Background/Purpose: The WOMAC questionnaire is one of most commonly used instruments to assess knee pain in persons with or at risk of knee osteoarthritis (OA). However, the summary score of the WOMAC pain subscale fails to distinguish pain experienced during activity from pain felt at rest. Yet these two symptomatic profiles may represent distinct phenotypes.  If recognizable patterns of pain are, in fact, distinguishable as unique groupings of the individual WOMAC item responses, then their delineation could valuably inform distinct underlying mechanisms and the subsequent development of phenotype-specific treatments for OA-related knee pain.

Methods: The Osteoarthritis Initiative (OAI) cohort consists of adults ≥ 45 years of age that have or are at risk for knee OA. The WOMAC pain subscale was administered at baseline and each annual follow-up visit. Using a latent class model (SAS software PROC LTA), we identified distinct pain phenotypes based on participants’ responses to each of the 5 WOMAC pain questions. The resultant pain phenotypes each consist of subjects with homogenous responses to the 5 WOMAC pain questions. We then examined the baseline characteristics of these pain phenotypes and estimated the transition probability of each pain phenotype to another over 24 months follow-up.

Results: Among 3,985 participants (7,906 knees) who had a WOMAC pain score available during the first 3 follow-up visits, we identified 5 distinct knee pain phenotypes at baseline: 1) No/minimal pain (No-Pain: 55.0%); 2) Mild pain during activity only (Mild-P-A: 20.0%); 3) Moderate pain during activity only (Mod-P-A: 12.3%); 4) Mild pain both during activity and at rest (Mild-P-A-R: 7.1%); and 5) Moderate pain both during activity and at rest (Mod-P-A-R: 5.8%). Age at baseline was similar across the 5 pain phenotypes; however women, higher BMI, higher depressive symptoms, and K/L score ≥3 tended to be higher among subjects with pain both during activity and at rest than pain during activity only. During follow-up, a majority of knees’ pain phenotypes remain stable; ~11% of knees with No-Pain at baseline later developed Mild-P-A, ~20% of knees with Mild-P-A transitioned to the No-Pain group, and ~15% of knees with Mild-P-A transitioned to Mod-P-A. Among knees with Mild-P-A-R, ~43% improved to either No-Pain (~32%) or Mild-P-A (~11%). Among knees with Mod-P-A-R, ~26% had their pain improved to Mod-P-A, while the probability of transitioning to either No-Pain (9.5%) or Mild-P-A (7.2%) was small.

Conclusion: We identified 5 distinct knee pain phenotypes based on responses to the WOMAC pain questionnaire. While the pain phenotypes remained stable in the majority of knees over time, there was substantial transitioning both to better and worse pain phenotypes. These findings may have implications for identification of pain phenotype-specific risk factors and development of treatment for distinct pain phenotypes.