INTRODUCTION
Textile vascular prostheses made from polyethylene terephtalate (PET) fibers
are commonly used today in the treatment of aneurismal or occlusive disease
of arteries with medium and large diameter (Chakfe et
al., 2004; Rae et al., 1991). Plain weaved
tubular prostheses have a relatively tight construction that ensures blood tightness
and durability, but limits the extent of the healing process, compliance and
flexibility (Pourdeyhimi, 1986; Guidoin
et al., 1982; Rae et al., 1991). Theses
limitations were responsible for several complications and longterm failures,
namely thrombogenic occlusion, dilatation and rupture, sutureline failure and
bleeding (Feldstein and Pourdeyhimi, 1990; Dieval
et al., 2003; Fontes et al., 2004;
Riepe et al., 1997).
Healing capacity and longterm stability are the most important in vivo performances
of a textile vascular prosthesis (Chakfe et al.,
2004). These performances are strongly linked to porosity level and mechanical
behavior of the prosthesis. On one hand, textile vascular prosthesis must be
porous enough in order to show a quick integration after implantation, by promoting
a normal cells proliferation from the native vessel or surrounding tissue (Pourdeyhimi,
1986; Guidoin et al., 1987). However, it
must be tight to blood to avoid postoperative risks of hemorrhage (Chu
and Rawlinson, 1994; Guidoin et al., 1982,
1987). On the other hand, vascular prosthesis must be
enough resistant and compliant to support blood pressure as long as possible
(Dieval et al., 2003) and to avoid high shear
stress and turbulence of blood flow at anastomoses (Ballyk
et al., 1998; Ben Abdessalem et al., 2001,
2005; Salacinski et al.,
2001). Furthermore, flexible grafts are usually easy to handle and suture
by surgeons. All these properties are related to the prosthesis textile parameters
such as yarn properties, fabric design and density and graft steps such as crimping,
cleaning and sterilization (Chu and Rawlinson, 1994;
Dieval et al., 2003; Feldstein
and Pourdeyhimi, 1990).
In order to manufacture woven prostheses responding simultaneously to all the mentioned requirements i.e., quick healing and longterm stability, a compromise has to be established between their porosity, permeability, mechanical durability and compliance. This compromise can be obtained by adopting appropriate textile construction parameters.
The aims of the present study are the identification of the most influencing
textile manufacturing parameters on plain weaved vascular prosthesis performances
and the determination of optimal manufacturing parameters values. The simultaneous
optimization methodology, using fractional factorial experimental design analysis
and overlaid contour plots based on multiple linear regression models was used
(Douglas, 2004).
MATERIALS AND METHODS
Prostheses samples manufacturing: Texturized Dacron^{®}
polyethylene terephtalate (PET) filament yarns were used to weave seamless tubes
according to plain taffeta pattern (Fig. 1, 2)
by using a shuttle loom specially built for straight and bifurcated tubular
structures fabrication (Ben Abdessalem et al., 2006).
Table 1: 
Performance of prostheses samples according to twolevel fractional
factorial design (2^{(52)}) 

*Values of responses are mean of five repetitions 

Fig. 1: 
Tubular plain woven structure (a cross section) 

Fig. 2: 
Sample of tubular plain woven structure 

Fig. 3: 
Experimental device for water permeability measurement 
Yarns characteristics as well as fabric parameters were varied following a
twolevel fractional factorial design of five selected factors: Linear density
of longitudinal (warp) yarns X_{1} (110, 167 dtex (10^{4} g
m^{1}), number of filaments in the warp yarns X_{2} (34, 100
filaments), linear density of transversal (weft) yarns X_{3} (110, 167
dtex), number of filaments in the weft yarns X_{4} (34, 100 filaments)
and warp density X_{5} (24, 40 ends/cm). Eight (2^{(52)}) different
samples of weaved prostheses (Table 1) were manufactured and
their physical and mechanical properties were tested.
Prostheses properties testing
Porosity measurement: Porosity P (%) which is a measure of void fraction
within the prosthesis was determined gravimetrically according to the standard
ISO 7198 (1998) by using the following equation:
where, M (g) is the total mass of the sample, A (mm^{2}) is total area, t (mm) is the thickness of the fabric and ρ (g cm^{3}) is fibers density (ρ = 1.37 g cm^{3} for polyester fibers).
Permeability measurement: Tightness to blood of prostheses samples was
evaluated by measuring water permeability, which is a measure of water can flow
through prosthesis wall under physiological pressure conditions, according to
the standard ISO 7198 (1998). We used a permeability test
device (Fig. 3) permitting water flow through a circular sample,
having an area of 1.0 cm^{2} under a hydrostatic pressure equal to 16
kPa (120 mmHg). Flow rate measured by a flow meter (McMillan Co, model 1016)
corresponds to prosthesis water permeability W (mL/min/cm^{2} at 120
mmHg).
Mechanical characterization: In order to evaluate prosthesis strength,
durability and compliance, we performed circumferential tensile test and bursting
test simulating respectively dilatation and multidirectional efforts induced
in vivo by blood flow (Dieval et al., 2001).
Tubular samples having a length L (mm) and a nominal diameter D (mm) were placed
over two hemicylindric pins and stretched at a steady rate of 100 mm/min until
break (ISO 7198, 1998) on Lloyd tensile machine (Lloyd
LR 5k, Lloyd instruments Ltd) with 1 kN load cell. Circumferential tensile strength
σ_{r} (N mm^{1}) and dilatation percentage at rupture
%dilat_{r} (%) were then determined according to Eq. 2
and 3, respectively:
where, F_{r} (N) is the maximum load and A_{r} (mm) is the maximum elongation.
In order to evaluate mechanical durability of the prosthetic device, we calculated
the circumferential secant modulus (Nmm%) according to Eq. 4.
Indeed, Dieval et al. (2001) demonstrated that
the risk of longterm rupture of a textile vascular graft increases when its
circumferential secant modulus decreases.
Salacinski et al. (2001) defined vascular prosthesis
compliance as a structural property depending on dimensions and modulus. They
reported that vascular graft is well compliant when its circumferential stiffness
modulus is low. We determined therefore the circumferential initial modulus
(in N/mm%) from stressdilatation curve of each prosthesis to evaluate graft
compliance.
Prostheses samples were subjected to a probe bursting test (ISO
7198, 1998) by using Lloyd tensile machine operating in compression mode
with 1 kN load cell. A Flat sample clamped over a specimen holder, having a
circular orifice with area of 1 cm^{2}, is traversed by a cylindrical
probe with hemispherical head (diameter of 11 mm) at a steady rate of 100 mm
min^{1} until it bursts. Bursting strength F_{b} (in N cm^{2})
is then recorded. According to the British standard, the bursting strength of
a vascular graft has to be higher than 120 N cm^{2} in order to provide
an adequate safety factor for prosthesis long term implantation.
Statistical analysis: Factorial analysis permits to identify most important
textile manufacturing variables (factors) for each performance (response) of
the plain weaved vascular prostheses: porosity (Y_{1}), water permeability
(Y_{2}), circumferential secant modulus (Y_{3}), circumferential
initial modulus (Y_{4}) and bursting strength (Y_{5}). Multiplelinear
regression and analysis of variance was performed to model adequately the relationship
between each prosthesis response and their significant factors. Overlaid contour
plots were plotted from developed models to determine optimal values for significant
factors allowing the manufacture of woven prosthesis having simultaneously high
porosity, low permeability to blood, high mechanical durability and low stiffness.
For statistical analyses, we used Minitab14 software (MINITAB Ltd, Coventry,
United Kingdom).
RESULTS AND DISCUSSION
Factors selecting and models development: The responses Y_{j} (j = 1,2,3,4,5) corresponding to the eight manufactured samples are shown in Table 1.
Factorial analysis showed that the five studied factors exhibit different effects
on each plain weaved prosthesis response. Most important factors for each response
were identified from effects plots (Fig. 4ae)
and main effects plots (Fig. 5ae), generated
by factorial analysis. It can be noticed that warp and weft yarns linear density,
number of filament in warp yarns and warp density (X_{1}, X_{2},
X_{3} and X_{5}) were important factors for porosity (Y_{1}).
Whereas only warp yarns linear density and warp density (X_{1} and X_{5})
seem to be the most significant factors on plain weaved graft's water permeability
(Y_{2}). It can also be seen that only warp and weft yarns linear density
(X_{1} and X_{3}) have significant effects on prostheses circumferential
secant modulus (Y_{3}) of our prostheses, whereas only weft yarns linear
density (X_{3}) did not have significant effect on circumferential initial
modulus (Y_{4}). Finally, bursting strength (Y_{5}) of plain
woven prostheses depended mainly on warp linear density and warp density (X_{1}
and X_{5}), as shown in Fig. 4 and 5.
We developed then linear models showing the relationship between each response and their corresponding significant factors (Table 2). The analysis of variance for every response demonstrated that models are adequate (pvalue <0.05) and explain the 98.6, 88.4, 86.9, 96.7 and 89.5% of the variability in porosity, water permeability, circumferential secant modulus, circumferential initial modulus and bursting strength respectively, at the 95% confidence level. Predicted values agree quite well with measured values (high correlation coefficient R value (Table 2) for all the responses. Table 2 also demonstrates that the selected factors for each response are all significant (p<0.05).

Fig. 4: 
Effect plots for plain woven prostheses responses. (a) significant
effect plot for porosity, (b) significant effect plot for water permeability,
(c) significant effect plot for circumferential secant modules, (d) significant
effect plot for circumferential initial modules and (e) significant effect
plot for bursting strength 
Table 2: 
Regression coefficients, analysis of variance and correlation
between predicted and measured values for responses 

a_{i}: Coefficients of the models, p: Significance
level for the hypothesis that the coefficient is zero, R^{2}: Coefficient
of determination for the models, F: Fisher ratio Fvalue, pvalue: Significance
level for the hypothesis that al the coefficients in the model are zero,
R: Coefficient of correlation between predicted and measured values, p and
pvalue <0.05 are significant at 95% confidence level 

Fig. 5: 
Main effects plots for plain woven prostheses responses. (a)
significant effect plot for porosity, (b) significant effect plot for water
permeability, (c) significant effect plot for circumferential secant modules,
(d) significant effect plot for circumferential initial modules and (e)
significant effect plot for bursting strength 
Simultaneous optimization of plain woven vascular prostheses performances:
Graphical optimization method (Douglas, 2004) was adopted
to determine the optimum levels of textile construction parameters for plain
woven vascular prostheses performances. The contour plots performed from models
of every responses were overlaid and the region that best satisfied the requirements
(high porosity: >64%, low water permeability: <50 mL/min/cm^{2},
high mechanical durability (circumferential secant modulus >0.5 N/mm% and
bursting strength >120 N cm^{2}) and low stiffness (circumferential
initial modulus <0.1) expresses optimum conditions. Compromise regions were
obtained only for porosity and water permeability (Fig. 6),
for porosity and circumferential secant modulus (Fig. 7a,
b), for porosity and circumferential initial modulus (Fig.
8a, b) and for porosity and bursting strength (Fig.
9a, b).

Fig. 6: 
Overlaid contour plot of porosity (Y_{1}) and water
permeability (Y_{2}) as function of warp yarns linear density (X_{1})
and warp density (X_{5}) while the other factors (X_{2},
X_{3} and X_{4}) were set at high levels 

Fig. 7: 
Overlaid contour plots of porosity (Y_{1}) and circumferential
secant modulus (Y_{3}) when warp and weft yarns linear density change
while the other two factors (X_{2} and X_{5}) were set at
(a) low levels and also (b) at high levels 

Fig. 8: 
Overlaid contour plots of porosity (Y_{1}) and circumferential
initial modulus (Y_{4}); (a) when warp yarns linear density and
warp density change while the other factors (X_{2}, X_{3}
and X_{4}) were set at low levels and (b) when warp yarns linear
density and warp filament count change while the other factors (X_{3},
X_{4} and X_{5}) were set at low levels 

Fig. 9: 
Overlaid contour plots of porosity (Y_{1}) and bursting
strength (Y_{5}) as function of warp yarns linear density (X_{1})
and warp density (X_{5}) while the other factors (X_{2},
X_{3}) were set at (a) low levels and also at (b) high levels 
The unshaded area in Fig. 6 showed warp yarns linear density and warp density values leading to simultaneous optimal values of porosity and water permeability for plain woven vascular prostheses. The other construction parameters of the tubular fabric i.e. number of filament in the yarns and weft yarns linear density, should be at 100 filaments and 167 dtex, respectively.
Figure 7 showed warp and weft yarns linear density values enabling simultaneously high porosity and high circumferential secant modulus, while the other two factors i.e. number of filament in warp yarns and warp density were fixed at low values (Fig. 7a) and also at high values (Fig. 7b). As can be seen from Fig. 7, yarns of large range of linear density could be used to produce plain woven prostheses with both good porosity and high durability. But according to Fig. 8, in order to minimize the circumferential initial modulus of the tubular fabric, warp yarns with linear density higher than 150 dtex and warp density lower than 30 ends cm^{1} are required, when weft yarns linear density is 110 dtex and number of filament in weft yarns is 34 filaments. Furthermore, as shown by Fig. 9, warp yarns linear density of 150 dtex and warp density of 24 ends cm^{1} are sufficient conditions to manufacture plain woven vascular prosthesis having at the same time good porosity and satisfactory bursting strength.
Guidoin et al. (1982) and Pourdeyhimi
(1986) only compared the performances of textile vascular prostheses having
various textile structures. In this study, we investigate the dependence of
most important properties of plain woven vascular prostheses on their textile
manufacturing parameters. Effects plots and main effects plots allow to identify
the most significant textile manufacturing variables for each plain woven vascular
prostheses performance. Warp yarns linear density X_{1} had significant
effect on every performance, whereas weft yarns linear density X_{3}
was significant only for porosity Y_{1} and circumferential secant modulus
Y_{3}. Warp density X_{5} was significant for porosity Y_{1},
water permeability Y_{2}, circumferential initial modulus Y_{4}
and bursting strength Y_{5}. Number of filament in warp yarns X_{2}
had significant effect only on Y_{1}. Finally, number of filament in
weft yarns X_{4 }was significant only for Y_{4}. The proposed
models permit to predict performances of a plain woven vascular prosthesis having
known textile parameters and make easier its quality evaluation.
In this study we determined values for all textile parameters permitting to
obtain a plain woven vascular prosthesis with simultaneously good porosity,
low permeability, high mechanical strength and good flexibility. Compromise
regions were obtained only for porosity and water permeability, circumferential
secant modulus, circumferential initial modulus and bursting strength. This
confirmed the importance of the porosity for vascular prostheses overall quality
(Guidoin et al., 1982).
CONCLUSIONS
In this study, the effects of five textile manufacturing parameters on main performances of plain weaved vascular prostheses were investigated. Linear models were developed to establish relationship between each response and its significant factors. These models could be used to predict performances of plain weaved vascular prostheses.
Simultaneous optimization methodology based on overlaid contour plots permitted to determine optimal values for significant factors of each response in order to manufacture plain weaved prosthesis having at the same time high porosity, low permeability to blood, high mechanical durability and low stiffness. We concluded that the use of warp yarns having linear density of 150 dtex and weft yarns of 110 dtex with warp density of 24 ends cm^{1} could be a good compromise to obtain compliant plain woven vascular grafts with high porosity level, reasonable blood permeability and minimum longterm rupture risk.
Further work will focus on the study of the influence of plain weaved prosthesis textile manufacturing parameters on the adhesion of coating proteins to the prosthesis wall. In fact, the success of textile prostheses is also linked to the nature of interactions between textile structure and coating substances.