OSTEOARTHRITIS and CARTILAGE

Transcription

1 Osteoarthritis and Cartilage (1993) 1, O 1993 Osteoarthritis Research Society / $08.00/0 OSTEOARTHRITIS and CARTILAGE Precision of joint space width measurement in knee osteoarthritis from digital image analysis of high definition macroradiographs BY JOHN A. LYNCH*, J. CHRISTOPHER BUCKLAND-WRIGHT* AND DIANA G. MACFARLANEt *Division of Anatomy and Cell Biology, United Medical and Dental Schools of Guy's and St Thomas's Hospitals, London and tkent and Sussex Hospital, Tunbridge Wells, Kent, U.K. Summary The precision of joint space width (JSW) measurements from plain film radiographs of the knee is limited by poor radio-anatomical positioning of the joint and/or insensitive methods of measurement. These limitations were overcome by establishing a precise radio-anatomical position for standing loaded and weight-bearing tunnel views of the knee, using the advantages of high definition macroradiography and a new computerized method for automatic JSW measurement from digitized macroradiographs of the knee. Reproducibility of JSW measurements was obtained from macroradiographs of knees of five post-mortem subjects and 12 patients with osteoarthritis (OA). JSW measurements were altered more by vertical than by horizontal misalignment of the X-ray beam relative to the joint space. In OA knees with medial compartment involvement, JSW measurements were more reproducible in the medial than lateral compartments in both radiographic views. In the medial compartment, the coefficient of variation for joint repositioning was 1% for minimum and 2% for average JSW, and for inter= and intraobserver errors, it was < 1% for both JSW measurements. The precision of this method will permit quantification of changes in JSW associated with disease progression and the treatment of OA. Key words: Knee, Osteoarthritis, Joint space width, Macroradiography, Measurement. Introduction ASSESSMENT of articular cartilage thickness is important for the evaluation not only of disease progression [1] in osteoarthritis (OA) but also for therapeutic response [2]. The most complete visualization of the cartilage is achieved with MR imaging [2, 3]. However, comparative studies of different imaging modalities [3, 4[ have shown that plain film radiography provides the most precise method of determining hyaline cartilage thickness from measurements of the joint space width (JSW). Two approaches have been used to measure OA knee JSW from plain film radiographs. First, joint position is very precisely controlled with the aid of custom built apparatus. This method allows standardization of the position of the knee and reproducible repositioning of the joint on successive examinations [3-5]. However, in these studies JSW measurements were carried out using a simple but imprecise method such as a ruler [4]. Conversely, the second approach has used an automatic system Submitted 11 June 1993; accepted 17 September Supported by a grant from Ciba Geigy and The University of London Central Research Fund. Address reprint requests and correspondence to: Dr J. C. Buckland-Wright, UMDS (Guy's Hospital), London Bridge, London SE1 9RT, U.K. for increased precision in JSW measurement, employing computerized analysis of digitally stored radiographic images [6, 7]. In these investigations there has been no specific control in standardizing the knee joint position, the absence of which can result in real errors for JSW measurement [8]. The work of our group, using quantitative microfocal radiography, is based on a precise and reproducible radiographic procedure forming the basis for accurate and reproducible measurements of radiographic features [9-11], This permits quantitation of the extent and progression in the hands of patients with OA [10-12] and rheumatoid arthritis [13-15]. In studying OA of the knee, JSW was assessed from~.weight bearing views taken in the standing and tunnel views, positions which most accurately reflect the articular cartilage thickness [16-19]. Here, we describe a new automated method for precisely measuring JSW in the tibio-femoral compartments, using computerized image analysis of digitally stored images of the knee macroradiographs. The precision of both the radiographic procedure and computerized method of JSW measurements was evaluated, together with that for inter- and intraobserver variability for JSW measurements. 209

2 210 Lynch et al.: Measuring OA in the knee Materials, patients and methods MATERIALS The post-mortem human knees of five women (mean age 65 years (range 61-69)) with no gross joint abnormality were obtained from the dissecting room. PATIENTS The macroradiographs of 12 patients with OA of the knee were examined. The patients comprised five men and seven women, with a mean ± S.D. age of 64.5+_3.7 years and a mean ±S.D. disease duration, based on knee pain, of 3.0 ± 2.9 years. On conventional radiography all had evidence of osteophytes and had either subchondral sclerosis and/or joint space narrowing. All patients had medial tibio-femoral compartment involvement. PREPARATION OF MACRORADIOGRAPHS Post-mortem knees. Stereopair macroradiographs at x 5 magnification [20, 21] were prepared of the knees with the joint positioned as if in the semiflexed standing view described below for patients. The knees were supported on an adjustable platform and held in position by sand bags. Difficulties in flexing the post-mortem knees beyond 30 led to distortions in the anatomical arrangement of the joints, preventing their use in repositioning tests for views of the knees in the modified tunnel position. Patient knees. Stereopair macroradiographs at 5 magnification [20, 21] were prepared of the knees in both the semi-flexed standing and weight bearing or modified tunnel views. The patient on the stereotaxic platform was placed close to the source (250 mm) and the whole platform displaced vertically by 8 mm between successive X-ray exposures [20]. Radiographic magnification for computing the size of X-ray features within the joint was determined by placing fine wire meshes on the knee's anterior and posterior surfaces, prior to radiography, and from measurements taken of the focus to knee and focus to film distance following radiography. Weight-bearing semi-flexed standing view. The semi-flexed standing view [22] was adopted in preference to the conventional radiographic view of the patient's knee in full extension [17], since in the former not only is the surface of the tibial plateau perpendicular to the X-ray film, but also the joint is closer to the normal anatomical standing position and to the region of major contact stresses in the tibio-femoral articulation [23]. Further, arthroscopic findings have revealed that in this region there is a greater incidence of cartilage damage [22] associated with osteoarthritic changes than at the sites associated with the knee in full extension. The macroradiographs of the standing view were taken postero-anteriorly so as to limit the degree of radiographic magnification of the patella and hence the extent to which it obscured the central region of the tibio-femoral joint. The center of the joint, defined by the joint space, was aligned with the center of the X-ray beam with the aid of a cross-optic laser and the vertical displacement of the patient relative to the X-ray source [20]. Each knee was flexed until the tibial plateau was horizontal relative to the floor, parallel to the central X-ray beam and perpendicular to the X-ray film. The degree of flexion varied between patients and ranged between 1 and 20. The Precise position of the knee was obtained visualiy with the help of image intensifier screening. With the heel fixed, the foot was internally or externally rotated until the tibial spines appeared centrally placed relative to the femoral notch [Fig.' l(a)]. Immediately following this procedure, the outline of the foot was drawn on a large sheet of paper placed on the platform of the stereotaxic apparatus, to facilitate joint repositioning at subsequent visits. Weight-bearing or modified tunnel view. The term 'modified' was applied to distinguish this weight bearing view of the knee from the usual unloaded tunnel view. The angle of flexion, used in this study, of 130 subtended by the tibia and femur at the joint, is similar to that described for the standard tunnel view [24]. The view was used to assess alteration of cartilage thickness over the popliteal surface of the condyles. The macroradiographs of the modified tunnel view were taken with the patient sitting on the edge of a stool with weight transmitted through the leg under X-ray examination. The joint was X-rayed in the antero-posterior position with the patient's foot placed in a slot on the patient table [20]. The angle of flexion for the knee of 130 at the tibio-femoral joint was checked with the aid of a perspex template. With the aid of the laser and image intensifier screening, the patient's position was adjusted to ensure that the tibial plateau was horizontal and perpendicular to the X-ray plate and that the tibial spines were centrally placed relative to the femoral notch [Fig. l(b)]. In the

3 Osteoarthritis and Cartilage Vol. 1 No FIG. 1. Macroradiograph of an OA knee joint in (a) the semi-flexed standing view and (b) of the same OA knee in the weight bearing or modified tunnel view. Note that the joint space is centrally placed on the film, the tibial plateau is level and the tibial spines centrally located relative to the intercondylar notch. The grids, placed on the anterior and posterior surfaces of the knee, were used to determine radiographic magnification (radiographic magnification x 5, reproduced at 0.95, i.e. with no magnification). tunnel view, the patient's foot placed within the slot on the stereotaxic platform permitted accurate repositioning of the joint at subsequent X-ray visits. IMAGE ANALYSIS Digitization of macroradiographs. The macroradiographs were digitized using a CCD camera (Videk Megaplus, Eastman Kodak Co, San Diego, CA) and a 55 mm Nikon Nikkor lens. The camera output was recorded via a Univision UPX1000 frame grabber board on a Univision UDC2600 display board in an IBM PC-AT computer. OPTIMAS software (Bioscan Inc, Edmonds, WA) was used to control the process and to write the data to disc. The camera digitized the X-ray image at a resolution of 1280 pixels (horizontally) by 1024 pixels (vertically) with 256 grey levels, linearly related to the optical density of the film. With both the optical magnification, i.e. the distance between the camera and film, and radiographic magnification known, the pixel size was set at (a) (b) 0.06 mm (e.g mm on the::i film when radiographic magnification was x 5). The digital images were written to a SUN SPARCstation IPX (SUN Microsystems, Mountain View, CA) via the Medical School ethernet using PC-NFS software. Programs written in the 'C' programming language and running on a SUN SPARCstation IPX were used to measure JSW. Joint space width measurement. Stereoscopic examination of the macroradiographs identified the following bony margins used for measuring the interbone distance separately in the standing and tunnel views of the knee: Femur: the distal convex margin of the condyle in the medial and lateral compartments respectively (Fig. 1). Tibia, medial compartment: a line extending from the near tibial spine to the medial or outer margin, across the center of the floor of the articular fossa in the mid-coronal plane of the joint. This line is defined by the superior margin of the bright radiodense band of the subchondral cortex, and appears below the anterior and posterior articular margins of the tibial plateau (Fig. 1). Tibia, lateral compartment: the proximal margin of the articular surface, defined by the superior margin of the bright radiodense band of the subchondral cortex extending from near the tibial spine to the lateral or outer margin (Fig. 1). Measurement. For each compartment, the algorithms used to outline the margins of the femoral condyle required the operator to use the 'mouse' to mark the inner (adjacent to the intercondylar notch) and outer edge of the condyle. The operator did not need to mark the edges precisely since the computer automatically corrected the position of the mark so that it coincided with the site of greatest edge-strength. The computer automatically traced the edge of the condyle, following the greatest edge-strength, from the position of the inner to the outer mark [Fig. 2(a)]. The algorithms developed to trace across the middle of the tibial articular surface followed a similar procedure in that the position of the inner (adjacent to the tibial spine) and outer edges of the tibial plateau were identified by a mark. However, in this instance, the algorithm followed the line of the brightest pixels, i.e. the superior margin of the subchondral cortex [Fig. 2(a)]. Using the mouse, the position of the tip of the medial tibial spine was also recorded as a reference point. From the computer-generated lines defining the joint space within a compartment, the symmetric axis of these lines [25] was used to calculate a

4 212 Lynch et al.: Measuring OA in the knee (a) (b) series of points across the joint space. These points lie equidistant between the proximal and distal articular margins of the joint space, and are the centers of the largest circle, the circumference of which does not cross the lines defining the joint margins [Fig. 2(b)-(d)]. The diameter of these circles of maximum radius defines the JSW, and corresponds to the articular cartilage thickness at each of these points. This data provides, for the standing and tunnel views of the knee, separate values for the minimum [Fig. 2(b)-(d)] and average JSW for the medial and lateral tibio-femoral compartments, respectively. On average, the JSW measurement procedure, for both compartments, takes 90 s per film. REPRODUCIBILITY TESTS :! In this study precision is defined as the reprodueibility of a measurement: i.e. how close is the agreement between different measurements of the same quantity [4]. POST~MORTEM KNEES (c) Joint repositioning. One post-mortem knee was repositioned and X-rayed 10 times. The radiographic magnification was obtained for each film. The films were digitized and the minimum and average joint space widths measured. Two principal sources of error occur in repositioning the knee joint, arising from variations in the vertical and rotational displacement of the knee relative to the central ray of the X-ray beam. (d) Vertical displacement of the knee relative to the X-ray source can result in the central region of the joint space lying either above or below the center of the X-ray beam. The effect alters the configuration of the joint space, as the tibial FIG. 2. (a) Computer screen showing a digitized macroradiograph of the medial tibio-femoral compartment of an OA knee with the margins of the JSW outlined; white: femoral condyle, black: floor of the middle of tibial plateau. The anterior and posterior margins of the tibial plateau are to be seen above the black line. (b) Computer image of the same knee as in (a). The diameter of the circles placed between the two lines is used by the computer to calculate the width of the joint space across the compartment. This plate only shows a small number of the circles used by the computer to calculate joint space width. The solid circle is the narrowest point between the two lines and gives the minimum JSW measurement. (c) This shows the application of the computer program to JSW measurement in the lateral tibio-femoral compartment. (d) Computerized JSW measurement in the medial compartment with marked loss of cartilage at the medial or outer margin; note also the large osteophytes on the tibial spines.

5 Osteoarthritis and Cartilage Vol. 1 No plateau no longer appears level. Macroradiographs of three post-mortem knee joints were obt~ained with the joint space centered on the central X-ray. Using an hydraulic platform, each knee was then displaced by 5 mm, 10 mm and 15 mm above and below the initial position. Macroradiographs were taken after each displacement. Rotation of the knee about the vertical axis occurred during screening when positioning the tibial spines centrally relative to the femoral notch. Macroradiographs of three post-mortem knee joints were obtained initially with the joint space centered on the central X-ray. Using a rotating platform, each knee was rotated internally and externally by 5, 10 and 15 with respect to the initial position, and re-x-rayed. For both the tests described above, the JSW was computed from each of the macroradiographs and the change in JSW relative to that of the initial position calculated. PATIENT KNEES Joint repositioning. Problems associated with the film processor resulted in five patients, one man and four women, having repeat macroradiographs of one knee within a 2-week period, comprising four knees radiographed in the standing and one in the tunnel view, respectively. The average and minimum JSW measurements in the medial and lateral compartments were computed from digitized macroradiographs of each knee, these were used to calculate the coefficient of variation for joint repositioning in patients. The macroradiographic films of the knees of seven patients, four men and three women, obtained from one visit were used for the following reproducibility tests. Intraobserver reproducibility. Joint space width was computed six times in both the medial and lateral tibio-femoral compartments, by one observer (J.A.L.), for each macroradiograph of the seven patients. For each compartment in the standing and tunnel views the mean, standard deviation and mean coefficient of variation were calculated for the average and minimum JSW, respectively, using the method described in Jonsson et al. [4]. Interobserver reproducibility. JSW was measured once in each compartment, for the standing and tunnel views, by six observers. The mean and standard deviation and mean coefficient of variation of each measurement were calculated for the average and minimum JSW. The Pearson correlation was used to assess the degree of association between measurements taken by the different observers. Results POST-MORTEM KNEES Joint repositioning. The automated measurement of JSW from digitized macroradiographs of one knee, repositioned 10 times, gave a mean (±S.D.) for the minimum and average JSW in the medial compartment of 4.12(_+0.13)mm and 4.84 (_+ 0.07) mm; and a mean (_ S.D.) for the minimum and average JSW in the lateral compartment of 6.00 (±0.22) mm and 7.37 (±0.07) mm, respectively. The coefficient of variation for repeat measures was large for the minimum JSW measurements in the medial (3.2%) and lateral (3.7~/o) compartments and less for the average JSW (1.4% and 1.0%, respectively). Vertical displacement. The effect on JSW measurement of positioning the center of the X-ray beam above or below the joint's center, defined by the joint space, is given in Table I. In the medial compartment both the minimum and average JSW narrow as the knee is displaced vertically above and the JSW widens as the knee is displaced below the center of the X-ray beam. In the lateral compartment, similar changes were observed only for the average JSW measurement. The rate of increase in the error of JSW measurement was approximately 0.02 mm per 1 mm of vertical displacement of the center of the knee away from the central ray of the X-ray beam. Rotation about a vertical axis. Changes in the degree of internal and external rotation of the knee joint did not systematically alter the JSW measured, as observed with vertical displacements. However, the average root mean square (RMS) of the change in JSW measured relative to the reference film, given in Table II, shows that, in the medial compartment, the random errors in the minimum and average JSW measurement increase with the greater angle of rotation. For minimum JSW the error increased at approximately 0.1 mm per 10 of rotation and for the average JSW measurement the error increased at approximately

6 214 Lynch et al.: Measuring OA in the knee Table I The mean difference in the minimum and average JSW measured (in mm) in the medial and lateral compartments, for varying amounts of vertical displacement of the knee above and below the plane of the center of the X-ray beam passing through the center of the joint space Vertical displacement relative to the center of the X-ray beam Below Above 10 mm 5 mm 5 mm 10 mm Minimum JSW Medial compartment (0.091) (0.080) (0.085) (0.152) Lateral compartment (0.102) (0.105) (0.551) (0.560) Average JSW Medial compartment (0.083) (0.074) (0.093),0.220 (0.099) Lateral compartment (0.190) (0.162) (0.191) (0.086) Table II The average (S.D.) root mean square (RMS) errors for the minimum and average JSW measured (in mm) in the medial and lateral compartments for the different degrees of internal and external rotation of the knee, away from their initial reference position _+ 5 rotation +_ 10 rotation + 15 rotation Minimum JSW Medial compartment Lateral compartment Average JSW Medial compartment Lateral compartment (0.129) (0.171) (0.107) (0.116) (0.268) (0.267) (0.071) (0.162) '(0.178) (0.058) (0.103) (0.130) 0.15 mm per 10 of rotation. In the lateral compartment, the degree of rotation of the knee did not affect the JSW measured (Table II). PATIENT KNEES Joint repositioning. In the medial compartment, the coefficient of variation for the minimum JSW measurement of 1.19% (Table III) was more precise than that for the average JSW measurement (2.08%). The coefficient of variation for both JSW measurements in the lateral compartment was large (Table III). Intra- and interobserver reproducibility for both the minimum and average JSW measurements in the medial compartment showed greater precision than in the lateral compartment (Tables IV and V). Not only were the standard deviations of the means within the dimensions of 1 pixel (0.06 ram), but also the coefficients of variation for the JSW measurements in the medial were better than those for the lateral compartment. Measurement of minimum JSW in the lateral compartment was least reproducible (Tables IV and V). Agreement between observers in JSW measurements was good, even between the observers with the greatest difference. In the medial and lateral compartments of the latter, the interobserver reproducibility had a correlation coefficient of r = and r = 0.965, respectively. Discussion The radiographic measurement of JSW can be considered, at best, to be an inference of the combined thickness of the cartilage and synovial fluid layers that separate the subchondral bone of the two opposing surfaces [26]. However, in a recent study designed to compare, in the OA knees (N = 20) of a group of patients, measurements of the interbone distance with that of the cartilage thickness visualized using double contrast arthrography, there was such a close agreement between the two measures that the observations [27] leave few grounds for concern regarding the reliability of this radiographic measurement. Additionally, the study revealed that JSW reliably measured cartilage thickness in the medial but not the lateral compartment of the knees with medial

7 Osteoarthritis and Cartilage Vol. 1 No Table III Reproducibility of knee joint repositioning for the minimum and average JSW of the medial and lateral compartments, measured (in mm) for macroradiographs of the joint repeated within a 2-week period Medial compartment Minimum JSW (mm) Average JSW (mm) Initial film Repeat film Difference Initial film Repeat film Difference Mean JSW (ram) Mean of differences (mm) Coefficient of variation (%) Lateral Compartment Minimum JSW (mm) Average JSW (mm) Initial film Repeat film Difference Initial film Repeat film Difference Mean JSW (mm) Mean of differences (mm) Coefficient of variation (%) compartment OA. The difference was due to a variable pattern of joint space narrowing in the lateral compartment. In the present investigation of patients who also had medial compartment OA, the JSW measurement results showed a small coefficient of variation for repeat JSW measurements in the medial but a large coefficient of variation in the lateral compartment, findings that Table IV Results of the intraobserver reproducibility tests showing the mean and S.D. (in mm) and coefficient of variation (CV) for repeat measurement of JSW from digitized macroradiographs by a single observer repeated six times Knee CV (%) Minimum JSW Medial compartment Lateral compartment Average JSW Medial compartment Lateral compartment Mean S.D Mean S.D Mean S.D Mean S.D

8 216 Lynch et al.: Measuring OA in the knee Table V Results of the interobserver reproducibility tests showing the mean and S.D. (in mm) and coefficient of variation (CV) for measurement of JSW from digitized macroradiographs by six observers Knee CV (%) Minimum JSW Medial compartment Lateral compartment Average JSW Medial compartment Lateral compartment Mean S.D Mean S.D Mean ~ ~ S.D Mean ,682 S.D A tend to confirm those of the arthrographic study. Although further investigations are required, the difference in precision for JSW measurements between the two compartments may be a result of alterations in joint alignment in knees with medial compartment OA [18]. The low coefficient of variation (< 1%) for repositioning patients' knees, in this study, demonstrated that it is possible for the difficulties associated with positioning this joint to be overcome. This was achieved by controlling precisely the anatomical position of the joint during radiography and by obtaining accurate and reproducible measurements of the features from the radiological images. In the former, procedures for defining the joints' radio-anatomical position in three planes resulted in virtually the same radiological landmarks being used to measure JSW, both between and within patients. Precision in repositioning the joint was helped considerably by the use of a foot map drawn at the time of the first X-ray visit. Conversely, in the post-mortem amputated knees, the poor reproducibility in joint repositioning occurred from an absence of any definite anatomical structure to facilitate accurate repositioning. Inaccuracies in positioning the knee joint did affect the precision of JSW measurement. When the two principal planes in which the joint was moved during positioning were assessed separately, it was found that rotation about the joint's central axis had a smaller effect than that due to the alignment of the center of the X-ray beam relative to the center of the joint space. JSW increased or decreased by 0.2 mm for every one centimeter displacement of the center of the X-ray beam below or above the plane passing ~hrough the center of the joint space. The extent of this discrepancy would increase proportionally with progressive joint space narrowing and with variations in joint repositioning on subsequent visits. Similar observations to the above were noted~by Fife and her colleagues [8], who found that JSW decreased when centering the tube at increasing distances above the patella's inferior margin. During this study we attempted to check our radiographic measurements of JSW obtained from post-mortem knees against the corresponding interbone distance measured from frozen serial sections of the same joint. Two factors prevented us from obtaining satisfactory results. The freezing process resulted in shrinkage of the articular cartilage owing to water crystallizing out into the space between aposing articular surfaces. From the sections it was not possible to define the boundary of the mineralized tissue which corresponded precisely to the joint space margin recorded in the macroradiograph. The latter presented a particular problem and is the subject of further study. As far as we can ascertain, the only other studies that have assessed the coefficient of variation for JSW measurements from repositioning patients' joints are those of Martel and colleagues [3, 4] using standard radiography of knees and hips and our macroradiographic studies of hand joints [9, 10]. In both studies, joint repositioning was controlled by the use of special apparatus. In the former, the results indicated that the large coefficient of variation of 6.5~/o can be attributed to the

9 Osteoarthritis and Cartilage Vol. 1 No combined limitations of a large X-ray focal spot size and variability, in manual measurements of JSW using a ruler. In the latter, the small-focalspot radiography characteristic of this method was described as being superior to other imaging modalities in reproducibility of JSW measurements due to improved visualization of joint space margins [4]. Thus, the indication was that the coefficient of variation of 3.7% obtained in this study can be attributed to the manual method of measurement employing the cross-wire cursor and digitizing tablet. The advantage of automated computer methods for JSW measurement over manual techniques are their independence from observer variability, as shown not only by the very low coefficient of variation for intra- and interobserver variability in the medial compartment JSW in this study, but also by the precision in remeasuring the same image as previously demonstrated by Dacre [7]. None the less, one of the principal limitations of these computational systems is that errors can arise in the margin selected by the computer for JSW measurement. The margin selection is based on changes in the grey level scale within the image. Variations in the mineral density of the tibial plateau result in part or all of the anterior or posterior margin of the plateau being selected for measurement as opposed to the central plane of the articular surface. In the present system, the use of an operator, who does not require any computational skill, overcomes this difficulty. Precision was assured by the operator identifying for the computer a point close to the correct margin, permitting the edge-following program of the computer to complete the procedure by tracing along the margin identified. Cartilage loss in the involved compartment in OA is not uniform. Joint space narrowing occurs predominantly towards the outer region of the compartment causing alterations in joint alignment leading to a tendency for the adjacent compartment to open up [18]. Thus any JSW measurement based on an assessment of the mean interbone distance across the compartment would average out the effect of localized cartilage loss and reduce the sensitivity of the technique for assessing disease progression. The advantage of the symmetric axis transform program described here is that it permitted both the average and minimum JSW to be measured within each compartment, thereby ensuring the precision of this method for quantifying cartilage loss at any region across the compartment. The accuracy [9, 10, 20] and precision of quantitative microfocal radiography supports the view that radiographic evaluation remains the most precise method, over other imaging modalities, for assessing progression of pathological changes in OA [3, 4, 8]. With the current interest in determining whether pharmacological agents retard the progression of articular cartilage breakdown in OA [8, 28, 29], it is anticipated that the present method may offer the opportunity for detecting and quantifying changes in JSW associated with the treatment of OA. Acknowledgments We would like to thank Charles Bird, Peter Liepins, Judy Vlahovic and Joy Pollard for carrying out the measurements employed in the calculations for interobserver variability, to Judy Vlahovic for her technical assistance and to Kevin Fitzpatrick and Sarah Smith for preparing the photographic plates. References 1. Altman R, Fries, Block DA et al. Radiological assessment of progression in osteoarthritis. Arthritis Rheum 1987;30: Adams ME, Wallace CJ. Quantitative imaging of osteoarthritis. Semin Arthritis Rheum 1991;20(Suppl 2): Martel W, Adler RS, Chan K, Niklason L, Helvie MA, Jonsson K. Overview: new methods in imaging OA. J Rheumatol 1991;18(Suppl 27): Jonsson K, Buckwalter K, Helvie H, Niklason L, Martel W. Precision of hyaline cartilage thickness measurements. Acta Radiol 1992;33: Siu D, Peng T, Cooke DV et al. A standardized technique for lower limb radiography, practice, applications and error analysis. Invest Radiol 1991;26: Dacre JE, Huskisson EC. The automatic assessment of knee radiographs in osteoarthritis using digital image analysis. Br J Rheumatol 1989;28: Dacre JE, Coppock JS, Herbert KE, Perret D, Huskisson EC. Development of a new radiographic scoring system using digital image analysis. Ann Rheum Dis 1989;48: Fife RS, Brandt KD, Braunstein EM et al. Relationship between arthroscopic evidence of cartilage damage and radiographic evidence of joint space narrowing in early osteoarthritis of the knee. Arthritis Rheum 1991;34: Buckland-Wright JC, Carmichael I, Walker SR. Quantitative microfocal radiography accurately detects joint changes in rheumatoid arthritis. Ann Rheum Dis 1986;45: Buckland-Wright JC, Macfarlane DG, Lynch JA, Clark B. Quantitative microfocal radiographic assessment of progression in osteoarthritis of the hand. Arthritis Rheum 1990;33: Buckland-Wright JC, Macfarlane DG, Lynch JA. Relationship between joint space width and subchondral sclerosis in the osteoarthritic hand: a

10 218 Lynch et al.: Measuring OA in the knee quantitative microfocal radiographic study. J Rheumatol 1992;19: Buckland-Wright JC, Macfarlane DG, Lynch JA. Osteophytes in the osteoarthritic hand: their incidence, size, distribution and progression. Ann Rheum Dis 1991;5{}: Buckland-Wright JC, Walker SR. Incidence and size of erosions in the wrist and hand of rheumatoid patients: a quantitative microfocal radiographic study. Ann Rheum Dis 1987;46: Buckland-Wright JC, Clarke GS, Walker SR. Erosion number and area progression in the wrist and hand of rheumatoid patients: a quantitative microfocal radiographic sttidy. Ann Rheum Dis 1989;48: Buckland-Wright JC, Clarke GS, Chikanza IC, Grahame R. Quantitative microfocal radiography detects changes in erosion area in patients with early rheumatoid arthritis treated with myocrisine. J Rheumatol 1993;20: Ahlback S. Osteoarthritis of the knee, a radiographic investigation. Acta Radiol 1968;277(Suppl): Leach RE, Gregg T, Siber FJ. Weight-bearing radiography in osteoarthritis of the knee. Radiology 1970;97: Thomas RH, Resnick D, Alazraki NP, Daniel D, Greenfield R. Compartmental evaluation of osteoarthritis of the knee. A comparative study of available diagnostic modalities. Radiology 1975;116: Resnick D, Vint V. The "tunnel" view in assessment of cartilage loss in osteoarthritis of the knee. Radiology 1980;137: Buckland-Wright JC. A new high definition microfocal x-ray unit. Br J Radiol 1989;62: Buckland-Wright JC, Bradshaw CR. Clinical applications of high definition microfocal radiography. Br J Radiol 1989;62: Messieh SS, Fowler PJ, Munro T. Anteroposterior radiographs of the osteoarthritic knee. J Bone Joint Surg [BR] 1990;72: Maquet P. Biomechanics of the knee. Berlin: Springer-Verlag Sartoris DJ, Resnick D. Plain film radiography: routine and specialized techniques and projections. In: Resnick D, Niwayama G, Eds. Diagnosis of bone and joint disorders, 2nd ed~ Philadelphia, PA: W.B. Saunders 1988: Blum H. Biological shape and visual science. J Theor Biol 1973;38: Adams ME. Microfocal radiography with macroradiographs in osteoarthritis~does it hit the spot and show the big picture? Editorial. J Rheumatol 1992;19: Buckland-Wright JC, Macfarlane DG, Jasani MK, Lynch JA. Joint space width measures cartilage thickness in knee OA: plaim film and double contrast macroradiographic investigation. Trans Orthop Res Soc 1993;18: Burkhardt D, Ghosh P. Laboratory evaluation of anti-arthritic drugs as potential chondroprotective agents. Semin Arthritis Rheum 1'987;17: Rejholec V. Long-term studies of antiosteoarthritic drugs: an assessment. Semin Arthritis Rheum 1987;17:35-53.