Background/Purpose: A delay in diagnosis and management of patients with PsA leads to poor radiographic and functional outcomes [1]. The need to identify which patients might progress radiographically has been recognised by the Group for Research and Assessment of Psoriasis and Psoriatic Arthritis (GRAPPA) as a key area of unmet need within PsA [2]. It is anticipated that biomarkers for radiographic joint damage will help in patient stratification so those with greater likelihood of poor outcome may be treated more timely and aggressively. The identification of such biomarkers would also be useful in clinical research to evaluate novel treatment efficacy in selected subjects at higher risk of rapid progression. The SPIRIT-P1 (NCT01695239) Phase 3 randomized controlled trial (RCT) of ixekizumab, a high-affinity IL-17A antagonist antibody, in active PsA patients resulted in reduced progression of structural damage [3], however 5-10% of patients who progressed may have benefitted from an earlier or more aggressive treatment if identified using biomarkers at the outset. The aim of this study was to use mass spectrometry-based proteomics to identify protein biomarkers which might distinguish at baseline those patients who progress to joint damage from those who will not. Top-ranking protein biomarkers were then combined with key clinical parameters to try to improve the discrimination of progressors (P) from non-progressors (NP) to joint damage.

Methods: Baseline serum samples from 83 PsA patients (28 P and 55 NP) were obtained from the SPIRIT-P1 RCT. Radiographic P showed a >0.5 change from baseline modified total Sharp score at week 24 or 52. Two proteomic analyses were performed: 1) unbiased discovery using mass spectrometry of the 83 baseline samples; and 2) targeted analysis of in-house panel of 206 proteins originally developed to distinguish between arthropathies. Univariate and multivariate machine learning random forest modelling was undertaken on the 2 proteomic datasets.

Results: Unbiased discovery proteomics resulted in the identification of 74 unique peptides which were significantly differentially expressed (ANOVA p< 0.01). Random forest modelling identified 15 top-ranking peptides which could distinguish NP from P with a ROC AUC of 0.94. Univariate analysis of the 206 proteins measured by targeted proteomics revealed 4 differentially expressed peptides (ANOVA p< 0.01) and random forest modelling revealed the top 15 candidate peptides could distinguish P from NP with a ROC AUC of 0.85. The baseline clinical data was combined with candidate peptides biomarkers in additional random forest models and this revealed improved model performance.

Conclusion: Data from 2 complimentary proteomic approaches was subjected to univariate and multivariate machine learning statistical analysis which revealed a total of 103 candidate biomarker peptides corresponding with 69 proteins that can potentially discriminate PsA patients who will progress to radiographic damage from those who will not. Random forest models produced convincing ROC AUCs which were improved by inclusion of patient clinical baseline data. The data, whilst promising, requires further validation using separate cohorts of similar patient samples.

« Back to ACR Convergence 2021

ACR Meeting Abstracts - https://acrabstracts.org/abstract/identification-of-serum-protein-biomarkers-at-baseline-to-distinguish-radiographic-progressors-from-non-progressors-in-patients-with-active-psoriatic-arthritis/