IMAGING-BASED CHARACTERIZATION OF 3 DIFFERENT OA ENDOTYPES DERIVED FROM BIOCHEMICAL MARKERS - DATA FROM TWO OBSERVATIONAL COHORTS

Abstract

A recent machine-learning study identified 3 different endotype clusters, based on serum and urine biomarker concentration patterns that were consistent between both the APPROACH and the OAI-based FNIH cohort. The clusters were characterized as systemic inflammation (SI), low tissue turnover (LTT), and structural damage of bone and articular cartilage (DAM, Fig. 1). To evaluate whether the prevalence of joint tissue pathologies (MRI Osteoarthritis Knee Scores (MOAKS)) and femorotibial joint (FTJ) cartilage thickness differ between these clusters, and whether 2-year changes in MOAKS scores and FTJ cartilage thickness differ between these 3 clusters. APPROACH enrolled 297 participants (age: 66.5±7.1y, BMI: 28.1±5.3 kg/m², 77.5% women, 55% KLG 2-4) and FNIH 600 OA patients (age: 61.5±8.8y, BMI: 30.7±4.8 kg/m², 58.8% women, 87.5% KLG 2-4). Of 295 APPROACH / 600 FNIH participants with biochemical marker levels available, 93/178 were assigned to the LTT cluster, 92/241 to the SI cluster, and 110/181 to the DAM cluster, respectively. Semi-quantitative MOAKS scores were assessed at months 0 and 24 by an experienced radiologist. FTJ cartilage thickness was measured at months 0 and 24 by manual, quality-controlled segmentations. Frequencies were compared using X² tests between the clusters, and cartilage thickness measures using t-tests. In APPROACH, baseline cartilage damage, osteophytes, effusion- and Hoffa-synovitis did not differ between the 3 clusters, while bone marrow lesions (BML) were more severe in the DAM cluster (Fig. 2). Medial meniscus damage was more severe in the LTT than the SI cluster (Fig. 2). In the FNIH cohort, baseline cartilage damage and BMLs were more severe in the DAM and SI clusters than in the LTT cluster (Fig. 2). Moreover, Hoffa-synovitis was more severe in the DAM than in the LTT cluster; no differences were observed for osteophytes or meniscus MOAKS scores (Fig. 2). In APPROACH, FTJ cartilage thickness was lower in the DAM than in the other clusters, while in FNIH, FTJ cartilage thickness was lower in both the DAM and SI than in the LTT cluster (data not shown). In APPROACH, the DAM cluster showed more 2-year worsening of full thickness cartilage damage, osteophytes, and medial meniscus extrusion than the LTT cluster and more worsening of osteophytes than the SI cluster (data not shown). In FNIH, the DAM cluster displayed more 2-year worsening of cartilage damage extent, lateral meniscus extrusion, and Hoffa-synovitis than the LTT cluster (data not shown). The LTT cluster showed more worsening in BMLs than the DAM cluster, and more worsening in osteophytes than the SI cluster. In both cohorts, change in FTJ cartilage thickness did not differ between clusters (data not shown). Differences in the severity of MR imaging features were more pronounced between clusters in the FNIH than in the APPROACH cohort, potentially because of the greater percentage of knees with radiographic knee OA (KLG2-4). Still, imaging features did not appear to be highly specific for the clusters derived from baseline biochemical markers, most likely because systemic biochemical markers are not only influenced by the selected target knee but also by other joints and tissue changes.