www.elsevier.com/locate/clinbiomech

Clinical Biomechanics 18 (2003) 523–536

Experimental investigation of bone remodelling usingcomposite femurs

V. Waide a, L. Cristofolini a,b,*, J. Stolk c, N. Verdonschot c, A. Toni a

a Laboratorio di Tecnologia Medica, Istituti Ortopedici Rizzoli, Via di Barbiano 1/10, 40136 Bologna, Italyb DIEM, Facolt�aa di Ingegneria, Universit�aa degli Studi di Bologna, Bologna, Italy

c Orthopaedic Research Laboratory, University Medical Center Nijmegen, P.O. Box 9101, 6500 HB Nijmegen, The Netherlands

Received 12 December 2001; accepted 25 March 2003

Abstract

Objective. To determine the load transfer patterns of femurs in the intact, immediate post-operative and long-term (remodelled)

post-operative implanted conditions for Lubinus SPII and M€uuller-Curved cemented hip prostheses, and to examine to what extent

remodelling may influence the long-term outcome.

Design. Experimental and finite element (FE) methods were applied to composite femurs under loads representing the heel-strike

phase of gait, determining cortical bone and cement strains for the different femur conditions.

Background. The authors previously developed protocols to measure bone and cement strains, and to produce remodelled femur

specimens, yet these have not been applied together to compare strain patterns of different femur conditions. The Lubinus SPII is

clinically more successful than the M€uuller-Curved stem, with failure mainly due to aseptic loosening.

Methods. Cortical bone strains were determined in intact femurs. Six femurs each were implanted with the two stem types and

cortical bone and cement strains were measured. Bone remodelling was recreated using a validated CAD–CAM procedure to re-

move a layer of proximal cortical bone, replicating a typical scenario found in stable clinical retrievals. Strains were remeasured. FE

methods were used to compliment the experiments.

Results. Stress shielding was reduced with remodelling, though bone strains did not return to their intact values, particularly

around the calcar. Cement strains increased with remodelling. Differences occurred between the two stems; the M€uuller-Curvedproduced a more severe strain transition.

Conclusions. Procedures were successfully combined together to investigate in vitro the performance of two cemented stems, in

immediate and long-term post-operative conditions. The increase of cement strains with remodelling is a potential indicator for in

vivo cement failure.

Relevance

The consequences of femoral bone remodelling on the long-term success of joint replacements are not well understood, where

remodelling may lead to increased bone and cement stresses.

� 2003 Elsevier Science Ltd. All rights reserved.

Keywords: Bone remodelling; Composite femurs; Cemented implants; In vitro simulations; Strain measurement; Finite element analysis

1. Introduction

Revision of cemented hip joint replacements is mainly

attributed to aseptic loosening. This is typically associ-

ated with breakdown of the cement mantle (Verdons-

chot and Huiskes, 1997), where the clinical findings of

*Corresponding author.

E-mail address: cristofolini@tecno.ior.it (L. Cristofolini).

0268-0033/03/$ - see front matter � 2003 Elsevier Science Ltd. All rights res

doi:10.1016/S0268-0033(03)00072-X

the Swedish Hip Register (Herberts and Malchau, 2000;

Malchau et al., 2000) have shown certain cemented

prostheses to be more successful than others. For ex-

ample, the Lubinus SPII implant (with an anatomically

shaped, collared stem) has been found to have a clini-

cally better outcome than the M€uuller-Curved prosthesis

(with a symmetrically curved, collared stem), where a

greater percentage of revisions have been performed forthe M€uuller-Curved implants than for the Lubinus SPII

implants within a fixed time period. Recent work by

erved.

524 V. Waide et al. / Clinical Biomechanics 18 (2003) 523–536

some of the authors has involved studying these twoimplants (Stolk et al., 2002; Cristofolini and Viceconti,

2000), where the long-term objective is to determine

whether the different clinical results can be explained

biomechanically.

An important step in the process of biomechanically

assessing hip prostheses is to determine the load transfer

characteristics of the different stem designs. This should

not only provide a profile of the bone strains, but also thestrains within the cement mantle. Indeed, both experi-

mental and finite element (FE) methods have shown a

close correlation in the bone and cement strain profiles

for these two stems implanted in commercially available

synthetic femurs (Pacific Research Laboratories, Vashon

Island,Washington,USA) (Stolk et al., 2002).While such

synthetic femurs may be limited in not being able to

represent the biological response of natural femurs, nev-ertheless, their geometrical and mechanical characteris-

tics have been validated to fall within the range of those of

cadaveric specimens (Cristofolini et al., 1996b; Szivek

and Gealer, 1991; McNamara et al., 1994). Furthermore,

the inter- and intra-subject variability of the synthetic

femurs is considerably less than those of cadaveric ones,

as well as being advantageous in overcoming such prob-

lems as availability, handling, and preservation.The in vitro tests of this nature (Stolk et al., 2002;

Cristofolini and Viceconti, 1999b) have consisted of

implanting the stems in synthetic femurs and measuring

the strains at predetermined positions on the composite

‘‘cortical’’ bone surface. In addition, a method was de-

veloped to provide a technique for strain measurements

within the polymethyl methacrylate bone cement mantle

(Cristofolini and Viceconti, 2000). This method involvedprecoating the stem with a thin layer of cement and

instrumenting the cement with strain gauges at specified

locations. However, these studies (Stolk et al., 2002;

Cristofolini and Viceconti, 1999b, 2000) only represent

the implanted stems in the immediate post-operative

condition, a condition which, though alters the strain

profile compared to the intact femur, does not represent

a long-term post-operative situation, such as remodel-ling of the femur. Remodelling may be more detrimental

to the long-term outcome of the prosthesis, though there

is clinical evidence of well-fixed cemented stems in bone

remodelled femora (Schmalzried et al., 1993; Maloney

et al., 1989; Maloney et al., 1996). Nevertheless, its

consequences on the long-term success of joint replace-

ments are not well understood (Rubash et al., 1998).

Another methodology was therefore developed toproduce experimental specimens suitable for investigat-

ing a long-term post-operative scenario (a well-fixed

implanted femur subject to bone resorption). This in-

volved using computer aided design and manufacturing

(CAD–CAM) methods to modify the composite femur

specimens to incorporate bone resorption features

(Waide et al., 2001a,b). These modifications are based on

clinical data (Maloney et al., 1996), not a bone remod-elling algorithm, and replicate a clinically based scenario

of an average amount of bone remodelling for a cemented

stem. However, though the methodology has been vali-

dated (Waide et al., 2001a,b), the load transfer charac-

teristics of these specimens have yet to be determined.

This study reports on the application of these pro-

tocols (Cristofolini and Viceconti, 2000; Cristofolini and

Viceconti, 1999b; Waide et al., 2001a,b; Cristofolini etal., 1997) to investigate the Lubinus SPII and M€uuller-Curved cemented hip prostheses. The objective was to

analyse the load transfer patterns (determining both

cortical bone and cement strains) for: the intact femur;

the immediate post-operative implanted femur; the long-

term post-operative implanted femur with clinically

relevant moderate cortical bone remodelling and a well-

fixed stem. Specifically, the following research questionswere addressed:

(1) to confirm that the separate protocols can be applied

together successfully to investigate stress shielding

under the different femur conditions;

(2) to analyse if the remodelling condition results in crit-

ical stress–strain conditions for the bone;

(3) to analyse if the remodelling condition results in crit-ical stress–strain conditions for the bone cement;

(4) to investigate if the stress–strain profile after remod-

elling is similar to that for the intact femur;

(5) to determine if different profiles exist between the

two different designs of prostheses.

While these methods may present a possible expla-

nation for subsequent aseptic loosening, attributed to aremodelling phenomena, it should be emphasised that

failure may also occur due to other biological and me-

chanical factors such as wear debris and the formation

of a fibrous tissue layer, loss of stem stability due to

breakdown of the stem–cement interface, cement frac-

ture, etc. The potential role of remodelling in the long-

term outcome of cemented prostheses is discussed based

on the findings of this study. These methods are pre-sented and discussed as a potential in vitro means for

pre-clinical evaluation of prostheses designs. Further-

more, it is hypothesised that when used together these

methods can discriminate between different post-opera-

tive states and between stem types. Such techniques

would be clinically relevant in examining, for example,

the conditions of strain within the cement and whether

when loaded in the remodelled femur it may be subjectto greater strains, likely to cause cement failure.

2. Methods

Full details of the validation and development of

the procedures used are given elsewhere (Cristofolini

V. Waide et al. / Clinical Biomechanics 18 (2003) 523–536 525

and Viceconti, 1999b, 2000; Waide et al., 2001a,b;Cristofolini et al., 1997). Therefore, the following de-

scription focuses on their combined application, de-

scribing the necessary details to present a coherent

explanation of the methods involved. The methodology

involved is summarised by the flowchart illustrated in

Fig. 1.

2.1. Preparation and testing of intact femurs

Twelve left composite femurs (Mod. 3103, Pacific

Research Laboratories) were investigated in this study.

Before testing, each femur was carefully marked with a

femur-aligned reference system (Cristofolini and Vice-

conti, 1999b; Ruff and Hayes, 1983) to assist repro-

ducibility in subsequent testing stages. The femurs were

prepared for strain measurement with 10 triaxial rosettesbonded to the outer cortical surface at five levels on the

medial and lateral sides (shown in Fig. 2), at the mid-

distance of the diaphysis at each level (Cristofolini and

Viceconti, 1999b; Cristofolini et al., 1997). The 12 fe-

murs were loaded, simulating the heel strike phase of

gait. Ten measurement repetitions were made for each

femur, following a validated protocol (Cristofolini and

Viceconti, 1999b); see Fig. 2.

Fig. 1. Flowchart illustrating the sequence of tests (clockwise). The composite

then implanted with instrumented stems and the composite cortical bone su

bedded gauges are shown for the Lubinus SPII stem and for the M€uuller-Cur

prior to bonding the strain gauges. For the Lubinus SPII, the thickness was 1

outer cortical surfaces of the femurs were machined, using CAD–CAM techn

the composite cortical bone substitute and cement strains for the remodelled

2.2. Preparation and testing of implanted femurs

Two different stem types (six specimens for each type)

were investigated: (i) Lubinus SPII of length: 150 mm

and width: narrow 135� (Waldemark Link GmBH,

Hamburg, Germany); and (ii) M€uuller-Curved with

standard stem and medium neck (supplied by JRI Ltd.,

London, UK). For both stem types, three locations were

selected for cement strain measurement with embeddedgauges. These locations were selected in order to stay

clear of high stress gradients, based on a preliminary FE

study. The positions, determined as a fraction of the

stem length, are illustrated in Fig. 1. Briefly, the proce-

dure consisted of pre-coating the stem (at the selected

positions) with a thin cement layer of fixed thickness.

Strain rosettes were bonded to the cement layers so that

the instrumented stems could be implanted followingroutine practice. Full details on the preparation and

validation of the embedded gauges are reported else-

where (Cristofolini and Viceconti, 2000).

The femora were subsequently prepared by an expe-

rienced surgeon. The femoral cavities were filled with

vacuum mixed (Optivac system, Scandimed, Sj€oobo,Sweden) bone cement (Cemex Rx, Tecres Spa, Som-

macampagna (VR), Italy) and implanted with the in-strumented stems. Strain measurements were made for

cortical strains were determined for the intact femurs. The femurs were

bstitute and cement strains were measured. The positions of the em-

ved stem. The layer of bone cement was prepared on the stem surface

� 0.1 mm. For the M€uuller-Curved, it was 0.5� 0.1 mm. The proximal

iques, to represent bone resorption. The femurs were retested to obtain

condition.

Fig. 2. The heel strike loading set-up is shown. The direction and magnitude of the hip and abducting forces are such as to balance the resultant force

applied to the top of the cantilever. Their ratio and direction is controlled by adjusting the lever arms. The following components are connected to the

cantilever (1): a movable hip socket (2) whose position can be adjusted with the screw (3); an abducting rod (4), free to pivot together with the angle

indicator (5); its length is adjusted by tightening the bolt (6); a plumb line (7) indicates if the cantilever (1) is horizontal; a slide (8) is adjusted by

means of the screw (9); a load cell (10) monitor the abducting force. Also indicated are the strain gauge locations on five levels, on the lateral and

medial aspects.

526 V. Waide et al. / Clinical Biomechanics 18 (2003) 523–536

the femurs subjected to the same heel strike loading

conditions as for the intact state. Again, 10 measure-

ment repetitions were made for each femur, with strains

being recorded for both the cortical gauges and the

embedded gauges. Great care was taken to ensure that

the implanted femur was subjected to the same system of

loads, by selecting the suitable neck length, and further

correcting for the lever arms (Cristofolini and Viceconti,1999a).

2.3. Preparation and testing of remodelled femurs

Experimental modelling of bone resorption of the

proximal femur with a well-fixed implant required

modifying the femoral cortical surfaces of the composite

femur specimens. Therefore, to develop suitably accu-rate experimental specimens, clinical data (Maloney

et al., 1996), based on a detailed autopsy study of re-

modelling for well-fixed cemented implants (13 cases of

seven different stem designs with a mean duration of 7.9

years), have been used to define the altered (remodelled)

surfaces of the composite femurs. In this way, the av-

erage values from the clinical data were used to repre-

sent of an average amount of bone remodelling in a

successfully implanted femur. The clinical data give

quantitative results for the mean percentage changes in

cortical bone thickness, cortical bone areas, and bone

mineral densities. Measurements were reported for fivecross-sectional levels (defined by the prosthesis stem

length) and at four regions around the cross-section

(anterior, posterior, medial, and lateral).

Starting with a CAD model of the intact composite

femur (this is freely available on the Internet (Viceconti

et al., 1996)), CAD–CAM software (Unigraphics V15.0,

Electronic Data Systems Co., Maryland Heights, MO,

USA) was used to design the remodelled cortical ge-ometry for the implanted femurs, where two separate

remodelled geometries were developed for the two im-

plant designs. The procedure first involved identifying

the five cross-sectional levels corresponding to the levels

V. Waide et al. / Clinical Biomechanics 18 (2003) 523–536 527

of Maloney et al. (1996). The proximal external corticalgeometry was subsequently altered, based on the clinical

measurements at these levels, to represent the appro-

priate morphological features of bone resorption. This

alteration involved reducing the cross-sectional geome-

try of the external perimeter, initially using the per-

centage values given for changes in cortical bone

thickness and area. However, the change in bone min-

eral density cannot be easily reproduced in the com-posite femur (it would be difficult to perform this

experimentally). Therefore, the reduction in bone min-

eral density was accounted for by a further reduction of

the cross-sectional geometry to match the flexural stiff-

ness of the composite femur to that of bone with a re-

duced density. Table 1 reports the average thickness of

cortical bone removed at each of the five strain gauge

levels (shown in Fig. 2) for the two stem designs. TheCAD model was checked to ensure that the resultant

data closely matched the clinical data, falling within the

range of values (mean and standard deviation) reported

(Maloney et al., 1996). Though the clinical data was

based on a small number of specimens and different stem

designs, therefore producing a large inter-stem varia-

tion, using clinical based data was considered a more

positive approach than deriving a geometry from a re-modelling theory.

The CAD models with the remodelled cortical sur-

faces were subsequently used to control the cutting

movements of a milling machine (ProLight 2000 Ma-

chining Center, Light Machines Co., Manchester, NH,

USA). Each femur was mounted on the milling machine

using a custom-designed jig. The accuracy of the align-

ment of each femur was ensured by using a digitiser inconjunction with the milling machine. Following ma-

chining of the external cortical surfaces, the specimens

were redigitised to confirm the accuracy of the process.

In all cases, the mean error achieved was in the order of

0.2 mm. Full details of the development and validation

of these techniques are detailed elsewhere (Waide et al.,

2001a,b). Since the machining process involved removal

of the existing cortical strain gauges, the remodelled

Table 1

Reduction of the cortical bone thickness applied to the composite femur CA

Strain gauge Lubinus

Total (mm)a %b

Level 1 1.27 (0.75) 70

Level 2 1.55 (0.27) 59

Level 3 2.16 (0.29) 35

Level 4 2.03 (0.21) 41

Level 5 0.62 (0.17) 29

aMean (standard deviation), based on measurements made at approximat

strain gauge levels.bAmount of the reduction representing the adjustment in flexural stiffnes

percentage of the total reduction.cNo reduction was applied since the stem length did not extend to this le

femurs were reinstrumented with strain gauges atthe same levels as for the intact femur, at mid-distance

of the diaphysis (neutral axis) at each level. The speci-

mens were retested following the same methodology as

before.

2.4. Processing of strain gauge data

The strain measurements recorded from the samefemur were averaged over the 10 measurement repeti-

tions, to give average values for each femur for each of

the 10 cortical surface strain gauges and for the three

rosettes embedded in the cement. The data from one of

the M€uuller-Curved implanted femurs were discarded

based on the Chauvenet criterion for outliers (Dally,

1989). In addition, data for the most proximal medial

gauge of one of the remodelled Lubinus femurs was alsofound to be an outlier and subsequently disregarded.

The principal strains were computed and used in sub-

sequent calculations. The inter-specimen variability

(expressed by the coefficient of variation for principal

strain values of the cortical bone and cement strains)

was determined for the two different stem types and for

each of the femur conditions tested. An ANOVAANOVA was

performed for the resultant strains for the lateral andmedial cortical gauges and the cement gauges using the

Statistica analysis package kernel release 5.5A (StatSoft

Inc., OK, USA). Where significant differences (P < 0:05)were observed the post hoc Scheff�ee test was applied to

determine the significant interactions.

Two methods were used to analyse and describe the

strain distribution for the two stem types under the

different femur conditions. The first method involved athree-parameter model, previously validated as being

suitable to describe synthetically and interpolate the

axial strain distribution in the femoral diaphysis (Cris-

tofolini et al., 1996a). This model was applied to

the cortical strain data for each of the femurs under the

different femur conditions. It was used to quantify the

strain gradients, analysing the load transfer character-

istics and stiffening effects of the stem. One parameter

D models to represent the bone remodelled femurs

M€uuller-Curved

Total (mm)a %b

1.16 (0.57) 66

1.67 (0.25) 35

2.10 (0.25) 41

0.12 (0.03) 29

Not applicablec Not applicablec

ely 50 points around the circumference of the bone, at each of the five

s to account for the change in bone mineral density, given in terms of

vel.

528 V. Waide et al. / Clinical Biomechanics 18 (2003) 523–536

represents the compliance for the axial load and theother two the compliance for the bending moment in the

medial–lateral plane. For each of the three parameters,

an ANOVAANOVA was performed to assess if significant differ-

ences existed for stem type and femur condition, and

their interaction. A second method was used to analyse

the strain distribution for the two stem types, being

applied to each of the three strain gauge regions (medial

cortical strains, lateral cortical strains, and cementstrains). This method was used to identify high gradi-

ents, where the difference in strain between adjacent

regions is large, this may indicate regions where possible

failure of the bone or cement may occur. The details of

the calculation of the root mean square strain difference

(an indicator of the strain distribution) are given in

Appendix A.

2.5. Finite element analysis of remodelled femurs

As indicated above, some of the authors previously

carried out a separate FE study, obtaining a close cor-

relation with the experimental strain profiles (of both

bone and cement) observed in the implanted femurs

(Stolk et al., 2002). In the current study, to compliment

the experimental strain analysis with remodelling, thesame FE techniques (Stolk et al., 2002) were subse-

quently applied to the CAD geometry of the remodelled

Fig. 3. Mean axial strains and 95% confidence intervals for the six Lubinus SP

the medial and lateral composite cortical bone surfaces, as shown. The strain

tested intact. The corresponding microstrain value for the intact femur (equ

base of each group of data bars.

femurs. It was intended that these FE models wouldprovide additional stress–strain data on the loading

characteristics, and in particular to confirm agreement

on the trends observed with the experimental models.

3. Results

The principal strains in the rosettes were found to liewithin 10� of the long axis of the bone, with the longi-

tudinal strains being larger in magnitude compared to

the hoop strains. The principal strain values are re-

ported for all cases. Seven of the grids (of the 108 em-

bedded grids) were damaged during implantation or

testing; therefore, data were not obtained from five

gauges.

The cortical strains for the two stem types are plottedin Figs. 3 and 4. These are reported as the percentage of

strains measured in the same femurs when intact. No

strain reduction has occurred given a value of 100%,

whereas complete stress shielding may be attributed to a

value of 0%. Each bar represents the mean and the 95%

confidence interval for the femurs with the same stem

type, under the implanted and remodelled conditions.

Stress shielding occurred for both stems, with strainsfalling below those of the intact values. The percentage

values for the stress shielding condition for the Lubinus

II implanted and remodelled femurs at the strain gauges positioned on

s are expressed as a percentage of the strain in the same femurs when

ivalent to 100%) is also reported for each strain gauge position, at the

Fig. 4. Mean axial strains and 95% confidence intervals for five of the six M€uuller-Curved implanted and remodelled femurs at the strain gauges

positioned on the medial and lateral composite cortical bone surfaces, as shown. The strain data from one of the M€uuller-Curved implanted femurs

were discarded based on the Chauvenet criterion for outliers (Dally, 1989). The strains are expressed as a percentage of the strain in the same femurs

when tested intact. The corresponding microstrain value for the intact femur (equivalent to 100%) is also reported for each strain gauge position, at

the base of each group of data bars.

Fig. 5. Mean axial strains and 95% confidence intervals for the six

Lubinus SPII implanted and remodelled femurs at the strain gauges

embedded within the cement mantle, for the positions shown. The

values are expressed in microstrain.

V. Waide et al. / Clinical Biomechanics 18 (2003) 523–536 529

stem (Fig. 3) were slightly less than those for the M€uuller-Curved stem (Fig. 4), for all cortical bone strain levels.Stress shielding was most prominent in the medial

proximal region (gauge position M1) where the mean

cortical strains were found to be 33% and 34% of the

intact values for the Lubinus and M€uuller-Curved stems,

respectively. In the remodelled condition, all cortical

bone strains increased over those of the implanted con-

dition, for both stem types. As for the implanted con-

dition, the percentage values for the stress shieldingcondition for the Lubinus stem (Fig. 3) remained

slightly less than those for the M€uuller-Curved stem (Fig.

4). Stress shielding was still noted, again being most

prominent in the medial proximal region; the mean

cortical strains were 65% and 70% of the intact values

for the Lubinus and M€uuller-Curved stems, respectively.

The resultant strains for the embedded gauges are

shown in Figs. 5 and 6, for the two stem types. Each barrepresents the mean and the 95% confidence interval, in

microstrain, for the femurs with the same stem type,

under the implanted and remodelled conditions. The

cement strain values for the Lubinus stem (Fig. 5) were

greater than those for the M€uuller-Curved stem (Fig. 6)

for the implanted condition, and at the proximal and

mid-stem levels (gauges C1 and C2) for the remodelled

condition. The cement strains increased from proximalto distal for both stem types, where, for the implanted

condition, the mean strain values near the stem tip(gauge E3) were 2–3 times greater than the values below

the collar (gauge C1). The remodelling condition was

found to increase the cement strains by an average of

35% for the Lubinus stem and 31% for the M€uuller-Curved stem (Table 2). This was especially evident for

the distal part of the stem (gauges C2 and C3), where,

for both prostheses, the differences between the mean

strain values of the implanted and the remodelled con-ditions were of the order of 200 microstrain or greater.

Fig. 6. Mean axial strains and 95% confidence intervals for five of the

six M€uuller-Curved implanted and remodelled femurs at the strain

gauges embedded within the cement mantle, for the positions shown.

The strain data from one of the M€uuller-Curved implanted femurs were

discarded based on the Chauvenet criterion for outliers (Dally, 1989).

The values are expressed in microstrain.

530 V. Waide et al. / Clinical Biomechanics 18 (2003) 523–536

Near the stem tip (gauge C3), the mean strain values

were greater than 1000 microstrain.

The statistical analysis ANOVAANOVA of the resultant strains

for the medial and lateral cortical regions and the ce-

ment mantle showed significant differences (P < 0:05)for the majority of factors (stem type, gauge position,

and femur condition) and their interactions, as given inTable 3. The post hoc three-way interaction Scheff�ee testdemonstrated significant differences (P < 0:05) between

Table 2

Cement strains and stresses obtained from both the experimental and FE m

Lubinus

Implanted Remodelled %

C1 [microstrain] 343 457

C2 [microstrain] 689 928

C3 [microstrain] 765 1047

Mean

FE peak stress (MPa) 5.3a 7.4a

Volume (%)c 2.8 8.4

a Located medially around stem tip.b Located laterally around stem tip.c Indicates percentage volume of cement mantle with a maximal principal

Table 3

P -values for factorial ANOVAANOVA analysed for the three strain regions indicated

Effect Lateral

Stem typea <0.0001

Gauge positionb <0.0001

Femur conditionc <0.0001

Stem type· gauge position <0.0001

Stem type· femur condition <0.0001

Gauge position· femur condition <0.0001

Stem type· gauge position · femur condition <0.0001

a Lubinus SPII or M€uuller-Curved.b L1 to L5, or M1 to M5, or C1 to C3, as appropriate.c Intact, implanted, remodelled for cortical bone strains; implanted, remo

the two stems at the mid-level positions of the diaphysis(gauges M3, L3, and L4 for the implanted condition,

and gauges M2, M3, and L3 for the remodelled condi-

tion); no significant differences were found between the

stems for all the other (gauge) positions. Significant

differences (P < 0:05) were found between femur con-

ditions, for the two different stems, as indicated at the

relevant locations in Fig. 7. For the Lubinus SPII stem,

such differences occurred at the mid-position corticalstrain region (levels 2 and 3), between the intact and

implanted conditions, and between the intact and re-

modelled conditions––see Fig. 7. There were fewer oc-

currences for the M€uuller-Curved stem, though again the

mid-position cortical strain region (level 3) was involved

(Fig. 7). The proximal medial region (rosette M1) had

significantly different strains between the intact and

implanted conditions, for both stems. Significant dif-ferences between the implanted and remodelled condi-

tions were found at the mid-position cortical region

(level 3) for the M€uuller-Curved stem and at the lateral

mid-distal region (rosette L4) for the Lubinus SPII stem.

The post hoc test found that for the cement strains a

significant difference only occurred between the im-

planted and remodelled conditions near the stem tip

(embedded rosette C3) of the M€uuller-Curved stem.The inter-specimen variability for the cortical strains

in the intact and implanted femurs ranged from 3% to

17% for both stem types. For the remodelled condition,

ethods, with corresponding percentage change between conditions

M€uuller-Curved

Change Implanted Remodelled % Change

33 266 302 14

35 603 799 33

37 756 1111 47

35 31

40 4.6b 6.4b 39

2.7 7.8

stress greater than 2 MPa.

Medial Cement

<0.0001 0.034

<0.0001 <0.0001

<0.0001 <0.0001

<0.0001 0.096

<0.0001 0.794

<0.0001 0.0097

<0.0001 0.564

delled for cement strains.

Fig. 7. Diagram illustrating where significant differences (P < 0:05) were observed for the two stem types (Lubinus SPII, left and M€uuller-Curved,right). Differences between femur conditions (Int/Imp, intact and implanted; Int/Rem, intact and remodelled; Imp/Rem, implanted and remodelled)

are indicated at the relevant gauge positions. Differences between the principal strain values of adjacent gauges, under the different femur conditions

(Int, intact; Imp, implanted; Rem, remodelled), are indicated by the bracketed arrows between the relevant strain positions.

V. Waide et al. / Clinical Biomechanics 18 (2003) 523–536 531

the Lubinus implant was within the same range (3–14%),

whereas the M€uuller-Curved implant showed a greater

variability (7–24%). A greater variability was also noted

for the cement strains. However, in comparing the ce-

ment strains for the two stems, the Lubinus stem had a

lower variability (10–14% for implanted and 10–25% forremodelled) than the M€uuller-Curved (12–22% for im-

planted and 15–33% for remodelled).

The analysis of cortical strain distribution using the

three-parameter model showed significant differences

(ANOVAANOVA, P < 0:0001) for all three parameters, for bothstem type and femur condition, and their interactions.

The post hoc Scheff�ee test revealed that the incidences

where no significant differences occurred (P < 0:05) onlyinvolved one parameter (one of the two compliance

parameters for the bending moment). For this parame-

ter, there was no difference between the two stem types,

and neither between the remodelled femur condition and

both the intact and implanted femur conditions. The

analysis of the strain distribution, based on the root

mean square strain difference (described in Appendix A),

showed that for the Lubinus SPII a smoother transitionoccurred between the strain values of adjacent gauges,

when compared to the transition for the M€uuller-Curvedstem––for further details see root mean square strain

Table 4

Root mean square difference in principal strains between adjacent strain gau

Root mean square strain differences: mean (standard

Composite cortical surface bone strain gauges

Intact Implanted

Medial Lateral Medial Latera

Lubinus SPII 476 (144) 586 (66) 385 (47) 296 (5

M€uuller-Curved 574 (117) 644 (89) 732 (62) 597 (7

difference values in Table 4. For the cortical bone, the

M€uuller-Curved femurs had strain gradients about twice

those of the Lubinus SPII, in both the implanted and

remodelled conditions; these were statistically significant

(ANOVAANOVA, P < 0:05). There was no significant difference

between the two groups of specimens in the intact state.Neither was any significant difference found between the

two stem designs for the cement strains, in both the im-

planted and remodelled conditions. In addition, the post

hoc three-way interaction Scheff�ee test of the original

strain data was used to determine where significant

differences occurred between the principal strain values

of adjacent gauges for the two different stem types. The

M€uuller-Curved stem showed more incidences of statis-tical differences between adjacent gauges than the

Lubinus SPII stem confirming the presence of a less even

stress/strain distribution; for the cortical bone strains of

the Lubinus SPII stem, differences only occurred be-

tween the gauges at the mid-position of the diaphysis

(levels 2 and 3)––see Fig. 7.

Similar to the results previously obtained for the FE

analysis of the implanted femurs (Stolk et al., 2002), theroot mean square error between experimental and nu-

merical strains for the remodelled femurs were within

10%. The FE models followed the same trends as the

ges for the different femur conditions and stem types

deviation) [microstrain]

Cement strain gauges

Remodelled Implanted Remodelled

l Medial Lateral

2) 435 (73) 403 (60) 234 (78) 327 (100)

3) 993 (185) 983 (325) 236 (77) 453 (133)

532 V. Waide et al. / Clinical Biomechanics 18 (2003) 523–536

experimental models with a corresponding increase inboth cortical bone and cement strains with remodelling.

In particular, peak stresses (Table 2) in the cement

mantle were determined to occur medially around the

stem tip for the Lubinus SPII stem and laterally around

the tip for the M€uuller-Curved stem, increasing by about

40% with remodelling. This increase was similar to the

increase in experimental strains (Table 2), especially at

the distal gauge position (C3). However, the level ofstress was similar for both stem types, both before and

after remodelling. The change in cement stresses with

remodelling was also considered by expressing the per-

centage volume of the cement mantle with a maximal

principal stress greater than 2 MPa, before and after

remodelling (Table 2). There was little difference be-

tween stem types.

4. Discussion

The aim of this work was to determine the load

transfer patterns of femurs under different pre- and post-

surgery (intact, immediate post-operative implanted,

and long-term post-operative implanted with clinically

relevant moderate cortical bone remodelling and a well-fixed stem) conditions, where these were investigated

with two different designs of cemented implant. A

number of research questions were addressed in this

work, as outlined below.

4.1. Combination of separate protocols?

It has been verified that the individual protocols,previously developed within the authors� laboratory, canbe successfully combined together to investigate the

stress shielding of different implant designs, under dif-

ferent femur conditions. These methods were shown to

distinguish statistical differences between the load

transfer characteristics of two different stem designs and

for the different femoral conditions. While the compos-

ite femur models used in this work have been frequentlyused for in vitro investigations in the immediate post-

operative condition, this is the first time that these

specimens have been used to experimentally study re-

modelling. It is important to note that the remodelled

patterns applied to the femoral specimens were devel-

oped based on clinical findings. The resultant remod-

elled geometries for the composite femurs were

confirmed to fall within the clinical data reported byMaloney et al. (1996), with the remodelled specimens

accurately reproduced to within an average of 0.2 mm of

the desired geometry. Therefore, the remodelled speci-

mens can be assumed to be mechanically representative

of a human specimen with clinically observed bone re-

modelling features around a typically stable implant.

The inter-specimen variability, determined for differ-ent femur conditions, is an indicator of the reproduc-

ibility of the methods. The variability of the cortical

strains in the intact and implanted femurs (both stem

types) fell within the lower end of the range for those

reported in the literature, therefore indicating good re-

producibility (Cristofolini and Viceconti, 1999b; Cris-

tofolini, 1997). The Lubinus stem in the remodelled

condition performed within the same range, whereas theM€uuller-Curved stem demonstrated slightly less repro-

ducibility, with the range extending to up to 24%. The

cement strains also showed a slightly greater variability

than those of cortical strains, which may be attributed to

small difference in the cement thickness (Fisher et al.,

1997). The greater variability for the M€uuller-Curvedstem may be attributed to its surgical instrumentation;

this is relatively basic (having been developed in the1970s) and consists of a single rasp size that only pro-

vides for a thin cement mantle. Since all the implants

tested here were inserted by the same experienced sur-

geon, a larger study, involving multiple surgeons, would

be required to establish inter-surgeon variability.

4.2. If remodelling is critical for bone?

In analysing whether the remodelling condition pro-

duces critical stress–strain conditions for the bone, the

resultant strains for the remodelled condition increased

from those for the immediate post-operative implanted

condition, yet typically remained below those of the

intact femur. Exceptions occurred for the M€uuller-Curved stem at the mid-level regions on the diaphysis,

below the level of the calcar; strains on the medial andlateral sides were approximately 20% and 10% greater

than the intact case, respectively. Nevertheless, consid-

ering that the yield strain of cortical bone is in the region

of 6000–8000 microstrain in tension and compression

(Carter et al., 1981; Biewener, 1993), the strain values

for all the femur conditions from this study are well

below these thresholds. Indeed, the principal tensile

strains at the proximal lateral aspect of the intact femurs(mean (standard deviation) of 1329 (164) microstrain at

gauge position L1) compared well with those measured

at a similar location in vivo (Aamodt et al., 1997) for the

stance phase of walking (approximately 1200 micro-

strain), especially when taking into consideration the

influence of strain gauge positioning (Cristofolini and

Viceconti, 1999a). The strain transitions between adja-

cent levels, though different for the two stem designs andthe femur condition, also compared to those typically

observed for bone (Biewener, 1993) and for similar in

vitro tests involving implanted femurs (Cristofolini and

Viceconti, 1999b; Maloney et al., 1989). However, the

larger strains observed at the mid-level regions on the

diaphysis of the remodelled M€uuller-Curved femurs were

close to the ranges that have been postulated to give rise

V. Waide et al. / Clinical Biomechanics 18 (2003) 523–536 533

to a degradation of the mechanical properties of corticalbone, a possible stimulus for bone remodelling (Pattin

et al., 1996). Therefore, for the remodelled conditions

examined in this study (a well-fixed stem under heel

strike loading), the stress–strains conditions may not be

considered critical for the bone.

4.3. If remodelling is critical for cement?

The remodelling condition resulted in an increase in

cement strains over that of the immediate post-operative

implanted condition. Though shown to be significantly

different for the individual effects (as listed in Table 3),

no significant differences were obtained for the two-way

and three-way interactions. This may be due to the

larger ranges of inter-specimen variability for the cement

strains, as opposed to those of the cortical strains, asdiscussed above. Nevertheless, the cement strains

around the tip of the M€uuller-Curved stem were signifi-

cantly greater in the remodelled condition. An increase

also occurred for the strain gradients of the cement

mantle (Table 4), indicating a greater change in strain

between adjacent gauges. The increase in strains with

remodelling, similarly observed in both experimental

and FE parts of the study, indicates a possible source offailure of the cement mantle; the increased cement

stresses in the remodelled femur could contribute to

crack initiation and propagation. This would be more

evident in the case of thin cement mantles (Fisher et al.,

1997; Jasty et al., 1991; Ebramzadeh et al., 1994; Estok

et al., 1997) and pores or defects in the cement (James

et al., 1992; McCormack and Prendergast, 1999). Other

in vitro strain gauge and FE studies have found similarpeak strains, greater than 1000 microstrain, around the

tip (Estok et al., 1997; O�Connor et al., 1996). Failure ofbone cement has been found to occur at this strain level

in the presence of porosity, within an average of 1.8

million cycles (Davies et al., 1987). This is equivalent to

approximately two years of patient use, where hip re-

placements can be expected to average 0.9 million cycles

per year (Schmalzried et al., 1998). In this study, theremodelling condition has been shown to increase the

distal cement strains to over 1000 microstrain and in

the region of 6.4–7.4 MPa peak stresses for a typical

daily loading activity (heel strike in the gait cycle).

Therefore, it may be reasonably argued that a more

strenuous loading condition coupled with a thin or flawed

cement mantle could produce critical stress–strain con-

ditions, leading to failure of the cement mantle.

4.4. If profile after remodelling is similar to intact?

The stress–strain profile after remodelling could not

be described as being similar to the intact femur. While

the remodelled cortical bone strains were closer to the

intact strains than the immediate post-operative im-

planted strains, as illustrated in Figs. 3 and 4, the strainsdid not return to the intact (100%) value. For both stem

types, the mean implanted cortical strains at the calcar

region (gauge position M1) were less than 40% of the

corresponding intact strains. Furthermore, when the

remodelling conditions were applied (based on clinical

data), the resultant strains approached those of the in-

tact femur but were still not normalised (at least 30%

difference at the calcar region). In addition, the analysisof the strain gradients showed statistically significant

differences between the femur conditions, on applying

both the three-parameter model (Cristofolini et al.,

1996a) and the root mean square strain difference method

(Appendix A). This further indicates that the remodel-

ling condition did not reproduce the strains for the in-

tact femur.

It should be highlighted that the results obtained hereare in agreement with the findings of Maloney et al.

(1989), who studied cortical bone strains in autopsy-re-

trieved specimens. While their methods involved a dif-

ferent procedure and different prostheses designs to

those applied in the current work, the trends observed

are similar. The autopsy-retrieved implanted specimens

examined in their study were taken from patients who

died from two weeks to 17 years after having receivedthe replacement, therefore including cases with long-

term in vivo adaptive bone remodelling. Comparisons

were made between the contralateral intact and im-

planted femurs. Their results showed that stress shield-

ing remained evident, especially in the medial proximal

cortex, even after up to 17 years of implantation. The

results of this study, coupled with those of Maloney et al.

(1989), would not support the theory that bone resorp-tion continues until the restoration of normal strain

patterns is attained. Other factors are probably in-

volved, where a certain amount of bone mass remains

present, regardless of the resulting strain patterns. In

view of the similar results to Maloney et al. (1989), and

the representative nature of the composite femur as a

human specimen (Cristofolini et al., 1996b), the testing

protocols described in this paper can be said to modelcorrectly the stress shielding trends observed from

in vivo specimens.

4.5. If different profiles exist between prostheses designs?

A number of techniques were used to assess the

stress–strain profiles of both stems. While, at a first

glance, significant differences were found between thetwo different stem types, this was only noted at three of

the thirteen strain gauge positions for both implanted

and remodelled conditions. The post hoc statistics re-

vealed that the diaphysis region of the femur, below the

calcar, was that region where such differences were sig-

nificant. In a further examination of the strain profiles,

an analysis of the gradients indicated that the Lubinus

534 V. Waide et al. / Clinical Biomechanics 18 (2003) 523–536

SPII stem had a more gradual transition between thestrain values of adjacent gauges, when compared to the

transition for the M€uuller-Curved stem. This trend oc-

curred in both the implanted and remodelled states, with

the M€uuller-Curved stem demonstrating a greater differ-

ence between adjacent gauges (see mean strain difference

values in Table 4 of Appendix A). This was also evident

from the results of the three-way interaction Scheff�ee testof the original strain data, showing more incidences ofstatistical differences between the principal strain values

of adjacent gauges for the M€uuller-Curved stem (Fig. 7).

The FE results confirmed the trends reported for the

experimental strains of both bone and cement. The ap-

plication of the FE methods to the implanted femurs is

described in further detail elsewhere (Stolk et al., 2002),

nevertheless, the methods applied to the remodelled

femur are included to assist in analysing the resultantdata. In particular, it was desired to use these FE

models, once validated against the experimental data, to

indicate whether differences could be observed between

the two stems, especially after remodelling. The results

of the FE method showed that the two stems had fairly

similar stress distributions and changes with remodel-

ling, with only small differences in the locations of the

peaks.Further to these analyses, it may be concluded that

subtle differences were observed between the two stems

for the loading conditions examined. What is important

to note is that remodelling was found to have the same

effect on cement strains (an increase in strain), regardless

of the stem type. Though the two stems tested have

substantially different clinical outcomes (Malchau et al.,

2000), the similar response of both stems to remodellingmay indicate that remodelling is not necessarily the main

reason for the greater incidences of failure of the M€uuller-Curved stem. Indeed, it should be noted that this was

not an assumption made at the outset of this work.

5. Conclusions

A series of procedures were successfully combined

together and used to investigate in vitro the performance

of two designs of cemented stems, subjected to different

femoral conditions. The remodelling condition was

shown to increase cement strains, a potential indicator

of subsequent in vivo cement failure in fatigue. How-

ever, remodelling did not restore bone strains to their

original intact values, with stress shielding still evident,particularly in the medial proximal region, a finding in

agreement with the literature. In addition, under the

conditions examined, the resultant strains were not

deemed to reach a level critical for the bone. The

methods were shown to be only able to discriminate

subtle differences between two different stem designs,

with an analysis of the strain gradients showing that the

M€uuller-Curved prosthesis has a more severe straintransition along the femoral diaphysis than the Lubinus

SPII prosthesis. Nevertheless, as noted above, remod-

elling may not necessarily be the main cause for failure

with clinical evidence of well-fixed cemented stems

(Schmalzried et al., 1993; Maloney et al., 1989; Maloney

et al., 1996). Given that, to the best of the authors�knowledge, no other comparable in vitro method has

been reported for examining bone remodelling, themethods described are proposed as one of a series of

techniques to assist in the process of pre-clinical evalu-

ation of prostheses designs, e.g. for assessing some

‘‘what if?’’ scenarios.

Acknowledgements

This work is part of a project financed by the Euro-

pean Commission under the Standards Measurement

and Testing programme (SMT4-CT96-2076 and SMT4-CT98-9033). Thanks are due to Waldemar Link for

donation of the Lubinus SPII stems, to Tecres for the

Cemex bone cement, and to Scandimed for the OptiVac

mixing and injection system, used for the implantations.

The authors also gratefully acknowledge the input of a

number of people: Marco Viceconti for suggestions and

advice; Mauro Ansaloni for CAD details on the Stan-

dardized Femur; Roberta Fognani, Stefano Muratori,Donald Nosari, and Chiara Bertaccini for technical

support at various stages during the development of the

protocols; Cristina Ancarani and Barbara Bordini for

statistical assistance; and Luigi Lena for completion of

the graphical work.

Appendix A

A.1. Calculation of root mean square strain difference

The term, the root mean square strain difference, is

defined as a measure of the strain transition, for the two

different prostheses, in each of the three regions instru-

mented with strain gauges. Therefore, in the following

equations the subscript P denotes the prosthesis type

(P¼Lub or Mul, indicating the Lubinus SPII or M€uuller-Curved stems, respectively). The subscript R defines the

region being examined (R¼M, L, or C, denoting the

medial and lateral cortical regions, and the embedded

cement mantle region, respectively).

Therefore, the mean principal strain value, for a

femur (i) examined with a stem design (P ), at a particulargauge location (j) is defined by eðPRÞi;j and consequently,

the term eðPRÞi;jþ1 refers to the mean principal strainvalue at the successive adjacent gauge. Given results for

n femurs for stem design (P ), and m strain gauges in the

region concerned (m ¼ 5 for both the medial and lateral

V. Waide et al. / Clinical Biomechanics 18 (2003) 523–536 535

composite cortical bone regions, and m ¼ 3 for the ce-ment mantle), then the root mean square strain difference

for each femur (RMSeðPRÞi) is defined by,

RMSeðPRÞi ¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiPm�1j¼1 ðeðPRÞi;jþ1 � eðPRÞi;jÞ2

m� 1

sðA:1Þ

The mean and standard deviation of the root mean

square strain difference was calculated for the n femursfor each of the femur conditions investigated, with the

results reported in Table 4.

References

Aamodt, A., Lund-Larsen, J., Eine, J., Andersen, E., Benum, P.,

Husby, O.S., 1997. In vivo measurements show tensile axial strain

in the proximal lateral aspect of the human femur. Journal of

Orthopaedic Research 15 (6), 927–931.

Biewener, A.A., 1993. Safety factors in bone strength. Calcified Tissue

International 53 (Suppl. 1), S68–S74.

Carter, D.R., Caler, W.E., Spengler, D.M., Frankel, V.H., 1981.

Fatigue behavior of adult cortical bone: the influence of mean

strain and strain range. Acta Orthopaedica Scandinavica 52 (5),

481–490.

Cristofolini, L., 1997. A critical analysis of stress shielding evaluation

of hip prostheses. Critical Reviews in Biomedical Engineering 25

(4–5), 409–483.

Cristofolini, L., Cappello, A., McNamara, B.P., Viceconti, M., 1996a.

A minimal parametric model of the femur to describe axial elastic

strain in response to loads. Medical Engineering and Physics 18 (6),

502–514.

Cristofolini, L., McNamara, B.P., Freddi, A., Viceconti, M., 1997.

In vitro measured strains in the loaded femur: quantification of

experimental error. Journal of Strain Analysis for Engineering

Design 32 (3), 193–200.

Cristofolini, L., Viceconti, M., 1999a. In vitro stress shielding

measurements can be affected by large errors. Journal of Arthro-

plasty 14 (2), 215–219.

Cristofolini, L., Viceconti, M., 1999b. Towards the standardization of

in vitro load transfer investigations of hip prostheses. Journal of

Strain Analysis for Engineering Design 34 (1), 1–15.

Cristofolini, L., Viceconti, M., 2000. Development and validation of a

technique for strain measurement inside polymethyl methacrylate.

Journal of Strain Analysis for Engineering Design 35 (1), 21–33.

Cristofolini, L., Viceconti, M., Cappello, A., Toni, A., 1996b.

Mechanical validation of whole bone composite femur models.

Journal of Biomechanics 29 (4), 525–535.

Dally, J.D., 1989. Statistical analysis of experimental data. In: Doyle,

J.F., Phillips, J.W. (Eds.), Manual on Experimental Stress Anal-

ysis, fifth ed. SEM Inc., Bethel, pp. 15–33.

Davies, J.P., Burke, D.W., O�Connor, D.O., Harris, W.H., 1987.

Comparison of the fatigue characteristics of centrifuged and

uncentrifuged Simplex P bone cement. Journal of Orthopaedic

Research 5 (3), 366–371.

Ebramzadeh, E., Sarmiento, A., McKellop, H.A., Llinas, A., Gogan,

W., 1994. The cement mantle in total hip arthroplasty. Analysis of

long-term radiographic results. Journal of Bone and Joint Surgery.

American Volume 76 (1), 77–87.

Estok, D.M., Orr, T.E., Harris, W.H., 1997. Factors affecting cement

strains near the tip of a cemented femoral component. Journal of

Arthroplasty 12 (1), 40–48.

Fisher, D.A., Tsang, A.C., Paydar, N., Milionis, S., Turner, C.H.,

1997. Cement-mantle thickness affects cement strains in total hip

replacement. Journal of Biomechanics 30 (11–12), 1173–1177.

Herberts, P., Malchau, H., 2000. Long-term registration has improved

the quality of hip replacement: a review of the Swedish THR

Register comparing 160,000 cases. Acta Orthopaedica Scandinav-

ica 71 (2), 111–121.

James, S.P., Jasty, M., Davies, J., Piehler, H., Harris, W.H., 1992. A

fractographic investigation of PMMA bone cement focusing on the

relationship between porosity reduction and increased fatigue life.

Journal of Biomedical Materials Research 26 (5), 651–662.

Jasty, M., Maloney, W.J., Bragdon, C.R., O�Connor, D.O., Haire, T.,Harris, W.H., 1991. The initiation of failure in cemented femoral

components of hip arthroplasties. Journal of Bone and Joint

Surgery. British Volume 73 (4), 551–558.

Malchau, H., Herberts, P., S€ooderman, P., Od�een, A., 2000. Prognosis of

total hip replacement. Update and validation of results from the

Swedish National Hip Arthroplasty Registry 1979–1998. In: 67th

Annual Meeting of the American Academy of Orthopaedic

Surgeons, Orlando, USA.

Maloney, W.J., Jasty, M., Burke, D.W., O�Connor, D.O., Zalenski,E.B., Bragdon, C., et al., 1989. Biomechanical and histologic

investigation of cemented total hip arthroplasties. A study of

autopsy-retrieved femurs after in vivo cycling. Clinical Orthopae-

dics and Related Research 249, 129–140.

Maloney, W.J., Sychterz, C., Bragdon, C., McGovern, T., Jasty, M.,

Engh, C.A., et al., 1996. The Otto Aufranc Award. Skeletal

response to well fixed femoral components inserted with and

without cement. Clinical Orthopaedics and Related Research 333,

15–26.

McCormack, B.A., Prendergast, P.J., 1999. Microdamage accumula-

tion in the cement layer of hip replacements under flexural loading.

Journal of Biomechanics 32 (5), 467–475.

McNamara, B.P., Cristofolini, L., Toni, A., Taylor, D., 1994.

Evaluation of experimental and finite element models of synthetic

and cadaveric femora for pre-clinical design-analysis. Clinical

Materials 17 (3), 131–140.

O�Connor, D.O., Burke, D.W., Jasty, M., Sedlacek, R.C., Harris,

W.H., 1996. In vitro measurement of strain in the bone cement

surrounding the femoral component of total hip replacements

during simulated gait and stair-climbing. Journal of Orthopaedic

Research 14 (5), 769–777.

Pattin, C.A., Caler, W.E., Carter, D.R., 1996. Cyclic mechanical

property degradation during fatigue loading of cortical bone.

Journal of Biomechanics 29 (1), 69–79.

Rubash, H.E., Sinha, R.K., Shanbhag, A.S., Kim, S.Y., 1998.

Pathogenesis of bone loss after total hip arthroplasty. Orthopaedic

Clinics of North America 29 (2), 173–186.

Ruff, C.B., Hayes, W.C., 1983. Cross-sectional geometry of Pecos

Pueblo femora and tibiae––a biomechanical investigation: I.

Method and general patterns of variation. American Journal of

Physical Anthropology 60 (3), 359–381.

Schmalzried, T.P., Maloney, W.J., Jasty, M., Kwong, L.M., Harris,

W.H., 1993. Autopsy studies of the bone–cement interface in well-

fixed cemented total hip arthroplasties. Journal of Arthroplasty 8

(2), 179–188.

Schmalzried, T.P., Szuszczewicz, E.S., Northfield, M.R., Akizuki,

K.H., Frankel, R.E., Belcher, G., et al., 1998. Quantitative

assessment of walking activity after total hip or knee replacement.

Journal of Bone and Joint Surgery. American Volume 80 (1), 54–59.

Stolk, J., Verdonschot, N., Cristofolini, L., Toni, A., Huiskes, R.,

2002. Finite element and experimental models of cemented hip joint

reconstructions can produce similar bone and cement strains in

pre-clinical tests. Journal of Biomechanics 35, 499–510.

Szivek, J.A., Gealer, R.L., 1991. Comparison of the deformation

response of synthetic and cadaveric femora during simulated one-

legged stance. Journal of Applied Biomaterials 2 (4), 277–280.

Verdonschot, N., Huiskes, R., 1997. The effects of cement-stem

debonding in THA on the long-term failure probability of cement.

Journal of Biomechanics 30 (8), 795–802.

536 V. Waide et al. / Clinical Biomechanics 18 (2003) 523–536

Viceconti, M., Casali, M., Massari, B., Cristofolini, L., Bassini,

S., Toni, A., 1996. The �standardized femur program� proposal for areference geometry to be used for the creation of finite ele-

ment models of the femur. Journal of Biomechanics 29 (9),

1241.

Waide, V., Cristofolini, L., Toni, A., 2001a. A CAD–CAM method-

ology to produce bone remodelled composite femurs for pre-

clinical investigations, Proceedings of the Institution of Mechanical

Engineers. Journal of Engineering in Medicine, Part H 215, 459–

469.

Waide, V., Cristofolini, L., Toni, A., 2001b. Development of exper-

imental bone remodelled femur specimens using CAD/CAM

techniques. In: Middleton, J., Jones, M.L., Shrive, N.G., Pande,

G.N. (Eds.), Computer Methods in Biomechanics & Biomedical

Engineering, vol. 3. Gordon and Breach Science Publishers,

Singapore, pp. 135–140.