Osteoarthritis Initiative

July 12, 2001 (historical)

Surrogate “White” Paper on Imaging
Summary of January 11, 2000 Imaging Meeting

Meeting Agenda

The meeting began with presentations reviewing the current status of the most promising surrogate endpoints in OA, in terms of the roles and performance criteria listed in a manuscript from Dr. Peterfy. Presentations were limited to 20 min and followed by 10 min of discussion. After lunch, there was a general round-table discussion of the points covered in the morning, followed by the drafting of a list of research priorities and imaging protocol options for the final study.

9:00 AM Welcome and Introductions
Gayle Lester and Charles Peterfy
9:10 AM Presentation on performance metrics for imaging markers in clinical trials
Charles Peterfy
9:40 AM Presentation(s) on radiographic markers of OA in the knee, hip (hand):
Christopher Buckland-Wright / Kenneth Brandt
10:10 AM Presentation on DXA, CT and MR markers of subarticular bone: Sharmila Majumdar
10: 40 AM Presentation on MRI markers of articular cartilage
Laurence Hall
11:10 AM Presentation on MRI whole-organ scoring: Health ABC experience Michael Nevitt
11:40 AM Presentation on optical coherence tomography in OA Mark Brezinski
12:10 AM Lunch Break
1:30 PM General round-table discussion
2:30 PM Prioritize list of research questions
3:50 PM Draft imaging protocol options
4:50 PM Assign any additional tasks needed to prepare for the Winter Meeting
5:00 PM Adjourn


Participants: Charles Peterfy, Michael Nevitt, Sharmila Majumdar, Marissa Laserre, Jim Witter, Christopher Buckland-Wright, Laury Hall, Tamara Harris, Steve Einstein, Colin Miller, Wayne Carter, Mark Brezinski, Gayle Lester, Joan McGowan, Gregory Downing

The meeting began with an overview of the OA Initiative, describing the history of the Initiative and summarizing past meetings and discussions. This meeting is a preliminary for the big meeting on February 28-29, 2000 at Lister Hill in Bethseda, Maryland. One of the major purposes of these endeavors is to develop surrogate endpoints for OA that would facilitate the development of new treatments for disease and in combination with the companies serve the public health. This initiative is not to develop products but to facilitate their future development. Materials had been sent out earlier in the month to serve as discussion points for this meeting; these included a number of objectives for this meeting and issues to be addressed related to imaging.

Dr. Charles Peterfy went over these issues one by one and drafted a series of protocols that have been useful in clinical trials that could easily be applied to the OA Initiative. The objectives will have different priorities depending on what biomarkers one is investigating.

Initially, one should decide which joints will be considered. Will we focus on knee, hip, spine, shoulder?

We need to agree on the performance criteria by which different imaging alternatives will be compared. The metrics should be similar between imaging and biochemical markers. They could synergize each other so both should be evaluated by same metrics. The presentations of the day will review and summarize current knowledge of imaging techniques available today along the lines of new performance criteria. We would like to identify where the holes are in the knowledge base and what are the most important questions for the Initiative to address. We will need to prioritize the research questions and develop imaging protocols to address these questions. Perhaps one or two. These "straw proposals" will be presented at the February meeting to the Steering Committee; the data from this meeting will be presented as well. There may be ancillary studies to accompany the initiative.

Dr. Peterfy then reviewed the metrics for imaging markers based on his PowerPoint presentation.

Markers can be used diagnostically or to select subgroups of patients to be studied or to select those most likely to proceed to further disease. Markers can be used to characterize the disease process versus the current disease state and they can be used to monitor progression and response to treatment. These are not always the same markers. How can one measure changes in the marker; how sensitive is the marker; imaging for safety monitoring is not always done but is also important. Clinical trial decisions are focused on identification of agents that modify course of disease. Should agents be taken to further levels of testing such as for monitoring progression, response to treatment, and identification of complications. Metrics should be correlated with performance criteria.
Performance Criteria:

  1. Surrogate validity - how well does marker correlate with disease outcome (is marker linked to clinical outcome)? Its importance still depends on the context in which it is used, i.e., what questions are being asked.

  2. Predictive power and diagnostic accuracy can be expressed in terms of dynamic range of the marker. This is more a concept than a number that we usually arrive at. It represents the proportion of the true clinical outcome that is captured in the change of the surrogate endpoint. If the change falls within the clinically relevant range, it would be found to have suitable sensitivity.

  3. The other more important aspect of the marker is responsiveness to disease and therapy - theoretically determines the length of the time of follow up. A marker that can reliably detect small changes over relative short periods of time would allow shorter study time and a quicker turn around to get an agent to clinical market. This aspect is critical in optimizing patient care to be able to detect responders and non-responders.

  4. Variability-
    Biological variation - targeted patient recruitment can be used to limit this
    Technical variation - very important in multi-centered studies
    At the time of data acquisition - standardization of instrumentations; stability of protocol Measurement process - 1 reading facility versus multiple readers; use of computer technology Improving precision of measurement will decrease the number of subjects needed and the duration of the study. Once two markers are shown to be equivalent on these fronts, cost and convenience can be considered.

One must establish the change criteria above which one has 95% confidence that the changes is real. This can be done by multiplying the precision error by 2.8 (based on data from osteoporosis research) and called the least significant change - analogous to the minimum detectable difference used by OMERACT. In certain contexts it is possible that one might find adequate decision-making power at only 80% confidence (multiply precision error by 1.8). This is called the trend assessment margin.

Follow up time interval is the change criterion value divided by the median change per year. The numbers presented are for JSN at an average rate of change of 0.2mm/year (based on meta analysis). This then dictates the monitoring time interval. Markers can be ordered as more or less sensitive based on these evaluations. This approach is subject to cohort biases and so the concept of standardized precision errors has been raised (again, initially in osteoporosis research). Standardized precision error would allow comparison of quite different types of markers (i.e., radiography, MRI, or biochemical markers); the standard deviation or precision error of a technique is divided by the response ratio (the ratio of the rate of change of the index marker to that of the reference marker). A technique that has twice the precision error of the reference technique but 4 times the responsiveness would thus have a standardized precision error half that of the reference standard. This approach combines both responsiveness and precision error in a single measure of longitudinal sensitivity, and is less group dependent than simple change rates because part of the cohort effects cancel out in the response ratio. This may be a good parameter for us to consider in evaluation of appropriate techniques to be used. There are examples shown in the presentation.

This is a cohort study and will most likely require several centers to collect sufficient data. There may be 100's of sites included. The larger the study the simpler the protocol must be to allow reproducibility. Multi-center data acquisition requires wide availability, stable performance, and good reproducibility. The protocol should be designed to insure this and sites must have the technical competence to carry out the techniques. Easy to use, tolerable to patients, and cost effective. The reading can be done later at a centralized facility. That improves the precision of the technique because standardization can be optimized. One reader or a series of trained readers can be used - the volume of reading needs to be large enough to keep the performance high. This can be combined with very powerful tools and software simply for that purpose. This is why many studies use a centralized facility for reading.

Surrogate validity is the target of most of these markers we are concerned with. It assumes that the surrogate marker, which measures a particular morphological, anatomical or process-related feature, lies directly on the disease pathway and on the intervention pathways as well. The extent to which there are alternate pathways of disease and alternate mechanisms of therapy is the extent to which this surrogate marker will not register those adverse or beneficial effects, respectively. In addition to this theoretical validity it is critical that the technique used to measure the surrogate marker also be valid. This depends on the conditions of data acquisition and reading / analysis. Sometimes it is confusing because there are these two aspects of validity.

The search for earlier readouts causes us to look for techniques with higher responsiveness. If one technique can measure cartilage loss over a 2 year period and another technique can measure cartilage loss over a 1 year period, then the second technique would provide better responsiveness based on techniques looking at the same structural outcome. One can further increase responsiveness by looking up stream at earlier markers. For example, matrix loss from articular cartilage might be able to demonstrate change over a period of a few months. However, a biochemical marker might pick this up in only one month. The farther we move away from the clinical endpoint it seems the more responsive the marker is and the more desirable from that standpoint, but there is greater opportunity for alternate disease pathways that could undermine that marker's surrogate validity.

In contrast to this simplistic model of disease, the reality in OA is more complex. There are multiple determinants of pain and mechanical malfunction and there may be many more we do not know. This is a complex, interactive matrix. Accordingly, the tendency to focus on only 2 or 3 structural features gives only a keyhole view of disease. We need to be sure that we use a broader palate of imaging markers and a whole organ evaluation of the joint, which is consistent with the disease being one of organ failure. The concept of whole organ evaluation was a dream in the past but is now reality. There are current technologies available to evaluate various aspects of disease changes in the same patient and to monitor these over time. So that we can evaluate the status of each of these articular components in a patient and monitor them longitudinally to begin to understand the relationship between joint changes and the clinical and functional outcome.

Here is a list of the most promising markers for OA. The markers available for evaluating cartilage and specifically cartilage loss are radiographic joint space width, MRI cartilage score; MRI cartilage volume, MRI cartilage thickness; ultrasound cartilage thickness, ultrasound backscattering, and optical coherence tomography. Compositional markers include evaluation of collagen with MR T2 relaxation, magnetization transfer effect, and optical coherence tomography with polarization; and evaluation of PG matrix with gadolinium enhanced MR, sodium MR, or water diffusion coefficient. Bone markers include osteophyte analysis with X-ray or MR; technetium uptake indicating bone synthesis; and a host of trabecular measures that can be applied to sub-articular bone. Marrow edema can be see with scintigraphy or MRI. Sub-articular cysts can be observed with MR or X-ray; contrast-enhanced MRI or Doppler ultrasound can be used to assess synovitis. Other joint structures have been imaged for years but not examined in this context include meniscus; disc; cruciate ligaments; and bursa. These can be imaged with MRI or ultrasound. These techniques paint a picture of where work needs to be done. What needs to be validated with such a large cohort study?

Group Discussion:

  • Everyone understands that this initiative is in the interest of the public. So we are interested in patients but there are many people who are not patients who have disease (asymptomatic or just untreated) and need to be captured in this type of study. Many will become patients due to disability but are not at present.

  • We need to apply these techniques to such groups so that they can be better defined.

  • Our charge is to develop the research tools and define them properly so that they can support a diversity of study design.

  • The OA Initiative could achieve that . It would be of interest to put together these structural changes to see what might be predictive of what is going to happen down the road with disease.

  • You can validate a surrogate but you always make compromises - we are using JSN but it may not be the most valid in all circumstances.

  • How does imaging relate to clinical outcomes? The really important outcome is how the patient is doing? How are these assessments being done?

  • It would be nice to come up with something biological processes that relate to the imaging changes.

  • Are we sure what ties in with pain? We must understand the pathology and biology of the joint.

The second presentation was by Dr. Christopher Buckland-Wright. His major points addressed "Radiographic Markers of OA in the Knee, Hip and Hand".
Christopher Buckland-Wright

Radiographic Procedures for OA Joints
  1. Primary outcome measure:-
    The validity of joint space width (JSW) as a surrogate measure for articular cartilage thickness was based upon the following:-
    • Buckland-Wright JC, Macfarlane DG, Lynch JA, Jasani MK, Bradshaw CR. (1995) Joint space width measures cartilage thickness in osteoarthritis of the knee: high resolution plain film and double contrast macroradiographic investigation. Ann Rheum Dis 54, 263-268.


    This investigation confirmed that JSW reliably and accurately measures articular cartilage thickness in the medial diseased but not the lateral tibio-femoral compartment in OA knees. In addition, JSW measurements in the standing weight-bearing knee measures the status, i.e. compressibility and the tissue thickness of articular cartilage

  2. Optimum radio-anatomical position for the knee joint.
    The different radiographic positions used to assess the knee in the standing position were reviewed for radiography of the knee:
    1. without the use of fluoroscopy:-

      • Buckland-Wright JC, Wolfe F, Ward RJ, Flowers N, Hayne C. Substantial superiority of semiflexed (MTP) views in knee osteoarthritis: a comparative radiographic study, without fluoroscopy, of standing extended, semiflexed (MTP), and schuss views. J Rheumatol 1999;26:2664-74.


    2. The results showed that in the semi-flexed MTP view: i) joint positioning, ii) joint repositioning and iii) JSW measurement were significantly more reproducible than the in the other views.

    3. with the use of fluoroscopy for precise positioning repositioning of the joint:-

      • Buckland-Wright JC, Macfarlane DG, Williams S A, Ward RJ. (1995) Accuracy and precision of joint space width measurements in standard and macroradiographs of osteoarthritic knees. Ann Rheum Dis; 54: 872-880.

    The results showed that in the semi-flexed view, JSW measurement was significantly more reproducible than the in the standing extended knee view.

  4. Optimum radio-anatomical position for the hip joint.
    Reviewing the published radiographic procedures for the hip:-

      • Conrozier T, Lequesne M, Tron AM, Mathieu P, Berdah L, Vignon E. The effects of position on the radiographic joint space in osteoarthritis of the hip. Osteoarthritis Cart 1997; 5: 17-22.

      • Buckland-Wright JC. Quantitation of radiographic changes. In Brandt KD, Lohmander S, Doherty M, eds., Osteoarthritis: Oxford, University Press, 1998, 459-472.

      • Buckland-Wright JC. (1998) Protocols for radiography. In: Osteoarthritis. KD Brandt, S Lohmander, M Doherty (eds). Oxford, University Press. pp. 578-580.


    The results indicated that the patient should be radiographed with the foot medially rotated and the x-ray beam centred upon the hip joint.

  5. Optimum radio-anatomical position for the hand joint.
    As with the radiographic procedures for the joints described above, those described in published protocols are required to ensure reproducible reposition within and between patients.

      • Buckland-Wright JC. Quantitation of radiographic changes. In Brandt KD, Lohmander S, Doherty M, eds., Osteoarthritis: Oxford, University Press, 1998, 459-472.

      • Buckland-Wright JC. (1998) Protocols for radiography. In: Osteoarthritis. KD Brandt, S Lohmander, M Doherty (eds). Oxford, University Press. pp. 578-580.


  6. Summary on radio-anatomical position for all joints.
    The above methods are reviewed in the following publication:-

      • Buckland-Wright JC. Radiographic assessment of osteoarthritis: comparison between existing methodologies. Osteoarthritis Cart 1999;7:430-3.


    This review concluded with the following principals for radio-anatomical positioning of OA joints, in order that features recorded in the radiographs can be reliably and reproducibly measured.

    • The joint should be weight bearing (for the hand under muscle tension) and in a position consistent with normal functional or cyclic loading of the joint.

    • The plane of JSW measurement should be i) perpendicular to the articular surfaces and to the central of the x-ray beam and parallel to the film, and ii) coincident with the principal region of load transmission.

    • Minimum JSW measurement is more likely to assess articular cartilage compressibility and thickness as the measurement is taken at or close to the site of load transmission in a joint, than JSW area or mean JSW area. The latter includes the entire zone of the interbone distance in their assessment, including tissue outside the main region of load transmission.

    • Computerised JSW measurement is preferable to manual methods as they are more accurate and reproducible, and overcome the limitations of observer based variability.

    • Correction for radiographic magnification is desirable since it significantly reduces the number of patients required for a study [5].

    • The increased spatial resolution of images obtained with high definition microfocal radiography improves JSW measurement precision, reducing the study numbers [3].


    Quantifying Disease Progression

  7. Measurement of disease progression in knee OA
    Published results of quantitative progression have been obtained from microfocal radiographic studies:-

      • Buckland-Wright JC, Macfarlane DG, Lynch JA, Jasani MK. (1995) Quantitative microfocal radiography detects changes in OA knee joint space width in patients in placebo controlled trial of NSAID therapy. J Rheumatol 22, 937-943.

    1. Joint space width. This study showed that JSW narrowed at different rates depending upon the disease status. In knees with early disease the rate was slower (0.18 mm/yr [5] than when the joint had lost 50% or more of normal healthy articular cartilage thickness. Changes in joint space width were significantly different between patients receiving NSAID and placebo.

    2. Osteophytes changed in number and size with disease progression.

    3. Subchondral cortical plate did not alter during the course of the study.

    4. Subarticular trabecular bone The horizontal trabeculae were found to thicken up before articular cartilage, measured as JSW, had narrowed significantly, as reported in the following publication.

      • Buckland-Wright JC, Lynch JA, Macfarlane DG. (1996) Fractal signature analysis measures cancellous bone organisation in macroradiographs of patients with knee osteoarthritis. Ann Rheum Dis; 55:749-755.

  8. Measurement of disease progression in hip OA
    Results based upon manual methods of measurement have revealed that the rate of joint space loss to range between 0.59 - 2.0 mm/yr. This rate is faster in joints with rapid deteriorating forms, as described by Lequesne:

      • Lequesne M. Quantitative measurements of joint space during progression of osteoarthritis: "Chondrometry". In: Osteoarthritic Disorders, KE Kuettner, VM Goldberg (eds), American Academy of Orthopaedic Surgeons, Rosemont, Illinois. 427-44.

  9. Measurement of disease progression in hand OA
    Published results from microfocal radiographic studies have provided information similar to that described above for the knee:

      • Buckland-Wright JC, D G Macfarlane, J A Lynch, B Clark. (1990) Quantitative microfocal radiographic assessment of progression in osteoarthritis of the hand. Arthritis Rheum; 33: 57-65.


    Quantification of disease related changes in the radiographic features of OA joints is dependent upon published and validated protocols, that permit:

    1. reliable, reproducible radio-anatomic positioning of weight bearing joint and in a position consistent with its normal functional or cyclic loading.

    2. the plane of measurement for each feature to be perpendicular to the central of the x-ray beam and parallel to the film.

    Discussion: What influences changes in JSN across the compartments? Osteophytes increase with loss of cartilage but once cartilage loss is stabilized, osteophytes stabilize. Are there data on microfocal techniques? The Bayer data speak to these somewhat.

Dr. Kenneth Brandt presented evaluations of radiological assessment of osteoarthritis. Following are some of the notes and material from slides used in his presentation: "Joint Space Narrowing in the Knee"

What predicts X-ray progression? Increased knee pain may alter evaluation of JSN. Changes in body weight can alter space from film and/or compression. Effusions will also alter image. On average the JSN annually decrease more in patients than in population based groups. The data from the following studies show this.

Descriptions of Medial Joint Space Narrowing (JSN)
Measured in Standing AP Radiographs
Annual Rate of JSN (mm/yr)
Lethbridge-Çejku et al BLSOA cohort
Felson et al Framingham cohort
Dieppe et al Bristol OA500
Lesquesne et al Clinical OA sample
Ravaud et al Clinical OA sample
Kirwan et al Clinical OA sample


Studies of Biomarkers/Mediators of OA Progression
Using Standing AP Radiographs
Study Variable
Association with
OA Progression
Sharif et al Serum COMP
(GT/E 3.17µg/ml increase over 12 months)
Predicts 5-yr progression:
0% sensitivity
8% specificity
Dieppe et al
99mTC-MDP uptake

(Baseline bone scans rated as normal or abnormal)
of 55 OA knees with normal scans progressed over 5 yrs
2 (60%) of 87 OA knees with abnormal scans progressed over 5 yrs
cAlindon et al
Vitamin D (dietary intake and serum concentration)
3-4-fold increase in risk of progression over 8 yrs in lower vs. upper tertiles
cAlindon et al
Vitamin C intake 3-fold increase in risk of progression over 8 yrs in lower vs. upper tertiles*


  • Similar association with incident knee pain
  • GT/E (Greater than or equal)
    Pharmacologic Modification of OA Progression
    Measured in Standing AP Radiographs
    Study Treatment
    Reginster et al Glucosamine sulfate
    Vs. placebo
    1. JSN prevented
    2. Pain reduced


    Technical Limitations of the Bilateral Standing AP Radiograph

    • Medial tibial plateau (MTP) and x-ray beam are often misaligned.

    • Uncontrolled rotation of the knee alters osteophyte profiles.

    • CV of repeated measures of medial JSW GT/E 15-20%.

    • Full extension exaggerates JSW.

    • Increasing knee pain may confound interpretation of JSN.


    First Generation (Fluoroscopically Assisted) Protocols
    For Standardized Knee Radiography
    Key Procedures
    Knee Flexion/Extension
    # of Knees
    Ravaud et al
    Full extension
    Angle of x-ray beam adjusted to bring the MTP* into sharpest focus
    20 (N)
    Buckland-Wright et al
    1.5-10 degrees flexion
    Knee flexed to super-impose the anterior and posterior margins of the MTP
    10 (N)
    25 OA
    Piperno et al
    ~30 degrees flexion (schuss)
    1. Angle of x-ray beam adjusted to bring the MTP into sharpest focus
    10 (N)
    10 OA

    * MTP = medial tibial plateau
    † CV of JSW LT/E 10% for all
    LT/E (Less than or equal)

    Second Generation (Non-Fluoro-Assisted) Protocols
    For Standardized Knee Radiography
    Key Procedures
    Knee Flexion/Extension
    MTP/Beam Alignment
    Foot Rotation
    N of Knees
    Ravaud et al
    Full extension of knee
    5 degrees caudal angulation of x-ray beam
    15 degrees external foot rotation
    20 (N)
    36 OA
    Buckland-Wright et al
    -10 degrees flexion of knee
    (1st metatarsophalangeal joint aligned with patella) Horizontal beam
    15 degrees foot rotation
    1. Peterfy et al
    1. PA
    1. ~20 degrees flexion of knee
    2. 10 degrees caudal angulation
    3. 10 degrees foot rotation
    18 (N)

    19 OA



    Future Directions for Research:
    Effects of Uncontrolled Variables in Existing Databases
    Effects of non-standardized knee position
    Effects of failure to control for severity of knee pain

    Relationship Between Knee Pain (WOMAC) and
    Extended View JSW in OA Knees
    All knees (WOMAC score 5-25)
    Painful knees (WOMAC score 10-25)

    *P < 0.05; † P < 0.01

    Note: correlation between WOMAC pain and JSW in semiflexed views = -0.05 (n.s.)

Future Directions for Research:
Continued Development of Standardization Protocols

    Direct comparisons of 1st & 2nd generation protocols
    Field studies of the implementation of research protocols

Field Test of the Reproducibility of Automated JSW Measurements in
Semiflexed AP Views of OA and Normal Knees
OA Severity (K&L Grade)
Standard Error
of Measurement
Laboratory Standard (Buckland-Wright et al)
5 hospital radiology
depts. (Mazzuca et al)
1. 2 satisfactory films
2. 1 satisfactory film
3. 0 satisfactory films





Recommendations for Reporting of Radiologic
Methods in Future Studies of OA Progression

  1. AP or PA view

  2. +/- fluoroscopic positioning

  3. Degree of flexion (if not full extension)

  4. Measures taken to note and adjust for knee pain
    (if knee is fully extended)

  5. Standard error of measurement (SEm) of JSW

Group Discussion: What is meant by standard error of the JSW measurement? It is a measure of intra patient variability? Actually, it is the relationship between noise and variation. We need to look at the standard deviation of the scores. It is actually an intra class coefficient of variation and is heavily biased by the denominator.

Dr. Sharmila Majumdar's presentation was entitled "Cartilage-bone Interactions in Osteoarthritis".

The objectives of this research were as follows.

To develop high-resolution in vivo MR techniques to depict and quantify cartilage and trabecular bone structure in the knee joint.

To segment articular cartilage images using Immersion-based Watershed Algorithm.

To determine :
Cartilage Thickness
Cartilage Volume / Epicondylar Distance
Correlation of above with T2

To quantify spatial variation of the trabecular bone structure along the distal femur and proximal tibia.

To examine the difference between
tibia and femur
presence of disease and absence of disease

To derive relationships between stereological bone parameters, cartilage morphometry and clinical findings of OA as determined by the radiography-based Kellgren- Lawrence (KL) scale.

To determine the reproducibility of cartilage and bone measurements.

To correlate cartilage morphometry with subchondral bone changes

To derive correlations between cartilage morphometry and stereological bone structure parameters.

Her conclusions were as follows:

  1. High resolution MRI has the potential to show variations in cartilage and trabecular structure.

  2. An increased variation in cartilage structure is observed in the tibio femoral than in the patellar region.

  3. Impact of OA on trabecular bone is different in the tibia versus the femur.

  4. Changes in cartilage and trabecular structure depend on the extent of the disease.

  5. These changes may be used to assess the relationship between bone and cartilage degeneration.

  6. Further studies will establish if high resolution MRI can provide an efficient technique for early detection of changes in trabecular bone structure relative to cartilage alteration and arthrosis.

Group Discussion: In the description of the total joint, are we taking into account the possibility of disease in other joints? Will that have an impact?

Dr. Laurance Hall gave his presentation on MR Imaging of cartilage entitled: "MRI Markers of Articular Cartilage". Summary and highlight slides are shown below.

Human Joints Dimensions Quality
Knee Thickness, volume Image contrast
Hip Thickness Image contrast
Hand Thickness MRI parameters
Animal Knees Dimensions Quality
Guinea pig Thickness Relaxation times
Rat Thickness Relaxation times
Rabbit Thickness Image contrast
Dog Thickness Image contrast
Biopsy Internal structure Relaxation times
Enzymatic degradation Dimensions 23Sodium


Current Radiological Practice
Soft tissues (cartilage, menisci, tendons, ligaments)
Bone (cortical, trabecular)
Extra capsular (musculature, vasculature)
Available on all clinical scanners
Read by orthopaedic radiologists
Cartilage lesions; damage to menisci, ligaments; bone erosions

Research Protocols

Measure thickness, total volume

Patient Screening and Recruitment

Pathological status of whole organ

Exclusion criteria

Radiological scoring of cartilage and bone

Initial Measurement

Cartilage dimensions

MRI parameters

Serial Measurements of Progression

Regional changes


MRI of the Human Knee
1999 2000
3D MRI: 15 minutes 3 D MRI: 2-5 minutes
3D MRI: 15 minutes 3D MRI: 5-10 minutes
Resolution: 0.6 x 0.6 x 0.6 mm Resolution: 0.6 x 0.3 x 0.3 mm
Radiological reporting Automated thickness
Arbitrary scoring Local volumes


Variations in image contrast Fully automated: 10-30 minutes
Radiological reporting 2D - MRI parameters
  3D - MRI parameters
  23Sodium MRI
  Gadolinium contrast


Goal~Multi-centre, 3D-MRI Radiology

selection of scan centres

cross-validation of scan protocols

radiological scoring of "whole organ"

Questions~ Which make of scanner:?

Which scoring system?

USA, only or worldwide

Which clinical measurement?

Will there be a central scoring facility?

Goal~Measurement of Cartilage Dimensions

Thickness, volume, localized volumes


Coefficient of variation

Questions~ Which regions will be measured?

How to maintain objectivity and accuracy.

Single measurement centre or distributed software?

Goal~Measurement of Cartilage Quality

Protocols for water MRI parameters

Cross validation

Questions~ Which parameters?

Analysis of data?

Statistical summary

Spatial heterogeneity

Biochemical conversion

Goal~Proteoglycan Measurement

23Sodium MRI

Gadolinium uptake

Questions~ What protocols for scanners?

Which hardware for scanners?

Reference samples for standardization

Patient compliance


Group Discussions: How much can be done on actual clinical images? Many of the techniques described require higher resolution.

Dr. Michael Nevitt presented MRI Whole organ scoring in the Health ABC experience. Below are highlights from his presentation: "Knee OA Imaging in The Health ABC Study".
Co-authors are Tamara Harris, MD and Charles Peterfy, MD, Ph.D.

Importance of Knee OA in Health ABC
Knee OA a major weight-related disease affecting physical function
How do knee pain and OA affect change in body composition and decline in strength, fitness and physical function and how do age-related changes in weight, body composition and strength affect the development of knee pain and OA?
Focus on Sx knee OA (Sx knee OA=pain + X-ray findings)


Baseline Knee OA Assessment
Knee symptoms

frequent pain (most days of month - NHANES)

>= mod pain with activity (WOMAC)


X-ray and MRI

in those with knee pain (frequent or activity)

random sample without pain

Sx knee OA = pain + x-ray findings


Follow-up Assessments

Annual knee symptom assessment

X-ray in those with new knee pain
5-year FU x-ray in those with baseline pain
Periodic measurement of quad strength, fitness, physical function


HABC and validation of surrogate markers for knee OA outcomes
MRI, x-ray and other predictors of 5-year outcome

progression: x-ray, Sx, disability

body comp, quad strength, fitness, physical function

marker of early OA

Feasibility of MRI in large studies


reading and analysis

Missing pieces

interim imaging


Knee Radiography
Bilateral, PA TF view

'fixed' flexion (`20-25°)

fixed ext. rot. (10°)

beam centered btw knees on a line with the plane of the joint space

beam angle 10°

CV JSW 2.6% (Peterfy, 98)

Axial weighted PF view

standing with support

knee flexed (30-40°)

beam aligned vertically through the PF joint space


MRI Acquisition: Short (15 min/knee) Protocol
1.5-T Signa scanners
Axial localizer T2 FSE (1 min)

PF cartilage, osteophytes

Sagittal T2 FSE (entire synovial cavity) with fat suppression (4.5 min)

TF cartilage, osteophytes, marrow edema, cysts, bone attrition, ligaments

Coronal T2 FSE (4.0 min)

TF osteophyes, collateral ligament


MRI Exams: Completion and Quality
MRI / pain eligible knees: 955 / 1364 (70%)

AAF (76%); WF (75%)

AAM (61%); WM (61%)

QA center (UCSF) review

unacceptable quality, repeat (2.7%)

patella incomplete

motion artifact

missing sequence

missing scans (0.2%)


Reason for incomplete MRI Men Women
Contraindication* (60%) 62% 58%

claustrophobia (30%)

35% 24%

metal fragment (17%)

27% 4%

TKR (21%)

13% 25%

recent surg, implants (18%)

18% 18%

obese (3%)

2% 4%
Refuse, no show (40%) 38% 42%


MRI Reading Methods
Sun workstation/MRVision
Scroll stacked slices
On-screen scoring
15-20 minutes per knee
Train with 'expert reader'
CD-ROM atlas


Knee MRI: Whole-Organ Knee Score
5 articular surface features


marrow edema

subarticular cysts

bone attrition



Knee MRI imaging feasible in large studies
Limitations include cost of acquisition and reading, contraindications
Existing studies may provide framework for validation of surrogate imaging markers and could be enhanced by additions to protocol


Dr. Mark Brezinski presented his research on optical coherence tomography in osteoarthritis.

General Discussion: There were discussions of MRI use in RA assessment. There is a need for agreement and reliability between techniques. There is very little information on the metrics of instruments used in these studies. We need a true quantification not a scoring system. Reliability is also difficult to determine. There is not much in the RA literature. Percent change and Pearson correlations are not good enough. Validity is difficult to assess without a "gold standard" for comparison. Because of changes with disease, time and treatment, it is difficult to determine reliability of methods. We need metrics data for measures to be used. Surrogate markers are difficult to validate and cannot be ultimately validated but only approximated. There are markers for various aspects of disease. Many are used as surrogates. It is imperative that a true surrogate meets 4 of 6 criteria. After further discussion, the group agreed that a version of the following methodology should be recommended to the OA Initiative Steering Committee in February.

LEVEL I (performed on all enrolled subjects)
Joints to be examined: bilateral knees
Methods to be used:

X-ray - Weight bearing; partially flexed knees

Non-fluoroscopic positioning with positioning device

Sufficient resolution to allow cancellous bone analysis

Evaluation of osteophytes

Joint space width measurements

Magnetic Resonance Imaging

30 minute protocol per knee (modification of basic Health ABC protocol including meniscus)

Cartilage volume, thickness, and score


Bone sclerosis(?)

Osteophytes and bone edema

Intervals and length of study

X-ray and MR imaging at

Baseline, 6, 12, 24, 36, and 48 months


LEVEL II (Requires statistical power)
Additional joints to be evaluated: bilateral hips; bilateral hands
Modifications/additions to methods:

X-ray - hips and hands

Compare non-fluoro positioning to fluoroscopic positioning

MR Imaging


Intervals and length of study

X-ray and MR imaging at baseline and 6 month intervals for 4 years


LEVEL III (Does not require statistical power)
Additional joints: shoulder, spine, and temporomandibular joint
Modifications/additions to methods
X-ray - spine, jaw, shoulder

Micro focal X-ray techniques (knee)

MR Imaging

Gadolinium uptake

Sodium imaging

Water diffusion

Extremity magnets versus standard

Assisted optical computerized tomography cartilage imaging
Analysis of bone architecture with MR, X-ray and/or DXA
New scintographic techniques for bone or cartilage imaging
Analysis of synovial perfusion with gadolinium
Intervals and length of study

X-ray and MR imaging at baseline and 3 month (or sooner) intervals for 4 years


Below is the background and charge to the meeting participants written by Dr. Charles Peterfy prior to the January 11, 2000 meeting.

    The Osteoarthritis Initiative Imaging Subcommittee Meeting:
    Imaging in Drug Development and Clinical Trials
    January 11, 2000

  1. Objectives

    The objective of this meeting is to develop draft proposals of imaging protocols for the Osteoarthritis (OA) Initiative steering group to consider at the Winter Meeting (February 28-29, 2000). These imaging protocols serve as the foundation for research studies that will advance knowledge about, and evaluate the use of, imaging modalities in the identification of anatomic markers that may serve as surrogate endpoints in OA clinical trials, and open doors of opportunity for developing disease-modifying therapies or interventions.

  2. Research Aims

    The proposed research aims to provide the technology and knowledge to:

    • expand the understanding of the pathophysiology and epidemiology of the disease

    • evaluate the efficacy and safety of new therapies; and

    • guide patient management in clinical practice.

  3. Each of these three research aims has slightly different priorities and, therefore, places different demands on biomarkers. The charge for our group is to identify the common ground among all of these, and design a protocol that will serve their overlapping needs rather than the interests of any one group. This will require agreement on the different roles and objectives of imaging markers in each of these contexts, and consensus on the correlated performance criteria by which different imaging markers could be compared. Once these are established, the most promising markers currently available could each be appraised in terms of these predetermined criteria. Harmonization of these criteria with those for evaluating biochemical markers would facilitate integration of these two complementary and potentially synergistic types of markers into single study protocols.

    This exercise will point out areas where further validation and characterization are most needed and will allow prioritization of the research questions that should be included in the final study that will come out of the Winter meeting. The imaging protocols proposed for that study should incorporate as many of these questions as possible. Additional sub-studies could also be proposed to complement and augment the information derived from the core study.

    The following is provided as a framework for discussions at the January meeting about the various roles and contexts of imaging in clinical trials and the correlated performance characteristics for surrogate endpoints. These points are also covered in the attached manuscript (Scratching the Surface: Articular Cartilage Disorders in the Knee) currently in preparation for MRI Clinics of North America.

  4. Roles of Imaging in Clinical Trials

    Imaging is needed in clinical trials for three fundamental purposes:

    • To select subjects most appropriate for study or treatment

    • To monitor disease progression and therapeutic response

    • To monitor complications of the disease or therapy

    The rank order of valued attributes for imaging markers in each of these rolls depends on the context in which it will be used (i.e., clinical service vs. clinical trials; internal decision making vs. definitive testing for regulatory approval.) The priorities of clinical trials are slightly different from those of clinical service. However, they share two fundamental objectives: 1) to help predict which patients will progress more rapidly or be most likely to respond to therapy (in clinical trials this allows enrichment of the study population to accelerate the trial; in clinical practice it helps identify patients most in need of aggressive therapy; and 2) to provide a valid appraisal of efficacy and safety of a particular treatment (in clinical trials to gain regulatory approval; in clinical practice to properly manage individual patient therapy).

    In addition to supporting definitive testing in Phase III of drug development for regulatory approval, imaging also facilitates internal decision making in Phase II about which compounds to prioritize. This includes proof-of-concept studies, dose-selection studies, patient-typing studies, etc. Here, an early readout is particularly valuable. Accordingly, the more responsive to the disease and / or therapy the particular morphological, compositional or process-related feature being measured is, the shorter the duration of monitoring needed to answer the question. There is a huge upside to rapid decision making in the competitive arena of drug development. So, the responsiveness of a marker at the biological level is a key performance metric.

    The statistical power with which differences in the rate of progression between two groups (e.g., treatment vs. placebo or active comparitor) can be resolved is determined by the measurement precision (i.e., reproducibility). There are three sources of variability that affect this parameter: 1) biological variation, 2) variation related to the acquisition method, and 3) variation related to the measurement algorithm. Biological variation can only be controlled through careful patient selection. The other two sources of variability, however, depend on a) how the imaging is conducted (multi-center vs. single-center; stability of the protocol; competence of the sites), and b) how the image analyses are performed (centralized readings with computer-assisted image analysis tools offer the greatest precision). High measurement precision can be traded for reduced patient numbers and/or shorter study duration. Only if two different markers are sufficiently similar on all of these levels do convenience and cost become deciding factors.

  5. Classification of Imaging Markers in OA


    1. Morphological (thickness, volume, geometry)

    2. Compositional (collagen, proteoglycan, water, fat)

    3. Process-related (perfusion, catabolism, synthesis, joint kinetics)


    1. Bone

    2. Cartilage

    3. Synovial (tissue, effusion)

    4. Meniscus, labrum, intra-articular disc, intervertebral disc

    5. Ligament (intra-articular, capsular)

    6. Tendons

    7. Muscles

    8. Bursae


    1. X-ray

    2. Computed tomography

    3. Dual-energy x-ray absorptiometry

    4. Magnetic resonance imaging

    5. Ultrasonography

    6. Scintigraphy

    7. Optical coherence tomography

    Measurement type

    1. Quantitative (continuous)

    2. Semi-quantitative / subjective (ordinal)

      1. Likert

      2. Complex


    1. Knee

    2. Hip

    3. Hand

    4. Shoulder

    5. Spine

    6. TMJ

  6. Performance Metrics for Imaging Endpoints

    1. Surrogate validity (link of biomarker to "true" outcome)
      Diagnostic power (sensitivity, specificity, area under ROC curve, etc.)
      Predictive power (for structural outcomes; for clinical outcomes)
      Importance depends on context
      internal decision-making
      clinical use
      regulatory approval

    2. Dynamic range
      proportion of true outcome captured by surrogate endpoint
      floor / ceiling effects?

    3. Responsiveness
      to disease and to therapy

    4. Measurement precision
      minimally detectable difference

    5. Convenience & Cost