Full text of DOI:10.1053/ajkd.2001.27706

Buzz in the Axilla: A New Physical Sign in Hemodialysis Forearm
Graft Evaluation
Rajiv Agarwal, MD, and Grant McDougal, MD
● The value of physical signs in predicting flow rates or polytetrafluoroethylene (PTFE) hemodialysis graft survival
is not well known. Simultaneous physical examination and flow measurements were performed in 67 dialysis
patients with PTFE grafts over a period of 1 month. Physical signs included the presence and location of a palpable
thrill, location and auscultatory nature of the bruit, and presence or absence of a pulsation. Patients were followed
up for 1 year to detect the time to failure defined by angioplasty, clotting, or surgical revision. There was a fair
correlation between physical signs and graft flow rates. In a multivariate model, systolic blood pressure, location of
the thrill, and patient age were the best predictors of graft flow. During follow-up, there were 68 graft failures (30
grafts, angioplasty alone; 38 grafts, thrombosis followed by interventional or surgical revisions). The median
survival of a graft without a thrill in the axilla was 154 days compared with 321 days when a thrill was present (P ؍
0.05, log-rank test). Dialysis graft physical examination is a useful test to predict graft flows, as well as graft failure.
The single best test is the location of the thrill along the venous limb of the graft. Average flow rates for no thrill,
thrill but distal to the midarm, and axillary buzz were approximately 500, 750, and 1,000 mL/min, respectively. This
single physical sign also correlated with subsequent graft failure and should be incorporated into physical
examinations of all dialysis grafts.
© 2001 by the National Kidney Foundation, Inc.
INDEX WORDS: Physical examination; hemodialysis (HD) graft; hemodialysis (HD); vascular access.
machines (Fresenius Medical Care, Bad Homberg, Ger-
many) at the other two dialysis units with a median blood
flow of 400 mL/min for 3.0 to 4.5 hours (median, 4 hours)
three times weekly. CT 190 dialyzers (Baxter Healthcare
Inc, McGaw Park, IL) were used in those patients for whom
the dialyzer was reused.
ROSPECTIVE STUDIES on the natural his-
tory of graft patency in hemodialysis pa-
P
tients in the United States show poor patency
rates1 that entail costs to society, as well as
substantial patient morbidity.2 Numerous polytet-
rafluoroethylene (PTFE) graft monitoring tech-
niques have emerged in recent years,3-6 but physi-
cal examination of the graft that entails no
additional cost has been scarcely examined.7,8
Regular monitoring of interposition grafts to
detect early failure improves graft survival.9 Con-
versely, graft flow measurements require finan-
cial and time investments, which have precluded
the wide adoption of these technologies.2 Further-
more, the value of single or repeated graft flow
monitoring in clinical decision making has been
questioned.10-12 However, venous pressure moni-
toring in dialysis patients has yielded conflicting
results; some studies showed little utility in pre-
dicting failure,13 whereas others14,15 showed sub-
stantial benefit. Accordingly, we explored the
relationship between physical examination and
its correlation with graft flows and outcomes.
Physical Examination
A structured physical examination was performed by one
of the investigators (R.A.), who was blinded to the results of
graft flow measurements. Physical examination consisted of
the following factors: (1) presence or absence of a palpable
thrill near the venous anastomosis; (2) if the thrill was
palpable, it was ascertained whether the thrill was present in
the axilla or distal to the mid–upper arm; (3) presence or
absence of graft pulsation distal to the venous anastomosis;
(4) location of bruit in the axilla or distal to the mid–upper
arm; and (5) the nature of the bruit; whether it was continu-
ous or discontinuous.
Graft Flow Rate Measurement
Graft flow rates and recirculation percentages were deter-
mined by ultrasound dilution16 (Transonic Systems Inc,
From the Indiana University School of Medicine; and
Roudebush Veterans Affairs Medical Center, Indianapolis,
IN.
METHODS
Received March 19, 2001; accepted in revised form May
4, 2001.
Subjects
Address reprint requests to Rajiv Agarwal, MD, Indiana
University School of Medicine and Roudebush VA Medical
Center, 1481 West 10th St, Indianapolis, IN 46202. E-mail:
ragarwal@iupui.edu
© 2001 by the National Kidney Foundation, Inc.
0272-6386/01/3804-0016$35.00/0
Sixty-seven patients with PTFE grafts on chronic mainte-
nance hemodialysis therapy at three dialysis units affiliated
with Indiana University (Indianapolis, IN) were the subjects
of the study. All patients underwent dialysis using Cobe
Centry 3 dialysis machines (Cobe-Gambro Healthcare Inc,
Lakewood, CO) at one dialysis unit and Fresenius dialysis
doi:10.1053/ajkd.2001.27706
American Journal of Kidney Diseases, Vol 38, No 4 (October), 2001: pp 853-857
853

854
AGARWAL AND MCDOUGAL
Ithaca, NY) using standard procedures recommended by the
manufacturer. Graft flow monitoring was performed in dupli-
cate by a single operator (G.M.) over a 1-month period; the
average of two readings was used. Systolic blood pressure
was measured by an oscillometric cuff method using the
dialysis machine’s monitor immediately after each measure-
ment.
procedures with Statistica for Windows, Release 4.5 (Stat-
Soft Inc, Tulsa, OK). All P are two-sided and considered
significant at the 0.05 level. Mean Ϯ SD are reported
throughout the text, except when otherwise stated.
RESULTS
Average age of the subjects was 60 Ϯ 15
years, and there were 35 men (52%), 51 blacks
(76%), and 14 whites (21%). Causes of end-
stage renal disease were diabetes mellitus in 36
patients (54%), hypertension in 25 patients (37%),
and other causes in the remaining 9%.
Outcomes
Graft failure was defined as a composite outcome of graft
stenosis requiring angioplasty or surgical revision or clotting
of the graft. At the time of the study, no specific intervention
protocol was in place to maintain patency of dialysis grafts.
Patients had fistulograms performed based on the decisions
of physicians caring for these patients. Reasons included
difficult graft cannulation, excessive bleeding, or elevated
venous pressures.
A left forearm PTFE graft was present in 48
patients, and a right forearm graft, 19 patients.
The PTFE graft was in a straight configuration in
2 patients and a loop configuration in the rest.
Grafts were placed 42 Ϯ 26 months before the
study. Over the lives of the grafts, 2.97 Ϯ 4.27
clots (maximum, 29 clots; median, 2 clots) were
seen, yielding a rate of 0.99 Ϯ 1.16 clots/y
(maximum rate, 6 clots/y; median rate, 0.61
clots/y). In the previous 12 months, 0.97 Ϯ 1.15
clots were recorded (maximum, 4 clots; median,
1 clot). The number of procedures performed on
the graft over the life of the graft was 4.32 Ϯ
4.81 procedures (maximum, 27 procedures; me-
dian, 3 procedures), yielding a rate of 1.33 Ϯ
1.44 procedures/y (maximum, 6 procedures/y;
median, 0.94 procedures/y). In the previous 12
months, 1.35 Ϯ 1.44 graft procedures were re-
corded (maximum, 5 procedures; median, 1 pro-
cedure). All graft flow measurements and physi-
cal examinations were performed in January
2000. Average graft flow was 892 Ϯ 481 mL/
min.
Analysis
One-way analysis of variance was used to explore the
relationship between physical examination and graft flows
in a cross-sectional analysis. Based on the three components
of the physical examination, thrill, bruit, and pulsation, that
provided clinically independent information, a simple score
was created (Table 1).
One-way analysis of variance was used to explore the
relationship between this composite score and graft flow
measurements.
A multivariate model to predict graft flows was created
using multivariate regression. Those variables found to be
significant predictors of graft flow in a univariate analysis
were analyzed by forward stepwise multiple regression.
Tolerance was set at 1%. Standardized partial correlations,
which are estimates of the independent contribution of the
variable and adjusted coefficient of determination, an esti-
mate of goodness of fit, and the SD of residuals also were
calculated for the model.17
Kaplan-Meier survival curves were constructed from the
time of the end of graft flow monitoring to the time to
composite event, defined previously using the composite
graft score as a grouping variable.
All statistical analyses were performed using standard
On physical examination, there was no thrill
palpable in nine patients. In these patients, flow
averaged 499 Ϯ 274 mL/min. In those patients
(n ϭ 35) who had a thrill in the axilla, flow rate
averaged 1,057 Ϯ 455 mL/min, whereas those
patients who had a thrill palpable distal to the
midarm had a flow rate of 842 Ϯ 460 mL/min
(n ϭ 23; Fig 1). Patients who had a discontinu-
ous bruit had an average graft flow of 571 Ϯ 320
mL/min, whereas those who had a continuous
bruit had a graft flow of 954 Ϯ 472 mL/min (P ϭ
0.030). Similarly, patients without pulsatile grafts
had an average graft flow of 931 Ϯ 471 mL/min
(n ϭ 63), whereas those with pulsatile grafts had
an average graft flow of 555 Ϯ 372 mL/min (n ϭ
4; P ϭ 0.122).
Table 1. Scoring System of Graft
Physical Examination
Score
0
1
2
Thrill
Absent
Distal to the
midarm
In the axilla
(above
midarm)
Bruit
Pulsation
Discontinuous Continuous
Present Absent
NOTE. Graft function is graded on an ordinal scale in
which a sum of scores of thrill, bruit, and pulsation provides
a composite score of 0 through 4; a higher score repre-
sents better graft function by physical examination.

DIALYSIS GRAFT PHYSICAL EXAMINATION
855
DISCUSSION
Few investigators have addressed the value of
physical examination of the dialysis access to
make clinical decisions.7,8,18 The National Kid-
ney Foundation-Dialysis Outcomes Quality Ini-
tiative (NKF-DOQI) guidelines on vascular grafts
state, “Physical examination of an access graft
should be performed weekly and should include,
but not be limited to, inspection and palpation for
pulse and thrill at the arterial, mid, and venous
sections of the graft”9 (Opinion). There are nu-
merous other signs of graft dysfunction, such as
swelling of the arm or prolonged bleeding after
removal of the needle. The former is a late sign,
whereas the latter is difficult to observe by the
nephrologist. Therefore, only those signs readily
assessable at the bedside and used routinely in
clinical assessments of the graft were prospec-
tively evaluated. We did not find differences in
physical signs between arterial, venous, and loop
segments of the grafts, as recommended by the
NKF-DOQI. Conversely, clinical observations
during physical examination of the graft showed
that a more thorough assessment of the nature
and location of the thrill, bruit, and pulsation,
similar to what may be performed during a
cardiac physical examination (described in Meth-
ods) were found more useful. Because the PTFE
graft most often undergoes venous anastomotic
stenosis, we limited the examination to that seg-
ment because we believed there was maximal
information in that segment of the dialysis graft.
Preliminary studies showed that this test could
be readily performed and there was good agree-
ment between observers; therefore, a prospective
evaluation was undertaken.
Fig 1. Box plots (mean, SE, SD) of graft flow rates
versus the presence and location of the graft thrill by
palpation.
Results of the scoring system are shown in Fig
2. Patients with the maximum possible score of 4
had a graft blood flow rate of 1,059 mL/min,
which was almost twice the graft blood flow rate
(average, 542 mL/min) of patients who had scores
of 2 or less (Fig 2).
In univariate analysis, the following parame-
ters emerged as significant predictors of graft
blood flow rates: age (slope ϭ –9.6 mL/min/y;
intercept, 1,481 mL/min; r2 ϭ 0.09), systolic
blood pressure (slope ϭ 6.24 mL/min/mm Hg;
intercept, 0.646 mL/min; r2 ϭ 0.11), and all
physical signs, except pulsatility, noted previ-
ously. Cause of renal disease, sex, race, and
presence or absence of diabetes had no correla-
tion with graft blood flow. Results of the forward
stepwise multivariate analysis are listed in Table
2. Location of the thrill, patient age, and systolic
blood pressure at the time of examination
emerged as significant predictors and accounted
for 28.5% of the variance of graft flows.
The pathophysiological examination of physi-
Graft Failure Rates
Thirty-one patients remained event free dur-
ing follow-up. The remaining 36 patients under-
went 68 procedures, in which 24 patients had 38
clots requiring surgical revisions or interven-
tional declotting. In 17 patients, angioplasty alone
without declotting was the first event. The me-
dian event-free survival of a graft without a thrill
in the axilla was 154 days compared with 321
days when a thrill was present (P ϭ 0.05, log
rank test; Fig 3).
Fig 2. Box plots (mean, SE, SD) of graft flow rates
versus a simple score of the graft as listed in Table 1.

856
AGARWAL AND MCDOUGAL
Partial Correlation Coefficient
Table 2. Multivariate Model to Predict Graft Flows
B
SE of B
P
␤
R2
Age
Ϫ9.24
4.42
197
3.35
2.176
75
0.008
0.046
0.011
0.132
Ϫ.0295
0.231
0.299
0.008
0.125
0.128
Ϫ0.295
0.231
0.313
NA
Systolic blood pressure
Location of thrill
Intercept
542
355
NA
NA
NOTE. Overall R2 ϭ 0.285, adjusted R2 ϭ 0.251, F ϭ 8.38, P Ͻ 0.0001, and SD of residuals ϭ 408. Thrill is coded as listed
in Table 1.
Abbreviations: B, regression coefficient; NA, not applicable.
cal signs requires consideration of hemodynam-
ics. PTFE grafts are conduits between a high-
pressure arterial system and a low-pressure
venous system. The occurrence of significant
venous anastomotic stenosis will result in an
increase in outflow resistance that will increase
the magnitude of pulsation upstream and cause
the bruit to become discontinuous. The latter is
analogous to what may occur in a patient with
patent ductus arteriosus who develops pulmo-
nary hypertension; the machinery murmur
changes to a systolic murmur only. The location
of the thrill can be understood by consideration
of physical signs that appear in patients with
mitral regurgitation, which, in hemodynamic
terms, is a leak between a high-pressure (analo-
gous to the arterial segment of the access) and a
low-pressure system (analogous to the postanas-
tomotic vein). When the flow from high to low
pressures is unobstructed, as in severe mitral
regurgitation, the thrill is easily palpable. How-
ever, when the flow from a high- to low-flow
system is limited, such as in mild to moderate
mitral regurgitation, the thrill is barely felt or not
felt at all.
Our study included only PTFE grafts because
they are associated with high failure rates in
contrast to fistulas.19 We had a defined period
over which patients underwent physical examina-
tions and flow measurements (1 month) and a
1-year follow-up. This study design allowed us
to evaluate each patient’s outcome in a prospec-
tive manner. We show that physical examination
provides valuable information not only on predict-
ing PTFE graft flows, but also outcomes. Those
patients who had no thrill had almost half the
graft survival duration of those with a buzz in the
axilla. Conversely, pulsation of the graft was
uncommon; only five patients had this physical
sign and had substantially lower flows, as ex-
pected.
Graft flow measured by the Transonic device
is influenced by blood pressure and, as shown in
this study, the age of the patient. Operation of
this device entails costs and time of dialysis
personnel. Furthermore, Paulson et al10,11 noted
that single or even repeated graft flow monitor-
ing is insufficient to make clinical decisions
because of poor association with outcomes. Thus,
until we better understand the biological states of
venous stenosis and graft failure, we need to
develop tests that allow rapid, inexpensive, and
reliable evaluations of the graft. Results of the
current study reinforce the value of physical
examinations in monitoring dialysis grafts.
Several questions remain unanswered. Does a
change in physical examination results in a given
patient correlate with a decrease in graft flow
and, more importantly, predict graft thrombosis?
Fig 3. Kaplan-Meier survival curve of surviving
event-free grafts in patients who had an axillary thrill
versus those who did not. Patients with an axillary
thrill had a median graft survival of 321 days compared
with 154 days in those without an axillary thrill (P ؍
0.05, log-rank test).

DIALYSIS GRAFT PHYSICAL EXAMINATION
857
malfunction: Comparison of physical examination with US.
J Vasc Interv Radiol 7:15-20, 1996
Is a change in physical examination results suffi-
ciently reliable to intervene to prevent PTFE
graft thrombosis? What are the operating charac-
teristics of a change in physical examination
results and subsequent event rates? Can physical
examination refine the performance of graft flow
measurements? These questions need to be an-
swered in prospective adequately powered stud-
ies of patients with PTFE grafts.
In conclusion, our study emphasizes that physi-
cal examination is far from obsolete. We can gain
valuable information by integrating our sense of
touch, sight, and sound in evaluating the PTFE
graft, which not only predicts flows, but also
outcomes. Feeling the buzz in the axilla should
be integrated into all physical examinations of
PTFE grafts.
8. Safa AA, Valji K, Roberts AC, Ziegler TW, Hye RJ,
Oglevie SB: Detection and treatment of dysfunctional hemo-
dialysis access grafts: Effect of a surveillance program on
graft patency and the incidence of thrombosis. Radiology
199:653-657, 1996
9. National Kidney Foundation: DOQI Clinical Practice
Guidelines for Vascular Access. Am J Kidney Dis 30:S150-
S191, 1997 (suppl 3)
10. Paulson WD, Ram SJ, Birk CG, Zapczynski M,
Martin SR, Work J: Accuracy of decrease in blood flow in
predicting hemodialysis graft thrombosis. Am J Kidney Dis
35:1089-1095, 2000
11. Paulson WD, Ram SJ, Birk CG, Work J: Does blood
flow accurately predict thrombosis or failure of hemodialy-
sis synthetic grafts? A meta-analysis. Am J Kidney Dis
34:478-485, 1999
12. McDougal G, Agarwal R: Clinical performance char-
acteristics of hemodialysis graft monitoring. Kidney Int
60:762-766, 2001
REFERENCES
13. Bosman PJ, Boereboom FT, Eikelboom BC, Koom-
ans HA, Blankestijn PJ: Graft flow as a predictor of thrombo-
sis in hemodialysis grafts. Kidney Int 54:1726-1730, 1998
14. Besarab A, Sullivan KL, Ross RP, Moritz MJ: Utility
of intra-access pressure monitoring in detecting and correct-
ing venous outlet stenoses prior to thrombosis. Kidney Int
47:1364-1373, 1995
15. 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
1. Miller PE, Carlton D, Deierhoi MH, Redden DT, Allon
M: Natural history of arteriovenous grafts in hemodialysis
patients. Am J Kidney Dis 36:68-74, 2000
2. Hakim R, Himmelfarb J: Hemodialysis access failure:
A call to action. Kidney Int 54:1029-1040, 1998
3. Sands J: Hemodialysis access—The role of monitor-
ing. Introduction. ASAIO J 44:34, 1998
4. Besarab A, Lubkowski T, Frinak S, Ramanathan S,
Escobar F: Detecting vascular access dysfunction. ASAIO J
43:M539-M543, 1997
16. Krivitski NM: Theory and validation of access flow
measurement by dilution technique during hemodialysis.
Kidney Int 48:244-250, 1995
17. Zar JH: Biostatistical Analysis (ed 4). Upper Saddler
River, NJ, Prentice Hall, 1999
5. Lindsay RM, Blake PG, Malek P, Posen G, Martin B,
Bradfield E: Hemodialysis access blood flow rates can be
measured by a differential conductivity technique and are
predictive of access clotting. Am J Kidney Dis 30:475-482,
1997
18. Beathard GA: Physical examination of the dialysis
vascular access. Semin Dial 11:231-236, 1998
19. Agarwal R, Davis JL: Monitoring interposition graft
venous pressures at higher blood-flow rates improves sensi-
tivity in predicting graft failure. Am J Kidney Dis 34:212-
217, 1999
6. Sherman RA, Kapoian T: Recirculation, urea disequi-
librium, and dialysis efficiency: Peripheral arteriovenous
versus central venovenous vascular access. Am J Kidney Dis
29:479-489, 1997
7. Trerotola SO, Scheel PJ Jr, Powe NR, Prescott C,
Feeley N, He J, WatsonA: Screening for dialysis access graft