Graft geometry and venous intimal-medial hyperplasia in arteriovenous loop grafts M a r k F. Fillinger, M D , E m a n u e l R. Reinitz, M D , R o b e r t A. Schwartz, MD,
Dennis E. Resetarits, MD, Andrew M. Paskanik, David Bruch, MS, and Carl E. Bredenberg, MD, Syracuse, N.Y. This study explores graft geometry and hemodynamics in a reproducible canine arteriovenous loop graft model o f intimal-medial hyperplasia. Untapered 6 m m diameter polytetrafluoroethylene grafts (n = 10) were paired with 4 to 7 m m taper (n = 5) or 7 to 4 nun taper (n = 5) grafts for a 12-week period. Several hemodynamic variables were assessed at multiple locations, and venous intimal-medial thickness was measured at locations corresponding to the hemodynamic measurements. Color Doppler imaging demonstrated energy transfer out o f the vessel in the form of perivascular tissue vibration. This was quantitated by the distance required for Doppler signal attenuation or volume o f the detected vibration signal. Differences among graft types were noted for pressure, flow velocity, tissue vibration, and venous intimal-medial thickness. Hyperplasia was significantly decreased in 4 to 7 m m taper grafts. Stepwise deletion regression indicated volume o f the vibration signal had a better correlation with venous intimal-medial thickness than any other variable (r 0.9,p < 0.001). We conclude that graft geometry can have a significant impact on hemodynamic factors and venous intimal-medial hyperplasia in arteriovenous loop grafts. Flow disturbances appear to cause energy transfer through the vessel wall and into perivascular tissue. Kinetic energy transfer in the form ofperivasoalar tissue vibration was quantitated in vivo and correlates strongly with venous intimal-medial thickness. (J VAsc SURG 1990;11:556-66.)
Prosthetic arteriovenous grafts are more prone to failure as a result of anastomotic intimal hyperplasia than any other type of vascular graft. In polytetrafluoroethylene (PTFE) hemodialysis access grafts hyperplasia causes more than 50% of all failures, and graft patency at i year is less than 80%. 1 Numerous hemodynamic theories have been proposed to explain this phenomenon, but relatively few in vivo studies have been done. In a previous study we demonstrated a quantitative relationship between Reynolds number (a parameter related to flow stability and turbulence) and venous intimal-medial thickening in arteriovenous loop grafts? In that study turbulence was decreased by placing a flow-limiting band on 6 mm diameter grafts. However, in an ideal graft, flow disturbances would be minimized without reducing flow or requiring the creation of a severe graft stenosis. Toward that end, this study explores From the Department of Surgery, State University of New York Health Science Center, Syracuse. Presented at the Third Annual Meeting of the Eastern Vascular Society, Bermuda, May 4-7, 1989. Reprint requests: Robert A. Schwartz, MD, Department of Surgery, State University of New York Health Science Center, 750 E. Adams St., Syracuse, NY 13210. 24/6/18322
556
the hemodynamic patterns created by three different graft geometries and the resulting impact of these hemodynamic patterns on intimal hyperplasia. The noninvasive study of in vivo flow disturbance and energy related phenomena was made possible by the use of color Doppler ultrasound imaging technology. MATERIAL AND M E T H O D S Study design. Grafts were implanted bilaterall) in a paired fashion that allowed for a side-by-side comparison of three different graft geometries. A standard untapered 6 mm diameter graft was always paired with either a 7 to 4 nun taper graft (7 mm diameter at the arterial anastomosis, 4 mm at the distal venous anastomosis) or a 4 to 7 mm taper graft (4 mm at the arterial anastomosis) as shown in Fig. 1. Animal care complied at all times with the "Principles of Laboratory Animal Care" and the "Guide for the Care and Use of Laboratory Animals" (NIH Publication No. 80-23, revised 1985). Implantation protocol. Fourteen mongrel dogs weighing 17 to 23 kg were anesthetized with intravenous sodium pentobarbital (30 mg/kg), and bilateral femoral arteriovenous loop grafts were implanted with pairings as described in the study design section. All grafts were 25 cm in length and made of
Volume 11 Number 4 April 1990
Graft geometry and intimal-medial hyperplasia 557
i lilacvein..~ Femoralv e i n ~
/
I ~ . . l I l i a c artery ~ I n g u i n a l ligament ~ . ~ F / . ~ ~ ~-Femoralartery
~ - -
Graft
Vessel I
#1
#2 #3 #4
#5
LocationNumbers Fig. 2. Subdivision of vessels for histology. Note the location numbers for orientation in the text and other figures.
4-7 mm Taper Graft
6mm Untapered Graft
Fig. 1. Implant configuration. One side is always a 6 mm untapered graft. expanded PTFE (Gore-Tex, registered trademark of W. L. Gore & Assoc., Elkton, Md). After mobilization of the femoral artery and vein, internal vessel diameter was gently measured by use of coronary dilators. An end-graft to side-artery anastomosis was then created with a continuous suture of 6-0 polypropylene. The graft was tunneled in a subcutaneous loop before end-graft to side-vein anastomosis. Graft implantation on the opposite side was carried out in the same fashion. Local heparin flushing was the Only anticoagulation used. Prophylactic antibiotics were given up to 5 days after operation. The grafts were left in place for a period of 12 weeks. Graft patency was checked frequently with a portable, continuouswave 5 M H z Doppler transducer. Chronic hemodynamic measurements. After implantation of both grafts, systolic, diastolic, and mean volumetric flow rates were recorded with an electromagnetic flowmeter (Carolina Medical Electronics, King, N.C.). These measurements were repeated in anesthetized animals at the time they were killed. The Collected data allowed for the calculation of flow pulsatility, mean flow velocity, and Reynolds number. For the purposes of this study, flow pulsanity was defined by use of the volumetric flow rate: Flow pulsatility = (maximum flow r a t e minimum flow rate)/mean flow rate. The Reynolds number, an indicator o f flow stability, was calculated for the distal graft by use of the standard definition: Reynolds number = 0Vd/IX where 0 = fluid density, V = mean fluid velocity, d = diameter, and tx = fluid viscosity. Fluid density and viscosity were calculated from the hematocrit.
Color Doppler studies. Color Doppler ultrasound imaging was performed with a Quantum QAD1 scanner (Quantum Medical Instruments, Inc., Issaquah, Wash.) with a 7.5 M H z transducer o n 10 animals between 2 and 12 weeks after graft implantation. Doppler studies were not possible in acute subjects because of the air trapped in porous PTFE grafts. Twenty-four bilateral scans were completed, although three dogs with acute thrombosis of one or both grafts had only one scan and were excluded from the study. Subjects were given intravenous 2.5% sodium thiamylal titrated to effect. Detailed data acquisition was performed after the examination by means of ultrasound postprocessing of real time data stored digitally on videotape. Studies included measurement of volumetric flow rate, flow velocity, and vessel diameter in the iliac artery, iliac vein, graft (three locations), femoral artery (superior and inferior to the anastomosis), and femoral vein (locations 1 through 5 in Fig. 2). Spectral signal patterns and color flow images were examined at each location for visual evidence of turbulence or flow disturbance. Color Doppler ultrasound imaging portrays movement as a color map by processing phase and frequency shift information from reflected Ultrasound. Perivascular tissue vibration at the venous anastomosis is thus demonstrated visually. A method o f quantitating tissue vibration was developed as a measure of energy transfer out of the vessel. This was measured by the distance required for signal attenuation (the distance to the limits of the vibration signal at standard threshold levels). The color Doppler system can "sum" the digital signals over several frames of videotape onto a single image, thus enhancing the reproducibility of measurements. Internal software allowed direct measurements with a resolution o f 0.1 mm on the resulting image. The volume of the vibration signal (VVS) was
558
Journal of VASCULAR SURGERY
Fillinger et al.
T a b l e I. H e m o d y n a m i c data Group I 6 mm std.
Implant VFR FP FVVel DGrVel DGrRe Sacrifice VFR FP FVVel DGVel DGRe Acute VFR Press (V2) Pg-Pv
Group I I 4-7 taper
6 mm std.
7-4 taper
1340 0.180 60.0 79.1 1580
-+ 140 - 0.03 ~- 5.6 -+ 8.3 -+ 170
1230 0.206 54.3 53.2 1240
+ 140 + 0.02 + 4.5 + 6.1 ~ _+ 140
1040 0.196 50.8 61.2 1220
+_ 120 -- 0.02 --- 5.9 --- 6.9 -+ 140
1030 0.172 51.1 137 1830
___ 120 ~ 0.02 +__6.8 z 16" ~ 210"
765 0.134 34.0 45.1 901
+ 200 + 0.02 -+ 8.6 -4- 12 + 240
691 0.114 30.7 29.9 698
-4- 90 + 0.03 --- 3.4 -+ 3.9 -+ 91
819 0.124 40.4 48.3 965
-+ 270 + 0.01 + 13 _+ 16 --_ 310
598 0.144 29.6 79.4 1060
___67 ___0.01 -+ 4.1 ___8.9* _ 120
1160 -+ 230 19.4 + 2.7 22.5 -4- 12
1050 _+ 80 7.2 -4- 1.I* 16.6 + 5.8
1120 +- 190 16.7 --- 2.4 25.2 -4- 15
980 --_ 74 5.0 ___ 1.3~ 68.2 ~ 7.5*
*p < 0.05 versus paired 6 mm untapered grafts. Values are reported as mean + SE. N = 5 for each column. Implant and sacrifice refer to hemodynamic values for long-term subjects. Acute refers to values for acute pressure measurement subjects. VFR, Volumetric flow rate (cc/min); FP, flow pulsatility; FVVeL, femoral vein velocity (cm/sec); DGVel, distal graft velocity (cm/sec); DGRe, distal graft Reynolds number; Press (V2), mean femoral vein pressure at location 2 (mm Hg); Pg-Pv, pressure drop across the venous anastom,..~fis (mm Hg). calculated f r o m measurements obtained at three locations in transverse and longitudinal views. Data were obtained at distances o f 0, 1, and 2 cm f r o m the venous anastomosis in b o t h views. I n the transverse view the distance required for signal attenuation was measured in the dorsal, ventral, medial, and lateral directions. T h e distance to attenuation was always measured at right angles to the vessel. Measurements in the dorsal and ventral directions were repeated in the longitudinal view to prevent misalignment errors. T h e V V S could then be calculated over the 2 cm vessel segment. V o l u m e s were calculated separately for signals dorsal and ventral to the vessel, because signals ventral to the vessel were sometimes "cut off°' at skin level and therefore had a different shape. This calculation was performed for segments o n b o t h sides o f the venous anastomosis. P r e s s u r e studies. T o avoid additional vein dissection or catheter-related intimal injury, pressure measurements were performed in acute animals only. T e n m o n g r e l dogs weighing 17 to 21 kg underwent placement o f bilateral grafts by means o f a technique identical to that o f chronic studies. After graft implantation and collection o f volumetric fl0w data, the abdominal cavity was entered, and a branch o f the iliac vein was used to insert a 1.5 m m diameter Millar pressure catheter (Millar Instruments, Inc., H o u s t o n , Texas) connected to a strip chart recorder. Pullback pressures were measured at 1 c m intervals t h r o u g h o u t each graft configuration. H i s t o l o g y / m o r p h o l o g y . Specimens were ob-
tained for histologic and m o r p h o m e t r i c studies in 10 animals before they were killed. T h e femoral vessels were gently excised in continuity with the prosthetic graft, o p e n e d along the vessel wall opposite the anastomosis, pinned to in vivo dimensions, and fixed in buffered 10% formalin. After gross inspection the vessels were subdivided at locations 1 t h r o u g h 5 as illustrated in Fig. 2. T h e resulting segments were processed, and two sections f r o m each location were stained with a modified iron-hematoxylin stain. Histologic sections were then subjected to qualitative analysis via light microscopy and examined for intimal-medial thickening.Quantitative measurements o f c o m b i n e d intimal-mediai thickness were oK tained by projecting histologic sections o n t o a screen at × 17.5 magnification. Outlines o f the c o m b i n e d intima and media were traced o n t o paper, and planimetry was performed o n a digitizing tablet (Numonics M o d e l 2400, N u m o n i c s Corp., Lansdale, Pa.). Average intimal-medial thickness was then calculated f r o m the digitized vessel dimensions for the entire cross section. Statistical analysis. Statistical analysis was performed with Statview I I software (Abacus Concepts, Inc., Berkeley, Calif.) o n a Macintosh IIx microcomputer (Apple C o m p u t e r , Inc., Cupertino, Calif.). Patency data were examined with a contingency table by chi-square analysis. Differences between hemodynamic variables in paired graft configurations were assessed by means o f Student t tests. Data are expressed as mean +-- standard error. Potential differ-
Volume 11 Number 4 April 1990
Graft geometry and intimal-medial hyperplasia 559
1
2
4
A
B
C Fig. 3. Doppler spectral tracings 12 weeks after implantation. 1, 2, and 4 refer to location numbers (Fig. 2). A specimens are from a 4 to 7 mm taper graft, B specimens are from a 6 mm untapered graft in the same animal as A, and C specimens are from a 7 to 4 rnm taper graft. ences in intimal-medial thickness were evaluated by use of a two-factor analysis of variance (ANOVA) with graft type and location as factors. An F statistic corresponding to p < 0.05 was considered significant. If the A N O V A revealed a significant difference, differences between groups were examined with Dunnett's test for multiple comparisons to a control. Potential correlations between hemodynamic variables and intimal-medial thickness at the venous anastomosis were evaluated with stepwise deletion regression analysis. Data are expressed as the correlation coefficient r and probability p for the regression. RESULTS Chronic hemodynamic studies. Both grafts remained patent in 10 of the 14 chronic subjects. One 6 mm graft, one 4 to 7 mm taper graft and three 7 to 4 mm taper grafts thrombosed acutely (less than 4 weeks after implant). There were no chronic occlusions. Patency data were not significant by chi-square analysis (p > 0.05). All animals with occluded grafts were excluded from the study. Hemodynamic measurements described below will refer to Table I for values at implantation and death unless o~ ~erwise specified. Volumetric flow rate, flow pulsatility, and mean femoral vein velocity were all statistically equivalent for the three graft types at implant. Flow velocity in the distal graft was significantly decreased in 4 to 7 mm taper grafts and increased in 7 to 4 mm taper
grafts in comparison to standard 6 mm untapered grafts. Reynolds number in the distal graft was also significantly increased in 7 to 4 mm taper grafts, indicating less stable flow. Mean Reynolds number was lowest in 4 to 7 mm taper grafts but this difference was not statistically significant (p > 0.05). Mean values for all hemodynamic parameters were lower at the time of death. The drop in volumetric flow rate from implant to death was not statistically different for tapered versus untapered graft types. There were fewer significant differences between graft types at the time of death, but trends were similar (Table I). Color Doppler studies. Spectral signals from color Doppler ultrasound studies demonstrated spectral broadening at all points distal to the arterial anastomosis in all three graft types. However, spectral signals appeared to differ in the femoral vein (Fig. 3). Dramatic flow disturbances with severe aliasing were frequently seen at locations 1 through 3. Aliasing (velocities causing frequency shifts exceeding the Nyquist limit) and velocities exceeding the dynamic range of the equipment often prevented quantitative estimates in these three locations, but flow disturbances decreased qualitatively as distance from the anastomosis increased. Flow velocities and the degree of disturbance were consistently much lower in the inferior vein (locations 4 and 5), even at locations immediately adjacent to the heel of the anastomosis.
560
Journal of VASCULAR SURGERY
Fillinger et al.
Fig. 4. Color Doppler images of tissue vibration surrounding the superior vein (including locations 1 through 3) at 12 weeks after implantation. A, vein from a 4 to 7 mm taper graft; B, vein from a 6 mm untapered graft in the same animal as A; C, vein from a 7 to 4 mm taper graft; D, vein from a 6 mm untapered graft in the same animal as C. Grafts are at the top right and flow is from right to left in each picture. Vibration signals obscure portions of the vein near the graft.
It was possible to demonstrate tissue vibration in vivo in every graft studied with color Doppler equipment (Fig. 4). The amount o f vibration was least in 4 to 7 m m taper grafts and approximately equal in untapered and 7 to 4 m m taper grafts. Voltune o f the vibration signal for the dorsal aspect o f the superior vein was 4.81 + 1.6 cc, for 6 m m untapered grafts, 1.58 + 1.2 cc for 4 to 7 m m taper grafts and 3.82 +_ 1.1 cc for 7 to 4 m m taper grafts.
N values were not sufficient for the paired t tests used for other hemodynamic data. Vibration was minimal or nonexistent around the inferior vein (VVS was less than 0.1 cc in all inferior veins). Pressure studies. In acute pressure measurement subjects, volumetric flow rates were not significantly different in the different graft types (p > 0.05, Table I). Mean flow rates were similar to those of chronic subjects, Pressure patterns in the three graft types
Volume 11 Number 4 April 1990
Graft geometry and intimal-medial hyperplasia 561 I~p//,,
Venouslimb Femoralv.
Arteriallimb
A
4-7mm Taper Graft Venouslimb
~ lliacv.
~
}'jls// Femoralv.
Venouslimb
_ Ill ~....
IIiaca./ /I/I/~[ r~,
graft
Arteriallimb
6rnm U n t a p e r e d Graft
/l G
graft
Arteriallimb 7-4mm Taper Graft
Fig. 5. Mean pessure values (ram Hg) for the three graft configurations.
were quite different, however, as shown in Fig. 5. There were large pressure drops across the arterial anastomosis in 6 mm untapered grafts (27.1 _+ 10.2 mm Hg) and 4 to 7 mm taper grafts (34.1 + 6.5 mm Hg, p > 0.05 vs untapered), but not in 7 to 4 mm tapers (4.7 + 7.4 mm Hg, p < 0.05 vs untapered). The mean pressure drop across the venous anastomosis was significantly higher in 7 to 4 mm taper grafts (Table I). Overall pressure drops from the ilia.c artery to the iliac vein were not statistically different for the three graft types (p > 0.05). In the femoral vein, superior venous pressure (locations 1 and 2) was significantly lower in both tapered graft types (Fig. 5, Table I). Despite large differences in mean values, inferior venous pressures (locations 4 and 5) were not significantly different for the different graft types. Inferior venous pressure was significantly higher than superior venous pressure in taper grafts (location 4 vs 2 and 5 vs 1, p < 0.05) but not in untapered grafts. In the iliac vein, mean pressure was slightly higher in untapered grafts, but this was not significant (p > 0.05). H i s t o l o g y / m o r p h o m e t r y . Histologic technique allowed examination o f the luminal surface at the time of death. Irregular opaque white lesions were often obvious to the-naked eye at locations 2 and 3 on the femoral vein, especially for untapered grafts. Graft geometry and location relative to the anastomosis also had an easily recognized effect on venous
histology. A representative histologic section is shown in Fig. 6. Sections taken from venous location 2 in 4 to 7 taper grafts had mild or no hyperplasia, whereas specimens from similar locations in 7 to 4 taper or 6 mm untapered grafts displayed the marked thickening, hypercellularity, and excessive collagen deposition characteristic of intimal-medial hyperplasia. The term intimal-mcdial hyperplasia is used because the boundary between intima and media is so poorly defined in venous specimens. Specimens from locations 1 and 3 generally displayed less hyperplasia than specimens at location 2. Sections from the inferior vein (locations 4 and 5) displayed little or no hyperplasia or thickening. Hyperplastic lesions were rare in arterial specimens, although small focal areas ofintimal (not medial) hyperplasia were occasionally seen. The above observations are supported by quantitative measurements of intimal-medial thickness. Results for intimal-medial thickness separated by graft type and femoral vein location are shown in Figs. 7 and 8. Two-factor A N O V A was significant for graft type (p < 0.001) and location (p < 0.01) for venous specimens in Group I (4 to 7 mm taper and 6 mm untapered grafts). In contrast, twofactor A N O V A was not significant for graft type (p > 0.05) but was significant for location of venous specimens (p < 0.001) in Group II (7 to 4 taper and 6 mm untapered grafts). Only venous specimens in
Journal of VASCULAR SURGERY
562 Fillinger et al.
f
Fig. 6. Histologic specimen from a 7 to 4 mm taper graft configuration at venous location 2, demonstrating severe intimal-medial hyperplasia three months after implantation. I-M, lntimamedia; Ad, adventitia. (Modified iron-hematoxylin stain; original magnification x 200.) 4 to 7 taper configurations were not significantly different from control vein (Fig. 7). For arterial specimens, two-factor A N O V A was not significant for graft type or location in either group I or II. Arterial thickness values were not significantly different from control at any location in either group I or II. Regression analysis. Stepwise deletion regression analysis was performed for~VVS (dorsal to the vein), volumetric flow rate, flow pulsatility, mean femoral vein flow velocity, distal graft flow velocity, and distal graft Reynolds number versus venous intimal-medial thickness at locations 3 and 5. Volume of the vibration signal had the strongest correlation with intimal-medial thickening (r = 0.92, r 2 = 0.84, p < 0.001). The remaining variables were all deleted by the regression program and therefore did not strengthen the regression by their inclusion. The regression was slightly less significant if the dorsal and ventral vibration signal volumes were combined into a single variable. If VVS was not included in the analysis, Reynolds number in the distal graft was the only variable left in the regression. This regres-
sion was significant but the correlation was weaker (r = 0.61, r 2 = 0.37,p < 0.01). DISCUSSION
This study indicates that graft geometry can ha a significant impact on hemodynamic parameters and venous intimal-medial hyperplasia in arteriovenous loop grafts. This has strong implications regarding graft design and provokes an important question. Is the relationship between hemodynamics and hyperplasia one of coincidence or causality? The results presented here suggest that flow disturbances cause kinetic energy transfer through the vessel wall and into perivascular tissue where it becomes manifest as tissue vibration. This energy transfer has been quantitated in vivo and correlates strongly with venous intimal-medial thickness. There is clearly a reproducible pattern to the development of venous hyperplasia in arteriovenous loop grafts as seen clinically, 3 in our previous study, 2 and in the present study. The characteristic lesion occurs at the 'toe' of the anastomosis (location 2)
Volume 11 Number 4 April 1990
Grafi geometry and intimal-medial hyperplasia 563
0.30 *
E
0.20
p < 0.05 vs Control
[]
Control
[]
4-7 mm Taper
[]
6 mm Untapared paired with 4-7's
0.10
0.00 Cntrl Ufl
Uf2
Uf3
Uf4
U15
F1
F2
F3
F4
F5
Location on Femoral Vein
Fig. 7. Venous intimal-medial thickness in 4 to 7 mm taper grafts (F1-F5) and paired 6 mm untapered grafts (Ufl-Uf5), mean + SE. Cntrl, Normal control vein; numbers 1 through 5 refer to locations I through 5 (see Fig. 2).
0.30 *
0.20
0.10
0.00
/I iii Cntrl Us1
Us2 Us3
Us4 Us5
$1
$2
$3
$4
p < 0.05 vS Control
[]
Control
[]
7-4 mm Taper
[]
6 mm Untapered paired with 7-4's
$5
Location on Femoral Vein
Fig. 8. Venous intimal-medial thickness in 7 to 4 mm taper grafts (S1-$5) and paired 6 nun tmtapered grafts (Usl-Us5), mean _+ SE. Cntrl, Normal control vein; numbers 1 through 5 refer to locations 1 through 5 (see Fig. 2). with minimal to no change in vein, that is only a short distance away, including the heel of the anastomosis. Thus the detailed evaluation of hemodynamics in these locations will produce strong evidence regarding causation. While emphasizing local hemodynamics it is important to note that the bilateral implantation protocol ensures that systemic parameters (e.g., cardiac output, mean arterial blood pressure, metabolic and coagulation factors) are the same for each pair of grafts. Graft material, suture, and implantation technique were also identical for all grafts. Differences between graft types therefore iso-
lated the effect of graft geometry on local hemodynamic and mechanical parameters. It is commonly believed that a 4 to 7 mm taper graft will have a lower volumetric flow rate than a 6 mm untapered graft in the same location, but that was not the case in this study. Since most femoral arteries in the study were 4 lama in diameter it is likely that the 4 mm anastomosis was rarely more flow limiting than the 6 mm anastomosis, and pressure data confirm this. This underscores the importance of analyzing an entire flow system before assuming the impact o f a certain graft geometry. The similarity
564
Journal of VASCULAR SURGERY
Fillinger et al.
of flow rates for the three graft types also indicates that alteration of volumetric flow rate as such is not likely responsible for the differing degrees ofintimalmedial thickening. An interesting contrast is thus provided by our previous work. 2 In that study 6 mm untapered grafts were implanted bilaterally and a flow-limiting band was placed on one of the grafts so flow rates differed but anastomotic geometry was identical. We used Reynolds number as a quantitative measure of turbulence confirmed qualitatively by phonoangiography of exposed vessels (color Doppler technology was not yet available). Reynolds number was found to correlate strongly with intimal-medial thickening. In the present study geometries differ but flow rates do not. With differing geometries one cannot assume a simple correlation between preanastomotic Reynolds number and the degree of turbulence, but this parameter is nonetheless a useful indicator of flow stability. Higher numbers indicate less stable flow and Reynolds numbers >2300 indicate fully developed turbulence is likely. The range of Reynolds numbers in this study indicate flow disturbance is almost certain at an irregularity such as a distal end-to-side anastomosis, but that it will likely dampen out as flow moves downstream. Spectral tracings from in vivo Doppler studies were consistent with this analysis. Flow disturbances at locations 2 and 3 are often violent, but the disturbances progressively lessen as flow moves away from the anastomosis. Aliasing prevented the use of spectral signals to quantitate turbulence, but qualitatively disturbances appeared to be least severe in 4 to 7 mm taper grafts (Fig. 3). There was no obvious difference between untapered and 7 to 4 mm taper grafts in terms of postanastomotic turbulence. It is interesting that if vibration data are not used in the regression analysis, distal graft (preanastomotic) Reynolds number has the best correlation with venous intimal-medial thickening. Although the correlation is not as strong as in our previous work, this is not surprising with the complicating factor of differing anastomotic geometries. Flow disturbances were dramatically lower at locations 4 and 5 for all three graft types. Tissue vibration and intimal-medial hyperplasia was also dramatically lower at these locations. Thus the proposed relationship between turbulence, energy transfer through the vessel wall, and hyperplasia holds not only across different graft types, but also holds consistently for different locations within the vein itself. The vibration of perivascular tissue requires energy. Some component of the hemodynamic forces
within the lumen must be exerted at right angles to the vessel wall to set up the observed vibration patterns. Of the hemodynamic parameters studied, three components are logical candidates for such forces. The pulsatile nature of the flow is an obvious choice, but the high frequency "thrill" of an arteriovenous fistula is much higher than the frequency of pressure oscillations in pulsatile flow. Doppler technology indicated frequencies on the order of 1000 to 3000 Hz for the perivascular vibrations studied here. The second and third candidates are components of velocity--the mean component, which will have vectors at right angles to the vessel wall in disturbed flow of any kind, and a fluctuating component superimposed on the mean velocity. To be precise, the mean velocity does fluctuate, but it does so in an easily described, repeatable manner. The "fluctuating component" referred to here is a randomly fluctuating component that occurs only with actual turbulence. 4 When examining the hemodynamic fo:-%es in this way it seems clear that only the high freqhency, randomly fluctuating forces related to true turbulence could cause the vibrations observed in this study. It also seems likely that these forces would cause abnormal wall stresses and chronic vascular injury. It is interesting to note that Nathan and Imparato ~ also detected high frequency (1000 to 2000 Hz) signals using phonoangiography with frequency analysis in an arteriovenous model of intimal hyperplasia. In their study symmetric lesions were produced in preparations producing only low frequency signals, but they noted an association of high frequency signals at implant with a characteristic asymmetric lesion. They felt the high frequency signals resulted from turbulent flow and hypothesized that a relationship existed between intimal hyperplasia and the f~ces responsible for the high energy signals. Acute tissue injury is a known potentiator of intimal hyperplasia. 6"7 Cyclic stretch has also been implicated in stimulation of cell growth via biologic mediators. 8 The work of Fry 9 demonstrated acute cell damage in the face of either extremely high shear stress or turbulence, and the Reynolds numbers of this model are similar. Our previous study related turbulence to the development of hyperplasia, and we postulated that the relationship was one of chronic vessel injury and stress? The present study demonstrates in vivo that rapidly oscillating stresses and forces that could induce chronic vascular injury occur at the precise locations where hyperplasia occurs.
The present results must also be reviewed in terms of other prominent theories regarding i.nti-
Volume 1 l Number 4 April 1990
f
" •
mal hyperplasia. When discussing turbulence, onc must bring up theories regarding boundary layer separation ~° and low shear stress. 11 Boundary layer scparation is a phcnomenon in which flow streamlincs move in a regular, definable pattern that is not parallel to the vessel wall (unlike simple laminar flow). Flow with Reynolds numbers in the 500 to 2000 range combined with the geomctry of distal end-to-side anastomoses certainly favors such phcnomcna. Advcrse pressure gradients were also found in thc femoral vein for all graft types (Fig. 5). Color flow patterns did revcal somc areas of rcgularly repcating flow disturbances that wcre consistent with boundary layer separation, but thcse wcre found within the graft or in the inferior vein (locations 4 and 5) rather than at locations 1 and 2 on the vcin where hyperplasia is prominent. The patterns at locations 1 and 2 were chaotic and random, which is more indicative of actual turbulence than boundary layer ~¢paration. In addition, the relcase of energy in the form of tissuc vibration is further indication of a more scvere form of flow disturbance or actual turbulence. It has been noted that high shear and low shear are likely to coexist in areas of flow disturbance. ~2 Low flow vclocities (and thereforc low shcar stresscs) were common in thc infcrior vein but were rare or nonexistent at vein locations 1 through 3. The lowest flow velocities at locations 1 through 3 were observed in veins of the 4 to 7 mm taper configuration where hyperplasia was least severe. Thcse obscrvations point toward high shcar and turbulence as causcs for hyperplasia rathcr than low shear and the milder disturbance of boundary layer scparation. This does not necessarily imply that past theories are incorrect. It is q'-:,te possible that the cause of hyperplasia in the high flow milieu of arteriovenous loop grafts is different from that of arterial bypass grafts. If the results presented here do not strongly support low shcar as the causative factor in this form of hypcrplasia, what of high shear? Is this type of hypcrplasia simply the result of a high velocity "jet" at the vcnous anastomosis, with the highest velocities producing the most severe response? Although this is possible, observations do not strictly support this theory either. The highest flow velocities in the distal graft arc found in the 7 to 4 mm taper configuration, but the intimal-medial thickening was no worse than that of veins in untapered 6 mm graft configurations. One might suggest that anastomotic geometry is a confounding factor, but that would also argue against a simple "jet" effect. Furthermore, flow velocity in the distal graft did not correlate with intimal-
Graft geometry and intimal-medial hyperplasia 565
medial thickening as well as energy transfer or Reynolds number--again providing evidence against a simple "jet" theory. There is evidence that tangential pressure stress may be at least partially responsible for the development of intimal hyperplasia in veins subjected to elevated (arterial) pressures. 13,14Close examination of the pressure patterns found here (Fig. 5) would seem to argue against this theory for venous hyperplasia in arteriovenous loop grafts. Femoral venous pressures at location 2 were significantly increased in untapered grafts but not in 7 to 4 mm taper grafts. Also, femoral venous pressure was significantly higher in the inferior portions (locations 4 and 5) of tapered grafts where no significant hyperplasia was found. These patterns of pressure argue against tangential pressure stress as a major factor in this form of hyperplasia. Finally, it has also been suggested that kinetic energy transfer plays a role in the hyperplasia associated with compliance mismatch, is but compliance mismatch was not altered in the present study. Although the relationship between altered geometry, hemodynamics and histologic abnormalities was postulated 30 years ago, 16 technology has advanced to the point that hemodynamic mechanisms can be proved. The implication is that grafts can be designed to produce more favorable hemodynamic patterns and thereby reduce the development of venous intimal hyperplasia, at least in arteriovenous loop grafts. Current research emphasizes graft materials and surface phenomena such as platelet-graft interactions and endothelial cell seeding. Although these are indeed important areas, graft geometry can be altered cheaply and easily with currently available technology. Efforts in this area may well prove to be of value until breakthroughs are achieved in other areas of hyperplasia research. We conclude that graft geometry can significantly affect hemodynamic patterns and venous intimalmedial hyperplasia in arteriovenous loop grafts. Flow disturbances appear to cause kinetic energy transfer through the vessel wall and into perivascutar tissue. Kinetic energy transfer in the form of perivascular tissue vibration has been quantitatcd in vivo and correlates strongly with venous intimal-medial thickness. The postulated relationship is as follows: graft geometry affects several hemodynamic factors, of which flow stability (Reynolds number) is the most important. Flow stability determines the degree to which flow disturbance or turbulence occurs at irregularities such as the venous anastomosis. The larger the disturbance at the anastomosis, the greater
566
Journal of VASCULAR SURGERY
Fillinger et al.
the release of energy. The forces that cause perivascular energy transfer through the vessel wall also cause chronic stress and damage, triggering biochemical mediators and ultimately resulting in the development of intimal-medial hyperplasia. REFERENCES 1. Rizzuti RP, Hale JC, Burkart TE. Extended patency of expanded polytetrafluoroethylene grafts for vascular access using optimal configuration and revisions. Surg Gynecol Obstet 1988;166:23-7. 2. Fillinger MF, Reinitz ER, Schwartz RA, Resetarits DE, Paskanik AM, Bredenberg CE. Beneficial effects of banding on venous intimal-medial hyperplasia i n arteriovenous loop grafts. Am J Surg 1989;158:87-94. 3. Sottiurai VS, Yao JST, Flinn WR, Batson RC. Intimal hyperplasia and neointima: an ultrastructural analysis of thromhosed grafts in humans. Surgery 1983;93:809-17. 4. Whitmore RL. The flow of fluids. In: Whitmore RL, ed. Rheology of the circulation. 1st ed. New York: Pergamon Press, 1968:37-9. 5. Nathan I, Imparato AM. Vibration analysis in experimental models of atherosclerosis. Bull NY Acad Med 1977;53:84968. 6. Clowes AW. Pathobiology of arterial healing. In: Strandness Jr DE, Didisheim P, Clowes AW, Watson JT, eds. Vascular diseases: current research and clinical applications. 1st ed. Orlando: Grune & Stratton, 1987:351-62. 7. LoGeffo FW, Quist w e , Cantelmo NL, Haudenschild CC.
8.
9.
10.
11.
12.
13.
14.
15.
16.
Integrity of vein grafts as a function of initial intima, and medial preservation. Circulation 1983;68(suppl II): 117-24. Sumpio BE, Banes AJ, Levin LG, Johnson G. Mechanical stress stimulates aortic endothelial cells to proliferate. J VASe SURG 1987;6:252-6. Fry DL. Acute vascular endothelial changes associated with increased blood velocity gradients. Circ Res i968;22:16S97. Crawshaw HM, Quist WC, Serrallach E~ Valeri CR, LoGerfo FW. Flow disturbance at the distal end-to-side anastomosis. Arch Surg 1980; 115:1280-4. Rittgers SE, Karayannacos PE, Guy JF, Nerem RM, Shaw GM, Hostetler JR, Vasko JS. Velocity distribution and intimal proliferation in autologous vein grafts in dogs. Circ Res 1978;42:792-801. Nerem RM, Levesque MJ. Hemodynamics and the arterial wall. In: Strandness Jr DE, Didisheim P, Clowes AW, Watson JT, eds. Vascular diseases: current research and clinical applications. 1st ed. Orlando: Grune & Stratton, 1987:296-9. Kohler TR, Kirkman TR, Clowes AW. The effect of rigid external support on vein graft adaptation to the arterial circulation. J VASC SURG 1989;9:277-85. Karayannacos PE, Hostetler JR, Bond MG, et al. Late failure in vein grafts: mediating factors in subendothelial fi,~¢~muscular hyperplasia. Ann Surg 1978; 187:183-8. Suggs WE), Henriques HF, DePalma RG. Vein cuff interposition prevents juxta-anastomotic neointimal hyperplasia. Ann Surg 1988;207:717-23. Imparato AM, Lord JW, Texon M, Helpern M. Experimental atherosclerosis produced by alteration of blood vessel configuration. Surg Forum 1961;12:245-7.
Dennis E. Resetarits, MD, Andrew M. Paskanik, David Bruch, MS, and Carl E. Bredenberg, MD, Syracuse, N.Y. This study explores graft geometry and hemodynamics in a reproducible canine arteriovenous loop graft model o f intimal-medial hyperplasia. Untapered 6 m m diameter polytetrafluoroethylene grafts (n = 10) were paired with 4 to 7 m m taper (n = 5) or 7 to 4 nun taper (n = 5) grafts for a 12-week period. Several hemodynamic variables were assessed at multiple locations, and venous intimal-medial thickness was measured at locations corresponding to the hemodynamic measurements. Color Doppler imaging demonstrated energy transfer out o f the vessel in the form of perivascular tissue vibration. This was quantitated by the distance required for Doppler signal attenuation or volume o f the detected vibration signal. Differences among graft types were noted for pressure, flow velocity, tissue vibration, and venous intimal-medial thickness. Hyperplasia was significantly decreased in 4 to 7 m m taper grafts. Stepwise deletion regression indicated volume o f the vibration signal had a better correlation with venous intimal-medial thickness than any other variable (r 0.9,p < 0.001). We conclude that graft geometry can have a significant impact on hemodynamic factors and venous intimal-medial hyperplasia in arteriovenous loop grafts. Flow disturbances appear to cause energy transfer through the vessel wall and into perivascular tissue. Kinetic energy transfer in the form ofperivasoalar tissue vibration was quantitated in vivo and correlates strongly with venous intimal-medial thickness. (J VAsc SURG 1990;11:556-66.)
Prosthetic arteriovenous grafts are more prone to failure as a result of anastomotic intimal hyperplasia than any other type of vascular graft. In polytetrafluoroethylene (PTFE) hemodialysis access grafts hyperplasia causes more than 50% of all failures, and graft patency at i year is less than 80%. 1 Numerous hemodynamic theories have been proposed to explain this phenomenon, but relatively few in vivo studies have been done. In a previous study we demonstrated a quantitative relationship between Reynolds number (a parameter related to flow stability and turbulence) and venous intimal-medial thickening in arteriovenous loop grafts? In that study turbulence was decreased by placing a flow-limiting band on 6 mm diameter grafts. However, in an ideal graft, flow disturbances would be minimized without reducing flow or requiring the creation of a severe graft stenosis. Toward that end, this study explores From the Department of Surgery, State University of New York Health Science Center, Syracuse. Presented at the Third Annual Meeting of the Eastern Vascular Society, Bermuda, May 4-7, 1989. Reprint requests: Robert A. Schwartz, MD, Department of Surgery, State University of New York Health Science Center, 750 E. Adams St., Syracuse, NY 13210. 24/6/18322
556
the hemodynamic patterns created by three different graft geometries and the resulting impact of these hemodynamic patterns on intimal hyperplasia. The noninvasive study of in vivo flow disturbance and energy related phenomena was made possible by the use of color Doppler ultrasound imaging technology. MATERIAL AND M E T H O D S Study design. Grafts were implanted bilaterall) in a paired fashion that allowed for a side-by-side comparison of three different graft geometries. A standard untapered 6 mm diameter graft was always paired with either a 7 to 4 nun taper graft (7 mm diameter at the arterial anastomosis, 4 mm at the distal venous anastomosis) or a 4 to 7 mm taper graft (4 mm at the arterial anastomosis) as shown in Fig. 1. Animal care complied at all times with the "Principles of Laboratory Animal Care" and the "Guide for the Care and Use of Laboratory Animals" (NIH Publication No. 80-23, revised 1985). Implantation protocol. Fourteen mongrel dogs weighing 17 to 23 kg were anesthetized with intravenous sodium pentobarbital (30 mg/kg), and bilateral femoral arteriovenous loop grafts were implanted with pairings as described in the study design section. All grafts were 25 cm in length and made of
Volume 11 Number 4 April 1990
Graft geometry and intimal-medial hyperplasia 557
i lilacvein..~ Femoralv e i n ~
/
I ~ . . l I l i a c artery ~ I n g u i n a l ligament ~ . ~ F / . ~ ~ ~-Femoralartery
~ - -
Graft
Vessel I
#1
#2 #3 #4
#5
LocationNumbers Fig. 2. Subdivision of vessels for histology. Note the location numbers for orientation in the text and other figures.
4-7 mm Taper Graft
6mm Untapered Graft
Fig. 1. Implant configuration. One side is always a 6 mm untapered graft. expanded PTFE (Gore-Tex, registered trademark of W. L. Gore & Assoc., Elkton, Md). After mobilization of the femoral artery and vein, internal vessel diameter was gently measured by use of coronary dilators. An end-graft to side-artery anastomosis was then created with a continuous suture of 6-0 polypropylene. The graft was tunneled in a subcutaneous loop before end-graft to side-vein anastomosis. Graft implantation on the opposite side was carried out in the same fashion. Local heparin flushing was the Only anticoagulation used. Prophylactic antibiotics were given up to 5 days after operation. The grafts were left in place for a period of 12 weeks. Graft patency was checked frequently with a portable, continuouswave 5 M H z Doppler transducer. Chronic hemodynamic measurements. After implantation of both grafts, systolic, diastolic, and mean volumetric flow rates were recorded with an electromagnetic flowmeter (Carolina Medical Electronics, King, N.C.). These measurements were repeated in anesthetized animals at the time they were killed. The Collected data allowed for the calculation of flow pulsatility, mean flow velocity, and Reynolds number. For the purposes of this study, flow pulsanity was defined by use of the volumetric flow rate: Flow pulsatility = (maximum flow r a t e minimum flow rate)/mean flow rate. The Reynolds number, an indicator o f flow stability, was calculated for the distal graft by use of the standard definition: Reynolds number = 0Vd/IX where 0 = fluid density, V = mean fluid velocity, d = diameter, and tx = fluid viscosity. Fluid density and viscosity were calculated from the hematocrit.
Color Doppler studies. Color Doppler ultrasound imaging was performed with a Quantum QAD1 scanner (Quantum Medical Instruments, Inc., Issaquah, Wash.) with a 7.5 M H z transducer o n 10 animals between 2 and 12 weeks after graft implantation. Doppler studies were not possible in acute subjects because of the air trapped in porous PTFE grafts. Twenty-four bilateral scans were completed, although three dogs with acute thrombosis of one or both grafts had only one scan and were excluded from the study. Subjects were given intravenous 2.5% sodium thiamylal titrated to effect. Detailed data acquisition was performed after the examination by means of ultrasound postprocessing of real time data stored digitally on videotape. Studies included measurement of volumetric flow rate, flow velocity, and vessel diameter in the iliac artery, iliac vein, graft (three locations), femoral artery (superior and inferior to the anastomosis), and femoral vein (locations 1 through 5 in Fig. 2). Spectral signal patterns and color flow images were examined at each location for visual evidence of turbulence or flow disturbance. Color Doppler ultrasound imaging portrays movement as a color map by processing phase and frequency shift information from reflected Ultrasound. Perivascular tissue vibration at the venous anastomosis is thus demonstrated visually. A method o f quantitating tissue vibration was developed as a measure of energy transfer out of the vessel. This was measured by the distance required for signal attenuation (the distance to the limits of the vibration signal at standard threshold levels). The color Doppler system can "sum" the digital signals over several frames of videotape onto a single image, thus enhancing the reproducibility of measurements. Internal software allowed direct measurements with a resolution o f 0.1 mm on the resulting image. The volume of the vibration signal (VVS) was
558
Journal of VASCULAR SURGERY
Fillinger et al.
T a b l e I. H e m o d y n a m i c data Group I 6 mm std.
Implant VFR FP FVVel DGrVel DGrRe Sacrifice VFR FP FVVel DGVel DGRe Acute VFR Press (V2) Pg-Pv
Group I I 4-7 taper
6 mm std.
7-4 taper
1340 0.180 60.0 79.1 1580
-+ 140 - 0.03 ~- 5.6 -+ 8.3 -+ 170
1230 0.206 54.3 53.2 1240
+ 140 + 0.02 + 4.5 + 6.1 ~ _+ 140
1040 0.196 50.8 61.2 1220
+_ 120 -- 0.02 --- 5.9 --- 6.9 -+ 140
1030 0.172 51.1 137 1830
___ 120 ~ 0.02 +__6.8 z 16" ~ 210"
765 0.134 34.0 45.1 901
+ 200 + 0.02 -+ 8.6 -4- 12 + 240
691 0.114 30.7 29.9 698
-4- 90 + 0.03 --- 3.4 -+ 3.9 -+ 91
819 0.124 40.4 48.3 965
-+ 270 + 0.01 + 13 _+ 16 --_ 310
598 0.144 29.6 79.4 1060
___67 ___0.01 -+ 4.1 ___8.9* _ 120
1160 -+ 230 19.4 + 2.7 22.5 -4- 12
1050 _+ 80 7.2 -4- 1.I* 16.6 + 5.8
1120 +- 190 16.7 --- 2.4 25.2 -4- 15
980 --_ 74 5.0 ___ 1.3~ 68.2 ~ 7.5*
*p < 0.05 versus paired 6 mm untapered grafts. Values are reported as mean + SE. N = 5 for each column. Implant and sacrifice refer to hemodynamic values for long-term subjects. Acute refers to values for acute pressure measurement subjects. VFR, Volumetric flow rate (cc/min); FP, flow pulsatility; FVVeL, femoral vein velocity (cm/sec); DGVel, distal graft velocity (cm/sec); DGRe, distal graft Reynolds number; Press (V2), mean femoral vein pressure at location 2 (mm Hg); Pg-Pv, pressure drop across the venous anastom,..~fis (mm Hg). calculated f r o m measurements obtained at three locations in transverse and longitudinal views. Data were obtained at distances o f 0, 1, and 2 cm f r o m the venous anastomosis in b o t h views. I n the transverse view the distance required for signal attenuation was measured in the dorsal, ventral, medial, and lateral directions. T h e distance to attenuation was always measured at right angles to the vessel. Measurements in the dorsal and ventral directions were repeated in the longitudinal view to prevent misalignment errors. T h e V V S could then be calculated over the 2 cm vessel segment. V o l u m e s were calculated separately for signals dorsal and ventral to the vessel, because signals ventral to the vessel were sometimes "cut off°' at skin level and therefore had a different shape. This calculation was performed for segments o n b o t h sides o f the venous anastomosis. P r e s s u r e studies. T o avoid additional vein dissection or catheter-related intimal injury, pressure measurements were performed in acute animals only. T e n m o n g r e l dogs weighing 17 to 21 kg underwent placement o f bilateral grafts by means o f a technique identical to that o f chronic studies. After graft implantation and collection o f volumetric fl0w data, the abdominal cavity was entered, and a branch o f the iliac vein was used to insert a 1.5 m m diameter Millar pressure catheter (Millar Instruments, Inc., H o u s t o n , Texas) connected to a strip chart recorder. Pullback pressures were measured at 1 c m intervals t h r o u g h o u t each graft configuration. H i s t o l o g y / m o r p h o l o g y . Specimens were ob-
tained for histologic and m o r p h o m e t r i c studies in 10 animals before they were killed. T h e femoral vessels were gently excised in continuity with the prosthetic graft, o p e n e d along the vessel wall opposite the anastomosis, pinned to in vivo dimensions, and fixed in buffered 10% formalin. After gross inspection the vessels were subdivided at locations 1 t h r o u g h 5 as illustrated in Fig. 2. T h e resulting segments were processed, and two sections f r o m each location were stained with a modified iron-hematoxylin stain. Histologic sections were then subjected to qualitative analysis via light microscopy and examined for intimal-medial thickening.Quantitative measurements o f c o m b i n e d intimal-mediai thickness were oK tained by projecting histologic sections o n t o a screen at × 17.5 magnification. Outlines o f the c o m b i n e d intima and media were traced o n t o paper, and planimetry was performed o n a digitizing tablet (Numonics M o d e l 2400, N u m o n i c s Corp., Lansdale, Pa.). Average intimal-medial thickness was then calculated f r o m the digitized vessel dimensions for the entire cross section. Statistical analysis. Statistical analysis was performed with Statview I I software (Abacus Concepts, Inc., Berkeley, Calif.) o n a Macintosh IIx microcomputer (Apple C o m p u t e r , Inc., Cupertino, Calif.). Patency data were examined with a contingency table by chi-square analysis. Differences between hemodynamic variables in paired graft configurations were assessed by means o f Student t tests. Data are expressed as mean +-- standard error. Potential differ-
Volume 11 Number 4 April 1990
Graft geometry and intimal-medial hyperplasia 559
1
2
4
A
B
C Fig. 3. Doppler spectral tracings 12 weeks after implantation. 1, 2, and 4 refer to location numbers (Fig. 2). A specimens are from a 4 to 7 mm taper graft, B specimens are from a 6 mm untapered graft in the same animal as A, and C specimens are from a 7 to 4 rnm taper graft. ences in intimal-medial thickness were evaluated by use of a two-factor analysis of variance (ANOVA) with graft type and location as factors. An F statistic corresponding to p < 0.05 was considered significant. If the A N O V A revealed a significant difference, differences between groups were examined with Dunnett's test for multiple comparisons to a control. Potential correlations between hemodynamic variables and intimal-medial thickness at the venous anastomosis were evaluated with stepwise deletion regression analysis. Data are expressed as the correlation coefficient r and probability p for the regression. RESULTS Chronic hemodynamic studies. Both grafts remained patent in 10 of the 14 chronic subjects. One 6 mm graft, one 4 to 7 mm taper graft and three 7 to 4 mm taper grafts thrombosed acutely (less than 4 weeks after implant). There were no chronic occlusions. Patency data were not significant by chi-square analysis (p > 0.05). All animals with occluded grafts were excluded from the study. Hemodynamic measurements described below will refer to Table I for values at implantation and death unless o~ ~erwise specified. Volumetric flow rate, flow pulsatility, and mean femoral vein velocity were all statistically equivalent for the three graft types at implant. Flow velocity in the distal graft was significantly decreased in 4 to 7 mm taper grafts and increased in 7 to 4 mm taper
grafts in comparison to standard 6 mm untapered grafts. Reynolds number in the distal graft was also significantly increased in 7 to 4 mm taper grafts, indicating less stable flow. Mean Reynolds number was lowest in 4 to 7 mm taper grafts but this difference was not statistically significant (p > 0.05). Mean values for all hemodynamic parameters were lower at the time of death. The drop in volumetric flow rate from implant to death was not statistically different for tapered versus untapered graft types. There were fewer significant differences between graft types at the time of death, but trends were similar (Table I). Color Doppler studies. Spectral signals from color Doppler ultrasound studies demonstrated spectral broadening at all points distal to the arterial anastomosis in all three graft types. However, spectral signals appeared to differ in the femoral vein (Fig. 3). Dramatic flow disturbances with severe aliasing were frequently seen at locations 1 through 3. Aliasing (velocities causing frequency shifts exceeding the Nyquist limit) and velocities exceeding the dynamic range of the equipment often prevented quantitative estimates in these three locations, but flow disturbances decreased qualitatively as distance from the anastomosis increased. Flow velocities and the degree of disturbance were consistently much lower in the inferior vein (locations 4 and 5), even at locations immediately adjacent to the heel of the anastomosis.
560
Journal of VASCULAR SURGERY
Fillinger et al.
Fig. 4. Color Doppler images of tissue vibration surrounding the superior vein (including locations 1 through 3) at 12 weeks after implantation. A, vein from a 4 to 7 mm taper graft; B, vein from a 6 mm untapered graft in the same animal as A; C, vein from a 7 to 4 mm taper graft; D, vein from a 6 mm untapered graft in the same animal as C. Grafts are at the top right and flow is from right to left in each picture. Vibration signals obscure portions of the vein near the graft.
It was possible to demonstrate tissue vibration in vivo in every graft studied with color Doppler equipment (Fig. 4). The amount o f vibration was least in 4 to 7 m m taper grafts and approximately equal in untapered and 7 to 4 m m taper grafts. Voltune o f the vibration signal for the dorsal aspect o f the superior vein was 4.81 + 1.6 cc, for 6 m m untapered grafts, 1.58 + 1.2 cc for 4 to 7 m m taper grafts and 3.82 +_ 1.1 cc for 7 to 4 m m taper grafts.
N values were not sufficient for the paired t tests used for other hemodynamic data. Vibration was minimal or nonexistent around the inferior vein (VVS was less than 0.1 cc in all inferior veins). Pressure studies. In acute pressure measurement subjects, volumetric flow rates were not significantly different in the different graft types (p > 0.05, Table I). Mean flow rates were similar to those of chronic subjects, Pressure patterns in the three graft types
Volume 11 Number 4 April 1990
Graft geometry and intimal-medial hyperplasia 561 I~p//,,
Venouslimb Femoralv.
Arteriallimb
A
4-7mm Taper Graft Venouslimb
~ lliacv.
~
}'jls// Femoralv.
Venouslimb
_ Ill ~....
IIiaca./ /I/I/~[ r~,
graft
Arteriallimb
6rnm U n t a p e r e d Graft
/l G
graft
Arteriallimb 7-4mm Taper Graft
Fig. 5. Mean pessure values (ram Hg) for the three graft configurations.
were quite different, however, as shown in Fig. 5. There were large pressure drops across the arterial anastomosis in 6 mm untapered grafts (27.1 _+ 10.2 mm Hg) and 4 to 7 mm taper grafts (34.1 + 6.5 mm Hg, p > 0.05 vs untapered), but not in 7 to 4 mm tapers (4.7 + 7.4 mm Hg, p < 0.05 vs untapered). The mean pressure drop across the venous anastomosis was significantly higher in 7 to 4 mm taper grafts (Table I). Overall pressure drops from the ilia.c artery to the iliac vein were not statistically different for the three graft types (p > 0.05). In the femoral vein, superior venous pressure (locations 1 and 2) was significantly lower in both tapered graft types (Fig. 5, Table I). Despite large differences in mean values, inferior venous pressures (locations 4 and 5) were not significantly different for the different graft types. Inferior venous pressure was significantly higher than superior venous pressure in taper grafts (location 4 vs 2 and 5 vs 1, p < 0.05) but not in untapered grafts. In the iliac vein, mean pressure was slightly higher in untapered grafts, but this was not significant (p > 0.05). H i s t o l o g y / m o r p h o m e t r y . Histologic technique allowed examination o f the luminal surface at the time of death. Irregular opaque white lesions were often obvious to the-naked eye at locations 2 and 3 on the femoral vein, especially for untapered grafts. Graft geometry and location relative to the anastomosis also had an easily recognized effect on venous
histology. A representative histologic section is shown in Fig. 6. Sections taken from venous location 2 in 4 to 7 taper grafts had mild or no hyperplasia, whereas specimens from similar locations in 7 to 4 taper or 6 mm untapered grafts displayed the marked thickening, hypercellularity, and excessive collagen deposition characteristic of intimal-medial hyperplasia. The term intimal-mcdial hyperplasia is used because the boundary between intima and media is so poorly defined in venous specimens. Specimens from locations 1 and 3 generally displayed less hyperplasia than specimens at location 2. Sections from the inferior vein (locations 4 and 5) displayed little or no hyperplasia or thickening. Hyperplastic lesions were rare in arterial specimens, although small focal areas ofintimal (not medial) hyperplasia were occasionally seen. The above observations are supported by quantitative measurements of intimal-medial thickness. Results for intimal-medial thickness separated by graft type and femoral vein location are shown in Figs. 7 and 8. Two-factor A N O V A was significant for graft type (p < 0.001) and location (p < 0.01) for venous specimens in Group I (4 to 7 mm taper and 6 mm untapered grafts). In contrast, twofactor A N O V A was not significant for graft type (p > 0.05) but was significant for location of venous specimens (p < 0.001) in Group II (7 to 4 taper and 6 mm untapered grafts). Only venous specimens in
Journal of VASCULAR SURGERY
562 Fillinger et al.
f
Fig. 6. Histologic specimen from a 7 to 4 mm taper graft configuration at venous location 2, demonstrating severe intimal-medial hyperplasia three months after implantation. I-M, lntimamedia; Ad, adventitia. (Modified iron-hematoxylin stain; original magnification x 200.) 4 to 7 taper configurations were not significantly different from control vein (Fig. 7). For arterial specimens, two-factor A N O V A was not significant for graft type or location in either group I or II. Arterial thickness values were not significantly different from control at any location in either group I or II. Regression analysis. Stepwise deletion regression analysis was performed for~VVS (dorsal to the vein), volumetric flow rate, flow pulsatility, mean femoral vein flow velocity, distal graft flow velocity, and distal graft Reynolds number versus venous intimal-medial thickness at locations 3 and 5. Volume of the vibration signal had the strongest correlation with intimal-medial thickening (r = 0.92, r 2 = 0.84, p < 0.001). The remaining variables were all deleted by the regression program and therefore did not strengthen the regression by their inclusion. The regression was slightly less significant if the dorsal and ventral vibration signal volumes were combined into a single variable. If VVS was not included in the analysis, Reynolds number in the distal graft was the only variable left in the regression. This regres-
sion was significant but the correlation was weaker (r = 0.61, r 2 = 0.37,p < 0.01). DISCUSSION
This study indicates that graft geometry can ha a significant impact on hemodynamic parameters and venous intimal-medial hyperplasia in arteriovenous loop grafts. This has strong implications regarding graft design and provokes an important question. Is the relationship between hemodynamics and hyperplasia one of coincidence or causality? The results presented here suggest that flow disturbances cause kinetic energy transfer through the vessel wall and into perivascular tissue where it becomes manifest as tissue vibration. This energy transfer has been quantitated in vivo and correlates strongly with venous intimal-medial thickness. There is clearly a reproducible pattern to the development of venous hyperplasia in arteriovenous loop grafts as seen clinically, 3 in our previous study, 2 and in the present study. The characteristic lesion occurs at the 'toe' of the anastomosis (location 2)
Volume 11 Number 4 April 1990
Grafi geometry and intimal-medial hyperplasia 563
0.30 *
E
0.20
p < 0.05 vs Control
[]
Control
[]
4-7 mm Taper
[]
6 mm Untapared paired with 4-7's
0.10
0.00 Cntrl Ufl
Uf2
Uf3
Uf4
U15
F1
F2
F3
F4
F5
Location on Femoral Vein
Fig. 7. Venous intimal-medial thickness in 4 to 7 mm taper grafts (F1-F5) and paired 6 mm untapered grafts (Ufl-Uf5), mean + SE. Cntrl, Normal control vein; numbers 1 through 5 refer to locations I through 5 (see Fig. 2).
0.30 *
0.20
0.10
0.00
/I iii Cntrl Us1
Us2 Us3
Us4 Us5
$1
$2
$3
$4
p < 0.05 vS Control
[]
Control
[]
7-4 mm Taper
[]
6 mm Untapered paired with 7-4's
$5
Location on Femoral Vein
Fig. 8. Venous intimal-medial thickness in 7 to 4 mm taper grafts (S1-$5) and paired 6 nun tmtapered grafts (Usl-Us5), mean _+ SE. Cntrl, Normal control vein; numbers 1 through 5 refer to locations 1 through 5 (see Fig. 2). with minimal to no change in vein, that is only a short distance away, including the heel of the anastomosis. Thus the detailed evaluation of hemodynamics in these locations will produce strong evidence regarding causation. While emphasizing local hemodynamics it is important to note that the bilateral implantation protocol ensures that systemic parameters (e.g., cardiac output, mean arterial blood pressure, metabolic and coagulation factors) are the same for each pair of grafts. Graft material, suture, and implantation technique were also identical for all grafts. Differences between graft types therefore iso-
lated the effect of graft geometry on local hemodynamic and mechanical parameters. It is commonly believed that a 4 to 7 mm taper graft will have a lower volumetric flow rate than a 6 mm untapered graft in the same location, but that was not the case in this study. Since most femoral arteries in the study were 4 lama in diameter it is likely that the 4 mm anastomosis was rarely more flow limiting than the 6 mm anastomosis, and pressure data confirm this. This underscores the importance of analyzing an entire flow system before assuming the impact o f a certain graft geometry. The similarity
564
Journal of VASCULAR SURGERY
Fillinger et al.
of flow rates for the three graft types also indicates that alteration of volumetric flow rate as such is not likely responsible for the differing degrees ofintimalmedial thickening. An interesting contrast is thus provided by our previous work. 2 In that study 6 mm untapered grafts were implanted bilaterally and a flow-limiting band was placed on one of the grafts so flow rates differed but anastomotic geometry was identical. We used Reynolds number as a quantitative measure of turbulence confirmed qualitatively by phonoangiography of exposed vessels (color Doppler technology was not yet available). Reynolds number was found to correlate strongly with intimal-medial thickening. In the present study geometries differ but flow rates do not. With differing geometries one cannot assume a simple correlation between preanastomotic Reynolds number and the degree of turbulence, but this parameter is nonetheless a useful indicator of flow stability. Higher numbers indicate less stable flow and Reynolds numbers >2300 indicate fully developed turbulence is likely. The range of Reynolds numbers in this study indicate flow disturbance is almost certain at an irregularity such as a distal end-to-side anastomosis, but that it will likely dampen out as flow moves downstream. Spectral tracings from in vivo Doppler studies were consistent with this analysis. Flow disturbances at locations 2 and 3 are often violent, but the disturbances progressively lessen as flow moves away from the anastomosis. Aliasing prevented the use of spectral signals to quantitate turbulence, but qualitatively disturbances appeared to be least severe in 4 to 7 mm taper grafts (Fig. 3). There was no obvious difference between untapered and 7 to 4 mm taper grafts in terms of postanastomotic turbulence. It is interesting that if vibration data are not used in the regression analysis, distal graft (preanastomotic) Reynolds number has the best correlation with venous intimal-medial thickening. Although the correlation is not as strong as in our previous work, this is not surprising with the complicating factor of differing anastomotic geometries. Flow disturbances were dramatically lower at locations 4 and 5 for all three graft types. Tissue vibration and intimal-medial hyperplasia was also dramatically lower at these locations. Thus the proposed relationship between turbulence, energy transfer through the vessel wall, and hyperplasia holds not only across different graft types, but also holds consistently for different locations within the vein itself. The vibration of perivascular tissue requires energy. Some component of the hemodynamic forces
within the lumen must be exerted at right angles to the vessel wall to set up the observed vibration patterns. Of the hemodynamic parameters studied, three components are logical candidates for such forces. The pulsatile nature of the flow is an obvious choice, but the high frequency "thrill" of an arteriovenous fistula is much higher than the frequency of pressure oscillations in pulsatile flow. Doppler technology indicated frequencies on the order of 1000 to 3000 Hz for the perivascular vibrations studied here. The second and third candidates are components of velocity--the mean component, which will have vectors at right angles to the vessel wall in disturbed flow of any kind, and a fluctuating component superimposed on the mean velocity. To be precise, the mean velocity does fluctuate, but it does so in an easily described, repeatable manner. The "fluctuating component" referred to here is a randomly fluctuating component that occurs only with actual turbulence. 4 When examining the hemodynamic fo:-%es in this way it seems clear that only the high freqhency, randomly fluctuating forces related to true turbulence could cause the vibrations observed in this study. It also seems likely that these forces would cause abnormal wall stresses and chronic vascular injury. It is interesting to note that Nathan and Imparato ~ also detected high frequency (1000 to 2000 Hz) signals using phonoangiography with frequency analysis in an arteriovenous model of intimal hyperplasia. In their study symmetric lesions were produced in preparations producing only low frequency signals, but they noted an association of high frequency signals at implant with a characteristic asymmetric lesion. They felt the high frequency signals resulted from turbulent flow and hypothesized that a relationship existed between intimal hyperplasia and the f~ces responsible for the high energy signals. Acute tissue injury is a known potentiator of intimal hyperplasia. 6"7 Cyclic stretch has also been implicated in stimulation of cell growth via biologic mediators. 8 The work of Fry 9 demonstrated acute cell damage in the face of either extremely high shear stress or turbulence, and the Reynolds numbers of this model are similar. Our previous study related turbulence to the development of hyperplasia, and we postulated that the relationship was one of chronic vessel injury and stress? The present study demonstrates in vivo that rapidly oscillating stresses and forces that could induce chronic vascular injury occur at the precise locations where hyperplasia occurs.
The present results must also be reviewed in terms of other prominent theories regarding i.nti-
Volume 1 l Number 4 April 1990
f
" •
mal hyperplasia. When discussing turbulence, onc must bring up theories regarding boundary layer separation ~° and low shear stress. 11 Boundary layer scparation is a phcnomenon in which flow streamlincs move in a regular, definable pattern that is not parallel to the vessel wall (unlike simple laminar flow). Flow with Reynolds numbers in the 500 to 2000 range combined with the geomctry of distal end-to-side anastomoses certainly favors such phcnomcna. Advcrse pressure gradients were also found in thc femoral vein for all graft types (Fig. 5). Color flow patterns did revcal somc areas of rcgularly repcating flow disturbances that wcre consistent with boundary layer separation, but thcse wcre found within the graft or in the inferior vein (locations 4 and 5) rather than at locations 1 and 2 on the vcin where hyperplasia is prominent. The patterns at locations 1 and 2 were chaotic and random, which is more indicative of actual turbulence than boundary layer ~¢paration. In addition, the relcase of energy in the form of tissuc vibration is further indication of a more scvere form of flow disturbance or actual turbulence. It has been noted that high shear and low shear are likely to coexist in areas of flow disturbance. ~2 Low flow vclocities (and thereforc low shcar stresscs) were common in thc infcrior vein but were rare or nonexistent at vein locations 1 through 3. The lowest flow velocities at locations 1 through 3 were observed in veins of the 4 to 7 mm taper configuration where hyperplasia was least severe. Thcse obscrvations point toward high shcar and turbulence as causcs for hyperplasia rathcr than low shear and the milder disturbance of boundary layer scparation. This does not necessarily imply that past theories are incorrect. It is q'-:,te possible that the cause of hyperplasia in the high flow milieu of arteriovenous loop grafts is different from that of arterial bypass grafts. If the results presented here do not strongly support low shcar as the causative factor in this form of hypcrplasia, what of high shear? Is this type of hypcrplasia simply the result of a high velocity "jet" at the vcnous anastomosis, with the highest velocities producing the most severe response? Although this is possible, observations do not strictly support this theory either. The highest flow velocities in the distal graft arc found in the 7 to 4 mm taper configuration, but the intimal-medial thickening was no worse than that of veins in untapered 6 mm graft configurations. One might suggest that anastomotic geometry is a confounding factor, but that would also argue against a simple "jet" effect. Furthermore, flow velocity in the distal graft did not correlate with intimal-
Graft geometry and intimal-medial hyperplasia 565
medial thickening as well as energy transfer or Reynolds number--again providing evidence against a simple "jet" theory. There is evidence that tangential pressure stress may be at least partially responsible for the development of intimal hyperplasia in veins subjected to elevated (arterial) pressures. 13,14Close examination of the pressure patterns found here (Fig. 5) would seem to argue against this theory for venous hyperplasia in arteriovenous loop grafts. Femoral venous pressures at location 2 were significantly increased in untapered grafts but not in 7 to 4 mm taper grafts. Also, femoral venous pressure was significantly higher in the inferior portions (locations 4 and 5) of tapered grafts where no significant hyperplasia was found. These patterns of pressure argue against tangential pressure stress as a major factor in this form of hyperplasia. Finally, it has also been suggested that kinetic energy transfer plays a role in the hyperplasia associated with compliance mismatch, is but compliance mismatch was not altered in the present study. Although the relationship between altered geometry, hemodynamics and histologic abnormalities was postulated 30 years ago, 16 technology has advanced to the point that hemodynamic mechanisms can be proved. The implication is that grafts can be designed to produce more favorable hemodynamic patterns and thereby reduce the development of venous intimal hyperplasia, at least in arteriovenous loop grafts. Current research emphasizes graft materials and surface phenomena such as platelet-graft interactions and endothelial cell seeding. Although these are indeed important areas, graft geometry can be altered cheaply and easily with currently available technology. Efforts in this area may well prove to be of value until breakthroughs are achieved in other areas of hyperplasia research. We conclude that graft geometry can significantly affect hemodynamic patterns and venous intimalmedial hyperplasia in arteriovenous loop grafts. Flow disturbances appear to cause kinetic energy transfer through the vessel wall and into perivascutar tissue. Kinetic energy transfer in the form of perivascular tissue vibration has been quantitatcd in vivo and correlates strongly with venous intimal-medial thickness. The postulated relationship is as follows: graft geometry affects several hemodynamic factors, of which flow stability (Reynolds number) is the most important. Flow stability determines the degree to which flow disturbance or turbulence occurs at irregularities such as the venous anastomosis. The larger the disturbance at the anastomosis, the greater
566
Journal of VASCULAR SURGERY
Fillinger et al.
the release of energy. The forces that cause perivascular energy transfer through the vessel wall also cause chronic stress and damage, triggering biochemical mediators and ultimately resulting in the development of intimal-medial hyperplasia. REFERENCES 1. Rizzuti RP, Hale JC, Burkart TE. Extended patency of expanded polytetrafluoroethylene grafts for vascular access using optimal configuration and revisions. Surg Gynecol Obstet 1988;166:23-7. 2. Fillinger MF, Reinitz ER, Schwartz RA, Resetarits DE, Paskanik AM, Bredenberg CE. Beneficial effects of banding on venous intimal-medial hyperplasia i n arteriovenous loop grafts. Am J Surg 1989;158:87-94. 3. Sottiurai VS, Yao JST, Flinn WR, Batson RC. Intimal hyperplasia and neointima: an ultrastructural analysis of thromhosed grafts in humans. Surgery 1983;93:809-17. 4. Whitmore RL. The flow of fluids. In: Whitmore RL, ed. Rheology of the circulation. 1st ed. New York: Pergamon Press, 1968:37-9. 5. Nathan I, Imparato AM. Vibration analysis in experimental models of atherosclerosis. Bull NY Acad Med 1977;53:84968. 6. Clowes AW. Pathobiology of arterial healing. In: Strandness Jr DE, Didisheim P, Clowes AW, Watson JT, eds. Vascular diseases: current research and clinical applications. 1st ed. Orlando: Grune & Stratton, 1987:351-62. 7. LoGeffo FW, Quist w e , Cantelmo NL, Haudenschild CC.
8.
9.
10.
11.
12.
13.
14.
15.
16.
Integrity of vein grafts as a function of initial intima, and medial preservation. Circulation 1983;68(suppl II): 117-24. Sumpio BE, Banes AJ, Levin LG, Johnson G. Mechanical stress stimulates aortic endothelial cells to proliferate. J VASe SURG 1987;6:252-6. Fry DL. Acute vascular endothelial changes associated with increased blood velocity gradients. Circ Res i968;22:16S97. Crawshaw HM, Quist WC, Serrallach E~ Valeri CR, LoGerfo FW. Flow disturbance at the distal end-to-side anastomosis. Arch Surg 1980; 115:1280-4. Rittgers SE, Karayannacos PE, Guy JF, Nerem RM, Shaw GM, Hostetler JR, Vasko JS. Velocity distribution and intimal proliferation in autologous vein grafts in dogs. Circ Res 1978;42:792-801. Nerem RM, Levesque MJ. Hemodynamics and the arterial wall. In: Strandness Jr DE, Didisheim P, Clowes AW, Watson JT, eds. Vascular diseases: current research and clinical applications. 1st ed. Orlando: Grune & Stratton, 1987:296-9. Kohler TR, Kirkman TR, Clowes AW. The effect of rigid external support on vein graft adaptation to the arterial circulation. J VASC SURG 1989;9:277-85. Karayannacos PE, Hostetler JR, Bond MG, et al. Late failure in vein grafts: mediating factors in subendothelial fi,~¢~muscular hyperplasia. Ann Surg 1978; 187:183-8. Suggs WE), Henriques HF, DePalma RG. Vein cuff interposition prevents juxta-anastomotic neointimal hyperplasia. Ann Surg 1988;207:717-23. Imparato AM, Lord JW, Texon M, Helpern M. Experimental atherosclerosis produced by alteration of blood vessel configuration. Surg Forum 1961;12:245-7.
