Artery to vein configuration of arteriovenous fistula improves hemodynamics to increase maturation and patency

Abstract

Arteriovenous fistulae (AVF) are the preferred mode of hemodialysis access, yet 60% of conventional (vein-to-artery, V-A) AVF fail to mature and only 50% remain patent at one year. We previously showed improved maturation and patency in a pilot study of the radial artery deviation and reimplantation (RADAR) technique that uses an artery-to-vein (A-V) configuration. Here, we show that RADAR exhibits higher rates of maturation, as well as increased primary and secondary long-term patencies. RADAR is also protective in female patients, where it is associated with decreased reintervention rates and improved secondary patency. RADAR and conventional geometries were compared further in a rat bilateral carotid artery-internal jugular vein fistula model. There was decreased cell proliferation and neointimal hyperplasia in the A-V configuration in male and female animals, but no difference in hypoxia between the A-V and V-A configurations. Similar trends were seen in uremic male rats. The A-V configuration also associated with increased peak systolic velocity and expression of Kruppel Like Factor 2 and phosphorylated endothelial nitric oxide synthase, consistent with improved hemodynamics. Computed tomography and ultrasound-informed computational modeling showed different hemodynamics in the A-V and V-A configurations and improving the hemodynamics in the V-A configuration was protective against neointimal hyperplasia. These findings collectively demonstrate that RADAR is a durable surgical option for patients requiring radial-cephalic AVF for hemodialysis access.

One Sentence Summary:

Radial artery deviation and reimplantation exhibits improved patency rates in both sexes, and is associated with improved hemodynamics in a rat model.

Introduction

The epidemiology of chronic kidney disease (CKD) parallels the current worldwide epidemics of diabetes and hypertension (1); the United States Renal Data System estimates that 30 million Americans suffer CKD (2). Most patients with end-stage renal disease rely on hemodialysis to replace their lost kidney function (3). Clinical practice guidelines designate arteriovenous fistulae (AVF) as the preferred vascular access for hemodialysis since AVF have fewer complications, improved access patency, and lower risk of mortality compared to AV grafts or central venous catheters (4–6). The 1997 National Kidney Foundation-Kidney Disease Outcome Quality Initiative, followed by the Fistula First Breakthrough Initiative of 2003, have dramatically increased the rates of AVF use (7). Despite their advantages, 60% of AVF fail to mature and only 50% are primarily patent at one year, thus imposing a large burden of healthcare and costs on patients and the health care system (4, 5, 8). Further, fistula failures are higher in female patients when compared to males (9), and the majority of AVF are created in uremic patients, a hyper-proliferative state associated with increased smooth muscle proliferation, neointimal hyperplasia (NIH), and early fistula failure (10).

The high rate of AVF failure results from failures occurring within two distinct time frames after the procedure. Early failure occurs when the newly created fistula fails to mature; the vein must thicken and dilate within the new fistula environment to support the higher flow rates required for hemodialysis (11, 12). Late failure occurs after successful venous remodeling, which allows initiation of hemodialysis, but then fails secondary to juxta-anastomotic neointimal hyperplasia (13, 14). Although it remains unknown which variables are most important for predicting fistula failure, both local and systemic factors including inflammation, hypoxia, and chronic disease associate with formation of NIH (10, 15–18). In addition, fistula creation alters flow hemodynamics, exposing the outflow vein to high pressures and disturbed flow, both of which promote NIH (18, 19).

Multiple strategies to prevent NIH have been examined but have shown limited potential. Pharmacological treatments aimed at decreasing inflammation, cell recruitment, and platelet aggregation, as well as gene therapy to decrease smooth muscle cell proliferation have been evaluated, although none have gained clinical success (15, 20–24). Modifications of the initial surgical technique, including changes in fistula angles that decrease disturbed flow to promote maturation and decrease NIH, are by far the most promising (15, 20–22, 25). Clinical trials evaluating external support devices such as VasQ (Laminate Medical Technologies), designed to optimize fistula geometry and flow parameters based on computational fluid dynamics, are also ongoing (20).

Surgical arteriovenous fistulae between the radial artery and the adjacent cephalic vein in a side-to-side configuration were first described in 1966 (26). To bridge larger distances between the artery and the vein, the technique was modified to divide the cephalic vein and swing it onto the side of the radial artery (Fig. 1A); despite poor patency rates, this vein-to-artery (V-A) configuration has remained the predominant AVF geometry (27). To improve the hemodynamics and patency of human AVF, we previously described the radial artery deviation and reimplantation (RADAR) technique to create artery-to-vein (A-V) fistulae (28). RADAR transposes the radial artery onto the cephalic vein in the distal forearm to create a radial-cephalic AVF without venous dissection, avoiding the extensive venous mobilization required for conventional AVF surgical procedures (Fig. 1A) (28). In addition, RADAR minimizes tissue handling of both the artery and vein, using a tourniquet to avoid use of clamps, to reduce potential wall hypoxia near disrupted vasa vasorum. In the first report of RADAR in human patients, there was reduced NIH and juxta-anastomotic stenosis as well as increased fistula maturation and patency compared to the traditional vein-to-artery AVF (28).

Fig. 1. Improved fistula maturation and long-term patency with RADAR in human patients.

(A) Schematic showing traditional and RADAR fistula configurations. (B) Bar graph quantifying the percentage of mature fistulae in the control and RADAR groups at 6 weeks, 3 months, and last follow up; *, P = 0.002, P = 0.001, P < 0.0001 respectively. (C) Kaplan-Meier analysis of cumulative rate of intervention on juxta-anastomotic stenosis; P = 0.000001. (D) Kaplan-Meier analysis of primary patency; P = 0.000065. (E) Kaplan-Meier analysis of secondary patency; P < 0.0000001. n = 201 patients in the RADAR group and n = 73 in the control group. P values in panel A were calculated by the Chi square test and panels C-E by the log-rank test.

Importantly, however, it remains unknown whether the improved short-term results with RADAR in humans are secondary to diminished hypoxic injury or to improved hemodynamics. We hypothesized that long-term follow up of patients with RADAR procedures would provide a mechanistic clue; sustained improvement in patency suggests long-term improvement in hemodynamics of the fistula environment. To evaluate the role of hemodynamics as a mechanism of the increased early maturation and patency associated with RADAR, we developed a rat model of bilateral carotid artery-jugular vein AVF using two different configurations to model RADAR and conventional AVF, with both configurations in the same animal. We hypothesized that the artery-to-vein (A-V) configuration has more favorable hemodynamics and thus will show decreased NIH.

Results

Improved fistula maturation and long-term patency with RADAR

We previously reported a pilot study in 53 patients describing the RADAR (artery-to-vein) technique that showed short-term improvements in maturation and patency rates (28). We hypothesized that these clinical improvements were secondary to improved hemodynamics and thus would be durable during long-term follow-up. Herein, we compare 201 consecutive patients who had RADAR performed between October 2014 and April 2017 with 73 consecutive patients who received control (vein-to-artery) AVF between January 2013 and September 2014 (fig. S1). The two groups had similar baseline characteristics, with the exception of significantly smaller diameter arteries (P = 0.004) and veins (P < 0.0001) in the RADAR group (Table 1). The mean blood creatinine concentration at the time of access creation was 477.9 ± 179.2 µmol/L. In a pooled cohort that included both female and male patients, AVF performed with RADAR had significantly higher rates of maturation at 6 weeks (P = 0.002) and 3 months (P = 0.001) compared to control AVF, with more RADAR remaining in use for hemodialysis at the end of follow-up (P < 0.0001) (Fig. 1B). The cumulative reintervention rate in the juxta-anastomotic segment was significantly lower in patients with RADAR compared to controls (cumulative reintervention rate: 12.6% vs. 42.6% at 12 months and 17.3% vs. 49.1% at 36 months, P < 0.000001) (Fig. 1C). Both primary and secondary patencies were significantly increased at 12 and 36 months in patients treated with RADAR compared to controls (primary patency: 72.2% vs. 48.1% at 12 months and 62.1% vs. 37.6% at 36 months, P = 0.000065; secondary patency: 98.4% vs. 72.1% at 12 months and 94.9% vs. 66.8% at 36 months; P < 0.0000001; Fig. 1, D and E). Patients with RADAR required significantly fewer reinterventions (30.1 per 100 person-years vs. 39.9 per 100 person-years; P = 0.03); the incidence of reintervention for venous juxta-anastomotic stenosis was 1.6 per 100 person-years with RADAR compared to 17.1 per 100 person-years (P < 0.0001), whereas the RADAR group required more arterial reinterventions (P = 0.02; Table 2). There was no evidence of hand ischemia in any patient in the RADAR group. Since patients with RADAR had smaller diameter vessels compared with control patients, we performed a binomial logistic regression to determine the effect of AVF type and vessel diameters on the likelihood of developing juxta-anastomotic stenosis. Both traditional fistula geometry (P < 0.0001) and decreased venous diameter (P = 0.025) associated with higher odds of developing juxta-anastomotic stenosis (table S1). These long–term clinical results confirm the results of our pilot study and demonstrate that RADAR is associated with less juxta-anastomotic stenosis and improved rates of maturation and long-term patency in human patients.

Table 1. Baseline human patient characteristics.

Similar baseline characteristics in patients undergoing radial artery deviation and reimplantation (RADAR) compared to controls. N=number; SEM=standard error of the mean.

Variable	Control (N=73)	RADAR (N=201)	P value
	N or mean (% or SEM)	N or mean (% or SEM)
Age (years)	70 ± 1.52	67 ± 1.06	0.15
Sex
Male	54 (74)	132 (66)	0.19
Female	19 (26)	69 (34)
Comorbidities (at inclusion)
Diabetes	25 (34)	61 (30)	0.54
Hypertension	60 (82)	150 (75)	0.19
Dyslipidemia	16 (22)	42 (21)	0.86
Tobacco use	12 (16)	32 (16)	0.92
Treatment (at inclusion)
Anticoagulant	10 (14)	31 (15)	0.72
Antiplatelet	36 (49)	78 (39)	0.12
Statin	33 (45)	64 (32)	0.05
Vessel diameter (mm)
Artery	2.7 ± 0.08	2.5 ± 0.02	0.004
Vein	3.6 ± 0.13	3.1 ± 0.04	<0.0001

Table 2. Reintervention rate and stenosis in human patients.

Reduced reintervention rate to maintain or reestablish fistula patency in RADAR group compared to controls. Two patients had both venous and arterial juxta-anastomotic stenosis treated at the same time; N=number; OR=odds ratio; CI=confidence interval.

	Control n=158 person years		RADAR n=365 person years		Absolute risk reduction (95% CI)	OR (95% CI)	P value
	N	Rate per 100 person years	N	Rate per 100 person years
Fistula re-interventions (any segment, all occurrences)
	63	39.9	110	30.1	0.10 (0.01-0.19)	0.65 (0.44-0.96)	0.03
Juxta-anastomotic stenosis re-interventions (first occurrence)
All	30	19	29	7.9	0.11 (0.05-0.18)	0.37 (0.21-0.64)	0.0004
Venous	27	17.1	6	1.6	0.15 (0.10-0.22)	0.08 (0.03-0.20)	<0.0001
Arterial	1	0.6	25	9.9	−0.06 (−0.09- −0.02)	11.5 (1.54-85.8)	0.02

It has previously been reported that female sex is a predictor of primary fistula failure (9). To determine whether female and male patients benefit from RADAR to a similar degree, we performed subgroup analyses on both sexes and found similarly compelling results (Fig. 2). At 42 months, female patients undergoing RADAR had a significantly lower rate of cumulative reintervention and improved secondary patency (cumulative reintervention rate: 24.9% vs. 50.4% in RADAR vs. control; P = 0.03; secondary patency 89.6% vs. 50.4% in RADAR vs. control; P = 0.0003; Fig. 2, A and C), with no difference in primary patency rates (primary patency: 51.2% vs. 39% in RADAR vs. control; P = 0.12; Fig. 2B). At 42 months, male patients undergoing RADAR had a significantly lower rate of cumulative reintervention and improved primary as well as secondary patency rates (cumulative reintervention rate: 13.6% vs. 48.4% in RADAR vs. control; P < 0.0001; primary patency 67.7% vs. 42% in RADAR vs. control; P = 0.001; 97.6% vs. 66.9% in RADAR vs. control; P < 0.0001; Fig. 2, D to F).

Fig. 2. Female and male human patients benefit equally from RADAR.

(A) Kaplan-Meier analysis of cumulative reintervention rate on juxta-anastomotic stenosis in female patients; P = 0.03. (B) Kaplan-Meier analysis of primary patency in female patients; P = 0.12. (C) Kaplan-Meier analysis of secondary patency in female patients; P = 0.0003. (D) Kaplan-Meier analysis of cumulative reintervention rate on juxta-anastomotic stenosis in male patients; P < 0.0001. (E) Kaplan-Meier analysis of primary patency in male patients; P = 0.001. (F) Kaplan-Meier analysis of secondary patency in male patients; P < 0.0001. n = 69 female patients in the RADAR group and n = 19 in the control group in panels A-C; n = 132 male patients in the RADAR group and n = 54 in the control group in panels D-F. P values in panels A-F were calculated by the log-rank test.

Animal model of RADAR

To quantify hemodynamic and molecular changes associated with RADAR, we developed a new animal model: in a single rat, the right internal jugular vein was transposed onto the right carotid artery as a vein-to-artery (V-A) configuration mimicking the geometry of conventional AVF, whereas the left carotid artery was divided and transposed onto the left jugular vein as an artery-to-vein (A-V) configuration mimicking the geometry of RADAR (Fig. 3A). The operative ischemic time, from time of swing segment vessel ligation to restoration of flow, was similar in both configurations (fig. S2A).

Fig. 3. Decreased neointimal hyperplasia in an A-V fistula in male non-CKD rats.

(A) Schematic (top panels) and operative photographs (bottom panels) of the V-A (conventional geometry) and A-V (RADAR geometry) fistula configurations in a rat model. CCA, common carotid artery; CJV, common jugular vein; EJV, external jugular vein; IJV, internal jugular vein; blue arrow represents the direction of venous flow; red arrow represents the direction of arterial flow; * denotes the manipulated vessel; ruler markings 1 mm. (B) Left panels: representative high-power H&E staining showing outflow (internal jugular vein 1 mm distal from the anastomosis) neointimal thickness at days 7, 21, and 42; scale bar 100 µm. Right panels: merged immunofluorescence of vWF (green), α-actin (red), and DAPI (blue) showing neointimal thickness in outflow vessel; scale bar 100 µm. (C) Bar graph quantifying neointimal area at days 7, 21, and 42; *, P = 0.0139, P = 0.1654, and P = 0.0168 at days 7, 21 and 42 respectively. (D) Bar graph quantifying neointima/lumen area ratio at days 7, 21, and 42; *, P = 0.0138, P = 0.0464, and P = 0.0434 at days 7, 21, and 42 respectively. (E) Bar graph quantifying proliferative index in outflow limb at days 7, 21, and 42; *, P = 0.0038, P = 0.0155, and P = 0.0174 at days 7, 21, and 42 respectively. (F) Bar graph quantifying apoptotic index in outflow limb at days 7, 21, and 42; P = 0.6662, P = 0.5312, and P = 0.7526 at days 7, 21, and 42 respectively. n = 3-7 animals per group in panels C-F. Bar graphs in panels C-F represent mean ± SEM; P values were calculated using a two-tailed unpaired Student’s t test.

First, we examined configuration-related changes in male animals only. Bilateral fistulae were excised on days 7, 21, or 42, and neointimal hyperplasia was visualized using hematoxylin and eosin (H&E) staining or immunofluorescence to detect α-actin and von Willebrand factor (Fig. 3B). Neointimal area (Fig. 3C) was decreased in the A-V configuration compared to the V-A configuration at both 7 and 42 days after fistula creation, and the neointima/lumen ratio (Fig. 3D) was decreased in the A-V configuration at all time points. Consistent with decreased NIH, the A-V configuration was also associated with fewer proliferative cells (Fig. 3E, fig. S2B) with similar numbers of apoptotic cells (Fig. 3F, fig. S2B) at each time point, compared to the V-A configuration. Anastomosis diameter did not differ between groups at any time point (fig. S2, C and D), whereas anastomotic neointimal hyperplasia was significantly reduced in the A-V group at all time points (day 7, P = 0.0024; day 21, P = 0.0302; day 42 P = 0.0412) (fig. S2E). These results suggest that the A-V configuration of AVF is associated with decreased anastomotic and perianastomotic neointimal hyperplasia in a rat model of RADAR, consistent with the improved patency rates in human patients treated with RADAR.

To determine whether the artery to vein configuration is equally protective in female animals, we then conducted similar studies in female rats. Neointimal hyperplasia was similarly visualized using H&E staining or immunofluorescence to detect α-actin and von Willebrand factor (Fig. 4A), although only at 21 days after fistula creation. At this time, both neointimal area (Fig. 4B) and the neointima/lumen ratio (Fig. 4C) were decreased in the A-V configuration compared to the V-A configuration. Consistent with decreased NIH, the A-V configuration was also associated with a decreased number of proliferative cells (Fig. 4D) and similar numbers of apoptotic cells (Fig. 4E), compared to the V-A configuration. These data are consistent with the improved results in human female patients treated with RADAR.

Fig. 4. Decreased neointimal hyperplasia in an A-V fistula in female non-CKD rats.

Fistulae were created in 3 female animals; 3 survivors (100%) had patent fistulae that were analyzed. (A) Left panels: representative high-power H&E staining showing outflow (internal jugular vein 1 mm distal from the anastomosis) neointimal thickness at day 21; scale bar 100 µm; dashed yellow line delineates intima/media transition. Right panels: merged immunofluorescence of vWF (green), α-actin (red), and DAPI (blue) showing neointimal thickness in outflow vessel; scale bar 100 µm. (B) Bar graph quantifying neointimal area at day 21; *, P = 0.0039. (C) Bar graph quantifying neointima/lumen area ratio at day 21; *, P = 0.0008. (D) Bar graph quantifying proliferative index at day 21; *, P < 0.0001. (E) Bar graph quantifying apoptotic index at day 21; P = 0.548. n = 3 animals per group in panels B-E. Bar graphs in panels B-E represent mean ± SEM; P values were calculated using a two-tailed unpaired Student’s t test.

Because CKD results in metabolic derangements associated with increased neointimal hyperplasia and vessel calcification, we tested our fistula model in uremic animals (29, 30). We induced renal failure in male rats via surgical 5/6 nephrectomy, as well as with an adenine-enriched diet. The mean serum urea concentration on the day of fistula creation was 13.53 ± 1.88 mg/dL in the nephrectomy diet and 11.38 ± 1.92 mg/dL in the adenine-enhanced diet. Neointimal hyperplasia was visualized in the 5/6 nephrectomy group using H&E staining or immunofluorescence to detect α-actin and von Willebrand factor (Fig. 5A), 21 days following fistula creation. At this time, there was no significant decrease in neointimal area (Fig. 5B) or neointima/lumen ratio (Fig. 5C) in the A-V configuration compared to the V-A configuration. The A-V configuration was associated with fewer proliferative cells (Fig. 5D) and similar numbers of apoptotic cells (Fig. 5E) compared to the V-A configuration. Similar non-significant decrease in neointimal area and neointima/lumen ratio was also present in the fistulae of adenine-enriched diet animals (fig. S2, F to G). These data are consistent with the improved results in human patients with CKD treated with RADAR.

Fig. 5. Animal model of CKD using 5/6 nephrectomy in male rats.

Chronic kidney disease was created with 5/6 nephrectomy in 35 male rats; 5 survivors (14.3%) had 3 patent AVF that were analyzed. (A) Left panels: representative high-power H&E staining showing outflow (internal jugular vein 1 mm distal from the anastomosis) neointimal thickness at day 21 in nephrectomy animals; scale bar 100 µm; dashed yellow line delineates intima/media transition. Right panels: merged immunofluorescence of vWF (green), α-actin (red), and DAPI (blue) showing neointimal thickness in outflow vessel; scale bar 100 µm. (B) Bar graph quantifying neointimal area at day 21 in nephrectomy animals; P = 0.0999. (C) Bar graph quantifying neointima/lumen area ratio at day 21 in nephrectomy animals; P = 0.0708. (D) Bar graph quantifying proliferative index at day 21 in nephrectomy animals; *, P < 0.0001. (E) Bar graph quantifying apoptotic index at day 21 in nephrectomy animals; P = 0.8291. n = 3 animals per group in panels B-E. Bar graphs in panels B-E represent mean ± SEM; P values were calculated using a two-tailed unpaired Student’s t test.

Similar hypoxic stress in A-V and V-A configurations

We have previously showed that surgical creation of a fistula is associated with increased hypoxic stress including expression of reactive oxygen species (ROS) and hypoxia-inducible factor-1α (HIF-1α) (31); accordingly, we examined whether the decrease in NIH in the A-V configuration is associated with reduced hypoxic stress. Although both ROS (Fig. 6A) and HIF-1α (Fig. 6, B and C) were upregulated after fistula creation, there was no significant difference between the A-V and V-A configurations at either 3 or 7 days postoperatively. Since there was no difference in overall hypoxic stress between configurations, we determined whether hypoxia differed amongst different cell types. The percentage of pimonidazole-positive cells was similar between the A-V and V-A configurations and did not increase between days 3 to 7 after fistula creation (Fig. 6D); although the number of pimonidazole-positive endothelial cells, smooth muscle cells, and macrophages increased after fistula creation, there was no difference in the number of hypoxic cells between fistula configurations (Fig. 6, E to H, fig. S3A). We also evaluated heme oxygenase-1 (HO-1), which is a downstream target of HIF-1α signaling. Although there was a sustained increase in the number of pimonidazole/HO-1 dual-positive cells 3 and 7 days after fistula creation, there was no difference between the A-V and V-A configurations (Fig. 6I). Similarly, anastomotic HIF-1α and pimonidazole-positive cells were both upregulated on days 3 and 7, with no significant difference between the A-V and V-A configurations (fig. S3, B to D). These results suggest that the hypoxic stress associated with surgical fistula creation is not significantly different between the A-V and V-A configurations of this animal model.

Fig. 6. Similar hypoxic stress in V-A and A-V configurations in non-CKD male rats.

(A) Bar graph showing quantification of spectrophotometry-measured absorbance (560 nm) of tissue secreted H₂O₂ (picomoles per milligram of dry tissue). *, P = 0.0168. (B) Representative Western blot and densitometry of HIF-1α expression in outflow veins of V-A and A-V fistulae at days 3 and 7; P = 0.9920 and P = 0.8514 at days 3 and 7 respectively. (C) Bar graph showing percentage of HIF-1α positive cells in V-A and A-V outflow veins at days 3 and 7; P = 0.8509 and P = 0.8412 at days 3 and 7 respectively. (D) Bar graph showing percent of pimonidazole-positive cells in the V-A and A-V outflow veins at days 3 and 7; P = 0.7944 and P = 0.9266 at days 3 and 7 respectively. (E) Representative photomicrographs of merged immunofluorescence of vWF (red), pimonidazole (green), and DAPI (blue) in V-A and A-V outflow veins, as well as control vein, at days 3 and 7; yellow arrows show dual-positive cells; scale bar 100 µm. (F) Bar graphs quantifying ratio of pimonidazole-positive cells in all vWF positive cells; P = 0.6754 and P = 0.8651 at days 3 and 7 respectively. (G) Bar graphs quantifying ratio of pimonidazole-positive cells in all α-actin positive cells; P = 0.8149 and P = 0.1642 at days 3 and 7 respectively. (H) Bar graphs quantifying ratio of pimonidazole-positive cells in all CD68-positive cells; P = 0.8707 and P > 0.999 at days 3 and 7 respectively. (I) Bar graphs quantifying ratio of pimonidazole-positive cells in all HO-1-positive cells; P = 0.7540 and P = 0.7444 at days 3 and 7 respectively. n = 3 animals per group in panels A-D and F-I. Bar graphs in panels A-D and F-I represent mean ± SEM; P values were calculated using a two-tailed unpaired Student’s t test.

Different hemodynamics in A-V and V-A configurations

Surgical creation of AVF alters hemodynamics, increasing the magnitude and frequency of shear stress on the venous endothelium (32); these changes are associated with fistula remodeling, with NIH correlating inversely with the magnitude of shear stress (33, 34). Since there is less NIH in the A-V configuration compared with the V-A configuration, but no difference in hypoxic stress between the two, we compared the hemodynamics in both configurations in the rat models. Computational modeling was enabled by micro-computed tomography (CT) images and informed by ultrasound data for both AVF configurations at day 3 following fistula creation. The simulations predicted that pressures in the perianastomotic outflow veins, averaged over a cardiac cycle, were lower in the A-V configuration compared to the V-A configuration (46.98 ± 1.54 mmHg versus 66.06 ± 14.18 mmHg; P = 0.0814) (Fig. 7A). Further, time averaged wall shear stress (TAWSS) over a cardiac cycle was decreased in the perianastomotic outflow vein in the A-V versus V-A configuration (85.57 ± 22.94 Pa versus 156.75 ± 26.75 Pa; P = 0.0249) (Fig. 7B). The order of magnitude of the computed wall shear stress was comparable to measurements in the rat systemic circulation (35). Notwithstanding limitations associated with the assumptions of resistance boundary conditions and rigid walls, these data suggest that lower pressures and shear stress in the A-V vein precede reduced neointimal hyperplasia and improved long-term patency in this animal model of RADAR. Ultrasound of both V-A and A-V fistulae 7 days after creation showed increased peak systolic velocity in the A-V configuration (Fig. 8A). There was also increased expression of Krüppel-like factor 2 (Klf2) at day 21 after fistula creation, and a trend but no significant increase in phosphorylated endothelial nitric oxide synthase (p-eNOS) in the A-V configuration (fig. S4, A to C), consistent with improved hemodynamics (36–38). Expression of both Klf2 and p-eNOS normalized by day 42 in both configurations (fig. S4, A to C). Immunofluorescence analysis indicated increased p-eNOS expression in the A-V configuration in endothelial cells on days 7 and 21 (fig. S4, D and E). These data suggest improved hemodynamics in the A-V configuration compared with the V-A configuration.

Fig. 7. Different simulated hemodynamics in V-A and A-V configurations in non-CKD male rats.

(A) Spatial distribution of pressure (mmHg), averaged over a cardiac cycle, in both the V-A (top row) and A-V configurations (bottom row) in 3 different non-CKD male animals. Arrow indicates anastomotic site. (B) Time averaged (over a cardiac cycle) wall shear stress (TAWSS), in both the V-A (top row) and A-V configurations (bottom row), in 3 different non-CKD male animals.

Fig. 8. Ligation of V-A outflow is associated with improved hemodynamics and decreased neointimal hyperplasia in non-CKD male rats.

(A) Bar graph showing ultrasound-assessed peak systolic velocity in the inflow and outflow vessels of fistulae in V-A and A-V configurations, day 7; *, P = 0.2595 and P = 0.0112 for inflow and outflow respectively. (B) Schematic of the V-A fistula configuration with outflow ligation in a rat model. CCA, common carotid artery; CJV, common jugular vein; EJV, external jugular vein; IJV, internal jugular vein; blue arrow represents the direction of venous flow; red arrow represents the direction of arterial flow. (C) Bar graph quantifying peak systolic velocity in control and ligated V-A fistulae, day 14; *, P = 0.01. (D) Bar graph quantifying outflow diameter in control and ligated V-A fistulae, day 14; P = 0.3936. (E) Bar graph quantifying flow in control and ligated V-A fistulae, day 14; *, P = 0.0892. (F) Bar graph quantifying shear stress (SS) in control and ligated V-A fistulae, day 14; *, P = 0.0585. (G) Left panels: representative high-power H&E staining showing outflow (internal jugular vein 1 mm distal from the anastomosis) neointimal thickness at day 14 in control and ligated V-A fistulae; scale bar 100 µm; dashed yellow line delineates intima/media transition. Right panels: merged immunofluorescence of vWF (green), α-actin (red), and DAPI (blue) showing neointimal thickness in outflow vessel; scale bar 100 µm. (H) Bar graph quantifying neointimal thickness at day 14 in control and ligated V-A fistulae; *, P = 0.0027. (I) Bar graph quantifying proliferative index, day 14; *, P < 0.0001. (J) Bar graph quantifying apoptotic index, day 14; P = 0.4131. n = 2-3 animals per group in panel A; n = 5-6 animals in panels C-F and H; n = 3-4 animals per group in panels I-J. Bar graphs in panels A, C-F and H-J represent mean ± SEM; P values were calculated using a two-tailed unpaired Student’s t test.

Since computational modeling shows that the A-V configuration is associated with different hemodynamics as early as day 3 compared to the V-A configuration, and molecular data is consistent with improved hemodynamics by day 7 coincident with decreased NIH, we next determined whether improvement of hemodynamics in the V-A configuration would diminish NIH. After creating a V-A configuration of a fistula, the common carotid artery was ligated distally to the anastomosis to increase flow through the fistula, thereby potentially improving its hemodynamic environment (Fig. 8B). After 14 days, ultrasound was used to examine the fistulae; there was increased peak systolic velocity as well as flow volume and calculated shear stress in the V-A fistulae with ligated outflow compared to unligated V-A controls (Fig. 8, C to F), with values similar in magnitude to those of the A-V configuration. Neointimal hyperplasia was visualized using H&E staining or immunofluorescence to detect α-actin and von Willebrand factor (Fig. 8G); there was reduced neointimal thickness (Fig. 8H) and cellular proliferation (Fig. 8I), with no change in apoptosis (Fig. 8J) in the fistulae with ligated outflow compared with unligated controls. These data suggest that improved hemodynamics associated with increased flow through a V-A configuration AVF are associated with diminished neointimal hyperplasia, and therefore the improved patency in human patients with RADAR artery-to-vein configuration may be secondary to an improved hemodynamic environment compared with the vein-to-artery configuration of traditional radial-cephalic AVF.

Discussion

We present 3-year follow up of patients treated with the RADAR (artery-to-vein) configuration of AVF compared with a control group of patients treated with the traditional (vein-to-artery) configuration of radial-cephalic AVF. We demonstrate that the initially improved fistula maturation and patency and reduced reintervention rates persist 3 years after fistula creation, and are similarly impactful in female as well as male patients. (28) In a rat model that mimics the RADAR (A-V) and traditional (V-A) configurations, the A-V configuration was associated with less proliferation and NIH in both sexes, with similar trends in uremic male animals. Although we did not detect any difference in hypoxia between A-V and V-A configurations, there were several differences in the hemodynamic environments between the A-V and V-A configurations; improving the hemodynamics of the V-A configuration was associated with diminished neointimal hyperplasia. These results suggest that AVF in both the human RADAR and the rat A-V configuration have less NIH and improved rates of maturation and patency secondary to improved hemodynamics.

Our primary finding is that the artery-to-vein configuration of an AVF results in decreased neointimal hyperplasia compared with the traditional vein-to-artery configuration, both in human patients and in a rat model, independent of sex. Neointimal hyperplasia after fistula creation results in juxta-anastomotic venous stenosis secondary to aberrant cellular activation, proliferation and migration, as well as inappropriate extracellular matrix remodeling (17). Multiple strategies aimed at reducing neointimal hyperplasia in AVF including cryoplasty, brachytherapy, and adventitial wraps have been evaluated in clinical trials, with equivocal results and no long-term success (39–41). We report long-term safety and patency of our artery-to-vein surgical technique in humans, consistent with our previous pilot study (28).

Our data also show that creation of an artery-to-vein configuration of an AVF results in improved hemodynamics in the venous outflow, compared to the traditional vein-to-artery configuration. Venous remodeling that occurs during AVF maturation results predominantly in outward dilation in the high flow, high shear stress, and low pressure fistula environment, whereas venous remodeling during vein graft adaptation to the arterial environment results in predominantly increased wall thickening and is associated with higher pressures than the AVF environment (42, 43). Whereas conventional vein-to-artery AVF exposes the vein to a high-pressure, high-flow configuration, the artery-to-vein RADAR procedure exposes the vein to a moderate-pressure, high-flow configuration; previous computational simulations suggest that reduced pressure could be more favorable to long-term venous adaptation (44). Although increasing venous outflow pressures in our modified fistula result in decreased NIH, outflow ligation also increased shear stress that inversely correlates with formation of NIH (32, 34). Thus, although improved maturation and decreased NIH in the A-V configuration can be explained in part by decreased outflow venous pressure, increased shear stress likely plays a larger role (44).

Our data shows that the A-V configuration has less neointimal hyperplasia compared to the traditional V-A configuration of radial-cephalic AVF, both in humans and in a rat model. The A-V configuration may have more favorable hemodynamics that account for this, and importantly, the A-V configuration shows less neointimal hyperplasia in human women treated with RADAR as well as in female rats. Women traditionally have had reduced rates of AVF maturation and patency and accordingly less utilization of AVF, and these worse results are thought to be a function of the smaller diameter of blood vessels in women compared to men (5, 45, 46). Our data suggests that women achieve superior outcomes using the A-V configuration of AVF, a finding of translational importance. Similarly, our animal model shows that rats with CKD also have less neointimal hyperplasia in the A-V configuration; not only is this translationally relevant, that is the A-V configuration is of benefit in patients with CKD, but also these data in toto suggest that hemodynamic differences between the A-V and V-A configurations may be a mechanism underlying the different rates of neointimal formation.

We did not detect a difference in hypoxic stress between the A-V and V-A configurations in our animal model. Because vessel wall hypoxia stimulates cell proliferation and NIH, and surgical dissection during venous mobilization to perform the conventional vein-to-artery AVF disrupts venous vasa vasorum, leading to relative wall hypoxia and potentially NIH, the original concept of the RADAR technique was based on surgical techniques that minimized venous dissection to eliminate hypoxic injury (17, 47). Because both arterial and venous segments are mobilized to a similar extent in our animal model, we hypothesized that there is no difference in hypoxic stress between the A-V and V-A configurations, and thus this model is useful in determining the effects of altered hemodynamics resulting from differences in geometry, without confounding effects of hypoxia. Although our data suggest that hemodynamic changes are a mechanism of improved clinical efficacy of RADAR, the possibility of less wall ischemia in this configuration may still be another potential mechanism. Nonetheless, our animal model, although a useful model for studying hemodynamic changes associated with altered fistula geometry between the A-V and the V-A configurations, is not a perfect recapitulation of the human RADAR surgical procedure. Whereas the success of RADAR in humans may stem partly from decreased venous mobilization and therefore reduced wall ischemia, the arterial and venous components are completely mobilized in both configurations in our animal model, resulting in similar ischemic changes to the venous wall. As such, any improvements in maturation or patency observed in our animal model are secondary to altered hemodynamics and geometry, not less vessel wall ischemia. Another important limitation of the animal model is that although the rate for bilateral fistula creation in male and female wild type rats approached 100%, survival rates were much lower in uremic animals, limiting our sample size; the high mortality rate following fistula creation in uremic rats precluded sex comparisons, limiting our data to male rats.

There are several important limitations to the RADAR data. Our human study is a retrospective report of a prospectively maintained registry, lacking randomization and power calculations, and thus selection bias due to the non-randomized design of the study could have influenced the results. However, since there was no overlap between the two techniques, this bias should be limited; a prospective multicenter randomized controlled trial of RADAR compared with traditional V-A AVF is currently ongoing in France (www.clinicaltrial.gov, NCT02728817). The percentage of patients with diabetes in this study (control: 34% and RADAR: 30%) is lower than the percentage of patients with diabetes reported in France, and this discrepancy may have occurred due to selection bias. Since patients with diabetes and end stage renal disease often present with severe calcifications of distal arteries and diseased venous conduits, it is possible that some patients may not have been selected for the creation of a distal AVF, artificially inflating our observed inter group differences. Furthermore, reproducibility of the RADAR technique by surgeons outside of our group may vary. We recommend the following principles to ensure success: careful intraoperative vessel identification and anastomosis planning with duplex ultrasound prior to incision to ensure minimal dissection; minimal venous dissection extending only on the anterior and medial aspect and never circumferentially; dissection of the radial artery with its vascular pedicle for a distance 2 times that between the artery and vein; and creation of the anastomosis with little slack in the radial artery to avoid excess length. Lastly, our RADAR cohort had smaller diameter arteries and veins, compared to controls, and this was associated with higher odds of developing juxta-anastomotic stenosis; however, these findings agree with published findings suggesting a detrimental effect of smaller diameter vessels on fistula patency, and support the superiority of the RADAR configuration, especially in people with smaller diameter vessels such as women; therefore, these data suggest broad applicability of the RADAR technique (46, 48).

In summary, the RADAR technique confers a durable improvement in maturation and patency rates in humans that require a radial-cephalic AVF for hemodialysis access, regardless of sex; improved fistula hemodynamics is a mechanism for success of the artery-to-vein configuration, resulting in decreased neointimal hyperplasia. These findings are of translational importance and should influence surgical practice to inspire future clinical trials that evaluate outcome of different fistula geometry.

Methods

Study design

The aim of our study was to determine whether the RADAR configuration of AVF in humans is associated with improved maturation and long-term patency rates, compared to the traditional V-A configuration of AVF, as determined by ultrasound imaging. To confirm our human data, and to determine potential mechanisms for the improved patency rates in the RADAR configuration, we developed a rat bilateral carotid artery-internal jugular vein AVF model to assess both A-V and V-A configurations in a single animal and performed hemodynamic studies using ultrasound, CT, and computational modeling, in addition to histological analysis of tissue samples collected at varying time points. The human portion of our study is a single institution cohort study, with retrospective review of a prospectively maintained clinical database. Secondary to the retrospective nature of our study, randomization was not performed. The RADAR cohort consisted of all consecutive patients who had RADAR AVF constructed between October 2014 and June 2015, whereas the control group included all traditional radial-cephalic fistulae constructed between January 2013 and September 2014. This study was conducted in conformity with the Declaration of Helsinki. Informed consent was obtained for all procedures as well as for database registration. According to French law, the ethics committee declined review of this study secondary to its retrospective design. Where applicable, rats were assigned to treatment groups randomly and analysis was performed in a blinded fashion. Sample sizes for rat studies were determined based on the magnitude of change observed between groups and are defined in the figure legends. All animal experiments were approved by the Institutional Animal Care and Use Committee of Yale School of Medicine and were in strict compliance with federal guidelines. Primary data for small sample sizes is included in Data File S1.

Human subjects

All RADAR procedures performed between October 2014 and April 2017 were included in this study; the control group included all radial-cephalic AVF constructed between January 2013 and September 2014. No RADAR fistulae were constructed before October 2014. Side-to-side configurations were excluded from this analysis. Patients were referred for vascular access creation when the eGFR was <10 mL/min/1.73m². The Vascular Access registry of the Department of Vascular Surgery of the University Hospital of Nice is registered at the French National Information Commissioner’s Office under N°236 for the University Hospital of Nice; this registry maintains prospective follow-up of all patients having AVF, including those with RADAR.

Surgical procedure and follow up

Surgical procedures were performed as previously described (28). Patients were required to have a patent ulnar artery and palmar arch that were free from heavy calcifications, suitable vessels greater than 2 mm in diameter, and sufficient venous length to allow future puncture with dialysis access needles. Patients who did not satisfy one or more of these criteria were not eligible for any type of radial-cephalic fistula and were offered alternate access. The same vascular surgery team (N.S. and S.D.) performed all the surgical procedures together. Both surgeons were present and participated in each operation. Radial-cephalic AVF that were used as a comparative group were constructed by deviation of the cephalic vein towards the radial artery, with an end-to-side vein-to-artery anastomosis. RADAR AVF were created as previously described (28). Briefly, minimal dissection was used to prepare a 15 mm segment of the cephalic vein. The distal radial artery was then ligated using two surgical clips. The radial artery pedicle was gently turned toward the minimally dissected medial aspect of the vein to form a smooth loop. Under tourniquet control for hemostasis, the anastomosis was performed between the end of the artery and the side of the vein using a continuous suture of 8–0 polypropylene, without the use of any surgical clamps or mobilization of the vein.

After AVF creation, patients underwent a clinical examination by the surgeon at 4 to 6 weeks post-operatively, in addition to monthly nephrology visits for patients already on hemodialysis. Patients were monitored regularly with duplex ultrasound; the first examination occurred 6 weeks post operatively, followed by every 3 months for the first year, and twice per year thereafter; patients with clinical suspicion of stenosis were also examined.

Symptomatic stenosis was defined as an ultrasound-visualized lesion located anywhere in the outflow vein or inflow artery that caused thrombosis, failure to mature at 3 months, loss of thrill, >20% reduction in access flow, >20% access recirculation rate, inability to complete a dialysis session in 4 hours, or increased bleeding time after puncture. Juxta-anastomotic stenosis was defined as a stenosis located within 4 cm of the anastomosis, on either the venous or arterial limb of the fistula. Complications were adjusted for the duration of follow-up for each group and given as incidence per 100 patient-years.

Rat arteriovenous fistula model

Male and female Wistar rats, ages 6–8 weeks old, were used for AVF creation. Microsurgical procedures were performed aseptically in a dedicated facility using a dissecting microscope (Leica MZ 95), similar to previously described rat fistula models (49). After induction of anesthesia with inhaled isoflurane each animal received a subcutaneous injection of 10 mg/kg Xylazine prior to incision. A midline neck incision was then made; blunt and sharp dissection were used to expose the common carotid artery (CCA) and internal jugular vein (IJV). On the animal’s right side, the IJV was dissected and its tributaries were ligated to provide adequate length to perform a tension free anastomosis. The IJV was then clamped proximally and ligated distally in the neck; the CCA was clamped with atraumatic microvascular clamps and a 1 mm longitudinal arteriotomy was made on the CCA. The IJV was then anastomosed to the CCA in an end-to-side (vein-to-artery) configuration, using a running 10-0 polypropylene suture. Hemostasis was achieved with gentle manual pressure. In the left neck of the same animal, the distal CCA, close to the bifurcation, was ligated and divided; a 1-mm venotomy was made in the IJV. The CCA was then anastomosed to the IJV using the same end-to-side (artery-to-vein) configuration and suturing technique. In both configurations, the angle between the CCA and the IJV was approximately 90°. The skin incision was then closed in two layers using running 5-0 and 4-0 absorbable sutures. Each animal then received 0.05-0.1 mg/kg intramuscular buprenorphine every 6-12 hours for 24 hours post operatively. No heparin or antibiotics were used in this model. Survival was routinely >95% in male rats without renal failure and 100% in the 3 female rats.

In some animals an outflow ligation was performed; an IJV to CCA (vein-to-artery) fistula was created on the right side of the neck, without contralateral dissection. After fistula creation, the distal CCA outflow was suture ligated with a 6-0 polypropylene suture, without CCA transection, to maintain the geometry of the fistula without any alterations. 2 weeks following fistula creation animals were imaged with ultrasound and samples were collected for further histological processing.

Kidney disease was induced in some animals prior to fistula creation either via surgical 5/6 nephrectomy or with an adenine diet. The 5/6 nephrectomy model of CKD is an established model of CKD without associated hypertension (50). The 5/6 nephrectomy was performed first as a separate procedure. After induction of anesthesia with inhaled isoflurane each animal received a subcutaneous injection of 10 mg/kg Xylazine prior to incision. A midline laparotomy incision was then made; blunt and sharp dissection was used to expose bilateral kidneys. Each animal then underwent a right total nephrectomy and excision of the upper and lower poles of the left kidney. Hemostasis was achieved with a combination of manual pressure and, if needed, suture ligation of the renal hilum and/or poles with a polypropylene suture. The abdomen was then closed in two layers using running 5-0 and 4-0 absorbable sutures. Each animal then received 0.05-0.1 mg/kg intramuscular buprenorphine every 6-12 hours for 24 hours post operatively. One week following nephrectomy CKD was confirmed by measuring blood urea concentration via tail vein prick and uremic animals then underwent bilateral fistula creation as described above. Alternatively, some animals had kidney disease induced with an adenine diet; this group received 0.25% adenine suspended in drinking water continuously starting 3 days prior to fistula creation and for the duration of the study until 2 weeks after fistula creation. CKD was similarly confirmed by measuring blood urea concentrations via tail vein prick prior to bilateral fistula creation (51). 35 rats had 5/6 nephrectomy and subsequent AVF creation, of which 5 survived (14.3%), and 3 had patent bilateral fistulae. 20 rats had CKD induction via an adenine diet, of which 3 survived (15%), with all 3 fistulae being patent.

At the time of harvest, rats were anesthetized with isoflurane and the previous neck incision was opened with scissors. Meticulous subcutaneous dissection was then carried out to expose the unilateral or bilateral fistulae. Fistula patency was then confirmed by visualization of blood flow proximal and distal to the anastomosis as well as in the outflow jugular vein. The entire AVF was excised, and the outflow jugular vein was analyzed 1 mm distal to the anastomosis in each case. All nephrectomy and bilateral A-V and V-A operations were performed by a single surgeon (H.B.), and all unilateral V-A fistulae (without nephrectomy) were created by a different single surgeon (J.G.).

Western blot

After anesthesia with isoflurane, each of the outflow veins of bilateral AVF and control ipsilateral external jugular veins were excised and snap frozen in liquid nitrogen. Samples were then crushed and mixed with lysis buffer containing protease inhibitors (Roche, cOmplete Mini Protease inhibitor), prior to sonication (5 sec) and centrifugation (135000 rpm, 15 min). Equal amounts of protein from each sample were then loaded on acrylamide gels for SDS-PAGE. Gels were subsequently transferred to polyvinylidene difluoride membranes (Millipore). Membranes were incubated overnight at 4°C with one of the following primary antibodies: anti-HIF-1α (Novus biological, 1:1000), anti-Klf2 (Abcam, ab15580; 1:5000), anti-β-actin (Cell Signaling, 4970; 1:5000), anti-GAPDH (Cell Signaling, 14C10; 1:2000) and anti-phospho-eNOS (peNOS) (Santa Cruz Biotechnology, 612392; 1:2000). Secondary antibodies were HRP-linked (Santa Cruz Biotechnology). Protein signaling was then detected using ECL detection reagent (Life Technologies). Films were subsequently scanned digitally and band densitometry was performed using Image J software and were normalized to GAPDH or β-actin.

Immunohistochemistry/Immunofluorescence

After anesthesia with isoflurane, a thoracotomy was performed, and tissues were fixed by transcardial perfusion of phosphate buffered saline (PBS) followed by 10% formalin. Vessels were then explanted and fixed overnight in 10% formalin, followed by a 24-hour immersion in 70% ethanol. Vessel sections were then embedded in paraffin and sectioned (5 μm thickness). Tissue sections were de-paraffined using xylene and a graded series of alcohols. Sections were heated in citric acid buffer (pH 6.0) at 100 °C for 10 min for antigen retrieval. The sections were blocked with 5% bovine serum albumin PBS containing 0.05% Triton X‐100 for 1h at room temperature. Tissue was stained for morphological analysis with hematoxylin/eosin or elastin Van Gieson (EVG) stain (Dako).

To determine fistula neointimal hyperplasia area, high power images of the entire circumference of elastin Van Gieson (EVG)-stained samples were examined. Image J software was used to measure the area of neointimal hyperplasia by tracing first the outer diameter of the intima/media transition, followed by tracing the inner diameter of neointimal hyperplasia. The fistula diameter was determined by measuring the fistula diameter in the widest axis in each specimen and the neointima/lumen ratio was established by dividing neointimal area by fistula area. Neointimal thickness was measured by taking 6 high-powered field images of EVG-stained fistulae and averaging the thickness from the intima/media transition across samples. All measurements were completed in a blinded fashion, without the use of a pathologist.

For immunohistochemistry, sections were incubated overnight at 4°C with the following primary antibodies: anti-proliferating cell nuclear antigen (PCNA, M0879, Dako, 1:50), anti-cleaved caspase-3 (apoptosis) (Cell Signaling, 9661, 1:50), anti-HIF-1α (hypoxia) (Abcam, ab51608, 1:50). After overnight incubation, sections were incubated in EnVision reagents for 1 h at room temperature and treated with Liquid DAB Substrate Chromogen System (Dako). Finally, sections were counterstained with Dako Mayer’s Hematoxylin. Images were captured with light microscopy (Olympus Q-Color 5). Positively stained cells were manually counted in 5-8 high-power fields, then averaged. All cellular counts were performed in a blinded fashion, without the use of a pathologist.

For immunofluorescence, sections were incubated overnight at 4°C with the following antibodies: anti-CD68 (macrophages; Abcam, ab31630; 1:100), anti-α-actin (smooth muscle cells; Abcam, ab5694; 1:100), anti-vWF (endothelial cells; Abcam, ab11713, 1:100), anti-peNOS (Abcam, ab184154 1:100), anti-heme oxygenase 1 (Abcam, ab13248 1:100). Secondary antibodies were linked to Alexa Fluor 488 or 568 (Invitrogen, Carlsbad, CA). Sections were stained with 4,’6-diamidino-2-phelylindole (DAPI, Invitrogen) to mark nuclei. Digital fluorescence images were captured and positively stained cells alone or co-localized with cellular types of interest were manually counted in 5 high-power fields, then averaged. All cellular counts were performed in a blinded fashion, without the use of a pathologist.

Pimonidazole staining

Pimonidazole hydrochloride staining was used to localize ischemic tissues (52). Briefly, pimonidazole (60mg/kg) was injected intra-peritoneally 1-2 hours prior to vessel explantation; localization of pimonidazole was determined by immunofluorescence according to the manufacturer’s protocol (Hydroxyprobe, Inc). Positively stained cells alone or co-localized with cellular types of interest were manually counted in 5 high-power fields, then averaged. All cellular counts were performed in a blinded fashion, without the use of a pathologist.

Measurement of extracellular reactive oxygen species

The Amplex Red assay kit (Life Technologies) was used for quantitative detection of extracellular hydrogen peroxide (H₂O₂) as previously reported (31). Briefly, under anesthesia, rats were infused with a modified Krebs-Ringer solution and the outflow veins of the AVF were excised, sectioned into 2-mm rings, and transferred to a 96-well plate containing Amplex Red (50 mM 10-acetyl-3,7-dihydroxyphenoxazine) and horseradish peroxidase (0.1 U/mL). After a 1-hour incubation at 37°, vessels were removed and a spectrophotometer was used to measure solution absorbance at 560 nm. H₂O₂ standards were used as controls. H₂O₂ accumulation was normalized to dry tissue weight (picomoles per milligram of dry tissue).

Animal imaging and computational modeling

To obtain CT images, animals (post-operative day 3) were anesthetized with inhaled isoflurane (1-2%) and placed supine on a horizontal stage. Images were obtained with a micro-CT scanner (McroSPECT4CT; MILabs) with a cone beam filtered back projection algorithm, set to an 80-μm effective voxel size. Micro-CT was performed with 55-kVp X-ray tube voltage, 370-μA tube current, 40 milliseconds per frame, 360° angle, and 0.25° increments per view. A long circulating contrast agent, ExiTron nano 12000 (MiltenyiBiotech), was injected (0.5 mL) via tail vein and animals were imaged 7-15 min after contrast injection.

For Doppler ultrasound imaging, animals were anesthesized with inhaled isoflurane (1%) and placed prone on a horizontal stage. Ultrasound (40 MHz; Vevo770 High Resolution Imaging System; Visual Sonics Inc., Toronto, Ontario, Canada) was then used to obtain color flow images of bilateral neck vasculature, at baseline and over time up to post-operative day 28.

Rat-specific hemodynamic simulations were performed using the open-source software SimVascular (www.simvascular.org) (53, 54). Model construction and simulation were similar to prior work (55, 56). Briefly, a micro-CT image stack at day 3 post-surgery (fig. S5A) was used to construct pathlines along the center of the lumen proximal and distal to an anastomosis (fig. S5B). Lumens were segmented on 2D slices perpendicular to the pathlines, then segmentations were lofted to construct a 3D geometric model of the vasculature (fig. S5C). This 3D model was discretized into finite tetrahedral elements to simulate the hemodynamics over the cardiac cycle (fig. S5D) using a stabilized finite element method to solve the incompressible Navier-Stokes equations and to determine spatial and temporal distributions of pressure and velocity (fig. S5E) (57). Other metrics of interest can be determined from these basic quantities, including distributions of wall shear stress. Solution of the Navier-Stokes equations requires conditions at the boundary surfaces. These boundary conditions can be broadly divided into three types – a flow-boundary in which flow is prescribed at a surface, a resistance-boundary in which pressure is proportional to and in phase with the flow, and a rigid boundary for which the vessel wall does not deform. We used the average velocity derived from ultrasound measurements to prescribe flow conditions at appropriate boundaries. Specifically, temporal profiles of flow over a cardiac cycle were inferred from ultrasound and prescribed at the inlet of the proximal section of the vessel as well as at outlets having measured flow-reversal. Because ultrasound and micro-CT measurements were on separate rats, ultrasound measured flow into the carotid artery of a classical arteriovenous fistula was scaled such that the cycle-averaged inlet pressure was approximately 100 mmHg; all other prescribed-flow inlets/outlets were scaled proportionately. Consistent with electrical analog models for flow and pressure, we assigned resistance boundary conditions, to account for viscous dissipation, at the remaining outlets (58). Resistance values for vessels visible on ultrasound were tuned to match the measured percent flow split; resistance for vessels visible on micro-CT, but not on ultrasound, were based on their cross-sectional area. We used the average velocity derived from ultrasound measurements to inform flow conditions at the boundaries (fig. S5F). All vascular walls were assumed to be rigid. Simulations were run for six cardiac cycles (0.2 s/cycle) and results from the last cycle were used for post-processing. Blood was assumed to exhibit a Newtonian behavior, with a dynamic viscosity of 0.004 kg/(m-s) and a density of 1060 kg/m³.

Statistical analysis

For human studies, the primary endpoint was the cumulative intervention rate on the juxta-anastomotic segment at 12 months. Secondary end points included the AVF maturation rate at 6 weeks and 3 months, AVF primary and secondary patency at 12 months, and the post-operative complication rate. Primary patency was defined as the interval between fistula creation and any intervention to maintain or re-establish patency, fistula thrombosis, or time of patency measurement (59, 60). Secondary patency was defined as the interval between fistula creation and access abandonment or time of patency measurement, including interventions designed to re-establish functionality in a stenosed or thrombosed fistula (59, 60). Fistulae were considered to be mature after completion of a dialysis session in 4 hours using two needles or with an arterial flow of 500 ml/min and vein diameter ≥5 mm, as determined by duplex ultrasound. Kaplan-Meier survival analysis was used to estimate the cumulative intervention and patency rates, and the log rank test was used to compare groups. Comparison of maturation rates was made at each time point using the Chi-squared test. Patients who received a renal transplant were censored at the time of transplantation. Comparisons of outcomes between groups were reported as absolute risk reduction, 95% confidence interval (CI) and as odds ratio (OR), 95% CI. Statistical analysis was performed with SPSS (IBM Corp, Armonk, NY).

Categorical variables were described as number and frequency and compared with a Chi-square test while quantitative variables were described as mean and SEM. The Shapiro-Wilk test was performed to analyze normality, where applicable. For two group comparisons with normally distributed data, the unpaired Student’s t test was used and the unpaired Student’s t test (with Welch correction) was used for data with unequal variances. For multiple group comparisons with normally distributed data, the one-way ANOVA followed by the Tukey post-hoc test was used. For animal experiments, statistical analysis was performed using GraphPad Prism. Densitometry analysis after Western Blot was performed on scanned images using ImageJ software. P values less than 0.05 were considered significant.

Supplementary Material

Supplemental FIgures and Table

Fig. S1. Flow chart summarizing control and RADAR group selection in human patients.

Fig. S2. Supplementary data for the male rat models of arteriovenous fistulae.

Fig. S3. Similar hypoxic stress in V-A and A-V configurations in non-CKD male rats.

Fig. S4. Different hemodynamics in V-A and A-V configurations in non-CKD male rats.

Fig. S5. Methodological workflow for simulating hemodynamics in a non-CKD male rat-specific geometry using the open-source software SimVascular.

Table S1. Binomial logistic regression to determine likelihood of juxta-anastomotic stenosis based on fistula type, venous and arterial diameters in human patients.

Table of primary data

Data File S1. Individual subject-level data for small sample sizes