Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptNIH Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Arthritis Care Res (Hoboken). Author manuscript; available in PMC Jul 1, 2011.
Published in final edited form as:
PMCID: PMC2937275
Comparison of radiographic joint space width to MRI cartilage morphometry: analysis of longitudinal data from the Osteoarthritis Initiative (OAI)
Jeffrey Duryea,1 Gesa Neumann,1 Jingbo Niu,2 Saara Totterman,3 Jose Tamez,3 Christine Dabrowski,4 Marie-Pierre Hellio Le Graverand,5 Monica Luchi,6 Chan R. Beals,7 and David J. Hunter8
1Brigham and Women’s Hospital / Harvard Medical School, Boston, MA
2Boston University, Boston MA
3VirtualScopics Inc., Rochester, NY
4GlaxoSmithKline Pharmaceutical Company, King of Prussia, PA
5Pfizer Global Research and Development, Ann Arbor, MI.
6Novartis Inc., East Hanover, NJ
7Merck Research Laboratories, Rahway, NJ
8New England Baptist Hospital, Boston, MA
Reprints and Correspondence Jeff Duryea, Ph.D. Assistant Professor Radiology Brigham and Women’s Hospital / Harvard Medical School 75 Francis Street, Boston, MA 02115, USA phone +1-617-525-3332; fax +1-617-525-3330 ; jduryea/at/
Magnetic resonance imaging (MRI) and radiography are established imaging modalities for the assessment of knee osteoarthritis (OA). The objective of our study is to compare the responsiveness of radiographic joint space width (JSW) to MRI derived measures of cartilage morphometry for OA progression in participants from the Osteoarthritis Initiative (OAI).
This study examined the baseline and 12 month visits of a subset of 150 subjects from the OAI. Measurement of radiographic JSW was facilitated by the use of automated software that delineated the femoral and tibial margins of the joint. Measures of medial compartment minimum JSW and JSW at fixed locations were compared to cartilage morphometry measures derived from MRI. The results were stratified by Kellgren and Lawrence (KL) grade and by tibio-femoral anatomical angle. In order to examine the relative responsiveness of various techniques, we calculated the standardized response mean (SRM) between the two visits.
The SRM value for radiographic JSW measured at the optimal location was −0.32 compared to −0.39 for the most responsive MRI measure. For a subgroup with KL = 2 or 3, the most responsive SRM values were −0.34 for radiographic JSW and −0.42 MRI.
Our study demonstrates that new measures using a software analysis of digital knee radiographic images are comparable to MRI in detecting OA progression, and potentially superior when considering the cost effectiveness of the two imaging modalities.
Radiography and magnetic resonance imaging (MRI) are the imaging modalities most frequently used to monitor OA progression(1). MRI offers visualization of cartilage and other soft tissues and is a three-dimensional modality. However, radiography is considerably less expensive than MRI due to both the acquisition and reader costs. Radiography provides a proven low-cost method to monitor OA progression and continues to be more widely accepted than MRI to measure structural change for clinical trials of OA(2). Radiography is used in several large OA studies (3, 4) including the Osteoarthritis Initiative (OAI), for which approximately 48,000 radiographic images of the knee will be acquired(5). The spatial resolution of radiography is substantially better than MRI and radiography is ideal for visualizing bone margins.
Radiography can be used to quantify cartilage thinning indirectly by measuring the narrowing of the joint space width (JSW) between the adjacent bones of the knee. Minimum joint space width (mJSW) between the projected femur and tibia margins on a knee radiograph is the currently accepted metric to assess OA longitudinally (6). mJSW can be quantified by visually determining the location of the minimum distance while viewing the film and making the measurement with a hand-held graduated lens.
Software methods to detect change in knee OA have been developed for both radiography(7-10)and MRI (11-15). With these techniques, the computer generally determines the location and size of the structures that change with the disease such as the joint margins for radiography and the outline of the articular cartilage in the case of MRI.
Measurement of radiographic mJSW by software tools is usually accomplished by delineating the margins of the opposing bones in the joint. The software then determines the minimum distance between the femur and tibia margins in each compartment. Measurement of mJSW, however, probes only a single location along the joint interface and does not take full advantage of the delineated joint margins provided by the software.
To investigate the potential of more general measures of JSW, a previous study examined the reproducibility of JSW at fixed locations along the joint using digitized duplicate knee radiographs (16). We denote this measure as JSW(x). This study demonstrated an improvement of JSW(x) over mJSW(17). A second study using longitudinal (baseline and 36 month follow-up) OA subjects established an improved responsiveness of JSW(x) over mJSW for measuring OA progression. The goal of our current work is to examine radiographic JSW changes over a one year time period and make a direct comparison between radiographic JSW and MRI by comparing the responsiveness of the two methods on the same sub-cohort of subjects. Additionally we examine radiographic JSW at fixed locations in the lateral compartment and examine the different measures as a function of the tibio-femoral anatomical angle to study the effect of knee alignment.
Study Subjects
Serial bilateral posteroanterior (PA) conventional radiographs and MRI data were obtained from the baseline and 12 month visits of subjects participating in the OAI, an ongoing National Institutes of Health and industry sponsored study of OA (5). Data were selected from the Progression subcohort, which includes individuals who have frequent symptoms of knee OA and pain, aching, or stiffness on most days of a month during the past year in the same knee. The sample used for this study consisted of 160 subjects from clinical Data Set 0.1.1 and Image Releases 0.B.1 and 1.B.1. These data are available online at
Bilateral MRI exams from 160 participants were provided by OAI but only one knee from 150 patients (one knee per subject) was identified for analysis. The rationale for the reducing the sample from 160 to 150 was that the budget for processing the images was limited and in addition we wanted to optimize the use of subjects more likely to progress. The selection of the index knee for this analysis was based on the presence of both symptoms (frequent knee pain) and radiographic OA (ROA) in the same knee. 100 patients had unilateral symptomatic ROA, and this knee was chosen for analysis, regardless of radiographic severity. For the remaining participants with bilateral symptomatic ROA one knee was selected, favoring the knee with moderate disease more likely to undergo disease progression (for further details on knee selection see prior publication(18). Subjects were enrolled at the five OAI clinical centers based on the enrolling center’s radiographic assessment of findings of OA in at least one knee.
Bilateral PA view radiography was performed using a SynaFlexer™ frame (Synarc, Inc., San Francisco, CA) to position the subject’s feet reproducibly. The beam and subjects were positioned according to a fixed-flexion protocol which required the x-ray beam to be aimed at a 10 degree caudal angle(19, 20).
Baseline and follow-up radiographs of the 150 subjects’ indexed knees were read independently by two study readers, one a bone and joint radiologist, and the other a rheumatologist (DH). Readers evaluated the KL grade on a 0-4(21) scale as well as individual radiographic features, i.e. osteophytes; for the KL grade we used adjudicated readings that were determined by a consensus of the readers.
DH also assessed the anatomic axis at the time of these readings. Digital imaging software (eFilm Workstation (Version 2.0.0) software) was used to compose reference lines and calculate these measures. The anatomic axis was defined as the angle formed by the intersection of two lines originating from points bisecting the femur and tibia and converging at the center of the tibial spine tips, consistent with the methods described by Kraus et al where varus is negative(22). The origin of these lines was 10 cm from the knee-joint surfaces when included in the field of view on the radiograph.
As part of the OAI data release, the radiographic images are provided in digital format for viewing on computer monitors or to input to software analysis applications. For our study each indexed knee image was analyzed by custom software developed in our laboratory to delineate the femoral and tibial joint margins. In practice a reader identified and corrected the computer-determined contours in cases where the software failed to correctly delineate the true joint margins. mJSW was defined as the minimum distance between the two contours in the medial compartment. The software also provided a measure of the JSW at fixed locations along the joint, denoted as JSW(x), by finding positions along the joint interface on a coordinate system defined in Figure 1. The x-axis (Line A), defined as the line tangent to both femoral condyles, was placed automatically by the software. The x variable is a dimensionless quantity and represents the position of the JSW(x) measurement along the projected surface of the joint. The y-axis (Line B) was placed manually as a line perpendicular to the x-axis and tangent to the greatest prominence of the medial epicondyle. The line x = 1 (Line C) was defined as the tangent to the greatest prominence of the lateral epicondyle of the knee. The software displayed cropped images of the epicondyles for both visits simultaneously so that the reader could verify consistent landmark placement for both baseline and follow-up images.
Figure 1
Figure 1
Landmarks and definition of coordinate system.
All images of the knee joint were placed in a consistent orientation with the medial compartment on the left (x < 0.5) and the lateral compartment on the right (x > 0.5). Measurements were made of the medial compartment mJSW and JSW(x) at 0.025 intervals for x = 0.15 - 0.30 (medial compartment) and x = 0.70 - 0.90 (lateral compartment). Due to subluxation for several knees, it was impossible to produce a consistent measurement for all knees at x < 0.15.
The subjects were scanned on a 3 Telsa Siemens Magnetom (Erlangen, Germany) MRI scanner with a quadrature transmit-receive knee coil (USA Instruments, Aurora, OH). The study used a sagittal 3D DESS pulse sequence with a slice thickness of 0.7 mm, 16.3ms TR, 4.7ms TE, 25° FA, 160 slices, 140mm Field of View (FOV), 384×307 matrix, and an in-plane resolution 0.37mm ×0.46mm (interpolated to an isotropic in-plane resolution of 0.37mm×0.37mm). More details of the MRI protocol are given in a separate publication(23).
Segmentation of the cartilage from the femur, tibial plateaus, and patella was performed using a semi-automated software technique (13). This method quantified the total cartilage volume (V), and the cartilage volume normalized to the surface area as a measure of thickness (Th). Measurement was made for the total femur, lateral and medial tibial plateaus, and patella as well as the following subregions: central lateral compartment femur (LF), central medial compartment femur (MF), central lateral compartment tibia (LT), and central medial compartment tibia (MT). The complete results from this study are provided in a separate publication (23), but a subset of the most response measures is also included in this report so that a direct comparison can be made between radiographic JSW and MRI cartilage morphometry. Specifically, we have chosen to provide measures for the central portion of the medial and lateral femur and tibial plateau (MF, LF, MT, and LT) since these are most likely to correlate with radiographic JSW and are the most appropriate measures for comparison.
Statistical Analysis
All readings were performed viewing baseline and follow-up images paired but blinded as to time point order. To compare the responsiveness of the different variables we used the standardized response mean (SRM) defined as mean change divided by the standard deviation of the change. The data were also stratified by K&L grade and anatomical angle.
Table 1 gives the characteristics of the 150 subjects. Table 2 provides a breakdown of the number of knees by KL grade and knee alignment. The neutral full-limb (mechanical) alignment in those without OA is approximately 1.0 degree varus and as a result by convention neutral is typically categorized as 0-2 degrees varus(24). In delineating the categorization of anatomic axis alignment for this study, we used the 4 degree offset from mechanical axis advocated by Kraus et al. (22) Therefore, we define valgus alignment as AA < −4°, neutral alignment as −4° ≥ AA ≥ −2°, and varus as AA > −2°.
Table 1
Table 1
Descriptive characteristics of study sample (n= 150)
Table 2
Table 2
Breakdown of the number of knees by KL grade and knee alignment. We define valgus alignment as AA < −4°, neutral alignment as −4° ≥ AA ≥ −2°, and varus as AA > −2°. (more ...)
Table 3 provides results of the most responsive JSW and cartilage morphometry measures, for the entire 150 subject data set and for the subgroup of subjects with KL = 2 or 3. Location specific JSW is more responsive that mJSW with SRM = −0.32 for the optimal location at x = 0.275 for all subjects. The most responsive location for the subset of KL = 2 or 3 is at x = 0.250 (SRM = −0.34) as compared to SRM = −0.18 for mJSW. The SRM values are, in general, higher for the medial versus lateral compartments, but there is evidence of improved lateral compartment responsiveness for the KL = 2 or 3 group. The most responsive MRI morphometry measure is V (MF) with SRM = −0.39 (all data) and SRM = 0.42 (KL = 2 or 3). To investigate the correlation between the radiographic and MRI measures, We also performed a linear regression fit between the change in cartilage volume and the change in mJSW and JSW (x = 0.275). The R values were 0.28 (p < 0.0001) for both mJSW and JSW(x=0.275) Figure 2 provides the SRM results for three groups stratified by anatomical angle, and for the entire data set An examination of Figure 2 reveals several interesting features. The pattern of increased medial compartment responsiveness in the center of the knee (locations closer to the tibial spines) is more pronounced for the varus group while the opposite may be true for the valgus knees. In the lateral compartment, one can observe a similar trend of increased central joint responsiveness for the valgus group and there evidence of pseudo-widening (positive SRM values) with the varus knees in the outer part of the lateral compartment.
Table 3
Table 3
Baseline, change, and SRM values for all 150 subjects and a subset of KL = 2 or 3 (N = 116).
Figure 2
Figure 2
Graphs of SRM versus x stratified by anatomical angle (AA) for the (a) medial and (b) lateral compartments. We define valgus alignment as AA < −4, neutral alignment as AA = −4 or −2, and varus as AA > −2. (more ...)
Our data confirm results from a previous study that examined longitudinal radiographs with a 36 month follow-up time and also used the Synaflexer positioning frame and the fixed flexion protocol(17). In that study, JSW(x) also showed greater sensitivity to detect change than mJSW in some regions and there was a trend toward improved disease sensitivity in the more central portion of the joint and for more diseased knees. In our current study, we examine a one year follow-up from a different cohort of OA subjects and provide a direct comparison between radiographic JSW and cartilage morphometry measures from MRI. Table 2 demonstrates that the measure of JSW at x = 0.275 (SRM = −0.32) performs better than all cartilage morphometry measures for all subjects with the exception of Th(MF) (SRM = −0.34) and V(MF) (SRM = −0.39). For KL = 2 and 3 subjects, JSW at x = 0.250 (SRM = −0.34) performs better than all cartilage morphometry measures with the exception of V(MF) (SRM = −0.42).
Our results suggest that radiography may be a valuable modality for clinical trials of knee OA given the general availability of radiology facilities for image acquisition, and the expense required for MRI image acquisition and evaluation. Our study presents the results of a head-to-head comparison between quantitative MRI and radiography measures of OA progression. For KL = 2 or 3, a population that is likely to be used for clinical trials, the difference between the most responsive measures (0.32 and 0.42) is modest. These data imply that an equally powered radiography study would require approximately 1.5 times as many subjects as for a study that used MRI since the minimum sample size to see a statistically significant effect is inversely related to the square of the SRM(25). The expense of additional subjects would most likely offset by the decrease in imaging costs.
Although located in a generally less weight bearing portion of the joint, increased performance for the more central portion (higher x for the medial compartment) may be explained by the difficultly associated with visualizing the margins in the more damaged portion of the joint. While the best region for longitudinal assessment may not be located at the exact site of the joint damage, JSW(x) at the optimal location may provide an indirect, but more precise, measure of the joint damage that occurs less centrally.
The measures of lateral compartment JSW progression suggest that different JSW metrics could be used to monitor progression depending on the knee alignment to maximize the responsiveness. For example, it may be optimal to use JSW at x = 0.250 for varus knees while JSW at x = 0.725 (lateral compartment) may be the preferred location for valgus knees. Both these sub-groups show improved responsiveness over measures of JSW for the entire set and for all MRI measures. It should also be noted that it was sometimes necessary to use the tibial rims for the delineation of the lateral compartment tibial margin in cases where the plateau could not be visualized.
We found a relatively weak correlation between the radiographic and MRI metrics. While a stronger correlation might indicate an improved understanding of knee OA progression, the results also suggest that each method potentially probes independent OA changes. There are unknown factors affecting structural progression that each method fails to detect. Once these effects are better understood, there is reason to hope that both methods can be further improved. One possible explanation for the low correlation may be that, while radiography was performed with weight-bearing joints, the MRI acquisitions are collected with the subject supine. In addition, the leg is fully extended during the MRI scan while the radiograph is acquired with the knee flexed. A study using an open MRI system where individuals can be imaged while standing would probe this question.
This study has several limitations. While the total number of subjects is substantial, the number of individuals in the valgus group is only 16. Future studies should use larger numbers of valgus aligned knees to confirm these results. We made no attempt to correct for a any possible rotation of the femur which could affect the consistency of the coordinate system; in theory, the use of the positioning frame should minimize this effect. Our study used images from the OAI that were obtained using a fixed-flexion protocol (20). Previous work has demonstrated that different knee positioning protocols may produce dissimilar results(26). One study which compared the fixed flexion to a fluoroscopically guided protocol using the same subjects found increased reproducibility and responsiveness of mJSW using the Lyon schuss protocol and a similar software approach(26). We cannot necessarily generalize our conclusion for other protocols aimed at aligning the tibial rims on the radiograph such as fluoroscopically guided knee radiography(6).
We have performed a longitudinal study of 150 subjects comparing the ability of software-based methods of MRI and radiography to detect changes due to knee OA. Our results demonstrate that JSW(x) measures on radiography compares very favorably with the responsiveness of MRI cartilage morphometry measurement and suggest that knee radiography will continue to play an important role in studies of knee OA.
We would like to thank the Principal Investigators (Michael Nevitt, Kent Kwoh, Charles B. Eaton, Rebecca Jackson, Marc Hochberg, Joan Bathon), Co-Investigators and staff of the OAI. We would also like to acknowledge the following persons who contributed to this work: Piran Aliabadi (performed Kellgren and Lawrence scoring of the knee radiographs) and David Felson (chaired the x-ray adjudication sessions). We would also like to acknowledge the input of Merck statisticians Yevgen Tymofyeyev and Amy Ko, David Raunig from Pfizer, and Randall Smith from GlaxoSmithKline.
The OAI is a public-private partnership comprised of five contracts (N01-AR-2-2258; N01-AR-2-2259; N01-AR-2-2260; N01-AR-2-2261; N01-AR-2-2262) funded by the National Institutes of Health, a branch of the Department of Health and Human Services, and conducted by the OAI Study Investigators. Private funding partners include Merck Research Laboratories; Novartis Pharmaceuticals Corporation, GlaxoSmithKline; and Pfizer, Inc. Private sector funding for the OAI is managed by the Foundation for the National Institutes of Health. This manuscript has received the approval of the OAI Publications Committee based on a review of its scientific content and data interpretation.
1. Song IH, Althoff CE, Hermann KG, Scheel AK, Knetsch T, Schoenharting M, et al. Knee Osteoarthritis Efficacy of a new method of contrast-enhanced musculoskeletal ultrasonography in detection of synovitis in patients with knee osteoarthritis in comparison with magnetic resonance imaging. Ann Rheum Dis. 2007 [PubMed]
2. Altman R, Brandt K, Hochberg M, Moskowitz R, Bellamy N, Bloch DA, et al. Design and conduct of clinical trials in patients with osteoarthritis: recommendations from a task force of the Osteoarthritis Research Society. Results from a workshop; Osteoarthritis Cartilage; 1996. pp. 217–43. [PubMed]
3. Visser M, Newman AB, Nevitt MC, Kritchevsky SB, Stamm EB, Goodpaster BH, et al. Reexamining the sarcopenia hypothesis. Muscle mass versus muscle strength. Health, Aging, and Body Composition Study Research Group. Ann N Y Acad Sci. 2000;904:456–61. [PubMed]
4. Felson DT, Niu J, Guermazi A, Roemer F, Aliabadi P, Clancy M, et al. Correlation of the development of knee pain with enlarging bone marrow lesions on magnetic resonance imaging. Arthritis Rheum. 2007;56(9):2986–92. [PubMed]
5. Osteoarthritis Initiative. 2007. (checked 9/2007)
6. Buckland-Wright C. Protocols for precise radio-anatomical positioning of the tibiofemoral and patellofemoral compartments of the knee. Osteoarthritis Cartilage. 1995;(3 Suppl A):71–80. [PubMed]
7. Dacre JE, Coppock JS, Herbert KE, Perrett D, Huskisson EC. Development of a new radiographic scoring system using digital image analysis. Annals of the Rheumatic Diseases. 1989;48(3):194–200. [PMC free article] [PubMed]
8. Lynch JA, Buckland-Wright C, Macfarlane DG. Precision of joint space width measurment in knee osteoarthritis from digital image analysis of high definition macroradiographs. Osteoarthritis and Cartilage. 1993;1:209–18. [PubMed]
9. Duryea J, Li J, Peterfy CG, Gordon C, Genant HK. Trainable rule-based algorithm for the measurement of joint space width in digital radiographic images of the knee. Med Phys. 2000;27(3):580–91. [PubMed]
10. Vignon E, Piperno M, Le Graverand MP, Mazzuca SA, Brandt KD, Mathieu P, et al. Measurement of radiographic joint space width in the tibiofemoral compartment of the osteoarthritic knee: comparison of standing anteroposterior and Lyon schuss views. Arthritis Rheum. 2003;48(2):378–84. [PubMed]
11. Kshirsagar AA, Watson PJ, Tyler JA, Hall LD. Measurement of localized cartilage volume and thickness of human knee joints by computer analysis of three-dimensional magnetic resonance images. Invest Radiol. 1998;33(5):289–99. [PubMed]
12. Peterfy CG, van Dijke CF, Janzen DL, Gluer CC, Namba R, Majumdar S, et al. Quantification of articular cartilage in the knee with pulsed saturation transfer subtraction and fat-suppressed MR imaging: optimization and validation. Radiology. 1994;192(2):485–91. [PubMed]
13. Tamez-Pena J, Barbu-McInnis M, Totterman S. Unsupervised definition of the tibia-femoral joint regions of the human knee and its applications to cartilage analysis. Proceedings of SPIE. 2006;6144:1465–1475.
14. Eckstein F, Hudelmaier M, Wirth W, Kiefer B, Jackson R, Yu J, et al. Double echo steady state magnetic resonance imaging of knee articular cartilage at 3 Tesla: a pilot study for the Osteoarthritis Initiative. Ann Rheum Dis. 2006;65(4):433–41. [PMC free article] [PubMed]
15. Duryea J, Neumann G, Brem MH, Koh W, Noorbakhsh F, Jackson RD, et al. Novel fast semi-automated software to segment cartilage for knee MR acquisitions. Osteoarthritis Cartilage. 2006;15:487–92. [PubMed]
16. Duryea J, Zaim S, Genant HK. New radiographic-based outcome measures for osteoarthritis of the knee. Osteoarthritis and Cartilage. 2003;11(2):102–10. [PubMed]
17. Neumann G, Hunter D, Nevitt M, Chibnik LB, Kwoh K, Chen H, et al. Location specific radiographic joint space width for osteoarthritis progression. Osteoarthritis Cartilage. 2008 [PMC free article] [PubMed]
18. Hunter DJ, Niu J, Zhang Y, Totterman S, Tamez J, Dabrowski C, et al. Change in cartilage morphometry: a sample of the progression cohort of the Osteoarthritis Initiative. Ann Rheum Dis. 2009;68(3):349–56. [PMC free article] [PubMed]
19. Peterfy C, Li J, Zaim S, Duryea J, Lynch JA, Miaux Y, et al. Comparison of Fixed-Flexion Positioning with Fluoroscopic Semi-Flexed Positioning for Quantifying Radiographic Joint-Space Width in the Knee: Test-Retest Reproducibility. Skeletal Radiology. 2003;3:128–32. [PubMed]
20. Kothari M, Guermazi A, Von Ingersleben G, Miaux Y, Sieffert M, Block JE, et al. Fixed-flexion radiography of the knee provides reproducible joint space width measurements in osteoarthritis. Eur Radiol. 2004;14(9):1568–73. [PubMed]
21. Kellgren JH, Lawrence JS. Radiological assessment of osteoarthritis. Ann.Rheum. Dis. 1957;16(11):494–501. [PMC free article] [PubMed]
22. Kraus VB, Vail TP, Worrell T, McDaniel G. A comparative assessment of alignment angle of the knee by radiographic and physical examination methods. Arthritis Rheum. 2005;52(6):1730–5. [PubMed]
23. Hunter DJ, Jingbo Niu J, Zhang Y, Totterman S, Tamez J, Dabrowski C, et al. Change in Cartilage Morphometry: a sample of the progression cohort of the Osteoarthritis Initiative. Submitted to Annals of Rheumatic Diseases. 2007 [PMC free article] [PubMed]
24. Cooke TD, Sled EA, Scudamore RA. Frontal plane knee alignment: a call for standardized measurement. J Rheumatol. 2007;34(9):1796–801. [PubMed]
25. Snedecor G, Cochran W, Cox D. Statistical Methods. 8th edition The Iowa State University Press; 1989.
26. Hellio Le Graverand MP, Vignon EP, Brandt KD, Mazzuca SA, Piperno M, Buck R, et al. Head-to-head comparison of the Lyon schuss and fixed flexion radiographic techniques. Long-term reproducibility in normal knees and sensitivity to change in osteoarthritic knees. Ann Rheum Dis. 2008 [PubMed]