0325: Knee Symptom Phenotyping Incorporating Pain and Function across Different Levels of WOMAC Function in Two Cohorts: Data from the Multicenter Osteoarthritis Study (MOST) and the Osteoarthritis Initiative (OAI)

Background/Purpose: Knee pain and functional limitations are the most common symptoms of knee OA. Understanding the various patterns of symptoms (i.e., pain and function changes over time) can help identify individuals most in need of targeted interventions. Using OAI data, we have previously shown that there are distinct symptom trajectories for knees that start with no symptoms at baseline and there are specific predictors of a worse symptom trajectory. Our objective was to analyze knee symptoms among individuals with or at risk for knee OA with varying degrees of symptoms to identify temporal knee pain and function phenotypes using two large cohorts.

Methods: We selected 2965 knees from MOST, and 9491 knees from OAI that had the following knee pain/function measures available at baseline to develop the knee symptom phenotypes based upon trajectories of these measures over time: WOMAC function, WOMAC pain, and the Numerical Rating Scale (NRS) for knee pain severity. We used data from 0 to the 7^th year in MOST, and from 0 to the 9^th year in OAI to develop the trajectories. Knees that were already replaced at the baseline visit were excluded. Latent Class Mixed Models (LCMM) were employed to cluster the selected knees into distinct knee pain/function progression trajectories, while simultaneously using logistic regression to identify predictors associated with each trajectory. The response covariance matrix was allowed to be different for each latent class. We used tertiles of the baseline WOMAC function score as strata for LCMM analyses for each of the symptom measures.

Results: In both cohorts, two trajectories for each model were identified as optimal. For each tertile of baseline WOMAC function, the patterns of the trajectories were similar for each of the pain and function outcomes. As a sample illustration, the clusters for WOMAC function by tertile for MOST and OAI are shown in Figure 1. For the lowest tertile, there was a stable trajectory (blue) vs. a rapidly worsening trajectory (red). For the middle tertile, there was an improving trajectory (blue) vs. a worsening trajectory (red). For the highest tertile, both the red and blue clusters showed improving trajectories, but starting at different baseline scores. The ORs for being in the red (i.e., worse/worsening) vs. the blue (i.e., stable or improving) clusters for each tertile are summarized in Figure 2 for MOST and Figure 3 for OAI. The predictors in common across models associated with being in the red (i.e., worse) cluster were widespread pain, overweight/obese, higher Kellgren-Lawrence Grade (KLG), and more depressive symptoms (i.e., Center for Epidemiologic Studies Depression scale, CES-D).

Conclusion: For both MOST and OAI, similar symptom trajectories were identified when the cohorts were stratified using the tertile of the baseline WOMAC function score. We have also identified readily assessable predictors of the worse symptom trajectories. The trajectories with worse symptom progression may represent a group to target with interventions earlier in their disease course.