flow rates was
skewed, particularly for native fistulae. Log transformation
was used to help normalize flow rate data. ANOVA was used
RESULTS
Patient Characteristics
Patient characteristics at the time of first ac-
cess placement (204 patients) and for all access
placements combined (322 accesses) are listed in
Table 1. In comparison to incident patients re-
ported to the USRDS in 1995, our incident
patients overall were younger (56 versus 60
years), more frequently men (63% versus 53%),
mostly white (93% versus 62%), and had a
greater incidence of diabetes (47% versus 40%).23
When patients were further stratified by the type
of access they received, it was noted that patients
who received an LAF were more likely to be
younger and men than patients who received
either a UAF or AVG. Conversely, there was no
difference in demographic or comorbid character-
istics of patients who received either an AVG or
UAF.
Demographic and comorbid factors at the time
of all access placements combined are listed in
Table 2 and were generally similar to those for
first access placements. However, a diagnosis of
hyperlipidemia was significantly more common
in patients who received a native fistula (either
UAF or LAF) than patients who received an
AVG.
Table 1. Patient Characteristics Stratified by Access Type at Time of the First Access Placement
AVG
UAF
LAF
Total
(56 patients)
(75 patients)
(73 patients)
(204 patients)
Age (y)
58.1 Ϯ 2.3
59.1 Ϯ 1.9
51.6 Ϯ 2.3*
56.2 Ϯ 1.2
Men (%)
Race W/B/O (%)
Diabetes (%)
48
91/7/2
43
55
95/1/4
52
82†
92/4/4
45
63
93/4/3
47
Hypertension (%)
Peripheral vascular disease (%)
Coronary disease (%)
Tobacco (%)
77
21
38
20
87
28
51
15
92
15
37
21
86
22
42
18
Hyperlipidemia (%)
Body mass index (kg/m2)
21
29
37
30
27.3 Ϯ 0.9
25.4 Ϯ 0.7
26.3 Ϯ 0.6
26.2 Ϯ 0.4
NOTE. N ϭ 204 patients.
Abbreviations: W, white; B, black; O, other.
*P Ͻ 0.001 for LAFs compared with UAFs or AVGs.
†P Ͻ 0.01 for LAFs compared with UAFs or AVGs.

LONG-TERM SURVIVAL OF UPPER-ARM FISTULA
95
Table 2. Patient Characteristics Stratified by Access Type at Time of Each Access for All
Access Placements Combined
AVG
UAF
LAF
Total
(117 accesses)
(117 accesses)
(88 accesses)
(322 accesses)
Age (y)
59.3 Ϯ 1.6
57.8 Ϯ 1.6
50.9 Ϯ 1.9*
56.4 Ϯ 1.0
Men (%)
Race W/B/O (%)
Diabetes (%)
53
88/5/7
47
56
92/4/4
56
81†
89/6/5
47
61
90/5/5
50
Hypertension (%)
Peripheral vascular disease (%)
Coronary disease (%)
Tobacco (%)
81
26
43
15
89
28
49
13
90
16
39
18
86
24
44
15
Hyperlipidemia (%)
Body mass index (kg/m2)
17‡
26.9 Ϯ 0.6
26
36‡
26.8 Ϯ 0.6
26
25.5 Ϯ 0.6
26.3 Ϯ 0.3
NOTE. N ϭ 322 accesses.
Abbreviations: W, white; B, black; O, other.
*P Ͻ 0.001 for LAFs compared with UAFs or AVGs.
†P Ͻ 0.01 for LAFs compared with UAFs or AVGs.
‡P Ͻ 0.01 for LAFs or AVGs compared with UAFs.
Access Selection
surgeries between 1994 and 1995; rather, the
shift from LAFs to UAFs between 1994 and
1995 represented a shift in the choice of first
access placement by staff surgeons.
For first accesses, 36% were LAFs, 37% were
UAFs, and 27% were AVGs. For all subsequent
accesses, 52% were AVGs, 35% were UAFs, and
13% were LAFs. When first access placement
was analyzed by study year, it was apparent that
a significant shift in type of access placed oc-
curred between 1994 and 1995. Over the dura-
tion of the study, the number of grafts placed
remained relatively stable from year to year at
approximately 26% of the total. Conversely, be-
tween 1994 and 1995, there was a reversal in the
type of native AVF placed, away from LAFs to
more UAFs (Fig 1). There was no major change
in university surgical staff performing access
Access Survival
For the first access, median primary unassisted
patency was 11.5 months, and median cumula-
tive patency was 27.1 months. When patency of
the first access was stratified by access type, it
was found that both the primary unassisted and
cumulative patencies for UAFs were signifi-
cantly longer than for AVGs or LAFs (Fig 2).
Cumulative patency rates for UAFs were 71% at
1 year and 57% at 5 years compared with 54%
and 0% for AVGs, respectively (Fig 2B). The
survival curve for LAFs was intermediate be-
tween those for UAFs and grafts. Access patency
data for all 322 access placements combined showed
similar results as for first accesses (Table 3).
Primary access failure (failure to mature), de-
fined in the Methods, occurred in 31.5% of LAFs
(n ϭ 23), 22.6% of AVGs (n ϭ 13), and 28% of
UAFs (n ϭ 21; P ϭ not significant [NS] for
comparison between access types). By logistic
regression analysis, none of the measured covari-
ates was found to be a significant predictor of
primary failure. However, for first accesses, we
found a trend for age (relative risk [RR], 1.202
per decade; confidence interval [CI], 0.966 to
1.496; P ϭ 0.099), female sex (RR, 1.938; CI,
Fig 1. Variation in distribution of first access placed
by year of study. Data expressed as percentage of the
total number of first accesses placed each year. There
were significant differences in the distribution of fistu-
lae between LAFs and UAFs in 1993, 1994, 1995, and
1997 (P < 0.05). ( ) AVG; (■) UAF; ( ) LAF.

96
DIXON, NOVAK, AND FANGMAN
Although we did not identify differences in
demographic or comorbid factors between pa-
tients receiving a UAF compared with anAVG as
the first access, it is possible that differences in
access patency were caused by other unmeasured
factors. To further assess this possibility, we
examined the database and found that 11 patients
had received a UAF after having had an AVG as
either the first or second access that had failed.
We examined the cumulative patency of UAFs
(second or third access) placed after an AVG
(first or second access) and compared this with
the cumulative patency of all first or second
UAFs and found no difference in the patency of
UAFs placed after a previous AVG (access pa-
tency at 5 years, 70% for UAFs after AVGs
versus 59% for first and second UAFs; P ϭ NS).
This further suggests that underlying patient co-
morbid or demographic factors do not account
for the difference in patency between UAFs and
AVGs.
To determine whether the effect of access type
on access patency was independent of other
measured covariates, we analyzed the data using
a Cox proportional hazards model. As listed in
Table 4, the only factors that significantly influ-
enced primary unassisted patency of the first
access were type of vascular access and surgeon,
particularly those with fewer than 12 access
procedures. The strong effect of access type and
surgeon was observed for both primary and cumu-
lative access patency for both first accesses and
all accesses combined (Table 5). When all ac-
cesses were combined, a history of hyperlipid-
emia also was found to significantly affect cumu-
Fig 2. Kaplan-Meier analysis of access patency for
first vascular accesses stratified by access type. (A)
Primary unassisted access patency and (B) cumula-
tive (assisted) access patency. Dotted lines highlight
the median access patency for each of the three ac-
cess types. By log-rank test, primary unassisted pa-
tency was significantly better for UAFs compared with
either LAFs (P < 0.01) or AVGs (P < 0.001). Cumulative
patency for UAFs was significantly better than for
either LAFs or AVGs (P < 0.05).
Table 3. Primary Unassisted and Cumulative Access
Patency Rates at 1, 2, 3, and 5 Years After Creation for
All Accesses Combined
0.915 to 4.115; P ϭ 0.084), and hypertension
(RR, 2.881; CI, 0.858 to 9.670; P ϭ 0.087) to
predict primary failure. Diabetes and access type
were not found to be significant. Because ac-
cesses that fail to mature by definition have no
useful life, we reanalyzed survival data after
setting access survival for accesses that failed to
mature to zero. For first accesses, the cumulative
survival of UAFs was still better than either
AVGs or LAFs (1-, 3-, and 5-year survival rates
for UAFs, 69.7%, 59.5%, and 59.5%; for LAFs,
54.8%, 47%, and 37%; and for AVGs, 53.3%,
28.2%, and 0%, respectively; P ϭ 0.05 by log-
rank test).
Year 1
(%)
Year 2
(%)
Year 3
(%)
Year 5
(%)
Primary patency
AVG (117)
UAF (117)
27
62
44
44
13
48
40
33
7
37
31
24
0
32
22
19
LAF (88)
Overall (322)
Cumulative patency
AVG (117)
UAF (117)
LAF (88)
54
69
52
59
42
59
48
50
31
53
43
42
19
53
34
35
Overall (322)

LONG-TERM SURVIVAL OF UPPER-ARM FISTULA
97
Table 4. Analysis of Covariates for Primary Access
Patency by Cox Proportional Hazards Model
Table 1 that significantly influenced overall pa-
tient survival were older age (RR, 1.274 per
decade; CI, 1.079 to 1.503; P Ͻ 0.004) and the
presence of diabetes (RR, 1.785; CI, 1.073 to
2.950; P Ͻ 0.023). The presence of coronary
disease had an effect on survival that was of
borderline significance (RR, 1.52; CI, 0.933 to
2.488; P ϭ 0.093).
Covariate (all tested)
RR (95% CI)
P
Access type (reference ϭ
UAF)
AVG
LAF
2.61 (1.54-4.42)
1.82 (1.10-3.02)
0.0004
0.02
Surgeon (reference ϭ
surgeon A)
Surgeons Յ 12
procedures
Surgeon B
Surgeon C
Surgeon D
Access Procedures
3.00 (1.62-5.56)
1.37 (0.61-3.04)
1.23 (0.58-2.59)
1.24 (0.60-2.57)
1.61 (0.90-2.88)
1.10 (0.97-1.25)
1.08 (0.69-1.69)
0.0005
NS
NS
NS
NS
Despite shorter access survival, AVGs re-
quired more total access procedures than either
UAFs or LAFs (total procedures, including diag-
nostic fistulograms: 289 for 117 AVGs compared
with 116 for 117 UAFs and 53 for 88 LAFs).
Excluding diagnostic fistulograms, numbers of
procedures were 228, 76, and 26 for AVGs,
UAFs, and LAFs, respectively. When the total
number of procedures attributed to each access
per year is calculated, results are strikingly
skewed, with many accesses requiring no proce-
dures, whereas a few accesses require several
procedures over a very short life span. The me-
dian number of access procedures per year of
cumulative patency for an AVG was 2 compared
with 0 for both a UAF and LAF. The third
quartile was 6.19 procedures per year for an
AVG compared with 1.71 for a UAF and 0.95 for
an LAF. By Poisson regression analysis, the
number of access procedures for an AVG per
year was 3.84 (CI, 3.71 to 3.97) compared with
2.71 (CI, 2.65 to 2.78) for a UAF and 1.44 (CI,
1.39 to 1.50) for an LAF.
Surgeon E
Age (each 10 y)
Sex (women)
Race
Other
Black
Diabetes (present)
Hypertension (present)
PVD/carotid disease
(present)
Coronary disease
(present)
Tobacco (present)
Hyperlipidemia (present)
BMI
NS
NS
1.82 (0.63-5.29)
0.83 (0.29-2.46)
1.21 (0.81-1.83)
1.49 (0.81-2.73)
NS
NS
NS
NS
0.85 (0.51-1.40)
NS
1.10 (0.70-1.72)
0.82 (0.47-1.43)
0.88 (0.56-1.38)
0.99 (0.96-1.02)
NS
NS
NS
NS
NOTE. Data show RR with 95% CIs and P for the
contribution of various covariates to primary unassisted
patency of the first accesses.
Abbreviations: PVD, peripheral vascular disease; BMI,
body mass index.
lative access patency (Table 5). Diabetes had an
adverse effect on cumulative access patency that
did not reach criteria for statistical significance.
Table 5. Analysis of Covariates for Cumulative
Patency of All Accesses by Cox Proportional
Hazards Model
Patient Survival
We analyzed overall patient survival to deter-
mine whether patient survival was different when
stratified by access type. Overall cumulative pa-
tient survival from initiation of dialysis therapy
decreased nearly linearly, with a median survival
of 3.14 Ϯ 0.15 years (unadjusted death rate, 16.7
deaths/100 patient-years). Median survival for
patients who received an AVG as the first access
(2.59 Ϯ 0.52 years) was shorter than for those
who received a UAF (3.08 Ϯ 0.28 years) or LAF
(3.46 Ϯ 0.04 years). However, using a propor-
tional hazards model to correct for covariates,
patient survival was not significantly different
for any type of access. The only factors listed in
Covariate (significant only)
RR (95% CI)
P
Access type (reference ϭ
UAF)
AVG
LAF
1.84 (1.21-2.81) 0.005
2.20 (1.39-3.49) 0.0008
Surgeon (reference ϭ
surgeon A)
Surgeons Յ 15 procedures 1.76 (1.00-3.08) 0.05
Surgeon F
Hyperlipidemia (present)
Diabetes (present)
3.43 (1.59-7.39) 0.002
0.60 (0.39-0.94) 0.025
1.42 (0.98-2.05) 0.064
NOTE. Data show RR with 95% CIs and P for only the
statistically significant covariates.

98
DIXON, NOVAK, AND FANGMAN
patients; P Ͻ 0.01 for UAFs compared with
LAFs or AVGs). These log-transformed values
correspond to mean flow rates of 1,247, 938, and
851 mL/min for UAFs, LAFs, andAVGs, respec-
tively. Several patients with UAFs had measured
flow rates up to 4 L/min, and 15% of all UAFs
had access flow rates greater than 2,500 mL/min
compared with none of the patients with an AVG
or LAF (P Ͻ 0.01; Fig 3). However, when we
analyzed our data looking for the number of
accesses that were either banded or ligated for
steal or high-output syndromes, we found that
only 3 of 117 UAFs (2.6%) compared with 4 of
117 AVGs (3.4%) and 2 of 88 LAFs (2.3%) had
undergone this procedure (P ϭ NS).3
Access flow rates less than 400 mL/min were
found in 7.2% of all access measurements (11%,
4.4%, and 8.7% for AVGs, UAFs, and LAFs,
respectively). To determine whether there was an
effect of flow rate on adequacy of dialysis, we
analyzed spKt/V and machine blood flow rate.
spKt/V was measured the same month as the
access flow measurement, and the machine blood
flow rate was measured at the same time as the
access flow measurement. Comparing accesses
with flow rates greater than 400 mL/min (1,281 Ϯ
25 mL/min) versus those with flow rates of 400
mL/min or less (mean flow, 276 Ϯ 12 mL/min)
showed that both machine blood flow and spKt/V
were slightly but significantly lower in patients
with lower access flow rates (machine blood
flow, 407.4 Ϯ 1.2 mL/min; n ϭ 756 compared
with 395.2 Ϯ 6.3 mL/min; n ϭ 65; spKt/V,
1.51 Ϯ 0.007; n ϭ 880 versus 1.45 Ϯ 0.034; n ϭ
63; P Ͻ 0.05).
For grafts, the ability to maintain a good
machine flow rate and Kt/V even at a low access
flow rate was documented to be secondary to
midgraft stenoses. The midgraft stenosis is by-
passed by the arterial and venous needles during
dialysis, but is detected when the needles are
reversed to measure access flow rate. Similarly,
low flow rates in native fistulae may be caused
by collateral vessels that steal blood from the
vein when configured for access flow measure-
ment, but permit adequate delivery of blood flow
during dialysis. Overall, most patients with a low
access flow rate were receiving adequate dialy-
sis. This confirms that access blood flow is a
much more sensitive index of access dysfunction
than either machine blood flow or Kt/V.
Fig 3. Comparison of access flow rates for each
access type. Each point represents the average
monthly flow rate for a patient with the specified ac-
cess type. The horizontal line represents the median,
upper and lower limits of the box include the first and
third quartiles, and bars enclose 90% of data. Fifteen
percent of UAFs had access flow rates greater than
2,500 mL/min. Analysis of log-transformed data by
ANOVA showed that access flow rates for UAFs were
significantly greater than for either AVGs or LAFs (P <
0.001).
Catheter Requirements
Twenty-eight accesses failed before starting
dialysis therapy (14 LAFs, 7 UAFs, and 7 AVGs)
and did not require a catheter. In accesses that
were patent at the time of starting dialysis therapy,
a catheter was required most frequently forAVGs
(74%), followed by UAFs (64%) and LAFs (50%;
P Ͻ 0.05 for AVGs compared with UAFs or
LAFs and for UAFs compared with LAFs). For
accesses that required a catheter, the catheter
remained in for a longer time in UAFs compared
with AVGs (median number of catheter days, 36,
53, and 56 days for AVGs, LAFs, and UAFs,
respectively; P Ͻ 0.05 by Kuskal-Wallis test).
There was no significant difference in total cath-
eter days between LAFs and UAFs.
Access Flow Rate
One concern regarding UAFs is the presence
of greater flow rates, leading to possible in-
creased problems with high-output cardiac fail-
ure or vascular steal syndromes.18,21,24,25 We mea-
sured access flow rates in our patients on a
monthly basis. Log transformation of the mean
monthly flow rate for each patient was used for
statistical analysis. As shown in Fig 3, flow rates
for UAFs were significantly greater than those
for LAFs or AVGs (log average flow rate: UAF,
3.096 Ϯ 0.43 [SEM]; 48 patients; LAF, 2.972 Ϯ
0.046; 31 patients; AVG, 2.930 Ϯ 0.032; 30

LONG-TERM SURVIVAL OF UPPER-ARM FISTULA
DISCUSSION
99
ure rates for LAFs can be substantial, particu-
larly in women, older patients, and those with
diabetes.1,12-15 Given the trend for increasing age
and prevalence of diabetes in incipient hemodi-
alysis patients, this has raised questions about
whether LAFs should be the first choice of vascu-
lar access, particularly in these higher risk pa-
tients.12 A recent study by Miller et al15 reported
adequacy rates for forearm fistulae of only 7%
for women, 12% in patients 65 years or older,
and 21% for patients with diabetes. Conversely,
they reported adequacy rates for UAFs of 56% in
women, 54% in older patients, and 48% in pa-
tients with diabetes.15 Our results are generally
consistent with these observations and suggest
that in selected patients, a UAF may be the most
appropriate choice for a first access and is a
better alternative than a graft for a second access
attempt.
A recent review of hemodialysis vascular ac-
cess management recommended the UAF as an
alternative to a graft for a second vascular access
procedure.11 However, concern also was ex-
pressed over the lack of knowledge on long-term
survival and possible complications related to
high access flow rates.11 Subsequent studies have
shown that UAFs can be used effectively to
increase the prevalence of native fistulae and
improve short-term outcome, even in women,
older patients, and those with diabetes, for whom
an LAF is difficult to establish.15,26 The present
prospective observational study further confirms
these recent findings and provides additional
information on long-term access survival and
complications of using a UAF, specifically the
end-to-side brachiocephalic fistula.
Survival rates for native fistulae in the present
study are similar to those reported in the recent
literature. Patency rates for UAFs have been
reported in several studies.16-18,20,27,28 For native
brachiocephalic UAFs, cumulative 1- and 3-year
patency rates have been reported to be between
67% to 84% and 50% to 78%, respectively.16-18,27
Similar patencies have been reported with the
transposed brachiobasilic fistula.17,20,28 These ob-
servations are similar to those reported for UAFs
in the present study (71% and 57% for 1 and 3
years, respectively; Table 3) and suggest that the
observed patency for UAFs may be broadly
attainable. Patency of LAFs depends on patient
selection. However, several recent studies report
1- and 3-year patency rates (primary unassisted
and cumulative) of 48% to 69% and 36% to 48%
for LAFs, respectively.5,12,14,29 These are similar
to results found in the present study for LAFs
(Table 3). By comparison, 1-year cumulative
patency rates for prosthetic grafts have been
reported to be 40% to 71% in recent stud-
ies.3,5,26,30,31
Because none of the available studies are
randomized, concerns may be raised about the
potential for selection bias to influence the com-
parison of patency data. In the present study, bias
was clearly present in the selection of patients
(younger men) for an LAF. However, there was
no evidence for any difference in demographic
characteristics between patients receiving a UAF
or AVG. It is possible that vessel anatomy or
other patient characteristics that were not re-
corded nevertheless influenced access selection
and might have affected access survival. How-
ever, the patency of a UAF placed after a failed
AVG was the same as a UAF placed without a
previous AVG. These observations suggest that
selection bias was not the major reason for the
better access patency of UAFs compared with
AVGs.
Using a Cox proportional hazards model to
correct for possible differences in measured co-
variates, the only factors that significantly influ-
enced access patency in the present study were
type of access selected and surgeon. The influ-
ence of the operating surgeon was previously
reported by Prischl et al14 to be the major factor
in determining both short-term and long-term
patency of radiocephalic LAFs. The effect of the
surgeon on access patency has not been analyzed
in most recent series; however, it likely contrib-
utes in part to the variation in access patency
reported in the literature. Other factors that may
influence access patency include patient age,
There are no randomized trials comparing the
patency of native fistulae to grafts. However,
numerous (but not all) observational studies have
reported improved patency for native fistulae
compared with grafts.1-4,6,19,21,26,32-34 The most
commonly used native fistula in these studies has
been the LAF. Once established, the long-term
complication-free patency of an LAF is much
better than grafts.1,3,4,19,21,32,35 However, early fail-

100
DIXON, NOVAK, AND FANGMAN
diabetes, and sex; however, the effect of these
factors has not been consistent.1,5,6,14,19,29,30,33 In
the present study, none of these factors was
found to affect primary unassisted patency (Table
4). However, the presence of diabetes was found
to have a borderline effect to adversely affect
cumulative patency, particularly in subsequent ac-
cess attempts, many of which were grafts (Table 5).
Currently, prosthetic AVGs account for 65%
of accesses used in the United States.8,9 If this
trend was reversed and the majority of accesses
placed were native fistulae, what survival benefit
would result? Woods et al1 recently analyzed 784
randomly selected incident hemodialysis pa-
tients in the USRDS database (69% AVGs and
31% native fistulae, primarily LAFs) and found
an overall 1-year primary unassisted patency rate
of 53%. Their study understates the overall mag-
nitude of the access problem because only pa-
tients with a functional access within 30 to 120
days after starting dialysis therapy were included
on the study.1 We reanalyzed our data to exclude
patients who did not have a functional access at
30, 60, or 120 days after starting dialysis therapy,
and the revised average 1-year primary unas-
sisted patency rates were 68.9%, 71.6%, and
75.7%, respectively. This suggests that an in-
crease in the percentage of first accesses that are
native fistulae from the current level of 31%
reported by Woods et al1 to 72.5% in the present
study would result in an overall increase in the
1-year primary unassisted access patency rate of
16% to 23%. Moreover, given that survival curves
for native fistulae continued to diverge beyond 1
year (Fig 2), overall long-term access patency
would be even better. Finally, the significantly
fewer number of procedures required to maintain
a native fistula compared with an AVG would
further reduce overall access-related costs.
are obese. However, given the large difference in
patency and complications for the two types of
accesses, we do not believe that this is a good
argument against using a UAF. Regarding the
problem of high-output failure and distal isch-
emia, Zibari et al21 reported that 8% of brachioce-
phalic UAFs required ligation for steal compared
with only 1.7% for LAFs and 1.8% for AVGs. A
more recent report found that 4% of native fistu-
lae (53% UAFs and 47% LAFs combined) were
ligated for steal syndrome compared with 0.8%
for grafts.26 However, several other studies found
no increased risk for high-output failure or steal
syndromes with UAFs.16,27 Although our prospec-
tive flow monitoring over the last 2 years identi-
fied a subgroup of patients with very high fistula
flow rates, there was no increase in the percent-
age of UAF accesses ligated for steal syndromes
in the preceding 7 years of this study. Moreover,
after correcting for covariates, the survival rate
for patients with a UAF was no worse than those
for patients with anAVG or LAF. Further prospec-
tive investigation is needed into the long-term
outcome of patients with these high access flow
rates. However, in our experience, access-related
ischemia is not sufficiently common to proscribe
routine placement of UAFs.
ACKNOWLEDGMENT
The authors thank Dr Lawrence G. Hunsicker for
assistance with statistical analysis; Drs You Min Wu and
Stephen C. Rayhill for helpful discussions on surgical
techniques; and Drs William J. Lawton and John B.
Stokes for enthusiastic encouragement and guidance for
these studies and analysis.
REFERENCES
1. Woods JD, Turenne MN, Strawderman RL, Young
EW, Hirth RA, Port FK, Held PJ: Vascular access survival
among incident hemodialysis patients in the United States.
Am J Kidney Dis 30:50-57, 1997
Several concerns have been raised regarding
the use of UAFs, including loss of more distal
access sites, difficulty cannulating the access,
and the potential for greater access flow rates,
leading to high-output failure or problems with
ischemia in the distal extremity.10,18,21,24,25 In
general, the most distal sites should be used for
placement of the native access, but an LAF is not
possible in many current patients. In these pa-
tients, use of a UAF would be preferred over an
AVG. Cannulation of a UAF can be more diffi-
cult than an AVG, particularly in patients who
2. Bender MH, Bruyninckx CM, Gerlag PG: The brachio-
cephalic elbow fistula: A useful alternative angioaccess for
permanent hemodialysis. J Vasc Surg 20:808-813, 1994
3. Burger H, Kluchert BA, Kootstra G, Kitslaar PJ, Ub-
bink DT: Survival of arteriovenous fistulas and shunts for
haemodialysis. Eur J Surg 161:327-334, 1995
4. Chazan JA, London MR, Pono LM: Long-term sur-
vival of vascular accesses in a large chronic hemodialysis
population. Nephron 69:228-233, 1995
5. Kalman PG, Pope M, Bhola C, Richardson R, Snider-
man KW: A practical approach to vascular access for hemo-
dialysis and predictors of success. J Vasc Surg 30:727-733,
1999
6. Churchill DN, Taylor DW, Cook RJ, LaPlante P, Barre

LONG-TERM SURVIVAL OF UPPER-ARM FISTULA
101
P, Cartier P, Fay WP, Goldstein MB, Jindal K, Mandin H,
McKenzie JK, Muirhead N, Parfrey PS, Posen GA, Slaugh-
ter D, Ulan RA, Werb R: Canadian Hemodialysis Morbidity
Study. Am J Kidney Dis 19:214-234, 1992
21. Zibari GB, Rohr MS, Landreneau MD, Bridges RM,
DeVault GA, Petty FH, Costley KJ, Brown ST, McDonald
JC: Complications from permanent hemodialysis vascular
access. Surgery 104:681-686, 1988
7. Windus DW: Permanent vascular access: A nephrolo-
gist’s view. Am J Kidney Dis 21:457-471, 1993
8. Hirth RA, Turenne MN, Woods JD, Young EW, Port
FK, Pauly MV, Held PJ: Predictors of type of vascular
access in hemodialysis patients. JAMA276:1303-1308, 1996
9. Pisoni RL, Young EW, Greenwood RN, Hecking E,
Goodkin DA, Wolfe RA, Held PJ: Comparison of vascular
access use in the US and Europe: The Dialysis Outcomes
and Practice Patterns Study (DOPPS). J Am Soc Nephrol
10:215A, 1999 (abstr)
10. National Kidney Foundation: DOQI Clinical Practice
Guidelines for Vascular Access. Am J Kidney Dis 30:S150-
S191, 1997 (suppl 3)
11. Feldman HI, Kobrin S, Wasserstein A: Hemodialysis
vascular access morbidity. J Am Soc Nephrol 7:523-535,
1996
12. Leapman SB, Boyle M, Pescovitz MD, Milgrom ML,
Jindal RM, Filo RS: The arteriovenous fistula for hemodialy-
sis access: Gold standard or archaic relic? Am Surg 62:652-
656 1996
13. Ifudu O, Macey LJ, Homel P, Hyppolite JC, Hong J,
Sumrani N, Distant D, Sommer BG, Friedman EA: Determi-
nants of type of initial hemodialysis vascular access. Am J
Nephrol 17:425-427, 1997
14. Prischl FC, Kirchgatterer A, Brandstatter E, Wallner
M, Baldinger C, Roithinger FX, Kramar R: Parameters of
prognostic relevance to the patency of vascular access in
hemodialysis patients. J Am Soc Nephrol 6:1613-1618,
1995
15. Miller PE, Tolwani A, Luscy CP, Deierhoi MH,
Bailey R, Redden DT, Allon M: Predictors of adequacy of
arteriovenous fistulas in hemodialysis patients. Kidney Int
56:275-280, 1999
22. Garred LJ, Barichello DL, DiGiuseppe B, McCready
WG, Canaud BC: Simple Kt/V formulas based on urea mass
balance theory. ASAIO J 40:997-1004, 1994
23. US Renal Data System: Excerps From the US Renal
Data System 1997 Annual Data Report. Am J Kidney Dis
30:S40-S53, 1997 (suppl 1)
24. Young PR Jr, Rohr MS, Marterre WF Jr: High-output
cardiac failure secondary to a brachiocephalic arteriovenous
hemodialysis fistula: Two cases. Am Surg 64:239-241, 1998
25. Burkhart HM, Cikrit DF: Arteriovenous fistulae for
hemodialysis. Semin Vasc Surg 10:162-165, 1997
26. Ascher E, Gade P, Hingorani A, Mazzariol F, Gunduz
Y, Fodera M, Yorkovich W: Changes in the practice of
angioaccess surgery: Impact of Dialysis Outcomes Quality
Initiative recommendations. J Vasc Surg 31:84-92, 2000
27. Bender MH, Bruyninckx CM, Gerlag PG: The Gracz
arteriovenous fistula evaluated. Results of the brachioce-
phalic elbow fistula in haemodialysis angio-access. Eur J
Vasc Endovasc Surg 10:294-297, 1995
28. HakaimAG, Nalbandian M, Scott T: Superior matura-
tion and patency of primary brachiocephalic and transposed
basilic vein arteriovenous fistulae in patients with diabetes. J
Vasc Surg 27:154-157, 1998
29. Golledge J, Smith CJ, Emery J, Farrington K, Thomp-
son HH: Outcome of primary radiocephalic fistula for haemo-
dialysis. Br J Surg 86:211-216, 1999
30. Cinat ME, Hopkins J, Wilson SE: A prospective
evaluation of PTFE graft patency and surveillance tech-
niques in hemodialysis access. Ann Vasc Surg 13:191-198,
1999
31. Bosman PJ, Blankestijn PJ, van der Graaf Y, Heintjes
RJ, Koomans HA, Eikelboom BC: A comparison between
PTFE and denatured homologous vein grafts for haemodialy-
sis access: A prospective randomised multicentre trial. The
SMASH Study Group. Study of Graft Materials in Access
for Haemodialysis. Eur J Vasc Endovasc Surg 16:126-132,
1998
16. Livingston CK, Potts JR III: Upper arm arteriovenous
fistulas as a reliable access alternative for patients requiring
chronic hemodialysis. Am Surg 65:1038-1042, 1999
17. Cantelmo NL, LoGerfo FW, Menzoian JO: Brachio-
basilic and brachiocephalic fistulas as secondary angioac-
cess routes. Surg Gynecol Obstet 155:545-548, 1982
18. Dunlop MG, Mackinlay JY, Jenkins AM: Vascular
access: Experience with the brachiocephalic fistula. Ann R
Coll Surg Engl 68:203-206, 1986
32. Winsett OE, Wolma FJ: Complications of vascular
access for hemodialysis. South Med J 78:513-517, 1985
33. Culp K, Flanigan M, Taylor L, Rothstein M: Vascular
access thrombosis in new hemodialysis patients. Am J Kid-
ney Dis 26:341-346, 1995
19. Lazarides MK, Iatrou CE, Karanikas ID, Kaperonis
NM, Petras DI, Zirogiannis PN, Dayantas JN: Factors affect-
ing the lifespan of autologous and synthetic arteriovenous
access routes for haemodialysis. Eur J Surg 162:297-301,
1996
20. Coburn MC, Carney WI Jr: Comparison of basilic
vein and polytetrafluoroethylene for brachial arteriovenous
fistula. J Vasc Surg 20:896-902, 1994
34. Palder SB, Kirkman RL, Whittemore AD, Hakim
RM, Lazarus JM, Tilney NL: Vascular access for hemodialy-
sis. Patency rates and results of revision. Ann Surg 202:235-
239, 1985
35. Kherlakian GM, Roedersheimer LR, Arbaugh JJ,
Newmark KJ, King LR: Comparison of autogenous fistula
versus expanded polytetrafluoroethylene graft fistula for
angioaccess in hemodialysis. Am J Surg 152:238-243, 1986