Background/Purpose: The multi-biomarker disease activity (MBDA) test uses a validated algorithm with 12 serum protein biomarkers to assess disease activity in patients with RA. The MBDA score has previously been found to be a predictor of risk for radiographic progression (RP).  We evaluated data from 6 cohorts and performed a meta-analysis to collectively establish the relationship between the MBDA score and risk for RP.

Methods: Clinical, MBDA score and radiographic data were analyzed for 6 cohorts with N>100: Leiden, SWEFOT Year 1, SWEFOT Year 2, OPERA Year 1, and AMPLE Year 1 (abatacept and adalimumab arms). Analyses used data on file when published data were not available (ie, for Leiden and for OPERA CRP analyses). Frequency of RP over 1 year was determined by category of MBDA score (low, moderate, high on a scale of 1−100) at the start of the year, as reported for 4 cohorts and at the end of the year as reported for the 2 AMPLE cohorts. RP was defined using the threshold for change in modified total Sharp score (ΔmTSS) specific to each study (2 to >5 TSS units). Positive and negative predictive values (PPV and NPV) were determined for each study by comparing patients with high MBDA score (>44), DAS28-(ESR/CRP) (>5.1 or >4.09) or CRP (>3 mg/dL) vs those in a low/moderate category. Relative risk (RR) for RP was determined for each study. RR values were integrated in a meta-analysis that included only the 3 non-overlapping cohorts that had reported radiographic analyses using MBDA scores at the start of the year. Results of published multivariate analyses and analyses that combined MBDA score with other risk factors for RP were summarized.

Results: The 6 study cohorts included patients receiving csDMARDs alone or with a biologic. Overall rates of RP were 10−26%. In each study, RP was most frequent among patients with a high vs low/medium MBDA score (>44 vs ≤44). For high MBDA scores, NPVs were 93−97% and PPVs were 18−32%, with RR values of 3.6−9.5 (P=0.002 to <0.0001) (Figure). In a meta-analysis of the Leiden, SWEFOT Year 1 and OPERA Year 1 cohorts, RR was 5.1 (P<0.0001) for MBDA categories, and 1.4 (P=0.23) and 1.6 (P=0.01) for categories of DAS28-CRP or CRP, respectively. Published multivariate analyses in the Leiden and SWEFOT Year 1 cohorts showed MBDA score was an independent predictor of RP after accounting for the effect of other predictors. In the Leiden cohort, MBDA score was the strongest predictor and high MBDA score discriminated between high and low risk for RP among patients with high SJC (>5) or high DAS28-CRP, with PPV as high as 57%.

Conclusion: High MBDA scores were associated with increased risk for RP in 6 study cohorts, including patients treated with csDMARDs, TNFi and abatacept. Based on high NPVs (≥93%), the MBDA score used alone had clinical value for identifying patients with little or no risk of RP. Combining the MBDA score with clinical measures yielded PPVs approaching 60%, suggesting that biomarkers can help stratify patients by their risk for RP.