Study design and participant selection
We identified individuals for this study using radiographic data from the OAI baseline and first 4 annual follow-up visits. The OAI is a multicenter (Memorial Hospital of Rhode Island, The Ohio State University, University of Maryland and Johns Hopkins University, and the University of Pittsburgh) cohort study that recruited 4796 adults with or at risk for symptomatic knee OA between February 2004 and May 2006 [7]. Institutional review boards at all OAI clinical sites and the OAI coordinating center (University of California, San Francisco) approved the OAI study. Participants provided informed consent prior to participation.
For this study, readers assessed 12 knee features on MR images at the OAI baseline visit, as well as at specific timepoints relative to the onset of disease (i.e. 2 and 1 years prior to onset of disease). Key features include semi-quantitative readings (i.e. collateral ligaments, cruciate ligaments, extensor mechanism, gastrocnemius tendons, infrapatellar fat pad signal intensity alteration, menisci) and quantitative measures (i.e. effusion-synovitis, bone marrow lesion [BML], and cartilage).
Participant selection
Participants in all groups were identified based on annual radiographs from the baseline to the 48-month OAI visit [3]. All groups had at least one knee with no radiographic knee OA at baseline (Kellgren-Lawrence [KL] < 1). Individuals who developed AKOA were defined as having one knee progress to advanced-stage knee OA (KL Grade = 0/1 to 3/4, definitive osteophyte and joint space narrowing) within 48 months (n = 125) [3]. Individuals with typical knee OA experienced a more gradual onset of OA and were defined as having one knee increase in KL grade within 48 months (i.e. KL = 0 to 1, 0 to 2, 1 to 2; n = 187). Individuals were defined as having no knee OA if both knees had no change in KL grade from baseline to the 48-month OAI visit (n = 1325). Individuals in the typical and no knee OA group were randomly matched to the AKOA group based on sex. Each group had 125 participants. For data analysis, we combined the typical knee OA and no knee OA groups into a single “no AKOA” group to allow for a comparison between individuals who would and would not develop AKOA [8].
Index knee
The index knee in individuals with AKOA or typical knee OA was defined as the first knee to meet the definition of AKOA or typical knee OA, respectively. The index knee in individuals with no knee OA was the same knee as that person’s matched member of the AKOA group.
Index visit
For individuals with AKOA or typical knee OA, the index visit was defined as the visit when a person first met the definition for AKOA or typical knee OA. For someone with no knee OA the index visit was the same visit as that person’s matched member of the incident AKOA group. The index visit could be at a 12-, 24-, 36-, or 48-month OAI visit.
Knee radiographs
To determine group assignment, we used readings of bilateral weight-bearing, fixed-flexion posteroanterior knee radiographs obtained at baseline and each annual follow-up visits [3]. Central readers blinded to group assignment scored the KL Grade of each knee (KL = 0 to 4). The intrarater reliability agreement for the KL grades was good (weighted κ = 0.70 to 0.80) [9]. These data are publicly available (files: kXR_SQ_BU##_SAS [versions 0.6, 1.6, 3.5, 5.5, and 6.3]) [10].
MR imaging
MR acquisition
All semi-quantitative and quantitative analyses were conducted in index knees at the OAI baseline visit, as well as at 2 and 1 years prior to the index visit. MR images were acquired with one of four identical Siemens (Erlangen, Germany) Trio 3-Tesla MR systems at each clinical site using the OAI MR imaging protocol [10, 11]. The two musculoskeletal radiologists (RW, JM) performing semi-quantitative scoring were provided all the sequences acquired on each index knee at each visit (e.g., sagittal intermediate-weighted, turbo spin echo, fat-suppressed MR sequence; coronal intermediate-weighted, turbo spine echo, sequence without fat suppression, 3-dimensional dual-echo steady-state sequence). The quantitative measures of BML and effusion-synovitis were performed using a sagittal intermediate-weighted, turbo spin echo, fat-suppressed MR sequence: field of view = 160 mm, slice thickness = 3 mm, skip = 0 mm, flip angle = 180 degrees, echo time = 30 ms, recovery time = 3200 ms, 313 × 448 matrix, x resolution = 0.357 mm, y resolution = 0.511 mm, and total slice number = 37. Cartilage damage index was quantified using a 3-dimensional dual-echo steady-state sequence: field of view = 140 mm, slice thickness = 0.7 mm, skip = 0 mm, flip angle = 25 degrees, echo time = 4.7 ms, recovery time = 16.3 ms, 307 × 384 matrix, x resolution = 0.365 mm, y resolution = 0.456 mm, and total slice number = 160. These sequences have been described in detail elsewhere [10].
Semi-quantitative structural features
For all semi-quantitative and quantitative outcomes, the readers were blinded to group assignment and were unblinded to the order of time. Two musculoskeletal radiologists (RW:255 cases, JM:120 cases) performed the semi-quantitative MR readings. Readers had good agreement on the presence of each pathology among 25 cases: prevalence-adjusted and bias-adjusted kappa were 0.41 to 0.75 except for the posterior horn of the medial meniscus where the prevalence-adjusted and bias-adjusted kappa was fair at 0.25 (50% agreement).
The radiologists assessed the integrity of anterior/posterior cruciate ligaments, medial/lateral collateral ligaments, extensor mechanism, and gastrocnemius proximal tendons and noted if the structures appeared normal or degenerative. Degenerative tissue was defined as the presence of abnormal intrinsic high-signal intensity within the substance of the ligaments or tendon without discrete tear. Degenerative cruciate ligament pathology combined the presence of anterior or posterior cruciate ligament degenerative pathology. Degenerative collateral ligament pathology combined the presence of medial or lateral collateral ligament degenerative pathology.
The radiologists scored infrapatellar fat pad signal intensity alteration using the MR Imaging Osteoarthritis Knee Score grading system (i.e., normal, mild, moderate, and severe) [12]. Infrapatellar fat pad signal intensity was recoded as absence (i.e., normal) or presence (i.e., mild, moderate, and severe).
The radiologists scored medial and lateral meniscus extrusion using the MR Imaging Osteoarthritis Knee Score grading system (i.e., Grade 0: < 2 mm, Grade 1: 2 to 2.9 mm, Grade 2: 3 to 5 mm, and Grade 3: > 5 mm) [12]. Meniscal extrusion was recoded as absence (i.e., Grade 0) or presence (i.e., > Grade 1).
The radiologists used the International Society of Arthroscopy, Knee Surgery, and Orthopaedic Sports Medicine meniscal tear classification, which was modified for MR imaging [13], to assess the body, posterior/anterior horn of each meniscus as: normal, degeneration, horizontal, flap horizontal, vertical longitudinal, radial, morphologic deformity, maceration, complex, or vertical flap tear. Meniscal pathology was recoded as absence (i.e., normal or degeneration without tear) and presence (i.e., horizontal, flap horizontal, vertical longitudinal, radial, morphologic deformity, maceration, complex, or vertical flap tear). For the medial/lateral menisci, pathology in the three regions meniscal tears of different morphologies were combined into the same variable. The medial/lateral menisci were considered pathologic if pathology was present in any of the three regions.
Quantitative structural features
Effusion-synovitis volume
We used a customized semi-automatic software to measure knee effusion-synovitis. Two readers (JBD and a visiting fellow) used the software to mark the first and last MR slice that included bone, the proximal border of the patella, and the apex of the fibular head on a central slice. The software then automatically segmented effusion-synovitis between these limits based on an existing threshold. The senior reader (JBD) then manually adjusted the threshold to change the effusion-synovitis boundaries and removed areas of high signal intensity that were not effusion-synovitis (e.g., subchondral cysts, blood vessels). The senior reader demonstrated excellent intra-reader reliability (ICC3,1 = 0.96). A total knee effusion-synovitis volume (in cm3) was used for data analysis.
Bone marrow lesion volume
One reader (ACS) measured tibiofemoral BML volume with a semi-automated segmentation method [14, 15]. The only manual step required the reader to identify crude boundaries of the tibia and femur in each slice of the MR images. The boundary furthest from the articular surfaces was marked just prior to the epiphyseal line or at the edge of the bone and soft tissue. The program then automatically identified the precise bone boundaries and performed a thresholding and curve evolution process twice to segment areas of high signal intensity, which may represent a BML. We eliminated false-positive regions by operationally defining a BML based on 2 criteria: 1) the distance between a BML to the articular surface should be < 10 mm; 2) a BML needed to span more than one MR image. The study principal investigator (JBD) reviewed all measurements with both timepoints on screen simultaneously. Our reader demonstrated excellent intra-reader reliability (ICC3,1 = 0.91). A total tibiofemoral BML volume (in cm3) was used for data analysis.
Cartilage damage index
The validated cartilage damage index (CDI) was used to quantify tibiofemoral cartilage size [16, 17]. One reader (JED) manually marked the bone-cartilage boundary on specific knee slices that are automatically selected based on the width of the knee. The reader then measured cartilage thickness at predefined informative locations, which the software automatically located. The software then computed the CDI for the medial femur, lateral femur, medial tibia, and lateral tibia by summing the products of cartilage thickness, cartilage length (anterior-posterior), and voxel size from 9 informative locations in each compartment. All measurements were reviewed by study principal investigator. Our reader demonstrated excellent intra-reader reliability (ICC3,1 = 0.86 to 0.99). The sum of all four tibiofemoral compartment CDI values was divided by the participant’s height to calculate a normalized total tibiofemoral CDI that was used for data analysis.
Clinical data
Demographic and other participant characteristics were acquired based on a standard protocol. We extracted age, body mass index, global impact rating, frequent knee pain and Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain at the OAI baseline visit. The data are publicly available (Files: allclinical0#; version 0.2.2, 1.2.1, 3.2.1, 5.2.1, and 6.2.1) [11].
Data analysis
For the continuous quantitative outcomes, the variables in the entire cohort were separated into tertiles and converted to a dichotomous variable to compare the worst tertile (i.e. largest BML and effusion-synovitis, smallest CDI) to the combination of the other two tertiles to facilitate the interpretation of the odds ratios.
Statistical analysis
Primary analysis
Are Early Pre-Radiographic Structural Features Associated with the Onset of Accelerated Knee Osteoarthritis?
Separate logistic regression models were used to determine which pre-radiographic structural features at OAI baseline were more likely to antedate the development of AKOA compared to individuals not developing AKOA (i.e. referent group). Additionally, we conducted the same analyses for each structural outcome at 2 and 1 years prior to the index visit. The results are presented as odds ratios (ORs) and 95% confidence intervals (95% CIs). To control for multiple comparisons, we utilized a statistically significant p-value corrected for the number of structural features utilized in the primary OAI baseline analysis (p < 0.05/12 = 0.004).
Secondary analysis
Which Combination of Baseline Pre-Radiographic Structural Features Most Associate with the Onset of Accelerated Knee Osteoarthritis?
To explore which combination of pre-radiographic structural features characterize AKOA, we performed a backward stepwise logistic regression where the outcome was AKOA or no AKOA (referent group) at the OAI baseline. Separate models also were conducted for 2 and 1 years prior to the index visit. All 9 semi-quantitative and 3 quantitative pre-radiographic structural features were included in the analysis at each time point. The ability for the combination of pre-radiographic structural features to discriminate between AKOA status was quantified with the C statistic [18]. The discriminatory ability of a model based on the C statistic was classified as: very poor (C < 0.50), poor (0.50 < C < 0.70), good (0.70 < C < 0.80), and strong (0.80 < C < 1.00) [19].
All analyses were performed unadjusted, as the purpose of this investigation was to specifically determine the prognostic capability of baseline structural features at associating with future development of incident AKOA. Due to missing MR images at different OAI visits, there are different sample sizes depending on the analysis: OAI baseline (n = 354), 2 years prior to onset (n = 248), 1 years prior to onset (n = 354). There are uneven sample sizes at the different time points because some participants are unable to have a 2 year prior to onset visit (i.e., index visit at the 1-year OAI visit). We conducted a sensitivity analysis for the OAI baseline and 1 year prior to onset analyses limiting the sample to the 248 participants in the 2 years prior to onset analysis. All analyses were performed with SAS Enterprise 7.15 (Cary, NC, USA).