Background/Purpose:

Treatment of RA patients is guided by measures of disease activity such as DAS28, best practice recommendation, and less often by a treat-to-target approach. This is due to a lack of diagnostic tools that can make an objective endotypic profile of the patient allowing for better targeted treatment. Another challenge is that traditionally used biomarkers reflect the level of systemic inflammation (e.g. cytokines) rather than the affected tissue. The aims were to test a novel combination of blood-based biomarkers reflecting tissue turnover and inflammation to identify patients with different forms of RA. In addition, we investigated whether such endotypes are associated with clinical disease activity and structural progression.

Methods:

Post-hoc analysis was conducted on a cohort of patients with active and moderate-severe RA from a biomarker sub-study of LITHE, a phase III clinical trial (N=741). Only patients from the placebo arm were considered, who had serological biomarkers measured at both baseline (BL) and week 4, as well as bone erosion (ERN) measured at BL and week 52 (n=69). Progression was defined as a positive absolute change in ERN from BL to week 52. The following biomarkers reflecting tissue metabolite were measured in BL samples: PIINP and C2M (cartilage formation/degradation); CTX-I, OC PINP and ICTP (bone resorption and formation); C1M and C3M (interstitial matrix degradation); C4M and C6M (basement membrane degradation; and CRPM and VICM (inflammation).

All serum measurements were log transformed and normalized to have values between zero and one. Unsupervised hierarchical clustering was then performed using serological biomarkers taken at BL and week 4. The significance of change in ERN of each group was tested using a Mann-Whitney U test.

Results:

Hierarchical clustering revealed two main clusters. Cluster A (see figure) is defined by low levels of collagen biomarkers and varying levels of other biomarkers. Cluster B displays high level of bone, connective tissue and basement membrane markers, and low levels of the cartilage markers. Ten of the 12 biomarkers were significantly lower in cluster A than in cluster B (p< 0.5). Cluster A can be divided into several subgroups characterised by high bone biomarkers and low bone biomarkers respectively. Due to the small population size in this study, the significance of these clusters was not investigated.

There is a trend showing that patients in cluster B have a higher DAS28 score at BL (p=0.08). The change in ERN was significantly different between the clusters (p=0.04) indicating group B (55% progressors) progresses faster than group A (29%).

Conclusion:

Using hierarchical clustering we were able to identify different endotypes of structural progression, including faster progressors in most need of treatment. Other likely endotypes were also identified, which shall be investigated further.