Arthroscopic estimation of the extent of chondropathy

Article (PDF Available)inOsteoarthritis and Cartilage 15(5):506-15 · June 2007with23 Reads
DOI: 10.1016/j.joca.2006.10.013 · Source: PubMed
Abstract
Arthroscopy has been used to evaluate articular cartilage (AC) pathology in osteoarthritis (OA) for outcome measurement and validation of non-invasive imaging. However, many fundamental aspects of arthroscopic assessment remain un-validated. This study evaluated arthroscopic estimates of extent of chondropathy. Serial arthroscopic assessments were performed in a group of 15 sheep before and after bilateral stifle medial meniscectomy (MMx). Post-mortem assessments were performed in un-MMx sheep and 4 and 16 weeks post-MMx. Arthroscopic assessments of the extent of each grade of chondropathy were compared with a non-arthroscopic hybrid assessment that incorporated biomechanical, thickness and macroscopic assessments. Arthroscopy evaluated only 36% of AC and missed significant pathological changes, softening and chondro-osteophyte, occurring in peripheral regions. The patterns of change in arthroscopic assessments were similar to those of the non-arthroscopic assessment but there was a very strong tendency to over-estimate the extent of softened AC after MMx. In spite of these limitations arthroscopic assessments were responsive to change. Estimates of the extent of normal and softened AC were most responsive to change over time followed by estimates of superficial and deep fibrillation. Arthroscopy was as an excellent discriminator between normal and OA. Assessments of chondro-osteophyte and exposed bone were not responsive to change. Arthroscopic estimates of extent of chondropathy are prone to substantial error. While experience and training may reduce these errors other approaches may more effectively improve performance.
This article was originally published in a journal published by
Elsevier, and the attached copy is provided by Elsevier for the
author’s benefit and for the benefit of the author’s institution, for
non-commercial research and educational use including without
limitation use in instruction at your institution, sending it to specific
colleagues that you know, and providing a copy to your institution’s
administrator.
All other uses, reproduction and distribution, including without
limitation commercial reprints, selling or licensing copies or access,
or posting on open internet sites, your personal or institution’s
website or repository, are prohibited. For exceptions, permission
may be sought for such use through Elsevier’s permissions site at:
http://www.elsevier.com/locate/permissionusematerial
Author's personal copy
Arthroscopic estimation of the extent of chondropathy
S. P. Oakley M.B.B.S., F.R.A.C.P., Grad Dip Clin Epi, Ph.D.y*, I. Portek M.B.B.S., F.R.A.C.P.z,
Z. Szomor M.D., F.R.A.C.S. (Ortho)x, R. C. Appleyard B.E. Mech (Hons), Ph.D.k,
P. Ghosh D.Sc., Ph.D., B.Sc. (Hons), F.R.A.C.I., F.R.S.C.{,
B. W. Kirkham M.B., Ch.B., M.D., F.R.C.P., F.R.A.C.P.y,
G. A. C. Murrell M.B.B.S., D.Phil. (Oxon)x and
M. N. Lassere M.B.B.S. (Hons), Grad Dip Epi, F.R.A.C.P., F.A.F.P.H.M., Ph.D.z
y Rheumatology Department, Guys & St. Thomas’ NHS Foundation Trust, London, United Kingdom
z Department of Rheumatology, St. George Hospital (University of New South Wales), Gray St.
Kogarah, NSW 2217, Australia
x Orthopaedic Research Institute, St. George Hospital (University of New South Wales), Gray St.
Kogarah, NSW 2217, Australia
k Orthopedics and Traumatic Surgery, Level 4, B36, Royal North Shore Hospital,
St Leonards, NSW 2065, Australia
{ The Institute of Nutraceutical Research, PO Box 35, Brookvale, NSW 2100, Australia
Summary
Introduction: Arthroscopy has been used to evaluate articular cartilage (AC) pathology in osteoarthritis (OA) for outcome measurement and
validation of non-invasive imaging. However, many fundamental aspects of arthroscopic assessment remain un-validated.
Objectives: This study evaluated arthroscopic estimates of extent of chondropathy.
Methods: Serial arthroscopic assessments were performed in a group of 15 sheep before and after bilateral stifle medial meniscectomy
(MMx). Post-mortem assessments were performed in un-MMx sheep and 4 and 16 weeks post-MMx. Arthroscopic assessments of the extent
of each grade of chondropathy were compared with a non-arthroscopic hybrid assessment that incorporated biomechanical, thickness and
macroscopic assessments.
Results: Arthroscopy evaluated only 36% of AC and missed significant pathological changes, softening and chondro-osteophyte, occurring in
peripheral regions. The patterns of change in arthroscopic assessments were similar to those of the non-arthroscopic assessment but there
was a very strong tendency to over-estimate the extent of softened AC after MMx. In spite of these limitations arthroscopic assessments were
responsive to change. Estimates of the extent of normal and softened AC were most responsive to change over time followed by estimates of
superficial and deep fibrillation. Arthroscopy was as an excellent discriminator between normal and OA. Assessments of chondro-osteophyte
and exposed bone were not responsive to change.
Conclusions: Arthroscopic estimates of extent of chondropathy are prone to substantial error. While experience and training may reduce these
errors other approaches may more effectively improve performance.
ª 2006 Osteoarthritis Research Society International. Published by Elsevier Ltd. All rights reserved.
Key words: Arthroscopy, Histology, Biomechanical, Validity, Reliability, Responsiveness, Ovine, Meniscectomy.
Introduction
Arthroscopy has been used as a tool to diagnose and quan-
titate the severity of knee osteoarthritis (OA) and to validate
non-invasive assessments such as MRI
1e10
and is likely to
continue in this role. Quantitative arthroscopic assessments
of OA chondropathy that require the observer to estimate
the extent of each grade of chondropathy prior to calculation
of a composite score
11e13
have construct validity, reliability
and responsiveness but the fundamental elements of ar-
throscopic assessment have not been evaluated compre-
hensively. Arthroscopy is as good as histological and
biomechanical tests for detection of the chondropathy of
very early OA
14
but quantitative estimates of the severity
of chondropathy have only undergone limited evaluation
15
and preliminary work in an artificial knee model suggests
that arthroscopic estimates of extent of chondropathy can
have considerable error
16,17
. In vivo measurements have
added complexity as there may be multiple grades of
chondropathy with varying border definition, the boundaries
of articular surfaces can be difficult to view and not all com-
partments of the knee can always be assessed
11,13,14
.A
significant obstacle to validating arthroscopic assessments
*Address correspondence and reprint requests to: Dr. Stephen P.
Oakley, M.B.B.S., F.R.A.C.P., Grad Dip Clin Epi, Ph.D., Department
of Rheumatology, Level 3, Lambeth Wing, St Thomas Hospital,
Lambeth Palace Road, London SE1 7EH, United Kingdom. Tel:
44-20-7188-5901; Fax: 44-20-7188-5868; E-mail: stephen.oakley@
gstt.nhs.uk
Received 14 March 2006; revision accepted 23 October 2006.
OsteoArthritis and Cartilage (2007) 15, 506e515
ª 2006 Osteoarthritis Research Society International. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.joca.2006.10.013
International
Cartilage
Repair
Society
506
Author's personal copy
in vivo is the destructive nature of most non-arthroscopic
assessments. The ovine (sheep) medial meniscectomy
(MMx) model of OA is well suited for the study of arthros-
copy
18
because a spectrum of disease develops rapidly
throughout the joint after MMx and non-arthroscopic as-
sessments can be performed in OA and normal controls.
Objectives
The aim of this paper was to comprehensively evaluate
the performance of arthroscopic estimates of the extent of
chondropathy in the ovine MMx model of OA.
METHODS
This paper included data from two arthroscopic studies.
Study 1 evaluated feasibility of performing arthroscopic as-
sessments in an ovine MMx model. Study 2 was designed
to evaluate criterion validity by comparing arthroscopic as-
sessments to post-mortem non-arthroscopic assessments
in this model. Arthroscopic assessments from both the
studies were used to evaluate reliability, content validity,
construct validity and responsiveness, while criterion valid-
ity was specifically evaluated using data from Study 2. Part I
of the methods describes the collection of arthroscopic and
non-arthroscopic data in both studies and the development
of a hybrid non-arthroscopic assessments based on a com-
bination of biomechanical assessments of AC modulus,
histological and macroscopic assessments of gross struc-
tural damage. Part II describes the analytic methods used
to evaluate reliability, validity and responsiveness.
Part I: Collection of Data and Development
of Hybrid Non -arthroscopic Assessm ent
of Articular Cartilage
Induction of OA and Performance of
Arthroscopic Procedures
Approval was given by UNSW Animal Ethics Committee
to perform open MMx and repeated arthroscopic assess-
ment of both stifle joints in sheep in two studies. Study 1
evaluated the feasibility of performing serial arthroscopic
assessments in five second generation in-bred Border
LeicestereMerino whether sheep (aged 12e15 months)
before and then at 4 weekly intervals up to 12 weeks (three
sheeps) or 24 weeks (two sheeps) post-MMx. Study 2 eval-
uated validity of arthroscopic assessments. Eight second
generation in-bred DorseteMerino whether sheep (aged
12e15 months) had knee arthroscopy before MMx and
then at 4, 10 and 16 weeks after MMx. Two sheep were
sacrificed at 4 weeks and six at 16 weeks for non-arthro-
scopic assessment of AC. Arthroscopy was performed
using a one-chip Smith and Nephew (North Ryde, Aus-
tralia) 2.7 mm diameter 30
angle ‘‘small joint’’ arthroscope,
‘‘white-set’’ on white surgical gauze and inserted through an
incision infero-medial to the patella under general anaes-
thesia. AC was probed using a 3 mm nerve hook via a portal
infero-lateral to the patella. Arthroscopic assessments were
performed in the medial compartment (medial tibial plateau
e MTP, medial femoral condyle e MFC), the lateral
compartment (lateral tibial plateau e LTP, lateral femoral
condyle e LFC) and the patello-femoral compartment (pa-
tellar e PAT, femoral trochlear surfaces e TRO). A single
observer (SO) performed and recorded the arthroscopic
procedures to magnetic video tape. These were digitised
(Digital Video Creator, Pinnacle Systems, USA) as
MPEG files, edited into small clips of individual compart-
ments (medial, lateral and patello-femoral) and randomised
for blinded assessment. The presence/absence of medial
meniscus precluded blinding of arthroscopy to operative
status in the medial compartment.
Assessment of Arthroscopic Video Footage
Arthroscopic observers viewed video clips and estimated
the extent of each grade of chondropathy. AC was graded
as normal, softened, superficially fibrillated (>50% AC
thickness remaining), deeply fibrillated (<50% AC thickness
remaining) and exposed bone
19
. Observers 1 (ZS) and 2
(IP) viewed medial compartment arthroscopic recordings
from Study 1 estimating the extent of each grade as the
percentage of the articular surface. Arthroscopic video
recordings were selected to include examples of the full
disease spectrum and were evaluated twice within a month.
Observer 3 (SO) then evaluated all recordings from all com-
partments in both studies and repeated the assessments in
a randomly selected subset several months later. Assess-
ments by observer 3 were recorded as maps ‘‘drawn’’ on
two-dimensional computer bitmap images using Microsoft
Paint
20
. Maps were drawn on templates that were shaped
specific to each articular surface. The extent of chondro-
osteophyte cartilage (a raised, smooth, firm rim of cartilage
at the edge of the articular surface) and AC ‘‘not seen’’ were
also marked on these maps. Osiris computer software
21
was used to measure the areas of each grade of chondrop-
athy and AC ‘‘not seen’’. The extent of each grade of
chondropathy was calculated as a percentage of the AC
within the area ‘‘seen’’.
Non-operated Controls
In Study 1 (feasibility) arthroscopic assessments could
only be performed in a minority of lateral and patello-femo-
ral compartments. Blinding to MMx status was not possible
in the medial compartment so Study 2 (validation) was
designed to evaluate only post-MMx arthroscopic assess-
ments. However, assessments were more successful in
Study 2 due to experience and learning acquired in Study
1. Assessments could therefore be blinded to MMx status
in the lateral and patello-femoral compartments. Four
post-mortem stifle joints were obtained from two un-MMx
second generation in-bred DorseteMerino whether sheep
aged 12e15 months in an unrelated study. The assumption
was made that arthroscopic findings would have been the
same as those in other pre-MMx sheep.
Non-arthroscopic Assessments
These methods have been described in detail previ-
ously
18
and are summarised below. Sheep stifle joints
were examined blind to identity and MMx status after a pe-
riod of storage at 70
C. A grid of reference points was
marked with dye on each articular surface. At each refer-
ence point AC was given a macroscopic grade as for
arthroscopy. AC stiffness was quantified as dynamic shear
507
Osteoarthritis and Cartilage Vol. 15, No. 5
Author's personal copy
modulus (G*)
22
using a hand-held device
23
(reliability ICC
0.76)
18
and AC thickness measured using a micrometer-
needle penetration method
24
(reliability ICC 0.77)
18
. Histo-
logical sections were cut through the area of worst AC
damage for each articular surface, mounted on glass slides,
stained with toluidine blue O and counter-stained with fast
green
25
. Histological photomicrographs were taken of AC
at each grid reference point and graded blind for structural
damage graded using a five-point scale analogous to the
arthroscopic grade (reliability ICC 0.74)
18
. Grade 1 was
described as superficial zone abnormalities rather than soft-
ening and AC was graded as ‘‘chondro-osteophyte’’ if it was
thickened, highly cellular, contained large swirls of collagen
fibres, stained intensely with toluidine blue and had no
tidemark.
Development of ‘Hybrid Non-arthroscopic’ Grades
Non-arthroscopic assessment analogous to the arthro-
scopic assessments was used to grade the severity of
chondropathy at each grid reference point. Macroscopic
assessments defined AC as non-fibrillated, fibrillated or
chondro-osteophyte, G* discriminated between normal
and softened AC while superficial and deep fibrillations
were defined on the basis of AC thickness.
Discrimination Between Fibrillated and Non-fibrillated
AC. Macroscopic assessments were used to discriminate
between fibrillated and non-fibrillated AC and to define
chondro-osteophyte. The suitability of macroscopic assess-
ments for this role was determined by comparing
macroscopic and histological assessments in the subset
of 702 grid reference points where both assessments
were available. This was summarised in terms of sensitivity
and specificity.
Discrimination Between Normal and Softened AC. Dy-
namic shear modulus (G*) was used to discriminate be-
tween normal and softened AC. The cut-off point for G*
was determined by a ReceivereOperator Characteristic
curve. This method is used to determine the cut-off point
for a score on a continuous scale (G*) that results in the
best sensitivity and specificity for predicting the state of
the AC (normal or softened) as defined by some reference
standard. We used MMx status as the reference standard
(state variable) assuming that all pre-MMx AC was normal
while all post-MMx AC was softened due to OA. G* mea-
surements at individual reference points were included if
AC was not fibrillated.
Discrimination Between Superficially and Deeply Fibril-
lated AC. Existing arthroscopic grading systems
11,19
cate-
gorise AC as superficially fibrillated (<50% loss of AC
thickness) or deeply fibrillated (>50% loss of AC thickness).
This method is conceptually problematic for a number of
reasons. The observer must estimate how thick the AC
would have been prior to the development of fibrillation
and then estimate how much of the AC thickness
remains. This is made more challenging because AC thick-
ness varies within the joint and within articular surfaces and
thickens considerably (40%) after MMx
18
. We therefore
opted for the simple approach of defining superficial fibrillated
AC as being thicker and deeply fibrillated AC thinner than the
median thickness of fibrillated AC. Exposed bone was arbi-
trarily defined as having AC less than 0.2 mm AC thick.
Part II: Analysis of Data
Content Validity
We defined content validity as the proportion of AC pathol-
ogy ‘‘seen’’ arthroscopically. Grid reference points on the
post-mortem stifle joints were retrospectively defined as
falling into regions ‘‘seen’’ or ‘‘not seen’’ arthroscopically by
observer 3. Pathological changes were described separately
for the two regions. Arthroscopic assessments by observer 3
from both Study 1 and Study 2 were used for this analysis.
Construct Validity
Our construct was that the extent of chondropathy in-
creases over time after MMx while the extent of normal
AC decreases. This was evaluated graphically in the form
of bar charts indicating the mean extent of each grade of
chondropathy at each time point on each articular surface.
Arthroscopic assessments by observer 3 from both Study
1 and Study 2 were used for this analysis.
Criterion Validity
Arthroscopic assessments and hybrid non-arthroscopic
assessments from Study 2 were used to evaluate criterion
validity. Estimates of the extent of each grade of chondrop-
athy by observer 3 (SO) were compared with the extent of
each grade by hybrid non-arthroscopic assessment at
each articular surface. The four non-operated control joints
were allocated mean pre-MMx arthroscopic scores from
other joints. Accuracy Difference Scores (ADS) were calcu-
lated as the arthroscopic estimate minus the true extent by
hybrid non-arthroscopic assessment. Positive ADS values
indicated over-estimation and negative values under-esti-
mation. The ADS scores were presented graphically.
Reliability
Intra-observer reliability was calculated for each observer
using estimates of all grades of chondropathy. There were
six estimates of the extent of chondropathy per articular
surface, one for each grade of chondropathy. Then, for ob-
server 3 (SO), reliability was calculated separately for each
grade of chondropathy. Inter-observer reliability was deter-
mined for assessments in the medial compartment. AC
graded as chondro-osteophyte by observer 3 (SO) was con-
sidered ‘‘normal’’ in this analysis since observers 1 and 2
had not used this grade. Reliability was summarised using
Intra-class Correlation Coefficients (ICC 2,1)
26
and 95%
limits of agreement
27
(mean of the difference scores
1.96 standard deviations of the difference scores).
Responsiveness
Absolute Responsiveness. We estimated the Smallest
Detectable Difference (SDD) from repeated measurements
based on the 95% LOA (SDD
LOA
)
28
. We then determined
SDD empirically by the evaluating changes over time in our
508
S. P. Oakley et al.: Arthroscopic assessment of chondropathy
Author's personal copy
longitudinal studies by performing multiple t tests with Bon-
ferroni corrections for Multiple Comparisons
29
(SDD
BMC
).
Relative Responsiveness. The responsiveness of
different grades of chondropathy was compared using
relative measures of responsiveness. Relative SDD
[%SDD
LOA
] is a measure of responsiveness relative to the
magnitude of measurements and was calculated from
repeated measurements as (SDD
LOA
/mean arthroscopic
estimate of extent) 100%. Relative responsiveness was
also determined from longitudinal trends using ANOVA
and multiple comparisons. F statistics of one-way ANOVA
for change over time represent the ratio of between-group
variability (change over time) to within-group variability
and are a measure of relative responsiveness. Higher F sta-
tistics indicate better responsiveness. Relative responsive-
ness is the ratio of the F statistics (RR
F
¼ F
1
/F
2
). We also
summarised relative responsiveness as the number of sig-
nificant differences detected (out of a total possible 36) for
each grade using t tests with Bonferroni corrections for Mul-
tiple Comparisons (RR
BMC
). This approach identified the
time points at which significant changes had occurred.
Results
PART I: COLLECTION OF DATA AND DEVELOPMENT
OF HYBRID NON-ARTHROSCOPIC ASSESSMENT
Arthroscopic Procedures
Thirty-nine arthroscopic video recordings were available
from Study 1 and 52 from Study 2. The medial compartment
was assessed arthroscopically in 100% of procedures. Suc-
cess rates were lower in the patello-femoral (41%) and lat-
eral (15%) compartments in Study 1 but improved in Study
2 (90% and 80%). Observers 1 (ZS) and 2 (IP) assessed 46
medial compartment recordings on the first viewing and
repeated the assessments in a subset of 22. Observer 3
(SO) assessed 400 recordings on first viewing and 108 on
second viewing and assessed all three compartments. Ar-
throscopic and hybrid non-arthroscopic assessments were
compared at 120 articular surfaces.
Development of ‘Hybrid Non-arthroscopic Grades
Macroscopic assessments had a sensitivity of 0.76 and
specificity for the 0.88 detection of fibrillation compared to
histological assessments in the 632 grid points not graded
as chondro-osteophyte or exposed bone. ReceivereOpera-
tor Characteristic curve analysis (Fig. 1) showed that a cut-
off point of 1.375 MPa provided the best sensitivity (0.55)
and specificity (0.75) when dynamic shear modulus (G*)
was used to diagnose softening due to OA. The median
thickness of macroscopically fibrillated AC was 1.23 mm
(range 0.01e3.94 mm). Macroscopic assessments had
a sensitivity of 0.52 and specificity of 0.98 for the diagnosis
of chondro-osteophyte AC. The hybrid non-arthroscopic
assessment algorithm is described in Fig. 2.
PART II: ANALYSIS OF DATA
Content Validity
Most of the patellar articular surface was seen (85%)
compared to the TRO (58% ‘‘seen’’) and MFC (38%
‘‘seen’’). More MTP AC was seen after MMx (76%) than
before MMx (50%). The lateral articular surfaces were not
easily assessed (LTP 20% and LFC 13% ‘‘seen’’). Areas
of the articular surface ‘‘not seen’’ typically lay in the periph-
eral regions of articular surfaces or under meniscal fibro-
cartilage. Overall, 1956 (36%) grid reference points were
thought to have been ‘‘seen’’ arthroscopically while 3482
(64%) were ‘‘not seen’’. Extensive pathological changes
occurred in both regions (Fig. 3). Sixty-two percent of AC
that was ‘‘seen’’ was normal before MMx the remainder
being soft. This fell to 17% by 16 weeks after MMx being
replaced largely by softened, superficially and deeply fibril-
lated AC. Eighty-one percent of AC that was ‘‘not seen’’ was
normal before MMx (81%) the remainder being soft. Normal
AC fell to below 40% after MMx replaced by softened AC
(42%) and chondro-osteophyte (11%).
Construct Validity
The arthroscopic observations by observer 3 [Fig. 4] were
consistent with the hypothetical construct. OA chondropathy
was observed arthroscopically at all articular surfaces and
increased in severity and extent over time after MMx. The
patterns of change were similar to those of the hybrid non-
arthroscopic assessments. An exponential pattern (rapid
initial change followed by slower change) was seen at the
femoral trochlea, MFC and MTP while a linear (constant
rate) was seen at other locations. Significant reductions in
the extent of normal AC were observed at all articular
surfaces. The most extensive pathological change was soft-
ening (significant at patella, femoral trochlea, MTP and LTP).
Small areas of superficial fibrillation developed at the TRO,
MFC and MTP. Similar trends were seen for deep fibrillation
(significant at TRO). Exposed bone appeared at TRO and
MFC and chondro-osteophyte developed at MTP, MFC
and TRO but these trends were not significant.
Evaluation of Criterion Validity
The mean ADS show that the largest systematic errors
occurred with estimates of normal and softened AC. These
estimates were mirror-images of each other so we have
Fig. 1. ReceivereOperator characteristic curve for determining the
optimal cut-off of dynamic shear modulus (G*) for discriminating
normal AC from AC softened due to OA. At bottom-left a lower
AC stiffness cut-off (G* ¼ 0.50) yields high specificity (>0.95) but
low sensitivity (<0.10). At top-right a hig her AC stiffness cut-off
(G* ¼ 3.50) yields high sensitivi ty (>0.95) but low speci ficity
(<0.30). The optimal performance is the point on the curve closest
to the top-left corner (G* ¼ 1.375, sensitivity 0.55, specificity 0.75).
509
Osteoarthritis and Cartilage Vol. 15, No. 5
Author's personal copy
presented only estimates of softening. Before MMx arthros-
copy slightly over-estimated the extent of softening at MFC
(5%) and LFC (10%), slight under-estimated at the patella
(10%) and LTP (10%) and greater under-estimation at MTP
(50%). However, after MMx there was strong general ten-
dency to over-estimate the extent of softening at all locations
(patella 20e45%, femoral trochlea 40e50%, MTP 10e15%,
LFC 25e40%, LTP 25e40%). The one exception was the
MFC where there was a tendency to under-estimate the
extent of softening (30e40%).
Reliability
Overall intra-reliability of assessments was good for ob-
servers 1 (ICC 0.75) and 2 (ICC 0.74). Observer 3 had
only moderate reliability (ICC 0.59 medial compartment,
0.63 all articular compartments). The 95% LOA were
around 30 for observers 1 and 2. For observer 3 95%
LOA were similar (28, all compartments, 54 medial com-
partment). Agreement was better between observer 1 and
the other two observers (ICC 0.63, 0.70) but was only mod-
erate between observers 2 and 3 (ICC 0.43).
Responsiveness
Empirical estimates of absolute responsiveness (Table I)
showed that estimates of extent normal AC (SDD
BMC
38)
and softened AC (SDD
BMC
41) were less responsive than
those of superficial fibrillation (SDD
BMC
7) and deep
fibrillation (SDD
BMC
7). Estimates of exposed bone and
Fig. 2. Hybrid non-arthroscopic grade of the severity of chondropathy. At each grid reference point macroscopic assessments were used to
distinguish between fibrillated and non-fibrillated articular cartilage and chondro-osteophyte AC. Dynamic shear modulus (G*) was then used
to distinguish between normal and softened AC while superficial and deep fibrillation and exposed bone were defined by micrometer mea-
surements of AC thickness.
Fig. 3. Content vali dity: the mean extent of each grade of chondropathy within areas ‘‘seen’’ and ‘not seen’’ arthroscop ical ly are shown.
Extensive pathological changes occurred in both areas ‘‘seen’’ and ‘‘not seen’’ arthroscopically.
510
S. P. Oakley et al.: Arthroscopic assessment of chondropathy
Author's personal copy
Table I
Reliability and responsiveness of APAEs
Observer Articular surfaces Grade of chondropathy Mean [SD]* Ny Reliability Responsiveness
ICC 95% LOA SDD
LOA
%SDD
LOA
SDD
BMC
%SDD
BMC
F RE
F
RR
BMC
1 (ZS) MFC, MTP All grades 20 [24] 230 0.75 30
2 (IP) MFC, MTP All grades 20 [26] 230 0.74 32
3 (SO) MFC, MTP All grades 17 [28] 192 0.59 54
3 (SO) All Surfaces All grades 7 [18] 648 0.63 28
Intra-observer reliability e different grades
3 (SO) All surfaces Normal 45 [45] 106 0.59 84 84 187 38 84 43.69 1.00 10/36
Softened 43 [42] 106 0.56 82 82 191 41 95 24.05 0.55 8/36
Superficial fibrillation 7 [13] 106 0.38 26 26 371 8 114 7.93 0.18 7/36
Deep fibrillation 3 [8] 106 0.40 12 12 400 7 233 5.70 0.13 4/36
Exposed bone 1 [3] 106 0.57 3 3 300 NS 1.56 0.04 0/36
Chondro-osteophyte 2 [6] 106 0.09 23 23 1150 NS 2.16 0.05 0/36
Inter-observer e overallz
1 vs 2 MFC, MTP All grades 20 [23] 230 0.63 42
1 vs 3 MFC, MTP All grades 19 [25] 230 0.70 42
2 vs 3 MFC, MTP All grades 19 [23] 230 0.43 59
ICC, Intra-class Correlation Coefficient (1,2); 95% LOA, Bland Altman 95% Limits of Agreement; SDD
LOA
, Smallest Detectable Difference based on 95% LOA; %SDD
LOA
, relative SDD
(SDD
LOA
/mean APAE) 100; F, one-way ANOVA F statistics for changes over time; RE
F
, relative efficiency based of one-way ANOVA F statistics; SDD
BMC
, Smallest Detectable Difference
determined empirically from Bonferroni multiple comparisons; All grades, reliability was calculated using arthroscopic estimates of the extent of all five grades of chondropathy; Ref, reference
value.
*Mean and standard deviation of APAEs from the first set of assessments. For inter-observer the mean of all APAE by first and second observers.
yNumber of pairs for calculation of agreement.
zObservers 1 and 2 did not estimate APAEs for chondro-osteophyte. To facilitate comparison, chondro-osteophyte APAEs were counted as ‘‘normal’’.
511
Osteoarthritis and Cartilage Vol. 15, No. 5
Author's personal copy
chondro-osteophyte were non-responsive. Cross-sectional
estimates of responsiveness (SDD
LOA
) exhibited similar
trends but were conservative by comparison (SDD
LOA
normal 82, softened 84, superficial and deep fibrillation 26
and 12) and estimates of exposed bone and chondro-osteo-
phyte were more responsive (SDD
LOA
3, 23). The opposite
trend was seen for relative measures of responsiveness.
Longitudinally-derived relative estimates found
measurements of normal AC to be most responsive (F
43.69, RE
F
1.00, RR
BMC
10/36) followed by softened (F
24.05, RE
F
0.55, 8/36), superficially fibrillated (F 7.93,
RE
F
0.18, RR
BMC
7/36), deeply fibrillated AC (F 5.70, RE
F
0.13, RR
BMC
4/36). Estimates of chondro-osteophyte and
exposed bone were non-responsive. Cross-sectionally-
based estimates (%SDD
LOA
) also found that responsive-
ness was better for normal AC (187%) and softened AC
Fig. 4. (a) Arthroscopic estimates (top row) and hybrid non-arthroscopic assessments (middle row) of the extent of chondropathy and the diff-
erences between them (accuracy difference scores, bottom row) at the patella, femoral trochlea and medial femoral condyle. Significance of
changes over time for arthroscopic estimates (top row) is indicated as one-way ANOVA P values and Bonferroni comparisons with pre-MMx
assessments [
#
P < 0.05,
##
P < 0.01 and
###
P < 0.001 and NS not significant]. Error bars have not been included for ‘‘true extent of chondr-
opathy’’ and ‘‘accuracy of arthroscopy’’ because numbers were too small to assess statistical significance. (b) Arthroscopic estimates (top row)
and hybrid non-arthroscopic assessments (middle row) of the extent of chondropathy and the differences between them (accuracy difference
scores, bottom row) at the medial tibial plateau, lateral femoral condyle and lateral tibial plateau. Significance of changes over time for arthro-
scopi c estimates (top row) is indicated as one-way ANOVA p values and Bo nferroni comparisons with pre-MMx assessments [
#
P < 0.05,
##
P < 0.01 and
###
P < 0.001 and NS not significant]. Error bars have not been included for ‘‘true extent of chondropathy’’ and ‘‘accuracy of
arthroscopy’’ because numbers were too small to assess statistical significance.
512
S. P. Oakley et al.: Arthroscopic assessment of chondropathy
Author's personal copy
(191%) than for superficially (371%), deeply fibrillated
AC (400%) or exposed bone (400%) and that estimates
of chondro-osteophyte were quite unresponsive to
change (1150%). BMC showed that the great majority of
significant changes occurred between pre- and post-MMx
time points.
Conclusions
While the study was conducted in an animal model it
included truly normal subjects, responsiveness was tested
empirically using longitudinal data and criterion validity
was evaluated by comparison with a sophisticated non-ar-
throscopic assessment of AC [Fig. 4] designed to objec-
tively assess the same AC characteristics as arthroscopy.
The study identified major strengths and weaknesses of
the performance of arthroscopic assessments of AC.
VALIDITY
This study suggested that the content validity of chon-
droscopy was far from complete. It was thought that, on
average, only 36% of the articular surface was ‘‘seen’’.
While arthroscopy is highly operator dependent and
content validity may be better in the hands of more expe-
rienced operators it is likely that significant pathological
changes are missed during arthroscopy. This is supported
by our finding that the success of arthroscopic assess-
ments of the lateral compartment increased markedly
from Study 1 to Study 2 as the experience of the arthro-
scopist increased.
Fig. 4. (continued).
513
Osteoarthritis and Cartilage Vol. 15, No. 5
Author's personal copy
Arthroscopic assessments agreed with the hypothetical
construct of change over time and the temporal and spa-
tial patterns of change closely resembled those of the hy-
brid non-arthroscopic assessment. There was a striking
tendency of arthroscopy to over-estimate the extent of
AC softening (criterion validity) compared to the hybrid
non-arthroscopic assessments after MMx. This apparent
error may be explained by the superior sensitivity of
arthroscopy for the detection of early OA compared to
the biomechanical assessments
14
. The exception to this
tendency was at the MTP where arthroscopy under-
estimated softening before MMx. This probably occurred
because normal AC was biomechanically softer at this lo-
cation before MMx.
The Hayes equation
22
used to calculate G* incorporates
a measurement of thickness assuming that cartilage
behaves as an isotropic, linear elastic material during
dynamic indentation. It is widely accepted as a measure
of intrinsic AC modulus which is independent of thickness.
However, we have found a strong negative correlation
(rho 0.47) between AC thickness and G* calculated using
the Hayes equation indicating that this method does not
provide a true indication of intrinsic AC modulus
18
. This
highlights the limitations of the Hayes equation and explains
why portions of AC were defined as ‘‘softened’’ in pre-MMx
subjects when the AC was normal.
Conversely, a substantial portion of AC graded arthro-
scopically as softened in post-MMx joints was found to
have ‘‘normal’’ G*. False negative results will occur when
any continuous assessment is converted to a dichotomous
one. However, it is possible that arthroscopists assessing
AC stiffness also base their judgement on changes in AC
lustre and colour that occur after MMx due to AC thicken-
ing, collagen disorganisation
18
and changes in molecular
structure and content. While arthroscopic assessments
cannot quantify the severity of softening they are more
sensitive for the detection of the earliest stages of
chondropathy.
RELIABILITY AND RESPONSIVENESS
Intra- and inter-observer reliability by ICC was good,
but not as good as that of previous studies
12,30,31
. These
other studies used composite arthroscopic scores which
would have the effect of raising the ICC and were consid-
erably smaller (N ¼ 5e10). Intra-observer reliability (ICC)
was better for estimates of normal and softened AC.
Our results clearly illustrate the advantages of estimating
responsiveness empirically from longitudinal studies.
SDD
LOA
give more conservative estimates than the
empirically-based SDD
BMC
. BMC were the single most
useful test providing empirical estimates of both absolute
and relative responsivenesses, identifying the time points
at which the most significant changes occurred and con-
firming previous findings that arthroscopic is an excellent
discriminator between normal and OA
14
. While the abso-
lute methods suggested that estimates of the extent of
normal and softened AC were less responsive, the rela-
tive methods showed that estimates of these grades
were more responsive. Relative measures of responsive-
ness permit comparison of different methods. %SDD
LOA
is a conservative estimate of relative responsiveness
that can be based on cross-sectional data and used to
compare different measurement tools. Our best results
(SDD
LOA
187%) indicated that three-fold changes in soft-
ening/normal AC could be detected. While this seems
quite poor, these results are comparable to other forms
of assessment such as radiographic Sharpe (%SDD
LOA
91%) and Larsen (%SDD
LOA
70%) scores in rheumatoid
arthritis
32
.
FUTURE DIRECTIONS
While the study was conducted in an animal model we be-
lieve the results would have been similar in humans. Arthros-
copy is invasive, operator dependent, has incomplete
content validity and may be subject to large measurement er-
ror. In spite of this, changes were detectable over a short
time period. Furthermore, it is still performed widely for re-
search
33
and in clinical practice, it has been used to
validate non-invasive imaging methods and is likely be
used in this way in the future. The major limitations to content
validity, criterion validity and responsiveness may be partly
addressed by experience and rigorous training but three-di-
mensional mapping is needed to eliminate distortion and so-
phisticated technology may be required to localise
assessments. Measurements of the extent of individual
grades of chondropathy have incomplete content validity
and composite scores should have superior content validity
and be more responsive. Important work has already been
done in the development of arthroscopic composite scoring
algorithms
11,13
. However, weighting coefficients may be ex-
pected to vary depending on the context of the assessment.
Thus further work will see refinements in composite scoring
methods. Our results pave the way for further advances in
arthroscopic assessment of AC for the purpose of validation
of non-invasive imaging and use as a measure of OA severity.
Acknowledgements
This Research was carried out at the Department of Rheu-
matology (University of New South Wales) and Orthopaedic
Research Institute at St George Hospital, Kogarah, NSW
2217 Australia and The Institute of Bone and Joint Re-
search, Royal North Shore Hospital, St Leonards, NSW,
NSW 2065, Australia. Arthroscopic and VCR equipment
was provided on loan by Smith and Nephew Surgical
(North Ryde, Australia). Dr Oakley received a three-year
scholarship from the Arthritis Foundation of Australia as Re-
search Fellow at St George Hospital Department of Rheu-
matology. The St. George Department of Rheumatology
provided funds for sheep and their medical costs. The
Kempsey Valley ARKS have made regular financial contri-
butions to this study. Dr Fredrick Joshua from St George
Hospital Dept. of Rheumatology randomised arthroscopic
video clips and post-mortem stifle joint specimens.
References
1. Yulish BS, Montanez J, Goodfellow DB, Bryan PJ,
Mulopulos GP, Modic MT. Chondromalacia patellae:
assessment with MR imaging. Radiology 1987;164:
763e6.
2. Adams M, Li D, McConkey P, Davidson R, Day B,
Duncan C, et al. Evaluation of cartilage lesions by
magnetic resonance imaging at 0.15T: comparison
with anatomy and concordance with arthroscopy.
J Rheumatol 1991;18:1573e80.
3. Nakanishi K, Inoue M, Harada K, Ikezoe J, Murakami T,
Nakamura H, et al. Subluxation of the patella: evalua-
tion of patellar articular cartilage with MR imaging. Br J
Radiol 1992;65:662e7.
514
S. P. Oakley et al.: Arthroscopic assessment of chondropathy
Author's personal copy
4. Blackburn WDJ, Bernreuter WK, Rominger M, Loose LL.
Arthroscopic evaluation of knee articular cartilage:
a comparison with plain radiographs and magnetic
resonance imaging. J Rheumatol 1994;21:675e9.
5. Broderick LS, Turner DA, Renfrew DL, Schnitzer TJ,
Huff JP, Harris C. Severity of articular cartilage abnor-
mality in patients with osteoarthritis: evaluation with
fast spin-echo MR vs arthroscopy. AJR Am J Roent-
genol 1994;162:99e103.
6. Recht MP, Piraino DW, Paletta GA, Schils JP,
Belhobek GH. Accuracy of fat-suppressed three-di-
mensional spoiled gradient-echo FLASH MR imaging
in the detection of patellofemoral articular cartilage
abnormalities. Radiology 1996;198:209e12.
7. Drape JL, Pessis E, Auleley GR, Chevrot A,
Dougados M, Ayral X. Quantitative MR imaging evalu-
ation of chondropathy in osteoarthritic knees. Radiol-
ogy 1998;208:49e55.
8. Kawahara Y, Uetani M, Nakahara N, Doiguchi Y,
Nishiguchi M, Futagawa S, et al. Fast spin-echo MR
of the articular cartilage in the osteoarthrotic knee.
Correlation of MR and arthroscopic findings. Acta Ra-
diol 1998;39:120e5.
9. Potter HG, Linklater JM, Allen AA, Hannafin JA,
Haas SB. Magnetic resonance imaging of articular
cartilage in the knee. An evaluation with use of fast-
spin-echo imaging. J Bone Joint Surg Am 1998;80:
1276e84.
10. Bredella MA, Tirman PF, Peterfy CG, Zarlingo M,
Feller JF, Bost FW, et al. Accuracy of T2-weighted
fast spin-echo MR imaging with fat saturation in de-
tecting cartilage defects in the knee: comparison with
arthroscopy in 130 patients. AJR Am J Roentgenol
1999;172:1073e80.
11. Dougados M, Ayral X, Listrat V, Gueguen A,
Bahuaud J, Beaufils P, et al. The SFA system for
assessing articular cartilage lesions at arthroscopy of
the knee. Arthroscopy 1994;10:69e77.
12. Ayral X, Dougados M, Listrat V, Bonvarlet J,
Simonnet I, Amor B. Arthroscopic evaluation of
chondropathy in osteoarthritis of the knee. J Rheuma-
tol 1996;23:698e706.
13. Ayral X, Gueguen A, Listrat V, Bahuaud J, Beaufils P,
Beguin J, et al. Simplified arthroscopy scoring system
for chondropathy of the knee (revised SFA score). Re-
vue du Rhumatisme (English Edition) 1994;61:88e90.
14. Oakley SP, Portek I, Szomor Z, Appleyard RC,
Ghosh P, Kirkham BW, et al. Arthroscopy e a potential
‘‘gold standard’’ for the diagnosis of the chondropathy
of early osteoarthritis. Osteoarthritis Cartilage 2005;
13:368e78.
15. Oakley SP, Lassere MN. A critical appraisal of quan-
titative arthroscopy as an outcome measure in osteo-
arthritis of the knee. Semin Arthritis Rheum 2003;33:
83e105.
16. Oakley SP, Portek I, Szomor Z, Turnbull A, Murrell GA,
Kirkham BW, et al. Poor accuracy and interobserver
reliability of knee arthroscopy measurements are
improved by the use of variable angle elongated
probes. Ann Rheum Dis 2002;61:540e3.
17. Oakley SP, Portek I, Szomor Z, Turnbull A,
Murrell GA, Kirkham BW, et al. Accuracy and reli-
ability of arthroscopic estimates of cartilage lesion
size in a plastic knee simulation model. Arthroscopy
2003;19:282e9.
18. Oakley SP, Lassere MN, Portek I, Szomor Z, Ghosh P,
Kirkham BW, et al. Biomechanical, histologic and
macroscopic assessment of articular cartilage in
a sheep model of osteoarthritis. Osteoarthritis Carti-
lage 2004;12:667e79.
19. Beguin J, Locker B. Chondropathie rotullienne. 2’eme
journee d’arthroscopie du genou 1983;1:89e90.
20. Microsoft Corp. Microsoft Paint 1981 (Windows Millen-
ium Edition Copyright Microsoft Corp.).
21. Osiris version 2.0, 2003, University Hospital of Geneva,
http://www.uni-mainz.de/Cardio/incis/Data/p3_5.htm.
22. Hayes WC, Keer LM, Herrmann G, Mockros LF. A
mathematical analysis for indentation tests of articular
cartilage. J Biomech 1972;5:541e51.
23. Appleyard RC, Swain MV, Khanna S, Murrell GA. The
accuracy and reliability of a novel handheld dynamic
indentation probe for analysing articular cartilage.
Phys Med Biol 2001;46:541e50.
24. Swann AC, Seedhom BB. Improved techniques for
measuring the indentation and thickness of articullar
cartilage. Proc Inst Mech Eng [H] 1989;203:143e50.
25. Getzy L, Malemud C, Goldberg V, Moskowitz R. Fac-
tors influencing the metachromatic staining in paraf-
fin-embedded sections of rabbit and human articular
cartilage: a comparison of the safranin O and toluidine
blue O techniques. J Histotechnol 1982;5:111e6.
26. Shrout P, Fleiss J. Intra-class correlations: uses in
assessing rater reliability. Psych Bull 1979;86:420e8.
27. Bland JM, Altman DG. Statistical methods for assess-
ing agreement between two methods of clinical
measurement. Lancet 1986;307e10.
28. Lassere M, Boers M, van der Heidje D, Boonen A,
Edmonds J, Saudan A, et al. Smallest detectable
difference in radiological progression. J Rheumatol
1999;26:731e9.
29. Pagano M, Gauvreau K. Multiple Comparisons Proce-
dures. In: Pagano M, Gauvreau K, Eds. Principles of
Biostatistics. 1st Edition. California, Duxbury: Belmont
1993:263e5.
30. Ayral X, Dougados M, Listrat V, Bonvarlet JP,
Simonnet J, Poiraudeau S, et al. Chondroscopy:
a new method for scoring chondropathy. Semin Arthri-
tis Rheum 1993;22:289e97.
31. Ayral X, Gueguen A, Ike RW, Bonvarlet JP, Frizziero L,
Kalunian K, et al. Inter-observer reliability of the arthro-
scopic quantification of chondropathy of the knee.
Osteoarthritis Cartilage 1998;6:160e6.
32. Lassere MN, Van Der Heijde D, Johnson K,
Bruynesteyn K, Molenaar E, Boonen A, et al. Robust-
ness and generalizability of smallest detectable differ-
ence in radiological progression. J Rheumatol 2001;
28:911e3.
33. Kane D, Veale DJ, FitzGerald O, Reece R. Survey of
arthroscopy performed by rheumatologists. Rheuma-
tology (Oxford) 2002;41:210e5.
515
Osteoarthritis and Cartilage Vol. 15, No. 5
  • Full-text · Article · · Cartilage
  • [Show abstract] [Hide abstract] ABSTRACT: Arthroscopic and particularly histopathological assessments have been used to evaluate alterations of knee cartilage in osteoarthritis (OA). The aim of this study was to examine the correlation between an arthroscopic method to grade the severity of chondropathies and the histological/histochemical grading system (HHGS) applied to the corresponding articular cartilage areas in knee OA. The articular cartilage surface was examined by chondroscopy using the Beguin and Locker severity criteria, analysing the lesions in 72 chondroscopic areas. Afterwards, samples were obtained by dividing the cartilage surface of the medial tibiofemoral compartment of three OA knee joints into equal squares and they were evaluated histologically using the HHGS. The correlation between both grading methods was assessed using the weighted Kappa coefficient (K(w)). The results obtained with both scores showed good agreement (K(w): mean+/-standard deviation, 0.619+/-0.071). While the average HHGS scores of the chondral samples showed a better agreement with arthroscopic grades 0, I and II, the arthroscopic evaluation has a tendency to overestimate chondral lesions for histological grades III and IV. The intra- and inter-observer reliability of the HHGS evaluation of chondral lesions was excellent (Intraclass Correlation Coefficient: 0.909 and 0.941, respectively). In this study, we found a good quantitative correlation between established arthroscopic severity and histopathological scoring systems, particularly in less advanced lesions. Our results suggest that the arthroscopic method is a valuable tool in clinical research to score chondropathies in the medial femorotibial compartment of the OA knee, although some limitations should not be overlooked.
    Article · Sep 2008
  • [Show abstract] [Hide abstract] ABSTRACT: To present a new method of arthroscopic measurement of the surface and location of condylar lesions. We propose measuring the height of the condylar lesion by using the lesion's arc (Δ°) obtained from the difference between the angle of flexion at the beginning of the lesion and the angle of flexion at the end of the lesion. The first goal of the study was to determine the intra and inter reliability of the lesion's arc. Experiment 1: 20 deep lesions were evaluated using the lesional arc by two arthroscopists. Experiment 2: In a second series of 20 lesions, the flexion angles of the knees were recorded using a goniometer. All 10 knees (5 in each series) were then disarticulated and the true lesion arc was checked with a goniometer to assess the validity of the scopic measurements. The second goal was to obtain the height of the lesion from the lesion's arc. The lesion arc Δ° of the condylar is converted into height (millimeters) on the basis of a table obtained from 5 standard profiles of the lateral X-ray of the knee. Experiment 1: The intra observer reliability was good but the inter observer reliability was poor. Experiment 2: The intra and inter observers' reliability were good. On the anatomic control after disarticulating the knee, the confidence interval was narrower when using the goniometer. We propose a simple, reliable method to measure the height of a condylar lesion with the lesion's arc during arthroscopy.
    Full-text · Article · Jun 2011