Study population
Subjects with knee OA were recruited by advertising through local newspapers and the Victorian branch of the Arthritis Foundation of Australia and in collaboration with general practitioners, rheumatologists, and orthopedic surgeons. The study was approved by the ethics committee of the Alfred and Caulfield Hospitals in Melbourne, Australia. All subjects gave informed written consent [16].
One hundred thirty-two subjects entered the study. Inclusion criteria were age older than 40 years, knee symptoms (at least one pain dimension of Western Ontario and McMaster University Osteoarthritis Index (WOMAC [17]) score >20% and osteophytes present), and radiographic knee OA (ACR radiographic and clinical criteria [18]). Subjects were excluded if any other form of arthritis was present, MRI was contradicted (for example, pacemaker, cerebral aneurysm clip, cochlear implant, presence of shrapnel in strategic locations, metal in the eye, and claustrophobia), inability to walk 50 feet without the use of assistive devices, hemiparesis of either lower limb, or planned total knee replacement.
Anthropometric and clinical data
Weight was measured to the nearest 0.1 kg (shoes and bulky clothing removed) by using a single pair of electronic scales. Height was measured to the nearest 0.1 cm (shoes removed) by using a stadiometer. Body mass index (BMI; weight/height2 (kg/m2)) was calculated. Function and pain were assessed with WOMAC (VAS, 10 cm) [17].
Radiograph
At baseline, each subject had a weight-bearing anteroposterior tibiofemoral radiograph of the symptomatic knee in full extension. Where both knees had OA and were symptomatic, the knee with least severe radiographic OA was used. These were independently scored by two trained observers who used a published atlas to classify disease in the tibiofemoral joint according to the Kellgren and Lawrence (K-L) scale. The radiologic features of tibiofemoral OA were graded in each compartment, on a 4-point scale (0 to 3) for individual features of osteophytes and joint space narrowing [19]. In the case of disagreement between observers, the films were reviewed by a third independent observer, and consensus values were used. Intraobserver reproducibility (κstatistic) for agreement on features of OA was 0.93 for osteophytes (grade 0, 1 versus 2, 3) and 0.93 for joint-space narrowing (grade 0, 1 versus 2, 3). Interobserver reproducibility was 0.86 for osteophytes and 0.85 for joint-space narrowing [20].
Magnetic resonance imaging
Each subject had an MRI performed on the symptomatic knee at baseline and ~2 years later. Knees were imaged in the sagittal plane on the same 1.5-T whole-body magnetic resonance unit (Signa Advantage HiSpeed; GE Medical Systems, Milwaukee, WI) by using a commercial receive-only extremity coil. The following sequence and parameters were used: a T1-weighted fat-suppressed 3D gradient recall acquisition in the steady state; flip angle, 55 degrees; repetition time, 58 msec; echo time, 12 msec; field of view, 16 cm; 60 partitions; 512 × 192 matrix; one acquisition time, 11 min 56 sec. Sagittal images were obtained at a partition thickness of 1.5 mm and an in-plane resolution of 0.31 × 0.83 mm (512 × 192 pixels).
Knee cartilage volume was determined by means of image processing on an independent work station by using the software program Osiris, as previously described [16, 20]. Two trained observers read each MRI. Each subject's baseline and follow-up MRI scans were scored unpaired and blinded to subject identification and timing of MRI. Their results were compared. If the results were within ± 20%, an average of the results was used. If they were outside this range, the measurements were repeated until the independent measures were within ± 20%, and the averages were used [16, 20]. Repeated measurements were made blind to the results of the comparison of the previous results. The coefficients of variation (CVs) for the measurements were 3.4% for the medial, 2.0% for the lateral, and 2.6% for the total tibial cartilage volume [16]. Tibial plateau area was determined by creating an isotropic volume from the three input images closest to the knee joint, which were reformatted in the axial plane. The area was directly measured from these images. The CVs for the medial and lateral tibial plateau area were 2.3% and 2.4%, respectively [16, 20].
A subchondral bone cyst was defined as a well-demarcated hypersignal, whereas a BML was an ill-defined hypersignal. The assessments of subchondral bone cysts and BMLs were performed on the MRI slice that yielded the greatest lesion size. The intensity and extent of cysts and BMLs were assessed in the medial and lateral tibiofemoral compartments and were graded as 0, absence of lesion; 1, mild to moderate lesion; and 2, severe (large) lesion. A reliability study done by using a two-reader consensus measure of a specific lesion size twice at a 6-week interval showed an r = 0.96, p < 0.0001 for subchondral bone cysts and r = 0.80, p < 0.001 for BMLs (test-retest Spearman correlation) [6]. The medial and lateral cyst and BML scores were each calculated as a sum of the scores for the tibial, femoral, and femoral posterior sites (scores 0 to 6). As a low prevalence of subjects was found with cyst scores >3 for the medial and >1 for the lateral compartment, we collapsed the scores to give a range of 0 to 3 for the medial and 0 to 1 for the lateral compartment.
Identification of knee replacement
At year 4, all subjects were contacted and asked whether they had undergone a knee replacement because of OA of the same knee in which they had a baseline MRI. This was confirmed by contacting the treating physician in all cases.
Statistical analysis
Descriptive statistics for characteristics of the subjects were tabulated. Annual percentage change in cartilage volume was calculated by cartilage change (follow-up cartilage volume subtracted from initial cartilage volume) divided by initial cartilage volume and time between MRIs. Outcome variables (baseline tibial cartilage volume and annual percentage change in tibial cartilage volume) were initially assessed for normality and were found to approximate normal distribution. Estimated marginal means was used to explore the cross-sectional relationship between subchondral bone cysts and tibial cartilage volume at baseline, and longitudinally, the relationship between baseline subchondral bone cysts and annual percentage tibial cartilage volume loss. Logistic regression was used to examine the relationship between baseline subchondral bone cysts and risk of knee-joint replacement over a 4-year period. All analyses were performed by using the SPSS statistical package (version 16.0.0; SPSS, Cary, NC), with a P value < 0.05 considered statistically significant.