Mechanical behaviour of a new acrylic radiopaque iodine-containing bone cement

ARTICLE IN PRESS

Biomaterials 25 (2004) 2657–2667

Mechanical behaviour of a new acrylic radiopaque iodine-containing bone cement Catharina S.J. van Hooy-Corstjensa,*, Leon E. Govaertb, Anne B. Spoelstrac, Sjoerd K. Bulstraa,d, Gwendolyn M.R. Wetzelsa,d, Leo H. Koolea b

a Centre for Biomaterials Research, University of Maastricht, P.O. Box 616, 6200 MD Maastricht, The Netherlands1 Department of Mechanical Engineering, Eindhoven University of Technology, P.O. Box 513, 5600 MB Eindhoven, The Netherlands c Department of Chemical Engineering, Eindhoven University of Technology, P.O. Box 513, 5600 MB Eindhoven, The Netherlands d Department of Orthopaedic Surgery, Academic Hospital Maastricht, P.O. Box 5800, 6202 AZ Maastricht, The Netherlands

Received 20 February 2003; accepted 4 September 2003

Abstract In total hip replacement, fixation of a prosthesis is in most cases obtained by the application of methacrylic bone cements. Most of the commercially available bone cements contain barium sulphate or zirconium dioxide as radiopacifier. As is shown in the literature, the presence of these inorganic particles can be unfavourable in terms of mechanical and biological properties. Here, we describe a new type of bone cement, where X-ray contrast is obtained via the introduction of an iodine-containing methacrylate copolymer; a copolymer of methylmethacrylate and 2-[4-iodobenzoyl]-oxo-ethylmethacrylate (4-IEMA) is added to the powder component of the cement. The properties of the new I-containing bone cement (I-cement) are compared to those of a commercially available bone cement, with barium sulphate as radiopacifier (B-cement). The composition of the I-cement is adjusted such that similar handling properties and radiopacity as for the commercial cement are obtained. In view of the mechanical properties, it can be stated that the intrinsic mechanical behaviour of the I-cement, as revealed from compression tests, is superior to that of B-cement. Concerning the fatigue behaviour it can be concluded that, though B-cement has a slightly higher fatigue crack propagation resistance than I-cement, the fatigue life of vacuum-mixed I-cement is significantly better than that of B-cement. This is explained by the presence of BaSO4 clumps in the commercial cement; these act as crack initiation sites. The mechanical properties (especially fatigue resistance) of the new I-cement warrant its further development toward clinical application. r 2003 Elsevier Ltd. All rights reserved. Keywords: Iodine copolymer; Acrylic bone cement; PMMA; Radiopacifier; Mechanical properties

1. Introduction The long-term success of total hip replacement strongly depends on the mechanical fixation of the metallic stem in the femoral channel. Two fixation strategies are commonly used: one based on methacrylic bone cement, and the other on the basis of a bioactive coating on the metallic surface to which bone adheres strongly and rapidly after the implantation. Numeri*Corresponding author. Tel.: +31-43-38-81-278; fax: +31-43-3884-159. E-mail address: [email protected] (C.S.J. van HooyCorstjens). 1 The Centre for Biomaterials Research of the University of Maastricht is part of the Dutch Research School ‘‘Integrated Biomedical Engineering’’ (IBME). 0142-9612/$ - see front matter r 2003 Elsevier Ltd. All rights reserved. doi:10.1016/j.biomaterials.2003.09.038

cally, bone cements are dominant; cementless fixation is more expensive and used primarily for treatment of relatively young patients. Methacrylic bone cements are prepared in the operation theatre, from a powder consisting of polymethylmethacrylate (PMMA) and an initiator, and a liquid component, generally methylmethacrylate (MMA) (or a mixture of MMA and butylmethacrylate). Mixing of powder and liquid results, after a few minutes, in a mouldable material, which is injected into the femoral channel. This is followed by implantation of the femoral stem and the subsequent self-curing of the cement results in anchoring of the prosthesis. To follow the healing process after surgery, it is important that the cement is radiopaque. This is often achieved by the addition of a contrast agent, like barium

ARTICLE IN PRESS C.S.J. van Hooy-Corstjens et al. / Biomaterials 25 (2004) 2657–2667

2658

sulphate (BaSO4) or zirconium dioxide (ZrO2) [1]. The presence of these inorganic radiopacifying particles influences both the mechanical and biological behaviour of the bone cements. It is shown that small BaSO4 particles result in an improved fatigue crack propagation resistance of the bone cement [2,3]. On the other hand, the presence of barium sulphate reduces the tensile strength of the cement compared to radiolucent cement [3–5] and, moreover, barium sulphate clumps are known to act as fatigue crack initiation sites [6]. This implies that the presence of these particles can favour the aseptic loosening of the prosthesis [7]. From the biological point of view it is proven that BaSO4 wear particles may enhance the differentiation of macrophages into bone resorbing osteoclasts [8]. In view of the negative influence of the BaSO4 particles on both the mechanical and biological behaviour of the cement, it is desirable to find an alternative for the use of radiopaque additives. A promising approach is the use of (co)polymers having covalently bound heavy atoms, for example iodine. The synthesis of such biomaterials has been reported several years ago [9–13] and the biocompatibility and stability on the long term in vivo was demonstrated [14]. Here we describe a new type of bone cement, where the X-ray contrast is introduced via such iodine-containing copolymer (I-copolymer). Use is made of a copolymer of MMA and 2-[4-iodobenzoyl]-oxo-ethylmethacrylate (4-IEMA) (Fig. 1), which is present in the powder component of the cement. It should be noted that the new bone cement in this work differs conceptually from other recently described bone cements based on iodine-methacrylates, where the contrast agent is introduced via the liquid component [3,15]. In the latter case, the I-copolymer is formed during self-curing, where the I-containing methacrylate reacts with MMA. Since the conversion of monomers during self-curing is incomplete, free monomer will always be present and as a result, there is always a risk for in situ release of iodine-containing methacrylates. Because nothing is known about the toxic effects of such a monomer, this poses an unknown and presumably unacceptable risk for the patient. Our new bone cement does not suffer from this drawback, since the Icopolymer, free of residual monomers (as judged by high-field NMR), is introduced via the powder compoO Me O O I O Fig. 1. Structural formula of the I-containing methacrylate 4-IEMA.

nent of the cement and hence there is no risk of free rest I-containing monomer in the body. In this article, the new I-containing bone cement is compared to a commercial bone cement with BaSO4 as radiopacifier. Since fatigue plays an important role in cement failure, the fatigue life and the fatigue crack propagation behaviour of both cements are determined. To compare the intrinsic behaviour of the material, compression tests are performed. The homogeneity and the fatigue fracture surfaces of the bone cements are studied by scanning electron microscopy (SEM).

2. Materials and methods 2.1. Materials 2.1.1. Preparation of I-copolymer Commercial MMA was purified by distillation at atmospheric pressure. Commercial 2-hydroxyethylmethacrylate (HEMA) was purified by distillation at reduced pressure (oil pump). The monomer 2-[4iodobenzoyl]-oxo-ethylmethacrylate (Fig. 1) was synthesised from 4-iodobenzoyl chloride and purified HEMA [16]. MMA (150.0 g), 4-IEMA (150.0 g), azoisobutyronitril (AIBN, 472 mg), and 1-methyl-2-pyrrolidinone (NMP, 500 ml) were transferred into a 1-l round bottom flask. The mixture was stirred until a homogeneous solution was obtained. The flask was immersed in an oil bath that was interfaced with a time–temperature control system, and the temperature profile as described in Table 1 was run, while stirring was continued. The resulting viscous copolymer solution was added dropwise to water (approximately 2 l) contained in a 4-l beaker under vigorous mechanical stirring. After completion of the addition, stirring was continued for 1 h. The product, a white powder, precipitated rapidly and completely when stirring was stopped. The supernatant was separated and discarded, and the I-copolymer was washed extensively with water and ethanol. After each washing step, the I-copolymer was sieved and the final product was dried through lyophilisation. The dry Table 1 Temperature–time profile as used in the synthesis of the I-copolymer (Tt¼0 ¼ 40 C) Step

Process

Temperature ( C)

Time (h)

1 2 3 4 5 6 7 8 9

Heating Isothermal Heating Isothermal Heating Isothermal Heating Isothermal Cooling

60 60 80 80 100 100 130 130 40

0.5 8 1 4 1 4 1.5 2 4

ARTICLE IN PRESS C.S.J. van Hooy-Corstjens et al. / Biomaterials 25 (2004) 2657–2667

material was powdered and stored in a glass vial. Gel permeation chromatography revealed that the weightaveraged molar mass (Mw ) was 8  104 g/mol, and that the polydispersity ratio was 3. 1H-NMR spectroscopy in solution (CDCl3) confirmed the identity and the purity of the material and, moreover, showed that the material did not contain any residual monomer (MMA or 4IEMA). This could be concluded from the absence of any signals in the spectra region d ¼ 5:5  6:2 ppm; both MMA and 4-IEMA have characteristic singlet signals in this region, due to their vinylic protons. Differential scanning calorimetry tests showed that the glass-transition temperature was 90 C. The material was thermally stable up to 280 C (thermal gravimetry test). 2.1.2. Preparation of bone cements The I-copolymer was applied as a radiopacifying material for the new type of bone cement (I-cement). The behaviour of this cement was compared to that of a commercial cement (C-ment3; kindly provided by EMCM Nijmegen, The Netherlands) with BaSO4 as radiopacifier (B-cement). The composition of the Icement was optimised such that a similar curing temperature–time curve as for the B-cement was obtained. The composition of both cements is compiled in Table 2. Cement preparation was performed by hand mixing of the powder and liquid components. After mixing for about 2 min with a spatula, the material was left for a few minutes. Once a non-sticking mixture was obtained (approximately 5 min after beginning of mixing) the material was moulded by hand to remove as much as possible air bubbles. The latter makes our hand mixing route different from the normal hand mixing, where only mixing with a spatula is applied. The obtained homogeneous mixture was subsequently filled in a Teflon mould. After curing of the cements, the specimens were machined into the desired shape. For compression tests the mixtures were filled in a Teflon tube with an inner diameter of about 10 mm and Table 2 Composition of B-cement and I-cement Powder

Liquid

B-cement

34.92 g polymethylmethacrylate 1.08 g benzoyl peroxide 4 g barium sulphate

13.84 g methylmethacrylate 2.16 g butylmethacrylate 0.39 g N,N-dimethyl-ptoluidine 20 ppm hydroquinone

31.04 g polymethylmethacrylate 0.96 g benzoyl peroxide 8 g I-containing copolymer

15.44 g methylmethacrylate 2.41 g butylmethacrylate 0.44 g N,N-dimethyl-ptoluidine 20 ppm hydroquinone

I-cement

the samples were cured at atmospheric pressure. From the rods, cylindrical specimens with a height and diameter of 6 mm were machined. Samples for fatigue crack propagation measurements were prepared by filling the material in a mould of 40  40  6 mm3 that was closed by screws. The dimensions of the machined specimen, which were prepared according to the ASTM E647-93, are given in Fig. 2a. Just before testing, a sharp, cooled razor blade was used to place a cut at the tip of the notch, in order to facilitate the initiation of the crack from that point when the specimen was loaded subsequently. For fatigue tests, two kinds of samples (round and rectangular) were prepared. For the round bars, curing was performed in a Teflon tube with an inner diameter of 10 mm and for the rectangular bars material was filled in a mould of 77  13  4 mm3, which was closed by screws. To minimise the formation of voids (which are very critical for fatigue tests) the samples were, in the first instance, prepared by vacuum mixing of the components (rectangular bars). However, this resulted in a poor dispersion of the BaSO4 particles in the cement. Therefore, hand mixing followed by moulding has been applied (round bars). To prevent the formation of voids due to the evaporation of MMA, the open1.25 W

1.25 W

an a

W

(a) d = 8.5

d=5

Type

2659

(b)

18

10 61.8 22.3 3.2

4.0 76.2

(c) Fig. 2. Test specimens with selected dimensions in mm (a) fatigue crack propagation W ¼ 32 mm, an ¼ 6:4 mm, a  an ¼ 122 mm; (b, c) fatigue.

ARTICLE IN PRESS 2660

C.S.J. van Hooy-Corstjens et al. / Biomaterials 25 (2004) 2657–2667

ended tubes were placed in an autoclave and subjected to a pressure of 3.2 bar. Figs. 2b and c show the shape and dimensions of the machined samples. 2.2. Methods 2.2.1. Temperature–time profile The time at which the powder and liquid components were put together was taken as t ¼ 0: After mixing the components for 2 min, approximately 25 g of the material was packed in aluminium foil. The exothermic polymerisation temperature was monitored at certain time intervals. 2.2.2. Scanning electron microscopy SEM analysis of the fatigue fracture surfaces was performed on an R.J. Lee PSEM75. The morphology of the cements (Fig. 5) has been studied with a Philips XL30 FEG ESEM, equipped with a solid-state backscattered electron detector. All samples were supplied with a thin conductive layer. 2.2.3. X-ray visibility To compare the radiopacity of both cements the samples (15 per cement) were X-ray irradiated with 1 mAs at 77 kV, resembling clinical practice [13].2 2.2.4. Compression tests Compression tests were performed on an MTS 810 servo-hydraulic tensile-compression tester. The machined cylinders (48 specimens) were compressed at ambient temperature at a true strain rate of 3  103 s1, and force and displacement were recorded continuously. Though the ASTM F451-99a standard specification for acrylic bone cements prescribes the sample height to be twice the diameter, we used samples with h=d ¼ 1: This low ratio proved to be favourable to study large strain deformations since buckling of the sample is avoided [17]. From both the I-cement and B-cement half of the specimens was incubated in an aqueous medium at 37 C for at least 2 weeks before testing. From the 12 curves per set-up (B-cement dry, B-cement wet, I-cement dry and I-cement wet) the mean yield stress and elastic modulus were determined. The non-parametric Mann– Whitney U-test was performed to make a pairwise comparison between the different samples, at a 5% level of significance. 2.2.5. Fatigue crack propagation tests Fatigue crack propagation tests were conducted on an MTS 810 servo-hydraulic tensile-compression tester, using the settings as proposed by Lewis and Nyman 2

Note: the authors of [13] propose to use a non-ionic water-soluble iodine compound as contrast agent; i.e. their approach differs from the concept of using an iodine-containing copolymer.

[18]. The samples were loaded in tension–tension using a and sinusoidal wave function (Fmax ¼ 170 N Fmin ¼ 10 N) at a frequency of 2 Hz. The tests were performed at room temperature in air at a relative humidity of approximately 67%. Images of the crack were recorded with a digital camera and an analysis software program was used to determine the crack length at each time interval (magnification 10  ). Like for the compression tests, part of the samples was incubated in an aqueous medium at 37 C before testing. The crack length (a) was plotted against the number of cycles (N) and the crack growth rate was calculated by fitting a bi-exponential curve through the data points (r2 > 0:99) and determining the derivative of this curve. Once the crack starts to propagate, the fatigue crack propagation can be presented with the help of the Paris– Erdogan equation [19], which implies that the crack propagation rate (da=dN) depends on the stress intensity factor range (DKI ): da=dN ¼ C DKIm ;

ð1Þ

where C and m are material constants. The mode I stress intensity factor range DKI can be calculated according to F ðaÞ pffiffiffiffiffiffi ; DKI ¼ DF ð2Þ ðB W Þ where DF (Fmax  Fmin ) is the load range of the fatigue cycle, F ðaÞ is a geometric factor, B is the thickness of the sample and a is defined as a=W : The geometric factor for the compact tension geometry, as calculated from the elastic theory, is [19] ð2 þ aÞ F ðaÞ ¼ ð1  aÞ1:5  ð0:886 þ 4:64a  13:32a2 þ 14:72a3  5:6a4 Þ: ð3Þ For each specimen, the crack growth rate was plotted on a log–log scale as a function of DKI and a linear regression was fit to the data to determine the coefficients of the power-law relationship. Additionally, the data from all specimens of a cement (four specimens for the dry samples and three for the wet samples) were combined and a single linear regression was calculated for each data set. The Mann–Whitney U-test was applied to determine whether the differences between the regressions were significant at the p ¼ 0:05 level. 2.2.6. Fatigue tests During the tension–tension fatigue tests, all performed on an MTS 858 Mini Bionix tensile-compression tester, a sinusoidal wave function with a frequency of 2 Hz was applied. The loading was force-controlled and the applied forces were set as such that smax was 30 MPa and smin =smax was 0.1. (We are aware that the applied stress is much higher than the stress experienced by the

ARTICLE IN PRESS C.S.J. van Hooy-Corstjens et al. / Biomaterials 25 (2004) 2657–2667

where b is the Weibull slope and Na the Weibull characteristic fatigue life (36.8% survival of the population). PðNf Þ is the probability of fracture ½¼ ðY  0:3Þ=ðn þ 0:4Þ ; with Y being the failure order number of the specimen in the set of n test specimens (1=shortest fatigue life to n=longest fatigue life).

3. Results 3.1. Curing behaviour The curing properties of both cements have been compared by measuring the temperature development of the powder/liquid mixtures in triplicate. The median temperature–time curves of both cements are depicted in Fig. 3. The results of both cements are very similar. The maximum temperature (Tmax ) is 86.5 C for B-cement and 85.5 C for I-cement. Also the setting times (time at which Tset ðTset ¼ 0:5ðTamb þ Tmax Þ is reached) are nearly equal; 12.7 min for the B-cement and 13 min for the Icement. This implies that for the compositions as described in Table 2 the handling properties of both cements are similar. 3.2. X-ray visibility Fig. 4 shows a fluoroscopic image of the fatigue test specimens of both radiopaque cements. From this figure it is clear that the radiopacity of I-cement (a) is similar to that of B-cement (b).

90 80 70

Temperature [°C]

cement mantle in cemented arthroplasty (2–11 MPa) [7]. However, the high stress levels are applied to get a comparison of the fatigue behaviour on the cements at a relatively short time interval. In future, fatigue tests will be performed at the more clinically realistic stress levels.) Before starting the tests, the machined specimens were incubated in an aqueous medium for at least 2 weeks at 37 C. Testing was performed in an aqueous medium at 25 C. After failure, the fracture surfaces have been studied by SEM. Fatigue failure often starts at a defect. To get a good impression about the real material parameters and to eliminate the effect of unrepresentative defects, samples with an air bubble equal to or larger than 0.5 mm at the fracture surface were rejected from analysis. To perform pairwise comparisons of the mean fatigue lives of the fatigue samples, the non-parametric Mann–Whitney U-test has been applied, at a 5% level of significance. The fatigue tests were also analysed using the two-parameter Weibull equation. The linearised form of this model is given by the expression   1 ln ln ð4Þ ¼ b ln Nf  b ln Na ; 1  PðNf Þ

2661

60 50 40 30 20 4

6

8

10

12

14

16

18

20

Time [min]

Fig. 3. Temperature–time curves upon cement preparation: (m) Icement; (J) B-cement.

Fig. 4. X-ray image of tensile bars: (a) I-cement; (b) B-cement.

3.3. Morphology The morphology of the I-cement and the B-cement has been studied by SEM using atom number contrast visualised in the backscattered electron images (Fig. 5). In both figures (a) and (b) the PMMA particles of the powder component are clearly visible as dark spheres with a diameter between 15 and 55 mm. Between the spheres the monomer has polymerised. For the Bcement (b), barium sulphate particles are distinguishable in these regions. Micrograph (a) shows that for the Icement the contrast agent is much more homogeneously dispersed in the continuous phase. 3.4. Intrinsic mechanical behaviour Compression tests have been performed to investigate the intrinsic mechanical behaviour of the cements. The slope of the first part of the curve characterises the stiffness of the material (E-modulus) and the maximum of the curve (yield point) defines the stress and strain beyond which the material will deform plastically. The drop after the maximum is a measure for the amount of localisation within the sample; the larger the drop, the more strain will tend to localise during tensile tests [20]. In Fig. 6 the median results for I-cement and B-cement are given, both for as prepared samples and for samples

ARTICLE IN PRESS 2662

C.S.J. van Hooy-Corstjens et al. / Biomaterials 25 (2004) 2657–2667

Compressive true stress [MPa]

100

80

60

40

20

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Compressive true strain [-]

(a)

Fig. 6. Average compressive true stress–true strain behaviour: (n) Icement dry; (J) B-cement dry; (m) I-cement wet; ( ) B-cement wet. Table 3 Material properties of the two bone cements in dry and wet form, as derived from the 48 compression tests B-cement

E-modulus (GPa) Yield stress (MPa)

(b) Fig. 5. Backscattered electron images of the two different bone cements: (a) I-cement; (b) B-cement.

that have been incubated in an aqueous medium for at least 2 weeks at 37 C. Table 3 compiles some data resulting from statistical analyses of the 48 compression tests. The data reveal that the I-cement has a longer range of linear elastic behaviour than the B-cement and that it exhibits a significantly higher yield stress. From the drop in both E-modulus and yield stress it follows that water clearly acts as a plasticiser for both cements. The B-cement is more influenced by incubation in the aqueous medium than the I-cement. This can be explained by the slightly higher water uptake of the former (2.0 wt%) compared to the I-cement (1.7 wt%). The post-yield behaviour is similar for all four samples. 3.5. Fatigue crack propagation The fatigue crack propagation rate of both cements in dry and wet form is depicted in Fig. 7. Table 4 represents the values for the C and m coefficients of the Paris–

I-cement

dry

Wet

Dry

Wet

2.0 8672.2

1.6 6771.8

2.1 9472.9

1.8 7972.1

Erdogan equation (Eq. (1)) obtained by fitting a linear regression for each specimen along with the correlation coefficients to the regressions. The C and m coefficients for the combined data of each material are given in bold. For both I-cement and B-cement the regression line of the combined data for the dry samples lies slightly above that of the wet samples. However, taking into account the data spread for the various samples, these differences are negligible. Also statistical analysis of the fitting results of each specimen separately reveals no statistical difference between the dry and wet samples of I-cement or B-cement. Obviously, incubation in an aqueous medium does not influence the fatigue crack propagation rate of the bone cements. By comparing the Icement to the B-cement, no statistically significant difference is found between the crack growth regression slopes (m) of the two samples. The crack growth regression intercept (C) of the I-cement is considerably higher than that of the B-cement for both the dry (56.4  107 vs. 11.6  107) and the wet (39.4  107 vs. 9.31  107) samples. According to the Mann–Whitney U-test this difference is significant for the dry samples, whereas for the wet samples this difference is not significant (p ¼ 0:13). However, due to the limited number of samples, statistical analyses are difficult to apply. Overall, the fatigue crack growth rate da=dN of B-cement will be slightly lower than for I-cement.

ARTICLE IN PRESS C.S.J. van Hooy-Corstjens et al. / Biomaterials 25 (2004) 2657–2667

2663

Table 4 Fatigue crack propagation rate parameters for each specimen (from Eq. (1))

-4.5

-5.0

log da/dN [m/cycles]

Specimen

C  107

m

-5.5

B-cement dry B-cement dry B-cement dry B-cement dry B-cement dry

-6.0

-6.5

8.17 13.7 10.3 14.8 11.6

Correlation coefficient

7.10 6.04 6.35 7.04 6.13

0.992 0.992 0.996 0.981 0.969

9.10 6.64 9.30 7.11 7.92

0.988 0.991 0.966 0.965 0.936

-7.0

-7.5 -0.20

-0.15

-0.10

-0.05

0.00

0.05

0.10

log ∆K [MPa m0.5]

(a)

I-cement dry I-cement dry I-cement dry I-cement dry I-cement dry

108 33.5 88.0 38.5 56.4

-4.5

-5.0

log da/dN [m/cycles]

-5.5

-6.0

-6.5

B-cement wet B-cement wet B-cement wet B-cement wet

9.41 15.0 7.0 9.31

11.65 6.17 6.87 6.40

0.987 0.976 0.994 0.917

I-cement wet I-cement wet I-cement wet I-cement wet

51.7 13.2 25.4 39.4

9.24 12.65 9.14 9.52

0.988 0.959 0.968 0.893

Note: In bold the results upon fitting the combined data are given.

-7.0

-7.5 -0.20

(b)

-0.15

-0.10

-0.05

0.00

0.05

0.10

0.15

0.20

log ∆K [MPa m0.5]

Fig. 7. Fatigue crack propagation behaviour and regression lines of (n; upper line) dry specimens; ( ; lower line) wet specimens. (a) I-cement; (b) B-cement.

In Fig. 8 typical examples of the fatigue crack propagation fracture surface of I-cement and B-cement are depicted (dark regions are PMMA spheres of the original powder). The fracture surface of I-cement is smooth, whereas that of the B-cement is rougher. Another difference is that for the I-cement the cracks proceed throughout the whole cement, whereas for the B-cement it seems that the crack preferentially grows through the interbead matrix. 3.6. Fatigue To determine the fatigue life of the samples, the samples have been loaded cyclically. To study the real material parameters and to eliminate as much as possible the effect of defects, samples with an air bubble larger than 0.5 mm at the fracture surface have been rejected from analysis. In the first instance, vacuum mixing was applied, since this is known to reduce the cement porosity [21]. The results of the fatigue tests are given in Fig. 9. For seven of the 13 vacuum-mixed Bcement samples, fracture started from a clearly distin-

guishable BaSO4 particle (approximately 0.5 mm or larger). This means that for the conditions we applied, vacuum mixing was not appropriate to obtain a homogeneous dispersion of the radiopacifier. The results obtained for the I-cement (N ¼ 3533) are, at the p ¼ 0:05 level, significantly better than the results for the Bcement (N ¼ 2017). When for the B-cement only the samples without a BaSO4 defect of approximately 0.5 mm or larger at the fracture surface are taken into account (n ¼ 6; N ¼ 33917975) no significant difference is found between B-cement and I-cement. Hand mixing of the samples followed by moulding proved to result in a much better dispersion of the BaSO4 particles. Now only two of the 20 B-cement samples contained a BaSO4 clump at the fracture surface (six samples contained an air bubble). The cycles till failure for the hand-mixed samples of both cements are not significantly different. This confirms the result of the vacuum-mixed samples that B-cement with homogeneously dispersed radiopacifier has a similar fatigue resistance as I-cement. By comparing the vacuum-mixed samples to the hand-mixed sample for I-cement no significant difference is found. From the literature it is known that the different mixing techniques can also result in a different porosity, but since in the analysis all samples with an air bubble at the fracture surface of 0.5 mm or larger are rejected, this difference is likely to be (partially) eliminated here. For B-cement, hand mixing followed by moulding results in a better dispersion of the BaSO4

ARTICLE IN PRESS 2664

C.S.J. van Hooy-Corstjens et al. / Biomaterials 25 (2004) 2657–2667 Table 5 Fitting results of the linearised form of the two-parameter Weibull model

(a)

(b) Fig. 8. Scanning electron micrographs of the fatigue crack propagation fracture surface: (a) I-cement; (b) B-cement; scale bar=20 mm.

Specimen

Na

b

I

Correlation coefficient

Vacuum-mixed B-cement Vacuum-mixed I-cement Hand-mixed B-cement Hand-mixed I-cement

2302 3917 4249 3050

0.8448 3.7375 1.5890 3.2853

2115 7572 5356 5528

0.97696 0.95949 0.95082 0.96534

particles and as a result a statistically improvement of the fatigue life is obtained compared to vacuum mixing. Besides comparison of the mean and standard deviation of Nf ; also the two-parameter Weibull model has been applied to determine some characteristics of the different cements. The estimated parameters are summarised in Table 5. The Weibull slope, b; features the degree of scatter within the fatigue data; a larger b indicates that the variability in the results is small. So, good fatigue performance requires both a long fatigue life (i.e. a high value of Na ) and a high predictability of Nf (i.e. a high value of b). The combination of these two factors is expressed in the fatigue performance index Ið¼ Na ObÞ [22]. From Table 5 it follows that the vacuum-mixed Bcement has the lowest fatigue performance index, whereas the vacuum-mixed I-cement demonstrates the highest value for this index. It is also seen that the handmixed samples do not exhibit a large degree of variation. The latter is in agreement with the Mann–Whitney Utest, where no significant difference between the two hand-mixed cements was found. As mentioned, the inferior performance of the vacuum-mixed B-cement is caused by poorly dispersed BaSO4 particles. The higher value I of the vacuum-mixed I-cement compared to the hand-mixed cement suggests that for the former a lower porosity is obtained. Since the hand mixing method is improved by moulding by hand and all samples with an air bubble of 0.5 mm and larger are rejected from analysis, the difference between the two mixing methods is not as drastic as is found normally. According to the U-test the difference is not even significant.

4. Discussion

Fig. 9. Mean fatigue life with standard deviation.

Since the introduction of the first commercial bone cement in 1964, intensive research to optimise the handling characteristics and mechanical properties of these cements has proceeded. To assure that the Icement has the desired handling properties, a comparison with a commercially available cement has been made. The similarity in the temperature–time curves of both cements proves that the handling properties of I-cement are indeed comparable to those of the

ARTICLE IN PRESS C.S.J. van Hooy-Corstjens et al. / Biomaterials 25 (2004) 2657–2667

commercial cement. From the beginning of mixing, it takes about 10 min before the cement starts to harden. Another important factor is that the new cement is clearly visible on an X-ray image. To achieve this, the amount of I-copolymer has been chosen such that a similar, or even slightly better, radiopacity as for the commercial cement is obtained. Once the composition of the I-cement is optimised for both the handling properties and the radiopacity, investigations on the mechanical behaviour can be performed. In the previous section the results of several mechanical tests are given. Although the mechanism of cement failure is not well defined, the cyclic nature of joint loading suggests that fatigue will play an important role. For fatigue life, both crack initiation and crack propagation are essential factors. As mentioned, the fatigue crack propagation is measured by pre-cracking the samples. During the other set of fatigue tests the total of fatigue crack initiation and propagation is determined. Our measurements of fatigue crack propagation show that, in contrast to the compression tests, no influence of water absorption on the crack propagation rate is observed. It is also concluded that B-cement possesses a slightly higher crack propagation resistance than Icement. A possible explanation for the better performance of the B-cement compared to the I-cement is based on the morphology of the fracture surfaces. Icement exhibits a smooth fracture surface, characteristic of indiscriminate crack propagation through PMMA and the interbead matrix. For the B-cement, the crack seems to propagate preferentially through the interbead matrix. This results, on the microscopic scale, in a longer crack path. Though this not definitely the only determining factor, it certainly contributes to a lower macroscopic crack growth rate. Incubation in water will not influence this preference and it seems therefore logical that no difference between dry and wet samples is found. A possible explanation for the preference to grow through the matrix in case of the B-cement could be that the molecular weight of the matrix is lower than that of the PMMA particles and this would result in a decreased fatigue fracture resistance of the interbead matrix [23]. However, gel permeation chromatography has shown that the molecular weight of the original powder component is close to the molecular weight of the eventual cement. This implies that the interbead matrix should have a similar molecular weight as the PMMA particles. Therefore, the observed crack propagation path cannot be explained based on differences in molecular weight. Another, and based on the difference with the I-cement more logical, explanation can be found in the presence of the BaSO4 particles. It is known that voids or microcracks around BaSO4 particles weakens the interbead matrix and therefore the crack grows preferentially through this region [2].

(a)

2665

(b)

Fig. 10. B-cement with a poor dispersion of BaSO4 due to vacuum mixing: (a) X-ray image; (b) scanning electron micrograph of the total failure fatigue fracture surface; scale bar=500 mm.

Thus, though the effect is not very dramatic, the presence of homogeneously dispersed BaSO4 particles results in a decreased crack growth rate. However, as is described earlier, the dispersion of BaSO4 is not always homogeneous and therefore clumps of the inorganic radiopacifier are often present. For example, for the vacuum-mixed samples, more than half of the test specimens without an air bubble contained large clumps of BaSO4 at the fracture surface. An example of the Xray image (white dots are clumps of BaSO4) and the scanning electron micrograph of the fracture surface of such sample are given in Fig. 10. When the average fatigue life of these samples (8397631, n ¼ 7) is compared to that of the more homogeneous vacuummixed B-cement (33917975, n ¼ 6) a significant decrease in the fatigue life of the samples is observed. This implies that clumps of BaSO4 clearly act as fatigue crack initiation sites. When the more homogeneous B-cement samples are compared to the I-cement no significant difference is observed. This means that, though the fatigue crack propagation rate slightly increases by replacing BaSO4 with the I-copolymer, the total fatigue life of the cement stays unaltered. This suggests that the resistance against crack initiation is larger for I-cement than for B-cement. A possible explanation for this difference can be obtained from the compression tests. From these measurements it is clear that the yield stress of the I-cement is considerably higher than that of Bcement. As a result, the I-cement will be more resistant to crack initiation than B-cement.

ARTICLE IN PRESS 2666

C.S.J. van Hooy-Corstjens et al. / Biomaterials 25 (2004) 2657–2667

Apparently, microscopically dispersed BaSO4 particles lower the fatigue crack propagation rate but not the total fatigue life. However, clumps of the inorganic radiopacifier lower the total fatigue life of the cement drastically. From the measurements it can be concluded that two important factors contribute to premature fatigue failure of B-cement. These factors are: (i) porosity due to air bubbles and (ii) relatively large clumps of contrast agent. For I-cement only factor (i) is relevant: the distribution of the iodine-containing copolymer is more homogeneous in comparison with the distribution of BaSO4. This difference between B-cement and I-cement may be important, especially since it is unknown whether large contrast clumps are also generated under clinical conditions. We presume that this may be the case, especially since, in our hands, the application of vacuum (in order to decrease porosity) during preparation of the B-cement, was associated with the formation of relatively large clumps. In future, fatigue experiments at the more clinically realistic stress levels will be performed. To further characterise the material, also creep and pull-out experiments can be done. Finally, we would like to mention that the presence of BaSO4 particles promotes bone resorption [8] and thereby increases the risk of aseptic loosening of the prosthesis. We will therefore perform tests to investigate possible differences of both cements on bone resorption. 5. Conclusions 1. For the chosen composition of the I-cement, both handling properties and radiopacity are comparable to those of the B-cement. 2. Compression tests reveal that the intrinsic behaviour of I-cement is better than that of B-cement. 3. Concerning the tension–tension fatigue behaviour it can be concluded that, though the fatigue crack propagation rate of I-cement is slightly higher than that of B-cement, the total fatigue life vacuum-mixed I-cement is significantly better than that of vacuummixed B-cement. On the other hand, the fatigue life of homogeneous I-cement and B-cement is similar. This implies that the dispersion of the radiopacifier plays a very important role. Whereas I-cement shows a very homogeneous morphology, B-cement quite often contains clumps of BaSO4. These clumps act as a crack initiation site and as a result, the fatigue life of the B-cement is drastically lowered.

Acknowledgements We would like to thank the mechanical workshop of the IDEE Maastricht for machining the test specimens.

References [1] Lewis G. Properties of acrylic bone cement: state of the art review. J Biomed Mater Res 1997;38:155–82. [2] Molino LN, Topoleski LD. Effect of BaSO4 on the fatigue crack propagation rate of PMMA bone cement. J Biomed Mater Res 1996;31:131–7. [3] Ginebra MP, Albuixech L, Fernandez-Barragan E, Aparicio C, Gil FJ, San Roman J, Vazquez B, Planell JA. Mechanical performance of acrylic bone cements containing different radiopacifying agents. Biomaterials 2002;23:1873–82. [4] Haas SS, Brauer GM, Dickson G. A characterization of polymethylmethacrylate bone cement. J Bone Jt Surg Am 1975;57:380–91. [5] Vazquez B, Deb S, Bonfield W. Optimization of benzoyl peroxide concentration in an experimental bone cement based on poly(methylmethacrylate). J Mater Sci: Mater Med 1997;8: 455–60. [6] Bhambri SK, Gilbertson LN. Micromechanisms of fatigue crack initiation and propagation in bone cements. J Biomed Mater Res 1995;29:233–7. [7] Krause W, Mathis RS. Fatigue properties of acrylic bone cements: review of the literature. J Biomed Mater Res 1988;22:37–53. [8] Sabokbar A, Fujikawa Y, Murray DW, Athanasou NA. Radioopaque agents in bone cement increase bone resorption. J Bone Jt Surg Br 1997;79:129–34. [9] Benzina A, Kruft MA, Bar F, van der Veen FH, Bastiaansen CW, Heijnen V, Reutelingsperger C, Koole LH. Studies on a new radiopaque polymeric biomaterial. Biomaterials 1994;15: 1122–8. [10] Benzina A, Kruft MA, van der Veen FH, Bar FH, Blezer R, Lindhout T, Koole LH. A versatile three-iodine molecular building block leading to new radiopaque polymeric biomaterials. J Biomed Mater Res 1996;32:459–66. [11] Mottu F, Rufenacht DA, Doelker E. Radiopaque polymeric materials for medical applications. Current aspects of biomaterial research. Invest Radiol 1999;34:323–35. [12] Davy KWM, Anseau MR, Odlyha M, Foster GM. X-ray opaque methacrylate polymers for biomedical applications. Pol Int 1997;43:143–54. [13] Kjellson F, Almen T, McCarthy I, Lidgren L. A simple method of comparing radiopacity in different cements 17th European Conference on Biomaterials, Barcelona, Spain, 11–14 September 2002. [14] Aldenhoff YB, Kruft MA, Pijpers AP, van der Veen FH, Bulstra SK, Kuijer R, Koole LH. Stability of radiopaque iodinecontaining biomaterials. Biomaterials 2002;23:881–6. [15] Vazquez B, Ginebra MP, Gil FJ, Planell JA, Lopez Bravo A, San Roman J. Radiopaque acrylic cements prepared with a new acrylic derivative of iodo-quinoline. Biomaterials 1999;20: 2047–53. [16] Kruft MA, Benzina A, Blezer R, Koole LH. Studies on radio-opaque polymeric biomaterials with potential applications to endovascular prostheses. Biomaterials 1996;17: 1803–12. [17] van Melick HGH, Govaert LE, Meijer HEH. On the origin of Strain hardening in glassy polymers. Polymer 2003;44(8): 2493–502. [18] Lewis G, Nyman JS. Toward standardization of methods of determination of fracture properties of acrylic bone cement and statistical analysis of test results. J Biomed Mater Res 2000;53:748–68. [19] Anderson TL. Fracture mechanics-fundamentals and applications. Boston: CRC Press; 1991.

ARTICLE IN PRESS C.S.J. van Hooy-Corstjens et al. / Biomaterials 25 (2004) 2657–2667 [20] van Melick HGH, Govaert LE, Meijer HEH. Localisation phenomena in glassy polymers: influence of thermal and mechanical history. Polymer 2003;44(12):3579–91. [21] Lidgren L, Drar H, Moller J. Strength of polymethylmethacrylate increased by vacuum mixing. Acta Orthop Scand 1984;55: 536–41.

2667

[22] Dunne NJ, Orr JF, Mushipe MT, Eveleigh RJ. The relationship between porosity and fatigue characteristics of bone cements. Biomaterials 2003;24:239–45. [23] Topoleski LD, Ducheyne P, Cuckler JM. Microstructural pathway of fracture in poly(methyl methacrylate) bone cement. Biomaterials 1993;14:1165–72.