A 3D upright MRI study of patellar alignment in participants with patellofemoral osteoarthritis and matched controls

Purpose: In patellofemoral (PF) osteoarthritis (OA), patellar alignment using magnetic resonance imaging (MRI) has typically been measured in traditional closed-bore scanners, with participants positioned in supine, with the knee in or near full extension, and often in non-weightbearing. However, individuals with PF OA tend to report pain during tasks that increase load at the patellofemoral joint (e.g. stairs, squatting). It may be more relevant to evaluate patellar alignment in functional, upright, full-weightbearing positions. We recently validated methods to evaluate 3D alignment in upright positions using an open-bore MRI scanner. In the present study, we aimed to evaluate 3D patellar alignment in upright (standing and squatting) and supine positions in participants with PF OA compared to matched asymptomatic controls. Methods: We recruited 15 participants with PF OA (aged ≥40 years, pain rated ≥3/10 on most days of the previous month during activities that increase PF joint load, radiographic PF OA ≥ Grade 1 on Kellgren & Lawrence) and 15 asymptomatic controls (matched by age, sex, body mass index [BMI], ethnicity, and current physical activity level). We imaged one knee per participant at four knee flexion angles (0, 15, 30, 45) in upright and supine positions. We scanned each participant in a weightbearing 0.5T vertical gap open scanner (MROpen, Paramed Genoa, Italy) using a sagittal GE sequence (SE:415:10, 300 mm FOV, 4.0 mm/0.4 mm, scan 58 seconds) with a commercial knee coil. We also acquired a higher resolution scan in a 3.0T scanner (Achieva, Philips, Netherlands) using a sagittal T1 weighted TSE sequence (SE:700:10, 320 mm FOV, 2.0 mm/2.0 mm, scan ∼16 min and a dual SENSE Flex-M coil set-up. Using Analyze 10.0 (Analyze Direct, KS, US), we manually segmented bony outlines of the femur and patella on the 3T images to create accurate participant-specific 3D anatomical surface models. We segmented the 0.5T images in a similar way. Using a custom MATLAB program, we then registered the lower-resolution data to the high-resolution models and calculated 3D patellofemoral alignment (key measures based on the literature were medial tilt, lateral translation, and proximal translation). We constructed mixed effects models to evaluate differences by group allocation, across all positions and knee flexion angles in a single model, one model for each alignment variable. We considered all possible interaction terms and square terms for the fixed portion of each model, and fit a random intercept and random slope for knee flexion angle at the pair level for the random portion. We set the level of significance as 0.05. Statistical analyses were done in Stata 13 (StataCorp, TX, US). Results: The full sample had a mean (SD) age of 56 (7) years, BMI 23.5 (3.7) kg/m2, and comprised 24 (80%) women. Almost all measures differed by position and by group (see Figure). Within-group (standing compared to supine), standing alignment was 1.9 (95% CI 0.7, 3.1) mm more laterally translated and 1.4 (0.1, 2.7) mm more proximally translated compared to supine alignment (tilt did not differ). Between-group in standing, PF OA knees were 5.8 (6.9, 4.8) degrees more laterally tilted, 2.5 (1.8, 3.2) mm more laterally translated, and 3.9 (1.7, 6.1) mm more proximally translated than controls. Proximal translation differed more with greater knee flexion. All between-group differences exceeded previously published smallest detectable differences (tilt 4.4 degrees; lateral translation 1.3 mm; proximal translation 2.4 mm). Conclusions: Patellar alignment differed between PF OA and matched controls, and in upright compared to supine positions. The recommended position and knee angle selected for assessing alignment in the future should be tailored to the specific research or clinical question. Measuring alignment in an upright position may better estimate functional malalignment in this important subgroup of participants with knee OA.