Abstract

Purpose: Pain is the cardinal symptom of knee osteoarthritis and its course is known to be variable over time and heterogeneous between individuals. Using data from two large population cohorts with repeated measures of pain severity and several years of follow-up we sought to identify empirically distinct and reproducible long-term symptom trajectories. Methods: Participants were members of the Clinical Assessment Study of the Knee (CAS-K) - a prospective population-based cohort of adults aged 50 years and over reporting knee pain in the last 12 months at baseline and followed up at 18-month intervals over a 6-year period. Latent class growth (LCG) modelling of WOMAC LK3.0 Pain subscale scores (0-20; high score, worse pain) was used to identify subgroups of participants with similar knee pain trajectories over time. Model selection was based on goodness-of-fit statistics and clinical interpretability. Descriptive statistics were used to describe key characteristics of each subgroup including health care use over 6-years. Analysis of variance (ANOVA) was also used to explore whether trajectory group membership predicted rate of decline in locomotor function over 6-years. Model replication was explored within the Osteoarthritis Initiative (OAI) cohort (using data at baseline and five follow-up time points over five years) after propensity score matching had been used to select a subgroup of participants with baseline characteristics that reflected CAS-K for age, gender, body mass index, WOMAC knee pain and function scores and Kellgren-Lawrence x-ray score in the index tibiofemoral joint. Results: CAS-K participants were analysed if they had no existing diagnosis of inflammatory arthritis, had baseline data present for the matching variables and WOMAC pain at two or more of the four available follow-up time points (N=600; mean age 65 years (SD 8.0); 54% female). A linear 1-class LCG model (i.e. for the sample as a whole) showed that the average trajectory of knee pain was relatively stable over time (increase of 0.07 WOMAC points per year). Model fit, however, was suboptimal and masked important clinical subgroups that only emerged when an optimal 5-class model was fitted to the data. Subgroups were: “Mild, but stable” (N=208, 35%), “Progressively deteriorating” (N=170, 28%), “Moderate, but stable” (N=137, 23%) “Improvers” (N=65, 11%), “Severe, but stable (N=20, 3%)”. Substantial differences in patterns of health care use were observed between subgroups and subgroup membership predicted the rate of decline in locomotor function. The 5-class model was a good fit to the matched OAI data. The model gave some support for the replication of the subgroups identified in CAS-K however a subgroup representing those that were “Progressively deteriorating” did not emerge in the OAI data. This may reflect that on average, participants improved, rather than deteriorated over time in the matched OAI data (decrease of 0.25 WOMAC points per year as estimated from a 1-class linear LCG model). Conclusions: The average view of slowly progressive knee pain is shown to be composed of distinct classes of longitudinal trajectories. Correctly identifying class membership at baseline could support targeted intervention, monitoring, and confident reassurance. Differences between cohorts in the rate of improvement and progression of pain deserve further exploration.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call