Wear 251 (2001) 1213–1217

Do current wear particle isolation procedures underestimate the number ofparticles generated by prosthetic bearing components?

Marcus Scott∗, Kirstin Widding, Shilesh JaniOrthopaedic Division, Smith and Nephew Inc., 1450 Brooks Road, Memphis, TN 38116, USA

Abstract

Hip simulator serum samples containing ultra-high molecular weight polyethylene (UHMWPE) wear debris were digested in acid, andreplicate digests were filtered through either a 0.2 or a 0.05�m pore size membrane. The recovered particles were characterized usingFourier transform-infrared spectroscopy (FT-IR) and scanning electron microscopy (SEM). Debris deposited on both the 0.2 and 0.05�mmembranes were identified as UHMWPE by FT-IR and were predominantly submicron and round, with occasional elongated fibrils. Themean and median diameters of the particles on the 0.05�m membranes were significantly lower than those of the particles on the 0.2�mmembranes. Over half of the particles on the 0.2�m membranes had diameters which were below the specified pore size, whereas only asmall percentage (2.8%) of particles on the 0.05�m membranes were smaller than the specified pore size. More than twice as many particleswere recovered on the 0.05�m membranes than the 0.2�m membranes. These findings indicate that a substantial number of wear particlespassed freely through the pores of the 0.2�m membranes, which resulted in an underestimation of particle number and an overestimation ofparticle size. Because the cellular response to wear debris has been found to be dependent upon particle number and size, among other factors,the introduction of a new orthopaedic bearing material should be supported by an accurate description of wear particle parameters. To ensurean accurate description of particle characteristics, it is recommended that filter membranes with very fine pore sizes (at most 0.05�m) beused to isolate UHMWPE wear debris from joint simulator serum and periprosthetic tissue. © 2001 Elsevier Science B.V. All rights reserved.

Keywords:Wear; Debris; UHMWPE; Joint replacement

1. Introduction

Osteolysis is a common long-term complication in totalhip replacement (THR) [1] and has been linked to wear de-bris generated from ultra-high molecular weight polyethy-lene (UHMWPE) acetabular liners [2–5]. While the responseof periprosthetic tissue to wear debris is not fully under-stood, macrophage response to particulate wear debris isbelieved to be an important factor in osteolysis [6–9]. Itis well established that the cellular response to wear debrisis dependent upon particle number, shape, size, surface area,and material chemistry, among other factors [10–12]. Theintroduction of new bearing materials, such as crosslinkedUHMWPE, should therefore be supported by accurate de-scriptions of the number, size distribution, surface area, andvolume of wear particles generated.

Various techniques have been developed to isolateUHMWPE wear particles from periprosthetic tissue andjoint simulator serum. Common protocols involve digestionof tissue or serum samples in either a strong base or acid,followed by filtration of the digests through filter mem-

∗ Corresponding author. Tel.:+1-901-399-5553; fax:+1-901-399-6020.E-mail address:marcus.scott@smith-nephew.com (M. Scott).

branes with a pore size of 0.2�m [13–15]. A scanningelectron microscope (SEM) is used to determine the num-bers, sizes, and shapes of particles deposited on the filtermembrane. Previous analyses of particles recovered fromthe periprosthetic tissues of THR patients and from hipsimulator serum indicated that the mode of the particle sizedistribution was at or below the pore size (0.2�m) of thefilter membrane [13]. Thus, a significant number of particleshaving a diameter below 0.2�m may have passed throughthe filter pores and not been detected during the analyses.It is therefore hypothesized that the number of UHMWPEparticles generated by THR bearing components are un-derestimated by particle isolation techniques which involvefiltration of debris through a 0.2�m filter membrane.

To test the above hypothesis, UHMWPE liners were artic-ulated against CoCrMo heads on an anatomic hip simulatorup to 12 million cycles. Serum samples were periodicallyharvested and subjected to a validated acid digestion method[16]. Replicate digests were vacuum filtered through eithera 0.2 or 0.05�m filter membrane, and the wear particles de-posited on the filter membranes were characterized using anSEM. Relative differences in the particle number and sizedistribution were determined for debris deposited on 0.2 and0.05�m filter membranes.

0043-1648/01/$ – see front matter © 2001 Elsevier Science B.V. All rights reserved.PII: S0043-1648(01)00762-1

1214 M. Scott et al. / Wear 251 (2001) 1213–1217

2. Materials and methods

2.1. Hip simulator specimens and parameters

Commercially available acetabular liners were machinedfrom ram-extruded GUR 1050 UHMWPE (Poly-Hi Solidur,Ft. Wayne, IN) and sterilized using ethylene oxide. Theliners were inserted into Ti–6Al–4V (ISO 5832) acetabu-lar shells and tested against 32 mm diameter CoCr femoralheads (ASTM F799). The bearing couples (n = 3) weretested under physiological loading and motion conditions[17–19] on a 12-station AMTI (Watertown, MA) hip simu-lator. Testing was conducted to 12 million cycles at a cyclicfrequency of 1 Hz. The test lubricant was bovine serum(Hyclone Laboratories, Logan, UT) which contained 0.2%sodium azide and 20 mM EDTA. The test serum was re-placed approximately every 500,000 cycles.

2.2. Isolation of UHMWPE particles

Eight serum samples were harvested during the 12 mil-lion cycle test. The test interval (in cycles) for each serumsample is listed in Table 1. Each sample was collected in anErlenmeyer flask containing a stirbar and stirred overnightat 350 rpm. An amount of 10 ml of serum was then addedto 40 ml of 37% w/v HCl and stirred for 1 h at 50◦C. Anamount of 1 ml of the digested solution was added to 100 mlof methanol, which was then filtered through a 0.2�m poly-carbonate filter membrane. A replicate digest was filteredthrough a 0.05�m membrane for each serum sample.

2.3. Characterization of particle size and number

Each filter membrane was mounted on an aluminumstub, sputter-coated with gold, and examined using an SEM(S360, Leica Inc., Deefield, IL). Images were taken at amagnification of 10,000× and transferred to a digital imag-ing system (eXL II, Oxford Instruments Ltd., England).A minimum of 20 fields of view were analyzed per filter

Table 1Data on serum samples harvested from a hip simulator

Serum sample Testing interval (106 cycles) NFa NC

b × 106

0.2�m filters 0.05�m filters 0.2�m filters 0.05�m filters

1 0.580–1.064 49.6 157.5 2.7 8.72 4.021–4.461 48.2 173.5 2.9 10.53 8.006–8.628 54.7 117.9 2.3 5.14 9.228–9.717 49.2 98.5 2.7 5.45 10.289–10.837 47.8 137.3 2.3 6.76 10.289–10.837 44.7 105.3 2.2 5.17 11.369–11.972 64.9 102.8 2.9 4.58 11.369–11.972 54.6 117.1 2.4 5.2

Average 51.7 126.2 2.6 6.4S.D. 6.3 27.4 0.2 2.2

a Mean number of particles per field of view (10,000× magnification).b Mean number of particles generated per cycle.

membrane. Particle diameter (DP) was calculated based onthe projected area (A) of each particle. This parameter wasbased on circular geometry and calculated as follows:

DP = 2

(A

π

)1/2

. (1)

For each filter membrane, the mean number of particlesper field of view (NF) was determined, and the number ofparticles generated per cycle of testing (NC) was calculatedas follows:

NC = NFAFILTER

AFIELD

d

c, (2)

whereAFILTER is the area of filter membrane (962 mm2),AFIELD the area of field of view (9.0 × 10−5 mm2), d thedilution ratio (2500), andc the number of test cycles.

For each type of filter membrane, the data from the differ-ent digests were pooled. The particle diameter data were pre-sented as the mean, median, mode, and standard deviation.The number of particles per cycle was presented as the meanand standard deviation. Significant differences (ANOVA;α = 0.05) in mean particle parameters were determined be-tween the particles deposited on the 0.2 and 0.05�m filtermembranes. The Kruskal–Wallis test was used to determinesignificant differences in median particle diameter betweenthe two filter membranes.

2.4. Reproducibility of particle isolation method

For the acid digestion/vacuum filtration method used inthis study, a strong linear correlation has been demonstratedbetween measured wear volume (as determined gravimetri-cally) and the volume of particles recovered from a total jointsimulator [16]. For digests filtered through a 0.05�m filtermembrane, interobserver reproducibility has been found tobe within 10% of the mean value for each of the followingparameters: (i) number of particles generated per cycle oftesting; (ii) mean particle diameter (Table 2).

M. Scott et al. / Wear 251 (2001) 1213–1217 1215

Table 2Wear particle data demonstrating interobserver reproduciblity for two serum samples harvested from a hip simulator (mean± S.D.)

Serum Sample NFa Mean particle diameter (�m)

Observer 1 Observer 2 Observer 1 Observer 2

A 123.6 ± 20.8 123.8± 45.0 0.11± 0.12 0.12± 0.14B 76.3 ± 12.0 78.1± 21.2 0.20± 0.29 0.22± 0.32

a Mean number of particles per field of view (10,000× magnification).

2.5. Verification of particle identity

Fourier transform infrared spectroscopy (FT-IR) wasperformed to determine the identity of wear debris fromthree of the serum samples. In each case, approximately1 mg of particles was transferred from the filter membranesonto a KBr disk and identified using an FT-IR spectrometerwith an attached infrared microscope (FTS165 spectrome-ter, UMA250 microscope, Bio-Rad Laboratories, Hercules,CA). The FT-IR spectra of the particles isolated from serumwere compared with that of a commercially available HDPEpowder (Shamrock Technologies Inc., Newark, NJ).

3. Results

SEM analysis revealed that the recovered wear particleswere distributed uniformly on both types of filter mem-branes (Figs. 1 and 2). Minimal agglomeration of particleswas observed. The wear particles deposited on both the 0.2and 0.05�m filter membranes were predominantly submi-cron and round in appearance. Elongated fibrils (3–10�min length) were occasionally observed on both types of fil-ter membranes. For the data pooled from all eight serumsamples, the total number of particles imaged on the 0.2and 0.05�m filters were 8272 and 20,197, respectively.

Fig. 1. Scanning electron micrograph of wear debris isolated from hipsimulator serum and recovered on a 0.2�m pore size filter membrane.Scale marker represents a length of 2�m.

Fig. 2. Scanning electron micrograph of wear debris isolated from hipsimulator serum and recovered on a 0.05�m pore size filter membrane.Scale marker represents a length of 2�m.

The size distributions for the particles deposited on the0.2 and 0.05�m filters are presented in Figs. 3 and 4, re-spectively. For particles isolated on the 0.2�m filters, themean (0.23�m) of the size distribution was above the spec-ified membrane pore size, whereas the median (0.19�m)was below the specified pore size. The peak (mode) of thedistribution occurred well below the 0.2�m pore size at0.13�m. Over half (52.2%) of the total number of particlesdetected had diameters below the filter pore size of 0.2�m.For the particles deposited on the 0.05�m filters, the mean(0.19�m) and median mean (0.18�m) diameters were well

Fig. 3. Size distribution of the particles recovered on the 0.2�m filtermembranes.

1216 M. Scott et al. / Wear 251 (2001) 1213–1217

Fig. 4. Size distribution of the particles recovered on the 0.05�m filtermembranes.

above the specified membrane pore size. No single dominantpeak occurred in the size distribution, with the majority ofparticles evenly distributed between 0.08 and 0.25�m. Only2.8% of the total particles detected had diameters below thefilter pore size of 0.05�m. The mean and median diametersof the particles deposited on the 0.05�m membranes weresignificantly lower (P < 0.001) than those of particles onthe 0.2�m filter.

The 0.05�m filter membranes contained a greater numberof wear particles than the 0.2�m membranes (Table 1). Themean number of particles generated per cycle was 6.4 ×106 ± 2.2 × 106 for the serum digests passed through the0.05�m filters and was 2.6×106±0.2×106 for the digestsfiltered through the 0.2�m membranes. This difference wasstatistically significant (P = 0.002).

The FT-IR spectra of the particles recovered from the hipsimulator sera were similar to that of the reference HDPEmaterial and consistent with UHMWPE in that they hadmajor peaks around 2917, 2850, 1470, and 721 cm−1 (Fig. 5)[20]. Extraneous peaks and valleys were found to correspondwith the peak positions of the KBr disk. No evidence ofimpurities, such as filter material, debris from the tubingthrough which serum was circulated, or undigested proteins,was found.

Fig. 5. FT-IR spectra of particles recovered from serum, HDPE referencematerial, and the KBr background.

4. Discussion

In the current study, an acid digestion/vacuum filtrationprotocol was used to recover UHWMPE wear debris fromhip simulator serum. The wear particles recovered on the 0.2and 0.05�m filter membranes were spatially distributed ina uniform manner. Agglomeration of particles, as seen in aprevious wear debris isolation study [21], was minimal. Be-cause individual particles were clearly discernible from otherparticles and distributed uniformly on the filter membranes,sampling errors were minimized, leading to a more accuratedetermination of particle count and size distribution.

Previous studies which utilized base digestion techniqueshave shown that UHMWPE wear particles from THRperiprosthetic tissue and hip simulator serum are predomi-nantly submicron and round, with occasional elongated fib-rils [13,14,21]. The particles recovered in the current studydemonstrated similar morphologies, suggesting that the aciddigestion method used here and base digestion techniquesyield UHMWPE particles with similar shape. McKellopet al. [13] also reported the size distribution of particles re-covered from hip simulator serum using a 0.2�m filtrationstep and found that the majority of the particles were lessthan 0.2�m in diameter. In the current study, the size dis-tribution of the particles isolated from hip simulator serumand captured on 0.2�m filter membranes was very similarto that reported by McKellop et al. [13]. In both cases, themedian and mode of the particle size distribution was belowthe specified filter pore size. It therefore appears that boththe acid digestion technique and base digestion techniquesyield UHMWPE particles with similar size distributions.

Over twice as many particles were recovered on the0.05�m membranes as on the 0.2�m membranes. Overhalf the particles deposited on the 0.2�m membranes haddiameters that were below the specified pore size, whereasonly a small fraction (2.8%) of the particles on the 0.05�mwere below the specified pore size. These results indicatethat a substantial portion of wear particles passed freelythrough 0.2�m filter membranes, which led to a consid-erable underestimation of the number, and consequentlythe surface area and volume, of the UHMWPE particles.Recently, Green et al. [22] found that smaller UHMWPEparticles (0.24�m) produced bone resorbing activity invitro at a lower volumetric dose than larger particles (0.45and 1.71�m). This evidence suggests that finer wear par-ticles may elicit a greater macrophage response than largerparticles. Thus, finer wear debris generated at orthopaedicbearing couples should be fully characterized to accuratelypredict macrophage response. This is particularly importantfor new bearing materials, such as crosslinked UHMWPE,which have been reported to generate smaller wear particles(mean diameter of less than 0.1�m) than conventional UH-WMPE liners when tested in a hip simulator [23]. This un-derscores the importance of using finer pore size (≤0.05�m)filter membranes to isolate and characterize wear debrisgenerated from new orthopaedic bearing materials.

M. Scott et al. / Wear 251 (2001) 1213–1217 1217

The acid digestion/filtration method employed hereand base digestion/centrifugation/filtration techniques[13,14,21] used by other laboratories appear to yield pureUHMWPE particles with similar morphologies and sizedistributions. The isolation protocol described here did notrequire the use of centrifugation equipment, and thus in-volved fewer steps and less time. The entire digestion andfiltration procedure took approximately 75 min when thedigests were filtered through 0.2�m filter membranes. Fil-tering the digests through 0.05�m filter membranes addedonly 5 min to the procedure with no significant increase inthe cost of materials and equipment. This increase in proce-dural time was well justified due to the fact that the numberof particles recovered from hip simulator serum, and conse-quently the size distribution, was strongly dependent uponthe pore size of the filter membrane used. To ensure accuratecharacterization of particle number and size, it is thereforerecommended that filter membranes with very fine poresizes (at most 0.05�m) be used to isolate UHMWPE weardebris from joint simulator serum and periprosthetic tissue.

5. Conclusions

When vacuum filtration is used to isolate UHMWPE weardebris from digested hip simulator serum, the number andsize distribution of recovered particles is strongly dependentupon the pore size of the filter membrane. Filtering digestedserum through a 0.2�m filter underestimates the number,and consequently the surface area and volume of finer-sizedparticles. A significantly greater number of finer-sized parti-cles can be isolated and analyzed by filtering digested serumthrough 0.05�m filters. Because fine wear particles havebeen found to greatly influence the macrophage responseto wear debris, it is recommended that filter membraneswith very fine pore sizes (at most 0.05�m) be used to iso-late UHMWPE wear debris from joint simulator serum andperiprosthetic tissue.

Acknowledgements

Grateful acknowledgement is extended to Willard Sauer,former manager of Tribology at Smith and Nephew. WillardSauer developed tribological testing capabilities and exper-tise, and made significant contributions to this study.

References

[1] W.H. Harris, Clin. Orthopaedics Rel. Res. 311 (1995) 46–53.[2] H.C. Amstutz, P. Campbell, N. Kossovsky, I.C. Clarke, Clin.

Orthopaedics Rel. Res. 276 (1992) 7–18.[3] T.P. Schmalzried, M. Jasty, W.H. Harris, J. Bone Joint Surg. 74A

(1992) 849–863.[4] H.G. Willert, H. Bertram, G.H. Buchhorn, Clin. Orthopaedics Rel.

Res. 258 (1990) 95–107.[5] S.R. Goldring, A.L. Schiller, M. Roelke, J. Bone Joint Surg. 65A

(1983) 575–584.[6] S.B. Goodman, P. Huie, Y. Song, D. Schurman, W. Maloney, S.

Woolson, R. Sibley, J. Bone Joint Surg. 80B (1998) 531–539.[7] M. Jasty, C. Bragdon, W. Jiranek, H. Chandler, W. Maloney, W.H.

Harris, Clin. Orthopaedics Rel. Res. 308 (1994) 111–126.[8] J. Chiba, H.E. Rubash, K.J. Kim, Y. Iwaki, Clin. Orthopaedics Rel.

Res. 300 (1994) 304–312.[9] W.A. Jiranek, M. Machado, M. Jasty, D. Jevsevar, H.J. Wolfe, S.R.

Goldring, M.J. Goldberg, W.H. Harris, J. Bone Joint Surg. 75A(1993) 863–879.

[10] T.R. Green, J. Fisher, M. Stone, B.M. Wroblewski, E. Ingham,Biomaterials 19 (1998) 2297–2302.

[11] O. Gonzalez, R.L. Smith, S.B. Goodman, J. Biomed. Mater. Res. 30(1996) 463–473.

[12] A.S. Shanbhag, J.J. Jacobs, J. Black, J.O. Galante, T.T. Giant, J.Biomed. Mater. Res. 28 (1994) 81–90.

[13] H.A. McKellop, P. Campbell, S.H. Park, T.P. Schmalzried, P. Grigoris,H.C. Amstutz, A. Sarmiento, Clin. Orthopaedics Rel. Res. 311 (1995)3–20.

[14] P. Campbell, S. Ma, B. Yeom, H. McKellop, T.P. Schmalzried, H.C.Amstutz, J. Biomed. Mater. Res. 29 (1995) 127–131.

[15] S. Niedzwiecki, J. Short, S. Jani, W. Sauer, C. Klapperich, M. Ries,L. Pruitt, in: Proceedings of the Transactions of the 25th Societyfor Biomaterials, Providence, Rhode Island, USA.; April 28–May 2,1999, 150 pp.

[16] M. Scott, H. Forster, S. Jani, K. Vadodaria, W. Sauer, M. Anthony,in: Proceedings of the Transactions of the Sixth World BiomaterialsCongress, Kamuela, Hawaii, USA.; May 15–20, 2000, 177 pp.

[17] G. Bergmann, R.F. Grachen, A. Rohlmann, J. Biomech. 26 (1993)969–990.

[18] R.C. Johnston, G.L. Smidt, J. Bone Joint Surg. 51 (1969) 1083–1094.[19] ISO/CD 14242-1.2, Implants for Surgery — Wear of Total Hip

Prostheses. Part I: Loading and Displacement Parameters of WearTesting Machines and Corresponding Environmental Conditions forTest, Draft Standard, October 1997.

[20] P.C. Painter, M.M. Coleman, J.L. Koenig, The Theory of VibrationalSpectroscopy and its Application to Polymeric Materials, Wiley, NewYork, 1982, 252 pp.

[21] A.S. Shanbhag, J.J. Jacobs, T.T. Glant, J.L. Gilbert, J. Black, J.O.Galante, J. Bone Joint Surg. 76B (1994) 60–67.

[22] T.R. Green, J. Fisher, J.B. Matthews, M.H. Stone, E. Ingham, J.Biomed. Mater. Res. 53 (2000) 490–497.

[23] S. Bhambri, M. Laurent, P. Campbell, L. Gilbertson, S. Lin, in:Proceedings of the Transactions of the 45th Orthopaedic ResearchSociety, 1999, 838 pp.