Tradeoffs Amongst Fatigue Wear and Oxidation Resistance of Cross-linked Ultra-high Molecular Weight Polyethylene

J O U R N A L O F T H E M E C H A N I C A L B E H AV I O R O F B I O M E D I C A L M A T E R I A L S 4 ( 2 0 1 1 ) 1 0 3 3 – 1 0 4 5

available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/jmbbm

Research paper

Tradeoffs amongst fatigue, wear, and oxidation resistance ofcross-linked ultra-high molecular weight polyethylene

Sara A. Atwooda,∗, Douglas W. Van Cittersb, Eli W. Pattena, Jevan Furmanskia,Michael D. Riesc, Lisa A. Pruitta

aUniversity of California, Berkeley, Department of Mechanical Engineering, 2121 Etcheverry Hall, Berkeley, CA 94720, USAb Thayer School of Engineering, Dartmouth College, 8000 Cummings Hall, Hanover, NH 03755, USAcUniversity of California, San Francisco, Department of Orthopaedic Surgery, 500 Parnassis Ave., MU 320-W, San Francisco, CA 94143, USA

A R T I C L E I N F O

Article history:

Received 27 November 2010

Received in revised form

1 March 2011

Accepted 4 March 2011

Published online 11 March 2011

Keywords:

Fatigue

Wear

Oxidation

Polyethylene (UHMWPE)

Microstructure

A B S T R A C T

This study evaluated the tradeoffs amongst fatigue crack propagation resistance, wear

resistance, and oxidative stability in a wide variety of clinically-relevant cross-linked ultra-

high molecular weight polyethylene. Highly cross-linked re-melted materials showed good

oxidation and wear performance, but diminished fatigue crack propagation resistance.

Highly cross-linked annealed materials showed good wear and fatigue performance, but

poor oxidation resistance. Moderately cross-linked re-melted materials showed good oxidation

resistance, but moderate wear and fatigue resistance. Increasing radiation dose increased

wear resistance but decreased fatigue crack propagation resistance. Annealing reduced

fatigue resistance less than re-melting, but left materials susceptible to oxidation. This

appears to occur because annealing below the melting temperature after cross-linking

increased the volume fraction and size of lamellae, but failed to neutralize all free

radicals. Alternately, re-melting after cross-linking appeared to eliminate free radicals, but,

restricted by the network of cross-links, the re-formed lamellae were fewer and smaller

in size which resulted in poor fatigue crack propagation resistance. This is the first study

to simultaneously evaluate fatigue crack propagation, wear, oxidation, and microstructure

in a wide variety of clinically-relevant ultra-high. The tradeoff we have shown in fatigue,

wear, and oxidation performance is critical to the material’s long-term success in total joint

replacements.c⃝ 2011 Elsevier Ltd. All rights reserved.

n

.)

d

∗ Corresponding address: Elizabethtown College, Department of EngiTel.: +1 817 301 6501; fax: +1 717 361 4767.

E-mail addresses: [email protected] (S.A. Atwood), [email protected] (E.W. Patten), [email protected] (J. Furmanski(L.A. Pruitt).

1751-6161/$ - see front matter c⃝ 2011 Elsevier Ltd. All rights reservedoi:10.1016/j.jmbbm.2011.03.012

eering and Physics, One Alpha Drive, Elizabethtown, PA 17022, USA.

[email protected] (D.W. Van Citters),, [email protected] (M.D. Ries), [email protected]

.

http://dx.doi.org/10.1016/j.jmbbm.2011.03.012

www.sciencedirect.com



http://www.elsevier.com/locate/jmbbm

mailto:[email protected]






http://dx.doi.org/10.1016/j.jmbbm.2011.03.012

1034 J O U R N A L O F T H E M E C H A N I C A L B E H AV I O R O F B I O M E D I C A L M A T E R I A L S 4 ( 2 0 1 1 ) 1 0 3 3 – 1 0 4 5

1. Introduction

Ultra-high molecular weight polyethylene (UHMWPE or ultra-high) often limits the longevity of total joint replacements dueto excessive wear and associated clinical complications suchas osteolysis (Bozic et al., 2009a,b; Harris, 2001). To mitigatesuch wear-related failure, manufacturers produced ultra-highthat was highly cross-linked, typically by gamma radiation(Muratoglu et al., 2001). Cross-linking was also coupled withsubsequent re-melting to neutralize free radicals that canlead to oxidative degradation of the material (Collier et al.,1996; Costa et al., 1998a; Edidin et al., 2000). However,cross-linking and re-melting decreased the resistance tofatigue crack propagation (Baker et al., 2003). This decreasedresistance to fatigue crack propagation has been implicatedin recent reports of catastrophic fractures of cross-linked re-melted ultra-high in total hip replacements (Furmanski et al.,2009; Tower et al., 2007). In an attempt to preserve adequateresistance to fatigue and fracture while maintaining wearresistance and oxidative stability, manufacturers producedultra-high that was either moderately cross-linked and re-melted, highly cross-linked and annealed below the meltingtemperature, or sequentially cross-linked and annealed. Thesuccess of such treatments with respect to fatigue crackpropagation, wear, and oxidation remains a subject of debate(Collier et al., 2003; Crowninshield and Muratoglu, 2008;Currier et al., 2007a; Dumbleton et al., 2006; Gencur et al.,2006; McKellop et al., 1999; Morrison and Jani, 2009; Wanget al., 2008) due to the paucity of full-spectrum mechanicalcharacterization studies that provide controlled comparisonsamongst multiple clinically-relevant ultra-high materials.

The purpose of this study was to evaluate the perfor-mance and elucidate the tradeoffs amongst fatigue crackpropagation resistance, wear resistance, and oxidative sta-bility in clinically-relevant cross-linked ultra-high. Addition-ally, we seek to provide insight into relationships amongstprocessing treatments, microstructure, and mechanical per-formance. For this purpose, we evaluated nine distinctcombinations of resin, radiation dose, and subsequent heattreatment. We compared the treatment groups on the basisof fatigue crack propagation resistance, wear resistance, ox-idative stability, tensile properties, and microstructure. Thiswork is the first study to simultaneously evaluate fatiguecrack propagation, wear, and oxidation in a large number ofultra-high groups similar to marketed materials from a vari-ety of device manufacturers.

2. Materials and methods

2.1. Study design

We evaluated nine distinct ultra-high groups, of which tworepresented untreated controls, three represented highly cross-linked re-melted materials (9 or 10 Mrad), two representedmoderately cross-linked re-meltedmaterials (5 and 7.5 Mrad),and two represented highly cross-linked annealed materials(9 Mrad). On these nine material groups we performedthe following tests in parallel: (1) fatigue tests using afracture mechanics approach to assess resistance to fatigue

Table 1 – Ultra-high molecular weight polyethylenegroups tested. Material groups are labeled with the key:resin-radiation dose-heat treatment.

Material group Resin Radiationdose Mrad

Heattreatment ◦C

1020-0-0 1020 None None1050-0-0 1050 None None1020-3 × 3-130 1020 3 × 3 3 × 130 for 8 h1020-9-130 1020 9 130 for 8 h1020-5-147 1020 5 147 for 2 h1020-7.5-147 1020 7.5 147 for 2 h1020-9-147 1020 9 147 for 2 h1020-10-147 1020 10 147 for 2 h1050-10-147 1050 10 147 for 2 h

crack propagation, (2) multidirectional pin on disk tests toevaluate wear rate, (3) artificial aging followed by absorbanceinfrared spectroscopy to measure oxidation, (4) tensile teststo determine yield strength, elastic modulus, and ultimatetrue tensile strength and strain, and (5) scanning electronmicroscopy, digital image analysis, and differential scanningcalorimetry to characterize the lamellar microstructure andcrystallinity. We also performed a statistical analysis onthe results of the mechanical tests and the microstructuralcharacterization to determine relationships between themechanical performance and microstructure.

2.2. Materials

We evaluated nine distinct groups of medical-grade ultra-high molecular weight polyethylene (ultra-high) that hadundergone clinically-relevant processing treatments at a de-vice manufacturer (Table 1). The groups include combinationsof base resin, radiation dose, and heat treatment that aresimilar to clinical materials from four major device man-ufacturers. Of the nine material groups, two are untreatedpolyethylene controls made from different orthopaedic graderesins (GUR 1020 and GUR 1050). These control materials dif-fer inmolecular weight and in consolidationmethod: the GUR1020 material has a molecular weight of 2–4 million g/moland is formed by compression molding, while the GUR 1050material has a molecular weight of 4–6 million g/mol and isformed by ram extrusion. The remaining ultra-high groupswere gamma-irradiated in one or multiple doses (with a dosetotaling 5–10 Mrad), and then heat-treated either above orbelow ultra-high’s melting temperature of 135 ◦C (Kurtz andSteven, 2009) (130 ◦C for 8 h or 147 ◦C for 2 h). All cross-linked groups were also either compression-molded GUR 1020or ram-extruded GUR 1050. The material groups are referredto throughout this work as RESIN—RADIATION DOSE (Mrad)—HEAT TREATMENT (◦C), for example 1020-9-130 represents aGUR 1020 resin irradiated to a dose of 9 Mrad and subse-quently annealed at 130 ◦C.

2.3. Methods

2.3.1. Fatigue crack propagation testingTo evaluate the materials’ resistance to fatigue crack propa-gation, we performed fatigue tests using a fracture mechanicsapproach on four to six compact tension specimens for each

J O U R N A L O F T H E M E C H A N I C A L B E H AV I O R O F B I O M E D I C A L M A T E R I A L S 4 ( 2 0 1 1 ) 1 0 3 3 – 1 0 4 5 1035

material group. A defect-tolerant approach measuring fatiguecrack propagation was considered rather than a total-life ap-proach including crack initiation because conditions exist inultra-high orthopaedic implants making them vulnerable toreadily initiating cracks. These conditions include design fea-tures that serve as stress concentrations (Furmanski et al.,2009), Hertzian contact that produces subsurface cracks dur-ing the wear process (Wang et al., 1995), and a known sus-ceptibility of ultra-high to crack propagation (Furmanski andPruitt, 2007).

The compact tension specimens (Baker et al., 2003) usedto measure crack propagation were machined with a 1 mmdeep, 40◦ side groove on both sides of the specimen in thecrack plane to allow for more accurate crack measurementand a more even distribution of stress through the thicknessof the specimen (Shih et al., 1977). The compact tensionspecimens used in the fatigue testing meet the ASTM E-647standard for fatigue characterization. Independent studies inour laboratory have shown that the specimen geometrymeetsplane strain conditions. The tip of the notch was sharpenedwith a razor blade before testing. The crack propagationdirection corresponded to the ram extrusion direction in theGUR 1050 materials. Fatigue tests were performed on anInstron 8871 servohydraulic load frame (Norwood, MA) usinga load-controlled sinusoidal wave function at a frequency of5 Hz. The fatigue tests were run under ambient conditionswith a room-temperature air-jet directed at the crack tip tomitigate hysteretic specimen heating. The sinusoidal loadwas applied at a constant load ratio of 0.1 (defined as theratio of the minimum load to the maximum load of thefatigue cycle). After 10,000 load cycles, the load was increased,maintaining a load ratio of 0.1 (for example, 30–300 N, then40–400 N, etc.). This process was repeated throughout thestable crack growth regime until the specimen fractured.Crack advance was quantified after each 10,000 cycles bymeasuring the distance between the crack tip and fiduciallinesmarked on the specimen surface using a high-resolutiondigital microscope consisting of a variable magnificationoptical system (Infinivar CFM-2/S, Boulder, Colorado, pixelsize 5 µm) and a digital CCD video camera (Sony XCD-SX910,Tokyo, Japan).

The crack advance data along with the prescribed loading,known specimen geometry, and measured number of cyclesallowed us to relate the crack growth per cycle (da/dN inmm/cycle) to the range of stress intensity driving the crackpropagation (1K in MPa

√m). The range of stress intensity is

defined as

1K = F1σ√

(πa) (1)

in which F is a specimen-specific geometrical factor describedpreviously (Baker et al., 2000), 1σ is the range of far-fieldapplied stress (MPa), and a is the crack length (m). In thestable crack growth regime, the Paris equation relates thestress intensity to the crack growth per cycle according to:

da/dN = C(1K)m (2)

in which C and m are parameters that depend on thematerial,environment, frequency, temperature and stress ratio. Ona bi-logarithmic plot of crack velocity versus the stressintensity range, m and C represent the slope and intercept,respectively.

Fig. 1 – Custom pin-on-disk tribotester: the vertical tableon top holds the ultra-high pins and is kept stationarywhile the horizontal table below holds the CoCr disk andmoves with X–Y motion control from a retrofitted CNCmilling machine. Loading is controlled using independentpneumatic actuators and is monitored by load cellsmounted below the disk holders.

We used linear regression to relate the logarithm of crackgrowth per cycle to the logarithm of stress intensity, andto statistically compare values of C and m (STATA v. 9,College Station, TX). A full linear regression model wasinitially fit including indicators and cross-products to allowfor statistical differences in intercept and slope amongstthe groups. Full versus restricted F-tests were performed todetermine whether various intercepts and slopes should bekept in the model. A cutoff of p < 0.05 was used to concludethat particular intercepts or slopes were statistically differentfrom others.

2.3.2. Wear testingTo evaluate the wear resistance of the materials, we per-formed multidirectional sliding wear tests with a custom pin-on-disk tribotester (Fig. 1) (Patten, 2008). The bearing couplecomprises a spherically-tipped ultra-high pin (3.28 mm ra-dius) against a flat CoCr disk (127 mm diameter). The tri-botester is a retrofitted vertical-knee milling machine withthe drilling head replaced with a vertical mounting table.The pins and loading system are attached to the vertical ta-ble, while the disks and load cells are attached to the hor-izontal table below. The pins are held in collets on verticalrails and the load is adjusted using individually-controlledpneumatic actuators. The ultra-high pins are held stationarythroughout the test while the horizontal table on which theCoCr disks are mounted moves in a defined x-y motion usingcomputer numerical control (National Instruments LabVIEWv8.5 and Motion Assistant v2.2, Austin, TX). The CoCr diskstranslate along a circular path (8 mm diameter) without ro-tation, achieving multidirectional sliding with cross-shear onthe ultra-high bearing surface.

The CoCr disks were polished to an arithmetic averageroughness of less than 0.03 µm as measured at multiplelocations using a stylus profilometer (Dektak IID, Sloan


Table 2 – Wear test parameters for multidirectionalpin-on-disk testing.

Wear test parameters

Wear path Circularly translating, 8 mmdiameter

Normal load 10–15 NMean contact pressure 20–30 MPaLinear speed 35 mm/sSliding distance 12.5 kmNumber of cycles 500,000Lubricant Bovine seruma

Environment ∼25 ◦C, Ambient

aDiluted 1:1 with deionized water, preserved with 0.1 wt% sodiumazide.

Technology Co., Santa Barbara, CA). Before testing, the ultra-high pins and CoCr disk were ultrasonically cleaned inacetone and deionized water. Two pins were tested for eachmaterial group.

The wear test conditions were chosen to be clinicallyrelevant and were previously validated (Table 2) (Klapperichet al., 1999; Zhou and Komvopoulos, 2005). The normalload of 12 ± 2.5 N results in a mean contract pressure of25 ± 5 MPa, which is similar to conditions found in totaljoint replacements (Bartel et al., 1986). The linear speedof 35 mm/s simulates speeds found in joint replacementsduring normal activity such as walking and running (Fisheret al., 1994). The lubricant was bovine serum diluted 1:1 withdeionized water and preserved with 0.1 wt% sodium azide. Alltests were performed in an ambient laboratory environment.Air and serum temperatures were monitored throughoutthe test; serum temperature was consistently 1◦–2◦ warmerfrom frictional heating. Tests were run for 500,000 cycles toestablish a steady-state wear rate after the run-in period.

The diameter of the wear scar on the pin wasmeasured every 50,000 cycles using the high-resolutiondigital microscope described above. The wear was calculatedvolumetrically as the volume of material lost from thespherically-ended ultra-high pin. The wear rate was thencalculated as the wear volume divided by the number ofcycles. Wear resistance was based on the steady-state wearrate achieved by the materials beyond 300,000 cycles. Thesteady-state wear data taken after 300,000 cycles (roughly72 h) is substantially beyond ultra-high’s transient creepperiod of about 24 h (Klapperich et al., 1999).

Analysis of variance (ANOVA) was used to determinestatistically significant differences amongst the steady-statewear rates of the material groups beyond 300,000 cycles(STATA v. 9). We adjusted for multiple post-hoc comparisonsusing the Student-Newman–Keuls (SNK) procedure. TheStudent-Newman–Keuls procedure increases the power ofdetecting a difference by arranging the means in increasingorder and performing one-tailed t-tests, while still controllingthe overall false-positive error (α = 0.05) for the family ofcomparisons spanning a given number of means (Glantz,2005). The minimum power to detect a 0.6 × 10−7 mm3/cycledifference (about twice the typical standard deviation) in wearrate is about 0.85, given 9 material groups, 6–8 data points pergroup, and an overall α = 0.05 (Glantz, 2005; Lenth, 2006–2009).

2.3.3. Oxidation following artificial agingTo assess the oxidative stability of the materials, we per-formed artificial aging in a previously-validated environmentfollowed by absorbance spectroscopy (Currier et al., 2007b).One sample for each group (10 mm × 10 mm × 10 mm)was placed in a pressure vessel with 3 atm O2 at 63 ◦Cfor 28 days (an environment less aggressive than ASTM-2003 Method A). Subsequently, oxidation in the materialwas measured by Fourier transform infrared spectroscopy on200 µm-thick cross-sections of each sample (Jung microtome,Heidelberg, Germany). Incorporation of oxygen into the ma-terial was measured using a Perkin Elmer AutoImage InfaredMicroscope (Waltham, MA) with 32 scans per 100 µm depthinterval, wavelength 2 cm−1, and aperture 100 µm2. The oxi-dation index was defined as themeasured 1715 cm−1 (ketone)peak height normalized to the 1368 cm−1 peak height (Currieret al., 2007b).

2.3.4. Tensile testingTo evaluate the materials’ mechanical properties, weperformed tensile tests on three dog-bone specimens for eachmaterial group. The dog-bone specimens were machined tothe ASTM type V geometry with a thickness of 1.5 mm. Beforetesting, dimensions of each specimen were measured usingdigital calipers (±0.01 mm). Tensile tests were performed onan Instron 8871 servohydraulic load frame (Norwood, MA)using displacement-control at a rate of 5 mm/min (Bakeret al., 2003). The tests were performed at room temperaturewith air jet cooling directed at the gage section of thespecimen. The load and displacement data were convertedinto engineering stress and strain using measured initialdimensions. The engineering stress and strain data were usedto determine the yield strength (where the stress decreasedslightly with increasing strain) and the elastic modulus(the secant modulus at 2% strain). The digital microscopedescribed above captured a sequence of images during thetest with a resolution of 5 µm, taken at a rate of approximatelyone image per second. Using the specimen dimensions in theimage captured just before failure, the ultimate true tensilestrength and ultimate true strain at failure were determined.

2.3.5. MicrostructureTo elucidate the relationship amongst microstructure,processing treatments, and performance, we assessed thematerials’ lamellar structure and crystallinity qualitativelyusing scanning electron microscopy and quantitatively usingimage analysis and differential scanning calorimetry. Themicrostructural parameters in this study are limited tolamellar properties and do not include amorphous propertiessuch as cross-link density. To obtain scanning electronmicrographs of the crystalline lamellae, cross-sections fromtwo samples of each material group (2 mm × 2 mm × 6 mm)were microtomed using a glass knife (Reichert Ultracut E,Depew, NY) and then subjected to a potassium permanganateetching procedure that preferentially removes the amorphousphase that occupies the space between the lamellae (Olleyand Bassett, 1977; Simis et al., 2006). The samples werethen sputter-coated with gold–palladium (Tousimis SputterCoater, Rockville, MD) and imaged using a field-emission


Fig. 2 – Image analysis software was used to digitally threshold and filter original scanning electron micrographs tovisualize cross-sections of representative lamellae. On each processed image, pixels were counted and scaleddimensionally to obtain distributions of lamellar cross-sectional area, thickness, and length.

scanning electron microscope (Hitachi S-5000, Pleasanton,CA) with an accelerating voltage of 30 kV. Digital images ofthe lamellae were taken at a magnification of 20,000 timeswith a resolution of 4 nm.

To quantify the lamellar dimensions observed in thescanning electron micrographs, we performed image analysison one representative image (5×5 µm) for eachmaterial group(National Instruments LabVIEW v. 8.5). The image analysisconsisted of a user-defined threshold of the image and aseries of standard filters that separated or eliminated largetangled clumps of lamellae, small round corners of lamellae,and lamellae touching the edge of the image (Fig. 2). A best-fit rectangle was fit to each remaining lamellar object andthe pixel dimensions of the rectangle were converted tonanometers using spatial calibration of the scale bar on theimage to determine lamellar thickness and length (resolutionof 4 nm). The individual objects were two-dimensional cross-sections of plate-like lamellae that were characterized by athickness (short dimension of best-fit rectangle), a length(long dimension of best-fit rectangle), and an area (number ofpixels of the object). Each image contained about 200 lamellarobjects upon which relative magnitudes and distributions forlamellar thickness, length, and area were based.

To assess the percent crystallinity, we performed differ-ential scanning calorimetry on three samples of each ma-terial group. Samples of approximately 10 milligrams weresubjected to a thermal scan from 50 to 180 ◦C at a rate of10 ◦C per minute (Perkin Elmer, Waltham, MA). The enthalpyof melting was determined by integrating the entire meltingendotherm from 80 to 160 ◦C and normalizing to the samplemass (TA Instruments Universal Analysis v. 3.1E, New Castle,DE). Percent crystallinity was calculated by normalizing theenthalpy of melting for a particular sample to that of a pureultra-high crystal (293 J/g) (Kurtz and Steven, 2009).

2.3.6. Statistical analysis of pair-wise correlations

To determine relationships amongst microstructural proper-ties and mechanical performance, we performed a statisticalanalysis using the non-parametric Spearman rank correla-tion coefficient (STATA v. 9). The non-parametric analysis ac-counts for small sample sizes and non-normal distributions

observed for some outcomes. For the estimates, we used me-dian values of mechanical and microstructural properties. Fa-tigue crack propagation resistance was quantified as the valuefor the range of stress intensity corresponding to a da/dN of10−5 mm/cycle (this captures the left-to-right shift of the fa-tigue curves in the Paris regime and is related to the stress in-tensity required for the inception of crack propagation (Bakeret al., 2003)).

3. Results

3.1. Fatigue crack propagation

Fatigue crack propagation resistance of the moderatelycross-linked re-melted materials and the highly cross-linkedannealed materials was greater than that of highly cross-linked re-melted materials, but was worse than that ofuntreated controls (Fig. 3(A)). Statistically, the slopes of allthe fatigue resistance lines were the same (p > 0.10,95%confidence interval: 7.3–8.4), but the intercepts were not.The fitted intercepts represent the left-to-right shift in thefatigue curve and are related to the inception stress intensity.The intercepts of the moderately cross-linked re-melted andhighly cross-linked annealed materials were not differentfrom one another (p > 0.20,95% CI: 1.02 to 4.07 × 10−6),but were significantly different from highly cross-linked re-melted materials (p < 0.001,95% CI: 4.10 to 17.9 × 10−6),and from untreated controls (p < 0.001,95% CI: 0.078 to0.153×10−6). This means that, in an idealized large laboratoryspecimen, at a stress intensity range of 1.3MPa

√m for 100,000

cycles (roughly one month of service for a joint replacement),a crack in an untreated material would grow about 0.1 mm,a crack in a moderately cross-linked or annealed materialwould grow about 1 mm, and a crack in highly cross-linkedre-melted material would grow about 10 mm.

Within the cross-linked re-melted materials, the fatigueresistance decreased with increasing radiation dose (Fig. 3(B))(the intercept, C, is significantly different for each highlycross-linked group, p < 0.001 from an F-test). For example,at a stress intensity range of 1.0 MPa

√m for 100,000 cycles,

a crack in a GUR 1020 7.5 Mrad material would grow about0.4 mm, while a crack in a GUR 1020 10 Mrad material wouldgrow about 3 mm.


Fig. 3 – Fatigue crack propagation data showing (A) fatigue resistance of moderately cross-linked re-melted and highlycross-linked annealed materials is increased compared to highly cross-linked re-melted materials, but decreased comparedto untreated controls and (B) within re-melted materials, fatigue resistance decreased with increasing radiation dose. Key:resin — radiation dose (Mrad) — subsequent thermal treatment (◦C).

Fig. 4 – Wear data showing (A) wear rates were substantially lower for all cross-linked ultra-high compared to untreated(n = 2 tests) and (B) steady-state wear rate depends primarily on radiation dose (beyond 300,000 cycles, n = 6–8 wear ratemeasurements). Key as in Fig. 2.

3.2. Wear

All the cross-linked ultra-high groups had significantlylower wear rates than untreated control materials (Table 3).All materials reached a steady-state wear rate afterapproximately 300,000 cycles (Fig. 4(A)). The moderatelycross-linked re-melted 5 Mrad material had a significantlyhigher wear rate than materials with higher radiation doses.There was no significant difference in wear rate for 9 Mradmaterials subjected to re-melting versus annealing. However,the sequentially-dosed ultra-high demonstrated the lowestwear rate of all the materials—significantly lower than thesingle dose 9 Mrad annealed material (Fig. 4(B)). The GUR 1050and GUR 1020 materials were not statistically different; norwere any of the highly cross-linked re-melted materials.

3.3. Oxidation following artificial aging

Oxidative stability was achieved by some materials but notby others (Fig. 5). The cross-linked annealed materials were

susceptible to oxidation with an oxidation index greater than0.5 after artificial aging. The cross-linked re-melted materialswere chemically stable after artificial aging (oxidation indexless than 0.1), along with the untreated materials.

3.4. Tensile testing

The mechanical properties (yield strength, elastic modulus,and ultimate true stress and strain) of the cross-linkedmaterials depended on the type of heat treatment, theradiation dose, and the resin and consolidation method.Elastic modulus and yield strength were higher for annealedmaterials and lower for re-melted materials, with secondarydependence on radiation dose. Alternately, the ultimateproperties generally decreased with increased radiation dose,regardless of heat treatment. Mechanical properties alsodepended on the resin and consolidation method. TheGUR 1050 materials had lower yield strength and elasticmodulus compared to GUR 1020materials with the same heat


Table 3 – Statistical comparisons between wear rates show that all cross-linked materials had significantly lower wearrates than the untreated control materials, while the 5 Mrad re-melted material had a significantly higher wear rate thanmaterials with higher radiation doses.

Significant Not significantComparison p-value Comparison p-value

1020-0-0 vs. all cross-linked <0.0001 1020-0-0 vs. 1050-0-0 0.10051050-0-0 vs. all cross-linked <0.0001 1020-10-147 vs. 1050-10-147 0.38831020-5-147 vs. 1020-7.5-147 0.0023 1020-9-130 vs. 1020-9-147 0.07371020-5-147 vs. 1020 − 3 × 3 − 130 0.0006

All other comparisons between cross-linkedmaterials

>0.05

1020-5-147 vs. 1020-9-130 0.00101020-5-147 vs. 1020-9-147 0.01331020-5-147 vs. 1020-10-147 0.03581020-5-147 vs. 1050-10-147 <0.00011020-3 × 3-130 vs. 1020-9-130 0.00181020-3 × 3-130 vs. 1020-7.5-130 0.0003

The p-value from the initial analysis of variance F-test is <0.0001. Post-hoc multiple comparisons were performed using the Student-Newman–Keuls test, with one-way p-values reported above.

Fig. 5 – Artificial aging data showing that annealedmaterials have an oxidation index greater than 0.5 whilere-melted and untreated materials remain chemicallyunchanged with an oxidation index less than 0.1. Agingwas performed in 3 atm O2 at 63 ◦C for 28 days (Currieret al., 2007b). Oxidation index is defined as the measured1715 cm−1 peak height (ketone) normalized to the1368 cm−1 peak height. Key as in Fig. 2.

treatment and radiation dose, but ultimate properties wereabout the same (Fig. 6).

3.5. Microstructure

Microstructural parameters such as lamellar size andpercent crystallinity also depended on the heat treatment,the radiation dose, and the resin. Annealing producedlarger lamellae similar in size to those of the untreatedcontrols with respect to the magnitude and distributionof lamellar thickness, length, and cross-sectional area.Annealed materials showed increased percent crystallinityby 10% compared to the untreated controls. Alternately, re-melted materials demonstrated lamellae smaller than thoseof the untreated controls (Fig. 7) with decreased percentcrystallinity by about 5%. Furthermore, in the re-melted

Fig. 6 – Tensile properties for all material groups. Yieldstrength and elastic modulus of cross-linked annealedmaterials are higher than untreated controls, whilecross-linked re-melted materials are lower. Ultimateproperties decrease with increasing radiation dose. Eachsample is represented by an O. The median is representedby—(n = 3 samples).

materials, the percent crystallinity decreased with increasing

radiation dose. With respect to resin, the GUR 1020 materials

had slightly smaller lamellae but higher crystallinity than


Fig. 7 – Microstructural characterization of four representative material groups: (a) uncross-linked control, (b) highlycross-linked annealed material, (c) moderately cross-linked re-melted material, (d) highly cross-linked re-melted material.Note the lamellar structure of the annealed material is similar to that of the untreated control, while that of the re-meltedmaterials is generally smaller. Scanning electron micrographs were obtained after amorphous regions were removed byetching. Scale bar in the lower right corner represents 1 µm for all images.

the GUR 1050 materials with the same heat treatment andradiation dose (Fig. 8).

3.6. Statistical analysis of pair-wise correlations

Non-parametric statistical analysis showed that the elasticmodulus and yield strength were highly-correlated withcrystallinity, while ultimate true stress and strain werehighly-correlated with the lamellar size parameter area(Table 3). Resistance to fatigue crack propagation wascorrelated with both crystallinity and with lamellar size andlength. Wear rate was not significantly correlated with any ofthe microstructural parameters. Lamellar thickness was notcorrelated with any mechanical properties, likely due to thesmall range and wide distributions of the lamellar thicknessmeasurements.

4. Discussion

There is a tradeoff amongst fatigue crack propagationresistance, wear resistance, and oxidative stability inclinically-relevant cross-linked ultra-high (Fig. 9). Highcross-linking combined with re-melting produces goodwear resistance and oxidative stability but relatively poorresistance to fatigue crack propagation. High cross-linkingcombined with annealing produces good resistance to wearand fatigue crack propagation but can leave the materialsusceptible to oxidation. Moderate cross-linking combinedwith re-melting produces good oxidative stability butmoderate resistance to fatigue crack propagation and wear.

Material behavior was dependent on both total radiationdose and heat treatment. Total radiation dose decreasedfatigue resistance and ultimate properties, but increasedwearresistance. Annealing, as opposed to re-melting, reducedfatigue resistance and elastic properties less than re-melting,but left materials susceptible to oxidation.

Within the highly cross-linked annealed materials,sequential doses of radiation and annealing did notsignificantly change resistance to fatigue or oxidation beyond

that of an equivalent total single dose. However, sequentialdoses of radiation and annealing did appear to improvethe wear resistance compared to the equivalent single dose.It is well known that increased thermal energy increasescross-link density (Muratoglu et al., 1999), and that increasedcross-link density increases multidirectional sliding wearresistance (Muratoglu et al., 1999; Wang, 2001). The sequentialdose annealed material is subjected to more thermal energyduring processing, which would affect both cross-link densityand wear behavior.

Resin and consolidation method also affected material be-havior. The two GUR 1020 materials had higher crystallinity,yield strength, modulus, and fatigue crack propagation resis-tance than GUR 1050 materials with the same heat treatmentand radiation dose. No statistical difference was found in thewear behavior of GUR 1020 versus GUR 1050 materials. Thedifferences in elastic and fatigue behavior may be due to thelower molecular weight of the GUR 1020 (2–4 million g/mol)compared to GUR 1050 (3–6 million g/mol). With a lowermolecular weight, the long-chain molecules are more mobileand may be more easily incorporated into lamellae. This re-sults in higher percent crystallinity, which is correlated withhigher yield strength, modulus and fatigue resistance. Thedifferences in resin are also confounded by differences in con-solidation method, which has been shown to impart direc-tionality and to effect fatigue crack propagation resistanceand tensile properties (Pruitt and Bailey, 1998). This direc-tionality explains the lower fatigue crack propagation resis-tance of the highly cross-linked GUR 1050material, whichwastested in the weak orientation in which crack propagation di-rection is aligned with the ram extrusion direction.

The effects of the processing treatments on the material’sbehavior can be elucidated by examining the microstructure.Cross-linking the ultra-high forms covalent bonds betweenthe long-chain molecules; this results in a networkedstructure in the amorphous phase. Irradiation for cross-linking also produces residual free radicals, which can leadto in vivo oxidative degradation if the free radicals are noteliminated by heat treatment (Collier et al., 2003; Currieret al., 2007b). Our results show that applying thermal energybelow themelting temperature after cross-linking (annealing)


Fig. 8 – Microstructural properties for all material groups.Crystallinity is higher for annealed materials and lower forre-melted materials compared to uncross-linked controls.Lamellar thickness, length, and size show a trend ofsmaller median lamellae with narrower distributions inre-melted materials. For crystallinity each sample isrepresented by an O with the median representedby—(n = 3 samples). For lamellar properties the boxrepresents the middle two quartiles separated by a medianline. The whiskers represent the upper and lower quartilesand outliers are represented by O (from image analysisperformed on one representative scanning electronmicrograph per material).

increased the volume fraction of lamellae, likely due to acombination of chain scission from radiation and increasedmobility of amorphous chains from annealing (Viano et al.,2007). However, this annealing process failed to neutralize allfree radicals. Alternatively, re-melting the materials appearedto eliminate free radicals, but, restricted by the network ofcross-links, the re-formed lamellae were fewer and smaller insize. The fewer, smaller lamellae coupled with the networkedamorphous phase were associated with decreased fatigueresistance and tensile properties.

The statistical relationships between fatigue resistance,tensile properties, and lamellar properties also add to ourunderstanding of micromechanisms of damage in ultra-high.Our non-parametric statistical analysis suggests that elasticparameters are strongly correlated with percent crystallinitywhile ultimate parameters are strongly correlated with

Fig. 9 – Schematic showing the tradeoffs in fatigue crackpropagation resistance, wear resistance, and oxidationresistance. The vertices of the colored triangles on eachaxis represent the relative performance in that category.The untreated controls have good fatigue and oxidationperformance, but poor wear resistance. The highlycross-linked re-melted materials have good oxidation andwear performance, but poor fatigue resistance. The highlycross-linked annealed materials have good wear andfatigue performance, but poor oxidation resistance. Themoderately cross-linked re-melted materials have goodoxidation resistance, but moderate wear and fatigueresistance.

lamellar size, in agreement with the literature (Medelet al., 2007; Ries and Pruitt, 2005; Simis et al., 2006). Wealso found that resistance to fatigue crack propagationis correlated with both percent crystallinity and lamellarsize. By extension then, superior fatigue crack propagationresistance is associated with both high elastic and highultimate properties. A high elastic modulus and high yieldstress may contribute to limiting the strains at the cracktip while dissipating more energy before yielding. Higherultimate true stresses and strains may contribute to moreenergy dissipation before ultimately failing on the microscaleat the crack tip. Resistance to fatigue crack propagation wasalso highly-correlated with a large lamellar cross-sectionalarea and with percent crystallinity, suggesting that morelamellae, large in both dimensions, may require more energyfor cracks to travel through or around (Table 4).

Wear rate was not significantly correlated to any of themicrostructural parameters in this study and was dominatedby radiation dose. In a similar study on the relationship ofmicrostructural parameters to rolling-sliding wear rate, VanCitters et al. (Van Citters et al., 2007) found a correlationbetween wear rate and lamellar size normalized by yieldstrength. Our data did not exhibit the same trend. However,the Van Citters study is based on delamination wear whichis ultimately a crack nucleation and propagation mechanism.Our results are consistent in demonstrating a correlationbetween fatigue processes and lamellar size. The primary


Table 4 – Spearman rank correlation coefficients for microstructural properties and mechanical performance measures.Note the elastic properties are correlated with percent crystallinity, the ultimate properties with lamellar size, resistanceto fatigue with both, and wear rate with none.

Microstructural property Resistance tofatigue ρ

Yield stress ρ Elasticmodulus ρ

Ultimate truestrength ρ

Ultimate truestrain ρ

Wearrate ρ

Crystallinity 0.85* 0.93** 0.98** 0.60 0.53 0.18Lamellar thickness 0.71 0.23 0.33 0.70 0.85 0.48Lamellar length 0.85* 0.55 0.72 0.72 0.73 0.48Lamellar area 0.93** 0.62 0.77 0.85* 0.85* 0.61

∗p < 0.01.∗∗p < 0.001.

wear mechanism in our study was multidirectional slidingwear. It has long been thought that multidirectional slidingwear is due to the alignment of lamellae in the slidingdirection, resulting in a weakened material under cross-shear (Edidin et al., 1999; Wang, 2001). Cross-linking hasbeen shown to inhibit the ability of the lamellae to align(Klapperich et al., 1999; Zhou and Komvopoulos, 2005), whichis in agreement with our observation that sliding wear ratedecreased with increased radiation dose. Our results suggestthat the number, size, and thickness of the lamellae are notas important for inhibiting sliding wear as is cross-linking theamorphous phase.

The principal strength of this study was the large numberof distinct clinically-relevant processing treatments includedand the full spectrum of mechanical characterizationperformed. For the same group of materials, we evaluatedfatigue crack propagation, wear, oxidation, tensile properties,and microstructure. We evaluated two untreated controlmaterials, two highly cross-linked annealed materials, twomoderately cross-linked re-melted materials, and threehighly cross-linked re-melted materials. We were able tocompare amongst materials with the same total radiationdose (9 Mrad) but subjected to single and sequentialannealing, as well as re-melting. We were also able tocompare amongst materials that had all been re-melted, butwere subjected to various total radiation doses. A furtherstrength of this study is that by evaluating microstructurealong with fatigue resistance, wear resistance, oxidativestability, and tensile properties, we were also able to explainrelationships between the treatments, microstructure, andthe resulting mechanical performance.

While the fatigue, wear, oxidation, and tensile data providerobust experimental comparisons amongst the groups, themicrostructure results of this study should be consideredwith the caveat that image analysis is not an exact methodto determine microstructural parameters. It is likely thatthe filtering process results in lamellae that are skewedsmall, but this bias is the same for all groups and sodoes not affect comparisons or statistical methods basedon ranks. Furthermore, alternative methods such as ultra-small angle X-ray scattering also have limitations. X-rayscattering estimates lamellar size using an analysis basedon Bragg’s law, but assumes a stack of infinitely longand wide lamellae, introduces error from its dependenceon average crystallinity measurements, and reports only amedian value rather than distributions (Fruhwirth et al., 2004;Turell and Bellare, 2004). However, with our procedure we

are able to acquire the distributions of lamellar dimensions.Future research could use this tool to evaluate materialsin the context of the entire distribution rather than justa single value. While absolute measurements of lamellarmicrostructure are challenging to obtain, our image analysisapproach is appropriate for characterizing relative differencesin microstructures amongst the materials.

Second, we did not directly measure cross-link densityor free radical concentration. However, it is well-known thatcross-link density is directly related to the radiation dose(Muratoglu et al., 1999). It has also been established thatartificial aging is an appropriate method to evaluate theoxidative stability of materials and that aging results arein agreement with free radical concentration measured byelectron spin resonance (Collier et al., 2003, 1996; Costa et al.,1998b; Edidin et al., 2000; Jahan et al., 1991). Our artificialaging results are consistent with a recent study by Morrisonand Jani (Morrison and Jani, 2009) which reported free radicalconcentration for similar annealed materials above 1.0 ×

1015 spins/g compared to concentrations below the level ofdetection for a re-melted material.

A third caveat is the relatively small range of microstruc-tures found in the materials in this study, which limits ourability to quantitatively determine relationships amongst thetreatments, mechanical performance, and microstructures.However, our results are consistent with other studies inves-tigating ultra-highmicrostructure andmechanical properties.It has been similarly reported that crystallinity is highly cor-related with elastic modulus (Medel et al., 2007; Simis et al.,2006) and that fatigue threshold scales better with lamellarsize than with crystallinity (Simis et al., 2006). Dependenceof fatigue resistance on crystallinity has also been reportedin other studies that did not evaluate other lamellar size pa-rameters (Baker et al., 2000). Our results suggest that fatigueresistance is correlated with both lamellar size and with crys-tallinity, although experimentally these parameters are notindependent because a cross-linked amorphous phase ap-pears to limit both the number and size of re-formed lamel-lae. Lamellar thickness in our study was not correlated withfatigue resistance or with any other parameter, likely due tothe small range of lamellar thickness present in our materi-als (interquartile ranges from 25 to 40 nm). Our results takentogether with the literature suggest that treatments such asre-melting that decrease lamellar size and crystallinity aredetrimental to fatigue crack propagation resistance, and thatsliding wear resistance is more dependent on amorphousphase parameters governed by cross-linking than by lamellarparameters.


This study has important clinical implications. Primarily,none of the clinically-relevant materials was able to excelin all three areas: fatigue crack propagation resistance,wear resistance, and oxidative stability. Another implicationof this study is for tailoring ultra-high treatments andmicrostructures for better fatigue, wear, and oxidationperformance. Our results suggest that the next generation oftreatments could improve the oxidative stability of annealedmaterials, or could increase the lamellar volume fractionand lamellar size of re-melted materials. Some promisingtreatments are those which combine annealing with achemical antioxidant, which is thought to scavenge more freeradicals and inhibit long-term oxidation (Kurtz et al., 2009).Our previous work on antioxidants shows that includingthe antioxidant in the initial resin may inhibit some ofthe cross-linking, but does not disrupt the formation of thelamellarmicrostructure (Furmanski et al., 2007). Furthermore,combining cross-linking and re-melting with a secondaryhigh pressure or thermal treatment would likely preserve theoxidative stability while producing more and larger lamellaeassociated with improved fatigue resistance (Bellare et al.,2009; Bistolfi et al., 2009).

This is the first study to simultaneously evaluate fatiguecrack propagation, wear, and oxidation in a wide variety ofclinically-relevant ultra-high. The tradeoff we have shownin fatigue, wear, and oxidation performance is critical tothe material’s long-term success in total joint replacements.Excessive wear is a historical problem that results in largenumbers of failures (Bozic et al., 2009a,b; Harris, 2001).Poor resistance to fatigue crack propagation and fracturehas been implicated in recent reports of cross-linked re-melted hip liners fracturing in vivo (Furmanski et al., 2009;Tower et al., 2007). It is also well-established that highlyoxidized ultra-high cannot adequately withstand in vivodemands (Collier et al., 1996; Costa et al., 1998a; Edidin et al.,2000). Understanding the tradeoff amongst fatigue, wear, andoxidation performance of cross-linked ultra-high, as well asthe relationship of mechanical performance to treatment andmicrostructure, can provide important information needed toproduce materials and designs that can withstand rigorous invivo mechanical demands and improve the longevity of totaljoint arthroplasty.

5. Conclusions

• There is a tradeoff amongst fatigue crack propagationresistance, wear resistance, and oxidative stability inclinically-relevant cross-linked ultra-high. None of theclinically-relevant materials was able to excel in all threeareas.

• Material behavior was dependent on both total radiationdose and heat treatment.

◦ Total radiation dose decreased fatigue resistance andultimate properties, but increased wear resistance.

◦ Annealing, as opposed to re-melting, reduced fatigueresistance and elastic properties less than re-melting,but left materials susceptible to oxidation.

◦ Sequential doses of radiation and annealing did notsignificantly change resistance to fatigue or oxidationbeyond that of an equivalent total single dose, but didappear to improve the wear resistance.

• Examining the microstructure elucidated the effects of theprocessing treatments on the material behavior.◦ Applying thermal energy below themelting temperature

after cross-linking increased the volume fraction oflamellae. However, this annealing process failed toneutralize all free radicals.

◦ Alternatively, re-melting the materials appeared toeliminate free radicals, but, restricted by the networkof cross-links, the re-formed lamellae were fewer andsmaller in size. These fewer, smaller lamellae wereassociated with decreased fatigue resistance and tensileproperties.

◦ Wear rate was not significantly correlated to any ofthe microstructural parameters in this study and wasdominated by radiation dose.

Acknowledgements

The authors would like to thank Erik Feest, John Tang, MikeHoang, Stephanie Uhlich, Chris Chaplin, Perry Johnson, MikeWatson, Ingrid Chang, Robyn Shaffer, and Tim Hong forassistance with research. We would also like to thank Dr.Tony Keaveny for guidance in preparation of the manuscript,and the Electron Microscope Lab at UC Berkeley. This workwas enabled by a grant from the National Science Foundation(#CMS 0505272) to UC Berkeley.

We would like to draw your attention to the followingpotential conflicts of interest relating to the submitted article:Dr. Michael Ries receives royalties from Smith and Nephew,and Dr. Douglas Van Citters serves as a consultant for DePuyand receives funding for other scientific research (not directlyrelated to this work) from DePuy. Dr. Sara Atwood, Eli Patten,Dr. Jevan Furmanski, and Dr. Lisa Pruitt have no knownconflicts of interest associated with this publication.

R E F E R E N C E S

Baker, D.A., Bellare, A., Pruitt, L., 2003. The effects of degree ofcrosslinking on the fatigue crack initiation and propagationresistance of orthopedic-grade polyethylene. J. Biomed. Mater.Res. A 66 (1), 146–154.

Baker, D.A., Hastings, R.S., Pruitt, L., 2000. Compression andtension fatigue resistance of medical grade ultra highmolecular weight polyethylene: the effect of morphology,sterilization, aging and temperature. Polymer 41 (2), 795–808.

Bartel, D.L., Bicknell, V.L., Wright, T.M., 1986. The effect ofconformity, thickness, and material on stresses in ultra-highmolecular weight components for total joint replacement.J. Bone Joint Surg. Am. 68 (7), 1041–1051.

Bellare, A., Kurtz, S.M., Steven, M.K., 2009. High pressurecrystallized UHMWPEs. In: UHMWPE Biomaterials Handbook,second ed. Academic Press, Boston, pp. 277–289.

Bistolfi, A., Turell, M.B., Lee, Y.L., Bellare, A., 2009. Tensileand tribological properties of high-crystallinity radiationcrosslinked UHMWPE. J. Biomed. Mater. Res. B Appl. Biomater.90 (1), 137–144.


Bozic, K.J., Kurtz, S.M., Lau, E., Ong, K., Chiu, V., Vail, T.P., Rubash,H.E., Berry, D.J., 2009a. The epidemiology of revision total kneearthroplasty in the United States. Clin. Orthop. Relat. Res..

Bozic, K.J., Kurtz, S.M., Lau, E., Ong, K., Vail, T.P., Berry, D.J., 2009b.The epidemiology of revision total hip arthroplasty in theUnited States. J. Bone Joint Surg. Am. 91 (1), 128–133.

Collier, J.P., Currier, B.H., Kennedy, F.E., Currier, J.H., Timmins,G.S., Jackson, S.K., Brewer, R.L., 2003. Comparison of cross-linked polyethylene materials for orthopaedic applications.Clin. Orthop. Relat. Res. (414), 289–304.

Collier, J.P., Sutula, L.C., Currier, B.H., Currier, J.H., Wooding, R.E.,Williams, I.R., Farber, K.B., Mayor, M.B., 1996. Overview ofpolyethylene as a bearing material: comparison of sterilizationmethods. Clin. Orthop. Relat. Res. (333), 76–86.

Costa, L., Luda, M.P., Trossarelli, L., Brach del Prever, E.M., Crova,M., Gallinaro, P., 1998a. In vivo UHMWPE biodegradation ofretrieved prosthesis. Biomaterials 19 (15), 1371–1385.

Costa, L., Luda, M.P., Trossarelli, L., Brach del Prever, E.M.,Crova, M., Gallinaro, P., 1998b. Oxidation in orthopaedicUHMWPE sterilized by gamma-radiation and ethylene oxide.Biomaterials 19 (7–9), 659–668.

Crowninshield, R.D., Muratoglu, O.K., 2008. How have newsterilization techniques and new forms of polyethyleneinfluenced wear in total joint replacement? J. Am. Acad.Orthop. Surg. 16 (Suppl. 1), S80–S85.

Currier, B.H., Currier, J.H., Mayor, M.B., Lyford, K.A., Collier, J.P.,Van Citters, D.W., 2007a. Evaluation of oxidation and fatiguedamage of retrieved crossfire polyethylene acetabular cups.J. Bone Joint Surg. Am. 89 (9), 2023–2029.

Currier, B.H., Currier, J.H., Mayor, M.B., Lyford, K.A., Van Citters,D.W., Collier, J.P., 2007b. In vivo oxidation of gamma-barrier-sterilized ultra-high-molecular-weight polyethylene bearings.J. Arthroplasty 22 (5), 721–731.

Dumbleton, J.H., D’Antonio, J.A., Manley, M.T., Capello, W.N.,Wang, A., 2006. The basis for a second-generation highly cross-linked UHMWPE. Clin. Orthop. Relat. Res. 453, 265–271.

Edidin, A.A., Jewett, C.W., Kalinowski, A., Kwarteng, K., Kurtz, S.M.,2000. Degradation of mechanical behavior in UHMWPE afternatural and accelerated aging. Biomaterials 21 (14), 1451–1460.

Edidin, A.A., Pruitt, L., Jewett, C.W., Crane, D.J., Roberts,D., Kurtz, S.M., 1999. Plasticity-induced damage layer isa precursor to wear in radiation-cross-linked UHMWPEacetabular components for total hip replacement. Ultra-high-molecular-weight polyethylene. J. Arthroplasty 14 (5), 616–627.

Fisher, J., Dowson, D., Hamdzah, H., Lee, H.L., 1994. The effect ofsliding velocity on the friction and wear of UHMWPE for use intotal artificial joints. Wear 175 (1–2), 219–225.

Fruhwirth, T., Fritz, G., Freiberger, N., Glatter, O., 2004. Structureand order in lamellar phases determined by small-anglescattering. J. Appl. Crystallogr. 37, 703–710.

Furmanski, J., Anderson, M., Bal, S., Greenwald, A.S., Halley, D.,Penenberg, B., Ries, M., Pruitt, L., 2009. Clinical fracture ofcross-linked UHMWPE acetabular liners. Biomaterials 30 (29),5572–5582.

Furmanski, J., Atwood, S.A., Feest, E., Tang, J., Hoang, M., Pruitt, L.,2007. Effect of vitamin E on fatigue of irradiated UHMWPE. In:53rd Annual Meeting of the Orthopaedic Research Society. SanDiego, CA.

Furmanski, J., Pruitt, L.A., 2007. Peak stress intensity dictatesfatigue crack propagation in UHMWPE. Polymer 48 (12),3512–3519.

Gencur, S.J., Rimnac, C.M., Kurtz, S.M., 2006. Fatigue crackpropagation resistance of virgin and highly crosslinked,thermally treated ultra-high molecular weight polyethylene.Biomaterials 27 (8), 1550–1557.

Glantz, S.A., 2005. Primer of Biostatstistics. McGraw-Hill, NewYork.

Harris, W.H., 2001. Wear and periprosthetic osteolysis: theproblem. Clin. Orthop. Relat. Res. (393), 66–70.

Jahan, M.S., Wang, C., Schwartz, G., Davidson, J.A., 1991. Combinedchemical and mechanical effects on free radicals in UHMWPEjoints during implantation. J. Biomed. Mater. Res. 25 (8),1005–1017.

Klapperich, C., Komvopoulos, K., Pruitt, L., 1999. Tribological prop-erties and microstructure evolution of ultra-high molecularweight polyethylene. J. Tribol. 121 (2), 394–402.

Kurtz, S., Bracco, P., Costa, L., Steven, M.K., 2009. Vitamin-E-blended UHMWPE biomaterials. In: UHMWPE BiomaterialsHandbook, second ed. Academic Press, Boston, pp. 237–247.

Kurtz, S.M., Steven, M.K., 2009. A primer on UHMWPE.In: UHMWPE Biomaterials Handbook, second ed. AcademicPress, Boston, pp. 1–6.

Lenth, R.V., 2006–2009. Java applets for power and sample size.McKellop, H., Shen, F.W., Lu, B., Campbell, P., Salovey, R.,

1999. Development of an extremely wear-resistant ultra highmolecular weight polyethylene for total hip replacements.J. Orthop. Res. 17 (2), 157–167.

Medel, F.J., Pena, P., Cegonino, J., Gomez-Barrena, E., Puertolas, J.A.,2007. Comparative fatigue behavior and toughness of remeltedand annealed highly crosslinked polyethylenes. J. Biomed.Mater. Res. B Appl. Biomater. 83 (2), 380–390.

Morrison, M.L., Jani, S., 2009. Evaluation of sequentiallycrosslinked ultra-high molecular weight polyethylene.J. Biomed. Mater. Res. B Appl. Biomater. 90 (1), 87–100.

Muratoglu, O.K., Bragdon, C.R., O’Connor, D.O., Jasty, M., Harris,W.H., 2001. A novel method of cross-linking ultra-high-molecular-weight polyethylene to improve wear, reduceoxidation, and retain mechanical properties. Recipient of the1999 HAP Paul award. J. Arthroplasty 16 (2), 149–160.

Muratoglu, O.K., Bragdon, C.R., O’Connor, D.O., Jasty, M., Harris,W.H., Gul, R., McGarry, F., 1999. Unified wear model forhighly crosslinked ultra-high molecular weight polyethylenes(UHMWPE). Biomaterials 20 (16), 1463–1470.

Olley, R., Bassett, D., 1977. Molecular-conformations in polyethy-lene after recrystallization or annealing at high-pressures.J. Polym. Sci. Ser. B 15 (6), 1011–1027.

Patten, E.W., 2008. Multi-axial tribo-system: rolling, sliding androtation of UHMWPE in total joint replacements. University ofCalifornia at Berkeley. Berkeley, CA.

Pruitt, L., Bailey, L., 1998. Factors affecting near-threshold fatiguecrack propagation behavior of orthopedic grade ultra highmolecular weight polyethylene. Polymer 39 (8–9), 1545–1553.

Ries, M.D., Pruitt, L., 2005. Effect of cross-linking on themicrostructure and mechanical properties of ultra-highmolecular weight polyethylene. Clin. Orthop. Relat. Res. 440,149–156.

Shih, C.F., Lorenzi, H.G., Andrews, W.R., 1977. Elastic compliancesand stress-intensity factors for side-grooved compact speci-mens. Int. J. Fract. 13 (4), 544–548.

Simis, K.S., Bistolfi, A., Bellare, A., Pruitt, L.A., 2006. Thecombined effects of crosslinking and high crystallinity onthe microstructural and mechanical properties of ultrahigh molecular weight polyethylene. Biomaterials 27 (9),1688–1694.

Tower, S.S., Currier, J.H., Currier, B.H., Lyford, K.A., VanCitters, D.W., Mayor, M.B., 2007. Rim cracking of thecross-linked longevity polyethylene acetabular liner aftertotal hip arthroplasty. J. Bone Joint Surg. Am. 89 (10),2212–2217.


Turell, M.B., Bellare, A., 2004. A study of the nanostructure andtensile properties of ultra-highmolecular weight polyethylene.Biomaterials 25 (17), 3389–3398.

Van Citters, D.W., Kennedy, F.E., Collier, J.P., 2007. Rolling slidingwear of UHMWPE for knee bearing applications. Wear 263(7–12), 1087–1094.

Viano, A.M., Spence, K.E., Shanks, M.A., Scott, M.A., Redfearn, R.D.,Carlson, C.W., Holm, T.A., Ray, A.K., 2007. Structural and chem-ical changes in ultra-high-molecular-weight polyethylene dueto gamma radiation-induced crosslinking and annealing in air.Biomed. Mater. Eng. 17 (5), 257–268.

Wang, A., 2001. A unified theory of wear for ultra-high molecularweight polyethylene in multi-directional sliding. Wear 248(1–2), 38–47.

Wang, A., Sun, D.C., Stark, C., Dumbleton, J.H., 1995. Wearmechanisms of UHMWPE in total joint replacements. Wear181–183 (1), 241–249.

Wang, A., Yau, S.S., Essner, A., Herrera, L., Manley, M., Dumbleton,J., 2008. A highly crosslinked UHMWPE for CR and PS total kneearthroplasties. J. Arthroplasty 23 (4), 559–566.

Zhou, J., Komvopoulos, K., 2005. Wear mechanisms of untreatedand gamma irradiated ultra-high molecular weight polyethy-lene for total joint replacements. J. Tribol. 127 (2), 273–279.

Documents

Tradeoffs Amongst Fatigue Wear and Oxidation Resistance of Cross-linked Ultra-high Molecular Weight Polyethylene