OsteoArthritis and Cartilage (2004) 12, 177–190
`© 2003 OsteoArthritis Research Society International. Published by Elsevier Ltd. All rights reserved.
`doi:10.1016/j.joca.2003.11.003
`
`International
`Cartilage
`Repair
`Society
`
`Whole-Organ Magnetic Resonance Imaging Score (WORMS) of the knee
`in osteoarthritis
`C. G. Peterfy M.D. Ph.D.†*, A. Guermazi M.D.†, S. Zaim M.D.†, P. F. J. Tirman M.D.‡, Y. Miaux
`M.D.†, D. White Ph.D.†, M. Kothari Ph.D.†, Y. Lu Ph.D.§, K. Fye M.D.§, S. Zhao Ph.D.§ and
`H. K. Genant M.D.§
`†Synarc, Inc., San Francisco, CA, USA
`‡National Orthopedic Imaging Associates, San Francisco, CA, USA
`§Osteoporosis & Arthritis Research Group, University of California, San Francisco, CA, USA
`
`Summary
`
`Objectives: To describe a semi-quantitative scoring method for multi-feature, whole-organ evaluation of the knee in osteoarthritis (OA) based
`on magnetic resonance imaging (MRI) findings. To determine the inter-observer agreement of this scoring method. To examine associations
`among the features included in the scoring method.
`
`Methods: Nineteen knees of 19 patients with knee OA were imaged with MRI using conventional pulse sequences and a clinical 1.5 T MRI
`system. Images were independently analyzed by two musculoskeletal radiologists using a whole-organ MRI scoring method (WORMS) that
`incorporated 14 features: articular cartilage integrity, subarticular bone marrow abnormality, subarticular cysts, subarticular bone attrition,
`marginal osteophytes, medial and lateral meniscal integrity, anterior and posterior cruciate ligament integrity, medial and lateral collateral
`ligament integrity, synovitis/effusion, intraarticular loose bodies, and periarticular cysts/bursitis. Intraclass correlation coefficients (ICC) were
`determined for each feature as a measure of inter-observer agreement. Associations among the scores for different features were expressed
`as Spearman Rho.
`
`Results: All knees showed structural abnormalities with MRI. Cartilage loss and osteophytes were the most prevalent features (98% and
`92%, respectively). One of the least common features was ligament abnormality (8%). Inter-observer agreement for WORMS scores was
`high (most ICC values were >0.80). The individual features showed strong inter-associations.
`
`Conclusion: The WORMS method described in this report provides multi-feature, whole-organ assessment of the knee in OA using
`conventional MR images, and shows high inter-observer agreement among trained readers. This method may be useful in epidemiological
`studies and clinical trials of OA.
`© 2003 OsteoArthritis Research Society International. Published by Elsevier Ltd. All rights reserved.
`Key words: Osteoarthritis, Cartilage, MRI, Imaging, Scoring.
`
`Introduction
`The structural determinants of pain and mechanical dys-
`function in osteoarthritis (OA) are not well understood, but
`are believed to involve multiple interactive pathways1–4.
`Accordingly, OA is best modeled as a disease of organ
`failure, in which injury to one joint component leads to
`damage of other components, and collectively to joint
`failure and the clinical manifestations of OA. The current
`practice of monitoring only a few features – typically radio-
`graphic joint-space narrowing and osteophytes – therefore
`provides only a keyhole view of this disease process, and is
`limited in content validity as an assessment of disease
`severity. A broader panel of imaging markers, i.e., a whole-
`organ evaluation,
`is needed to evaluate properly the
`structural integrity of joints affected by OA.
`
`*Address correspondence to: Charles Peterfy, M.D., Ph.D.,
`Chief Medical Officer, Executive Vice President, Synarc,
`Inc.,
`575 Market Street, 17th Floor, San Francisco, CA 94105, USA.
`Tel.:
`+1-415-817-8901;
`Fax:
`+1-415-817-8999;
`E-mail:
`charles.peterfy@synarc.com
`Received 21 April 2003; revision accepted 2 November 2003.
`
`Radiography, while offering high contrast and resolution
`for cortical and trabecular bone, cannot directly visualize
`non-ossified joint structures, such as articular cartilage,
`marrow tissue, menisci, cruciate and collateral ligaments,
`synovial fluid, and periarticular tendons and muscles, and
`therefore lacks the scope required for whole-organ assess-
`ment of joints5. Moreover, morphological distortion, geo-
`metric magnification and superimposition of overlying
`structures caused by the projectional viewing perspective
`of radiography complicate dimensional measurements, and
`can obscure important findings. Magnetic resonance imag-
`ing (MRI), on the other hand, is ideally suited for imaging
`arthritic joints. Not only is it free of ionizing radiation, but its
`tomographic viewing perspective obviates morphological
`distortion, magnification and superimposition. More impor-
`tantly, however, MRI is unparalleled in its ability to discrimi-
`nate articular
`tissues, such as cartilage, menisci and
`ligaments, and therefore holds the greatest potential as a
`tool for whole-organ imaging of the joint.
`In this article, we present a semiquantitative, multi-
`feature scoring method (WORMS)
`for whole-organ
`evaluation of the knee that is applicable to conventional
`MRI
`techniques that are widely available and easy to
`
`177
`
`-177-
`
`Smith & Nephew Ex. 1068
`IPR Petition - USP 9,295,482
`
`

`
`178
`
`C. G. Peterfy et al.: Whole-organ MRI of knee osteoarthritis
`
`implement at most imaging centers and hospitals around
`the world.
`
`Methods
`
`SUBJECTS
`
`Nineteen consecutive patients (15 men, 4 women; age
`61 years±8 years) with symptomatic OA of the knee were
`recruited from the rheumatology clinic of the University of
`California San Francisco Medical Center. All patients com-
`plained of pain in the study knee (ten right, nine left) for at
`least half the days of the preceding month. None of the
`patients had a history of previous knee trauma, knee
`surgery or arthroscopy. Conventional standing, extended-
`joint
`radiographs of
`the knees showed changes of
`Kellgren–Lawrence grade 2 (N=4) to 3 (N=15) and a mean
`minimum medial femorotibial joint-space width of 2.2 mm
`(range=0.5 mm–4.5 mm, S.D.=1.3 mm, measured manually
`with a graduated lens by an experienced clinical-trials
`radiologist). Fourteen patients were taking analgesics. Five
`were not receiving any medication. The study protocol was
`approved by institutional board review and informed con-
`sent was obtained from each subject prior to enrollment.
`
`MRI
`
`MRI of the study knee of each patient was acquired with
`a 1.5 Tesla whole-body scanner using a commercial cir-
`cumferential knee coil. Imaging sequences included, axial
`T1-weighted spin-echo (SE: 700/11 [TR msec/TE msec],
`20 cm field of view [FOV], 5 mm/1 mm [slice thickness/
`interslice gap], 256×192 matrix, frequency encoding [FE]
`anterior-posterior, one excitation), coronal T1-weighted SE
`(600/11, 16 cm FOV, 4 mm/0.5 mm, 256×192, FE superior-
`inferior, two excitations averaged), sagittal T1-weighted SE
`(600/11, 16 cm FOV, 4 mm/0.5 mm, 256×192, FE anterior-
`posterior, two excitations averaged), sagittal T2-weighted
`fast spin echo (FSE: 2500/90; echo train length (ETL) of
`eight; 14 cm FOV, 4 mm/0 mm, 256×192, FE superior-
`inferior,
`two excitation averaged) with fat suppression
`(frequency-selective presaturation), and sagittal
`fat-
`suppressed T1-weighted three dimensional (3D) spoiled
`gradient echo (FS-3DSPGR: 58/6, 40° flip angle, 14 cm
`FOV, 256×128 matrix, 60 contiguous 2-mm slices covering
`all articular cartilage plates in the knee, FE, superior-
`inferior, one excitation, frequency-selective fat saturation,
`superior-inferior saturation bands to minimize pulsation
`artifacts). The total time required for MRI, including patient
`set up, was 60 min.
`
`WHOLE-ORGAN MRI SCORING (WORMS)
`
`images were transferred to a Sun Workstation and
`All
`analyzed using MRVision software (MRVision, Inc, Menlo
`Park, CA). Images were scored with respect to 14 indepen-
`dent articular features: cartilage signal and morphology,
`subarticular bone marrow abnormality, subarticular cysts,
`subarticular bone attrition, marginal osteophytes, medial
`and lateral meniscal
`integrity, anterior and posterior cru-
`ciate ligament integrity, medial and lateral collateral
`liga-
`ment
`integrity, synovitis,
`loose bodies and periarticular
`cysts/bursae. Readings were performed independently by
`two musculoskeletal radiologists (CP, PT) following two
`separate two-hour
`training sessions using different
`case material than the 19 subjects included in this study.
`Readers used all images to evaluate each feature.
`
`Fig. 1. Regional subdivision of the articular surfaces. The patella
`(left image) is divided into medial (M) and lateral (L) regions, with
`the ridge considered part of the M region. The femur and tibia are
`also divided into M and L regions (right image), with the trochlear
`groove of the femur considered part of the M region. Region S
`represents the portion of the tibia beneath the tibial spines. The
`femoral and tibial surfaces are further subdivided into anterior (A),
`central (C) and posterior (P) regions (middle image). Region A of
`the femur corresponds to the patellofemoral articulation; region C
`the weight bearing surface, and region P the posterior convexity
`that articulates only in extreme flexion. Region C of the tibial
`surface corresponds to the uncovered portion between the anterior
`and posterior horns of the meniscus centrally and the portion
`covered by the body of the meniscus peripherally.
`
`the features examined (cartilage signal and
`Five of
`morphology, subarticular bone marrow abnormality, sub-
`articular cysts, subarticular bone attrition, marginal osteo-
`phytes) related to the articular surfaces. These features
`were evaluated in 15 different
`regions subdivided by
`anatomical landmarks in the fully extended knee (Fig. 1).
`Subdivisions were determined independently by each
`reader. The patella was divided into the lateral facet (LP)
`and medial facet (MP). The patellar ridge was considered
`part of
`the MP. The subchondral component of each
`patellar region extended the full thickness of the bone to the
`opposite cortex. The femoral articular surface was divided
`into medial (MF) and lateral (LF) condyles, with the troch-
`lear groove considered part of MF. The boundary between
`MF and LF was defined by a plane aligned with the lateral
`wall of the femoral notch (Fig. 2). MF and LF were each
`divided into three regions: (1) anterior (a): extending from
`the anterior-superior osteochondral junction to the anterior
`margin of the anterior horn of the meniscus; (2) central (c):
`extending from the anterior margin of the anterior horn of
`the meniscus to the posterior capsular attachment of the
`posterior horn of
`the meniscus; and (3) posterior (p):
`extending from the posterior capsular attachment of the
`posterior horn of the meniscus to the posterior-superior
`osteochondral
`junction. The subchondral component of
`each femoral region extended perpendicularly from the
`articular surface to the level of an imaginary line connecting
`the anterior and posterior osteochondral
`junctions. The
`medial tibial plateau (MT) and lateral tibial plateau (LT)
`were each divided into three equal regions: anterior (a),
`central (c) and posterior (p). Based on these subdivisions,
`the patellofemoral
`joint (PFJ) comprised regions MP, LP,
`MFa and LFa; the medial femorotibial
`joint (MFTJ) com-
`prised regions MFc, MFp, MTa, MTc and MTp; and the
`lateral
`femorotibial
`joint (LFTJ) comprised regions LFc,
`LFp, LTa, LTc and LTp. The non-articulating portion of the
`
`-178-
`
`

`
`Osteoarthritis and Cartilage Vol. 12, No. 3
`
`179
`
`Table I
`Maximum scores attainable with WORMS
`
`MFTJ LFTJ PFJ S Region
`
`Total
`
`3
`3
`
`30
`15
`15
`15
`35
`
`110
`6
`
`30
`15
`15
`15
`35
`
`110
`6
`
`24
`12
`12
`12
`28
`
`88
`
`Cartilage
`Marrow abnormality
`Bone cysts
`Bone attrition
`Osteophytes
`
`Compartment total
`Menisci
`Ligaments
`Synovitis
`
`Total
`
`84
`45
`45
`42
`98
`
`12
`3
`3
`
`332
`
`Fig. 4. Subarticular marrow abnormality score. This score is based
`on the extent of regional marrow involvement by areas of free
`water signal with ill-defined margins.
`
`the region;
`the region; 2=25% to 50% of
`1=<25% of
`3=>50% of the region. The maximum scores for MFTJ,
`LFTJ, PFJ, region S and the entire knee were 15, 15, 12, 3
`and 45 respectively (Table I).
`Subarticular cysts were identified as foci of markedly
`increased signal
`in the subarticular bone with sharply
`defined, rounded margins and no evidence of
`internal
`marrow tissue or trabecular bone on the fat-suppressed
`T2-weighted FSE images. Bone cysts were graded in each
`region, including the subspinous region of the tibia (S), from
`0 to 3 based on the extent of regional involvement, as for
`bone marrow abnormality (Fig. 5): 0=none; 1=<25% of the
`region; 2=25% to 50%of the region; 3=>50% of the region.
`
`Fig. 2. Defining anterior-lateral femoral region. The medial border
`of
`the lateral condyle on an axial
`image marks the boundary
`between regions FMa and FLa.
`
`tibial plateau beneath the tibial spines was designated
`region ‘S’ (subspinous). The subchondral component of
`each tibial region extended 2 cm beneath the articular
`surface.
`Cartilage signal and morphology was scored in each of
`the 14 articular-surface regions (excluding region S) using
`the fat-suppressed T2-weighted FSE images and the
`FS-3D SPGR images with an eight-point scale (Fig. 3):
`0=normal thickness and signal; 1=normal thickness but
`increased signal on T2-weighted images; 2.0=partial-
`thickness focal defect <1 cm in greatest width; 2.5=full-
`thickness focal defect <1 cm in greatest width; 3=multiple
`areas of partial-thickness (Grade 2.0) defects intermixed
`with areas of normal thickness, or a Grade 2.0 defect wider
`than 1 cm but <75% of the region; 4=diffuse (≥75% of the
`region) partial-thickness loss; 5=multiple areas of
`full-
`thickness loss (grade 2.5) or a grade 2.5 lesion wider than
`1 cm but <75% of the region; 6=diffuse (≥75% of the region)
`full-thickness loss. The maximum cartilage scores for
`MFTJ, LFTJ, PFJ and the entire knee were 30, 30, 24 and
`84, respectively (Table I).
`Subarticular bone marrow abnormality was defined as
`poorly marginated areas of increased signal
`intensity in
`the normally fatty epiphyseal marrow on fat-suppressed
`T2-weighted FSE images. This feature was graded in each
`of the 14 articular surface regions as well as the region of
`the tibia beneath the tibial spines (S) from 0 to 3 based
`on the extent of regional
`involvement (Fig. 4): 0=none;
`
`Fig. 3. Eight-point scale for scoring articular cartilage signal and morphology. Each region of the knee surface is scored independently.
`
`-179-
`
`

`
`180
`
`C. G. Peterfy et al.: Whole-organ MRI of knee osteoarthritis
`
`The maximum scores for MFTJ, LFTJ, PFJ, region S and
`the entire knee were 15, 15, 12, 3 and 45 respectively
`(Table I ).
`Flattening or depression of the articular surfaces was
`termed bone attrition and graded from 0 to 3 based on the
`subjective degree of deviation from the normal contour:
`0=normal; 1=mild; 2=moderate; 3=severe. For example,
`the osseous articular surfaces of the medial and lateral
`femoral condyles and medial facet of the patella are all
`normally slightly convex. Accordingly, flattening=Grade 1,
`slight concavity=Grade 2, and marked concavity=Grade 3.
`Fig. 6 illustrates this for the lateral tibial plateau (Fig. 6).
`The maximum scores for the MFTJ, LFTJ, PFJ and the
`entire knee were 15, 15, 12 and 42 respectively (Table I).
`Osteophytes along 14 different margins of the knee, the
`anterior (a), central weight bearing (c) and posterior (p)
`margins of the femoral condyles and tibial plateaus, and the
`medial (M) and lateral (L) margins of the patella, were
`graded from 0 to 7 using the following scale: 0=none;
`1=equivocal; 2=small; 3=small-moderate; 4=moderate;
`5=moderate-large; 6=large; 7=very large (Figs. 7 and 8).
`The maximum scores for the MFTJ, LFTJ, PFJ and the
`entire knee were 35, 35, 28, and 98 respectively (Table I).
`The anterior cruciate ligament (ACL) and posterior cru-
`ciate ligament (PCL) were independently scored as intact
`(0) or torn (1) using the sagittal T2 FSE images. The medial
`collateral
`ligament (MCL) and lateral collateral
`ligament
`(LCL) were independently scored as intact (0), or torn (1)
`
`Fig. 5. Subarticular cyst score. Subarticular cyst score is based on
`the extent of focal bone loss through individual cysts (illustrated in
`central region) or multiple cysts (illustrated in posterior region)
`along the articular surface.
`
`Fig. 6. Subarticular bone attrition score. Bone attrition is scored on
`the basis of the degree of flattening or depression of the articular
`surface relative to normal.
`
`Fig. 7. Regional subdivision of the articular margins. A. Patellar medial (M) and lateral (L) margins are evaluated using axial images. B.
`Femorotibial anterior (A) and posterior (P) margins are evaluated by combining information from both axial and sagittal (left panel) planes,
`whereas the central (C) femorotibial margins are evaluated using the coronal images (right panel).
`
`-180-
`
`

`
`Osteoarthritis and Cartilage Vol. 12, No. 3
`
`181
`
`Fig. 8. Eight-point scale for scoring marginal osteophytes. Osteophytes are scored on an eight-point scale based on size and the extent of
`margin involvement by the bone spur.
`
`Table II
`Scheme for determining total scores for the medial and lateral
`meniscus
`
`Score
`
`0
`1
`2
`3
`4
`5
`6
`
`Grades*
`
`All 0
`At least one 1, but no >1
`2 in only one region
`2 in more than one region
`3 in one or more region
`4 in only one region
`4 in more than one region
`
`*Regions: Ant, Body, Post
`
`using the coronal images. A combined ligament score was
`calculated by adding the sum of the ACL and PCL scores to
`half the sum of the MCL and LCL scores.
`The anterior horn, body segment and posterior horn of
`the medial and lateral menisci were graded separately from
`0 to 4 based on both the sagittal and coronal
`images:
`0=intact; 1=minor radial tear or parrot-beak tear; 2=non-
`displaced tear or prior surgical repair; 3=displaced tear or
`partial
`resection; 4=complete maceration/destruction or
`complete resection. A cumulative grade for each meniscus
`was also determined using the scheme shown in Table II.
`This algorithm was needed in order to adjust
`for non-
`linearity among the regional grades, which could lead to
`inconsistencies if the grades were simply summed. For
`example, if a meniscus had a grade-2 tear in the body and
`posterior horn, simply summing these regional grades
`would yield the same total score (4) as for a meniscus that
`was completely missing its posterior horn, even though the
`latter abnormality would be a far greater biomechanical
`insult to the knee. The corresponding scores derived with
`the conversion algorithm, however, would be 3 and 5,
`respectively.
`Synovial
`thickening and joint effusion were not dis-
`tinguished from each other, but graded collectively from 0
`to 3 in terms of the estimated maximal distention of the
`synovial cavity: 0=normal; 1=<33% of maximum potential
`distention; 2=33%–66% of maximum potential distention;
`3=>66% of maximum potential distention.
`Loose bodies in the synovial cavity were scored from 0 to
`3 based on number: 0=none; 1=1 loose body; 2=2 loose
`bodies; 3=3 or more loose bodies.
`Synovial cysts or bursal collections about the knee were
`specified (e.g., popliteal, anserine, semimembranosis,
`
`tibiofibular, etc.) and
`infrapatellar, prepatellar,
`meniscal,
`graded 1 to 3 based subjectively on size.
`Any other findings (e.g., patellar tendon or quadriceps
`tendon abnormalities, avascular necrosis, stress fracture,
`insufficiency fracture, focal osteochondral fracture, bone or
`soft-tissue tumors) were specified.
`Technical
`limitations, such as failed fat suppression or
`metallic artifacts that interfered with the reliability of the
`scoring of a particular case were noted.
`The final WORMS scores were tabulated as (1) indepen-
`dent values for each feature in each of the three compart-
`ments of the knee (PFJ, MFTJ and LFTJ), (2) cumulative
`surface feature (cartilage, marrow abnormality, subarticular
`cysts, bone attrition, osteophytes) scores for each compart-
`ment, (3) cumulative scores for each feature throughout the
`knee, and (4) a total combined score for the entire knee.
`
`STATISTICAL ANALYSES
`
`Scores were summarized as means and standard devi-
`ations (SD), Inter-observer agreement was based on the
`exact rating of each feature, not
`just
`the presence or
`absence of each feature, and expressed as intraclass
`correlation coefficients (ICC) by treating the data as con-
`tinuous variables. The ICC was used because it combines
`a measure of correlation with a test of the difference of
`means and is therefore a more appropriate test
`than
`Pearson’s correlation coefficient, which does not take into
`account bias between readers6. Associations among
`features were expressed as Spearman’s Rho.
`
`Results
`Table III shows the frequency of involvement (score >0)
`of each feature in the study population. Each compartment
`showed abnormalities more than 90% of the time. Ninety-
`eight percent of knees showed cartilage abnormalities. This
`was most frequent (94%) in the PFJ, but involvement of the
`MFTJ and LFTJ was also very common (89% and 71%,
`respectively). Ninety-two percent of knees showed osteo-
`phytes. The compartmental prevalence was approximately
`80%. Fifty-seven percent of knees showed bone marrow
`abnormality, most commonly in the MFTJ (35%). Bone
`cysts were present in 77% of knees, most commonly in the
`PFJ (38%), and bone attrition was present 48% of the time,
`most commonly in the MFTJ (29%). Eighty percent of the
`
`-181-
`
`

`
`182
`
`C. G. Peterfy et al.: Whole-organ MRI of knee osteoarthritis
`
`Table III
`Percent involvement
`
`Table V
`Mean score as a percent of maximum possible score
`
`MFTJ LFTJ PFJ S Region
`
`Total
`
`MFTJ LFTJ PFJ S Region
`
`Total
`
`Cartilage
`Marrow abnormality
`Bone cysts
`Bone attrition
`Osteophytes
`
`Compartment total
`Menisci
`Ligaments
`Synovitis
`
`Total
`
`14%
`38%
`
`89% 71% 94%
`35% 17% 16%
`34% 18% 38%
`29% 13% 17%
`83% 79% 81%
`
`94% 97% 91%
`70% 36%
`
`98%
`57%
`77%
`48%
`92%
`
`80%
`8%
`75%
`
`Cartilage
`Marrow abnormality
`Bone cysts
`Bone attrition
`Osteophytes
`
`Compartment total
`Menisci
`Ligaments
`Synovitis
`
`100%
`
`Total
`
`50% 27% 46%
`7%
`3% 3%
`7%
`3% 7%
`4%
`2% 3%
`26%
`9% 21%
`
`24% 14% 20%
`50% 17%
`–
`–
`–
`–
`–
`–
`–
`
`–
`
`–
`
`–
`
`–
`7%
`17%
`–
`–
`
`–
`–
`–
`–
`
`–
`
`39%
`4%
`7%
`2%
`21%
`
`–
`33%
`3%
`33%
`
`18%
`
`knees showed meniscal abnormalities. Abnormalities were
`almost twice as common in the medial meniscus (70%) as
`they were in the lateral meniscus (36%). Two patients had
`lateral meniscal cysts (grades 2 and 3). Synovial distention
`was present in three-quarters of the patients, but ligament
`abnormalities were seen in only 8%. Eleven subjects had
`popliteal cysts. Four of these were grade 1, five were grade
`2, and two were grade 3 dissecting cysts. Two subjects had
`cysts of the proximal tibiofibular joint (grades 1 and 2). One
`of these also had a grade-3 popliteal cyst. One patient had
`grade-1 prepatellar bursitis. No cases of anserine bursitis
`were identified. No intra-articular
`loose bodies were
`identified.
`Table IV shows the mean scores and their standard
`deviations for each feature (not including bursae and cysts)
`in the study population. As indicated in Table V most of
`these scores were in the lower quarter of their possible
`range. MFTJ cartilage score (50%) and medial meniscal
`score (50%) were the highest in this regard, where as bone
`marrow abnormality (4%), bone cysts (7%), bone attrition
`(2%) and ligament (3%) scores were in the lower tenth of
`their respective severity ranges.
`Table VI shows the agreement (ICC) between the two
`readers in this study. All values were greater than 0.61, and
`most were greater than 0.80. ICCs for cartilage and osteo-
`phyte scores were greater than 0.9. The poorest agree-
`ment was for bone attrition, but the prevalence of this
`feature was extremely low, particularly in the PFJ and LFTJ
`where the frequency of positive scores was too low to
`calculate ICC reliably.
`the
`As shown in Table VII, scores among many of
`individual features, particularly cartilage, bone cysts, bone
`attrition, osteophyte and meniscus, were relatively strongly
`
`Interestingly, meniscal scores associated
`associated.
`strongly with ipsilateral FTJ scores but poorly with contra-
`lateral FTJ score or with PFJ scores (Tables VII and VIII).
`Associations between features in region S (marrow abnor-
`mality and bone cysts) and those elsewhere in the knee
`were poor.
`
`Discussion
`MRI offers a unique opportunity to evaluate all com-
`ponents of a joint simultaneously and therefore to provide a
`whole organ assessment of the status of structural damage
`in patients with OA. Whole-organ assessment could help
`discriminate different patterns of intra-articular involvement
`in OA; detect early, potentially preclinical, stages of OA;
`identify structural risk factors for developing clinical OA; or
`increase scope and sensitivity to change for monitoring
`disease progression and treatment response in patients
`with established OA. This would aid subject selection,
`treatment monitoring, and safety assessment
`in clinical
`trials of putative new therapies for OA and in studies
`exploring the pathophysiology and epidemiology of OA.
`The
`semiquantitative whole-organ MRI
`scoring
`(WORMS) method presented in this report offers an initial
`instrument for performing multi-feature assessment of the
`knee using conventional MRI. It takes into account a variety
`of features that are currently believed to be relevant to the
`functional integrity of the knee and/or potentially involved in
`the pathophysiology of OA. It scores each of these features
`with a sufficient number of increments to allow detection of
`what are speculated to be clinically relevant changes –
`although, this was not explicitly tested in the current cross
`
`Cartilage
`Marrow abnormality
`Bone cysts
`Bone attrition
`Osteophytes
`
`Compartment total
`Menisci
`Ligaments
`Synovitis
`
`Total
`
`Table IV
`Mean scores*
`
`LFTJ
`
`8 (9)
`0.4 (1)
`0.4 (1)
`0.3 (0.9)
`6 (6)
`
`15 (15)
`1 (2)
`
`MFTJ
`
`15 (9)
`1 (0.9)
`1 (2)
`0.6 (1)
`9 (7)
`
`26 (18)
`3 (2)
`
`S Region
`
`0.2 (0.6)
`0.5 (0.7)
`
`PFJ
`
`11 (6)
`0.4 (1)
`0.8 (1)
`0.3 (0.8)
`6 (5)
`
`18 (10)
`
`Total
`
`33 (17)
`2 (3)
`3 (3)
`1 (2)
`21 (16)
`
`4 (3)
`0.1 (0.3)
`1 (1)
`
`60 (33)
`
`*Values are mean score (standard deviation) for each feature
`
`-182-
`
`

`
`Osteoarthritis and Cartilage Vol. 12, No. 3
`
`183
`
`Table VI
`Inter-reader agreement (ICC)
`
`MFTJ LFTJ PFJ S Region
`
`Total
`
`1.0
`1.0
`
`0.98
`0.90
`0.98
`0.61
`0.93
`
`0.97
`0.94
`
`0.99 0.98
`0.96 0.82
`0.95 0.73
`–
`0.78
`0.94 0.98
`
`0.98 0.99
`0.81
`
`Cartilage
`Marrow abnormality
`Bone cysts
`Bone attrition*
`Osteophytes
`
`Compartment total
`Menisci
`Ligaments
`Synovitis
`
`Total
`
`0.99
`0.74
`0.94
`–
`0.97
`
`0.87
`1.0
`0.74
`
`0.98
`
`*The frequency of cases of bone attrition in the LFTJ was too
`small to reliably calculate ICC.P<0.01 for all other values.
`
`Table VII
`Spearman Rho for association between cartilage scores and the
`corresponding comparmental and total scores of other features
`
`Marrow abnormality
`Bone cysts
`Bone attrition
`Osteophytes
`Medial meniscus
`Lateral meniscus
`Ligament
`Synovitis
`
`MFTJ
`
`LFTJ
`
`PFJ
`
`0.73*
`0.61*
`0.56*
`0.81*
`0.75*
`0.08
`−0.17
`0.40
`
`0.43
`0.57*
`0.11
`0.47*
`0.08
`0.47*
`0.18
`−0.34
`
`0.40
`0.62*
`0.60*
`0.51*
`−0.16
`0.21
`0.00
`0.25
`
`Total
`
`0.21
`0.49*
`0.37
`0.66*
`0.50*
`0.50*
`0.00
`0.21
`
`*Prob>|Rho| was 0.05 or less.
`
`Table VIII
`Spearman Rho among meniscal and compartment-total scores.
`
`Meniscus
`
`Lateral
`
`0.07
`0.55*
`0.05
`
`Medial
`
`0.80*
`−0.04
`−0.02
`
`Total
`
`MFTJ
`LFTJ
`PFJ
`
`*P<0.01.
`
`sectional study design. Based on the conditions of this
`study, it shows high inter-reader agreement among trained
`radiologists experienced in MRI interpretation of the knee,
`and employs conventional MR images that can be pro-
`duced with clinical MRI systems available at most hospitals
`and imaging centers around the world. It must be empha-
`sized, however, that WORMS and its constituent subscores
`are not claimed to be the definitive instrument for whole-
`organ MRI evaluation of the knee, but rather a first step in
`what is hoped will be a process of continuous improvement
`and refinement of the basic scheme.
`WORMS incorporates 14 articular features. Of these,
`articular cartilage loss and osteophytes are the most
`broadly accepted as being central to the pathophysiology of
`OA. MRI is ideally suited to imaging these two cardinal
`features. Cartilage imaging with MRI,
`in particular, has
`received considerable attention in the past several years.
`Numerous studies have shown MRI to be highly sensitive
`and specific for detecting focal cartilage defects and
`thinning7–11. The most commonly used pulse sequences
`for examining articular cartilage have been fat-suppressed,
`
`T1-weighted, 3D gradient-echo7
`(Fig. 9) and fat-
`suppressed T2-weighted or intermediate-weighted FSE9–11
`(Figs. 10–12). Both of these techniques are easy to perform
`and widely available on conventional clinical MRI systems.
`In this study we combined information from both techniques
`to evaluate the cartilage. Several investigators have devel-
`oped and validated computer-assisted methods for quanti-
`fying articular cartilage volume and thickness12–16,18,20.
`WORMS, however, relies on semiquantitative assessment
`by the trained eye.
`A number of semiquantitative scoring methods for grad-
`ing cartilage loss on MRI have been developed19,21–23, but
`no single method has been accepted broadly as a standard
`for clinical research thus far. Most of these scoring stra-
`tegies derive from arthroscopy, and grade primarily the
`depth of focal cartilage loss over a four-point scale. The
`method employed in WORMS is also based on this classic
`scheme, but expands the scale to eight points in order to
`capture different patterns of regional cartilage loss and
`more information about extent of surface involvement
`(Fig. 3). Each point on the scale is one integer, except for
`grade 2.5. This point interval is smaller than the others are
`because the difference in cartilage loss between a small
`focal partial-thickness defect (2.0) and a small focal full-
`thickness defect (2.5) is proportionately smaller than the
`difference between the other intervals. The adjustment,
`accordingly, improves linearity of the scale. WORMS also
`incorporates changes in cartilage signal on T2-weighted
`images, which have been shown to represent areas of
`chondromalacia that may precede focal tissue loss24–27.
`Despite the expanded scale of
`the WORMS cartilage
`score, inter-reader reproducibility was high (Table VI). The
`longitudinal
`reproducibility of
`this method was not
`addressed in this cross sectional study design, but the
`extra increments in the scale would be expected to in-
`crease sensitivity to change. Exactly how much change is
`clinically relevant is not currently known. However, in the
`population of OA patients examined in this study, most
`knees fell into the midrange of possible scores for cartilage
`damage (Table V). One consequence of using a larger
`scale for cartilage than for other
`features – besides
`osteophytes – is that cartilage score is given greater weight
`in the total score for the knee. This may be appropriate,
`given the importance that articular cartilage is believed
`have in the pathophysiology of OA. However, this requires
`validation in longitudinal studies.
`Cartilage score was determined independently in each of
`14 articular surface regions. The femoral surface was
`divided into anterior, central and posterior regions based on
`the different relative functions of these regions (Fig. 1). The
`tibial surface was similarly subdivided into anterior, central
`and posterior regions in order to allow examination of
`associations with abnormalities of the anterior horn, body
`and posterior horn of
`the meniscus, respectively. The
`anterior and posterior osteochondral
`junctions and the
`insertions of the meniscal
`ligaments to the joint capsule
`were used as the anatomical landmarks for subdividing the
`articular surface. This was because these landmarks were
`easily identified on most MRI examinations and because
`they delimited surfaces of relatively different function within
`the knee. The anterior region of
`the femoral cartilage
`corresponded to the patellofemoral joint, the central region
`between the anterior and posterior meniscal
`ligaments
`corresponded to the area of the femoral cartilage loaded
`during standing and normal walking, and the posterior
`region corresponded to the area loaded during deep flexion
`of the knee. It was felt that this subdivision strategy yielded
`
`-183-
`
`

`
`184
`
`C. G. Peterfy et al.: Whole-organ MRI of knee osteoarthritis
`
`the optimal compromise between reproducibility and func-
`tional correlation.
`It should be noted that each reader
`subdivided the joints independently. Accordingly,
`this
`potential source of variability was included in the assess-
`ments of inter-observer reproducibility (Table VI). How this
`may be affected by meniscal subluxation or tear, or by
`longitudinal study designs was not, however, addressed.
`The imaging protocol used in this study was designed for
`whole organ assessment rather than optimal evaluation of
`any indiv