Characterising novel imaging biomarkers for use in knee osteoarthritis clinical trials

Knee osteoarthritis (OA) prevalence is increasing globally. Detailed understanding of structural deterioration in OA knees is hampered by lack of responsive, reliable imaging biomarkers. Magnetic resonance imaging (MRI) has shown that OA pathology involves multiple tissues. Machine-learning based 3D image analysis of MR images accurately quantifies individual tissues and their spatial and temporal relationships. This thesis tested the hypothesis that novel 3D quantitative measures would provide valid imaging biomarkers for knee OA in terms of construct validity, reliability and responsiveness. The Osteoarthritis Initiative provided a unique, large, longitudinal database of knee MRIs to enable detailed and novel statistical analyses of novel imaging biomarkers. A longitudinal study exploring a range of quantitative meniscal measures in 86 patients demonstrated that two exhibited responsiveness comparable to other MRI outcomes, and better than radiographic JSN. Cross-sectional analysis of 600 participants demonstrated a relationship between two potential bone imaging biomarkers, a relatively well studied bone pathology, bone marrow lesions (BMLs) and a novel 3D bone shape measure. Longitudinally, bone shape was more responsive than BMLs. Latent growth modelling on 37,583 knee measurements established that 3D bone shape changed linearly in all three knee bones, with greatest change in the femur, but all three were influenced by clinical covariates in a similar manner. Parallel process growth models showed that onset and rates of structural deterioration were interrelated among the femur, tibia and patella. Latent class growth analysis revealed that distinct trajectories of structural change exist in knee OA. Knee pain, obesity, ethnicity and knee surgery were associated with classification into the fastest trajectory group. In summary, novel quantitative imaging biomarkers of meniscus and bone shape are valid knee OA imaging biomarkers. The introduction of these measures should improve understanding of OA structural pathogenesis, improve clinical trial sensitivity and potentially enable better stratification for clinical trial inclusion.