Tibiofemoral centroid velocity correlates more consistently with Cartilage damage than does contact path length in two ovine models of stifle injury

Tibiofemoral Centroid Velocity Correlates More Consistently withCartilage Damage Than Does Contact Path Length in Two OvineModels of Stifle Injury

Jillian E. Beveridge,1 Bryan J. Heard,2 Nigel G. Shrive,3 Cyril B. Frank4

1Department of Mechanical and Manufacturing Engineering, McCaig Institute for Bone and Joint Health, University of Calgary, Calgary, Alberta,Canada, 2McCaig Institute for Bone and Joint Health, University of Calgary, Calgary, Alberta, Canada, 3Department of Civil Engineering, McCaigInstitute for Bone and Joint Health, University of Calgary, Calgary, Alberta, Canada, 4Department of Surgery, c/o McCaig Institute for Bone andJoint Health, University of Calgary, 3330 Hospital Drive, NW, Calgary, Alberta, Canada T2N 4N1

Received 17 October 2012; accepted 6 June 2013

Published online 6 July 2013 in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/jor.22429

ABSTRACT: Anterior cruciate ligament (ACL) rupture and/or meniscal injury are known risk factors for post-traumatic osteoarthritis.We tested the hypothesis that increasingly abnormal tibiofemoral centroid path lengths and velocities would correlate with the severityof cartilage damage in injured sheep. Six sheep underwent combined ACL/medial collateral ligament transection (ACL/MCLx), fivecomplete lateral meniscectomy (Mx), and four sham arthrotomy (Sham). Weighted centroids were used to estimate in vivo tibiofemoralcartilage contact path length during stance and the velocity of relative motion. Cartilage morphology was graded at dissection.Ligament transection significantly elongated plateau centroid path lengths and velocities, whereas condyle paths and velocities werereduced. Differences between plateau and femoral velocities (relative centroid velocity) were increased up to 10-fold over baselinevalues in the medial compartment. Reductions in Mx lateral compartment paths were significantly different from ACL/MCLx paths,but not relative to baseline or Sham values. Importantly, only centroid velocities consistently correlated with cartilage damage in eitherinjury model, suggesting that while path length is valuable in detecting changes in the envelope of joint motion, it may average outmeaningful peaks in the rate of relative motion that more closely relate to the mechanisms that might be damaging articular cartilagein these models. � 2013 Orthopaedic Research Society. Published by Wiley Periodicals, Inc. J Orthop Res 31:1745–1756, 2013

Keywords: ACL; meniscectomy; path length; velocity; osteoarthritis

Anterior cruciate ligament (ACL) rupture and/ormeniscal injury are known risk factors for post-traumatic osteoarthritis (PTOA).1 Because ligamentsand menisci function together to constrain knee mo-tion, and in the case of the menisci, to transmit anddistribute load across the tibiofemoral joint, injury tothese structures are thought to have a mechanical rolein the pathogenesis of PTOA. However, the relationbetween measures of mechanical changes in jointfunction, and specifically interactions between thearticulating surfaces, and cartilage damage etiology,have yet to be clearly defined.

Elongation of cartilage contact paths has been usedas a measure of instability in ACL-deficient humans2 anddogs.3 Although changes in contact path length havenot yet been reported following meniscectomy, contactlocation changes as a result of lateral meniscectomy insheep.4 It remains to be seen if other surface inter-actions may also be affected in this injury model.

Whereas abnormal contact path length may indi-cate changes to the normal envelope of joint motion,the rate of relative surface motion may also beimportant in the development of post-traumatic chon-dral lesions. Nickel et al.5,6 demonstrated that trac-

tional force (combined forces of kinetic friction andplowing force) correlated with increased relative sur-face velocity in vitro. Based on the in vivo rate ofinter-surface sliding following ACL transection,Anderst et al.7 could make a distinction between dogsthat developed minor chondral damage and those thatdeveloped major damage. However, a direct correlationbetween increased rate of relative tibiofemoral surfacemotion and the severity of cartilage damage on anindividual basis has not yet been reported, nor if onemechanical measure (contact path length or relativetibiofemoral velocity), relates more consistently tocartilage damage than the other. Therefore, we quanti-fied the magnitude of change in cartilage contact pathlength and the rate of relative tibiofemoral surfacemotion systematically, and determined whether arelationship exists between these surface interactionsand the severity of cartilage damage following injury.Two ovine injury models were used to test thehypothesis that increasingly abnormal in vivo tibiofe-moral path lengths and velocities would correlate withthe severity of cartilage damage following eithercombined ACL/medial collateral ligament (MCL) tran-section or complete lateral meniscectomy.

METHODSSubjects and Surgical ApproachFifteen skeletally mature female Suffolk-cross sheep (4 � 1years old; 81 � 7 kg) were obtained from Wolfitt Acres (Olds,AB) and were apportioned to one of three groups: combinedACL/MCL transection (ACL/MCLx, n ¼ 6), complete lateralmeniscectomy (Mx, n ¼ 5), or sham arthrotomy (Sham,n ¼ 4; five sheep were originally assigned to this group, butdata from one were omitted due to errors in data collection).

Grant sponsor: Canadian Institutes of Health Research;Grant sponsor: Alberta Innovates Health Solutions (AIHS);Grant sponsor: Canadian Arthritis Network; Grant sponsor: TheArthritis Society; Grant sponsor: Natural Sciences and Engineer-ing Research Council of Canada; Grant sponsor: AIHS Inter-disciplinary Team Grant in OA.Correspondence to: Cyril B. Frank (T: 403-220-6881; F: 403-283-7742; E-mail: [email protected])

# 2013 Orthopaedic Research Society. Published by Wiley Periodicals, Inc.

JOURNAL OF ORTHOPAEDIC RESEARCH NOVEMBER 2013 1745

Surgeries were performed on the right hind stifle while thecontralateral stifle was unaltered. Details of the surgicalprocedures were described previously,4,8 were reviewed andapproved by the University of Calgary Animal Care Commit-tee, and comply with the Canadian Council on Animal Careguidelines.

Kinematic Data Collection and Geometric Modeling ofTibiofemoral SurfacesIn vivo kinematics were measured for 100 non-consecutivestrides prior to surgical intervention (Intact), and thenserially at 2, 4, and 20 weeks post-operatively using rigidbone-mounted marker assemblies and a 4-camera video-based system (accuracy 0.4 � 0.4 mm, Motion Analysis,Santa Rosa, CA), as described previously.9 Global markerpositions were smoothed using a cubic spline with a low-passcut-off frequency of 6 Hz.10 Following the final 20-weekcollection session, animals were euthanized by intravenousinjection (Euthanyl, Bimeda-MTC, Cambridge, ON), and thehind limbs disarticulated and dissected. Tibiofemoral sub-chondral surface geometries were measured and anatomicalcoordinate systems created using a hand-held coordinatemeasuring machine (accuracy 0.025 mm, Faro Technologies,Lake Mary, FL). Stifle plateaus and condyles were modeledand smoothed by fitting a thin-plate spline11 to the measuredsurface coordinates, details of which are available in Bever-idge et al.12 Each of the 100 strides was normalized to 101points between successive hoofstrikes (HSs), and an averagestride calculated for each time point. Changes in kinematicswere expressed as the difference from Intact for each animal.

Weighted Centroid CalculationRelative tibiofemoral motion was quantified using weightedcentroids determined from in vivo inter-surface distance or“proximity” values.12 Tibiofemoral surface proximity wasmapped at every surface vertex for each point in the averagegait cycle and used to determine the coordinates of a centroidlocated within a region of close contact, as was donepreviously.13 To control the extent that small imprecisionsassociated with the video-based method in determiningweighted centroid location might inflate centroid pathlengths, centroid coordinates were smoothed using a cubicspline (details of sensitivity analyses for determining split fitparameters are provided in the online supplement).

Tibiofemoral Path Length and VelocityCentroid paths were quantified as the 3D Euclidian length ofthe centroid trace throughout the stance phase of theaverage gait cycle, and were assessed for each tibiofemoralsurface (i.e., medial and lateral femoral condyles and tibialplateaus). Changes in relative path length, or the differencebetween femoral and tibial paths, were additionally evaluat-ed to examine changes in inter-surface motion attributable tothe sliding of the condyle over the plateau versus the condylerotating relative to the tibia-meniscus complex.14 Positivevalues in relative paths indicate increased translation of thecondyle over the plateau (i.e., decreased femur path lengthand increased tibial path length), whereas negative valuesindicate increased rotation of the condyle.

The velocity of the weighted centroids relative to eachsubchondral surface (absolute centroid velocity) was alsodetermined to estimate the rate of inter-surface slidingthroughout gait, and was calculated using a first centraldifference approach15 by including centroid positions 1%

before and after the gait points of interest. These pointswere: HS, loading response (LR), mid-stance (MS), hoof-off(HO), initial swing (ISW), and terminal swing (TSW).Tangential velocity of tibiofemoral centroids was calculatedby using the vector normal to the surface at the centroidlocation to determine the plane lying tangential to thesurface. Centroid velocity vectors were then projected ontothis plane, and the magnitude of the velocity determined.Like the relative path length calculations, relative centroidvelocity was also determined as the difference between thetibial and femoral centroid velocities (femoral centroid veloci-ty magnitude subtracted from tibial centroid velocity magni-tude), where positive values indicate increased rate offemoral condyle translation over the tibial plateaus.

Path Length and Velocity PrecisionPrecision was assessed by perturbing the femoral surfaces by0.35˚ and 0.85 mm simultaneously in each degree of freedom,which corresponds to the average perturbation in jointkinematics when twice the root mean square error of thevideo-based motion analysis system9 is propagated throughthe calculations. This approach represents the unlikelyworst-case scenario that both tibia and femur recording erroroccurred in exactly opposite directions equal to the maximumerror recorded. Using this conservative estimate, the meandifference between measured and perturbed path lengths forall sheep was 1.8 � 1.5 mm. Centroid velocity precision was4.6 � 1.0 mm/s for stance and 6.6 � 3.2 mm/s for swing.

Hind Limb LoadingPeak vertical ground reaction force (pVGRF) was monitoredto quantify hind limb unloading. Sheep walked across anembedded force platform (Kistler Instrumente, Winterthur,CH) until a minimum of 20 right hind limb strides werecollected. Data were recorded at 1,200 Hz, filtered andsmoothed using a cubic spline,16,17 and normalized to bodyweight.

Cartilage GradingAt dissection, tibiofemoral stifle cartilage was graded at 10standardized locations (Supplementary Fig. 3) by a single,experienced observer whose grades were validated by com-paring gross scores with the corresponding histopathology(Supplementary Fig. 4). The observer was aware of theexperimental treatment that each animal received at thetime of grading, but was blinded to the kinematic outcomes.The scoring scheme used was modified from Drez et al.,18 toa 6-point scale ranging from 0—no damage to 5—large areasof exposed bone spanning >10% of the surface area (Supple-mentary Table 2). A maximum score of 10 was possible formedial and lateral femoral condyles (two regions: anteriorand posterior) and tibial scores were out of a maximum of 15(three regions: anterior, posterior, and central). The locationswere identified based on a combination of anatomical land-marks and subjective delineation of boundaries as describedpreviously.4 After ensuring that there were no significantdifferences in the left (un-operated) limbs between groups(Kruskal–Wallis test, p ¼ 0.5; data not shown), cartilagegrades were expressed as the difference from the contralater-al side. For each tibiofemoral surface, scores from all sub-regions were summed to yield a “composite score” thatreflects the overall magnitude of gross cartilage damage forthat surface. This approach of creating composite scores fromseveral tibiofemoral sub-regions increased the dynamic range

1746 BEVERIDGE ET AL.

JOURNAL OF ORTHOPAEDIC RESEARCH NOVEMBER 2013

and sensitivity of our measurement grading scheme to detectlesser focal changes that were expected 20 weeks post-surgically in these injury models.

Statistical AnalysesCartilage scores and surface interaction data were normallydistributed. Repeated measures ANOVAs with post hocBonferroni corrections were used to determine whetherchanges in centroid path lengths, velocities, and pVGRFswere significantly different over time and among groups.Correlations among path lengths, centroid velocities, andcomposite cartilage scores were assessed using Pearsoncorrelation coefficients. All analyses tests were consideredsignificant if p � 0.05.

RESULTSInitial exploration of the path length data revealedthat one ACL/MCLx sheep was almost always a mild(1.5� the inter-quartile range (IQR) beyond the groupmean) or extreme (3� the IQR beyond the mean)outlier (Supplementary Fig. 5). Parametric (as de-scribed above) and more robust non-parametric analy-ses (Kruskal–Wallis exact tests) confirmed that theextreme data points of the one ACL/MCLx subjectimpacted the results significantly (SupplementaryTables 4 and 5). Because the path length data were sodifferent (in magnitude and direction) for this oneACL/MCLx subject, met the criteria of a statisticaloutlier, and because they significantly impacted thepooled mean of the small sample size, they wereomitted from the path length analyses only.Changes Hind Limb LoadingpVGRF did not change significantly over time (Fig. 1),although both ACL/MCLx and Mx sheep tended tounload their hind right limbs immediately after injuryby �9.1% and �10.8% BW, respectively. Hind limbloading returned to baseline values by 20 weeks(difference of �0.9% and 1.8% BW, respectively). ThepVGRF of sham-operated sheep varied little over time.

Absolute Changes in Centroid Path Length and VelocitiesCombined ligament transection had the greatest effecton the medial compartment path lengths (Fig. 2c) andvelocities (Fig. 3a–c). Path elongation peaked at 2weeks post-transection (4.0 � 2.7 mm), the magnitudebeing equivalent to 88 � 57% of the Intact total pathlength (individual range: 5–145%). The greatest in-crease in medial plateau centroid velocity also oc-curred 2 weeks at HS (35 � 13 mm/s), and wasequivalent to a 145% increase over mean Intactvalues. Centroid velocities at LR and HO representeven greater increases, at 180% and 390% of meanIntact velocities, respectively. Conversely, medial fem-oral condyle velocity was significantly reduced duringweight acceptance (HS & LR in Fig. 3d–f), with peakdecreases at 4 weeks corresponding to a change ofabout �65% of Intact values during weight accep-tance.

Changes in lateral compartment paths and veloci-ties were not as severe as those of the medial compart-ment, with peak tibial plateau centroid pathelongation at 2 weeks corresponding to 81 � 77% ofIntact values (individual range: �12–165%; Fig. 2d).Increases in lateral tibial plateau centroid velocitiesduring weight acceptance were similarly significant,with peak 20-week values (22 � 18 mm/s) being equiv-alent to a 227% increase over Intact velocities (Supple-mentary Fig. 6a–c). Meanwhile, lateral condylecentroid velocities were significantly reduced at allpost-transection time points during weight acceptancecompared to Intact values (Supplementary Fig. 6d–f).

Conversely, lateral meniscectomy tended to reducecentroid paths, particularly those of the lateral com-partment (Fig. 2b and d). Even though Mx lateralplateau paths were significantly shorter than ACL/MCLx centroid paths at all post-operative time points(Fig. 2d), velocities of the lateral compartment cent-roids were not different from Intact, save for twoexceptions (Supplementary Fig. 6d and f).

Because reductions in femoral path lengths could bea function of reduced flexion range of motion (ROM)during stance, changes in flexion ROM over time andamong groups were assessed using a repeated meas-ures ANOVA. Although reduced flexion ROM acrossthe entire gait cycle was previously reported for theACL/MCLx model,8 small reductions in post-operativeflexion ROM during stance at 2 and 4 weeks in ACL/MCLx sheep (�1.6˚ � 2.2˚ and �0.3˚ � 1.4˚, respec-tively) were not significant among groups or over time(Fig. 4). Further, reduced femoral paths did notcorrelate with reductions in flexion ROM (p > 0.05).

Changes in Relative Tibiofemoral Path Length and VelocityPost-transection medial and lateral compartment rela-tive paths were significantly increased, indicatingthat ligament transection led to greater femoral con-dyle translation over the plateaus during stance(Fig. 5), which occurred at a greater rate duringweight acceptance relative to the uninjured state

0

20

40

60

80

100

Intact 2w 4w 20w

Pea

k V

GR

F (

% B

ody

Wei

ght)

Hind right peak vertical ground reaction force

ACL/MCLx (n=6) Mx (n=5) Sham (n=4)

Figure 1. Normalized right (surgical) hind limb peak verticalground reaction force (VGRF) over time. Peak VGRF was notdifferent from baseline values within and between groups. Errorbars are �1 SD of the mean.

TIBIOFEMORAL CENTROID VELOCITY CORRELATES MORE CONSISTENTLY 1747


(Fig. 6). Strikingly, changes in medial compartmentrelative velocity at HS represent increases of up to1,000% over Intact, while LR increases were as greatas 450% over Intact values (Fig. 6a–c).

Neither lateral compartment relative centroid pathsnor velocities were significantly altered following com-plete lateral meniscectomy, indicating that the overallenvelope of lateral compartment tibiofemoral motionwas reduced, but the nature of the rolling and slidingmotion of the surfaces was unaltered. As expected,tibiofemoral centroid path length and velocities werenot affected by sham arthrotomy.

Centroid Path Length, Velocity, and Cartilage DamageCorrelationsSummary statistics for cartilage composite scores areprovided in Supplementary Table 3. Despite abnormaland subject-specific changes in path length followingcombined ligament transection, cartilage scores werenot correlated with path lengths (Fig. 7), with theexception of lateral femoral condyle relative pathlength. However, increased plateau centroid velocitiesand relative velocities at HS and LR at the acute (2-

week) and sub-acute (4-week) time points in ACL/MCLx sheep were positively correlated with compositecartilage scores (Fig. 8 and Table 1). Medial femoralcondyle cartilage scores and medial compartment rela-tive velocities at LR also suggested a relationship, butcorrelation coefficients did not reach significance.Although pooled analyses generally did not reveal anysignificant effect of lateral meniscectomy on tibiofe-moral centroid velocities, correlative analyses indicatethat small increases in medial plateau centroid rela-tive velocity at HO 2 weeks post-meniscectomy weresimilarly correlated with cartilage damage (Fig. 9).

DISCUSSIONUsing two sheep models of joint injury, we intended toquantify the magnitude of changes in contact pathlength and the rate of relative tibiofemoral surfacemotion systematically and, importantly, to test thehypothesis that increasingly abnormal in vivo tibiofe-moral path lengths and velocities would correlate withthe severity of early cartilage damage. Indeed, somemeasures of tibiofemoral path length and velocitysurface interactions were significantly different from

2w 4w 20w-10

-5

0

5

Cha

nge

in P

ath

Leng

th (

mm

)a Change in medial femoral condyle

path length during stance

2w 4w 20w-10

-5

0

5

Cha

nge

in P

ath

Leng

th (

mm

)

b Change in lateral femoral condyle path length during stance

2w 4w 20w-10

-5

0

5

Cha

nge

in P

ath

Leng

th (

mm

)

c Change in medial tibial plateau path length during stance

2w 4w 20w-10

-5

0

5

Cha

nge

in P

ath

Leng

th (

mm

)

d Change in lateral tibial plateau path length during stance


** *

p=0.10

*p=0.08

p=0.10

** p=0.07p=0.09

†

†

p=0.08 p=0.12

† †

p=0.07

Figure 2. Change in weighted centroid path length for: (a) medial femoral condyle, (b) lateral femoral condyle, (c) medial tibialplateau, and (d) lateral tibial plateau. �Significant between group difference. † or p-value directly over (or under) the bar indicatessignificantly different from Intact, or strong trend. Error bars are �1 SD of the mean.



Intact in the ACL/MCLx model. Elongated centroidpaths demonstrate that combined ligament transectionenlarges the “envelope” of tibiofemoral motion, and isaligned with what has been suggested previously inACL-deficient humans.2 More specifically, abnormalrelative paths indicate that this change represents afundamental shift in how the surfaces move relative toeach other, where the convex femoral condyles trans-late over the tibial plateaus instead of rotating in theconcave meniscus-plateau complex post-transection.Further, analyses of the reduced femoral condyle pathlengths with respect to small reductions in joint ROM(maximum reduction of �1.6˚ � 2.2˚ at 2 weeks)indicated that the two measures were not correlated.These results suggest that femoral path length

changes in these sheep were not a function of reducedflexion–extension range of motion alone.

Based on our previous experience with the ACL/MCLx sheep model,19 we expected that not all individ-uals would respond similarly to the same injury—either functionally or morphologically—thereby creat-ing the potential for different measures to be correlat-ed. In line with our expectations, abnormal ACL/MCLx relative path length increases demonstratedsignificant inter-animal variation, with the magnitudeof individual change ranging from 5% to 165% overIntact values. However, abnormal path length meas-ures were generally not correlated with early cartilagedamage scores in this injury model. The sole lateralfemoral condyle correlation in ACL-deficient sheep is

HS LR MS HO ISW TSW-40

-20

0

20

40

60a 2w Change in medial plateau centroid velocity

Vel

ocity

(m

m/s

)


-20

0

20

40

60b 4w Change in medial plateau centroid velocity

Vel

ocity

(m

m/s

)


-20

0

20

40

60

Vel

ocity

(m

m/s

)

c 20w Change in medial plateau centroid velocity


**

** *

*

**

**

**

** *

*

*

† † †

† † †

† † †


-20

0

20

40

60d 2w Change in medial condyle centroid velocity

Vel

ocity

(m

m/s

)


-20

0

20

40

60e 4w Change in medial condyle centroid velocity

Vel

ocity

(m

m/s

)


-20

0

20

40

60

Vel

ocity

(m

m/s

)

f 20w Change in medial condyle centroid velocity

† †

† †

† †

†

†

Figure 3. Changes in medial tibial plateau centroid velocity (a–c) and medial femoral condyle centroid velocity (d–f). �Significantbetween group difference. †Directly over (or under) the bar indicates significantly different from Intact. Error bars are �1 SD of themean.



similar to the findings of Tashman et al.,3 whoreported a correlation between elongated medial femo-ral paths (but not relative paths) and cartilage damagein ACL-deficient dogs.

This is the first report on the effect of completelateral meniscectomy on in vivo weighted centroidpath lengths. Contrary to ACL/MCLx path changes,the equal reduction in Mx lateral compartment surfacepaths suggests that complete loss of the meniscusreduces inter-surface motion, but does not perturb therelative contributions of sliding and femoral rotation,as did ligament transection. Furthermore, the severityof cartilage damage was relatively consistent acrossthe five meniscectomized sheep, as was the abrupt“shift” in stifle kinematics of these animals that wecharacterized previously.4 Together, the abrupt reduc-tion in tibiofemoral path lengths and the consistentscores of the more severely damaged lateral compart-ment cartilage4 suggests that functional and morpho-logical abnormalities arising as a result of completelateral meniscectomy do not lie on a continuum.Rather, these responses may represent a step-changein joint function that contributes to a reproducibledegree of cartilage damage and/or that centroid pathlength does not represent the mechanism(s) drivingdamage in this model, which, based on in vitro experi-ments, is likely to include increased contact stress.20

Combined ligament transection profoundly and per-manently affected centroid velocities, with increasesas great as 10-fold over Intact. These increasescompare favorably with those described in ACL-defi-cient dogs,7 with some minor subtleties in velocitymagnitude that presumably reflect the differencebetween walking in sheep (0.9 m/s) and trotting(1.5 m/s) in dogs. Contrary to the path length changes,increased centroid velocities and relative velocitiesduring weight acceptance were positively correlatedwith cartilage damage scores at acute post-injury timepoints for all surfaces except for the medial femoralcondyle. These results support the hypothesis that arelationship exists between centroid velocity and carti-lage damage. However, lateral meniscectomy had littleto no effect on centroid velocity, although the minormedial tibial plateau increases in relative centroidvelocities was positively correlated with cartilagedamage at acute post-injury time points. The sharednature of the correlations between different injurymodels in meniscus-competent compartments suggeststhat similar mechanical mechanisms may be at work.Based on previous in vitro studies that examined theeffects of relative surface velocity and tractionalforce,5,6 damaging plowing friction may be occurringdue to the increased rate of contact patch translation.Speculated abnormal tractional forces may be evenfurther heightened by compromised joint lubrication,which occurs following ACL injury in both humans21

and other animal models.22–25 Together, these twomechanisms might contribute to the accelerated carti-lage damage we observed in these sheep.

0

5

10

15

20

Intact 2w 4w 20w

Deg

rees

(°)


Flexion range of motion during stance

Figure 4. Changes in flexion range of motion (ROM) over time.Flexion ROM during stance was not different from baselinevalues within and between groups. Error bars are �1 SD of themean.

-2

0

2

4

6

8

10

Cha

nge

in R

elat

ive

Pat

h Le

ngth

(m

m)

a Change in medial compartmentrelative path length during stance

-2

0

2

4

6

8

10

Cha

nge

in R

elat

ive

Pat

h Le

ngth

(m

m)

b Change in lateral compartmentrelative path length during stance

ACL/MCLx (n=5)Mx (n=5)Sham (n=4)

2w 4w 20w

2w 4w 20w

**

*p=0.08

*p=0.052

**

*p=0.06

†††

†††

Figure 5. Post-surgical changes in relative path lengths duringstance for medial (a) and lateral (b) tibiofemoral compartments.�Significant between group difference. †Directly over the barindicates significantly different from Intact. Positive valuesindicate increased tibial path length relative to femoral pathlength, and vice versa for negative values. Error bars are �1 SDof the mean.



The temporal symmetry of the significant correla-tions immediately after injury and known aberrationsin intra-articular lubricin concentrations21,22 are ofparticular interest: when correlations between centroidvelocities and cartilage damage were significant, theyrepresent post-injury time points when joint tribologymay be most aberrant. However, we must acknowledgethat it is unknown whether the chondral defects

observed at euthanasia follow directly from the abnor-mal centroid velocities identified at 2 and 4 weeks post-injury. The r2 values (Fig. 6) suggest that, although notsignificant, a relationship between centroid velocityand cartilage damage scores remains by 20 weeks post-transection (HS lateral plateau r2 ¼ 0.68; lateral con-dyle r2 ¼ 0.79, not shown). If chronic measures ofcartilage damage are in fact related to the acute


0

50

100

a 2w Change in medial compartmentrelative centroid velocity

Vel

ocity

(m

m/s

)


0

50

100

b 4w Change in medial compartmentrelative centroid velocity

Vel

ocity

(m

m/s

)


0

50

100

Vel

ocity

(m

m/s

)

c 20w Change in medial compartmentrelative centroid velocity


Femoral Translation

Femoral Rotation

Femoral Translation

Femoral Rotation

Femoral Translation

Femoral Rotation

**

**

**

**

**

**

† †

† †

† †


0

50

100

d 2w Change in lateral compartmentrelative centroid velocity

Vel

ocity

(m

m/s

)


0

50

100

e 4w Change in lateral compartmentrelative centroid velocity

Vel

ocity

(m

m/s

)


-50

0

50

100

Vel

ocity

(m

m/s

)

f 20w Change in lateral compartmentrelative centroid velocity

** *

*

** *

*

** *

*

† †

† †

† †

†

†

†

Figure 6. Change in medial (a–c) and lateral (d–f) compartment relative centroid velocity. Positive values indicate greater translationof the femoral condyle over the plateau, and negative values indicate greater rotation of the femur relative to the tibia. �Significantbetween group difference. †Directly over (or under) the bar indicates significantly different from Intact. Error bars are �1 SD of themean.



changes in relative tibiofemoral velocities, the acutepost-injury period may represent a critical window ofopportunity for clinical intervention.

Similarly, we quantified only mild (surface fibrilla-tion) to moderate (surface fibrillation with some carti-lage loss) gross morphological cartilage changes inthese sheep. We chose to evaluate gross morphology atonly 20 weeks in an effort to understand better the

relationship between abnormal joint motion and carti-lage damage early in the PTOA disease process.Although damage alone does not indicate that thesesheep had OA at the time of the assessment, it isconsidered a seminal event in PTOA pathogenesis.26

We have, nevertheless, demonstrated previously thatthese same ACL/MCLx subjects, and others, developedosteophytes that are consistent with OA,19,27 as did

-4-202468

0 1 2 3 4 5

Car

tilag

e S

core

Change in Path Length (mm)

d 4w LTP change in stance path length

-4-202468

-3 -2 -1 0 1 2

Car

tilag

e S

core


c 4w LFC change in stance path length

-4-202468

-4 -3 -2 -1 0 1

Car

tilag

e S

core


a 4w MFC change in stance path length

-4-202468

2 3 4 5 6 7

Car

tilag

e S

core


b 4w MTP change in stance path length

ACL/MCLx (n=5)

r2=0.53p=0.36

r2=-0.42p=0.48

r2=0.76p=0.14

r2=0.05p=0.94

Figure 7. Abnormal path length did not corre-late with composite cartilage scores in eitherligament-transected sheep (a–d), or laterallymeniscectomized sheep (not shown). MFC/MTP,medial femoral condyle and tibial plateau, re-spectively. LFC/LTP, lateral femoral condyleand tibial plateau, respectively.

0

1

2

3

4

5

6

0 20 40 60 80 100

Com

posi

te C

artil

age

Sco

re

Change in Relative Centroid Velocity (mm/s)

a 2w medial plateau at LR*

0

1

2

3

4

5

6

0 20 40 60 80 100

Com

posi

te C

artil

age

Sco

re


b 4w medial plateau at LR*

0

1

2

3

4

5

6

0 20 40 60 80 100

Com

posi

te C

artil

age

Sco

re


c 20w medial plateau at LR

0

1

2

3

4

5

6

0 20 40 60 80 100

Com

posi

te C

artil

age

Sco

re


d 2w medial condyle at LR

0

1

2

3

4

5

6

0 20 40 60 80 100

Com

posi

te C

artil

age

Sco

re


e 4w medial condyle at LR

0

1

2

3

4

5

6

0 20 40 60 80 100

Com

posi

te C

artil

age

Sco

re


f 20w medial condyle at LR

ACL/MCLx (n=6)

r2 = 0.84p = 0.04

r2 = 0.96p = 0.002

r2 = 0.81p = 0.054

r2 = 0.60p = 0.21

r2 = 0.75p = 0.09

r2 = 0.72p = 0.11

Figure 8. ACL/MCLx correlations between changes in relative centroid velocity and composite cartilage scores for medial tibial (a–c)and femoral (d–f) surfaces. Correlations indicate increased rate of translation of the femoral condyle over the adjacent plateau.�Significant correlation.



meniscectomized sheep. Further, the natural historyand progression to OA by 6 months post-meniscectomyhas been documented in this model.28 These observa-tions suggest that progression to OA may occur in theserelatively acute injury models; however, we cannot sayfor certain whether the early articular changes wouldhave progressed to full-thickness cartilage loss andend-stage OA. We can say only that a relationshiplikely exists between abnormal centroid velocity andearly cartilage damage 20 weeks post-injury.

Whereas changes in path length and relative surfacevelocities and their relationship to cartilage morpholo-gy were investigated, changes in contact location mayalso contribute to the pathogenesis of early cartilagedamage.29,30 We previously demonstrated reasonablespatial agreement between abnormal contact locationand the spatial distribution in the lateral meniscec-tomy model,4 and future investigations in the ACL/MCLx model may reveal similar relationships.

Additional limitations of this study include inter-subject variation in the pooled analyses, a small samplesize, the invasive nature of the bone-mounted markerassemblies, and the use of subchondral bone geometriesinstead of actual cartilage surfaces. We previouslyshowed that inter-subject variability is significant inthese sheep,31 and therefore attempted to mitigate someof the expected variation by assessing the change insurface interactions and cartilage morphology measuresas the individual difference from baseline. Whereas themagnitude of variation may have limited our ability todetect a significant group or time effect in the ANOVAanalyses, this individual variation gave rise to the moreimportant correlative relationship between the range ofmechanical responses to stifle injury and the subsequentindividual variation in morphological changes. However,the possibility exists that conclusions based on the smallsample size of four to six sheep per group may beT

able

1.

Summary

ofSignificantCorrelation

sforEach

TibiofemoralSurface

ofACL/M

CLxShee

p

ACL/M

CLx

MTP

LTP

MFC

LFC

2Wee

ks

4Wee

ks

20Wee

ks

2Wee

ks

4Wee

ks

20Weeks

2Wee

ks

4Wee

ks

20Wee

ks

2Wee

ks

4Wee

ks

20Wee

ks

Relativepath

length

——

——

——

——

—U

U—

Velocity

HS,LR

HS,LR

——

HS

——

——

HO

——

Relativevelocity

HS,LR

LR

——

HS

——

——

—HS

ISW

MTP/LTP,med

ialandlateralplateaus;

MFC/LFC,med

ialandlateralcondyles;

HS,hoofstrike;

LR,loadingresp

onse;HO,hoof-off;

ISW,initialsw

ing;U,significantcorrelation

ofstance

phase

path

length.

-1

0

1

2

3

4

5

6

-40 -20 0 20 40 60

Com

posi

te C

artil

age

Sco

re


2w MTP change in relative velocity at HO*

Mx (n=5)

r2=0.90p=0.03

Figure 9. Mx correlation between 2-week change in relativecentroid velocity and composite cartilage score for the medialtibial plateau (MTP) at hoof-off (HO). The significant correlationindicates increased rate of translation of the femoral condyleover the adjacent plateau. �Significant correlation.



different if the sample sizes are increased. These resultsshould therefore be interpreted cautiously until addi-tional studies with larger sample sizes are conducted.The bone-mounted marker assembly is, nonetheless, amore invasive method for quantifying in vivo bonykinematics than the more newly developed dual fluoros-copy and model-based techniques.32,33 However, giventhat Sham surface interactions were not different fromIntact values under the same conditions as thoseexperienced by injured sheep, any potential bias intro-duced by the fixation of the assemblies was likelyconsistent across groups and was smaller than thekinematic changes detected as a result of joint injury.Lastly, subchondral bone surfaces were used to approxi-mate tibiofemoral surfaces. While DeFrate et al. de-scribed perturbations in calculated centroid locationsdue to discrepancies in inter-surface distance based oncartilage versus subchondral bone geometry, humantibiofemoral cartilage thickness is twice that of sheep,34–36 and would therefore be expected to have a greaterinfluence on the calculated centroid location. Whereasfor our analyses, the range of cartilage thickness valuesacross the ovine tibial plateaus is equal to the measure-ment precision in calculating tibiofemoral proximity inthese animals,4 and would not be expected to influencecentroid location any more than would the methodsused to quantify in vivo dynamic motion.

Abnormal centroid velocities, but not cartilage con-tact path length, consistently correlated with the severi-ty of cartilage damage on an individual basis in twomodels of joint injury. Our results suggest that whilepath length is valuable in detecting changes in theenvelope of joint motion, it may be averaging outmeaningful peaks in the rate of relative motion, whichmight relate more closely to the mechanisms damagingthe articular cartilage. Further, the significant correla-tions between tibiofemoral velocities and cartilagescores were detected during weight acceptance at acutepost-injury time points, suggesting that the opportunityfor therapeutic intervention might be time-sensitive.

ACKNOWLEDGEMENTSWe also thank the following individuals for theirtechnical contributions: J. Ronsky, C. Sutherland, L.Jacques, S. Darcy, and T. Fung. No study sponsorswere involved in the study design, data collection, inthe analysis and interpretation of the data, or in thepreparation or decision to submit this manuscript forpublication that may be perceived as having biasedthe presentation of the data.

REFERENCES1. Lohmander LS, Englund M, Dahl LL, et al. 2007. The long-

term consequence of anterior cruciate ligament and menis-cus injuries. Am J Sports Med 35:1756–1769.

2. Li G, Moses JM, Papannagari R, et al. 2006. Anteriorcruciate ligament deficiency alters the in vivo motion of thetibiofemoral cartilage contact points in both the anteroposte-rior and mediolateral directions. J Bone Joint Surg Am88:1826–1834.

3. Tashman S, Anderst W. 2001. Predicting cartilage damagefrom knee kinematics in ACL-deficient dogs In: OrthopaedicResearch Societing Meeting. San Francisco: OrthopaedicResearch Society. p 0660.

4. Beveridge JE, Shrive NG, Frank CB. 2011. Meniscectomycauses significant in vivo kinematic changes and mechani-cally induced focal chondral lesions in a sheep model. JOrthop Res 29:1397–1405.

5. Nickel JC, Iwasaki LR, Beatty MW, et al. 2004. Laboratorystresses and tractional forces on the TMJ disc surface. JDent Res 83:650–654.

6. Nickel JC, Iwasaki LR, Beatty MW, et al. 2006. Static anddynamic loading effects on temporomandibular joint disctractional forces. J Dent Res 85:809–813.

7. Anderst WJ, Tashman S. 2009. The association betweenvelocity of the center of closest proximity on subchondralbones and osteoarthritis progression. J Orthop Res 27:71–77.

8. Tapper J, Fukushima S, Azuma H, et al. 2008. Dynamic invivo three-dimensional (3D) kinematics of the anteriorcruciate ligament/medial collateral ligament transectedovine stifle joint. J Orthop Res 26:660–672.

9. Tapper J, Ronsky J, Powers M, et al. 2004. In vivomeasurement of the dynamic 3-D kinematics of the ovinestifle joint. J Biomech Eng 126:301–305.

10. Woltring HJ. 1985. On optimal smoothing and derivativeestimation from noisy displacement data in biomechanics.Hum Mov Sci 4:229–245.

11. Boyd SK, Ronsky JL, Lichti DD, et al. 1999. Joint surfacemodeling with thin-plate splines. J Biomech Eng 121:525–532.

12. Beveridge JE, Shrive NG, Frank CB. 2013. Repeat-ability and precision of a weighted centroid method forestimating dynamic in vivo tibiofemoral surface inter-actions in sheep. Comput Methods Biomech Biomed Eng,in press.

13. Anderst WJ, Tashman S. 2003. A method to estimate in vivodynamic articular surface interaction. J Biomech 36:1291–1299.

14. Hoshino Y, Tashman S. 2011. Internal tibial rotation duringin vivo, dynamic activity induces greater sliding of tibio-femoral joint contact on the medial compartment. Knee SurgSports Traumatol Arthrosc 20:1268–1275.

15. Wood GA. 1982. Data smoothing and differentiation proce-dures in biomechanics. Exerc Sport Sci Rev 10:308–362.

16. Woltring HJ. 1986. Signal Processing Software [Internet].International Society of Biomechanics. http://isbweb.org/soft-ware/sigproc.html. (Accessed 2004).

17. Woltring HJ. 1995. Smoothing and differentiation techni-ques applied to 3D data. In: Allard P, Stokes IAF, Blanchi J-P, editors. Three-dimensional analysis of human movement.Champaign IL: Human Kinetics. p 79–99.

18. Drez DJ, DeLee J, Holden J, et al. 1991. Anterior cruciateligament reconstruction using bone-patellar tendon-boneallografts. A biological and biomechanical evaluation ingoats. Am J Sports Med 19:256–263.

19. Frank CB, Beveridge JE, Huebner KD, et al. 2012. CompleteACL/MCL deficiency induces variable degrees of instabilityin sheep with specific kinematic abnormalities correlatingwith degrees of early osteoarthritis. J Orthop Res 30:384–392.

20. Cottrell JM, Scholten P, Wanich T, et al. 2008. A newtechnique to measure the dynamic contact pressures on thetibial plateau. J Biomech 41:2324–2329.

21. Elsaid KA, Fleming BC, Oksendahl HL, et al. 2008.Decreased lubricin concentrations and markers of jointinflammation in the synovial fluid of patients with anteriorcruciate ligament injury. Arthritis Rheum 58:1707–1715.



22. Elsaid KA, Jay GD, Warman ML, et al. 2005. Associationof articular cartilage degradation and loss of boundary-lubricating ability of synovial fluid following injuryand inflammatory arthritis. Arthritis Rheum 52:1746–1755.

23. Elsaid KA, Machan JT, Waller K, et al. 2009. The impact ofanterior cruciate ligament injury on lubricin metabolism andthe effect of inhibiting tumor necrosis factor alpha onchondroprotection in an animal model. Arthritis Rheum60:2997–3006.

24. Goudoulas TB, Kastrinakis EG, Nychas SG, et al. 2010.Rheological study of synovial fluid obtained from dogs:healthy, pathological, and post-surgery, after spontaneousrupture of cranial cruciate ligament. Ann Biomed Eng38:57–65.

25. Teeple E, Elsaid KA, Fleming BC, et al. 2008. Coefficients offriction, lubricin, and cartilage damage in the anteriorcruciate ligament-deficient guinea pig knee. J Orthop Res26:231–237.

26. Mow VC, Hung CT. 2003. Mechanical properties of normaland osteoarthritic articular cartilage, and the mechanobiol-ogy of chondrocytes. In: Brandt KD, Doherty M, LohmanderLS, editors. Osteoarthritis, 2nd ed. New York: OxfordUniversity Press. p 106–197.

27. O’Brien EJ, Beveridge JE, Huebner KD, et al. 2013.Osteoarthritis develops in the operated joint of an ovinemodel following ACL reconstruction with immediate ana-tomic reattachment of the native ACL. J Orthop Res 31:35–43.

28. Little C, Smith S, Ghosh P, et al. 1997. Histomorphologicaland immunohistochemical evaluation of joint changes in amodel of osteoarthritis induced by lateral meniscectomy insheep. J Rheumatol 24:2199–2209.

29. Andriacchi TP, Mundermann A, Smith RL, et al. 2004. Aframework for the in vivo pathomechanics of osteoarthritisat the knee. Ann Biomed Eng 32:447–457.

30. Hosseini A, Van de Velde S, Gill TJ, et al. 2012. Tibiofemoralcartilage contact biomechanics in patients after reconstruc-tion of a ruptured anterior cruciate ligament. J Orthop Res30:1781–1788.

31. Tapper JE, Fukushima S, Azuma H, et al. 2006. Dynamic invivo kinematics of the intact ovine stifle joint. J Orthop Res24:782–792.

32. Li G, Van de Velde SK, Bingham JT. 2008. Validation of anon-invasive fluoroscopic imaging technique for the mea-surement of dynamic knee joint motion. J Biomech 41:1616–1622.

33. Anderst W, Zauel R, Bishop J, et al. 2009. Validation ofthree-dimensional model-based tibio-femoral tracking duringrunning. Med Eng Phys 31:10–16.

34. DeFrate LE, Sun H, Gill TJ, et al. 2004. In vivo tibiofemoralcontact analysis using 3D MRI-based knee models. J Bio-mech 37:1499–1504.

35. Appleyard RC, Burkhardt D, Ghosh P, et al. 2003. Topo-graphical analysis of the structural, biochemical and dynam-ic biomechanical properties of cartilage in an ovine model ofosteoarthritis. Osteoarthritis Cartilage 11:65–77.

36. Armstrong SJ, Read RA, Price R. 1995. Topographicalvariation within the articular cartilage and subchondralbone of the normal ovine knee joint: a histological approach.Osteoarthritis Cartilage 5:25–33.

SUPPORTING INFORMATIONAdditional supporting information may be found in theonline version of this article at the publisher’s web-site.

Figure S1. Comparison of Intact centroid displace-ments for one sheep when all, 50%, 33%, 25%, or 10%of the average gait cycle points were used to quantifythe centroid path for each tibiofemoral surface. MFC/LFC, medial/lateral femoral condyle centroids; MTP/LTP, medial/lateral tibial plateau centroids.

Figure S2. Data smoothing sensitivity analysis forthe sheep with the noisiest data. The performance ofthe cubic spline generated using a fit tolerance of “1”was reasonable, shown by the smoothed displacementdata compared to the spline generated using a toler-ance of 0.5, in (a) and (a.1). Further, the secondderivative of the smoothed centroid position data didnot deteriorate when smoothed with either a toleranceof “1” or “2” (b and b.1), but the latter tended to over-smooth the data (a.2 and b.2).

Figure S3. (a) Standardized locations of gross mor-phology grading. Anterior, posterior, medial, andlateral femoral condyle regions are indicated by FAM,FAL, FPM, and FPL. TAL, TCL, and TPL indicatethe anterior, central, and posterior regions of thelateral tibial plateau, and TAM, TCM, TPM indicatethe same regions of the medial plateau. (b) Theseregions were defined based on key landmarks (aversion of this figure was originally published inBeveridge et al.).

Figure S4. A grade score of “1” was assigned to theposterior aspect of the lateral femoral condyle (rectan-gle over image on the right). The histology micrograph(shown on the left) of the articular cartilage from thisregion confirms structural irregularities, mild cellularcloning, and depleted proteoglycan staining. Sectionstained with Safranin-O and fast green. Scale bar¼ 129 mm.

Figure S5. Bee swarm plots of the changes in pathlength data. Each circle indicates data for one subject;horizontal bars indicate group mean. Data from oneACL/MCLx sheep were often mild outliers (at 4 and 20weeks in (a) and at 4 weeks in (d)), and weresometimes extreme outliers (c). Therefore, data fromthis particular subject were omitted from the finalpath length analyses.

Figure S6. Changes in lateral tibial plateau centroidvelocity (a–c) and lateral femoral condyle centroidvelocity (d–f). �Significant between group difference.†Directly over (or under) the bar indicates significantlydifferent from Intact. Error bars are �1 SD of themean.

Table S1. Intact Path Length Was Sensitive to Boththe Number of Gait Points Used in Its Calculation,and Perturbations in Tibiofemoral Alignment Equiva-lent to Twice the Maximum Motion Analysis Error(0.35˚ and 0.85 mm). Values are Mean � SD (n ¼ 15).



MFC/LFC, Medial/Lateral Femoral Condyles; MTP/LTP, Medial/Lateral Tibial Plateaus

Table S2. Grading Scheme Used to Quantify GrossCartilage Damage

Table S3. Cartilage Composite Score Summary Statis-tics. There Was Significant Inter-Animal Variability inthe Severity of Cartilage Damage Observed 20 WeeksPost-Operatively, Particularly for Ligament-Trans-ected Sheep. MFC/LFC, Medial/Lateral Femoral Con-dyles; MTP/LTP, Medial/Lateral Tibial Plateaus

Table S4. Comparison of Non-Parametric Kruskal–Wallis Test p Values with (First Row), and Without(Second Row), Path Length Data from One ACL/MCLx

Subject. Inclusion of the Path Length Data from theOne ACL/MCLx Subject Significantly Impacted thePooled Mean. MFC/LFC, Medial/Lateral Femoral Con-dyles; MTP/LTP, Medial/Lateral Tibial Plateaus

Table S5. Changes in Pooled (Mean � SD) Tibiofe-moral Centroid Path Lengths for ACL/MCLx Sheepwith (n ¼ 6) or Without (n ¼ 5) Data from One ACL/MCLx Subject. Data from This Particular “Outlier”Were Almost Always a Mild (1.5 Times the Inter-Quartile Range (IQR) of the Group Mean), or ExtremeOutlier (Three Times the IQR Beyond the Mean).Changes in Path Length for the “Outlier” are Providedin the Third Row for Each Tibiofemoral Surface. MFC/LFC, Medial/Lateral Femoral Condyles; MTP/LTP,Medial/Lateral Tibial Plateaus



Documents

Tibiofemoral centroid velocity correlates more consistently with Cartilage damage than does contact path length in two ovine models of stifle injury