PAIN PHENOTYPE AS AN EFFECT MODIFIER: EXPLORING THE ROLE OF PAIN-DETECT IN THE ASSOCIATION BETWEEN WOMAC SCORES AND MRI-DETECTED STRUCTURAL DAMAGE

Abstract

INTRODUCTION

A fundamental challenge in treating patients with OA is the discordance between pain and structural abnormalities, reflecting the fact that numerous factors outside of joint pathology can contribute to the pain experience. Thus, pain in OA represents different phenotypes, including nociceptive and neuropathic-like pain. The Pain-DETECT (PD-Q) can be used to help distinguish between the two, with lower PD-Q scores suggesting nociceptive pain. While associations between different MRI-based MOAKS measures and traditional pain outcomes (such as WOMAC) have been reported, whether those associations are modified by pain phenotype (as assessed by PD-Q) is not known.

OBJECTIVE

To test whether the association between WOMAC scores and MRI-detected OA structural pathology is modified by PD-Q score.

METHODS

We performed a pooled cross-sectional analysis with repeated measures using data from all 4 visits of the IMI-APPROACH cohort (baseline, 6-, 12-, and 24-months), where participants were administered the total WOMAC and PD-Q. For each participant, an index knee with OA was selected based on ACR clinical criteria. MRI of the index knee was obtained for all participants and visits, and scored using MOAKS, including bone marrow lesions (BML); Hoffa’s synovitis and effusion-synovitis. WOMAC scores ranged from 0-96. Participants with no symptoms or functional limitation (WOMAC = 0) were excluded to avoid floor effect. We conducted the analysis in 3 steps: Step 1: We fit linear mixed-effects models with random intercepts for each participant to account for multiple observations per participant and included a MOAKS  ×  PD-Q interaction term to test for effect modification. The MOAKS features we tested included: presence of full-thickness cartilage loss, presence of osteophytes ≥ grade 2, total number of BML, sum of total scores of BML, presence of BML ≥ grade 2, presence of effusion ≥grade 2 and grade 3 separately, presence of synovitis ≥ grade 2 and grade 3 separately, presence of any meniscus tear. Step 2: for each MOAKS measure identified as significant in step 1 (using a liberal threshold of p<0.2) we conducted a Johnson–Neyman (J-N) analysis to locate PD-Q regions where the conditional association between MOAKS and square root of WOMAC (sqrtWOMAC) changed from being statistically significant (using the 95% CI) to non-significant. Step 3: for each MOAKS measure identified in step 1, we stratified the cohort into two PD-Q subgroups (at the J-N cutoff identified in step 2) and refitted stratified linear mixed-effects models to estimate the MOAKS–sqrtWOMAC association within each subgroup.

RESULTS

We included 287 participants (mean age 66.5 (SD=7.2), 77.7% female, mean/median PD-Q score 9.36 / 9), who contributed 949 observations. Significant effect modification was observed for sum of BML size, presence of grade ≥2 osteophytes, and grade 3 effusion synovitis (PD-Q cutoff using J-N were 17.2, 4.7, and 11.6, respectively) (Figure 1). In participants with lower PD-Q scores (based on threshold identified in Step 2), large effusion synovitis was positively associated with WOMAC (β = 1.073; 95% CI: 0.520–1.627). Higher sum of BML sizes was also associated with higher sqrtWOMAC (β = 0.076; 95% CI: 0.039–0.112) (Figure 2). These associations were not statistically significant among participants with higher PD-Q scores (higher likelihood of neuropathic-like pain).

CONCLUSION

The PD-Q score potentially modifies the relationship between WOMAC and MRI-detected sum of BML, and grade 3 effusion. Those with lower PD-Q scores (suggestive of nociceptive pain) show stronger associations between these structural features and WOMAC score.