An age- and gender-stratified subsample (OAI public-use data sets 0.1.1, 0.B.1 and 1.B.1) of the OAI progression subcohort was studied. Patients had been recruited at four clinical sites: the University of Maryland School of Medicine (Baltimore), the Ohio State University (Columbus), the University of Pittsburgh and the Memorial Hospital of Rhode Island (Pawtucket). The subsample included 79 women with a mean (SD) age of 60.3 (9.5) years, weight 79.6 (15.6) kg, BMI 30.3 (5.5); and 77 men, age 62.0 (10.2) years, weight 94.6 (13.6) kg, BMI 30.1 (3.7). The BMI for 18 participants was <25, for 59 it was 25–30, for 58 it was 30–35 and for 21 it was >35. The participants included a diversity of ethnic minorities, were 45–79 years old, and had both frequent knee symptoms (pain, aching or stiffness on most days of a month in the past year) and radiographic OA (definite osteophytes in the postero-anterior fixed flexion radiographs15 16
) in at least one of their knees, based upon initial radiograph reading at the clinical sites. In this analysis, however, we used the results of independent readings by a musculoskeletal radiologist and a rheumatologist at Boston University for Kellgren–Lawrence (K-L) grade, which in the case of discrepancy were adjudicated by consensus with a third reader.
The MRI sequence used to quantify cartilage morphology (see below) was only available in the right knees, whereas some participants displayed symptoms and radiographic OA in their left knee. Therefore, and because the adjudicated central radiographic readings may have differed from the initial screening readings at the imaging sites, not all knees analysed had symptomatic knee OA. Of the 156 knees analysed, 108 had frequent symptoms, 110 had definite radiographic OA (56 knees with K-L grade = 2, 47 with K-L grade = 3, and 7 with K-L grade = 4), and 87 had symptomatic and radiographic OA. Seventeen knees showed K-L grade = 0, and 29 K-L grade = 1. 87 knees had both frequent symptoms and radiographic OA. Exclusion criteria were rheumatoid or inflammatory arthritis, bilateral end-stage knee OA, inability to walk without aids and 3 T MRI contraindications.
Double oblique coronal 3D fast low angle shot (FLASH) MRIs with water excitation, a slice thickness of 1.5 mm and an in-plane resolution of 0.31 mm × 0.31 mm of the right knees were available, which had been acquired at 3 T (Siemens Magnetom Trio, Erlangen, Germany) using quadrature transmit–receive knee coils (USA Instruments, Aurora, Ohio, USA). Further technical details on this imaging sequence have been provided for the OAI pilot studies.17–19
Additionally, a sagittal 3D double echo steady-state (DESS) sequence with 0.7 mm slice thickness was available for both knees,17–19
but was not analysed in this study. The reasons for analysing the FLASH rather than the DESS were that: (1) published reports on the rate and SD of change of cartilage morphology have been based on spoiled gradient recalled echo sequences (FLASH or SPGR); (2) FLASH or SPGR are currently available for all scanner manufacturers and are therefore easier to implement in multicentre studies; (3) the double oblique coronal FLASH displays minimal partial volume effects in the weight-bearing femoro-tibial compartment of the knee,20
with the OAI being targeted to (radiographic) femoro-tibial OA); (4) the FLASH has a high and isotropic in-plane resolution (0.31 mm × 0.31 mm vs DESS: 0.37 mm × 0.46 mm); (5) segmentation and quality control of the segmentations are less time consuming and more efficient for the FLASH (1.5 mm slice thickness), as it has fewer slices than DESS (0.7 mm); and (6) the OAI pilot analyses have shown similar test–retest precision for both sequences.17–19 21
Limitations of the FLASH include the relatively long acquisition time, and inferior contrast between the cartilage and the joint capsule at the posterior femoral condyle and at the trochlea (in the region of Hoffa’s fat pad).17
However, these limitations are less important when investigating only the weight-bearing region of the femoro-tibial joint using coronal acquisitions.
The image data were made available on an external hard drive by the OAI coordinating centre and were quality controlled and converted to a proprietary format at the image analysis centre (Chondrometrics GmbH, Ainring, Germany). Segmentation of the femoro-tibial cartilages was performed by seven technicians with formal training and at least 3 years experience in cartilage segmentation. Images were read in pairs, with blinding to the order of acquisition. The total subchondral bone area (tAB) and the cartilage joint surface area (AC) of the medial tibia (MT), the lateral tibia (LT), the central (weight-bearing) medial femoral condyle (cMF) and the central lateral femoral condyle (cLF) were traced manually.22
The weight-bearing region of the femoral condyles was analysed between the intercondylar notch and 60% of the distance to the posterior end of the femoral condyles.17 18
Quality control of all segmentations was performed by a single person (SM), reviewing all segmented slices of all data sets.14 17
Computations of the tAB, the AC, the part of the subchondral bone covered with cartilage (cAB), the denuded subchondral bone area (dAB), the cartilage volume (VC), the mean cartilage thickness over the cAB (ThCcAB, not including denuded areas) and the mean cartilage thickness over the entire subchondral bone area (ThCtAB, including denuded areas as 0 mm cartilage thickness) were then performed.22
Changes were also described for the medial (MFTC) and lateral (LFTC) femoro-tibial compartments, by comparing the summed values of MT and cMF, and LT and cLF, respectively, between baseline and follow-up.18 19
Since changes in VC and ThCtAB may differ from one another under conditions where the tAB is not constant over time,23 24
both variables were reported. We also compared changes in ThCcAB (actual cartilage thickness) or changes in cAB (cartilaginous area), since both may contribute differently to cartilage loss (changes in ThCtAB).22
The mean change, SD of change, standardised response mean (SRM = mean change/SD of change) and the significance of change (two-sided paired t test, without correction for multiple testing) were calculated for each parameter and cartilage plate. The mean percentage change (MC%) was calculated by relating the mean change (in µl, mm or cm2) in all knees to the mean baseline values of all knees. Negative values indicate a decrease over time in the parameters. Differences in the rate of change between femoro-tibial cartilage plates (MFTC vs LFTC, MT vs cMF and LT vs cLF), and between different morphological parameters (ThCtAB vs VC and ThCcAB vs cAB) were tested using two-sided paired t tests. These tests were performed on the individual percentage changes in each parameter and plate/compartment, without correcting for multiple comparisons. Multifactorial analysis of variance (ANOVA) was used for categorical variables (sex, frequent symptoms (yes/no), radiographic OA (K-L grade 2–4 vs 0–1) and obesity (BMI <30 vs >30)), and general linear models (Statistica 6.1) for continuous variables (age, BMI), to test main and interaction effects and to identify risk factors of cartilage loss. The model was also used to evaluate whether estimates of loss were affected by confounders and needed adjustment.