Regular Monitoring of Access Flow Compared with Monitoring of Venous Pressure Fails to Improve Graft Survival

Abstract

ABSTRACT. Regular vascular access blood flow (Qa) surveillance is recommended to detect graft stenosis; however, there is little evidence that monitoring and correcting with angioplasty improves graft survival. This blinded, randomized, controlled trial of 112 patients studied time to graft thrombosis and graft loss, comparing monthly Qa plus standard surveillance (dynamic venous pressure and physical examination) (treatment group) to standard surveillance alone (control group). Only the treatment group was referred for angiogram if Qa <650 ml/min or a 20% decrease in Qa from baseline. Percutaneous angioplasty was performed for stenosis >50%. The rate of graft thrombosis per patient-year at risk was 0.41 and 0.51 in the control and treatment groups, respectively. Fifty-one interventions (0.93/patient-years at risk) were performed in the treatment group versus 31 interventions (0.61/patient-years at risk) in the control group. There was no difference in time to graft loss (P = 0.890). In a multivariate analysis, aspirin (ASA) therapy at baseline was associated with an 84% reduction in risk of graft thrombosis (odds ratio [OR], 0.14; P = 0.002). Higher baseline Qa (OR, 0.84; P = 0.05) and longer interval since graft insertion (OR, 0.97; P = 0.07) were associated with a decrease in graft thrombosis. Results reveal that graft surveillance that uses Qa increases the detection of stenosis, compared with standard surveillance; however, intervention with angioplasty does not improve the time to graft thrombosis or time to graft loss. E-mail: [email protected]

Thrombosis of arteriovenous grafts continues to be a major source of morbidity, mortality, and cost among hemodialysis patients. Graft thrombosis occurs at a rate of 0.5 to 2.5 events per patient-year. Between 63% to 90% of grafts still function at 1 yr, whereas only 42% to 60% remain patent at 3 yr (^1–3). The annual cost of access maintenance has been estimated at $1 billion in the United States, and access-related problems are responsible for 17% to 25% of all hospitalizations in hemodialysis patients (^1,4–6). Graft thrombosis is usually associated with neointimal hyperplasia at the venous anastomosis of the graft (⁷). This lesion causes progressive stenosis, leading to thrombosis and ultimately to loss of the graft. In uncontrolled studies, preemptive identification and correction of hemodynamically significant stenosis before thrombosis has been shown to prolong the life of the graft and reduce the rate of future thrombosis when compared with access repair after a thrombotic event (^8–10).

Studies have revealed a relationship between hemodynamic, functional and clinical indicators, and graft thrombosis. These indicators include increased venous pressure (^8,10–12), intra-access pressures at zero blood pump speed (^8,13), urea recirculation, unexplained decline in the delivered dialysis prescription (¹⁴), and vascular access blood flow (Qa) (^15–17). A low Qa or a decrease in Qa is associated with an increased risk of graft thrombosis (^15,18–21). On the basis of this evidence, the Canadian Society of Nephrology (CSN) Practice Guidelines for Vascular Access and the National Kidney Foundation Kidney/Dialysis Outcome Quality Initiatives Guidelines for Vascular Access recommend monthly Qa measurements as the preferred method for surveillance of grafts (^4,22).

Surveillance that uses Qa requires the use of additional technology, expertise, and time, and there is now some debate about whether Qa monitoring provides additional benefit above that of standard monitoring (²³). Smits et al. (¹⁸) compared static venous pressure (SVP) and dynamic venous pressure (DVP) monitoring to Qa monitoring alone, Qa plus DVP, and Qa plus SVP. He concluded that SVP alone, Qa alone, or a combination of the two reduced thrombosis rate to <0.5 per patient-year.

We conducted a prospective, randomized, double-blind study in prevalent hemodialysis patients with grafts comparing monthly Qa measurement that uses the Transonic Hemodialysis Monitor (Transonic System, Ithaca, NY) plus standard surveillance (DVP and physical examination) versus standard surveillance alone. The primary objective was to compare time to graft thrombosis. The hypothesis tested was that the addition of Qa monitoring to standard surveillance would lengthen the time to graft thrombosis. Secondary objectives were to compare graft thrombosis episodes per patient-year at risk; time to permanent graft loss; and intervention rates per patient-year at risk.

Materials and Methods

Patients were recruited from two outpatient hemodialysis programs, London Health Sciences Center, London, Canada, and St. Joseph’s Health Center, Hamilton, Canada, from January 2000 to December 2001. The study protocol met the approval of the institutional review boards at each center. Patients were included if they were able to provide written informed consent; if the graft had no clinical or functional abnormality; and if the screening Qa was >650 ml/min. Grafts with Qa <650 ml/min were referred for angiogram or angioplasty, and those with subsequent Qa> 650 ml/min were then eligible for randomization. The site study coordinator was responsible for patient screening, consent, and enrollment. An independent central coordinator randomized patients to a graft surveillance program including Qa measurement plus standard surveillance (treatment group) or to a group studied by standard surveillance alone (control group). A computerized randomization sequence was generated in blocks of four. Access flows were performed in a blinded fashion in both groups. Transonic Systems provided a software program so the Qa results were reported only as an alphanumeric code on the computer screen. The Qa data were downloaded from the computer by the site study coordinator on a weekly basis and sent electronically to the central coordinator, who decoded the Qa data. The Qa data for the control group were not decoded or reported until the end of the study.

A vascular access surveillance program was operational at both participating hemodialysis units at the time of initiation of the study. The surveillance program included DVP monitoring and regular examination of the graft by the dialysis nurse. The DVP monitoring protocol was adopted from the CSN Clinical Practice Guidelines for Vascular Access (²²). DVP was measured at each dialysis session within the first 5 min, with the blood pump set at 200 ml/min. Grafts were cannulated with 15-gauge needles at least 2 inches apart. The dialysis nurse monitored DVP measurements, and DVP that exceeded threshold (>125 mmHg for Baxter Tina system 1000 and >140 mmHg for the Gambro Integra machines) on three consecutive sessions was referred for angiogram.

The dialysis nurse examined the graft at each dialysis session. An angiogram was requested in patients with arm swelling on the same side of the access, prolonged bleeding from needle puncture sites after dialysis, or altered characteristics of pulse or thrill in the graft.

Access flow was measured by ultrasound velocity dilution technique with the Transonic HD01 hemodialysis monitor. The theoretical background, in vitro and in vivo validation, and technique have been described in detail previously (^24,25). A dialysis nurse performed all Qa measurements within 90 min of the initiation of the dialysis session at a blood flow of 300 ml/min. Access flow measurements were deferred to another dialysis session if the patient required a decrease in the programmed rate of ultrafiltration and/or the administration of normal saline. Qa measurements were done twice within a 15-min time period. Patients in both groups had monthly, blinded Qa measurements, and only patients in the treatment group had the Qa results decoded during the study by the study coordinator. The study coordinator took the average of the two Qa measurements and recorded this as the Qa. Patients in the treatment group were referred for an angiogram if the Qa was <650 ml/min or if there was a >20% decrease from baseline Qa.

Angiography was requested to determine the presence and location of suspected stenosis at any one of the four predefined sites, venous anastomosis, arterial anastomosis, intergraft, or in the draining vein. Patients with a hemodynamically significant stenosis, defined as a ≥50% reduction of normal vessel diameter (graft or draining vein system), were referred for percutaneous transluminal angioplasty (PTA). PTA was performed by means of standard techniques (²⁶). Anatomic success of the PTA was defined as <30% residual diameter stenosis at the site or sites of treatment as described in the Society of Interventional Radiology consensus document (²⁶). Patients were referred for surgical revision if residual stenosis was >30% after angioplasty or if PTA was deemed inappropriate by the radiologist. Patients returned to their assigned surveillance protocol after an intervention. All interventions and thrombotic events were recorded. Follow-up was censored for the following reasons: transplantation, change to alternate vascular access, withdrawal from the study, transfer to peritoneal dialysis, transfer to a nonstudy center, or death.

The primary outcome was time to graft thrombosis. Secondary outcomes included thrombotic episodes per patient-year at risk, time to graft loss, and interventions per patient-year at risk. Graft loss was defined as permanent loss of graft patency such that the graft was no longer used to administer dialysis.

The following assumptions were made for calculation of group size. Cayco et al. (⁶) reported a graft thrombosis rate of 0.30 events per graft-year using DVP. Allowing for an expected 1-yr event-free rate of 70%, 63 subjects per group were required to detect a hazard ratio of 0.4 with 80% power and a two-sided α of 0.05.

Time to thrombosis and time to graft loss were estimated by the Kaplan-Meier technique. The two groups were compared by the log-rank test. Cox regression was used to adjust for the effect of center. The primary analysis was an intent-to-treat analysis, in which patients were analyzed in the group to which they were assigned. A secondary per-protocol analysis of the time to thrombosis was performed; patients who had a delay of more than 2 wk between the identification of an abnormal surveillance parameter and the angiogram were excluded. All tests of significance were two sided, and differences were considered statistically significant at P < 0.05. A secondary analysis was performed in which the association between 15 baseline characteristics and the occurrence of a graft thrombosis within 1 yr was explored by logistic regression techniques. A multivariable model was developed in which entry of baseline characteristics was allowed entry at the 0.20 level of significance. SAS version 8.2 was used for all analyses (SAS, Cary, NC).

Results

A screening Qa was performed in 124 patients; 118 had Qa >650 ml/min. Angiogram and PTA was performed in the remaining six patients, and four of these had an increase in Qa >650 ml/min and were eligible for inclusion in the study. Of the 120 patients eligible for inclusion, seven patients refused to provide consent, and one patient had high DVP with no stenosis documented from a previous angiogram. A total of 112 patients were enrolled into the study, 53 in the control group and 59 in the treatment group. The flow of patients throughout the study is depicted in Figure 1.

Figure 1.: Flow diagram of randomized controlled trial of standard surveillance versus standard surveillance plus regular access flow (Qa) looking at time to graph thrombosis.

The between-group comparison of baseline demographics and clinical characteristics is presented in Table 1. In the control group, 75.5% of the grafts were in the lower arm (with the remainder in the upper arm), and 69.8% were curved. In the treatment group, 67.8% of the grafts were in the lower arm, 23.7% in the upper arm, and the rest in the leg.

Table 1: Baseline demographic and clinical characteristics reported as mean (SD) or percent

Eighteen episodes of thrombosis occurred in the control group and 26 in the treatment group. The rate of graft thrombosis per patient-year at risk was 0.41 and 0.51 in the control and treatment groups, respectively. Survival curves for time to first thrombosis are presented in Figure 2. There was no difference between groups in terms of the time to thrombosis (P = 0.572 unadjusted, P = 0.569 adjusted for center). Graft loss occurred in eight patients (18.6%) in the control group and nine patients (19.2%) in the treatment group. Survival curves for time to graft loss are presented in FIgure 3.. There was no difference in the time to graft loss between the two groups (P = 0.890 unadjusted, P = 0.735 adjusted for center). Figure 3 illustrates graft survival.

Figure 2.: Time to graft thrombosis comparing standard surveillance alone (control) to standard surveillance plus regular access flow (Qa) (treatment).

Figure 3.: Time to graft loss comparing standard monitoring (control) to standard monitoring plus regular access flow (Qa) (treatment).

A total of 76 angiograms were performed in the treatment group: 53 for abnormal Qa, 14 for DVP greater than threshold, and 9 for abnormal clinical findings. In the control group, 41 angiograms were performed, 31 for DVP greater than threshold and 10 for abnormal clinical findings. A positive angiogram was defined as graft stenosis ≥50% at any one of the four predefined sites, venous anastomosis, arterial anastomosis, intergraft, and draining vein. Stenosis at the venous anastomosis or in the draining vein accounted for 88% of the reported stenosis. In the treatment group, the positive predictive value of Qa to detect stenosis ≥50% was 87%, and the positive predictive value for DVP was 71%. In the control group, the positive predictive value for DVP was 94%.

The number of PTAs required to maintain patency was greater in the treatment group compared with the control group. Fifty-one PTAs (0.93 interventions per patient-year at risk) were performed in the treatment group versus 31 PTAs (0.61 interventions per patient-year at risk) in the control group. Angioplasty was successful in 85% of the grafts as defined by angiographic criteria of <30% residual stenosis.

There was a concern that there may have been a delay from the time of identification of an abnormal surveillance parameter and the angiogram or PTA. The data were therefore reanalyzed on a per-protocol basis and included events with <14 d from the time of identification of an abnormal surveillance to the time of angiogram. The results were consistent with the intent-to-treat analysis, showing no difference in time to thrombosis (P = 0.88).

The univariate odds ratio (OR) for predictors of graft thrombosis is listed in Table 2. Higher serum albumin (OR, 0.90; P = 0.066), higher Qa (per 100 ml) at baseline (OR, 0.88; P = 0.10), longer interval since graft insertion (30 d) (OR, 0.98; P = 0.98), and aspirin (ASA) therapy at baseline (OR, 0.16; P = 0.003) were associated with decreased graft thrombosis. In a multivariable analysis, greater baseline Qa (OR, 0.82; P = 0.03), longer interval since graft insertion (OR, 0.97; P = 0.08), and use of ASA (OR, 0.14; P = 0.002) were each independently associated with a decrease in graft thrombosis. Age, diabetes, hemoglobin, and erythropoietin and warfarin use were not associated with graft thrombosis.

Table 2: Predictors of graft thrombosis at 1 yr

Discussion

In this blinded, randomized, controlled study, there were four clinically significant findings. First, both Qa and DVP were successful in detecting venous stenosis; however, Qa monitoring detected stenosis more frequently, as indicated by the increased use of angiography in the treatment group. Second, there was no difference in the time to graft thrombosis comparing the use of monthly Qa plus standard surveillance that used thrice-weekly DVP and physical examination to standard surveillance alone. The rate of thrombosis per patient-year at risk was 0.51 and 0.41, respectively, either close to or within the accepted rate of 0.5 thrombotic episodes per patient-year at risk (⁴). Third, there was no difference in time to permanent graft loss between the two groups. This is an important finding because the intent of vigilant surveillance is to prolong the assisted patency of the graft. Fourth, the number of procedures required to maintain graft patency was greater in the treatment group with 51 PTA interventions (0.93 interventions per patient-year at risk) versus 31 (0.61 interventions per patient-year at risk) in the control group.

The observation that there is no significant difference in thrombosis rate or graft patency between groups in spite of an improvement in detection of graft stenosis calls into question not the monitoring technique but the success of the PTA intervention.

Several authors have suggested Qa is the surveillance method of choice to detect stenosis before thrombosis (^14–21,27). Grafts with a Qa <600 ml/min have a higher rate of thrombosis than grafts with Qa >600 ml/min. Several investigators have shown that a trend of decreasing Qa is more predictive of venous stenosis than a single measurement of Qa (^10,19–21). Most of these studies were retrospective or nonrandomized, used historical controls, or analyzed fistulae and grafts together (^14–17,20).

These studies highlight two clinically important questions. First, does graft surveillance with Qa improve the time to thrombosis and graft survival? And second, does correction of stenosis with angioplasty prolong overall graft survival. Addressing the first question, Smits et al. (¹⁸) conducted a randomized trial looking at graft thrombosis rates and compared SVP and DVP monitoring to Qa monitoring alone, Qa plus DVP, and Qa plus SVP. Venous pressures were measured weekly, and Qa was measured every 8 wk. An event was defined as a thrombotic episode without a preceding positive test; however, this definition resulted in the exclusion of 21 or 42 thrombotic events and may have underestimated the true thrombosis rate. The conclusion was that SVP alone, Qa alone, or a combination of Qa and SVP reduced thrombosis rate below 0.5 per patient-year.

By means of a randomized controlled study design, we found that the addition of Qa to DVP and physical examination was not associated with a longer time to graft thrombosis or improved graft survival. DVP greater than threshold has been shown to identify grafts with outflow tract stenosis but miss those with poor arterial inflow or intragraft stenosis (³²). We found that 88% of the significant stenoses were at the venous anastomosis or in the draining vein, possibly accounting for the high detection rate with DVP. Pump flow, needle size and position, and BP (⁸) may alter the DVP, and this variation was minimized by performing DVP at a standard pump flow of 200 ml/min with 15-gauge needles and performing the measurement within the first 5 min of initiating the dialysis treatment.

The second question is whether correction of stenosis by angioplasty improves graft survival. The availability of a screening test with a high sensitivity and specificity may be helpful in providing an early warning of access failure, but there needs to be an effective method to correct the problem and increase the life of the graft. The addition of Qa to standard surveillance increases the detection of stenosis, as indicated by the increased number of angioplasties performed in the treatment group. Fifty-one PTAs (0.93 interventions per patient-year) were performed in the treatment group versus 31 (0.61 interventions per patient-year) in the control group. In spite of the increase in number of interventions, there was not a difference in time to thrombosis or overall graft survival. This suggests that although Qa monitoring accurately detects stenosis, intervention with angioplasty does not improve overall graft survival. In addition, the intervention is accompanied by significant cost.

Angioplasty is the most common intervention once a significant stenosis is identified. Angioplasty has a high degree of initial success in reducing the stenosis, but it does not always improve the functional abnormality, such as Qa or venous pressure; nor does it necessarily improve the long-term patency (^21,33–35).

Van der Linden et al. (³³) assessed 65 grafts with Qa measurement and angiograms before and after angioplasty. Access flows increased to levels >600 ml/min in only 60% of grafts, and the unassisted patency rate in these grafts was only 25% 6 mo after angioplasty. Multiple procedures were required in 43% of grafts, and the mean time interval to repeat PTA was shorter with each subsequent intervention. This suggests that PTA itself may damage the graft endothelium and promote further stenosis.

In an observational intervention study, Schwab et al. (²¹) followed 27 patients dialyzed through a graft. A decrease in Qa resulted in a PTA in 28 grafts. Access flow was measured within 1 wk after the PTA. Failure to increase Qa by 20% after PTA defined failure of the PTA and this occurred in 21% of grafts (6 of 28 PTA). Six of eight thromboses occurred in patients who had an unsuccessful PTA. This suggests it is the success of the intervention that determines improved patency, not the success of the monitoring parameter.

Ahya et al. (³⁴) studied the angiographic assessment of venous stenosis to change in access flow after angioplasty in 22 hemodialysis patients with a decreased Qa. There was no significant correlation between change in Qa and the change in percentage of stenosis. Similarly, there was no correlation between postangioplasty flow and reduction of stenosis (r² = 0.015; P = NS). This study concluded that PTA can improve Qa, but the authors did not study whether this effect improved graft patency.

In our study, PTA was successful in 85% of the grafts as defined by angiographic criteria of <30% residual stenosis. Access flows were not performed immediately after PTA but were followed monthly according to the study protocol. Postangioplasty Qa was available in 81% of grafts in the control group and 91% of grafts in the intervention group. Postangioplasty Qa was not performed in 11 instances for reasons including graft thromboses, termination of the study, hospitalization of the patient, and death. Failure to increase Qa by 20% after PTA occurred in 36% of grafts in the control group and in 41% of the grafts in the treatment group. The percentage of patients who failed to increase Qa by 20% was greater in our study compared with the study of Schwab et al. (²¹) (40% versus 21%), but Qa was performed up to 4 wk after the PTA in our study versus within 7 d in the study of Schwab et al., and this likely contributed to some of the difference. Failure to improve flows to >650 ml/min occurred in 20% of the control group and 13% of the treatment group.

Access flows have a high positive predictive value to detect graft stenosis; that is, more than 90% of patients with a decline in Qa have significant stenosis. This study has shown that it takes more that a good screening tool to improve graft patency rates. The intervention to treat the stenosis must be proven to improve graft patency. To date, we have not clearly defined what the definition is for ‘successful“ angioplasty. The available National Kidney Foundation Kidney/Dialysis Outcome Quality Initiatives and CSN guidelines for vascular access suggest that Qa should return to baseline, but there is no clear recommendations to when the Qa should be repeated. There has been no randomized, controlled trial demonstrating that grafts with “successful” PTA have an improved long-term patency rate.

Other factors have been associated with thrombosis but are not related to progressive stenosis or changes in Qa. These include hypercoagulability, race, hypoalbuminemia, hypotension, and local compression (^36–38). In our study, 11 (25%) of 44 thrombotic events occurred without any abnormality in surveillance parameters. This is similar to the rate of unpredicted thrombosis of 20% described by Schwab et al. (²¹).

Univariate and multivariate analyses were performed in this study to identify predictors of graft thrombosis. Higher baseline serum albumin, higher Qa, and longer interval since graft insertion were associated with decreased graft thrombosis. Others have also reported the association between a low albumin and an increase in graft thrombosis (^36,38). Miller et al. (³⁶) postulates that hypoalbuminemia is a marker for an inflammatory process that promotes myointimal hyperplasia and increasing the likelihood of stenosis. Alternatively, hypoalbuminemia may just be the marker of poor health, malnutrition, or another cofactor that increases the risk of graft failure.

At baseline, ASA therapy was administered to 30.2% of the patients in the control arm and in 42.6% of patients in the treatment arm. In both the univariate and multivariable analyses, ASA use was associated with a decrease in the risk of graft thrombosis. Sreedhara et al. (³⁹) conducted a randomized trial comparing ASA to dipyridamole alone, ASA plus dipyridamole, and placebo and compared time to thrombosis. ASA was associated with an increased the risk of thrombosis (relative risk, 1.99; P = 0.18) in incident grafts versus the protective effect associated with use of dipyridamole (relative risk, 0.35; P = 0.02). The Dialysis Outcomes and Practice Patterns Study recently reported on the association between vascular access patency and the use of specific drugs (⁴⁰). ASA therapy was associated with better secondary patency (relative risk, 0.70; P < 0.001). Forty percent of the patients were receiving warfarin therapy at baseline, and there was no association between warfarin and thrombosis or graft survival rate. The majority of patients were receiving erythropoietin therapy, and again, there was no association with graft thrombosis.

There are several limitations to this study. The control group reflects the standard of access surveillance in many North American dialysis units. The dialysis nurse and the nephrologist monitor DVP and physical changes in the graft, and angiograms were requested when abnormalities in these surveillance parameters were detected. In the treatment group, requests for angiograms were more tightly defined because the study coordinator followed predefined criteria regarding Qa. If anything, this design may be biased in favor of the treatment group because the Qa criteria were rigorously applied, whereas the standard monitoring was part of the “accepted clinical practice.” This study was performed in prevalent grafts, and it is possible that we selected a series of grafts with different thrombotic and survival rates that would not be present in an incident population of grafts.

In addition, access flows were scheduled every month; however, there were instances of delay in obtaining the Qa measurement. Reasons for a delay of Qa measurement included equipment malfunction, hospitalization of the patient, or lack of experienced personnel to do the Qa measurement. This is likely to be an inherent part of this surveillance technique because it requires specialized technology and expertise. Our protocol included a request for an angiogram when an abnormal surveillance parameter was met. Angiograms were requested within a week of the abnormal parameter; however, there were instances of delay in obtaining the angiogram as a result of constraints within the radiology department. We performed a per-protocol analysis and included only events that had <14 d from the time of identification of an abnormal surveillance to the time of angiogram. The results were consistent with the intent-to-treat analysis and showed no difference in time to thrombosis (P = 0.88). This suggests that delay in obtaining the intervention was not the contributing reason for the lack of a difference between the two groups. Access flows were not performed immediately after PTA but were available as part of routine surveillance in 86% of grafts. It is possible that the PTA intervention was not performed successfully in the units studied.

In conclusion, this blinded, randomized, controlled trial demonstrates that there was no difference in the time to graft thrombosis or time to permanent graft loss when comparing the use of monthly Qa plus standard surveillance by using thrice-weekly DVP and physical examination to standard surveillance alone. Access surveillance by means of Qa monitoring detected more stenosis resulting in an increased number of procedures, but this did not result in a difference in the time to graft thrombosis or time to permanent graft loss. The ability to detect graft stenosis does not improve graft patency without a successful intervention. Future studies are needed to define a successful PTA and demonstrate that intervention with PTA improves graft patency.

We acknowledge the financial assistance of the physicians of Ontario through the PSI Foundation.