Eur J Vasc Endovasc Surg 18, 375–380 (1999)

Article No. ejvs.1999.0892

Influence of Crimping Textile Polyester Vascular Prostheses on the FluidFlow Kinetics

S. Ben Abdessalem1, N. Chakfe∗2, J-F. Le Magnen1, M. Beaufigeau2, D. Adolphe1, B. Geny2, S. Akesbi1,G. Riepe3, J. G. Kretz2 and B. Durand1 (Groupe Europeen de Recherche sur les Protheses appliquees

a la Chirurgie Vasculaire)

1Laboratoire de Physique et Mecanique Textile, Ecole Nationale Superieure de l’Industrie Textile de Mulhouse,Mulhouse, France; 2Department of Cardiovascular Surgery, Les Hopitaux Universitaires de Strasbourg, Strasbourg,

France; and 3Department of Thoracic and Vascular Surgery, Harburg, Hamburg, Germany

Objectives: to characterise the impact of the crimping of polyester prostheses on the fluid flow kinetics.Design: an experimental in vitro study.Materials and methods: we investigated four models of polyester vascular prostheses in a continuous laminar flowcircuit. The flow velocity was 80 ml/s for all experiments. We studied two fluids of different viscosity within the circuit.The speed of the particles was measured by a laser Doppler anemometer 2 to 52 mm from the prosthetic interface. Wefirst established a calibrated flow-velocity profile corresponding to the study of the support inside the circuit without anyprosthesis. We measured the velocity profiles for each prosthesis corresponding to four crimp densities obtained bystretching the grafts.Results: the crimping of PET textile prostheses led to a decrease of flow velocity especially closer to the prosthetic surface.The decrease of flow velocity was dependent on the model of prosthesis. This decrease of flow velocity is described by thefollowing negative exponential law: DV=a·b−x where (a) is the crimp density and (b) the fluid viscosity.Conclusions: flow velocity near a prosthetic surface is influenced by the morphology of the crimping. The impact ofcrimping on the flow velocity in a vascular prosthesis can be predicted by computer simulation models. This may providethe optimal shape of crimping for each prosthesis.

Key Words: Blood-vessel prosthesis; Polyester; Polyethylene terephthalate; Blood flow.

Introduction steam pressure in a mould designed according to the

required configuration. This process usually producesThe first generation of textile prostheses were made a circular crimp but may also produce helical con-from hand-sewn woven polyester.1 These prostheses figuration. It is especially suited for velour prostheses,demonstrated difficulties after the implantation be- since it preserves the velour feature. The secondcause of the lack of compression resistance and their method is the thread-crimping process. Here thetendency to kink. Crimping gave the grafts a radial crimping is produced by winding a wire spirallyresistance and a longitudinal compliance. It was around the prosthesis, filled on a rod before com-achieved by an “accordian”-pleat deformation. It per- pressing it longitudinally. The configuration is thenmitted easier implantation of the prosthesis, as it al- heat-seated, obtaining a helical crimping. This methodlowed the surgeon to control the longitudinal tension. tends to permanently flatten the filamentous featureCrimping also improved the resistance of the pros- of the velour. Different kinds of crimping, dependingthesis to kinking2 and to external compression.3

on their circular or helicoidal shape, their width andCrimping can be carried out in two main ways. their depth, have been proposed. However, the impact

The first method is the mould-crimping process. The of a specific choice of crimping on the main propertiescrimping is created by the application of an internal of a vascular prosthesis, such as its thrombogenicity,

its healing, and its resistance to kinking or to external∗ Please address all correspondence to: N. Chakfe, Department of compression, have not yet been precisely defined.Cardiovascular Surgery, Chirurgie A, Les Hopitaux Universitaires

The aim of the present study was to evaluate thede Strasbourg, 1, place de l’Hopital, BP No 426, 67091 StrasbourgCedex, France. impact of the crimping of polyester vascular prostheses

1078–5884/99/110375+06 $12.00/0 1999 Harcourt Publishers Ltd.

S. Ben Abdessalem et al.376

Fig. 2. Schematic representation of the setting of the prosthesis onthe support.

returned the fluid to the upper tank (Fig. 1). The flow

velocity inside the circuit was fixed at 80 ml/s for allFig. 1. Schematic representation of the flow circuit.

experiments. The experiments were conducted with

two fluids of different viscosity: mixture A made of

81% glycerine and 18% water (viscosity=45 mPa/s),on the blood-flow velocity. This knowledge is of majorand mixture B made of 75% glycerine and 25% waterimportance for the design of small-diameter vascular(viscosity=37 mPa/s). The mixtures were stabilisedprostheses, since a high blood-flow decrease mightfor two days before the experiments and the measure-alter their patency rates dramatically. We studied fourments.impregnated-polyester prostheses commonly used in

The prostheses were opened longitudinally andvascular surgery. We studied the impact of their crimp-fixed on a support placed in the flow field (Fig. 2).ing on flow velocities and also the impact of theThe flow velocity profiles were measured with a laserlowering of their crimp density by stretching them, asDoppler anemometer (Dantec Electronik, Skovlunde,would be performed during surgical implantation.Denmark). The laser beam intersection was computer-

driven and measurements were taken by steps of

2 mm, from 2 mm to 52 mm from the prosthetic in-

terface. At first, we established a calibrated flow-ve-Material and Methodslocity profile corresponding to the study of the support

inside the circuit without any prosthesis. We measuredWe studied four polyester vascular prostheses ofthe velocity profiles for each prosthesis correspondingwoven and knitted structures. The characteristics ofto four crimp densities randomly obtained by stretch-the prostheses were the following: prosthesis A: USCIing the prostheses. Prosthesis A was stretched to getWoven, a woven polyester prosthesis with a relaxedthe following crimp densities of 4.2, 3.3, 2.1, and 1.4.crimp density of 6.7 crimp/cm (Bard, Billerica, MA,Prosthesis B was stretched to get the following crimpU.S.A.); prosthesis B: Hemaguard W, a woven col-densities of 3.3, 2.7, 2.2, and 1.6. Prosthesis C waslagen-impregnated polyester prosthesis with a relaxedstretched to get the following crimp densities of 2.7,crimp density of 3.8 crimp/cm (Intervascular, La Cio-2.1, 1.7, and 1.2. Prosthesis D was stretched to get thetat, France); prosthesis C: VPK 1200, a warp-knittedfollowing crimp densities of 2.2, 1.7, 1.3, and 0.8.polyester prosthesis with a relaxed crimp density of

5.2 crimp/cm (Vascutek, Inchinahan, Scotland, U.K.);

and prosthesis D: Dialine II: a warp-knitted collagen-

impregnated polyester prosthesis with a relaxed crimp

density of 4.3 crimp/cm (Cardial, Saint-Etienne,

France). Crimp density was defined as the number of Resultscrimps per length, measured on a 10 cm length, at five

different areas of the prostheses. The measurements of the flow velocity depending on

the distance from the support described a parabolicWe constructed a laminar flow circuit with a circuit

of measurement of 500-mm length, 200-mm width and shape of plane Poiseuille flow. The maximum flow

velocity was measured in the middle of the device.60-mm depth. The flow inside the circuit was obtained

by gravity. The measures of velocities were performed Setting up a prosthesis on the support modified this

curve by a slight deformation at the left side of theunder a constant gradient of pressure which was

created by locating a tank collector at a constant DH curve corresponding to a decrease of the flow velocities

related to the prosthesis (Fig. 3). For each experiment,of 175 cm below the level of the circuit. A pump

Eur J Vasc Endovasc Surg Vol 18, November 1999

Crimping and Rheology of PET Grafts 377

Table 1. Calculation of the terms a and b of the theorical lawDV=a·b−x using the mixture A.

Prosthesis Crimp density Term a Term b(crimp/cm)

A 4.2 4.14 1.433.3 3.11 1.392.1 2.01 1.361.4 1.41 1.41

B 3.3 3.85 1.422.7 3.32 1.382.2 2.75 1.371.6 2.23 1.36

C 2.7 3.51 1.402.1 3.07 1.441.7 2.34 1.43

52

14

02

Position on Y (mm)

Vel

ocit

y (m

m/s

)

22

10

12

8

6

4

2

6 10 14 18 26 30 34 50

Re = 30.4

4 8 12 16 20 24 28 32 3638

4042

4446

48

1.2 1.57 1.36Fig. 3. Calibration of the flow velocity profiles with the mixture A. D 2.2 2.85 1.37The measures inside the device without any prosthesis provided a 1.7 2.21 1.42parabolic shape of plane Poiseuille flow. The setting of a prosthesis 1.3 1.61 1.39on the support modified this curve by the observation of a slight 0.8 1.07 1.41deformation at its left side corresponding to a decrease of the flowvelocities. (Curve of the prosthesis A with a crimp density of 4.2crimp/cm.) (—) Calibration; (--) prosthesis A with a crimp densityof 4.2 crimp/cm.

Table 2. Calculation of the terms a and b of the theorical lawDV=a·b−x using the mixture B.

Prosthesis Crimp density Term a Term b(crimp/cm)

A 4.2 4.23 1.183.3 3.35 1.212.1 2.43 1.201.4 1.36 1.17

B 3.3 3.45 1.192.7 2.53 1.232.2 1.75 1.241.6 1.19 1.20

C 2.7 3.24 1.162.1 2.65 1.171.7 1.88 1.18

14

2.5

02

Position on Y (mm)

Vel

ocit

y (m

m/s

)

12

1.5

2

1

0.5

4 6 8 10

Re = 30.4

1.2 1.12 1.22D 2.2 2.75 1.19

Fig. 4. Representation of the theoretical law DV=a·b−x using the1.7 2.05 2.16

mixture A and the prosthesis A with the crimp density of 4.2 crimp/1.3 1.73 1.23

cm. a=4.14; b=1.43.0.8 1.34 1.21

the results were expressed as DV corresponding to the

calibrated flow velocity minus the measured flow

velocity with the prosthesis for the same distance from crimp densities. We calculated the terms (a) and (b)

of the theoretical law for the mixture A (Table 1) andthe support.

The experiments showed a decreasing DV according for the mixture B (Table 2). The term (a) was found to

be related to the crimp density of the prostheses. Itto the distance of the point of measurement from the

prosthetic wall. The decrease of flow velocity was decreased with the crimp density. However, it was

not exactly the same for identical crimp densities ofhigher for the prostheses with a higher crimp density.

This decrease was highest at a distance between 2 to different types of prostheses in different states of

stretching. The term (b) was found to be more constant3 mm from the prosthesis for the four models. We

postulated that DV follows a decreasing exponential and to be influenced only by the viscosity of the

mixture used in the flow circuit. It was constantlylaw DV=a·b−x, where x was the distance between the

prosthesis and the point of measurement (Fig. 4). This about 1.4 for all the experiments done with the mixture

A, and about 1.2 for all the experiments done with thehypothesis led us to adjust the variation of Ln (DV)

with x to a regression straight line. The test was made mixture B. Stretching the prostheses led to a reduction

of the decrease of the flow velocities for the fourwith the four types of prostheses A, B, C, D, at different

Eur J Vasc Endovasc Surg Vol 18, November 1999

S. Ben Abdessalem et al.378

14

3.5

02

Position on Y (mm)

Vel

ocit

y (m

m/s

)

12

1.5

2.5

1

0.5

4 6 8 10

Re = 35.5 (B)n = 4.2

n = 3.3

n = 2.1

n = 1.4

3

2

14

3.5

02

Position on Y (mm)

Vel

ocit

y (m

m/s

)

12

1.5

2.5

1

0.5

4 6 8 10

Re = 30.4 (A)

n = 4.2n = 3.3n = 2.1

n = 1.4

3

2

14

3.5

02

Position on Y (mm)

Vel

ocit

y (m

m/s

)

12

1.5

2.5

1

0.5

4 6 8 10

Re = 35.5 (B)

n = 3.3

n = 2.7

n = 2.2

3

2

14

3.5

02

Position on Y (mm)

Vel

ocit

y (m

m/s

)

12

1.5

2.5

1

0.5

4 6 8 10

Re = 30.4 (A)

n = 3.3n = 2.7n = 2.2n = 1.6

3

2

n = 1.6

Fig. 5. Comparison of the theoretical law DV=a·b−x of the prosthesis Fig. 6. Comparison of the theoretical law DV=a·b−x of the prosthesisB depending on its stretching with the mixture A (A), and with theA depending on its stretching with the mixture A (A), and with the

mixture B (B). n=crimp density mixture B (B). n=crimp density

crimping may have adverse mechanical consequencesprostheses studied with both mixtures in the flow

by allowing the phenomenon of early dilatation of thecircuit (Figs 5, 6, 7, and 8).

textile structures after their implantation.4,5 Crimping

may also induce drawbacks for the prosthesis. The

thermofixation of the deformation applied to the fibres

may damage the polymer and the filaments, and con-Discussionsequently weaken the prosthesis. It has been shown

that the luminal corrugations of the prosthesis fillCrimping is accepted as obligatory to improve the

performances of polyester vascular prostheses and is rapidly with fibrin and cellular materials, and that

crimping resulted in delayed healing and a thickerused for all polyester prostheses. However, despite

various types of crimping, differing in circular or inner lining.6 Finally, the crimping may theoretically

impair the blood-flow kinetics near the wall.helicoidal shape, width or depth, the impact of a

specific type of crimping on the properties of a poly- The present study attempted to characterise the

impact of the crimping on in vitro laminar flow. Weester vascular prosthesis has not yet been precisely

defined. The main role of crimping is to enhance the showed that the crimped prosthetic interface had an

influence on laminar fluid-flow kinetics. Crimpinglongitudinal compliance and the radial resistance of

the prosthesis. Enhanced longitudinal compliance of decreased the flow velocity profile and this effect was

maximal closest to the wall of the prosthesis. Thesethe prosthesis is said to optimise its implantation by

controlling its tension. However, the optimal tension first results demonstrated that, in the experimental

conditions, it is possible to characterise the impact ofto prevent long-term elongation and/or dilatation is

not known. The radial resistance of the prosthesis is the crimping on the rheology. We also demonstrated

that the impact of the crimping on the flow velocitiesan important factor to reduce kinking and external

compression, but these properties have neither been can be mathematically described by a negative ex-

ponential law DV=a·b−x. We showed that the term (a)studied by in vitro tests nor by clinical investigation.

On the other hand, the radial compliance related to of this law was influenced by the crimp density since

Eur J Vasc Endovasc Surg Vol 18, November 1999

Crimping and Rheology of PET Grafts 379

14

3.5

02

Position on Y (mm)

Vel

ocit

y (m

m/s

)

12

1.5

2.5

1

0.5

4 6 8 10

Re = 35.5 (B)

n = 2.7

n = 2.1

n = 1.7

n = 1.2

3

2

14

3.5

02

Position on Y (mm)

Vel

ocit

y (m

m/s

)

12

1.5

2.5

1

0.5

4 6 8 10

Re = 30.4 (A)

n = 2.7n = 2.1n = 1.7n = 1.2

3

2

14

3.5

02

Position on Y (mm)

Vel

ocit

y (m

m/s

)

12

1.5

2.5

1

0.5

4 6 8 10

Re = 35.5 (B)

n = 2.2

n = 1.7

n = 1.3

n = 0.8

3

2

14

3.5

02

Position on Y (mm)

Vel

ocit

y (m

m/s

)

12

1.5

2.5

1

0.5

4 6 8 10

Re = 30.4 (A)3

2 n = 2.2

n = 1.7

n = 1.3

n = 0.8

Fig. 7. Comparison of the theoretical law DV=a·b−x of the prosthesis Fig. 8. Comparison of the theoretical law DV=a·b−x of the prosthesisD depending on its stretching with the mixture A (A), and with theC depending on its stretching with the mixture A (A), and with the

mixture B (B). n=crimp density mixture B (B). n=crimp density

best choice for the different clinical indications. Thisin all experiments (a) changed proportionally with it.

The values of (a) were different for the same crimp observation may be important when choosing a pros-

thesis for a site of implantation with a significantdensities obtained on models of prostheses differently

stretched. This observation is probably related to the risk of thrombosis such as infrainguinal bypasses.

Expanded polytetrafluoroethylene (ePTFE) prosthesesfact that (a) is not only influenced by the stretch

density of the crimping, but also by its morphological are preferred by most vascular surgeons for infra-

inguinal bypasses since ePTFE has been said to be lesscharacteristics and by the characteristics of the poly-

ester textile structure. On the other hand, the term (b) thrombogenic than polyester.7–10 However, retro-

spective11,12 and prospective randomised13 comparativeof the law was relatively constant for each experiment

and was found to be related to the fluid viscosity of studies have demonstrated that polyester prostheses

give similar patency rates for above-the-knee femoro-the mixture used for the experiment. In all experiments,

the modifications of the Reynolds number did not popliteal bypasses. The intrinsic thrombogenicity of

the biomaterial used for the construction of a vascularinfluence the DV curves shape. A decrease of the

Reynolds number led to a decrease of the DV value. prosthesis is, of course, an important factor to take

into account to allow the best patency rates. We haveThis is due to the fact that the Reynolds number

depends only on the flow characteristics and on the demonstrated that the geometric characteristics of the

prosthetic wall may induce alterations of the flowviscosity of the liquid inside the circuit. Since in our

experimental conditions the flow remained constant kinetics inside the prosthesis. Consequently, it seems

sensible to be able to measure and to predict thesefor all experiments, the Reynolds number varied in

the same way as the viscosity chosen for the liquid alterations in order to propose the best design of

crimping for each PET textile structure.inside the circuit.

We demonstrated that the decrease of fluid flow The present study had three main flaws when com-

pared to the physiological conditions of implantationkinetics depended on the model of prosthesis studied.

It would be of major interest to be able to compare in humans. We did not use blood inside the circuit

and used a continuous flow. We also studied a flatteneddifferent models of prostheses in order to propose the

Eur J Vasc Endovasc Surg Vol 18, November 1999

S. Ben Abdessalem et al.380

segment of prosthesis and not a tubular graft. These in a vascular prosthesis may be predicted by the useconditions were required the laser-Doppler an- of further numerical simulation models in order toemometry technique used for the flow-speeds meas- provide the best shape of crimping for each prosthesisurements as it measures the speed of particles within depending on its further application.a translucid fluid.

When developing a new kind of crimping for avascular prosthesis we must keep in mind not onlyits properties of radial resistance, compliance or lon- Referencesgitudinal compliance, but also its rheologic properties.

1 Wesolowski SA, Fries CC, Karlson KE, DeBakey MC, SawyerWe believe that all these parameters should be in-PN. Porosity: primary determinant of ultimate fate of synthetictegrated on the basis of the concept of usable length.vascular grafts. Surgery 1961; 50: 91–96.

The determination of the best tension to apply to a 2 Chakfe N, Jahn Ch, Nicolini Ph et al. The impact of knee jointprosthesis submitted to a physiological intraluminal flexion on infrainguinal vascular substitutes: an angiographic

study. Eur J Vasc Endovasc Surg 1997; 13: 23–30.pressure is fundamental. Too low a tension will induce3 Cavallaro A, Sciacca V, Di Marzo L, Bove S, Mingoli A.

a potential for dilatation and elongation of the pros- The effect of body weight compression on axillo-femoral bypassthesis, and will increase the crimp density and depth. patency. J Cardiovasc Surg 1988; 29: 476–479.

4 Alimi Y, Juhan C, Morati N, Girard N, Cohen S. DilatationToo high a tension will decrease the crimp density, itsof woven and knitted aortic prosthetic grafts: CT scan evaluation.

adverse impact on the flow, and the risk of elongation Ann Vasc Surg 1994; 8: 238–242.and dilatation. On the other hand, it may increase the 5 Urban E, Chakfe N, Zollner G et al. Helical computed tomo-

graphy with 3D reconstruction as a tool for the study of in vivorisk of anastomotic false aneurysms due to anastomoticdilatation of vascular prostheses. J Digit Imaging (submitted).disruption.

6 Takebayashi J, Kamatani M, Katagami Y et al. A comparativeThis study was necessary in order to propose a first study on the patency of crimped and non-crimped vascular

prostheses, with emphasis on the earliest morphological changes.mathematical characterisation of the rheology at theJ Surg Res 1975; 19: 209–218.level of the crimps, and was a part of a programme

7 Goldman M, Norcott HC, Hawker RJ et al. Femoropoplitealof characterisation of the mechanical properties of bypass grafts. An isotope technique allowing in vivo comparison

of thrombogenicity. Br J Surg 1982; 69: 380–382.textile prostheses. These first results will allow us to8 Shoenfeld NA, Connolly R, Ramberg K et al. The systemicstart a research programme of numerical simulation

activation of platelets by grafts. Surg Gynecol Obst 1988; 166:of these three properties in order to be able to propose 454–457.the best crimping for each textile structure. Such simu- 9 Eldrup-Jorgensen J, Mackey WC, Connolly RJ et al. Evaluation

of arterial prostheses in a baboon ex vivo shunt. The effect oflations of the arterial flow have been done previouslygraft material and flow on platelet deposition. Am J Surg 1985;

to predict the flow modifications related to mor- 150: 185–190.phological conditions such as arterial bifurcation or 10 Callow AD, Connolly R, O’Donnell TF et al. Platelet-arterial

synthetic graft interaction and its modification. Arch Surg 1982;bending14,15 and could be used to predict the blood117: 1447–1455.

flow inside a crimped polyester prosthesis. However,11 Rosenthal D, Evans RD, McKinsey J et al. Prosthetic above-knee

in our further models we will have to include the femoropopliteal bypass for intermittent claudication. J CardiovascSurg 1990; 31: 462–468.properties of the prosthetic wall and their changes

12 Pevec WC, Darling RC, L’Italien GJ, Abbott WM. Femoro-during the cardiac cycle and not to consider it as apopliteal reconstruction with knitted, nonvelour Dacron versus

fixed wall. expanded polytetrafluoroethylene. J Vasc Surg 1992; 16: 60–65.13 Abbott WA, Green RM, Matsumoto T et al. Prosthetic above-In conclusion, the crimping of polyester textile pros-

knee femoropopliteal bypass grafting: Results of a multicentertheses leads to a decreased flow velocity, especiallyrandomized prospective trial. J Vasc Surg 1997; 25: 19–28.

near to the surface of the prosthesis. The study of the 14 Thiriet M, Issa R, Graham JMR. A pulsatile developing flowin a bend. J Phys III 1992; 2: 995–1013.experimental results demonstrated that this decrease

15 Thiriet M, Pares C, Saltel E, Hecht F. Numerical model ofof flow velocity can be described by the followingsteady flow in a model of aortic bifurcation. ASME J Biomech

negative exponential law: DV=a·b−x where (a) was Eng 1992; 114: 40–49.related to the crimp density and (b) to the fluid vis-cosity. The impact of crimping on the flow velocities Accepted 26 March 1999

Eur J Vasc Endovasc Surg Vol 18, November 1999