Adsorption of drugs onto a poly(acrylic acid) grafted cation-exchange membrane

European Journal of Pharmaceutical Sciences 9 (1999) 137–143 www.elsevier.nl / locate / ejps

Adsorption of drugs onto a poly(acrylic acid) grafted cation-exchange membrane a b ˚ ˚ ¨ Koivu a , Annika Sundell c , Petteri Paronen a , Satu Akerman , Kari Akerman , Jouni Karppi b , Paivi a, * ¨ Kristiina Jarvinen b

a Department of Pharmaceutics, University of Kuopio, P.O. Box 1627, FIN-70211 Kuopio, Finland Department of Clinical Chemistry, Kuopio University Hospital and University of Kuopio, P.O. Box 1777, FIN-70211 Kuopio, Finland c ˚ ˚ , Finland Department of Polymer Technology, Abo Akademi University, Porthansgatan 3 -5, FIN 20500, Abo

Received 5 March 1999; received in revised form 8 June 1999; accepted 28 June 1999

Abstract The influence of pH, ionic strength and the concentration of albumin in the adsorption medium as well as the charge and lipophilicity of a model drug on their adsorption onto poly(acrylic acid) grafted poly(vinylidene fluoride) (PAA–PVDF) membranes was evaluated. The PAA–PVDF membrane is a responsive porous polymer membrane that we have studied for controlled drug delivery. Sodium salicylate (anionic), flunitrazepam (neutral), primidone (neutral), desipramine (cationic) and thioridazine (cationic) were used as model drugs. The extent of drug adsorption was dependent on pH. Drug adsorption was enhanced by the dissociation of the grafted PAA chains and by a positive charge and a high lipophilicity of the drug. Increasing the ionic strength of the medium retarded the adsorption of the cationic drugs. Interestingly, the present results showing that drugs are adsorbed onto the membrane while albumin is not adsorbed onto the membrane suggest that the PAA–PVDF membrane may be suitable for separating drugs from proteinaceous substances for subsequent monitoring and evaluation.  1999 Elsevier Science B.V. All rights reserved. Keywords: Ion-exchange membrane; Poly(acrylic acid); Drug adsorption; Albumin; pH; Ionic strength

1. Introduction Environmentally sensitive polymers are being developed for use in intelligent drug delivery systems so that the amount of the drug released can be adjusted in response to a physiological need (Kost and Langer, 1992; Heller, 1993). Externally controlled delivery systems rely on external factors, such as changes in the environmental pH or temperature, as the triggers for pulsed delivery whereas drug release from self-regulated delivery systems is controlled by the feedback information from the body. In addition to responding to environmental stimuli, interactions between the drug and the polymer may affect the drug release from responsive polymeric devices (Chang and Bodmeier, 1997; Miyajima et al., 1998). For example, salicylic acid release from acrylic resin films increased as the ionic strength of dissolution medium increased (Jenquin et al., 1990). This may be due to the *Corresponding author. Tel:. 1358-17-162488; fax: 1358-17-162252. ¨ E-mail address: [email protected] (K. Jarvinen)

fact that species that compete with the drug for binding sites on the polymer may reduce drug binding onto the polymer and thus, enhance drug diffusion through the polymer (Jenquin et al., 1990). In contrast, hydrophobic interactions between the drug and the polymer may retard drug diffusion through the polymer due to the strong interaction between the drug and the polymer (Bettini et al., 1995). Poly(acrylic acid) (PAA) is an environmentally sensitive polymer which is capable of undergoing conformational ¨ changes as a function of pH and ionic strength (Hautojarvi ˚ et al., 1996; Akerman et al., 1998). Iwata et al. (1988) and Ito et al. (1989) designed a glucose responsive insulin delivery system based on a porous PAA grafted membrane. We have shown that grafting of PAA onto the surface of an inert porous membrane results in a pH sensitive membrane which is able to control the transport of macromolecules across the membrane as a function of the environmental ˚ ¨ pH (Akerman et al., 1998; Jarvinen et al., 1998). In addition to pH, the environmental ionic strength controls the drug release from these membranes due to the cation¨ exchange nature of the membrane (Jarvinen et al., 1998).

0928-0987 / 99 / $ – see front matter  1999 Elsevier Science B.V. All rights reserved. PII: S0928-0987( 99 )00055-X

138

˚ S. Akerman et al. / European Journal of Pharmaceutical Sciences 9 (1999) 137 – 143

The aim of the present study was to evaluate the interactions between the drug and the PAA–poly(vinylidene fluoride) (PVDF) membrane.

2. Material and methods

2.1. Preparation of membranes Hydrophobic PVDF membranes (Millipore) with a pore size of 0.22 mm were graft modified with acrylic acid (Aldrich, Steinheim, Germany) stabilized with 200 ppm hydroquinone. Preirradiation grafting was accomplished by first irradiating the PVDF membranes under nitrogen atmosphere (,200 ppm O 2 ) using an Electrocurtain electron accelerator (Energy Sciences) operating at an acceleration voltage of 175 kV. The membranes were irradiated with 25 kGy. Immediately after irradiation, the membranes were immersed at ambient temperature in graft solution containing acrylic acid. This solution was continuously purged with nitrogen in order to remove oxygen. After grafting, the membranes were Soxhlet extracted with water to remove remaining monomer and dried in vacuo at 408C overnight. The degree of grafting was determined gravimetrically according to G 5 ((m 1 2 m 0 ) /m 0 ) 3 100% (wt%) where m 0 represents the mass of the original PVDF membrane and m 1 represents the mass of the grafted, extracted and dried PVDF membrane. All studies were performed using a 20 wt% grafted membrane. Membranes were pretreated in a drug-free 60 mM phosphate buffer solution (20 ml, pH was adjusted to pH of drug solution) for 15 h before immersing membranes in the drug solutions (pH 4.0–9.0).

2.2. Drug adsorption in the absence of albumin Solutions containing 5 mg / ml of desipramine (CibaGeigy, Basel, Switzerland), flunitrazepam (Hoffman-La Roche, Basel, Switzerland), primidone (Cambridge Research Biochemicals, Cheshire, UK) and thioridazine (Orion, Helsinki, Finland) were prepared by dissolving the required amount of drug in 60 mM phosphate buffer solution at pHs 4.0–9.0. Sodium salicylate (Aldrich) was dissolved (20 mg / ml) in 60 mM phosphate buffer solution only at pH 4.0 and 7.0. The ionic strength of the buffer solution was adjusted with NaCl (FF-Chemicals, Yli-Ii, Finland). Drug adsorption studies were performed by shaking flasks containing a 20 wt% grafted PAA–PVDF membrane (area 0.36 cm 2 , weight varying between 7 and 9 mg) and 20 ml of 5 mg / ml (desipramine, flunitrazepam, primidone, thioridazine) or 5 ml of 20 mg / ml (sodium salicylate) drug solutions at room temperature until equilibrium was

achieved. Control membranes that were shaken in the drug-free phosphate buffer solution (60 mM) were treated under identical experimental conditions. The unspecific adsorption of drugs onto flasks etc. was determined in the membrane-free solution. Samples of 100 ml were collected from the adsorption medium until the drug concentration in the adsorption medium was constant as a function of time. Drug concentrations were analyzed by HPLC as described in Analytical methods.

2.3. Drug adsorption in the presence of albumin Drug–albumin solutions were prepared by adding the drug to the bovine serum albumin (Sigma, St. Louis, MO, USA) solution in phosphate buffer (60 mM, pH 7.0) immediately prior to use. Concentrations of the drug and albumin in the final solution were 5 mg drug / ml and 20, 40 or 80 g albumin / l. Membranes were shaken at room temperature in 60 mM phosphate buffer solution (pH 7.0, 20 ml) containing 5 mg / ml of desipramine or thioridazine and 20 g / l, 40 g / l or 80 g / l of albumin until drug concentration in the adsorption medium was constant as a function of time. Samples of 100 ml were collected. Prior to the drug analysis, albumin was precipitated with 200 ml of acetonitrile, samples were centrifuged for 10 min at 14 000 rpm and supernatant was separated for analysis. Drug concentrations were measured by HPLC as described below. The unspecific drug adsorption onto flasks etc. in the presence of albumin was studied in drug–albumin solution in the absence of membrane. Adsorption of albumin onto the membrane was studied in the absence and presence of thioridazine. In the absence of thioridazine, adsorption of albumin onto the membrane was studied by soaking membranes in 1 ml of 0.005–50 g / l human serum albumin (SPR, Helsinki, Finland) solution in 60 mM phosphate buffer (pH 7.0). Nonspecific albumin adsorption was studied in the membrane-free solution. In the presence of thioridazine (5 mg / ml), adsorption of albumin onto the membrane was studied using 1 ml of 20 g / l bovine serum albumin solution.

2.4. Analytical methods Drug concentrations were measured by HPLC. The HPLC was performed with Hewlett Packard 1050 liquidchromatography controlled by Chemstation chromatography workstation (Hewlett Packard, USA). The chromatographic separations were achieved using a Select-B (4 mm, 12534 mm) C 8 analytical column (Merck, Darmstadt, Germany). The elution was isocratic with a mobile phase of acetonitrile: 50 mM dipotassium hydrogen phosphate, pH 4.7 (40:100) at a flow-rate of 1.2 ml / min. Drugs were detected at 220 nm and drug concentrations were calculated from the reduced peak areas. The assay was calibrated using six standard solutions. All standard curves

˚ S. Akerman et al. / European Journal of Pharmaceutical Sciences 9 (1999) 137 – 143

showed good linearity (R51.0). The peak purity analyses were performed at 210–365 nm. Concentration of sodium salicylate was measured by fluorescence polarization immunoassay with Abbot ADx analyzer (Abbot, Abbot Park, IL, USA). The concentration of albumin was measured by kinetic nephelometry with Beckman assay protein analyzer (Beckman, Brea, CA, USA).

139

anionic and neutral drugs (Table 2). Kapp -values reveal that the extent of the cation adsorption onto the membrane increased with drug lipophilicity (Tables 1–2) and Fig. 2 shows that also equilibrium was achieved faster in the case of thioridazine (the most lipophilic model drug) than for desipramine.

3.3. Effect of ionic strength on the drug adsorption 2.5. PAA–PVDF /phosphate buffer solution drug partition coefficient The apparent partition coefficient (Kapp ) is the ratio of amount of drug (mg) per gram of the membrane to the drug concentration (mg / ml) in the surrounding medium and it was calculated as follows (Miyajima et al., 1998) Kapp 5V 3 (C0 2 Ct ) /(W 3 Ct ) where V is the volume of adsorption medium (ml), C0 is the initial drug concentration (mg / ml), Ct is the equilibrium drug concentration (mg / ml), and W is the weight of the PAA–PVDF membrane (g).

3. Results

3.1. Effect of pH on drug adsorption The physicochemical properties of the studied drugs are shown in Table 1. Adsorption of all the studied drugs onto the membrane was affected by environmental pH. In all cases, increasing the pH from 4.0 to 6.0 did not have any major impact on the extent of the drug adsorption while the drug adsorption was increased strongly when the pH was increased from 6.0 to 7.0. Kapp -values clearly indicate that the interaction between the drug and the polymer was much stronger at pH 7.0 than at pH 4.0 (Table 2). A further increase in pH did not enhance drug adsorption. Adsorption of desipramine and thioridazine onto the membrane as a function of pH are shown in Fig. 1.

3.2. Effect of charge and lipophilicity of the drug on the adsorption The cationic model drugs (desipramine and thioridazine) were adsorbed onto the membrane to a considerably greater extent than anionic (sodium salicylate) and neutral (flunitrazepam and primidone) model drugs at each studied pH-value. No adsorption of salicylate, flunitrazepam or primidone was detected at pH-values ,6.0, while 2–15% of the initial drug dose was adsorbed at pHs $7.0. For comparison, at pH ,7.0 the amount of thioridazine and desipramine adsorbed onto the membrane was about 50% and 15% of the initial dose, respectively, and at pH 7.0, these values had increased to about 95% and 60%, respectively (Fig. 1). As a result, Kapp -values for the cationic drugs were considerably higher than those for the

Increasing the ionic strength of the adsorption medium from 0.2 to 1.0 did not substantially affect the total amount of desipramine and thioridazine adsorbed onto the membrane but tended to retard the drug adsorption (Fig. 2). Especially the adsorption of thioridazine was clearly retarded as the ionic strength increased (Fig. 2b).

3.4. Effect of albumin on the drug adsorption Albumin was not adsorbed onto the PAA–PVDF membrane. The extent of both desipramine and thioridazine adsorption onto the membrane was decreased in the presence of albumin (40 g / l), and adsorption of thioridazine was decreased even more (Fig. 3). Fig. 4 shows that adsorption of thioridazine onto the membrane decreased as the albumin concentration in the adsorption medium increased.

4. Discussion

4.1. Effect of conformation of PAA chains on the drug adsorption The conformation and charge of the membrane-grafted PAA chains are affected by the pH and ionic strength. At a low pH and at a high ionic strength, PAA chains are in a ¨ collapsed state (Hautojarvi et al., 1996). PAA chains become expanded at increasing pH and / or decreasing ionic strength. To study whether the conformation and charges of PAA chains affects drug adsorption, adsorption of the model drugs onto the polymer was studied at pHs 4.0–9.0 and at ionic strengths of 0.2 and 1.0. All the model drugs were adsorbed more efficiently at pHs $7.0 than at lower pHs indicating that the expanded conformation of PAA chains enhances drug adsorption. This conclusion is supported by the fact that decreasing the ionic strength of the adsorption medium from 1.0 to 0.2 (a more expanded conformation than at m 51.0) decreased the time required to achieve the equilibrium. These results can be explained by the fact that solute penetration within the expanded polymer is easier than solute penetration into a compact polymer. However, the fact that the pH induced changes in the conformation of PAA chains had clear effects on the extent of drug adsorption (Fig. 1) while ionic strength induced changes affected the rate but not extent of adsorption (Fig. 2) illustrates that there are

140

˚ S. Akerman et al. / European Journal of Pharmaceutical Sciences 9 (1999) 137 – 143

Table 1 Physicochemical properties of the model drugs MW a

pKa a

Log P a

Protein binding in plasma b

Desipramine

266.39

10.4

4.9 ; 4.09

70–90%

Flunitrazepam

313.29

1.8

2.06 ; 2.36

78%

Primidone

218.26

0.91 ; 1.74

,20%

Sodium salicylate

160.11

2.97

2.26 ; 2.19

50–90%

Thioridazine

370.58

9.5

5.90 ; 6.42

.99.5%

Drug

Structure

13 c

MW: molecular weight; Log P: log partition coefficient (oil–water). a Hansch, 1990. b Moffat, 1986. c Gal and Cone, 1988.

number of factors acting simultaneously to alter the drug adsorption onto the membrane.

4.2. Effect of drug charge and lipophilicity on adsorption Interactions between the cationic drugs and the mem-

brane were much stronger than the interactions between the anionic or neutral drugs and the membrane. In addition, the interactions between all the studied model drugs and the membrane were stronger at pH $7 than at lower pH-values. The pKa -value of PAA was calculated to be 6.72 at zero ionic strength and 4.88 at an ionic strength of 0.5 (Park and Robinson, 1987). Our results suggest that

˚ S. Akerman et al. / European Journal of Pharmaceutical Sciences 9 (1999) 137 – 143

141

Table 2 The apparent drug partition coefficients (Kapp ) between a 20 wt% grafted PAA–PVDF membrane and the phosphate buffer solution at pHs 4.0 and 7.0 (mean6S.D, n53) Drug

Kapp at pH 4.0

Kapp at pH 7.0

Sodium salicylate (anionic) Flunitrazepam (neutral) Primidone (neutral) Desipramine (cationic) Thioridazine (cationic)

Not detectable Not detectable Not detectable 144672 6546118

305640 1267 130625 10026487 956363798

dissociation of PAA chains occurred between pH 6.0 and pH 7.0, which is in good agreement with their reported pKa -values (Park and Robinson, 1987). We can conclude that the ionic interaction between the drug and the negatively charged carboxyl groups of PAA was the most important factor affecting drug adsorption. The low Kapp values of anionic and neutral drugs as well as the low Kapp -values of cationic drugs at pH 4.0 are due to absence of ionic interactions. Also the decrease in adsorption of desipramine (pKa 10.4) or thioridazine (pKa 9.5) that occurred when pH was increased from 8.0 to 9.0 can be explained by the fact that the ionic interaction between the drugs and the membrane became weaker due to the decreased fraction of ionized thioridazine and desipramine. However, it must be kept in mind that our results show that some drug adsorption occurred also via nonionic interactions. Although both cationic model drugs, desipramine and thioridazine, are fully dissociated at pH #8, Kapp -values of thioridazine were much higher than those of desipramine. This indicates that the physicochemical properties of a cationic drug can also affect drug adsorption. The present

Fig. 1. Effect of pH on adsorption of desipramine and thioridazine onto the PAA–PVDF membrane at ionic strength of 0.2. Mean values6S.D are shown (n53).

Fig. 2. Effect of ionic strength of the adsorption medium on (A) desipramine and (B) thioridazine adsorption onto the PAA–PVDF membrane at pH 7.0. Mean values6S.D are shown (n53).

results are in good agreement with the earlier study ˚ (Akerman et al., 1998) that suggested that adsorption of cationic drug onto the PAA–PVDF membrane is related to the lipophilicity of the drug. When the ion-exchange membrane binds ions present in the adsorption medium, an increasing concentration of these ions may reduce the adsorption of the drug onto the ion-exchange membrane due to the competition between the ions present in the adsorption medium and the drug for the adsorption sites in the membranes (Okada et al., 1987). It is known that PAA is able to bind sodium ions (Charman et al., 1991; Kriwet and Kissel, 1996). Increasing the ionic strength in the adsorption medium slowed down the adsorption of thioridazine but did not greatly affect the total extent of thioridazine adsorption (Fig. 2). In the case of desipramine, the ionic strength in the adsorption medium did not have any major effect on either the rate or the extent of adsorption (Fig. 2). The fact that the effect of the ionic strength in the adsorption medium on desip-

142

˚ S. Akerman et al. / European Journal of Pharmaceutical Sciences 9 (1999) 137 – 143

ramine and thioridazine adsorption was dissimilar may be attributed to the higher affinity of thioridazine for the membrane as discussed earlier.

4.3. Effect of albumin on drug adsorption

Fig. 3. Effect of albumin on (A) desipramine and (B) thioridazine adsorption onto the PAA–PVDF membrane at pH 7.0 and at ionic strength of 0.2. Mean values6S.D are shown (n53).

Albumin reduced adsorption of thioridazine and desipramine onto membrane (Fig. 3). Wassel and Embery (1997) have proposed that decreased adsorption of the charged drug onto the solid surface may be due to either the competition between binding sites on the oppositely charged solid surfaces or due to complex formation which may reduce the drug adsorption by the steric effects. In the present study, albumin was not adsorbed onto the membrane in the absence or presence of the drug. Consequently, it can be predicted that albumin did not affect the binding capacity of the membrane or reduce the drug adsorption via steric effects. Since albumin (pI 4.7–4.9) binds both desipramine and thioridazine at physiological pH (Table 1), the reduced drug adsorption onto the PAA– PVDF membrane in the presence of albumin at pH 7.0 was most probably due to the distribution of the drug between albumin and the PAA–PVDF membrane. Since albumin binds thioridazine more extensively than desipramine (Table 1), albumin had greater impact on the extent of thioridazine adsorption. However, the time required to achieve the equilibrium was about 35 h for both drugs. Assuming that the drug is bound by albumin, it is possible that accumulation of a cationic drug onto the surface of PAA–PVDF membrane may be avoided in vivo. The earlier results have shown that PAA–PVDF membrane binds many drugs and releases the bound drug from ˚ the membrane in acidic solutions (Akerman et al., 1998). The present study shows that PAA–PVDF membrane does not bind albumin suggesting that PAA–PVDF membrane might be a suitable membrane for separating drugs from proteinaceous substances for subsequent monitoring and evaluation.

5. Conclusions

Fig. 4. Effect of the concentration of albumin on thioridazine adsorption onto the PAA–PVDF membrane at pH 7.0 and at ionic strength of 0.2. Mean values6S.D are shown (n53).

The effects of conformation of the grafted PAA chains, physicochemical properties of the drug and the presence of albumin on interactions between the drug and the PAA– PVDF membrane were studied. The expanded conformation of PAA chains enhanced the adsorption of the drugs onto the membrane. In addition, the extent of dissociation of PAA chains affects strongly the extent of adsorption in the case of cationic drug that emphasize the importance of ionic interactions between the drug and the membrane in the adsorption process. Further, the lipophilicity of the cationic drug enhanced its adsorption onto the membrane. The presence of albumin in adsorption medium diminished the adsorption of lipophilic model drugs onto the membrane. Albumin was not adsorbed onto the membrane.

˚ S. Akerman et al. / European Journal of Pharmaceutical Sciences 9 (1999) 137 – 143

Acknowledgements The authors thank Professor Arto Urtti, University of Kuopio, for valuable discussions during this work. The financial support from TEKES (the Technology Development Centre in Finland) is gratefully acknowledged. This study was supported also by grants from The Savo Foundation for Advanced Technology, The Foundation for Finnish Inventions, Finnish Cultural Foundation and Re˚ search and Science Foundation of Farmos (SA).

References ˚ ¨ Akerman, S., Viinikka, P., Svarfvar, B., Jarvinen, K., Kontturi, K., ¨ Nasman, J., Urtti, A., Paronen, P., 1998. Transport of drugs across porous ion exchange membranes. J. Control. Release 50, 153–166. Bettini, R., Colombo, P., Peppas, N.A., 1995. Solubility effects of drug transport through pH-sensitive, swelling-controlled release systems: Transport of theophylline and metoclopramide monohydrochloride. J. Control. Release 37, 105–111. Chang, C.-M., Bodmeier, R., 1997. Binding of drugs to monoglyceridebased drug delivery systems. Int. J. Pharm. 147, 135–142. Charman, W.N., Christy, D.P., Geunin, E.P., Monkhouse, D.C., 1991. Interaction between calcium, a model divalent cation, and a range of poly(acrylic acid) resins as a function of solution pH. Drug. Dev. Ind. Pharm. 17, 271–280. Gal, P., Cone, M.H., 1988. Antiepileptic drugs. In: Taylor, W.J., Diers Caviness, M.H. (Eds.), A Textbook for the Clinical Applications of Therapeutic Drug Monitoring, Irving, Texas, p. 240. Hansch, C., 1990. Comprehensive Medical Chemistry, Vol. 6, Pergamon, Oxford.

143

¨ ¨ Hautojarvi, J., Kontturi, K., Nasman, J.H., Svarfvar, B.L., Viinikka, P., Vuoristo, M., 1996. Characterization of graft modified porous polymer membranes. Ind. Eng. Chem. Res. 35, 450–457. Heller, J., 1993. Modulated release from drug delivery devices. Crit. Rev. Ther. Drug Carrier Syst. 10, 253–305. Ito, Y., Casolaro, M., Kono, K., Imanishi, Y., 1989. An insulin-releasing system that is responsive to glucose. J. Control. Release 10, 195–203. Iwata, H., Amemiya, H., Hata, T., Matsuda, T., Takano, H., Akutsu, T., 1988. Development of novel semipermeable membranes for self regulated insulin delivery systems. Proc. Int. Symp. Control. Release Bioact. Mater. 15, 170–171. Jenquin, M.R., Liebowittz, S.M., Sarabia, R.E., McGinity, J.W., 1990. Physical and chemical factors influencing the release of drugs from acrylic resin films. J. Pharm. Sci. 79, 811–816. ˚ ¨ Jarvinen, K., Akerman, S., Svarfvar, B., Tarvainen, T., Viinikka, P., Paronen, P., 1998. Drug release from pH and ionic strength responsive poly(acrylic acid) grafted poly(vinylidene fluoride) membrane bags in vitro. Pharm. Res. 15, 802–805. Kost, J., Langer, R., 1992. Responsive polymer systems for controlled delivery of therapeutics. TIBTECH 10, 127–131. Kriwet, B., Kissel, T., 1996. Interactions between bioadhesive poly(acrylic acid) and calcium ions. Int. J. Pharm. 147, 135–145. Miyajima, M., Koshika, A., Okada, J., Kusai, A., Ikeda, M., 1998. The effects of drug physico–chemical properties on release from copoly(lactic / glycolic acid) matrix. Int. J. Pharm. 169, 255–263. Moffat, A.C., 1986. Clarke’s isolation and identification of drugs, second ed., Pharmaceutical Press, London. Okada, S., Nakahara, H., Isaka, H., 1987. Adsorption of drugs on microcrystalline cellulose suspended in aqueous solutions. Chem. Pharm. Bull. 35, 761–768. Park, H., Robinson, J.R., 1987. Mechanisms of mucoadhesion of poly(acrylic acid) hydrogels. Pharm. Res. 4, 457–464. Wassel, D.T.H., Embery, G., 1997. Adsorption of chondroidin-4-sulphate and heparin onto titanium: effect of bovine serum albumin. Biomaterials 18, 1121–1126.