Non-invasive in vivo quantification of the medial tibial cartilagethickness progression in an osteoarthritis rabbit modelwith quantitative 3D high resolution micro-MRI1

C. Boulocher M.Sc., Dr Med. Vet.yz*, E. Chereul Ph.D.x, J. B. Langlois B.Sc.x,M. Armenean Ph.D.x, M. E. Duclos M.Sc., Dr Med. Vet.yz, E. Viguier Pr, Ph.D., Dr Med. Vet.yz,T. Roger Pr, Ph.D., Dr Med. Vet.yz and E. Vignon Pr, Ph.D.yk{yUniversite de Lyon, UR RTI2B, Lyon F-69003, FrancezUniversite de Lyon, Ecole Nationale Veterinaire de Lyon (ENVL), Marcy l’Etoile, F-69280, FrancexUniversite de Lyon, Plateforme Animage, Genopole Rhone-Alpes, FrancekUniversite Lyon 1, 69008 Lyon, France{Hospices civils de Lyon, Service de Rhumatologie, Pierre Benite F-69495, France

Summary

Objective: To develop a quantitative non-invasive in vivo three-dimensional (3D) high resolution (HR) micro-magnetic resonance imaging(mMRI) protocol to measure the medial tibial cartilage thickness (MT.ThC) in the normal rabbit and in the anterior cruciate ligament transection(ACLT) rabbit model of osteoarthritis and quantify the progression of MT.ThC.

Methods: The left knee of 10 control and 40 operated rabbits was imaged in vivo with a 7 T mMRI system at 3 and 5 months after ACLT. A 3D fastlow angle short (FLASH) fat-suppressed MRI protocol was implemented leading to 44! 176 mm3 spatial resolution and to 44 mm3 isotropic voxelafter cubic interpolation. Semi-automatic MT.ThC measurements were made in 3D, in four different locations, in vivo and longitudinally in bothgroups. At 5 months, gross macroscopy, visual analogical evaluation of the cartilage and histology were compared to the MR-based MT.ThC.

Results: At 3 and 5 months, the MT.ThC measured in the minimum interbone distance area was the thinnest MR-based MT.ThC. It wassignificantly lower in the operated group and among the four evaluated MT.ThC, it was the most discriminative between the normal andthe operated groups (P< 0.05). The MT.ThC measured in the minimum interbone distance area was also the most sensitive to change inthe operated group (66.4% MT.ThC loss, P¼ 0.003) while no significant changes were observed in the control group.

Conclusion: Quantitative 3D HR mMRI allowed for non-invasive longitudinal MT.ThC measurements in four different locations in both thenormal and the operated rabbits. We concluded the MT.ThC measured in the minimum interbone distance area reflected the severity ofthe disease and was the most effective to measure the progression of the medial tibial cartilage destruction.ª 2007 Osteoarthritis Research Society International. Published by Elsevier Ltd. All rights reserved.

Key words: Osteoarthritis, Experimental model, MRI, Cartilage, MR-based thickness, In vivo.

Introduction

Although osteoarthritis (OA) is a very common cause ofdisability and pain in the elderly its natural history is poorly un-derstood1,2. During the development of OA, one observesa progressive deterioration in quality and a decrease in thick-ness of the articular cartilage ultimately causing denudationof the joint surface. Assessment of changes of morphologyof the cartilage is critical for monitoring OA progression3

and the evaluation of the therapeutic effects of chondropro-tective drugs4,5. Accurate and sensitive diagnostic methodsare necessary to monitor therapeutic effect and detect very

early stages of the disease when pathology is likely to bereversible and/or progression controlled6. Experimental stud-ies in animal models7 that closely mimic human degenerativejoint disease are essential for both understanding the dis-ease8 and facilitating development of therapies such asstructure-modifying OA drugs. Experimental studies in ani-mal models are required to prove the existence of a therapeu-tic effect, before clinical trials can be proposed.

In humans, conventional radiography is currently the ‘‘goldstandard’’ for joint space width measurements to evaluatecartilage thickness but gives only an indirect assessment ofthe hyaline articular cartilage and is insensitive to detect car-tilage fibrillation, cracking, or erosion. In contrast to radiogra-phy, magnetic resonance imaging (MRI) is capable of directlyvisualizing the articular cartilage. MRI has been increasinglyinvestigated in OA because it allows accurate morphologicassessment and reproducible quantitative measurementsof the cartilage9e14. Quantitative MRI was proven to be anaccurate method for cartilage volumetric and thickness mea-surements, and has been used in OA patients for over 10years15,16. MRI measurements of cartilage thickness were

1Financial support was provided in part by Vetoquinol, Rhone-AlpesGenopole, Fondation Rhone-Alpes Futur and the Institut ClaudeBourgelat (IClB).

*Address correspondence and reprint requests to: CarolineBoulocher, M.Sc., Dr Med. Vet., Ecole Nationale Veterinaire deLyon, 1 Avenue Bourgelat, 69280 Marcy l’Etoile, France. Tel: 33-4-78-87-25-31; Fax: 33-4-78-87-25-33; E-mail: c.boulocher@vet-lyon.fr

Received 18 October 2006; revision accepted 24 April 2007.

Osteoarthritis and Cartilage (2007) 15, 1378e1387ª 2007 Osteoarthritis Research Society International. Published by Elsevier Ltd. All rights reserved.doi:10.1016/j.joca.2007.04.012

InternationalCartilageRepairSociety

1378

similar to the ones obtained with histology17, radiogra-phy18,19, computed tomography10,18, arthroscopy10,19e20,ultrasonography21 and stereophotogrammetry22. MRI mea-surements of cartilage volume23 correlated well to volumesmeasured after surgical removal13.

In animal models, there is no radiographic standardizedprocedure under weight bearing conditions to evaluate thejoint space width. Histologic and macroscopic assessmentsremain the ‘‘gold standards’’ to evaluate OA progression inexperimental studies. However, they preclude longitudinalstudies as they can only be completed after animal sacri-fice. Increasing ethical and cost-related issues triggeredthe development of alternative methods for accurate andlongitudinal follow-up of experimental animals with OA,such as MRI8,24. MRI allows accurate evaluation of thearticular cartilage in the rabbit24e27, guinea pig28, Rhesusmacaque29, dog30, rat31 and porcine models32. In the ante-rior cruciate ligament transected (ACLT) rabbit model33,in vivo micro-magnetic resonance (mMR) scoring of carti-lage lesions was reported24,27 but no longitudinal in vivoquantitative measurement of cartilage using three-dimen-sional micro-magnetic resonance imaging (3D-mMRI) hasbeen reported so far.

The purpose of the present study was to assess quantita-tively the cartilage degradation at various time points andlongitudinally in the rabbit ACLT model of OA. We devel-oped the protocol based on previous study on the guineapig knee28. Medial tibial cartilage thickness (MT.ThC)34

sensitivity to change was compared in four locations ofthe medial tibial cartilage (MTC) in the operated and in thecontrol rabbits. MT.ThC discriminative power was evaluatedbetween both groups at 3 and 5 months after ACLT.

Materials and methods

EXPERIMENTAL ANIMAL MODEL

All work was conducted in full compliance with the ENVL(Ecole Nationale Veterinaire de Lyon) ethical committeeguidelines for animal protection and in accordance with thelegislation of the European Community. Fifty White New Zea-land rabbits were used. After 10 days for acclimation and quar-antine, experimental OA was surgically induced in the left kneeof 40 rabbits (3.110# 0.250 kg) by ACLT. Right knees wereleft intact and were not evaluated in this study. The remaining10 rabbits (2.990# 0.163 kg) were not surgically modified.The operated and the control rabbit groups were homogenousin growth and weight. Radiography showed opened proximaltibial physes prior to the surgery and at 3 months. They wereclosed in all rabbits at 5 months. Distal femoral physes wereopened prior to the surgery and were closed at 3 months inall rabbits. The control and the operated rabbits were allhoused over the same time period in our Experimental Medi-cine and Surgery Center and kept in individual cages witha grid floor (surface area: 3000 cm2, height: 70 cm).

SURGICAL PROCEDURE

Rabbits from the operated group were anesthetized byintramuscular sedation. Left knee arthrotomy was performedvia a medial parapatellar approach under sterile conditionsas described in the literature35. The left anterior cruciateligament (ACL) was sectioned without affecting thesurrounding structures. Joint capsule, subcutaneous tissuesand skin were closed layer by layer with running sutures andstrict attention to hemostasis. The anterior drawer test wasconsistently positive after the operation. After surgery,

analgesia was provided with Fentanyl which has no effecton OA progression [25 mg ear patch (Duragesic! Transder-mal Patch, Janssen Pharmaceutica)]. The leg was not immo-bilized and rabbits were allowed to move freely in their cages.Recovery was closely monitored by veterinarians.

NON-INVASIVE IN VIVO MRI

Non-invasive in vivo 7 T (m-qMRI) examinations weresequentially performed at 3 and 5 months following the sur-gery on the left knee of the 40 operated rabbits and on theleft knee of the 10 control rabbits.

MRI hardware

MRI was done on a 70/20 horizontal Biospec! supercon-ducting magnet system (Bruker!, Karlsruhe, Germany) witha 210 mm bore and 72 mm free access for imaging dia-meters. This magnet has a 7.05 T nominal magnetic fieldstrength and high performance unshielded gradients (Mini im-aging Bruker!) with a 400 mT/m maximum gradient strength.

Radiofrequency coils and animal handling

Separate receiving and emitting coils with an active decou-pling system were used to increase the signal-to-noise ratio(SNR). At 3 months, a commercial 15 mm diameter receptioncoil (Bruker!) and a rectangular 8! 10 cm custom madesurface emitting coil were used. At 5 months, the receivingcoil was customized to fit the enlarged osteoarthritic knees(25 mm diameter with a central hole) [Fig. 1(a)]. Anesthesiawas induced by intramuscular injection of a combination ofketamine and xylazine. The ‘‘rabbit-bed’’ had identificationmarks drawn on its side to provide necessary markers foraccurate and reproducible coil positioning. The left pelviclimb was shaved to facilitate the anatomical landmark recog-nition and a splint was adjusted on its lateral side to achievefull extension of the leg. A dedicated Styrofoam pad wasadjusted with surgical tape around the splint, the knee andthe receiving coil, and then both legs were immobilized withsurgical tape [Fig. 1(b)]. The emitting coil was fixed on thetop using surgical tape [Fig. 1(c)]. The end stop of the bedwas set to get the knee joint and receiving coil at the levelof the magnet isocenter. The external diameter of the bed fit-ted the internal bore magnet diameter, preventing motionartifacts. Rabbits were kept anesthetized with 1e2% isofluranein 0.6e1 L/min O2, and the anesthetic gases were suppliedwith a facemask. The temperature was maintained usinga warming plastic blanket with circulating hot water [Fig. 1(d)].

MRI protocol

An automatic shimming process was performed at thebeginning of each scan to homogenize the main magneticfield. Spectrometer control and data analysis were per-formed with the 3.0.2.ParaVisionª version (Bruker,Germany) software. The sagittal high resolution (HR) 3Dfast low angle short (FLASH) fat-suppressed sequence pa-rameters were 30$ flip angle, repetition time (TR) 22.1 ms,echo time (TE) 4.2 ms, 26 min acquisition time, 21 kHz re-ceiver bandwidth, 2.25! 2.25! 2.25 cm field of view(FOV) and 256! 140! 85 acquisition matrix (givinga 88! 161! 265 mm3 acquisition resolution). The automaticreconstruction immediately followed and resulted ina 256! 256! 128 reconstruction matrix with the sameFOV leading to an 88! 88! 176 mm3 resolution. Imageswere prepared by zero-filling process to achieve a final reso-lution of 44! 44! 176 mm3 (2.25! 2.25! 2.25 cm3 FOVand 512! 512! 128 reconstruction matrix) (Fig. 2).

1379Osteoarthritis and Cartilage Vol. 15, No. 12

MRI-BASED MT.THC MEASUREMENTS

The observer was unaware of the rabbit status and wasexperienced28 in MRI image processing and MR-based car-tilage thickness measurements. Several measurementswere extracted in 3D space after image segmentation.The main steps of the procedure were selection of regionof interest, semi-automatical bone segmentation andcartilage thickness measurements with CartiLapª (Creatis-Animage, Lyon, France). The segmentation step of thesoftware gave an accurate 3D delineation of the tibial andfemoral subchondral bone surfaces using two ‘‘3D Snake

active surface model’’36,37. Two initial meshed spheres eone for the tibia and one for the femur e were iterativelydeformed under a field of local forces depending on imageinformation content that pulled them toward features suchas lines and edges. At convergence, 3D meshes (for thetibial and femoral subchondral bone surfaces) were ob-tained (Fig. 3). Finally, the software detected automaticallythe minimal distance between the femoral and the tibial3D surfaces (Fig. 4). The observer measured each MT.ThCin 3D, along a line perpendicular to the bone surface. Delin-eation of femoral and tibial cartilage boundaries was

Fig. 1. Radiofrequency coils’ fixation and animal handling. (a) Receiving coil (2.5! 2.5 cm2) and rectangular surface (8! 10 cm2) emitting coil.(b) The emitting coil was maintained steady on the top with surgical tape. (c) A splint was adjusted on the lateral side of the left leg to achievefull extension. A Styrofoam pad was adjusted with surgical tape around the splint, the knee and the receiving coil. (d) Gaseous anesthesia was

supplied with a facemask and the temperature was maintained using a warming plastic blanket with circulating hot water.

Tibio-patellar ligament

Distal femoral physis

Proximal tibial epiphysis

Medial fabella

Distal femoral epiphysis

Tibial bone diaphysis

Infrapatellar fat padFemoral cartilage

Tibial cartilage

anterior posterior

Fig. 2. HR 3D FLASH fat-suppressed sagittal MR image.

1380 C. Boulocher et al.: Non-invasive in vivo quantification of the MT.ThC progression in an OA rabbit model

detected visually and with the line intensity profile (LIP) (ie.gray curve level). The LIP showed two high signal peaksrepresenting medial tibial and femoral cartilages and a char-acteristic signal drop at their boundaries.

‘‘Thick-1’’ parameter was the MT.ThC in the minimuminterbone 3D area28. ‘‘Thick-2’’ and ‘‘Thick-3’’ parameterscorresponded, respectively, to the anterior and posteriorseparation of the MTC from the femoral cartilage. ‘‘Thick-4’’ was measured at the thickest part of MTC.

Semi-quantitative evaluation of the technique

Semi-quantitative evaluation was done a posteriori andboth MR image and MR measurement qualities were scored.This evaluation was original and based on the experience ofthe observer.

A six grade scale was attributed to the image quality (car-tilage visibility and delineation, fat suppression efficiency)from grade 0 meaning low quality (cartilage boundariesdifficult to define) to grade 5 meaning high quality (clearvisualization of the cartilage interfaces).

According to the LIP a four grade scale was attributed tothe MT.ThC measurements quality from grade 0 meaninglow quality (no drop within the LIP) to grade 3 meaninghigh quality (important drop).

GROSS MORPHOLOGY, VISUAL ANALOGICAL EVALUATION(VAE) OF THE CARTILAGE AND HISTOLOGY(SEMI-QUANTITATIVE AND HISTOMORPHOMETRY)

Five months after surgery, the rabbits were euthanized byinjection of Pentobarbital Sodium. All left knees from oper-ated and control rabbits were dissected. Two investigatorsunaware of the rabbit status performed the macroscopicevaluations. Each compartment of the knee joint wasevaluated. Prior to the sampling for histological evaluation,photographs were collected and labeled for subsequentevaluation. Osteophytes’ production and meniscal lesionswere noted and recorded.

Gross morphological cartilage changes of the femoro-tibial compartment were scored using VAE. This scorewas based on the International Cartilage Repair Society(ICRS) recommendations for grading cartilage defects38.Briefly, the VAE score is the product of the percentage ofthe area involved and a factor based on the grade of thecartilage lesion38. The scale ranges from 0 indicating thatthe cartilage is intact to 100 meaning that full-thicknesscartilage erosion occurred across the entire MTC plate38.

After macroscopic grading, left proximal tibial bones wereisolated, fixed in buffered formalin solution and decalcifiedfor 1 month in formic acid. After decalcification these sam-ples were embedded in paraffin, sagittal sections of 6 mm

Medial tibial subchondral bone

Tibial cartilage

Femoral cartilageAnterior horn of the medialmeniscus

3D mesh model of the femoral subchondral bone surface

Initial sphere for the 3D mesh model of the tibial subchondral bone surface

anterior posterior

Fig. 3. CartiLapª Snake procedure giving the subchondral bone surface 3D mesh model. Based on the 3D Snake active surface model twoinitial spheres e one for the tibia and one for the femur e were iteratively deformed to fit the 3D surface of the subchondral bones.

Fig. 4. Automatical minimal distance detection between the femoral and the tibial 3D surfaces. (a) In a control rabbit (rabbit 5N) at 3 months,(b) in an operated rabbit (rabbit 2J) at 3 months and (c) in an operated rabbit (rabbit 2J) at 5 months.

1381Osteoarthritis and Cartilage Vol. 15, No. 12

were cut from the central part of the medial tibial plateau(where the MR region of interest was centered) and stainedwith hematoxylin and eosin.

Semi-quantitative histopathological grading was perform-ed by observers blinded to the rabbit status, based on theOsteoarthritis Research Society International (OARSI) grad-ing system39. Grades were assessed by noting the mostadvanced lesion present within the cartilage, irrespectiveof its horizontal extent (grade¼ depth of cartilage ero-sion). In the OARSI grading system39 grades 0e4 involvearticular cartilage changes only, whereas grades 5 and 6involve subchondral bone as well.

Finally, histomorphometry40 was performed (Novotec,Lyon, France). The thinnest and the thickest part of themedial tibial samples were measured and compared,respectively, with Thick-1 and Thick-4 values obtainedfrom MRI images analyses.

STATISTICAL ANALYSIS

Statistical analyses were performed using a statisticalpackage software (R software version 2.2.1, Ihaka, 1996).Unless stated otherwise, results are expressed as meanand standard deviation (mean#SD). Paired Student’st tests with non-equal variances (Welch two sample t tests)were performed to compare quantitative data (MT.ThCwithin and between groups). Scores were ranked andcompared using non-parametric statistics (the Wilcoxonrank-sum test to compare groups). Pearson tests wereused to correlate macroscopic, histological and MRI results.A P< 0.05 was considered statistically significant.

Results

MRI-BASED MT.THC

Semi-quantitative evaluation of the technique

Image quality. In the control group, image quality rangedfrom 3.1 (#0.8) to 4.3 (#0.7) at 3 and 5 months, respec-tively. In the operated group, image quality ranged from2.7 (#0.7) to 3.8 (#0.6) out of five at 3 and 5 months, re-spectively. Results are shown in Table I.

1. Cross-sectional analyses: At 3 months, the imagequality was not significantly different between the con-trol and the operated groups (P¼ 0.08). At 5 months,the image quality was significantly higher in the controlgroup than in the operated group (P¼ 0.045).

2. Longitudinal analyses: The image quality was signifi-cantly higher at 5 months than at 3 months in the con-trol group (P¼ 0.01) and in the operated group(P< 0.0001).

MT.ThC measurement quality. In the control group, MT.ThCmeasurement quality ranged from 2.6 (#0.60) to 3.0(#0.00) at 3 and 5 months, respectively. In the operatedgroup, MT.ThC measurement quality ranged from 2.7(#0.5) to 2.8 (#0.4) out of five at 3 and 5 months, respec-tively. Results are shown in Table I.

1. Cross-sectional analyses: At 3 months, the measure-ment quality of the control group was not significantlydifferent than in the operated group (P> 0.05). At 5months, the measurement quality was significantlyhigher in the control group than in the operated group(P< 0.05).

2. Longitudinal analyses: In both groups, the measure-ment quality was higher at 5 months than at 3 months,but not significantly (P> 0.05).

MRI-based MT.ThC results (Table II)

1. Cross-sectional MT.ThC analyses e comparison be-tween groups at 3 and 5 months.

e Thick-1 was lower in the operated group than in thecontrol group. This difference was not significant at3 months (P> 0.05) but was significant at 5 months(P< 0.0001). Full-thickness erosion (MT.ThC¼0 mm) was measured in eight of the 40 operatedrabbits (20%) at 3 months and in 17 of the 40 oper-ated rabbits (42.5%) at 5 months. If rabbits withMT.ThC¼ 0 mm were excluded, Thick-1 was stilllower in the operated group at 3 months (not signif-icantly) and significantly at 5 months (P¼ 0.05).

e Thick-2 was significantly higher in the operatedgroup than in the control group at 3 months(P¼ 0.005) and at 5 months (P¼ 0.009).

e Thick-3 was higher in the operated group than in thecontrol group but not significantly (P> 0.05) at 3and 5 months.

e Thick-4 was significantly higher in the operatedrabbits than in the control group at 3 months(P¼ 0.0002) and at 5 months (P¼ 0.0005).

2. Longitudinal MT.ThC changes e comparison betweengroups.

e Thick-1: In the control group, Thick-1 changes werenot significant (P> 0.05) between 3 and 5 months.In the operated group, Thick-1 decreased signifi-cantly (P< 0.05) with 66.4% of thickness loss.Thick-1 was still significantly decreased after beingscaled to the initial thickness and if rabbits withMT.ThC¼ 0 mm were excluded from the mean.Thick-1 absolute change was significantly higher(P¼ 0.003) in the operated group than in the controlgroup, even if rabbits with Thick-1¼ 0 mm were ex-cluded (P¼ 0.05).

e Thick-2 increased in both groups. In the operatedgroup, Thick-2 increase was significant (P< 0.0

Table IMR image quality score from 0 to 5 and MR-based MT.ThC mea-surement quality score from 0 to 3. Results are presented as

mean and (SD)

Qualitative evaluationof the technique

Images quality(0e5)

Measurement quality(0e3)

3 Months 5 Months 3 Months 5 Months

Operated rabbits(n¼ 40)

2.7(0.7)

3.8(0.6)

2.7(0.5)

2.8(0.4)

Longitudinalcomparison (P )

*** NS

Control rabbits(n¼ 10)

3.1(0.8)

4.3(0.7)

2.6(0.6)

3.0(0.0)

Longitudinalcomparison (P )

** NS

Cross-sectionalcomparison

NS * NS *

NS: nonsignificant P> 0.05, Student’s paired t test. *Significancelevel P< 0.05, Student’s paired t test; **Significance level P< 0.01,Student’s paired t test and ***Significance level P< 0.001,Student’s paired t test.

1382 C. Boulocher et al.: Non-invasive in vivo quantification of the MT.ThC progression in an OA rabbit model

001) but not in the control group (P> 0.05). Thick-2absolute change was significantly higher in the oper-ated group than in the control group (P¼ 0.027).

e Thick-3 and Thick-4 increased in both groups butchanges were not statistically significant (P> 0.05).Longitudinal changes were not statistically differentbetween groups and varied widely between theoperated rabbits.

GROSS MORPHOLOGY (OSTEOPHYTES PRODUCTION,MENISCAL LESIONS), VAE OF THE CARTILAGE ANDHISTOLOGY (SEMI-QUANTITATIVE ANDHISTOMORPHOMETRY)

Gross morphology (osteophytes’ production, meniscallesions) was significantly different between the operatedand the control groups (P< 0.001). In the control group, nogross abnormalities and no meniscal injuries were noticed.The ACL was intact in all control rabbits. In the operatedgroup, complete section of the anterior cranial cruciate liga-ment was present in all but one rabbit which was excludedfrom further analyses. Osteophytes and fibrosis were ob-served in all operated rabbit knees. Medial meniscal damagewas present in 25 of the 40 operated rabbits (62.5%) and in-cluded longitudinal tears, extrusion and bucket handle tear ofthe left medial meniscus. Furthermore, the medial tibial pla-teaus had undergone extensive remodeling, with prominentosteophytes at the posterior lip of the medial tibial condyle.

In the operated group, the MTC VAE score was signifi-cantly higher (45.8# 14%) than in the control group(29# 9.8%) (P< 0.001). Full-thickness cartilage erosionwas confirmed macroscopically, in the caudal part of the me-dial tibial plateau (i.e., in the weight bearing area uncoveredby the meniscus) of 17 of the 40 operated rabbits.

The semi-quantitative histological grading was signifi-cantly higher in the operated rabbits (1.7# 0.8 out of four)than in the control rabbits (0.3# 0.4 out of four) (P< 10%6).

The thinnest and the thickest histological MT.ThC weremarkedly lower in operated rabbits than in control rabbitsbut surprisingly no statistical difference was observed. Inthe control rabbits, there was a statistically significant positivecorrelation between the thinnest histological MT.ThC and thethinnest MR-based MT.ThC, Thick-1 (r¼ 0. 68, P< 0.05).

Discussion

A number of authors have described the use of MRI in theevaluation of OA in rabbit models8,24e27,41e43 with in planeresolution ranging from 156 to 1100 mm2 in two-dimensional(2D) protocols and from 156 to 468 mm2 in 3D images. Sub-jective or semi-quantitative grading scales were used toassess the changes. To the best of the authors’ knowledge,this is the first report of longitudinal quantification of theMT.ThC using non-invasive in vivo 3D MRI in the ACLTrabbit model of OA.

Bolbos et al.28 assessed at 7 T the cartilage thicknessevolution between meniscectomized (MNX group) andspontaneous (SHAM group) guinea pig OA models witha voxel resolution of 59! 59! 156 mm3 in only 45 min.The short-term precision of the technique was 8.9% for theSHAM group and 8.2% for the MNX group. Based on thisprevious study, we implemented a non-invasive 3D FLASHfat-suppressed MRI protocol to measure in vivo the MT.ThCin the normal rabbit and in the ACLT rabbit model of OA. MRscanning protocol resulted in 44! 44! 176 mm3 spatial

Ta

ble

IIM

R-b

ase

dM

T.T

hC

inth

eo

pe

rate

dg

rou

p(n¼

40

)a

nd

inth

eco

ntr

ol

rab

bit

gro

up

(n¼

10

)a

t3

an

d5

mo

nth

s.R

esu

ltsa

rep

rese

nte

da

sm

ea

na

nd

(SD

)in

mm

MT

.Th

CT

hic

k-1

Th

ick-

2T

hic

k-3

Th

ick-

4

3M

on

ths

5M

on

ths

3M

on

ths

5M

on

ths

3M

on

ths

5M

on

ths

3M

on

ths

5M

on

ths

Op

era

ted

rab

bits

(n¼

40

)2

14

.0(1

82

.3)

13

6.0

(15

5.2

)4

20

.6(1

21

.5)

49

0.8

(14

5.7

)7

73

.3(1

76

.0)

77

7.1

(18

4.1

)9

69

.3(1

73

.1)

96

1.0

(14

3.3

)L

on

gitu

din

al

com

pa

riso

n(P

)**

***

NS

NS

Co

ntr

ol

rab

bits

(n¼

10

)2

65

.1(6

6.2

)2

90

(53

.3)

34

2.6

(57

.7)

36

6.5

(56

.3)

67

2.4

(41

.6)

73

3.0

(17

6.0

)8

50

(10

4.9

)7

85

.8(1

15

.4)

Lo

ng

itud

ina

lco

mp

ariso

n(P

)N

SN

SN

SN

SC

ross

-se

ctio

na

lco

mp

ariso

n(P

)N

S**

***

**N

SN

S**

***

*

NS

:n

on

sig

nifi

can

tP>

0.0

5,

Stu

de

nt’s

pa

ire

dt

test

.*S

ign

ifica

nce

leve

lP<

0.0

5,

Stu

de

nt’s

pa

ire

dt

test

;**

Sig

nifi

can

cele

vel

P<

0.0

1,

Stu

de

nt’s

pa

ire

dt

test

an

d**

*Sig

nifi

can

cele

vel

P<

0.0

01

,S

tud

en

t’sp

aire

dt

test

.

1383Osteoarthritis and Cartilage Vol. 15, No. 12

resolution and 44 mm3 isotropic voxel after cubic interpola-tion. To the best of the authors’ knowledge, this protocol re-sulted in the highest spatial resolution of in vivo mMRIimages of the rabbit knee reported so far.

Longitudinal measurement of cartilage thickness is an im-portant indicator of the disease status and of the responseto therapy of osteoarthritic joints. The advantage of 3D MR-based MT.ThC is that measurements are made in vivo andin 3D along a line perpendicular to the subchondral bonesurface. Detection of subtle changes requires precision,reproducibility, high spatial resolution and good contrast25,44,45.In humans, Duryea et al.46 measured in vivo cartilage thick-ness and volume in the knee with an excellent reproducibil-ity. In OA animal models, MRI studies are inherently moredifficult due to limited spatial resolution and field strengthor resulting in imaging time incompatible with in vivo stud-ies. In the past, in vivo cartilage thickness measurementin small animal models could not be accurately performedwith MRI due to these technical limitations. High fieldstrength MRI systems currently available allow for sufficientSNR, contrast and spatial resolutions with scan times thatare compatible with in vivo imaging17.

2D quantitative MR-based MT.ThC was realized by Calvoet al.27 in the rabbit OA partial-meniscectomy model witha 4.7 T magnet. They found tendency of the cartilage thick-ness to increase up to 8 weeks after surgery and a decreaseat 10 weeks in the two operated knees imaged at this timepoint. Wachsmuth et al.26 used in vivo contrast-enhanced7.1 T MRI in two surgically induced rabbit models of OA(ACLT and medial meniscectomy) using a grading scalebased on qualitative articular cartilage assessment from 2Dand 3D gradient echo images. While Calvo et al.27 observeda global increase in the cartilage thickness, this was not ob-served in the study realized by Wachsmuth et al.26 The rela-tively low in plane resolution (195 mm2) and slice thickness(1 mm) used by Calvo et al.27 could partly explain these differ-ences. In addition, they evaluated mean cartilage thicknessbut local thickness variations were not taken into account.Volume and mean cartilage thickness are ‘‘global’’ parame-ters and may be relatively insensitive to focal changes12.

In our study, the 3D FLASH fat-suppressed MRI sequenceswere not primarily aimed at the qualitative assessment of thesurrounding joint structure but they allowed for sufficient rec-ognition of anatomical landmarks for exact depiction of theMTC surface. No fat signal was observed leading to an excel-lent discrimination between cartilage, menisci and sub-chondral bone. The fat suppression was accomplished byspectral fat saturation using a preparation saturation pulsetuned to the resonant frequency of fat47. Fat suppression elim-inates chemical-shift artifacts that can be superimposed to thecartilage bone interface thereby improving image contrast atthis interface and better delineation of the cartilage bound-aries47. The customized ‘‘rabbit-bed’’ optimized the reproduc-ibility of the experimental setup, minimized the experimenttime and optimized the efficiency of the coils.

Cartilage thickness measurements depend on the accu-racy of the computational algorithm and on the quality ofthe image itself48. The image quality was always sufficientto perform the measurements even in the presence ofsevere morphological OA changes in the operated group.Subjective evaluation of the technique showed that imagesand measurement qualities were better in the control groupat both examination points. Indeed, during the OA process,cartilages signal changes and it gets a lower contrast com-pared to surrounding structures.

The osteophytes’ production, the meniscal lesions, theVAE of the cartilage and the semi-quantitative histological

score were significantly higher in the operated rabbitsthan in the control group which confirmed that the ACLTsurgery induced OA in the operated group and validatedour experimental model. The MTC lesions were macroscop-ically evident in the operated rabbit group and ranged fromcartilage fibrillation to full-thickness erosion (27.5% of theoperated rabbits at 3 months and 42.5% at 5 months).

Histologically, MT.ThC was reported to increase afterACLT in cat51, dog30,52 and rabbit24,49,53 models of OA. Inthe ACLT rabbit model of OA, Yoshioka et al.32 found non-significant increase in histological MT.ThC at 4 and 8 weeksbut significant decrease at 12 weeks. Histologically, normalMT.ThC spatial variations have previously been reportedand were related to growth and mechanical factors32,49. Inaddition, histopathological grading scales and histomorph-ometry have generally been used without assessment ofsampling and spatial variations thus it is unclear how wellthey describe particular regions of the knee joint32.

In humans, techniques for displaying regional thicknesspatterns have been developed and Kshirsagar et al.48 sug-gested that analyzing subvolumes within the joint surfacereduce precision errors compared to analyses of the entirecartilage plate. In the present study, MR-based MT.ThC wascomputed in four locations and in 3D. The MT.ThC measuredin the minimum interbone distance area (Thick-1) was the thin-nestMT.ThC. In the control rabbits, Thick-1 did not change sig-nificantly over and had the smallest SD at both imaging points.This confirmed that technical errors on MR-based thicknessmeasurements were reduced with the 3D acquisition protocoland 3D data processing algorithm. Thick-1 was located in thepart of the central medial tibial plateau uncovered by themeniscus and was expected to correspond to the area ofthe tibial plateau affected early by OA changes32,49,53. Full-thickness erosion of the MTC (MT.ThC¼0 mm) was only ob-served at Thick-1 and in the operated group. Thick-1 in thecontrol group was significantly correlated with the macro-scopic changes, the VAE of the cartilage and the histologicalsemi-quantitative measurements. Therefore, Thick-1 re-flected the severity of the disease. In human, initial volumeof cartilage and rate of cartilage erosion were reported to bepositively correlated50. In the operated group, we did not dem-onstrate any correlation between the initial Thick-1 and theamount of changes (P> 0.05). However, we studiedcartilage thickness and not an average volume.

Thick-2 was significantly higher in the operated rabbits.Thick-2 was located at the anterior separation of the femoraland tibial cartilage and therefore was not affected by bio-mechanical changes. We assumed Thick-2 changes onlyreflected the global chemical changes of the joint. It isknown that the focal increase in cartilage thickness is oneof the earliest measurable changes in OA. Therefore, it ispossible that this increase might precede cartilage thinningand subchondral bone remodeling.

Thick-4 and Thick-3 varied widely between the operatedrabbits thus the significance of the observed differencesbetween both groups is unclear. Indeed, it is not clear whyconsiderable loss of cartilage occurred in some of the kneesexamined but not in others. Thick-3 and Thick-4 were mea-sured on the posterior medial tibial plate which correspondsto the normal weight bearing area and load changes second-ary to the ACLT might partly explain the results.

LIMITATIONS OF THE STUDY

At 5 months, due to the knee enlargement and ankylosis,the receiving coils had to be customized with a central holeproviding a better knee coverage and a higher filling of the

1384 C. Boulocher et al.: Non-invasive in vivo quantification of the MT.ThC progression in an OA rabbit model

coil thus a better SNR. Consequently, receiving coils weredifferent at 3 and 5 months but this did not preclude longitu-dinal MT.ThC comparisons. Neither the emitting coil nor theprotocol sequences were changed and similar LIP was ob-served in the control group at 3 and 5 months. The custom-ized receiving coil was made with similar characteristics tothe commercial coil except that it had a central hole. There-fore, technical improvement was achieved with this coil as itgave a better coverage and better filling of the coil and SNR.

Gross macroscopic and semi-quantitative results showedsignificant differences between the operated and the controlgroup and were correlated significantly with Thick-1. Thehistological MT.ThC was thinner in the operated group com-pared to the control group but no statistically significant dif-ference was observed. Histological quantification ofdegraded cartilage was difficult and could not be easilyquantified and only cartilage thickness in the remaining car-tilage areas was measured. Consequently, it overestimatedto the mean cartilage thickness and no full-thickness carti-lage erosion was detected histologically while observedboth macroscopically and with MRI. More work is neededto evaluate the accuracy of our MT.ThC measurementscompared to histology.

The poor correlation between MR-based and histologicalMT.ThC measurements highlights the inherent problems ofusing a unidimensional measure (histology) to indirectlyevaluate changes in 3D structures (the MTC) within a jointcompartment. Indeed, sectional 2D images cannot bealigned perpendicular to all parts of the joint surface but3D imaging enables the entire joint to be viewed in any di-rection and independently of the original sectionalorientation.

PERSPECTIVES

The natural history of OA progression involves more thanjust thickness defects. Fibrillation of the longitudinal colla-gen fibers that comprise the articular cartilage superficiallayer is one of the first signs of degeneration. MRI techniquesshould also account for the signal intensity aberrations of the tis-sue. Indeed, the signal intensity abnormalities visible on MRIappear to correlate with abnormal biochemical propertiesof OA cartilage. In our study, LIP was only used to evaluatethe quality of the MT.ThC measurements. CombiningMT.ThC accurate determination with quantitative signalinterpretation would be valuable for tracking early OAchanges.

Conclusion

The use of a non-invasive technique for detection andmeasurement of cartilage thickness is very appealing forpreclinical studies in experimental OA models. MR-basedMT.ThC measurements are a promising non-invasive andsensitive procedure for monitoring in vivo the therapeuticresponse to structure-modifying OA drugs that need to befurther developed.

In the present study, MR-based MT.ThC enabled toquantify the cartilage thinning longitudinally in the normaland in the ACLT rabbit model of OA. The MT.ThC measuredsemi-automatically in the minimum interbone distance area(Thick-1) was the thinnest of the measured MT.ThC. Thick-1was the most sensitive to change and the most discriminativebetween rabbit groups. We concluded that Thick-1reflected the severity of the disease and was the most effec-tive to measure the progression of the MTC destruction.

MR-based MT.ThC indicates that differences exist be-tween animals at the same time after surgery as observedhistologically by Vignon et al.53 This might indicate that inthe rabbit ACLT model of OA, individual variations existbetween animals. An index of OA severity based onMR-based MT.ThC might be more useful than time aftersurgery for the interpretation of biochemical data and tomonitor the effects of rehabilitation treatments.

Acknowledgements

The authors wish to thank R. Bolbos, M. Janier (Universitede Lyon, Plateforme Animage), D. Hartmann (Universite deLyon, UR RTI2B, Lyon F-69003, France), C. Farmer(ENVL) and Dr W. Mai (University of Pennsylvania) forthe linguistic help and M.L. Delignette (ENVL) for the statis-tical analyses. We also acknowledge the Institut ClaudeBourgelat (IClB, France) for the animal care.

References

1. Cooper C. Epidemiology of osteoarthritis. In: Klippel JH,Dieppe PA, editors. Rheumatology. 2nd ed. London:Mosby; 1998, p. 1e20.

2. Reginster JY. The prevalence and burden of osteo-arthritis. Rheumatology (Oxford) 2002;41:3e6.

3. Zhai G, Ding C, Cicuttini F, Jones G. Optimal samplingof MRI slices for the assessment of knee cartilagevolume for cross-sectional and longitudinal studies.BMC Musculoskelet Disord Feb 20, 2005;6:10.

4. Abadie E, Ethgen D, Avouac B, Bouvenot G, Branco J,Bruyere O, et al. Recommendations for the use of newmethods to assess the efficacy of disease-modifyingdrugs in the treatment of osteoarthritis. OsteoarthritisCartilage 2004;12(4):263e8.

5. Gandy SJ, Brett AD, Dieppe PA, Keen MC, Maciewicz RA,Taylor CJ, et al. Measurement of cartilage volumes in rheu-matoid arthritis using MRI. Br J Radiol 2005;78:39e45.

6. Peterfy CG. Scratching the surface. Articular cartilagedisorders in the knee. Magn Reson Imaging Clin NAm 2000;8(2):409e30.

7. Schwartz ER. Animal models: a means to study the patho-genesis of osteoarthritis. J Rheumatol 1987;14:101e3.

8. Batiste DL, Kirkley A, Laverty S, Thain LM, Spouge AR,Holdsworth DW. Ex vivo characterization of articularcartilage and bone lesions in a rabbit ACL transectionmodel of osteoarthritis using MRI and micro-CT.Osteoarthritis Cartilage 2004;12(12):986e96.

9. Baysal O, Baysal T, Alkan A, Altay Z, Yologlu S.Comparison of MRI graded cartilage and MRI basedvolume measurement in knee osteoarthritis. SwissMed Wkly 2004;134:283e8.

10. Eckstein F, Schnier M, Haubner M, Priebsch J, Glaser C,Englmeier KH, et al. Accuracy of cartilage volume andthickness measurements with magnetic resonance im-aging. Clin Orthop Relat Res 1998;352:137e48.

11. Eckstein F, Glaser C. Measuring cartilage morphologywith quantitative magnetic resonance imaging. SeminMusculoskelet Radiol 2004;8:329e53.

12. Eckstein F, Cicuttini F, Raynauld JP, Waterton JC,Peterfy C. Magnetic resonance imaging (MRI) ofarticular cartilage in knee osteoarthritis (OA): morpho-logical assessment. Osteoarthritis Cartilage 2006;14(1):46e75.

1385Osteoarthritis and Cartilage Vol. 15, No. 12

13. Graichen H, Von Eisenhart-Rothe R, Vogl T,Englmeier KH, Eckstein F. Quantitative assessmentof cartilage status in osteoarthritis by quantitativemagnetic resonance imaging: technical validation foruse in analysis of cartilage volume and further mor-phologic parameters. Arthritis Rheum 2004;50(3):811e6.

14. Link TM, Linder N, Haeussler M, Reimer P, Allkemper A,Jerosch J, et al. Artificially produced cartilage lesions insmall joints: detection with optimized MRI-sequences.Magn Reson Imaging 1997;15(8):949e56.

15. Paul PK, Wang JZ, Mezrich RS, Rakhit A, Dunton AW,Furst D. 3D-MRI: a novel approach to quantification ofarticular cartilage. Arthritis Rheum 1990;33:S91.

16. Raynaud JP, Kauffman C, Beaudoins G,Berthaumie MJ, de Guise JA, Bloch DA, et al. Reliabil-ity of quantification imaging system using magneticresonance images to measure cartilage thicknessand volume in human normal and osteoarthritic knees.Osteoarthritis Cartilage 2003;11:351e60.

17. Kladny B, Bail H, Swoboda B, Schiwy-Bochat H,Beyer WF, Weseloh G. Cartilage thickness measure-ment in magnetic resonance imaging. OsteoarthritisCartilage 1996;4:181e6.

18. Chan WP, Lang P, Stevens MP, Sack K, Majumdar S,Stoller DW, et al. Osteoarthritis of the knee: compari-son of radiography, CT, and MR imaging to assessextent and severity. AJR Am J Roentgenol 1991;157(4):799e806.

19. Blackburn WD Jr, Bernreuter WK, Rominger M,Loose LL. Arthroscopic evaluation of knee articularcartilage: a comparison with plain radiographs andmagnetic resonance imaging. J Rheumatol 1994;Apr21(4):675e9.

20. Disler DG, McCauley TR, Kelman CG, Fuchs MD,Ratner LM, Wirth CR, et al. Fat-suppressed three-dimensional spoiled gradient-echo MR imaging ofhyaline cartilage defects in the knee: comparisonwith standard MR imaging and arthroscopy. AJR AmJ Roentgenol 1996;Jul 167(1):127e32.

21. Tarhan S, Unlu Z. Magnetic resonance imaging andultrasonographic evaluation of the patients with kneeosteoarthritis: a comparative study. Clin Rheumatol2003;Sept 22(3):181e8.

22. Cohen ZA, McCarthy DM, Kwak SD, Legrand P,Fogarasi F, Ciaccio EJ, et al. Knee cartilage topogra-phy, thickness, and contact areas from MRI: in-vitrocalibration and in-vivo measurements. OsteoarthritisCartilage 1999;Jan 7(1):95e109.

23. Eckstein F, Westhoff J, Sittek H, Maag KP,Haubner M, Faber S, et al. In vivo reproducibility ofthree-dimensional cartilage volume and thicknessmeasurements with magnetic resonance imaging.AJR Am J Roentgenol 1998;170:593e7.

24. Batiste DL, Kirkley A, Laverty S, Thain LM, Spouge AR,Gati JS, et al. High-resolution MRI and micro-CT in anex vivo rabbit anterior cruciate ligament transectionmodel of osteoarthritis. Osteoarthritis Cartilage 2004;12(8):614e26.

25. Laurent D, Wasvary J, O’Byrne E, Rudin M. In vivoqualitative assessments of articular cartilage in therabbit knee with high-resolution MRI at 3 T. MagnReson Med 2003;50:541.

26. Wachsmuth, Juretschke HP, Raiss RX, Lindhorst E.In vivo contrast-enhanced MRI of experimental osteo-arthritis in the rabbit knee joint at 7.1 T. OsteoarthritisCartilage 2003;11(11A):891e902.

27. Calvo E, Palacios I, Delgado E, Ruiz-Cabello J,Hernandez P, Sanchez-Pernaute O, et al. High-resolution MRI detects cartilage swelling at the earlystages of experimental osteoarthritis. OsteoarthritisCartilage 2001;9(5):463e72.

28. Bolbos R, Benoıt-Cattin H, Langlois JB, Chomel A,Chereul E, Odet C, et al. Knee cartilage thicknessmeasurements using MRI: a 4-month longitudinalstudy in the meniscectomized guinea pig model ofOA. Osteoarthritis Cartilage 2007;15(6):656e65.

29. Gahunia HK, Babyn P, Lemaire C, Kessler MJ,Pritzker KP. Osteoarthritis staging: comparison be-tween magnetic resonance imaging, gross pathologyand histopathology in the rhesus macaque. Osteoar-thritis Cartilage 1995;3(3):169e80.

30. Libicher M, Ivancic M, Hoffmann V, WenzEarly W.Changes in experimental osteoarthritis using thePondeNuki dog model: technical procedure and initialresults of in vivo MR imaging. Eur Radiol 2005;15:390e4.

31. Loeuille D, Gonord P, Guingamp C, Gillet P, Blum A,Sauzade M, et al. In vitro magnetic resonance micro-imaging of experimental osteoarthritis in the rat kneejoint. J Rheumatol 1997;24(1):133e9.

32. Yoshioka M, Coutts RD, Amiel D, Hacker SA. Charac-terization of a model of osteoarthritis in the rabbitknee. Osteoarthritis Cartilage 1996;4:87e98.

33. Bauer JS, Krause SJ, Ross CJ, Krug R, Carballido-Gamio J, Ozhinsky E, et al. Volumetric cartilage mea-surements of porcine knee at 1.5-T and 3.0-T MRimaging: evaluation of precision and accuracy. Radiol-ogy 2006;241(2):399e406.

34. Eckstein F, Ateshian G, Burgkart R, Burstein D,Cicuttini F, Dardzinski B, et al. Proposal for a nomencla-ture for magnetic resonance imaging based measuresof articular cartilage in osteoarthritis. OsteoarthritisCartilage 2006;14:974e83.

35. Pond MJ, Nuki G. Experimentally-induced osteoarthritisin the dog. Ann Rheum Dis 1973;32:387e8.

36. McInerney T, Terzopoulos D. Deformable models inmedical image analysis: a survey. Med Image Anal1996;1(2):91e108.

37. Kass M, Witkin A, Terzopoulos D. Snakes: activecontour models. Int J Comput Vis 1987;1(4):321e31.

38. http://www.cartilage.org/_files/contentmanagement/ICRS_evaluation.pdf; 1987 (checked 05.07.2006).

39. Pritzker KPH, Gay S, Jimenez SA, Ostergaard K,Pelletier J-P, Revell PA, et al. Osteoarthritis cartilagehistopathology: grading and staging. OsteoarthritisCartilage 2006;14(1):13e29.

40. Pastoureau P, Leduc S, Chomel A, De Ceuninck F.Quantitative assessment of articular cartilage andsubchondral bone histology in the meniscectomizedguinea pig model of osteoarthritis. OsteoarthritisCartilage 2003;11(6):412e23.

41. Bacic G, Liu KJ, Goda F, Hoopes PJ, Rosen GM,Swartz HM. MRI contrast enhanced study of cartilageproteoglycan degradation in the rabbit knee. MagnReson Med 1997;37:764e8.

42. Dawson J, Gustard S, Beckmann N. High-resolutionthree-dimensional magnetic resonance imaging forthe investigation of knee joint damage during thetime course of antigen-induced arthritis in rabbits.Arthritis Rheum 1999;42:119e28.

43. Anastasios A, Hall LD. A novel RF coil configuration forin-vivo and ex-vivo imaging of arthritic rabbit kneejoints. Magn Reson Imaging 2003;21:61e8.

1386 C. Boulocher et al.: Non-invasive in vivo quantification of the MT.ThC progression in an OA rabbit model

44. Losch A, Eckstein F, Haubner M, Englmeier KH. A non-invasive technique for 3-dimensional assessment ofarticular cartilage thickness based on MRI. Part I:Development of a computational method. Magn ResonImaging 1997;15:795e804.

45. Glaser C, Faber S, Eckstein F, Fischer H, Springer V,Heudorfer L, et al. Optimization and validation ofa rapid high-resolution T1-w 3D FLASH water excita-tion MRI sequence for the quantitative assessmentof articular cartilage volume and thickness. MagnReson Imaging 2001;19(2):177e85.

46. Duryea J, Neumann G, Brem M, Koh W, Noorbakhsh F,Jackson R, et al. Novel fast semi-automated softwareto segment cartilage for knee MR acquisitions. Osteo-arthritis Cartilage 2007;15(5):487e92.

47. Peterfy CG, van Dijke CF, Janzen DL, Gluer CC,Namba R, Majumdar S, et al. Quantification ofarticular cartilage in the knee with pulsed satura-tion transfer subtraction and fat-suppressed MRimaging: optimization and validation. Radiology1994;192:485e91.

48. Kshirsagar AA, Watson PJ, Tyler JA, Hall LD. Measure-ment of localized cartilage volume and thickness of

human knee joints by computer analysis of three-dimensional magnetic resonance images. InvestRadiol 1998;33:289e99.

49. Chang DG, Iverson EP, Schinagl RM, Sonoda M,Amiel D, Coutts RD, et al. Quantitation and localizationof cartilage degeneration following the induction ofosteoarthritis in the rabbit knee. Osteoarthritis Carti-lage 1997;Sept 5(5):357e72.

50. Wluka AE, Stuckey S, Snaddon J, Cicuttini FM. Thedeterminant of change in tibial cartilage volume in os-teoarthritic knees. Arthritis Rheum 2002;46(8):2065e72.

51. Herzog W, Adams ME, Matyas JR, Brooks JG.Hindlimb loading, morphology and biochemistry ofarticular cartilage in the ACL-deficient cat knee.Osteoarthritis Cartilage 1993;1(4):243e51.

52. Adams ME, Brandt KD. Hypertrophic repair of caninearticular cartilage in osteoarthritis after anteriorcruciate ligament transection. J Rheumatol 1991;18(3):428e35.

53. Vignon E, Bejui J, Mathieu P. Histological cartilagechanges in a rabbit model of osteoarthritis. J Rheuma-tol 1987;14S:104e6.

1387Osteoarthritis and Cartilage Vol. 15, No. 12