Kidney International, Vol. 64 (2003), pp. 272–280

A randomized controlled trial of blood flow and stenosissurveillance of hemodialysis grafts

SUNANDA J. RAM, JACK WORK, GLORIA C. CALDITO, J. MARK EASON, ASLAM PERVEZ,and WILLIAM D. PAULSON

Interventional Nephrology Section, Division of Nephrology and Hypertension, Department of Medicine and Department ofBiometry, Louisiana State University Health Sciences Center, Shreveport, Louisiana; Renal Division, Emory University Schoolof Medicine, Atlanta, Georgia; U.S. Vascular Access Holdings, Fresenius Medical Care, Inc., Winter Park, Florida

A randomized controlled trial of blood flow and stenosis sur- though the main mode of therapy is currently correctionveillance of hemodialysis grafts. of stenosis by percutaneous transluminal angioplasty

Background. It is widely accepted that hemodialysis graft (PTA) [1–3], the optimum timing of PTA has not yetsurveillance combined with correction of stenosis reducesbeen determined. The traditional method has been tothrombosis and prolongs graft survival. Nevertheless, few ran-

domized controlled trials have evaluated this approach. refer for angiography based upon clinical suspicion ofMethods. In this randomized controlled trial, 101 patients stenosis, or to wait until thrombosis to correct stenosis.

were assigned to control, flow (Qa), or stenosis groups, and This approach has fallen into disfavor because it haswere followed for up to 28 months. All patients had monthly Qabeen associated with a high rate of graft thrombosis andmeasured by ultrasound dilution and quarterly percent stenosis

measured by duplex ultrasound. Referral for angiography was failure [1, 2]. This has led to the proposal that surveil-based on the following criteria: (1) control group (N � 34), lance programs be used to detect graft dysfunction orclinical criteria; (2) flow group (N � 32), Qa �600 mL/min stenosis so that PTA can be applied before thrombosisor clinical criteria; and (3) stenosis group (N � 35), stenosis

[1–4]. The most widely recommended surveillance method�50% or clinical criteria. Stenosis �50% during angiographywas corrected by preemptive percutaneous transluminal angio- has been monthly Qa measurements.plasty (PTA). The rationale for Qa surveillance depends upon two

Results. The preemptive PTA rate in the control group (0.22/ assumptions. The first is that monthly Qa measurementspatient year) was two thirds the rate in the flow group (0.34/accurately predict thrombosis. Although this assumptionpatient year), and was highest in the stenosis group (0.65/patient

year, P � 0.01). The percentage of grafts that thrombosed seems plausible, we [5, 6] and others [7] have found thatwas similar in the control (47%) and flow groups (53%), but Qa is an inaccurate predictor of thrombosis, and thisreduced in the stenosis group (29%, P � 0.10). Two-year graft conclusion has been supported by a meta-analysis [8].survival was similar in the control (62%), flow (60%), and

Moreover, reanalysis of two widely cited studies [9, 10]stenosis groups (64%) (P � 0.89).Conclusion. Qa and stenosis surveillance were not associated that have supported Qa surveillance shows their results

with improved graft survival, although thrombosis was reduced are similar to ours [11]. Qa is an inaccurate predictor ofin the stenosis group. The most important factors in this result thrombosis mainly because wide hemodynamic variationmay be that monthly Qa and quarterly stenosis measurements

is present throughout dialysis, and this impairs the repro-were not accurate or timely indicators of risk of thrombosis orprogressive stenosis. This study does not support the concept ducibility of Qa measurements [12, 13]. Failure to accu-that Qa or stenosis surveillance are superior to aggressive clini- rately predict thrombosis would impair any potentialcal monitoring.

benefit from preemptive intervention.The second assumption in surveillance is that timely

intervention reduces thrombosis and prolongs graft life.Progressive stenosis of hemodialysis grafts leads toMost studies that have evaluated surveillance, whetherdecreased blood flow (Qa), stasis, and thrombosis. Al-by Qa or venous pressure, have used historic control groups,sequential groups, or have been retrospective [14–23].

Key words: randomized control trial, hemodialysis, blood vessel pros-It is well documented that inadequate randomizationthesis, graft surveillance, blood flow, stenosis, thrombosis, cumulative

survival. and historic control groups promote overestimation ofa treatment benefit [24–28].Received July 17, 2002

The few randomized controlled trials have not settledand in revised form October 11, 2002, and January 16, 2003Accepted for publication March 4, 2003 this issue. In a study that supports Qa surveillance in

reducing thrombosis, the control group had a high throm- 2003 by the International Society of Nephrology

272

Ram et al: Hemodialysis graft surveillance 273

bosis rate because of multiple thromboses in a small The 101 study patients were randomly assigned tonumber of grafts [29]. Two studies have evaluated steno- control, flow, or stenosis groups, and were followed untilsis surveillance by duplex ultrasound [30, 31]. In a study graft failure or the end of the study in February 2001.of new grafts, surgical revision for more than 50% steno- Group assignments were supervised by a single investiga-sis improved graft survival [30]. In the second study, tor who was not involved in patient care (S.J.R.); no onepreemptive PTA failed to improve graft survival, but else was aware of group assignments. Randomizationthat study has been criticized because the surveillance was determined by a random number table that wasgroup had more prior interventions than the control generated for this study (Systat 5.2 for Macintosh, Rich-group [31]. In a subset analysis of 21 patients without mond, CA, USA). Blocked randomization was used toprior interventions, PTA was associated with improved ensure that numbers of subjects in each group were ap-graft survival [32]. Finally, in a preliminary report, Qa proximately equal [36]. Only one graft/patient was in-surveillance failed to improve graft survival (abstract; cluded in the study.Moist et al, J Am Soc Nephrol 12:34A, 2002). However, All patients underwent monthly Qa measurements bythe control group underwent dynamic venous pressure ultrasound dilution and quarterly stenosis measurementssurveillance, which may have obscured any potential by duplex ultrasound. Referral for angiography, how-benefit from Qa surveillance.

ever, was restricted to the following criteria: (1) controlIt is noteworthy that the accuracy of venous pressuregroup, clinical criteria alone; (2) flow group, Qa �600in predicting thrombosis is also controversial [33, 34]. InmL/min or clinical criteria; (3) stenosis group, stenosisaddition, in a preliminary report, a randomized con-�50% or clinical criteria.trolled trial found that static venous pressure surveillance

A single investigator (S.J.R.) was aware of the resultsdid not improve graft survival (abstract; Dember et al,of Qa and stenosis measurements. Clinical criteria forJ Am Soc Nephrol 13:600A, 2002).referral were determined by attending nephrologists andThus, despite wide acceptance that surveillance is ben-were based upon suspicion of graft dysfunction or steno-eficial, the foregoing shows that a benefit has not yetsis (Table 1). Stenosis was defined as percent reductionbeen established. Clearly, further studies are needed toin luminal diameter. For all groups, stenosis of �50% onevaluate the effectiveness of surveillance in reducingangiography was corrected by PTA. Thrombosed graftsthrombosis and improving graft survival. The relativeunderwent thrombolysis and angiography with correc-benefit of surveillance by Qa vs. stenosis is also unknown.tion of stenosis by PTA. Grafts with unsuccessful PTAIn this randomized controlled trial, we evaluated whetherwere defined as failures and were referred for surgicalintervention based upon Qa or stenosis provide a benefit

when added to intervention based upon clinical criteria revision or abandoned. Revision was considered failurealone. because the venous anastomosis was abandoned and the

graft was extended proximally up the arm.

METHODS Patient data, measurements, proceduresPatients and protocol

A computer database was maintained for recordingWe followed the Consolidated Standards of Reporting diagnoses, demographic information, graft measurements

Trials statement for documenting randomized controlled and events, and serum albumin levels. Albumin levelstrials (Fig. 1) [35]. Patients with polytetrafluoroethylene were measured by autoanalyzer at study entry. Qa was(PTFE) grafts from two dialysis units (Dialysis Clinic, measured monthly by ultrasound dilution (Transonic Sys-Inc. and Dialysis Center of Shreveport) were enrolled tems, Inc., Ithaca, NY, USA) [37]. Measurements werein the study at various times beginning in November taken in duplicate at any time during dialysis except1998. The study protocol was approved by the Institu- within the first and last 15 minutes. When two measure-tional Review Boards of Louisiana State University and

ments differed by more than 10%, a third measurementthe parent corporations of the dialysis units (Dialysiswas taken. Measurements were then averaged.Clinic, Inc., Nashville, TN, USA; Fresenius Medical Care,

A single qualified technician (J.M.E.) performed quar-Walnut Creek, CA, USA). After obtaining written in-terly duplex ultrasound studies with a Siemens Sonolineformed consent, patients underwent a Qa measurementVersa Scanner and a 7.5 or 5.0 MHz linear transducerby ultrasound dilution, and stenosis and Qa measurements(Siemens Medical Systems, Inc., Issaquah, WA, USA).by duplex ultrasound, in order to establish baseline values.He was blinded to group assignment and previous ultra-From a total of 128 eligible patients, 101 completed thesesound studies. The graft, arterial and venous anastomo-measurements and were enrolled. Analysis of enrolledses, inflow artery, and outflow vein were evaluated inpatients was by “intention-to-treat” (all enrolled patientssagittal and transverse planes with and without color.were included in the analysis). The remaining 27 patientsImages were recorded on videotape for subsequent anal-were not enrolled because graft failure, transfer, trans-ysis. Location and percentage of stenosis were recorded.plantation, or infected graft removal occurred before

entry requirements could be met. Flow velocity was measured with the transducer at 60�

Ram et al: Hemodialysis graft surveillance274

Fig. 1. Flow diagram of patient progressthrough phases of randomized trial that testedgraft survival by Kaplan-Meier analysis. *Cen-sored grafts include (1 ) current, grafts thatwere still being monitored at the time thestudy ended; or (2 ) other end point, graftscensored because of death, transfer, trans-plant, recovery of renal function, graft re-moval due to infection, or graft ligation unre-lated to stenosis or thrombosis.

Table 1. Reasons for preemptive percutaneous transluminal angioplasty (PTA) in each group

Control group Flow group Stenosis group

Number of preemptive PTAs 7 15 26Criteria for preemptive PTAa

Prescribed blood pump speed not attained 5 2 —High dynamic venous pressureb 3 2 —Excessive bleeding postdialysis 4 — —Swollen arm 1 1 1Qa �600 mL/min — 12 —Stenosis �50% — — 25

Qa is blood flow.a Some grafts had more than one criterion for preemptive PTAb Measured at prescribed or actual blood pump speed (if prescribed blood pump speed not attainable)

with respect to the plane of the graft. Qa was computed were analyzed by chi-square test. Kaplan-Meier survivalanalysis was used to compare outcomes among the threefrom luminal diameter and time-averaged flow velocity.groups. The proportional hazards model was used to

Analysis determine whether adjustment for the variables in Ta-ble 2 influenced outcome. A backward elimination proce-Data are reported as mean � SE. Normally distributed

data were analyzed by analysis of variance. Categories dure was used to determine which variables were signifi-

Ram et al: Hemodialysis graft surveillance 275

Table 2. Patient and graft characteristics of the three groups at study entry

Control group Flow group Stenosis group P value

No. of patients 34 32 35African American race 30 (88%) 32 (100%) 34 (97%) 0.07Female 21 (62%) 21 (66%) 17 (49%) 0.33Diabetes mellitus 19 (56%) 12 (38%) 17 (49%) 0.32Patient age years 53�2a 55�3 59�3 0.28Serum albumin g/dL 3.9�0.1 4.0�0.8 3.8�0.1 0.38Graft age days 276�76 502�135 414�85 0.29Median graft age days 94 225 196 0.20Range of graft ages days 21–2355 45–2918 24–2099No. of prior PTAs 1.1�0.3 0.8�0.2 0.8�0.2 0.64Grafts without prior procedures 22 (65%) 19 (59%) 26 (74%) 0.42Loop configuration of graft 15 (44%) 14 (44%) 14 (40%) 0.93Qa (mL/min), by ultrasound dilution 1333�123 1219�114 1253�106 0.77Qa (mL/min), by duplex ultrasound 1102�99 1147�101 985�76 0.44Percent stenosis 35�3 37 �4 35�4 0.94

Qa is blood flow.a Mean � SE

Table 3. Comparison of graft events in the three groups

Control group Flow group Stenosis group P value

Preemptive PTA rate number/patient year 0.22 0.34 0.65b

Thrombosis rate number/patient year 0.68 0.91a 0.50Percentage of grafts that thrombosed during study 47.1% 53.1% 28.6% 0.10Kaplan-Meier survival analysis

Probability grafts were thrombosis-free at 1 year 0.45 0.52 0.70 0.12Probability grafts were event-freec at 1 year 0.34 0.47 0.36 0.55Probability grafts survived 2 years 0.62 0.60 0.64 0.89a P � 0.02; b P � 0.01 comparison vs. control groupc Event-free denotes no thrombosis or preemptive percutaneous transluminal angioplasty (PTA)

cant. A Z statistic was used to compare rates (events/ were for group-specific criteria (Qa �600 mL/min or ste-patient year). Significance was set at P � 0.05, and P nosis �50%, respectively). Group-specific criteria domi-values were two-tailed. nated in these groups because these criteria tend to occur

earlier than clinical criteria.The high rate of preemptive PTA in the stenosis groupRESULTS

was associated with reduced thrombosis (Table 3). TheCharacteristics of control, flow, and stenosis groups

thrombosis rate was highest in the flow group, but thisTable 2 compares patient and graft characteristics of is misleading in that it was caused by multiple thrombo-

the three groups at study entry. There were no statisti- ses in three grafts. In contrast, when the percentage ofcally significant differences between the groups. Most grafts that thrombosed at least once was compared, thepatients were African Americans, although the percent- control and flow groups were similar, and the stenosisage in the control group was slightly less than in the group remained lower.other two groups. Mean graft age was lowest in the

Kaplan-Meier analysis shows there was a trend towardcontrol group, but this difference was not significant.

delayed first thrombosis after study entry in the stenosisThe three groups had similar mean Qas, stenoses, andgroup (P � 0.12) (Table 3, Fig. 2). We used the propor-numbers of prior procedures.tional hazards model to determine whether adjustmentfor the variables in Table 2 altered this result. The onlyPreemptive PTA and thrombosissignificant variable in the multivariable model was QaOverall PTA and thrombosis rates were substantial,by duplex ultrasound at study entry (P � 0.008), andindicating progressive stenosis characterized the graftsinclusion of this variable strengthened the trend of de-in this study (Table 3). The preemptive PTA rate in thelayed thrombosis in the stenosis group (P � 0.05). It iscontrol group was two thirds the rate in the flow group,notable that graft age, number of prior PTAs, whetherand was highest in the stenosis group. Several clinicala graft had prior procedures, and Qa by ultrasound dilu-criteria led to referral for PTA in the control group

(Table 1). In the flow and stenosis groups, most PTAs tion at study entry were not significant. Figure 3 shows

Ram et al: Hemodialysis graft surveillance276

Fig. 2. Kaplan-Meier comparison of thrombosis-free graft survival inthe three groups (P � 0.12). Circles indicate thrombosis (stepdowns) Fig. 4. Last monthly blood flow (Qa) values before thrombosis in con-or censoring (horizontal portions). trol (N � 18), flow (N � 31), and stenosis (N � 11) groups (P � 0.77).

Because of repeated thrombosis in some grafts, Qa was not availablebefore all thromboses. Solid circles indicate mean � SE. Dashed lineindicates Qa � 600 mL/min.

sis. The figure also shows that four thrombosed graftsin the flow group met criteria for angiography beforethrombosis. Three thrombosed before angiographycould be scheduled and one patient missed his appoint-ment. These thromboses did not appear to affect graftsurvival because three had successful PTAs postthrom-bosis, and the fourth was already unsalvageable in thata recent preemptive PTA had yielded a minimal im-provement in Qa.

Graft survival

Kaplan-Meier analysis was used to determine whethersurveillance improved graft survival, with failure definedas graft abandonment or surgical revision. There wereno significant differences in survival during follow-upperiods of up to 28 months (P � 0.89) (Table 3, Fig. 5).Fig. 3. Kaplan-Meier comparison of event-free graft survival in theIn the proportional hazards model, once again the onlythree groups (P � 0.55). Circles indicate events (thrombosis or preemp-

tive percutaneous transluminal angioplasty (PTA) at stepdowns) or significant variable was Qa by duplex ultrasound (P �censoring (horizontal portions). 0.001), and it did not change the conclusion that survival

was similar in the three groups (P � 0.45). Only onegraft in each group underwent revision; in each case ajump graft bypassed the venous anastomosis. Redefiningthat time to first event (thrombosis or preemptive PTA)graft failure solely as abandonment (but not revision)was similar in the three groups (P � 0.55).did not alter the proportional hazards result (P � 0.32).In order to understand why Qa surveillance failed to

reduce thrombosis, we examined the last monthly QasConfidence intervals (CIs) and significancebefore thrombosis in all three groups. Figure 4 shows

Because many of the differences in Table 3 were notthat the last Qa was generally well above the 600 mL/significant, we computed 95% confidence intervals (CIs)min threshold that was used to refer for PTA in the flow

group. Thus, Qa was an inaccurate predictor of thrombo- for differences between flow or stenosis group vs. control

Ram et al: Hemodialysis graft surveillance 277

Fig. 5. Kaplan-Meier comparison of cumulative graft survival in thethree groups (P � 0.89). Circles indicate graft failure (stepdowns) orcensoring (horizontal portions).

Fig. 6. 95% CIs for differences between flow or stenosis group andcontrol group. The following variables were analyzed. (A ) Proportionof grafts that thrombosed during study. (B ) Kaplan-Meier probabilitythat grafts remained thrombosis-free 1 year after study entry. (C )group (Fig. 6). We considered the proportion of graftsKaplan-Meier probability that grafts survived for 2 years after study

that thrombosed during the study, and the probabilities entry. For example, probability of graft survival in the stenosis groupminus survival in the control group may have been as low as �0.25 orthat grafts remained thrombosis-free at 1 year or survivedas high as �0.27. CIs that exclude zero would indicate differencesfor 2 years after study entry. For thrombosis (Fig. 6 Abetween groups were significant.

and B), differences between the stenosis and controlgroups were nearly significant (CIs nearly excluded 0).These CIs suggest stenosis surveillance reduced thrombo- studies [14–23] that have found a benefit for surveillancesis. For graft survival (Fig. 6C), the CIs indicate that if may have been biased because of inadequate randomiza-Qa and stenosis surveillance increased the probability of tion or historic control groups.survival, it was by no more than 0.24 and 0.27, respectively. While planning this study, we were strongly influenced

by the landmark studies of Schwab et al [15] and Besarabet al [16]. Both nonrandomized studies observed approx-DISCUSSIONimately fourfold reductions in access failure after imple-To our knowledge, this is the first randomized con-menting surveillance. Thus, we were surprised to find notrolled trial that has studied the influence of blood flowimprovement in graft survival in this study’s surveillance(Qa) surveillance on graft survival, and the first to com-groups. This raises an important question: how small apare Qa and stenosis surveillance. We found that Qa anddifference in 2-year graft survival could we have de-stenosis surveillance were not associated with improvedtected? Figure 6C shows 95% CIs for differences betweengraft survival, although thrombosis was reduced in thesurveillance groups and control group. The CIs show,stenosis group. The substantial thrombosis and PTAfor example, that we cannot exclude the possibility thatrates in this study indicate that progressive stenosis char-the probability of graft survival in the flow group was asacterized our graft population (Table 3, Figs. 2 and 3).much as 0.24 higher than in the control group. Thus, thisThus, our patients were well-suited for testing the hy-study shows that if a surveillance benefit exists, it ispothesis that surveillance with intervention prolongsmodest at best and far less than in the two landmarkgraft life.studies [15, 16]. Nevertheless, in our opinion, an improve-The results of this study underscore the need for ran-ment in survival probability smaller than 0.24 may be largedomization with concurrent controls when testingenough to justify surveillance. Thus, larger randomizedhypotheses [24–28]. The purpose of randomization is tocontrolled trials are needed to better define whetherhelp ensure an unbiased assessment of a treatment bene-surveillance provides a significant survival benefit.fit by selecting study groups that are equivalent in all

respects other than the treatment itself [38]. Previous In contrast to graft survival, this study supports the

Ram et al: Hemodialysis graft surveillance278

Table 4. Factors that may account for failure of surveillance to prolong graft life in this study

1. Timely intervention may reduce thrombosis rates without ultimately influencing graft survival;2. Surveillance may provide an improvement in graft survival that is modest at best and requires a larger sample size to detect;3. Monthly blood flow (Qa) measurements do not accurately detect progressive stenosis;4. Quarterly stenosis measurements are not sufficiently frequent to detect progressive stenosis in many grafts;5. Unnecessary or premature percutaneous transluminal angioplasty (PTAs) may adversely affect graft survival;6. Aggressive clinical monitoring may provide most or all of the benefit provided by preemptive intervention;7. The benefit of surveillance may be limited to a subset of grafts.

concept that aggressive preemptive PTA reduces throm- (N.M. Krivitski, personal communication) recommendedthat measurements should ideally be in the first hour,bosis. The stenosis group, which had the highest preemp-

tive PTA rate (0.65/patient year), had the lowest throm- but he also noted that this might be impractical [12]. Inorder to settle this issue, we evaluated seven consecutivebosis rate and lowest percentage of thrombosed grafts.

As the 95% CIs in Figure 6 A and B show, these differ- sessions in 51 patients [12]. We found wide hemodynamicvariation from the beginning of dialysis, indicating thatences were not quite significant. Given that we were

specifically seeking reduced thrombosis in the surveil- hemodynamic stability does not characterize the earlypart of a session as is so widely believed. We concludedlance groups, one-tailed rather than two-tailed P values

could be used. This would reduce P by one half and that any restriction on timing of measurements is withoutbenefit [12, 13]. Thus, it is unlikely that early Qa measure-thereby yield significance (CIs would exclude zero). Ste-

nosis surveillance was potentially beneficial because, by ments would have altered our results.Some have suggested that decrease in Qa (Qa) is areducing the unscheduled PTA rate, it may have reduced

hospitalization and use of central venous catheters. It is more accurate predictor of thrombosis than a single Qa

value [10]. Thus, it might be argued that surveillancepossible, however, that this advantage was offset by thestenosis group’s high rate of procedures, some of which based upon Qa might have been more successful. How-

ever, the assertion that Qa is superior was based uponmight have been unnecessary or premature (see below).It is important to explore why surveillance did not a comparison of two separate studies, rather than a com-

parison of Qa and Qa within the same study [9, 10]. Theimprove graft survival in this study (Table 4). The mostobvious explanation may be that timely intervention re- apparent superiority of Qa is misleading because many

grafts with normal Qas thrombose before Qa can beduces thrombosis without ultimately influencing graftsurvival. However, a number of other possibilities should measured [5, 10]. This has caused a bias in favor of Qa

[13]. We have used logistic regression to show that whenalso be considered.A key factor in the flow group may be that Qa is not Qa is combined with single Qa in the same model, Qa

is not a significant predictor of thrombosis [40]. Further,accurate in detecting progressive stenosis or predictingthrombosis [5–8, 11]. Review of monthly Qa values before combined prediction of thrombosis by either Qa �600

mL/min or Qa �20% has a sensitivity of only 59% withthrombosis shows most thromboses were preceded byQas that were well above the 600 mL/min threshold used a false positive rate of 23% [5, 11]. Thus, it is unlikely

that Qa would have enhanced the benefit of surveillanceto refer for preemptive PTA in the flow group (Fig. 4).This poor predictive accuracy is mainly caused by wide in this study.

In the stenosis group, quarterly stenosis measurementshemodynamic variation during dialysis, which impairsthe reproducibility of Qa measurements [12, 13]. Rapid may not have been sufficiently frequent to provide a

timely warning of progressive stenosis. For example, ap-progression of stenosis in young grafts is also probablya factor (abstract; Atray et al, J Am Soc Nephrol 12:281A, proximately one third of our patients’ grafts thrombose

within 90 days of placement. We have recently used2001; abstract; Zaman et al, J Am Soc Nephrol 10:233A,1999). duplex ultrasound and angiography studies to show that

early thrombosis is associated with rapid progression ofSome might argue that patients are hemodynamicallystable early in a dialysis session, so that if we had mea- stenosis at the venous anastomosis (abstract; Atray et al,

J Am Soc Nephrol 12:281A, 2001; abstract; Zaman et al,sured Qa in the first 30 minutes rather than during thegreater part of a session, we might have had greater J Am Soc Nephrol 10:233A, 1999}. Thus, stenosis measure-

ments may need to be more frequent than quarterly.accuracy in detecting stenosis. Concern that such a re-striction is impractical led Rehman et al [39] to study The possibility that unnecessary or premature PTAs

impaired graft survival should also be considered. In thethis issue. They concluded that measurements could bedone anytime in the first 90 minutes (and often in the stenosis group, intervention was based on stenosis �50%

without regard to whether stenosis was progressive. Withfirst 120 to 150 minutes) of a session. The senior scientistof the manufacturer of the ultrasound dilution device respect to the flow group, Qa measurements frequently

Ram et al: Hemodialysis graft surveillance 279

yield false positive predictions of thrombosis [5–8, 11]. should not be applied to all patients unless it is shownto be superior to clinical monitoring.For example, hemodynamic variation during dialysis

As a final consideration, recall that in the two random-may cause a temporary but misleading decrease in Qa.ized controlled trials that found stenosis surveillance im-Subsequent PTA of a stable stenosis might then shortenproves survival, the investigators studied new grafts [30]graft life by stimulating neointimal hyperplasia. [41]. Itor grafts without prior procedures [32]. This suggestsis noteworthy that Dember, Holmberg, and Kaufmanthat the benefit of surveillance might be restricted to a[33] have similarly found that static venous pressure hassubset of grafts. In contrast, our proportional hazardsa poor predictive accuracy that may result in many un-model showed no evidence of an influence of graft agenecessary, invasive, and costly procedures. This under-or prior procedures on thrombosis or survival. Thus, wescores the need to develop more accurate methods offound no evidence that a subset of grafts benefited fromassessing risk of thrombosis so that the benefit of proce-surveillance.dures can be assured.

Finally, a high baseline PTA rate might help explainwhy the control group’s graft survival was similar to the CONCLUSIONsurveillance groups. The preemptive PTA rate in the

This randomized controlled trial failed to show thatcontrol group (0.22/patient year) was two thirds the rateQa surveillance reduced thrombosis or prolonged graftin the flow group (0.34/patient year). By comparison,life, whereas stenosis surveillance did reduce thrombosisthis rate is moderately higher than the baseline rate ofwithout prolonging graft life. This study provides many0.09/patient year by McCarley et al [20]. It is possibleinsights that may assist in designing the larger (probably

that our higher rate provided most of the benefit that maymulticenter) randomized controlled trials that are needed

follow from preemptive intervention. This rate likely to further evaluate the costs and benefits of surveillance.reflects the fact that the nephrologists in this study were We believe that until the results of such trials are avail-members of an interventional nephrology program. able, it is not possible to make definitive evidence-basedThus, the nephrologists and dialysis staff made clinical recommendations concerning graft surveillance.recognition of graft problems a high priority.

If a high baseline PTA rate prolonged graft survival, ACKNOWLEDGMENTSthen survival in the control group should have been un-

The authors thank Dialysis Clinic, Inc., for a grant that fundedusually long. The cumulative 800-day survival in the con-part of this study, and Michael Perry (President) and Lori Ferrell

trol group was 62% after study entry. Since the median (Operations Manager), U.S. Vascular Access Holdings, Fresenius Med-ical Care, Inc., for their assistance in providing duplex ultrasoundgraft age in the control group at study entry was 94 days,procedures during this study. This study was presented in part at thethis survival applies to 894 days after graft placement.2001 ASN/ISN World Congress of Nephrology.

In comparison, National Kidney Foundation DialysisAddress correspondence to William D. Paulson, M.D., LouisianaOutcomes Quality Initiative (NKF K/DOQI) evaluated

State University Health Sciences Center, P.O. Box 33932, Shreveport,the summed data from a number of studies [2]. TheLA 71130.

summary found that grafts without surveillance had a E-mail: wpauls@lsuhsc.educumulative 800-day survival of 45% after first use. Thus,our control group had a longer survival than K/DOQI. REFERENCES

As Miller et al [42] have reviewed, however, compari-1. Hakim R, Himmelfarb J: Hemodialysis access failure: A call to

sons of graft survival are difficult. For example, consider- action. Kidney Int 54:1029–1040, 19982. Schwab SJ: Vascular access for hemodialysis. Kidney Int 55:2078–ing that the K/DOQI data are as much as one or two

2090, 1999decades old, our longer survival may be due to changes3. National Kidney Foundation: K/DOQI Clinical Practice Guide-

in medical practice apart from clinical monitoring. In lines for Vascular Access. Am J Kidney Dis 37(Suppl 1):S137–S181,2001addition, previous studies have been generally nonran-

4. Besarab A: Access monitoring methods. Blood Purif 18:255–259,domized, patient comorbidities and access histories may 2000differ, and there is little consistency in the way graft 5. Paulson WD, Ram SJ, Birk CG, et al: Accuracy of decrease in

blood flow in predicting hemodialysis graft thrombosis. Am J Kid-survival has been measured [42]. Nevertheless, it is possi-ney Dis 35:1089–1095, 2000ble that aggressive clinical monitoring accounts for the 6. Paulson WD, Ram SJ, Work J: Use of vascular access blood flow

control group’s improved graft survival as compared with to evaluate vascular access [letter]. Am J Kidney Dis 38:916, 20017. McDougal G, Agarwal R: Clinical performance characteristicsolder data. Thus, this study is consistent with the concept

of hemodialysis graft monitoring. Kidney Int 60:762–766, 2001that aggressive clinical monitoring with easy access to 8. Paulson WD, Ram SJ, Birk CG, et al: Does blood flow accuratelyPTA provides most of the benefit that may follow from predict thrombosis or failure of hemodialysis synthetic grafts? A

meta-analysis. Am J Kidney Dis 34:478–485, 1999preemptive intervention. This issue is important because9. May RE, Himmelfarb J, Yenicesu M, et al: Predictive measures

universal application of surveillance would be costly and of vascular access thrombosis: A prospective study. Kidney Int52:1656–1662, 1997burdensome. We believe that Qa or stenosis surveillance

Ram et al: Hemodialysis graft surveillance280

10. Neyra NR, Ikizler TA, May RE, et al: Change in access blood 27. Sacks J, Chalmers TC, Smith H: Randomized versus historicalcontrols for clinical trials. Am J Med 72:233–240, 1982flow over time predicts vascular access thrombosis. Kidney Int

54:1714–1719, 1998 28. Schulz KF, Chalmers I, Hayes RJ, Altman DG: Empirical evi-dence of bias. Dimensions of methodological quality associated11. Paulson WD, Ram SJ, Work J: Use of vascular access blood flow

to evaluate vascular access [letter]. Am J Kidney Dis 37:451–452, with estimates of treatment effects in controlled trials. JAMA273:408–412, 19952001

12. DeSoto DJ, Ram SJ, Faiyaz R, et al: Hemodynamic reproducibility 29. Sands JJ, Jabyac PA, Miranda CL, Kapsick BJ: Interventionbased on monthly monitoring decreases hemodialysis accessduring blood flow measurements of hemodialysis synthetic grafts.

Am J Kidney Dis 37:790–796, 2001 thrombosis. ASAIO J 45:147–150, 199930. Mayer DA, Zingale RG, Tsapogas MJ: Duplex scanning of ex-13. Paulson WD: Blood flow surveillance of hemodialysis grafts and

the dysfunction hypothesis. Semin Dial 14:175–180, 2001 panded polytetrafluoroethylene dialysis shunts: Impact on patientmanagement and graft survival. Vasc Surg 27:647–658, 199314. Schwab SJ, Oliver MJ, Suhocki P, McCann R: Hemodialysis

arteriovenous access: Detection of stenosis and response to treat- 31. Lumsden AB, MacDonald MJ, Kikeri D, et al: Prophylactic bal-loon angioplasty fails to prolong the patency of expanded polytetra-ment by vascular access blood flow. Kidney Int 59:358–362, 2001

15. Schwab SJ, Raymond JR, Saeed M, et al: Prevention of hemodialy- fluoroethylene arteriovenous grafts: Results of a prospective ran-domized study. J Vasc Surg 26:382–392, 1997sis fistula thrombosis. Early detection of venous stenoses. Kidney

Int 36:707–711, 1989 32. Martin LG, MacDonald MJ, Kikeri D, et al: Prophylactic angio-plasty reduces thrombosis in virgin PTFE arteriovenous grafts with16. Besarab A, Sullivan KL, Ross RP, Moritz MJ: Utility of intra-

access pressure monitoring in detecting and correcting venous out- greater than 50% stenosis: Subset analysis of a prospective study.J Vasc Intervent Radiol 10:389–396, 1999let stenoses prior to thrombosis. Kidney Int 47:1364–1373, 1995

17. Roberts AB, Kahn MB, Bradford S, et al: Graft surveillance 33. Dember LM, Holmberg EF, Kaufman JS: Value of static venouspressure for predicting arteriovenous graft thrombosis. Kidney Intand angioplasty prolongs dialysis graft patency. J Am Coll Surg

183:486–492, 1996 61:1899–1904, 200234. Frinak S, Zasuwa G, Dunfee T, et al: Dynamic venous access18. Safa AA, Valji K, Roberts AC, et al: Detection and treatment

of dysfunctional hemodialysis access grafts: Effect of a surveillance pressure ratio test for hemodialysis access monitoring. Am J KidneyDis 40:760–768, 2002program on graft patency and the incidence of thrombosis. Radiol-

ogy 3:653–657, 1996 35. Moher D, Schulz KF, Altman D: The CONSORT statement:Revised recommendations for improving the quality of reports of19. Cayco AV, Abu-Alfa AK, Mahnensmith RL, Perazella MA:

Reduction in arteriovenous graft impairment: Results of a vascular parallel-group randomized trials. JAMA 285:1987–1991, 200136. Hulley SB, Cummings SR, Browner WS, et al: Designing Clinicalaccess surveillance protocol. Am J Kidney Dis 32:302–308, 1998

20. McCarley P, Wingard RL, Shyr Y, et al: Vascular access blood Research (2nd ed), Philadelphia, Lippincott, Williams & Wilkins,2001, pp 149flow monitoring reduces access morbidity and costs. Kidney Int

60:1164–1172, 2001 37. Krivitski NM: Theory and validation of access flow measurementby dilution technique during hemodialysis. Kidney Int 48:244–250,21. Altman SD, Arthur TS, Altman NR, et al: Maintenance angio-

plasty of hemodialysis access stenoses in a mobile setting, in Vascu- 199538. Wittes J: Randomized treatment assignment, in Encyclopedia oflar Access for Hemodialysis–V, edited by Henry ML, Ferguson

RM, Chicago, Precept Press, 1998, pp 178–188 Biostatistics (vol 5), edited by Armitage P, Colton T, Chichester,Wiley, 1999, pp 3703–371122. Sands J, Young S, Miranda C: The effect of Doppler flow screen-

ing studies and elective revisions on dialysis access failure. ASAIO 39. Rehman SU, Pupim LB, Shyr Y, et al: Intradialytic serial vascularaccess flow measurements. Am J Kidney Dis 34:471–477, 1999J 38:M524–M527, 1992

23. Sands JJ, Miranda CL: Prolongation of hemodialysis access sur- 40. Paulson WD, Ram SJ, Faiyaz R, et al: Association between bloodpressure, ultrafiltration, and hemodialysis graft thrombosis: Avival with elective revision. Clin Nephrol 44:329–333, 1995

24. Carroll D, Tramer H, McQuay H, et al: Randomization is impor- multivariable logistic regression analysis. Am J Kidney Dis 40:769–776, 2002tant in studies with pain outcomes: Systematic review of transcuta-

neous electrical nerve stimulation in acute postoperative pain. Br 41. Bauters C, Isner JM: The biology of restenosis, in Textbookof Cardiovascular Medicine, edited by Topol EJ, Philadelphia,J Anaesth 77:798–803, 1996

25. Gordis L: Epidemiology, Philadelphia, WB Saunders, 2000, pp Lippincott-Raven Publishers, 1998, pp 2465–248942. Miller PE, Carlton D, Deierhoi MH, et al: Natural history of100–109

26. Mathews DE, Farewell VT: In Using and Understanding Medical arteriovenous grafts in hemodialysis patients. Am J Kidney Dis36:68–74, 2000Statistics (3rd ed), New York, Karger, 1996, pp 199–203