Prediction of response to therapy in rheumatoid arthritis: Lost in validation

Abstract

Introduction Rheumatoid arthritis (RA) is a chronic, disabling disease that mainly affects the synovial joints. With a large arsenal of treatments among which tumor necrosis factor-alpha inhibitors (TNFi’s), RA disease activity decreases sufficiently in most but not all cases. The identification of responders and non-responders before initiation of therapy would therefore aid in making strategic treatment decisions and improve clinical outcomes. Biomarker research frequently identifies new potential targets that predict response, yet validation across multiple cohorts fails so far. Methods In this thesis, data from the Biologicals and Outcome Compared and predicted Utrecht region in Rheumatoid Arthritis” (BiOCURA) study was used. BiOCURA is an observational cohort, in which RA patients eligible for biological treatment according to regular clinical practice were enrolled and followed after start of treatment, in one academic hospital and seven regional hospitals in the Netherlands. Trained nurses gathered all data during a dedicated visit, which included all clinical parameters, joint counts and collection of blood. In total, more than 400 patients were included. Within serum and isolated cells, we searched for potential biomarkers to predict the respons on biological treatment in RA, by screening DNA, mRNA, microRNA, proteins and metabolites. Results Potential interesting targets were identified in all “omics” platforms that were screened. Technical replication of these results with specific assays, e.g. a qPCR for the mRNA targets SEMA6B and GPR15, confirmed the predictive value of these markers in the measured patient population. Yet, true confirmation can only be established when results are replicated in a new set of patients, therefore the identified targets were subsequently measured in a separate selection of patients. None of the markers, however, could be validated. Conclusion In this thesis we were not able to identify predictors, ready to use in clinical practice. This does not mean that the possibility of prediction of response should be ruled out, as many reasons for the inability to find biologically relevant predictors can co-exist. It is therefore conceivable that an adaption of the strategy enables the field to find reproducible and relevant predictors, and the most relevant items are discussed in this thesis. However, the possibility exists that prediction of response is too difficult in practice: the biology of RA might be too complex and variable to get hold of and/or patients may respond very unpredictable to unknown parameters like the exposome. Still, leading scientists in rheumatology firmly believe that predictors for treatment response in RA will be available someday to use. In order to get there, a better understanding of the pathogenesis and pharmacological effects of therapy might be needed. This means that instead of the current ‘shotgun approach’ in which many biomarkers at once are measured and some might hit the target, energy and funding might possibly better be spent in understanding the pathophysiology and pharmacological response mechanisms. Using this knowledge, hypotheses can be made that can be specifically tested, while also taking into account all (biologically possible) influencing factors. Although this might seem a step backwards, it can actually help us forward.