Background/Purpose: Development of disease-modifying drugs for OA has been challenging, partly due to lack of predictive biomarkers. Our primary objective was to identify baseline (BL) biomarkers predicting greater treatment effects on WOMAC pain among knee OA subjects in the ABT-981 ILLUSTRATE-K trial (NCT02087904).

Methods: Subjects (N=347) with Kellgren-Lawrence (KL) grade 2–3 knee OA, synovitis on MRI or ultrasound, and knee pain score 4–8 (range, 0–10) were randomized to placebo (PBO) or ABT-981 25, 100, or 200 mg subcutaneously every 2 wk for 50 wk. The primary endpoints were change from BL (CFB) in WOMAC pain at wk 16 and CFB in MRI synovitis at wk 26. Demographics, patient-reported outcomes (WOMAC, ICOAP, global assessment [PGA]), x-ray joint space width, and Whole Organ MRI Score (WORMS) were determined at BL. The Patient Rule Induction Method, Sequential Batting, and the Adaptive Index Model were used to identify BL predictive biomarkers and OA subsets with greater ABT-981 treatment effects. Continuous efficacy endpoints were assessed using ANCOVA with treatment, age group, and KL grade as main factors and BL measurements as covariates with LOCF imputation for WOMAC pain.

Results: WORMS Global Total Osteophyte Score (GTOS), which semi-quantitatively summates osteophyte severity from 14 regions of the knee, identified a subset of subjects with a greater ABT-981 treatment effect vs PBO; the optimal GTOS cutoff for discriminating treatment effects was 14 (Figure 1). Among subjects with a GTOS ≥14, the PBO WOMAC pain response was markedly reduced and only marginally improved for ABT-981. At wk 16, among subjects with GTOS ≥14, the standardized mean difference (95% CI) of WOMAC pain for the ABT-981 100-mg dose group vs PBO was −0.62 (−0.16 to −1.09) vs −0.30 (0 to −0.61) for all subjects. Compared with the total study population, the 41% of subjects with GTOS ≥14 not only had a greater ABT-981 treatment effect vs PBO on WOMAC pain, but also other measures of OA symptoms (Table 1). BL systemic markers of synovitis (serum C1M and C3M) and potential markers of macrophage activation by IL-1 (serum alkaline phosphatase) were positively associated with greater ABT-981 treatment effects vs PBO but to a lesser extent than GTOS. Other data supported the robustness of the GTOS predictive marker (Table 2).

Conclusion: The GTOS biomarker predicted improvement of knee OA pain and other symptoms with ABT-981 treatment. We hypothesize that subjects with more severe osteophytes may have had more inflammation-dependent pain that was less responsive to PBO, suggesting that IL-1 inhibitors should be studied further as a treatment for knee OA symptoms in this subset of patients.