* Corresponding author. Tel.

: #44-151-706-5606; fax: #44-151-

706-5803.
E-mail address: thienhow@liv.ac.uk (T.V. How)
Journal of Biomechanics 32 (1999) 915}925
Flow patterns in the radiocephalic arteriovenous "stula:
an in vitro study
S. Sivanesan, T.V. How*, R.A. Black, A. Bakran
Department of Clinical Engineering, University of Liverpool, Liverpool L69 3GA, UK
Renal Transplant Unit, Royal Liverpool University Hospital, Prescot Street, Liverpool L7 8XP, UK
Received 15 April 1999
Abstract
A signi"cant number of late failures of arteriovenous "stulae for haemodialysis access are related to the progression of intimal
hyperplasia. Although the aetiology of this process is still unknown, the geometry of the "stula and the local haemodynamics are
thought to be contributory factors. An in-vitro study was carried out to investigate the local haemodynamics in a model of
a Cimino-Brescia arteriovenous (AV) "stula with a 303 anastomotic angle and vein-to-artery diameter ratio of 1.6. Flow patterns were
obtained by planar illumination of micro-particles suspended in the #uid. Steady and pulsatile #ow studies were performed over
a range of #ow conditions corresponding to those recorded in patients. Quantitative measurements of wall shear stress and turbulence
were made using laser Doppler anemometry. The #ow structures in pulsatile #ow were similar to those seen in steady #ow with no
signi"cant qualitative changes over the cardiac cycle. This was probably the result of the low pulsatility index of the #ow waveform in
AV "stulae. Turbulence was the dominant feature in the vein, with relative turbulence intensity '0.5 within 10 mm of the suture line
decreasing to a relatively constant value of about 0.10}0.15 between 40 and 70 mm from the suture line. Peak and mean Reynolds
shear stress of 15 and 20 N/m, respectively, were recorded at the suture line. On the #oor of the artery, peak values of temporal mean
and oscillating wall shear stress of 9.22 and 29.8 N/m, respectively. In the vein, both mean and oscillating wall shear stress decreased
with distance from the anastomosis. 1999 Elsevier Science Ltd. All rights reserved.
Keywords: Vascular access; Haemodialysis; Haemodynamics; Turbulence; Shear stress
1. Introduction
Long-term haemodialysis of patients with chronic
renal failure is dependent on maintaining a patent
and functioning vascular access. The Cimino-Brescia
arteriovenous (AV) "stula is widely accepted as the best
form of access in these patients. It involves the anastomo-
sis of the cephalic vein and radial artery at the wrist, and
is usually constructed in end vein-to-side artery con"g-
uration. The lifetime of such a "stula can be anywhere
between months and tens of years, in a few cases. One of
the main causes of late failure is thrombosis secondary to
the development of intimal hyperplasia, a "bro-muscular
thickening of the vessel wall that gives rise to localised
stenoses in the "stula (Kinnaert et al., 1971; Wedgewood
et al., 1984; Sivanesan et al., 1998). The behaviour of
vascular smooth muscle cells (SMCs) depends on a deli-
cate balance between growth factors and inhibitors se-
creted by cells in the circulation and those in the vessel
wall. The endothelium is thought to play a key role in
determining the response of the vessel wall, with intimal
hyperplasia being the result of endothelial dysfunction
(Ross, 1986). The tendency for stenotic lesions to develop
at speci"c sites suggests that the geometry of the "stula
and the resulting local haemodynamic conditions have
a role to play in the aetiology of this disease.
Endothelial cells are known to be sensitive to #uid
shear forces, and wall shear stress in particular. In dis-
turbed #ows, these cells * and those within the #owing
blood * are subjected to a wide range of shear stresses.
Both high and low shear in#uence the ways these
cells respond to local haemodynamic conditions. High
shear stress can damage the endothelial layer, which, in
turn, stimulates the underlying SMCs to proliferate.
0021-9290/99/$- see front matter 1999 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 2 1 - 9 2 9 0 ( 9 9 ) 0 0 0 8 8 - 3
Fig. 1. Schematic diagram of experimental set-up for #ow visualisation and LDA measurement. The directions of #ow through the "stula, for the cases
of retrograde and antegrade #ow in distal artery, are indicated by arrows. The pulsatile #ow consisted of a steady and oscillatory #ow component
which were combined in a calming chamber. The steady #ow was provided by a constant-head tank supplied by a peristaltic pump and the oscillatory
component was produced by a computer-controlled piston pump.
Moreover, removal of the endothelium also exposes the
vessel wall to leukocytes that adhere to the vessel wall
and act as sources of SMC mitogens. Cellular material is
more likely to be deposited in regions of #ow separation,
recirculation and stagnation, where the residence time of
#uid is excessive.
Postoperative follow-up of AV "stulae using duplex
ultrasonography can provide information on "stula
maturation, measure "stula #ow rates and highlight re-
gions of disturbed #ow. Because of the limitations in the
temporal and spatial resolution of current duplex scan-
ners, however, little can be learnt about the nature of
these #ow disturbances. Hence the need for in vitro
modelling, for a better understanding of the #ow struc-
tures in AV "stulae and their possible contribution to the
development of stenoses. This paper describes #ow vis-
ualisation and laser Doppler anemometer (LDA) studies
carried out in a model of an AV "stula under di!erent
#ow conditions.
2. Materials and methods
2.1. Fistula model
A three-dimensional computer model of the "stula was
created using a CAD/CAM package (DUCT 5.1, Delcam
International, Birmingham, UK). The model was based
on the geometrical data obtained from bi-planar "stulo-
grams and from casts of AV "stulae constructed in vitro.
The diameter of the vein and artery segments was 7.0 and
4.5 mm, respectively (diameter ratio of 1.6) and the anas-
tomotic angle was 303. From the geometrical data a solid
master was fabricated from epoxy resin in a laser
stereolithography system. The epoxy model was used to
create a silicone rubber mould into which molten wax
was injected to produce a male cast. This wax cast was
then embedded in clear silicone rubber (Sylgard, 184,
Dow Corning) and, after curing, the wax was melted out,
leaving the transparent "stula model used in the #ow
experiments.
2.2. Flow circuit
The "stula model was used in a #ow circuit shown
diagramatically in Fig. 1. Steady #ow was provided by
a constant-head tank supplied by a peristaltic pump and
oscillatory #ow was produced by a purpose-built servo-
controlled piston-pump (Hughes et al., 1996). The steady
and oscillatory components were added in a calming
chamber which supplied the artery. To simulate retro-
grade #ow in the distal artery (i.e. #ow direction from
hand to "stula), the #uid entered the "stula model
through both the proximal and distal artery segments
and exited through the vein, as shown in Fig. 1. For the
case of antegrade #ow in the distal artery (i.e. #ow direc-
tion towards the hand), the #uid entered the proximal
artery and returned to the reservoir through the distal
artery and the vein. Flow control valves were provided to
set the resistances in the artery and vein and to set the
#ow ratio. A two-channel transit-time ultrasound #ow-
meter (Transonic HT107, Transonic Systems Inc., NY,
USA) provided with two 6 mm diameter cannulating
#owprobes was used to monitor the #ow in the vein and
distal artery.
2.3. Fluids
The #uid used for #ow visualisation was a solution of
40% (w/w) glycerol and 7% (w/w) sodium chloride in
916 S. Sivanesan et al. / Journal of Biomechanics 32 (1999) 915}925
Fig. 2. Example of venous #ow waveform used in the pulsatile #ow
experiments. The waveform was derived from intraoperative blood #ow
recordings in patients with radiocephalic AV "stulae.
distilled water. Light-scattering particles (Rilsan Blue, Elf
AtoChem. UK Ltd.) of between 75 and 150 m in dia-
meter were suspended in the analogue #uid as tracer
particles (0.2%, w/w). The kinematic viscosity of this #uid
was 3.1;10\ m/s.
For LDA measurements, it was necessary to increase
the refractive index of the #uid to the same value as that
of the silicone rubber (RI"1.41) in order to minimise
errors in the positioning of the LDA measuring volumes.
This was achieved by increasing the concentration of
glycerol in the solution to 46% (w/w) and that of sodium
chloride to 14% (w/w). The kinematic viscosity and den-
sity of this #uid at room temperature increased to
5.7;10\ m/s and 1190 kg/m, respectively. To extend
the working life of the solution, 0.15% (w/w) of sodium
azide was dissolved in the #uid. The sodium azide sus-
pension was found to contain su$cient scattering par-
ticles for LDA measurement, resulting in high data rates
and high validation percentages.
2.4. Flow visualisation
Flow patterns were obtained by the system described
previously (Hughes et al., 1996). Brie#y, the model was
illuminated with a 16 mW HeNe laser provided with
a line generator and the #ow patterns were photo-
graphed with a 35 mm camera and also recorded on
S-VHS video tape using a CCD camera (Hitachi KP-
M1E/K) and framegrabber (IP8 2M-C, Matrox UK Ltd).
The model was illuminated in the symmetry plane (x}y)
of the anastomosis, and at right angles to the symmetry
plane but parallel to the long axis of the artery (x}z). The
#ow was visualised under steady and pulsatile #ow con-
ditions over a range of Reynolds numbers (Re) relevant
to those recorded in patients (i.e. 50(Re(1300). The
#ow waveform shown in Fig. 2 is typical of those re-
corded in actual AV "stulae. The pulsatility index (PI,
de"ned as the peak-to-peak #ow %mean #ow) was 0.44.
Experiments were performed using a pulse period of 1.0
and 1.6 s corresponding to Womersley's parameter in
the vein of 5.0 and 3.9, respectively.
2.5. Laser Doppler anemometry
A 2-component LDA system (55X FiberFlow Series,
Dantec) was used to measure the axial and radial velocity
components in the symmetry plane of the anastomosis,
along the #oor of the artery, along the centreline of the
vein and across the venous lumen at various locations
from the anastomosis. LDA experiments were performed
with pulsatile #ow with mean Reynolds number (Re) of
600 in the vein, PI of 0.44, pulse period of 1.0 s and
Womersley's parameter of 3.7. Velocity data were re-
corded for 100 cycles with at least 20,000 validated bursts
at each measurement position.
The near-wall velocity data were sorted into 10 ms
time bins. The data in each bin were averaged to produce
the mean velocity waveform. The wall shear rate at each
location was estimated from the gradient of the near-wall
velocity pro"le that was created by a second-order poly-
nomial "t of the "ve measurement points adjacent to the
wall. The latter were acquired with a spacing of 25 m.
Fig. 3 shows the locations in the AV "stula model where
velocity pro"les and wall shear estimates were obtained.
Turbulence is an important feature of #ow in AV
"stulae and the intensity of turbulence de"ned as the
root-mean-square value of the #uctuating component of
velocity was determined at various locations in the vein.
The #uctuating velocity component in pulsatile #ow is
normally determined by ensemble averaging (Casty and
Giddens, 1984). Because of the random arrival time of
particles passing through the sample volume and the
pulsatile nature of the #ow, the following procedure was
used to calculate turbulence intensity. The velocity data
were sorted into 5 ms bins and averaged in each bin to
produce the mean velocity waveform. The variation be-
tween the instantaneous and the mean velocity (as esti-
mated from a cubic spline "t of the mean velocity
waveform) was calculated for each velocity measurement;
the squared velocity variations were averaged in their
individual time bins and the square root value taken in
each bin to produce the temporal root-mean-square velo-
city variation, or the turbulence intensity. The calcu-
lation was repeated for both axial and radial components
of #ow.
The Reynolds shear stress (the theoretical shear stress
that acts tangentially on a surface of an element of the
#uid) was calculated using a similar process. Here, the
velocity data were "ltered using a narrow arrival time
coincidence window (43 s) to ensure that the axial
and radial data remaining were coincident in time. The
S. Sivanesan et al. / Journal of Biomechanics 32 (1999) 915}925 917
Fig. 3. Diagram of the AV "stula showing the locations of velocity measurement across the vein (A}E). Wall shear measurements on the #oor of the
artery extended 20 mm on either side of the centre of the anastomosis.
Fig. 4. Flow patterns recorded in x}y (top) and x}z (bottom) planes
under steady #ow conditions with the distal artery occluded:
(a) Re"100, (b) Re"300 and (c) Re"600.
"ltered velocity data were sorted into 5 ms time bins. The
variation between the instantaneous and mean velocities
was calculated for each of the "ltered velocity measure-
ments. The `simultaneousa #uctuating part of the velo-
city components were then multiplied together and
multiplied by the density of the #uid to yield the
Reynolds shear stress. The mean and peak values of
Reynolds shear stress were determined at various distan-
ces along the centreline of the vein.
3. Results
3.1. Flow visualisation
The results under steady #ow are "rst presented to
show the e!ects of Reynolds number, #ow ratio and #ow
direction on the #ow patterns. Pulsatile #ow patterns at
a Re of 600 in the vein (the average encountered in vivo)
and PI of 0.44 are then presented, followed by the results
of the velocity measurements by LDA under the same
#ow conditions.
3.1.1. Steady yow
Fig. 4 shows the steady #ow patterns in the x}y and
x}z planes at Reynolds numbers of 100, 300 and 600 in
the vein, with the distal artery (DA) occluded. At Re of
100 (Fig. 4a), the #ow is laminar and the #ow pattern is
highly stable. As the #uid from the proximal artery (left)
enters the anastomosis, it separates at the heel where
a vortex with anticlockwise rotation can be seen. It then
impinges on the outer wall of the vein, between the toe
and heel, forming a large tear-drop shaped vortex (clock-
wise rotation) which occupies a large part of the anasto-
mosis. In the occluded distal artery, a small vortex with
anticlockwise rotation can also be seen. Stagnation
points can be identi"ed on the #oor of the artery, on the
outer wall of the vein and also on the inner wall close to
the heel. When viewed in the x}z plane, the vortices can
be seen to be symmetrical about the mid plane of the
artery. As the Reynolds number increases, the helical
motion in the vein becomes more apparent and at Re
'190, a second vortex begins to form further into the
occluded distal artery. At Re of 300 (Fig. 4b) the overall
#ow pattern remains essentially unchanged except for the
#uid motion which becomes increasingly chaotic, as in-
dicated by the blurred pathlines. Above Re of 600
918 S. Sivanesan et al. / Journal of Biomechanics 32 (1999) 915}925
Fig. 5. Flow patterns captured under pulsatile #ow conditions with di!erent #ow distributions. The mean Reynolds number, Re, was 300 and the ratio
of proximal-to-distal artery #ow in each case was (a) 1 : 0, (b) 3 : 1, (c) 1 : 1, (d) 1 : 3.
(Fig. 4c) the #ow is highly disturbed and it becomes
di$cult to discern any organised structures.
3.1.2. The ewects of varying the yow distribution
Fig. 5 shows the #ow patterns with increasing amount
of retrograde #ow in the distal artery. The Reynolds
number in the vein was 300 and the ratio of proximal to
distal artery #ow was 1 : 0, 3 : 1, 1 : 1 and 1 : 3. When the
#ow ratio is 3 : 1 (Fig. 5b), the main di!erence in the #ow
pattern compared to the occluded distal artery case is
that no vortices are present in the distal artery. With
a #ow ratio of 1 : 1 (Fig. 5c), the #ow pattern is markedly
changed. The #uid entering through the proximal and
distal artery segments merge in the anastomosis, towards
the heel. One or two small transient vortices can be
observed near the artery #oor and #ow separation re-
gions present at both the toe and the heel extend further
into the vein. As the #ow ratio decreases to 1 : 3 (Fig. 5d),
the point where the inlet #ows merge is shifted more
towards the heel and the #ow separation at the toe
becomes more prominent than that at the heel. The #ow
pattern for a #ow ratio of 0 : 1 (Fig. 5a) is identical to that
of a proximal side-to-end bypass graft anastomosis
(Hughes and How, 1995).
In the case of antegrade #ow in the distal artery, the
#ow in the anastomosis follows the same general pattern,
with part of the proximal arterial #ow being drawn into
the large vortex on the #oor of the anastomosis and being
shed into the distal artery in a helical swirl. As a result,
the vortex is more elongated and spans the whole length
of the anastomosis. A horseshoe vortex caused by #ow
separation at the heel can also be clearly seen at the lower
Reynolds numbers.
3.1.3. Pulsatile yow
The #ow patterns recorded at four speci"c points in
the cycle (proximal to distal artery #ow ratio of 3 : 1,
period of pulsatile cycle of 1 s) are shown in Fig. 6. The
pulsatile #ow patterns are very similar to those recorded
in steady #ow and do not appear to change signi"cantly
during the cycle. Di!erences are only evident in the detail
of the main vortex which appears to #uctuate with the
#ow. No appreciable di!erences could be observed when
the period of the pulsatile #ow was increased to 1.6 s.
3.2. LDA velocity measurements
3.2.1. Wall shear stress on the artery yoor
Fig. 7 shows a plot of the mean and oscillatory (i.e.
variation over the cycle) of wall shear stress along the
#oor of the artery. Positive shear stress corresponds to
#ow directed towards the distal artery. The oscillatory or
S. Sivanesan et al. / Journal of Biomechanics 32 (1999) 915}925 919
Fig. 6. Pulsatile #ow patterns at four points in the cycle (Re"600, and the ratio of proximal-to-distal artery #ow was 3 : 1).
Fig. 7. Mean and oscillatory (vertical bars) wall shear stresses along the
arterial #oor of the anastomosis (Re"600; the ratio of proximal-to-
distal artery #ow was 3 : 1).
Table 1
Mean and oscillatory wall shear stress determined from near-wall
velocity measurements at di!erent locations along the vein. The distan-
ces quoted are relative to the suture line. The mean Reynolds number
was 600, and the ratio of proximal-to-distal artery #ow was 3 : 1
Location Distance Wall shear stress along vein (N/m)
(mm)
Inner wall Outer wall
Mean Oscillatory Mean Oscillatory
A 4 * * 40 52
B 10 1 36 14 57
C 24 12 17 7 8
D 50 5 3 9 8
E 70 4 2 14 9
cyclic variation in wall shear stress, at each measurement
point is denoted by the vertical bars. The magnitudes of
the mean and oscillatory wall shear stress are highest
within the anastomosis, reaching peak values of 27.3 and
89 N/m, respectively.
3.2.2. Wall shear stress in the vein
Values of mean and oscillatory wall shear stress in the
vein at various distances from the anastomosis are shown
in Table 1. The mean wall shear stress reaches a peak at
the inner wall 24 mm from the suture line whilst on the
outer wall, it is highest at 4 mm, with a value of 40 N/m.
A region of relatively low wall shear stress (1 N/m) was
recorded at 10 mm on the inner wall. The oscillatory
shear stress generally decreases with distance into the
vein. The mean and oscillatory wall shear stress tend to
be greater on the outer wall of the vein.
3.2.3. Turbulence intensity in the vein
Fig. 8 shows a plot of axial turbulence intensity, rela-
tive to the time average velocity across the vein, as
a function of distance along the centreline of the vein.
920 S. Sivanesan et al. / Journal of Biomechanics 32 (1999) 915}925
Fig. 8. Variation of axial turbulence intensity relative to the time
average velocity across the vein with distance along the centreline of the
vein. Measurements were made with di!erent artery #ow ratios
(Re"600; the ratio of proximal-to-distal artery #ow was 1 : 0, 3 : 1 and
1 : 1).
Near the anastomosis, the relative turbulence intensity
was greatest and appeared to be dependent on the prox-
imal-to-distal artery #ow ratio being highest for a #ow
ratio of 1 : 0 and lowest for 1 : 1. From about 25 mm
downstream of the suture line, the relative turbulence
intensities were not dependent on the artery #ow ratio.
At 70 mm the relative turbulence intensity was still signif-
icant, with a value of approximately 0.14. Fig. 9 shows
how the absolute turbulence and the convective velocity
vary across the lumen at di!erent positions along the
vein. Four millimetres away from the anastomosis, axial
and radial turbulence intensities peak towards the outer
wall. At 10 mm, however, the axial turbulence intensity
peak occurs more centrally while the radial turbulence
intensity pro"le #attens out. At 24 mm, the radial turbu-
lence intensity pro"le exhibits a broad peak, shifted
slightly towards the outer wall. The maximum axial
turbulence intensity, however, is near the inner wall,
although the pro"le is largely #at, away from the walls.
As the velocity pro"le becomes more symmetrical at
50 mm, the radial turbulence intensity pro"le is also
more symmetrical. The axial turbulence intensity pro"le
possesses a maximum near both walls and is lowest at the
centre of the vessel. By 70 mm from the anastomosis, the
peak near the inner wall moves towards the centre of the
vessel whilst in all other ways the pro"les are similar to
the previous set.
3.2.4. Reynolds shear stress in the vein
The variation in Reynolds shear stress in the centreline
of the vein seems to follow a similar pattern to that of the
axial turbulence pro"les (Fig. 10). Both peak and mean
Reynolds stresses are greatest at the suture line, reaching
values of 464 and 244 N/m, respectively. Both Reynolds
stress values decrease in magnitude with distance from
the suture line.
4. Discussion
There have been very few studies on the
haemodynamics of AV "stulae for haemodialysis access
and most have been concerned with the determination of
bulk #ow through the "stula, and the e!ect of the anasto-
mosis con"guration on #ow (Anderson et al., 1977;
Wedgewood et al., 1984). Shu et al. (1987a,b) have investi-
gated the #ow patterns in models of a prosthetic graft AV
shunt (loop graft) but there does not appear to have been
similar studies in the Cimino-Brescia AV "stula.
In this in vitro study, the geometry of the "stula was
simpli"ed (e.g. the vessel segments were uniformly circu-
lar, the artery was straight, the anastomosis was planar,
i.e. the vein and artery lay in the same plane, and the
compliance, if any, of the AV "stula was not reproduced).
Because of the relatively thick wall, there was very little
wall movement during the pulsatile #ow cycle. However,
this model incorporated a similar vein to artery diameter
ratio as seen in vivo (Sivanesan et al., 1998a) and the #ow
waveforms recorded in AV "stulae in haemodialysis pa-
tients were also reproduced in the pulsatile #ow experi-
ments.
A notable "nding of this study is that the #ow patterns
in pulsatile #ow were very similar to those in steady #ow.
This was most likely a result of the low pulsatility index
of the #ow. As expected there were di!erences in the #ow
structures when the proximal and distal artery #ow con-
tributions were altered. A wide range of #ow distribu-
tions has been observed in vivo, with an average
proximal to distal artery #ow ratio of approximately 3 : 1
(Sivanesan et al., 1998b). In this article, we present quantit-
ative pulsatile #ow measurements for a #ow ratio of 3 : 1.
The detection of a `thrilla resulting from blood #ow
turbulence is commonly used by clinicians to con"rm
a patent and functional AV "stula. Not surprisingly,
therefore, turbulence was the dominant feature in the
model for Reynolds numbers within the normal range
seen in patients. Turbulence, which is associated with
both high wall shear stress and high #uid shear stress,
occurs at much lower Reynolds numbers than the critical
Reynolds number for turbulence in a straight tube largely
because of the acute deviation of #ow into the vein.
The turbulence intensity and Reynolds shear stress
reached their maximum values during the early decel-
erative phase of the cardiac cycle. Except near the suture
line where strong swirling structures were present, the
turbulence intensity was generally higher in the axial
than radial directions. The axial and radial turbulence
intensity in the centreline of the vein decayed rapidly with
distance reaching a relatively constant value of 0.l0}0.15
between 40 and 70 mm from the suture line.
S. Sivanesan et al. / Journal of Biomechanics 32 (1999) 915}925 921
Fig. 9. Absolute axial and radial turbulence intensity and velocity pro"les at (a) 4 mm, (b) 10 mm, (c) 24 mm, (d) 50 mm and (e) 70 mm from the
anastomosis (Re"600; the ratio of proximal- to-distal artery #ow was 3 : 1).
Sallam and Hwang (1984) found that the critical
Reynolds shear stress level for incipient haemolysis was
400 N/m, while Sutera (1977) found that prolonged ex-
posure to Reynolds shear stress of 150 N/m could in-
duce haemolysis, and that Reynolds shear stresses of
above 10 N/m could cause platelet lysis. However, the
relationship between the magnitude of Reynolds shear
stresses and the minimum corresponding duration for
cellular damage is still unknown. Many reports cite tur-
bulence as being potentially detrimental to blood compo-
nents and the vascular wall, but very few of these studies
have measured turbulence intensity per se. Some studies
have intimated a direct relationship between the amount
of thrombus formation and axial turbulence intensity
(Stein and Sabbah, 1974; Smith et al., 1972). The peak
Reynolds shear stress values measured in the "stula
model were found to lie within the range that may be
detrimental to red blood cells and platelets, however
exposure to such levels of shear stress occurs for a
fraction of each cycle. The e!ects of prolonged
exposure to such #uctuating Reynolds shear stresses is
not known.
Fry (1968) showed that wall shear stress in the range
35}40 N/m can cause injury to endothelial cells within
922 S. Sivanesan et al. / Journal of Biomechanics 32 (1999) 915}925
Fig. 10. Mean and peak Reynolds shear stress along the centreline of
the model vein (Re"600; the ratio of proximal-to-distal artery #ow
was 3 : 1).
an hour of exposure. Lower values can increase the
permeability of the endothelial layer, damage eryth-
rocytes (Blackshear, 1972) and cause enhanced platelet
adhesiveness and platelet aggregation. Conversely, low
wall shear stress, below 1 N/m, has been shown to be
associated with the development of intimal hyperplasia
in arterial bypass graft anastomoses (Salam, 1996). Sites
of oscillating wall shear stress have been found to show
evidence of intimal cell proliferation and increased endo-
thelial permeability (Ku and Giddens, 1983; Ku et al.,
1985). In the AV "stula model, high mean and oscillating
wall shear stress were seen at the anastomosis on the
#oor of the artery. The peak value of mean wall shear
stress was 27.3 N/m while the peak-to-peak amplitude
was 89 N/m. The oscillatory wall shear stresses in the
region of the anastomosis were generally of an order of
magnitude larger than the #uctuating stresses seen more
proximally.
These results suggest that the levels of mean and oscil-
latory wall shear and Reynolds stresses measured in the
"stula exceed the threshold values for cell damage. How-
ever, the above results cannot be applied directly to the
situation that exists in vivo because the density and
viscosity of the #uid used in the LDA measurements were
higher than the values for normal human blood. Since
wall shear stress is directly proportional to the density
and the square of the viscosity (see the appendix), the wall
shear stress measured in the model must be scaled to
provide an estimate of the values that would be found in
vivo. Patients with chronic kidney failure on long-term
haemodialysis usually have low haemoglobin and hae-
matocrit (hence low blood viscosity) because of reduced
secretion of erythropoietin, and mechanical trauma to
red cells in haemodialysers exacerbates this problem.
Currently, most patients receive recombinant human
erythropoietin (rHuEPO) to treat their anaemia and this
has the e!ect of maintaining their haematocrit and blood
viscosity at near normal values. Fellner et al., (1993)
found that the mean whole blood viscosity increased
from 2.8 to 3.7;10\ Pa s after 8 months of rHuEPO
treatment. It is shown in the appendix that the following
scaling law applies:

?
"
K

,
where
?
and
K
are the wall shear stresses in vivo and
measured in the model, respectively, and
?
and
?
,

K
and
K
are the density and kinematic viscosity of
blood and the model #uid, respectively. If we assume that
a patient's blood has a viscosity of 3.7;10\ Pa s and
density of 1050 kg/m, the expression in brackets, which
represents the scaling factor for shear stress, is equivalent
to 0.337. Applying the above correction to the experi-
mental data obtained along the arterial #oor (Fig. 7), for
example, gives maximum time-averaged wall shear stress
of 9.2 N/m in vivo. Likewise, the Reynolds stress values
quoted above decrease to levels below that which are
known to cause haemolysis. The wall shear stress values
for the vein given in Table 1, which should also be scaled
accordingly, reveal a region of low (0.34 N/m) and high
(13.5 N/m) mean wall shear stress at 10 mm on the inner
wall and 4 mm on the outer wall.
A previous clinical study (Sivanesan et al., 1999) identi-
"ed three speci"c sites of stenosis in patients with pri-
mary access radiocephalic "stulae: at the anastomosis on
the #oor of the artery, on the inner wall of the curved
region of the vein and just proximal to this curved seg-
ment where the vein straightens out. The present study
revealed high wall shear stress on the #oor of the artery
and on the outer wall of the vein, near the anastomosis.
High turbulence intensity was also observed in the "stula
especially near the anastomosis which may be associated
with anastomotic stenosis. Although turbulence intensity
remained relatively high throughout the vein, the regions
of abnormally low or high wall shear stress were found
within 10 mm of the suture line. Thus wall shear stress
may not be a contributory factor in the development of
localised stenoses within the cephalic vein. This study
indicates that the #ow patterns change rapidly in the
segment of the vein near the anastomosis and thus much
more detailed measurements are needed before any con-
clusive observations can be drawn. Further studies of the
e!ects of diameter ratio and anastomosis angle are also
required in order to determine whether it is possible, by
modifying the geometry of the "stula, to reduce the
magnitude of wall shear stress on the artery #oor and
turbulence intensity in the vein and thereby improve
long-term "stula patency.
S. Sivanesan et al. / Journal of Biomechanics 32 (1999) 915}925 923
Acknowledgements
The authors would like to thank Mr. Malcolm W.
Brown, Consultant Renal Transplant Surgeon, for his
co-operation and assistance in the collecting of some of
the clinical data, and Mr. James Blackhurst for his tech-
nical help. We gratefully acknowledge the "nancial sup-
port of the National Kidney Research Fund.
Appendix A. Dimensional analysis
In order to relate the measurements made in the "stula
model to the situation in vivo, the following dimensional
analysis was used. Consider the following group of vari-
ables:
(, , u , d, , u( , ).
The magnitude of the shear stress () exerted by the #uid
at the vessel wall is determined by the density () and
dynamic viscosity () of the #uid, its mean velocity (u ),
amplitude (u( ) and frequency () as it #ows through an
anastomosis of diameter d.
Applying Buckingham's theorem (see Massey, 1979),
it can be shown that

u
"

u d

,
u(
u
,
d
u
. (A.1)
Wall shear stress is therefore a function of the Reynolds
number (ud/), the ratio of the amplitude of pulsatile
velocity to mean velocity (u( /u ) and the Strouhal number
(d/u ).
Another dimensionless quantity, Womersley's fre-
quency parameter, , is given by
"

d
u
,
u d

.
It follows that

u
"

u d

,
u(
u
, d

.
The scaling law is obtained as follows. Because
the Reynolds number, the ratio (u( /u ), i.e. pulsatility
index, and frequency were all kept the same, and the
model was geometrically similar and true-to-scale, we
can write

?
/(
?
u
?
)

K
/(
K
uN
K
)
"1,
where the subscripts m and a denote the parameters in
the model and those in vivo, respectively. Thus,

?
"
K

?
u
?

K
u
K

. (A.2)
We now wish to "nd an expression for uN
?
/uN
K
. From
dynamic similarity,

?
u
?
d
?

?
"

K
u
K
d
K

K
,
u
?
u
K
"

K
d
K
d
?
,
where "/ is the kinematic viscosity. Since the model
is geometrically similar and true-to-scale, d
K
/d
?
"1, and
so
u
?
"u
K

. (A.3)
Substituting Eq. (A.3) into Eq. (A.2) gives

?
"
K

. (A.4)
The expressions in brackets in (A.3) and (A.4) represent
the scaling factors for velocity and shear stress respective-
ly. In this study, the scale factor for velocity was 0.618
and for shear stress it was 0.882;(0.618)"0.337.
Since Reynolds stress is the product of two velocity
components and density, the scale factor for this quantity
is

?
u
?
v
?

K
u
K
v
K

which is also equivalent to 0.337.

References
Anderson, C.B., Etheredge, E.E., Harter, H.R., Gra!, R.J., Codd, J.E.,
Newton, W.T., 1977. Local blood #ow characteristics of arterio-
venous "stulas in the forearm for dialysis. Surgery, Gynecology and
Obstetrics 144, 531}533.
Blackshear, P.L., 1972. Mechanical hemolysis in #owing blood.
In: Fung, Y.C., Perrone, N., Anliker, M. (Eds.), Biomechanics:
Its Foundation and Objectives. Prentice-Hall, Englewood Cli!s,
pp. 501}528.
Casty, M., Giddens, D.P., 1984. 25#1 channel pulsed ultrasound
Doppler velocity meter for quantitative #ow measurement and tur-
bulence analysis. Ultrasound Medical Biology 10, 161}172.
Fellner, S.K., Lang, R.M., Newmann, A., Korcarz, C., Borow, K.M.,
1993. Cardiovascular consequences of correction of the anaemia of
renal failure with erythropoietin. Kidney International 44,
1309}1315.
Fry, D.L., 1968. Acute vascular endothelial changes associated with
increased blood velocity gradients. Circulation Research 22, 165}197.
Hughes, P.E., Shortland, A.P., How, T.V., 1996. Visualization of vortex
shedding at the proximal side-to-end artery-graft anastomosis. Bior-
heology 33, 305}317.
924 S. Sivanesan et al. / Journal of Biomechanics 32 (1999) 915}925
Kinnaert, P., Geens, M., Vereerstaeten, P., Heuse, A., Buchin, R.,
Leclerc, J.L., Toussaint, C., van Geertruyden, J., 1971. Experience
with arteriovenous "stulas for chronic maintenance haemodialysis.
British Journal of Surgery 58, 104}108.
Ku, D.N., Giddens, D.P., 1983. Pulsatile #ow in a model carotid
bifurcation. Arteriosclerosis 3, 31}39.
Ku, D.N., Giddens, D.P., Zarins, C.K., Glagov, S., 1985. Pulsatile #ow
and atherosclerosis in the human carotid bifurcation. Arteriosclerosis
5, 293}302.
Massey, B.S., 1979. In: Mechanics of Fluids, 4th Edition. Van Nostrand
Reinhold Co, Wokingham, England.
Ross, R., 1986. The pathogenesis of atherosclerosis * an update. N.
Eng. J. Med. 314, 488}500.
Sallam, A.M., Hwang, H.C., 1984. Human RBC hemolysis in a turbu-
lent shear #ow: contribution of Reynolds shear stresses. Biorheology
21, 783}979.
Salam, T.A., Lumsden, A.B., Suggs, W.D., Ku, D.N., 1996. Low shear
stress promotes intimal hyperplasia thickening. Journal of Vascular
Investment 2, 12}22.
Shu, M.C.S., Noon, G.P., Hwang, N.H.C., 1987a. Phasic #ow patterns
at a hemodialysis venous anastomosis. Biorheology 24, 711}722.
Shu, M.C.S., Noon, G.P., Hwang, N.H.C., 1987b. Flow pro"les and wall
shear stress distribution at a hemodialysis venous anastomosis: pre-
liminary study. Biorheology 24, 723}735.
Sivanesan, S., How, T.V., Bakran, A., 1999. Characterising #ow distri-
butions in AV "stulae for haemodialysis access. Nephrol. Dial.
Trans., 13, 3108}3110.
Sivanesan, S., How, T.V., Bakran, A., 1999. Sites of stenosis in AV
"stulae for haemodialysis access. Nephrol. Dial. Trans., 14, 118}120.
Smith, R.L., Blick, E.F., Coalson, J., Stein, P.D., 1972. Thrombus
production by turbulence. J. Applied. Physiology 32, 261}264.
Stein, P.D., Sabbah, H.N., 1974. Measured turbulence and its e!ect on
thrombus formation. Circulation Research 35, 608}614.
Sutera, S.P., 1977. Flow-induced trauma to blood cells. Circulation
Research 41, 2}8.
Wedgwood, K.R., Wiggins, P.A., Guillou, P.J., 1984. A prospective
study of end-to-side vs. side-to-side arteriovenous "stulas for
haemodialysis. British Journal of Surgery 71, 640}642.
S. Sivanesan et al. / Journal of Biomechanics 32 (1999) 915}925 925