Full text of DOI:10.1053/ajkd.2001.28577

ORIGINAL INVESTIGATIONS
Analysis of Variable Flow Doppler Hemodialysis Access Flow
Measurements and Comparison With Ultrasound Dilution
William F. Weitzel, MD, Jonathan M. Rubin, MD, PhD, Sean F. Leavey, MD,
Richard D. Swartz, MD, Rajnish K. Dhingra, MD, and Joseph M. Messana, MD
● The variable flow (VF) Doppler method determines access blood flow from the pump speed–induced change in
Doppler signal between the arterial and venous needles. This study evaluated 35 patients in two analyses to assess
VF Doppler measurement reproducibility (54 paired measurements) and compared VF Doppler and ultrasound
dilution flow measurements (24 paired measurements). VF Doppler measurement variations were 4% for access
flow less than 800 mL/min (n ؍ 17), 6% for access flow of 801 to 1,600 mL/min (n ؍ 22), and 11% for access flow
greater than 1,600 mL/min (n ؍ 15). The mean measurement coefficient of variation was 7% for VF Doppler
compared with 5% for ultrasound dilution. Correlation coefficients (r) between VF Doppler and ultrasound dilution
access flow measurements were 0.79 (n ؍ 24; P < 0.0001), 0.84 for access flow less than 2,000 mL/min (n ؍ 20; P <
0.0001), and 0.91 for access flow less than 1,600 mL/min (n ؍ 18, P < 0.0001). VF Doppler measurements using
indicated versus measured pump flow rates correlated highly (r ؍ 0.99; P < 0.0001). VF Doppler therefore yields
reproducible access volume flow measurements that correlate with ultrasound dilution measurements. The VF
Doppler method is dependent on the pump-induced change in access Doppler signal and therefore is inherently
most accurate and reproducible at lower access blood flow rates. This method appears capable of determining
access flow rates in the clinically useful range.
© 2001 by the National Kidney Foundation, Inc.
INDEX WORDS: Hemodialysis (HD); access; graft; fistula; Doppler; volume flow; blood flow; access monitoring.
IALYSIS ACCESS remains a major cause
tools for hemodialysis access evaluation. How-
D of morbidity among dialysis patients and a ever, both these methods have shortcomings that
frequent cause for hospitalization.1,2 The Na-
tional Kidney Foundation-Dialysis Outcomes
Quality Initiatives guidelines and many research-
ers and clinical professionals have advocated
access surveillance as a means to improve pa-
tient care and reduce access-related costs.3-5 Ap-
propriate access surveillance allows for timely
corrective intervention, and monitoring has been
shown to increase graft life and prevent thrombo-
sis.4 By preventing access thrombosis, improved
access surveillance may reduce morbidity and
significantly enhance the quality of life of hemo-
dialysis patients, as well as help control the $1
billion annual costs associated with access care
in the United States.1-4,6
limit their utility for access surveillance. Duplex
imaging is costly, labor intensive, and subject to
operator-dependent error.21 Indicator dilution
techniques are time consuming and require dialy-
sis blood line reversal.14-19
A novel method that measures blood flow
using a changing Doppler signal characteristic
as a function of blood flow rate in the access
during dialysis is being evaluated.21-23 This
variable flow (VF) Doppler measurement pro-
cedure has several potential advantages. First,
a small change in cross-sectional radius mea-
surement leads to a large change in volume
flow calculations by duplex because area var-
Several approaches have been developed to
aid in access surveillance. These methods in-
clude duplex ultrasound techniques,6-9 static and
dynamic access pressure monitoring,9-12 recircu-
lation,13 and various indicator dilution tech-
niques.14-19 It is known that decreasing access
blood flow rate predicts progression of access
stenoses, and definitive and timely intervention
prevents thrombosis.3,6-8,18,20 Both duplex ultra-
sound and indicator dilution techniques deter-
mine access blood flow rate and are important
From the Departments of Internal Medicine and Radiol-
ogy, University of Michigan School of Medicine, Ann Arbor,
MI.
Received February 12, 2001; accepted in revised form
June 29, 2001.
Address reprint requests to William F. Weitzel, MD,
Division of Nephrology, Department of Internal Medi-
cine, University of Michigan Medical Center, 3914 Taub-
man Center, Ann Arbor, Michigan 48109-0364. E-mail:
weitzel@umich.edu
© 2001 by the National Kidney Foundation, Inc.
0272-6386/01/3805-0002$35.00/0
doi:10.1053/ajkd.2001.28577
American Journal of Kidney Diseases, Vol 38, No 5 (November), 2001: pp 935-940
935

936
WEITZEL ET AL
circuit. The position of the Doppler probe for VF Doppler is
shown in Fig 1, and the measurement method was outlined
in detail in a previous study.21
ies with the square of the radius. Because the
VF Doppler method does not require area
measurement, this potential source of error is
eliminated. Second, the Doppler signal beam
angle is a potential source of error in duplex
flow calculations, requiring accurate knowl-
edge of the angle to calculate the velocity
required for duplex volume flow determina-
tion. Conversely, the VF Doppler method uses
any Doppler signal proportional to volume
flow and thereby corrects itself for Doppler
beam-angle variance, making the measure-
ment insensitive to this source of error pro-
vided the Doppler beam angle remains fixed
during the VF Doppler measurement. Third,
the Doppler signal used to calculate volume
flow by this method changes more dramati-
cally with variable pump speeds when an ac-
cess has low volume flow, making this method
inherently most accurate in patients with lower
access flow rates who are at greater risk for
access thrombosis. Fourth, unlike indicator
dilution techniques, there is no indicator or
blood line reversal.
Briefly, the access has two needles introduced into its
lumen during dialysis; one needle for the removal of blood
(arterial) to pass through the dialysis circuit and one needle
for the return of blood (venous) to the circulation. The flow
through the graft or fistula downstream (QR) from the
arterial needle decreases as a function of the blood flow rate
(QA) in the dialysis circuit (QB). To the extent that the net
flow through the system does not change during dialysis, this
flow rate through the portion of the access between the
dialysis needles during dialysis (QR) follows the relation-
ship QR ϭ QA – QB. Doppler signal S is measured down-
stream from the arterial needle (QR) at various values for
QB (ie, various dialysis blood pump speeds). A modeling
function then is constructed for this signal F(QB) using
measured values such that S ϭ F(QB). This modeling
function may take the form of any algebraic or numerical
one-to-one function. In this application, because Doppler
velocity (and Doppler signal uncorrected for transducer
angle) is linearly related to volume flow, a linear regression
modeling function was used. The X intercept for this linear
regression function represents the point at which the Dopp-
ler signal S is zero and yields the value at which QR equals
zero, and therefore at which QB ϭ QA, determining access
flow.21
Study Procedure
The prototype device was fashioned using a personal
computer–based spectral Doppler board and software (SPECS
USA Inc, Sarasota, FL) that allows calculation and export of
various spectral waveform parameters, including instanta-
neous mean and time-averaged mean Doppler signals, used
for the VF Doppler method. This device was connected to a
laptop computer for signal analysis. The SPECS Doppler
board performs spectral analysis, sampling Doppler data at
100 times/s, and displays the spectral waveform and derived
METHODS
The change in access Doppler signal with variable dialy-
sis pump rates can be used to determine the access blood
flow rate. This method exploits the decreasing access blood
flow within the access between the needles with standard
two-needle placement in the access graft, with the arterial
needle upstream from the venous (blood return) needle
during dialysis as blood is pumped through the dialysis
Fig 1. Access volume
flow (QA) can be determined
from the relationship be-
tween the Doppler signal re-
maining downstream (QR) in
the access as a function of
the dialysis blood flow rate
(QB).

VARIABLE FLOW DOPPLER ACCESS FLOW MEASUREMENTS
937
parameters needed to perform the measurement on the
laptop computer. Spectral data and derived parameters were
imported into a database programmed to perform real-time
calculation of the X intercept of the linear regression line
from the VF Doppler data, thus determining access blood
flow. The correlation coefficient of this regression line also
was measured to give an index of measurement reliability.
Measurements with r2 greater than 0.9 are considered reli-
able. For this protocol, each VF Doppler measurement was
performed during routine dialysis using data from four
spectral Doppler waveforms: two waveforms collected at
100 mL/min and two waveforms collected at 400 mL/min.
The prototype allows for visual examination of the spectral
waveform and audio output; therefore, the Doppler trans-
ducer can be centered easily over the graft for the signal to
be maximized for data collection. An 8-MHz disc-shaped
(2.0-cm diameter ϫ 0.4-cm thick) Doppler transducer was
used to facilitate Doppler sampling between the needles.
using indicated versus measured pump speeds at indicated
pump rates of 400 mL/min. Measured pump speeds were
determined using the Transonics ultrasound dilution blood
flow results, available for 22 patients.
RESULTS
The variation in VF Doppler measurements
was 7% compared with 5% for ultrasound dilu-
tion (Table 1). Mean percentages of coefficient
of variation of all paired VF Doppler measure-
ments (n ϭ 54) were compared in three measure-
ment ranges of access flow: less than 800, 800 to
1,600, and greater than 1,600 mL/min. These
values are listed in Table 1.
The reproducibility of all VF Doppler measure-
ments less than 2,000 mL/min is shown in Fig 2.
The correlation coefficient (r) between repeated
measurements with different probe positions was
0.93 (P Ͻ 0.001; n ϭ 13).
The comparison of VF Doppler and Transon-
ics ultrasound dilution measurements is shown in
Fig 3. The correlation coefficient (r) between VF
Doppler and Transonics methods was 0.79 for all
comparison measurements (n ϭ 24; P Ͻ 0.0001).
Correlation coefficients (r) between measure-
ments were 0.91 for VF Doppler less than 1,600
mL/min (n ϭ 18; P Ͻ 0.0001) and 0.84 for VF
Doppler less than 2,000 mL/min (n ϭ 20; P Ͻ
0.0001).
The measured pump rate in our patients was
1.7 mL/min less than the indicated pump rate
Ϯ29 mL/min (7.3%) SD at an indicated pump
rate of 400 mL/min. Although the difference
between true and indicated pump rates may be
most problematic in catheters with poor inflows
rather than grafts or fistulae, it could conceivably
influence the accuracy of the VF Doppler mea-
surement. To investigate the impact of this effect
Patients
Thirty-five patients with prosthetic bridge grafts were
studied at the University of Michigan Medical Center’s
Outpatient Hemodialysis Facility (Ann Arbor, MI). Patients
underwent dialysis using Fresenius 2008H (Fresenius Medi-
cal Care, Lexington, MA) and Cobe Century System 3
(Gambro Renal Care Products, Lakewood, CO) with 15 G
dialysis needles. We tested the reproducibility of experimen-
tal measurements in 17 patients using the correlation be-
tween the first and second measurements and coefficient of
variation between paired measurements. Paired measure-
ments were subsequently collected in 13 of these patients to
assess the impact of changing the probe position between
measurements on recorded access flow measurement. We
then measured access blood flow rates in 24 subjects (includ-
ing 6 patients from the reproducibility study) with prosthetic
grafts using VF Doppler and ultrasound indicator dilution
(Transonics, Ithaca, NY) during the same dialysis treatment
to compare these measurement techniques.
Data Analysis
The correlation between paired measurements was sought
using linear regression analysis. Intermeasurement variation
was expressed as a percentage of coefficient of variation.
Percentage of coefficient of variation is defined as the SD in
access flow rates divided by mean access flow rate multi-
plied by 100. The lower the intermeasurement of percentage
of coefficient of variation, the greater the reproducibility of
the measurement. The mean coefficient of variation of
paired VF Doppler measurements was compared in three
measurement ranges of access flow: less than 800, 800 to
1,600, and greater than 1,600 mL/min. The coefficient of
variation for ultrasound dilution also was measured as a
reference value for the coefficient of variation in access flow
measurement. The correlation between VF Doppler and
ultrasound dilution was determined. Finally, the indicated
pump rate is often less than delivered pump rate at greater
pump values depending on the degree of negative pressure
generated at greater pump rates.24 We therefore retrospec-
tively investigated the impact of this effect on VF Doppler
measurement accuracy. We determined VF Doppler results
Table 1. Measurement Coefficient of Variation for
the Comparison Study Between VF Doppler and
Ultrasound Dilution and All Paired VF Doppler
Measurements
Coefficient of
No. of
Variation (%) Measurements
Transonics (comparison study)
VF Doppler (comparison study)
VF Doppler (all)
VF Doppler Ͻ800 mL/min
VF Doppler 800–1,600 mL/min
VF Doppler Ͼ1,600 mL/min
5
7
7
4
6
24
24
54
17
22
15
11

938
WEITZEL ET AL
Fig 4. Comparison of VF Doppler with indicated
and measured pump rates. Line of identity and correla-
tion line are shown. Correlation coefficient (r) is 0.99
(P < 0.0001).
Fig 2. Reproducibility of all VF Doppler measure-
ments less than 2,000 mL/min (n ؍ 44). Line of identity
and correlation line are shown (r ؍ 0.92; P < 0.0001).
the patient’s graft for Doppler examination dur-
ing dialysis because dialysis needles were taped
during dialysis. This problem has largely been
solved with suitable transducer selection and
orientation. Because imaging is not required,
there are several suitable small Doppler transduc-
ers that can be positioned between the needles
for this measurement. One early solution was the
use of a probe designed for urological applica-
tions (Urometrics Inc, St Paul, MN). This trans-
ducer was fit with a detachable strap fashioned
by the investigators to secure it in place for
measurements during dialysis. The transducer
currently used is an 8-MHz, 60°, 0.4-cm thick by
2-cm diameter flat-disc transducer (SPECS USA
Inc). This transducer fits in low profile, often
under the venous dialysis tubing, between the
needles oriented downstream along the access.
There is no interference in the Doppler signal
from the dialysis tubing even when the trans-
ducer and tubing are in direct contact. This
transducer selection and orientation have made
inaccessibility of the graft for VF Doppler exami-
nation an uncommon problem. However, further
improvements in transducer ergonomics and
beam profile are being investigated.
on the VF Doppler measurement accuracy, we
retrospectively redetermined VF Doppler results
using both indicated and measured pump speeds
when the dialysis machine indicated 400 mL/
min. The Transonics measured pump speed at
400 mL/min indicated pump speed had been
stored for 22 of the 24 subjects. Because these
results were determined retrospectively, we did
not have pump measurements for 2 of the 24
patients studied. Figure 4 shows the comparison
between VF Doppler results using indicated and
measured pump rates.
DISCUSSION
Early in the evaluation of this method, investi-
gators encountered the relative inaccessibility of
Flow turbulence also has been raised as a
potential problem. The palpable thrill and au-
dible bruit are often attributed to access turbu-
lence. Although turbulence may contribute to
these physical findings, the thrill results from
energy from the access blood flow being transmit-
ted into the access wall, inducing palpable vibra-
tions. These vibrations are seen in varying de-
Fig 3. VF Doppler and Transonics ultrasound dilu-
tion measurements with regression line and line of
identity. Correlation coefficients (r) are 0.79 for all
comparison measurements (n ؍ 24), 0.84 for VF Dopp-
ler less than 2,000 mL/min (n ؍ 20), and 0.91 for VF
Doppler less than 1,600 mL/min (n ؍ 18).

VARIABLE FLOW DOPPLER ACCESS FLOW MEASUREMENTS
939
grees of intensity in the spectral Doppler signal
and are often referred to as “wall clutter.” Al-
though the Doppler signal intensity from the
thrill may be high and the amplitude of the signal
may be modulated with the cardiac cycle, the
frequency observed is usually separate and lower
than the Doppler signal from the access flow. In
some cases, these low-frequency Doppler signals
(from wall clutter) lower the instantaneous mean
Doppler signal. However, data suggest that their
contribution to the relative change in Doppler
signal with pump speed used to calculate flow is
not a significant limiting factor for this method.
In addition, the spectral Doppler equipment be-
ing used (SPECS USA Inc) allows filtering of
this low-frequency wall vibration, if necessary.
Turbulence is a separate finding that requires
additional investigation. The authors observed
that laminar waveforms are predominately seen
in prosthetic grafts (with or without access wall
vibration). However, in the presence of turbu-
lence, when there is net forward flux of blood
through the access, Doppler velocity can be
measured. Turbulence will increase the variance
of the Doppler signal; ie, broaden the spectrum,
but the mean should remain stable. When turbu-
lence disappears, the mean should be relatively
unchanged as long as the turbulence is superim-
posed on the average flux down the access.
Spectral broadening and turbulent waveforms
are seen more often in fistulae than grafts. Al-
though this turbulence makes measurement more
difficult in fistulae, the Doppler signal is influ-
enced by pump speed in fistulae, evidenced by
marked reversal of the Doppler flow signal when
the access flow is less than the dialysis pump
rate.25 Because thrombosis is less likely in fistu-
lae than grafts, the value of this type of monitor-
ing is more important for prosthetic grafts. How-
ever, we are performing additional studies to
examine technical factors and limitations in deter-
mining accurate Doppler signals in fistulae and
in the setting of turbulence.
character result from differences in graft geom-
etry and may provide additional independent risk
factors for thrombosis and markers of stenosis.
Recent data25 suggest that a simple Doppler test
for retrograde access flow on dialysis may be a
specific predictor for access stenosis and more
sensitive than access recirculation. However,
checking for retrograde access flow with this
method does not disrupt dialysis and takes only a
few seconds to perform. Additional Doppler char-
acteristics may increase the predictive value of
surveillance when used in conjunction with flow
measurements. This requires further study.
In summary, VF Doppler yields reproducible
access volume flow measurements that correlate
with ultrasound dilution measurements in pros-
thetic grafts. The indicated dialysis pump rate
yields reliable VF Doppler results when dialysis
pump rates are 400 mL/min or less. In our
facility, indicated pump rates at 400 mL/min
were near measured pump rates and differed
from measured pump rates by 7.3% (SD). The
VF Doppler method appears capable of accu-
rately determining access flow rates in the clini-
cally useful range of less than 1,600 mL/min.
Additional studies testing transducer design, sig-
nal processing, effects of turbulence, and clinical
correlation are in progress to further examine
this method of access evaluation.
REFERENCES
1. Feldman HI, Held PJ, Hutchinson JT, Stoiber E, Harti-
gan MF, Berlin JA: Hemodialysis vascular access morbidity
in the United States. Kidney Int 43:1091-1096, 1993
2. Mayers JD, Markell MS, Cohen LS, Hong J, Lundin P,
Friedman EA: Vascular access surgery for maintenance
hemodialysis: Variables in hospital stay. ASAIO J 38:113-
115, 1992
3. May RE, Himmelfarb J, Yenicesu M, Knights S, Iki-
zler TA, Schulman G, Hernanz-Schulman M, Shyr Y, Hakim
RM: Predictive measures of vascular access thrombosis: A
prospective study. Kidney Int 52:1656-1662, 1997
4. National Kidney Foundation: Clinical Practice Guide-
lines for Vascular Access. New York, NY, National Kidney
Foundation, 1997; pp 35–42
5. US Renal Data System: X. The cost-effectiveness of
alternative types of vascular access and the economic cost of
ESRD. Am J Kidney Dis 26:S140-S156, 1995 (suppl 2)
6. Strauch BS, O’Connell RS, Geoly KL, Grundlehner
M, Yakub YN, Teitjen DP: Forecasting thrombosis of vascu-
lar access with Doppler color flow imaging. Am J Kidney
Dis 19:554-557, 1992
7. Findley DE, Longley DG, Foshager MC, Letourneau
JG: Duplex and color sonography of hemodialysis arterio-
venous fistulas and grafts. Radiographics 13:983-999, 1993
The ability to detect turbulence and measure
other waveform characteristics related to the
character of flow (spectral broadening, pulsatil-
ity, and resistive indices) within the graft using
Doppler may prove to be a significant advantage
of using Doppler-based technology to measure
access flow. Low flow is an established risk
factor for thrombosis. Differences in waveform

940
8. Sands J: The role of color-flow Doppler ultrasound in
the management of hemodialysis accesses. ASAIO J 44:41-
43, 1998
WEITZEL ET AL
Malek P, Blake PG: A comparison of methods for the
measurement of hemodialysis access recirculation and ac-
cess blood flow rate. ASAIO J 44:62-67, 1998
9. Sullivan KL, Besarab A, Bonn J, Shapiro MJ, Gardiner
GA, Moritz MJ: Hemodynamics of failing dialysis grafts.
Radiology 186:867-872, 1993
10. Besarab A, Sullivan KL, Ross RP, Moritz MJ: Utility
of intra-access pressure monitoring in detecting and correct-
ing venous outlet stenosis prior to thrombosis. Kidney Int
47:1364-1373, 1995
11. BesarabA, Frinak S: The prevention of access failure:
Pressure monitoring. ASAIO J 44:35-37, 1998
12. Schwab SJ, Raymond JR, Saeed M, Newman GE,
Dennis PA, Bollinger RR: Prevention of hemodialysis fistula
thrombosis: Early detection of venous stenoses. Kidney Int
36:707-711, 1989
18. Wang E, Schneditz D, Nepomuceno C, Lavarias V,
Martin K, Morris AT, Levin NW: Predictive value of access
blood flow in detecting access thrombosis. ASAIO J 44:
M555-M558, 1998
19. Sands J, Glidden D, Miranda C: Hemodialysis access
flow measurement. Comparison of ultrasound dilution and
duplex ultrasonography. ASAIO J 42:M899-M901, 1996
20. Kirshbaum B, Compton A: Study of vascular access
by angiodynography. Am J Kidney Dis 25:22-25, 1995
21. Weitzel WF, Rubin JM, Swartz RD, Woltmann DJ,
Messana JM: Variable flow Doppler for hemodialysis access
evaluation: Theory and clinical feasibility. ASAIO J 46:65-
69, 2000
13. Besarab A, Sherman R: The relationship of recirculation
to access blood flow. Am J Kidney Dis 29:223-229, 1997
14. Depner TA: Hemodialysis access: In-line methods.
ASAIO J 44:38-39, 1998
22. Paun M, Beach K, Ahmad S, Hickman R, Meissner
M, Plett C, Strandness E: New ultrasound approaches to
dialysis access monitoring. Am J Kidney Dis 35:477-481,
2000
15. Krivitski NM: Theory and validation of access flow
measurement by dilution technique during hemodialysis.
Kidney Int 48:244-250, 1995
23. Keen M, Persig P, Gotch F: Non-invasive quantitive
measurement of hemodialysis access graft (HG) flow. J Am
Soc Nephrol 17:65A, 1985 (abstr 23)
16. Brosman PJ, Boereboom FTJ, Bakker CJ, Mali WPT,
Eikelboom BC, Blankestijn PJ, Koomans HA: Access flow
measurements in hemodialysis patients: In vivo validation
of an ultrasound dilution technique. J Am Soc Nephrol
7:966-969, 1996
24. Depner TA, Rizwan S, Sasi TA: Pressure effects on
roller pump flow during hemodialysis. ASAIO Trans 31:
M456-M459, 1990
25. Weitzel WF, Khosla N, Rubin JM: Retrograde hemo-
dialysis access flow during dialysis as a predictor of access
pathology. Am J Kidney Dis 37:1241-1246, 2001
17. Lindsay RM, Bradfield E, Rothera C, Kianfar C,